Fuel Cells for Transportation: Fundamental Principles and Applications [1 ed.] 0323994857, 9780323994859

Citation preview

Fuel Cells for Transportation Fundamental Principles and Applications

This page intentionally left blank

Fuel Cells for Transportation Fundamental Principles and Applications

Edited by

Prodip K. Das Kui Jiao Yun Wang Frano Barbir Xianguo Li

Woodhead Publishing is an imprint of Elsevier 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States The Boulevard, Langford Lane, Kidlington, OX5 1GB, United Kingdom Copyright © 2023 Elsevier Ltd. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). MATLABs is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not warrant the accuracy of the text or exercises in this book. This book’s use or discussion of MATLABs software or related products does not constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular use of the MATLABs software. Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. ISBN: 978-0-323-99485-9 (print) ISBN: 978-0-323-99486-6 (online) For information on all Woodhead Publishing publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Megan R. Ball Acquisitions Editor: Peter Adamson Editorial Project Manager: Zsereena Rose Mampusti Production Project Manager: Kamesh R. Cover Designer: Christian J. Bilbow Typeset by MPS Limited, Chennai, India

Contents List of contributors About the editors Preface

1.

Fuel cells for transportation—an overview

xv xvii xxi 1

Prodip K. Das, Frano Barbir, Kui Jiao, Yun Wang and Xianguo Li

2.

1.1 Introduction 1.2 Hydrogen and fuel cell 1.3 History of hydrogen and fuel cell development 1.4 Fuel cells for transportation 1.5 Present status of fuel cells for transportation 1.6 Future of hydrogen and fuel cells 1.7 Conclusions References

1 3 6 10 15 19 24 26

Fuel cell fundamentals

29

Lei Xing, Jin Xuan and Prodip K. Das 2.1 Introduction 2.2 Operation principle of proton-exchange membrane fuel cells 2.3 Reaction kinetics and transport processes 2.3.1 Electrode kinetics 2.3.2 Multicomponent mass transport 2.3.3 Heat transport 2.4 Electrode properties 2.4.1 Porosity of the catalyst layer 2.4.2 Agglomerate density 2.4.3 Thicknesses of the ionomer and liquid water films 2.4.4 Specific area 2.4.5 Deformation of porous electrode 2.5 Water management 2.5.1 Water phase-transfer and water transport through membrane 2.5.2 Diffusion of species in Nafion ionomer with different membrane water content 2.5.3 Two-phase flow of gas
water mixture

29 29 31 31 42 48 52 52 53 55 56 57 59 59 63 65

v

vi

3.

Contents

2.6 Summary Review questions References

68 69 69

Fuel cell modeling and optimization

73

Lei Xing, Xueguan Song and Prodip K. Das 3.1 Introduction 3.2 Fuel cell modeling approach and key physicochemical and operating parameters 3.2.1 Water formation and transport in fuel cells 3.2.2 PEMFC modeling approaches 3.2.3 Modeling of water transport through the membrane 3.2.4 Modeling of water transport through porous electrodes 3.2.5 Catalyst layer modeling 3.3 Numerical optimization of PEMFCs 3.3.1 Electrode optimization 3.3.2 Flow fields optimization 3.3.3 Fuel cell stack optimization 3.3.4 Operating condition optimization 3.3.5 Multivariable optimization and data-driven surrogate modeling 3.4 Summary References

4.

Lattice Boltzmann modeling and artificial intelligence

73 74 75 76 79 81 82 83 84 89 91 92 93 96 97

103

Xing Li, Yuze Hou, Nada Zamel and Kui Jiao 4.1 Overview of lattice Boltzmann method and artificial intelligence 4.2 Application of lattice Boltzmann method in fuel cells 4.2.1 Current status of pore-scale research in gas diffusion layer 4.2.2 Current status of pore-scale research in the microporous layer 4.2.3 Current status of pore-scale research in catalyst layer 4.3 Artificial intelligence method 4.3.1 Parameter optimization 4.3.2 Model predictive control 4.3.3 Prognostics and health management 4.3.4 Fault diagnosis 4.4 Combination of lattice Boltzmann method and artificial intelligence 4.5 Summary References

103 107 107 110 112 114 115 117 118 120 122 123 123

Contents

5.

Low platinum-based electrocatalysts for fuel cells: status and prospects

vii

127

Huiyuan Liu and Xianguo Li 5.1 Introduction 5.2 Functionalization of commercial carbon supports 5.2.1 Effects of the structure and surface properties of carbon support 5.2.2 Functionalization methods of commercial carbon supports 5.3 Methods for loading Pt-based electrocatalysts on carbon supports 5.3.1 One-pot synthesis method 5.3.2 Ex situ mixing method 5.4 Synthesis of Pt-based electrocatalysts 5.4.1 Synthesis of Pt-based spherical nanoparticles 5.4.2 Synthesis of Pt-based polyhedrons 5.4.3 Synthesis of Pt-based open nanostructures 5.4.4 Synthesis of 1D Pt-based nanostructures 5.4.5 Synthesis of 2D Pt-based nanostructures 5.5 Postsynthesis treatments of Pt-based electrocatalysts 5.6 Future direction and prospects 5.6.1 Studies on functionalization of commercial carbon supports 5.6.2 Production of Pt-based electrocatalysts 5.6.3 Postsynthesis treatment References

6.

Platinum group metal-free catalysts for fuel cells: status and prospects

127 128 128 132 136 137 137 138 140 141 145 149 152 153 155 156 157 158 158

177

Md Aman Uddin and Ahmed Imtiaz Rais 6.1 Introduction 6.2 Platinum group metal-free catalyst development 6.3 Integration of platinum group metal-free catalyst in membrane electrode assembly 6.3.1 Effect of ionomer loading, equivalent weight of ionomer, and dispersion of ionomer in solvent 6.3.2 Effect of primary particle size 6.3.3 Engineering cathode to improve water management 6.4 Stability and durability of platinum group metal-free cathode 6.4.1 Micropore flooding 6.4.2 Active site protonation 6.4.3 Demetallation, carbon oxidation, and attack of peroxide and associated radicals 6.5 Mitigation strategies

177 178 182 183 184 185 187 188 189 189 190

viii

7.

Contents

6.6 Summary Acknowledgments References

192 192 192

Effective transport properties for fuel cells: modeling and experimental characterization

199

Pablo A. Garcı´a-Salaberri and Prodip K. Das

8.

199

7.1 Introduction 7.2 Structure and composition of porous transport layers in fuel cells 7.2.1 Porosity and pore size 7.2.2 Wettability 7.3 Effective transport properties 7.3.1 Effective diffusivity 7.3.2 Local mass transport resistance 7.3.3 Permeability 7.3.4 Effective thermal conductivity 7.3.5 Effective electrical conductivity 7.3.6 Effective ionic conductivity 7.4 Modeling and experimental techniques 7.4.1 Modeling 7.4.2 Experimental 7.5 Summary References

201 202 205 208 209 210 212 212 212 213 213 213 215 219 219

Liquid water transport and management for fuel cells

225

Anthony D. Santamaria and Prodip K. Das 8.1 Water production 8.2 Two-phase flow basics 8.3 PEMFC architecture 8.3.1 Membrane 8.3.2 Catalyst layer 8.3.3 Microporous layer 8.3.4 Gas diffusion layer 8.4 Channels and flow fields 8.5 Liquid management concerns and strategies 8.6 Pressure and flow control 8.7 Thermal regulation and humidification 8.8 Startup/shutdown 8.9 Surface coatings 8.10 Ultrathin electrodes 8.11 Patterned and structured porous media 8.12 Summary References

225 226 233 234 234 235 235 237 241 242 243 243 245 245 246 247 248

Contents

9.

Fuel cell short stack testing

ix 253

Graham Smith and Katie McCay 9.1 9.2 9.3 9.4

Introduction Principles of fuel cell operation and testing Testing requirements Measurement techniques 9.4.1 Characterization of test station 9.4.2 Stack operation 9.4.3 Diagnostics 9.4.4 Durability testing 9.5 Summary Acknowledgments References

10. Power demand for fuel cell system in hybrid vehicles

253 254 259 262 262 264 267 274 275 276 276 279

Rui Ma, Elena Breaz and Fei Gao 10.1 10.2 10.3 10.4

Introduction to hybrid fuel cell powertrain Fuel cell hybrid electric vehicle road testing profiles Fuel cell hybrid electric vehicle body modeling Fuel cell power demand from hybrid powertrain 10.4.1 The importance of energy management strategy 10.4.2 Major influential factors of fuel cell operation in vehicle applications 10.4.3 State of the art of fuel cell hybrid electric vehicles energy management strategies 10.5 A case study for the fuel cell hybrid electric vehicles energy management strategy 10.6 Conclusion References

11. Bipolar plates and flow field design

279 281 284 287 287 288 289 296 301 301 305

Xianguo Li 11.1 Introduction 11.2 Bipolar plates 11.2.1 Functions 11.2.2 Requirements 11.3 Flow field design 11.3.1 Flow field without guided flow path 11.3.2 Flow field with guided flow path 11.4 Materials and manufacturing 11.4.1 Typical materials and classification 11.4.2 Graphite 11.4.3 Carbon composite 11.4.4 Metallic bipolar plates 11.5 Summary

305 307 308 309 311 312 313 330 331 332 333 334 336

x

Contents

Acknowledgments References

12. Heat transport and thermal management

336 336 339

Siyuan Wu, Kui Jiao and Jae Wan Park 12.1 Introduction 12.2 The heat in proton exchange membrane fuel cell 12.2.1 Heat generation 12.2.2 Heat transport 12.3 Proton exchange membrane fuel cell thermal management 12.3.1 The cooling of proton exchange membrane fuel cell 12.3.2 Thermal management subsystem 12.3.3 Control strategy 12.3.4 Cold start 12.4 Summary Nomenclature References

13. Mass transport in the cathode

339 340 340 343 348 348 356 358 361 361 363 364 367

Linhao Fan, Zhiming Bao, Daniela Fernanda Ruiz Diaz, Yun Wang and Kui Jiao 13.1 Mass transfer in cathode gas flow fields 13.1.1 Characterization of oxygen distribution and water removal 13.1.2 Mass transfer in conventional flow channels 13.1.3 Mass transfer in novel flow fields 13.1.4 Effects of gas diffusion layer and operation conditions 13.2 Mass transfer in cathode gas diffusion layer and microporous layer 13.2.1 Characterization of oxygen transport 13.2.2 Characterization of liquid water transport 13.2.3 Characterization of electron transport 13.2.4 Innovations in gas diffusion layer structure and material 13.3 Mass transfer in cathode catalyst layer 13.3.1 Mass transfer in local region near catalysts 13.3.2 Mass transfer in bulk catalyst layers 13.4 Summary Nomenclature References

14. Control-oriented computational fluid dynamics models for polymer electrolyte membrane fuel cells

367 367 368 370 373 373 374 376 378 379 380 382 385 386 386 387

393

Jian Zhao, Xianguo Li, Chris Shum and John McPhee 14.1 Introduction

393

Contents

xi

14.2 1D computational fluid dynamics model 14.3 Pseudo-2D computational fluid dynamics model 14.4 Model accuracy and computing speed 14.5 Summary Acknowledgments References

395 401 408 412 412 412

15. Fuel cell durability under automotive driving cycles—fundamentals and experiments

419

Jian Zhao and Xianguo Li 15.1 Introduction 15.2 Fundamental degradation mechanisms under automotive driving cycles 15.2.1 Polymer electrolyte membrane 15.2.2 Catalyst layer 15.2.3 Gas diffusion layer 15.2.4 Bipolar plate 15.3 Steady-state durability test 15.3.1 Steady-state durability test protocols 15.3.2 Degradation rate by steady-state durability test 15.4 In situ accelerated stress test 15.4.1 In situ accelerated stress test protocols 15.4.2 Degradation rate by in situ accelerated stress test 15.5 Ex situ accelerated stress test 15.5.1 Humidity cycling 15.5.2 Temperature cycling 15.5.3 Hygrothermal cycling 15.5.4 Liquid water wet
dry cycling 15.5.5 Freeze
thaw cycling 15.5.6 Clamping force 15.5.7 Vibration 15.6 Summary Acknowledgments References

16. Subzero startup of polymer electrolyte fuel cell— a battle between water and thermal management at low temperatures

419 422 422 424 430 430 431 431 432 435 436 444 447 447 448 448 449 452 452 453 453 454 454

463

Jianbo Zhang, Dechun Si and Kei Ono 16.1 Introduction 16.2 Overview of subzero experiment 16.2.1 Four categories of experiment in subzero study 16.2.2 Test fixtures 16.2.3 Test procedures and control of temperature 16.2.4 Characterization techniques

463 466 466 469 469 471

xii

Contents

16.3 Damage and mitigation in subzero scenarios 16.3.1 Damages to fuel cell components from freeze/thaw and subzero startup 16.3.2 Mitigation effects of material and design 16.3.3 Mitigation effects of operation control 16.3.4 Discussion on the cause and mitigation of damages in subzero 16.4 States and behavior of water at subzero 16.4.1 States of water in fuel cell at subzero 16.4.2 Water transport dynamics in fuel cell at subzero 16.4.3 Water fill test: contribution and limitation 16.5 Temperature-dependent properties and thermal behavior at subzero 16.5.1 Temperature-dependent properties 16.5.2 Thermal management issues highlighted from a lumped model 16.5.3 Discussion on heat source, temperature difference, and thermal BC control 16.6 Subzero startup strategies and techniques 16.6.1 Unassisted start: a battle between water and thermal management 16.6.2 Assisted start: decoupling thermal from water management 16.7 Directions for further study 16.8 Summary 16.9 Exercise questions References Further reading

17. Solid oxide fuel cells for vehicles

471 474 480 480 483 485 485 491 503 511 512 513 519 525 525 532 533 535 536 537 544 547

Haoyu Li, ThomasJae Garcia and Min Hwan Lee 17.1 Overview of fuel cell 17.2 Solid oxide fuel cells for transportation 17.2.1 Solid oxide fuel cells: advantages and shortcomings 17.2.2 Leveraging advantages and overcoming shortcomings 17.2.3 Cell configurations 17.3 Fuel types 17.3.1 Hydrogen 17.3.2 Ammonia 17.3.3 Hydrocarbons 17.3.4 Alcohols 17.4 Applications 17.4.1 Auxiliary power units for heavy-duty trucks 17.4.2 Aircraft 17.4.3 Maritime applications 17.4.4 Railways 17.4.5 Passenger vehicles

547 548 548 549 550 550 551 551 554 555 556 556 557 558 559 560

Contents

xiii

17.5 Challenges and related efforts 17.5.1 Mechanical integrity 17.5.2 Low-temperature operation 17.5.3 Metal-supported cells 17.5.4 Tubular configuration 17.5.5 Wet impregnation 17.5.6 Activity degradations 17.5.7 Sintering and loss of active area 17.5.8 Dopant segregation 17.5.9 Carbon coking 17.6 Conclusion References

562 562 562 563 563 564 564 565 565 566 567 568

18. Hydrogen refueling stations/infrastructure

575

Yiheng Pang, Andrew Martinez and Yun Wang 18.1 Introduction 18.2 Hydrogen fuel 18.2.1 Hydrogen properties 18.2.2 Energy for compression or liquefaction 18.3 Hydrogen refueling station 18.3.1 Hydrogen storage and source in hydrogen refueling stations 18.3.2 Refueling at hydrogen refueling stations 18.3.3 Hydrogen refueling stations capital cost 18.4 Hydrogen refueling station networks 18.5 Challenges in hydrogen refueling stations network development 18.6 Summary 18.6.1 Review questions/worked examples Nomenclature Acronyms References Index

575 576 576 579 581 582 583 584 586 589 591 592 592 592 593 599

This page intentionally left blank

List of contributors Zhiming Bao State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China Frano Barbir Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, Split, Croatia Elena Breaz University of Technology of Belfort-Monte´bliard, Belfort, France Prodip K. Das School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom Daniela Fernanda Ruiz Diaz Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States Linhao Fan State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China Fei Gao University of Technology of Belfort-Monte´bliard, Belfort, France ThomasJae Garcia Department of Mechanical Engineering, University of California, Merced, CA, United States Pablo A. Garcı´a-Salaberri Departamento de Ingenierı´a Te´rmica y de Fluidos, Universidad Carlos III de Madrid, Madrid, Spain Yuze Hou Fraunhofer Institute for Solar Energy Systems ISE, Freiburg, Germany Kui Jiao State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China Min Hwan Lee Department of Mechanical Engineering, University of California, Merced, CA, United States Haoyu Li Department of Mechanical Engineering, University of California, Merced, CA, United States Xianguo Li Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada Xing Li State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China Huiyuan Liu Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada Rui Ma School of Automation, Northwestern Polytechnical University, Xi’an, P.R. China Andrew Martinez California Air Resources Board, Sustainable Transportation and Communities Division, California Environmental Protection Agency, Sacramento, CA, United States

xv

xvi

List of contributors

Katie McCay SINTEF Industry, Sustainable Energy Technology, Trondheim, Norway John McPhee Department of Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada Kei Ono School of Vehicle and Mobility, Tsinghua University, Beijing, P.R. China Yiheng Pang Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States Jae Wan Park Department of Mechanical and Aerospace Engineering, University of California, Davis, CA, United States Ahmed Imtiaz Rais Department of Mechanical Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh Anthony D. Santamaria General Motors, Global Propulsion Systems Pontiac Engineering Center, Pontiac, MI, United States Chris Shum Department of Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada Dechun Si Beijing Huairou Laboratory, Beijing, P.R. China Graham Smith National Physical Laboratory, Teddington, United Kingdom Xueguan Song State Key Laboratory of High-Performance Precision Manufacturing, School of Mechanical Engineering, Dalian University of Technology, Dalian, P.R. China Md Aman Uddin Department of Mechanical Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh Yun Wang Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States Siyuan Wu Department of Mechanical and Aerospace Engineering, University of California, Davis, CA, United States Lei Xing School of Chemistry and Chemical Engineering, University of Surrey, Guildford, United Kingdom Jin Xuan School of Chemistry and Chemical Engineering, University of Surrey, Guildford, United Kingdom Nada Zamel Fraunhofer Institute for Solar Energy Systems ISE, Freiburg, Germany Jianbo Zhang School of Vehicle and Mobility, Tsinghua University, Beijing, P.R. China Jian Zhao Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada

About the editors Prodip K. Das is currently an associate professor in hydrogen energy systems at the University of Edinburgh. He received his Bachelor of Science in mechanical engineering degree from the Bangladesh University of Engineering & Technology (BUET) in 1998 with distinctions. Thereafter, he received two master’s degrees. The first one was from BUET in 2001, and the second one was from the University of Alberta in 2003. After graduating from BUET, he worked as a lecturer (1998
2001) and then as an assistant professor (2001
07) at BUET. He received his PhD degree in mechanical engineering with a specialization in hydrogen fuel cells from the University of Waterloo in 2010. Then, he was an NSERC Postdoctoral Fellow at Lawrence Berkeley National Laboratory from 2010 to 2013 investigating water management and dynamics of water transport in PEM fuel cells and real-time defect detection using infrared thermography for fuel cell electrodes. His research interests and areas of expertise include hydrogen energy systems, fuel cells, Li-ion batteries, flow batteries, transport phenomena, porous media, nanofluids, convective heat transfer, infrared thermography, and multiphysics modeling. Dr. Das is an associate editor of Frontiers in Energy Research, Journal of Electrochemical Energy Conversion and Storage, and Frontiers in Chemical Engineering, and he is on the editorial/advisory board for Energies, Batteries, Renewable and Sustainable Energy, Inventions, and Challenges. He led several special issues as a lead guest editor, including the special article collection: “Rising Stars in Fuel Cells: 2022” for Frontiers in Energy Research. He has more than 120 scholarly publications on hydrogen fuel cells, Li-ion batteries, heat transfer, and related mechanical engineering areas, and he has served as track chair/cochair, session chair/cochair, and technical/organizing committee member for many international conferences. He is the recipient of several prestigious awards, including the Dr. V.G. Desa Gold Medal (BUET, Bangladesh), the Dr. Chandrashekar Memorial Award in Sustainable Energy (Waterloo, Canada), and the 2018 Emerging Investigators in Electrochemical Energy Conversion and Storage (ASME, USA).

xvii

xviii

About the editors

Kui Jiao is currently a chair professor at the State Key Laboratory of Engines and the executive deputy director of the National Industry-Education Platform of Energy Storage, Tianjin University, China. He received his PhD degree in mechanical engineering from the University of Waterloo, Canada, in 2011. His research interest includes fuel cell, battery, thermoelectric generator, turbocharger compressor, and other energy conversion technologies. He has published several books and more than 200 articles in highly reputed international journals, including Nature. He served as the chair for several international conferences such as International Conference on Energy and AI. He was granted the “National Natural Science Foundation of China—Outstanding Youth Foundation” and the “UK Royal Society—Advanced Newton Fellowship.” He has led more than 30 national and industrial projects and provided modeling and design services in the development of fuel cell engines for several major fuel cell manufacturers such as FAW, SAIC Motor, Bosch, and Weichai Power. He serves as the founding editor of energy and AI, associate editor of the International Journal of Green Energy, and specialty chief editor of Frontiers in Energy Research. He is the chair of the Energy Storage Division of the International Association for Green Energy, and Fellows of the Royal Society of Chemistry and the Institution of Engineering and Technology. Yun Wang received his BS and MS degrees in mechanics and engineering science from Peking University in 1998 and 2001, respectively. He went to the Pennsylvania State University where he received his PhD degree in mechanical engineering in 2006. He joined the Mechanical and Aerospace Engineering (MAE) Faculty at the University of California, Irvine, in 2006. He has produced more than 90 publications on PEM fuel cells, Li-air batteries, and other energy systems, including two books on PEM Fuel Cell Water and Thermal Management Fundamentals in 2013 and 2021, respectively. He has received several awards, including the prestigious President’s Award and Outstanding Educator Award from Orange County Engineering Council, and the 2011
12 Applied Energy Certificate of Excellence: Most Downloaded Authors. Several of his seminal works are highly cited in major fuel cell journals. He served as track chair/cochair, session chair/cochair, conference chair, and committee member for many international conferences on fuel cells, thermal energy, and engineering. He received the 2018 Reviewer of

About the editors

xix

The Year from the Journal of Electrochemical Energy Conversion and Storage. He is currently a professor at UC Irvine, an ASME fellow, an RSC fellow, and an associate editor of the ASME Journal of Heat Transfer. Frano Barbir is a Professor Emeritus at the Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, Croatia. He has been actively involved in hydrogen and fuel cell technology R&D, engineering and applications since 1989, working in the United States as a researcher and R&D manager in both industries (Energy Partners, Proton Energy Systems) and universities (University of Miami, University of Connecticut), and as the associate director of Science and Technology of the UNIDO
International Center for Hydrogen Energy Technologies in Turkey. His research interests include heat and mass transfer in PEM fuel cells, effects of operational conditions on fuel cell performance and durability, design of fuel cells and fuel cell stacks and systems, fuel cell applications, and hydrogen energy concept and its role in the context of energy future. He has authored and/or coauthored more than 200 papers on hydrogen and fuel cells published in scientific and technical journals, books, encyclopedias, and conference proceedings, as well as eight US patents on various aspects of fuel cell stack and system design and operation. His book, PEM Fuel Cells: Theory and Practice, published by Elsevier/Academic Press in 2005 (second edition came out in 2013), is being used as a textbook at many universities all over the world. He is the President of the Croatian Hydrogen Association and a Vice President of the International Association for Hydrogen Energy. He is a member of the Croatian Academy of Engineering. He holds a Dipl.-Ing. degree in mechanical engineering and an MSc degree in chemical engineering both from the University of Zagreb, Croatia, and a PhD degree in mechanical engineering from the University of Miami, Coral Gables, FL, United States. Xianguo Li is a professor of mechanical and mechatronics engineering and a University Research Chair, at the University of Waterloo. He received his bachelor of engineering degree from Tianjin University, China, in 1982, and master’s and PhD degrees from Northwestern University, Evanston, Illinois, United States, in 1986 and 1989, respectively. His research interests include fuel cells, liquid fuel atomization and sprays, and green

xx

About the editors

energy systems, as well as the thermal management of power batteries for electric vehicles. His research activities in fuel cells encompass system integration and optimization; design and optimization of stacks, including bipolar plates and flow fields; design, optimization, manufacturing and processing, testing and modeling/simulation of membrane electrode assemblies (MEAs); synthesis, characterization, and testing of catalysts; modeling and simulation, performance and durability testing, and testing methods and instrumentations for key fuel cell materials, components, MEAs, stacks, and systems. He has published extensively, including over 230 peer-reviewed journal articles; over 250 conference publications; more than 10 patents filed or granted; 30 chapters;120 research contract reports; 3 edited books, and a single-authored book Principles of Fuel Cells, the world’s first textbook on fuel cells, used extensively around the world. He has delivered more than 230 invited keynote lectures at various research institutions and international conferences. He serves as the Editor-in-Chief for the International Journal of Green Energy, Field Chief Editor for the journal Frontiers in Thermal Engineering, and is also on the editorial/advisory board for many journals, book series, encyclopedias, and handbooks. He is the Vice President of Technical Program, Canadian Society for Mechanical Engineering (CSME), President of the Fuel Cell Division, International Association for Hydrogen Energy, and President of the International Association for Green Energy. He is a Fellow of the Canadian Academy of Engineering, Engineering Institute of Canada, and CSME. He is the founder and initiator of the International Green Energy Conference series (IGEC) and the World Fuel Cell Conference (WFCC) series.

Preface Hydrogen fuel cells have attracted much interest as a promising green energy technology for various transportation applications from small passenger cars to heavy-duty trucks, transport buses, and rail transportation due to their high efficiency, power density, and zero emission at the point of use. Even though fuel cells were invented more than 180 years ago and then went through a rapid development for the space program in the 1960s, which was followed by even more intensive development in automotive applications since the 1990s, currently, there are two major fuel cell car models available in the international market, with a third coming in 2024. Several technical barriers still require continual improvement, including their high cost, which stems from the need to balance durability, performance, and materials. Thus the question is: is there a future for fuel cells for transportation? The future of fuel cells, including the fuel cells for transportation, is tightly related to the energy transition that is already taking place. In such a transition, hydrogen produced from renewable energy sources would enable the decarbonization of otherwise hard-to-decarbonize sectors, including the transportation sector. Fuel cells are the technology that will enable hydrogen’s use in transportation. Technology is ready, and political decisions have been made, so in the next decades, fuel cells will be widely used in transportation, not only in automobiles but also in delivery vehicles, trucks, buses, coaches, trains, and even in ships and airplanes. Although several vehicle manufacturers and the drone industry are dedicated to hydrogen fuel cell technology for light-duty vehicles (LDVs), there will be other challenges for fuel cells in heavy-duty vehicles (HDVs). This book highlights the current challenges, latest market outlooks, and targets for fuel cells in LDVs and HDVs, provides new avenues to mitigate current technical challenges, and discusses the solutions to fuel cell system integration and optimizing the operating conditions and improvements needed for fuel cell materials with the latest and ongoing research across the globe. This book includes the contributions of many renowned hydrogen and fuel cell scientists and engineers from Bangladesh, the EU, Canada, China, the United Kingdom, and the United States. It covers both fundamental principles and the latest advances in fuel cell systems for transportation applications. The key areas that are addressed in this book are, but not limited to, fuel cell fundamentals, modeling and performance optimization, stack

xxi

xxii

Preface

characterization, design of bipolar plates and flow fields, advancement in water and thermal management, durability under automotive driving cycles, cold start, state-of-the-art characterization, and optimization of various components. The book also highlights the requirements for hydrogen refueling stations and infrastructure. We hope that academic researchers, scientists, and professional engineers working in the area of hydrogen and fuel cells will be benefiting from this book and will gain up-to-date knowledge and advances in fuel cell materials, components, stacks, and systems. This book consists of 18 chapters, as follows: Chapter 1 showcases an in-depth highlight of fuel cells for transportation along with the history of hydrogen, the hydrogen roadmap, and the strategy needed for hydrogen, or, more correctly, the “hydricity” economy. It also highlights the operation principle of the fuel cell power train along with its key components. The requirements for a fuel cell system for cars, buses, trucks, and related applications are discussed along with the key target metrics for performance, cost, and durability for fuel cell electric vehicles (FCEVs). In addition, the emerging transportation sectors for FCEVs are highlighted, and the future of hydrogen and fuel cells is critically analyzed. Chapter 2 is devoted to fuel cell fundamentals. In this chapter, the fundamentals and principles of a typical proton-exchange membrane fuel cell (PEMFC) operated with hydrogen and air are represented, and the governing equations are derived. All the important chemical and physical processes involved in the operation of PEMFCs are also described by proper differential and algebraic equations. These governing equations build the framework of multiphysics, nonisothermal, and two-phase flow PEMFC models. Chapter 3 provides the chronological development of PEMFC modeling approaches with a focus on those modeling the catalyst layer and water formation and transport inside the PEMFCs. Numerical optimizations of PEMFCs with respect to electrodes, flow fields, fuel cell stacks, and operating conditions are summarized. The multivariable optimization and datadriven modeling are also introduced in this chapter. Chapter 4 introduces lattice Boltzmann modeling (LBM) and artificial intelligence (AI) for fuel cells. It reviews advanced research in which LBM is used in gas diffusion layers, microporous layers, and catalyst layers that explore reactant diffusion, two-phase flow, and electrochemical reaction processes, as well as highlights the application of AI methods in fuel cells, including parameter optimization, model predictive control, prediction and health management, and fault diagnosis. Chapter 5 discusses low platinum (Pt)-based electrocatalysts for fuel cells and their status and prospects. The current development of the controlled synthesis of Pt-based electrocatalysts for PEMFCs is covered in this chapter, including the size, size distribution, functionalization of carbon supports, shape, and simple low-cost synthesis, by controlling the experimental

Preface

xxiii

parameters. Perspectives for future research and development of Pt-based electrocatalysts are also provided. Chapter 6 summarizes the state-of-the-art platinum group metal (PGM)
free catalyst development methods and the integration of these catalysts in fuel cells. Subsequently, an understanding of PGM-free catalysts’ stability challenges and mitigation strategies is highlighted. Chapter 7 highlights the effective transport properties of fuel cell components and their modeling and experimental characterization. The content is intended to provide the reader with an overview of porous transport layers in fuel cells, the structure and composition of porous components, effective transport properties, and modeling and experimental techniques that are widely used for fuel cell characterization. Moreover, the main characteristics and effective transport properties of porous layers are discussed. Chapter 8 examines liquid water transport and management, which is critical to PEMFC performance. This chapter highlights liquid transport mechanisms, two-phase flow behavior, operation strategies, and novel materials and manufacturing methods, which can offer new opportunities and challenges in this area. Chapter 9 explains the fuel cell stack characterization. It provides a practical guide to characterizing fuel cell short stacks in a laboratory environment with dedicated test equipment and outlines the principles, facilities, considerations, operating conditions, and available diagnostic techniques. Though the focus of this chapter is on PEMFC short-stack testing, readers will also find much that is directly applicable to smaller single-cell testing. Chapter 10 demonstrates the total power demand for fuel cell systems in hybrid vehicles. A general idea of how the total power demand profile of a fuel cell hybrid electric vehicle (FCHEV) can be determined and how this power profile can be distributed between different power sources are included. This chapter also presents the major challenges and influential factors of fuel cell operation for vehicle applications, and a case study of a simple energy management strategy based on fuzzy logic control for FCHEV is demonstrated. Chapter 11 covers the bipolar plates and flow field design. The basic functionalities and requirements for bipolar plates, various basic flow field designs and associated various design improvement features, proper materials, and associated manufacturing methods that have been developed to achieve the functions and requirements for the bipolar plates are highlighted and discussed. This chapter culminates with a description of the state-of-theart knowledge and technology employed in the two commercial fuel cell vehicles and a practical comparison of the carbon-based vs. metal-based bipolar plates. Chapter 12 presents the heat generation mechanisms for fuel cells. Different types of waste heat are explained along with the heat transport mechanisms within different components of a PEMFC. The cooling methods

xxiv

Preface

for fuel cell stacks and their pros and cons, as well as their corresponding application scenarios, are well elucidated. The functions of each component included in a representative thermal management subsystem of the PEMFC stack are explained. The control strategies applied in thermal management, like proportional
integral
derivative (PID) control, model predictive control (MPC) control, fuzzy control, artificial neural network (ANN) control, adaptive control, and robust control, are also presented. Chapter 13 discusses the mass transport in the cathode, which plays a vital role in fuel cell performance, especially under high current density. Important mass transfer mechanisms, such as oxygen transport in flow fields, porous media, ionomer film, and two-phase flows, are introduced. An outlook on future routes of cathode development is also discussed. Chapter 14 is devoted to automotive fuel cell control strategies. Controloriented models of PEMECs with good fidelity and computational speed are the core of model predictive control to optimize fuel cells’ operation for better performance and a longer lifetime. Computational fluid dynamics (CFD) models demonstrate a good balance between the two trade-off factors—accuracy and computing speed—and less dependence on a significantly large number of experimental data in comparison to data-driven models. In this chapter, the recent development of reduced-dimensional CFD models, 1D and pseudo-2D, which can be potentially used for fuel cell control, is highlighted. Chapter 15 is about fuel cell durability under automotive driving cycles. In this chapter, the aging phenomena in major PEMFC components, such as membranes, catalyst layers, gas diffusion layers, and bipolar plates, are systematically examined, and the causes of the component degradation are analyzed. This chapter also provides insights into the fundamental understanding of durability issues, experimental investigation on lifetime, component design, and control strategy development. Chapter 16 addresses the subzero startup from the perspective of water and thermal management at low temperatures. It first gives an overview of the subzero experimental studies, including four categories of experiments, test fixtures, test procedures, and featured characterization techniques. It then presents the study of the fuel cell damages in subzero scenarios, which has motivated the subsequent efforts to mitigate and resolve the issue. After that, it summarizes the current understanding of the states and behaviors of water, and the thermal properties and behavior at subzero. Next, it compares the unassisted and the assisted startup strategies. Finally, it comments on the directions for further study. Chapter 17 reviews recent efforts in propelling relevant research and deploying solid oxide fuel cell (SOFC)
based systems for transportation. It discusses the intrinsic advantages and representative applications of SOFCs for transportation, current technological issues, and recent efforts to address these challenges.

Preface

xxv

Chapter 18 highlights the requirements for hydrogen refueling stations and infrastructure, which is crucial to fuel cell transport applications. Fundamentals of hydrogen gas compression and liquification and the associated energy use are briefly described. The cost and challenges in hydrogen refueling stations and infrastructure development are explained.

This page intentionally left blank

Chapter 1

Fuel cells for transportation—an overview Prodip K. Das1, Frano Barbir2, Kui Jiao3, Yun Wang4 and Xianguo Li5 1

School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom, 2Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, University of Split, Split, Croatia, 3State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China, 4 Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States, 5Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada

1.1

Introduction

Climate change and greenhouse gas (GHG) emissions are one of the most challenging problems that humanity has ever faced. The GHGs, such as water vapor (H2O), carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and ozone (O3), are causing the Earth’s atmosphere to warm, resulting in changes to the climate that we have been observing over the past decades. Some GHGs are emitted into the Earth’s atmosphere through natural processes, but a significant amount of GHGs is coming from human activities, particularly from the combustion of fossil fuels. The amount of GHG emissions that come from the combustion of fossil fuels can be as high as 90%, and the transportation sector is responsible for a large chunk of that GHG emissions. Internal combustion engines (ICEs) running on fossil fuels like gasoline and diesel not only release CO2, a powerful GHG, but also release other GHGs, such as CH4, N2O, and hydrofluorocarbons (HFCs). According to the International Energy Agency (IEA), global energyrelated CO2 emissions in 2018 were 33.5 billion tonnes CO2, and the emissions from the transport sector were 8.2 billion tonnes CO2; thus, the transport sector was responsible for about 24.5% of global CO2 emissions in 2018 [1,2]. Road vehicles, such as cars, trucks, and buses, were responsible for nearly 75% of transport CO2 emissions, and emissions from aviation and shipping are continually rising, as highlighted in Fig. 1.1. The demand for transport will be growing in the coming decades due to the increasing global population and better affordability of cars, trains, and flights. It is therefore Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00013-7 © 2023 Elsevier Ltd. All rights reserved.

1

2

Fuel Cells for Transportation

FIGURE 1.1 Global CO2 emissions from the transport sector from the year 2000 to 2018 (top) and the donut chart showing the sector-wise breakdown of CO2 emissions in 2018 (bottom). Data are from the International Energy Agency (http://www.iea.org).

expected that global transport in passenger kilometers (represents the transport of one passenger by a defined mode of transport over 1 km) to double and the demand for passenger and freight aviation to triple by 2070, which could lead to a large increase in CO2 emissions from the transport sector [3]. A recent report published by the Rhodium Group highlights the GHG emissions by major emitting sectors in the United States from 2005 to 2021, as shown in Fig. 1.2 [4]. It shows that the transportation sector is the highest emitting sector since 2016, higher than the industry sector (including manufacturing) and the electric power generation sector—this is primarily because the transportation sector is the most difficult sector for decarbonization. So the transport sector should be at the front line of efforts in reducing GHG emissions and moving toward the net-zero target. Currently, two avenues are available as viable alternatives to fossil fuels that can meet the requirements of zero-emission vehicles: batteries and fuel cells.

Fuel cells for transportation—an overview Chapter | 1

3

FIGURE 1.2 Greenhouse gas emissions from major emitting sectors in the United States from 2005 to 2021. Data are from the Rhodium Group (http://www.rhg.com).

Given the major technological innovations in zero-carbon electricity sources and hydrogen technologies, it is possible to eliminate GHG emissions from the transport sector. However, reaching net zero by 2050 could be a formidable challenge for the transport sector. It would be impossible to achieve net zero by 2050 only through electrification of the energy sector or battery electric vehicles (BEVs) for the transport sector and without hydrogen fuel cells. While BEVs would be a viable solution for the decarbonization of light-duty vehicles, fuel cell electric vehicles (FCEVs) provide the best solution for the decarbonization of both light-duty and heavy-duty vehicles (such as buses, trucks, and rail transport) as well as ships, ferries, and the aviation sector.

1.2

Hydrogen and fuel cell

Hydrogen is not an energy source but rather a fuel or an energy carrier that can be produced from a variety of sources. Obviously, the largest available hydrogen source in nature (on Earth) is water. Hydrogen production from water—water electrolysis—is a relatively simple and efficient process. Hydrogen is a very good fuel. Its main feature is that the result of its use, either in an engine or in a fuel cell is water. Production of hydrogen from water and hydrogen use, thus, does not generate harmful emissions. Inevitably, the process of generation of hydrogen requires more energy than it can be harnessed when hydrogen is used as a fuel, either in an ICE or in a fuel cell [5]. Hydrogen conversion to electricity in fuel cells is by its nature more efficient than that in the ICE process. The efficiency of an ICE is

4

Fuel Cells for Transportation

limited by the Carnot cycle, that is, by the temperatures of the thermal reservoirs (or an internal combustion engine, which is a 2T heat engine). A fuel cell, as an electrochemical energy converter, is a 1T engine. The maximum theoretical efficiency of fuel cells is given by the ratio of Gibbs free energy, ΔG, and enthalpy of reaction, ΔH, that is, η 5 ΔG/ΔH. It should be noted that when hydrogen is used as a fuel, the enthalpy of the reaction has the same value as the hydrogen higher heating value (HHV) for low-temperature fuel cells. Many fuel cell developers/manufacturers use the efficiency values based on the hydrogen’s lower heating value (LHV), to facilitate the comparison with ICEs, which traditionally use the efficiencies based on the LHV of the fuel used. Efficiencies based on both LHV and HHV are acceptable, as long as it is clearly stated which value is used in the calculation [6]. There is a similarity between the Carnot efficiency—the maximum theoretical efficiency a heat engine can achieve—and the fuel cell’s maximum theoretical efficiency. Both are theoretical, and both are achieved when the device, that is, an engine or a fuel cell, produces no power and as such has no practical value. The efficiency of a heat engine, as well as the efficiency of a fuel cell, has significantly lower values when the device generates power. According to the IEA, it is the right time to tap into hydrogen’s potential to play a key role in a clean, secure, and affordable energy future. Hydrogen can help tackle various critical energy challenges—not only for the transport sector but also for other sectors such as the chemical, iron, cement, and steel industries and the residential heating and cooling sector. Hydrogen along with fuel cells can not only power our vehicles but also help the decarbonizing heavy industry through carbon capture and storage (CCS). A hydrogen fuel cell works much like an electric battery, while batteries are energy storage devices and fuel cells are energy converters. Unlike batteries, the reactants in a fuel cell are supplied from an external source; thus, it can operate as long as fuel is supplied into it and does not run down or need recharging. In a hydrogen fuel cell, the chemical energy of hydrogen or other fuels is converted into electrical energy through electrochemical reactions and using the movement of charged ions (positive or negative) across an electrolyte. A single fuel cell consists of two electrodes (an anode and a cathode) sandwiched around an electrolyte. The anode is the negative electrode where hydrogen or other fuels release electrons into the external circuit, and the cathode is the positive electrode associated with reductive chemical reactions that gain electrons from the external circuit. The electrolyte acts as a separator for electron flow but provides ionic conductivity between the positive and negative electrodes. Both the anode side and the cathode side include a gas diffusion layer (GDL) and a catalyst layer (CL). Often a microporous layer (MPL) is added in between the GDL and CL for better water management, mass transport, and physical contact [7,8]. A single fuel cell hardware

Fuel cells for transportation—an overview Chapter | 1

5

assembly and the components of a fuel cell, such as GDL, MPL, CL, and electrolyte, are shown in Fig. 1.3. The power produced by a fuel cell depends on the fuel cell type, size, temperature at which it operates, and pressure at which gases are supplied/ consumed. A single fuel cell produces roughly 0.5
0.7 V with a power density of about 1 W/cm2 or less, which is insufficient to power even some small electrical devices. Thus cells are always combined in series to form a stack to achieve higher output voltage and desired power. Depending on the application, a fuel cell stack may contain only a few individual cells layered together or as many as hundreds of cells or higher. This scalability allows fuel cells to be constructed into a wide array of sizes to fit the required amount of energy desired for various applications, as listed below: 1. Warehouse logistics—to power clean trucks, forklifts, and pallet jacks. 2. Global distribution—several companies, including Nikola, Hyundai, Toyota, Kenworth, and UPS, are building medium- and heavy-duty fuel cell trucks for long-haul trucking and local distribution.

FIGURE 1.3 A single fuel cell hardware assembly (top) and schematic illustration of the fuel cell’s components (below). Credit: Dr. Deepashree Thumbarathy, Sustainable Energy Systems Lab, Newcastle University, UK.

6

Fuel Cells for Transportation

3. Public transportation—many major cities, including Chicago, Vancouver, London, and Beijing, have fuel cell buses for public transport. 4. Rail transportation—Alstom SA has demonstrated the world’s first hydrogen fuel cell train (Coradia iLint) in Germany. Outside of Germany, deployments of fuel cell trains are scheduled for Japan and South Korea in the future. 5. Passenger vehicles—major auto manufacturers, including Toyota, Honda, Hyundai, and BMW, have already developed FCEVs. 6. Aviation—Airbus unveiled the ZEROe concept aircraft with hydrogen fuel cells to provide electrical power. 7. Backup power generation—stationary fuel cells are being used as uninterruptible power supply systems. 8. Portable power generation—NASA is using hydrogen fuel cells to provide electricity for rockets and shuttles in space. 9. Marine transportation—fuel cells offer marine industries (boats, ships, and ferries) the perfect solution to the ever-growing need for clean onboard power to reduce GHG emissions on the water and in ports. 10. Unmanned aerial and underwater vehicles—Intelligent Energy in the United Kingdom using fuel cells in drones to extend flight times.

1.3

History of hydrogen and fuel cell development

The idea of using hydrogen as fuel was first presented by Jules Verne, a famous science fiction writer, in his book The Mysterious Island [9], where one of his characters says: Water decomposed into its primitive elements. . .and decomposed doubtless, by electricity. . .will one day be employed as fuel,. . .hydrogen and oxygen which constitute it, used singly or together, will furnish an inexhaustible source of heat and light, of an intensity of which coal is not capable,. . .Water will be the coal of the future.

The concept of a solar-originated hydrogen economy was first set down almost two centuries later by Bockris (1962), developed and diagrammed by Justi (1965), named a Hydrogen Economy by Bockris and Triner (1970), formulated by Bockris (1971) and Bockris and Appleby (1972), and quantified by Gregory (1972) and Marchetti (1972) [10]. Bockris and Veziroglu outlined the solar hydrogen energy system and discussed the real economics of potentially competitive energy systems of the future [11]. They showed that if hydrogen utilization efficiency advantage and total fuel costs (i.e., cost of production plus the cost of environmental damage done in every step of the fuel cycle) are taken into account, the solar hydrogen energy system is the most economical energy system possible. David Scott pointed out that both electricity and hydrogen will be the energy carriers in a future energy system

Fuel cells for transportation—an overview Chapter | 1

7

based on renewable energy and suggested that such an economy should not be called a hydrogen economy but rather a “hydricity economy” [12]. One of the key features of hydrogen as an energy carrier is that it does not need to be burned in order to produce useful energy. A fuel cell represents a very elegant way to convert hydrogen to electricity, and therefore the fuel cell may be considered as a device that would enable hydrogen or, more correctly, “hydricity” economy. The discovery of the fuel cell effect is attributed to Swiss-German scientist Christian F. Sch¨onbein, who in the January 1839 issue of the London and Edinburgh Philosophical Magazine and Journal of Science published an article entitled “On the voltaic polarization of certain Solid and fluid substances” [13]. Describing his experiments and his findings, he concluded that “The current in question is caused by the combination of hydrogen and oxygen (contained in dissolved water) and not by contact.” Welsh lawyer and experimental scientist, William R. Grove in the February 1839 issue of the same London and Edinburgh Philosophical Magazine and Journal of Science published an article on voltaic series and the combination of gases by platinum [14] with very similar findings as Sch¨onbein. The two of them started exchanging correspondence [15] and continued working on oxygen/hydrogen pile as Sch¨onbein called it, or gaseous voltaic battery as Grove called it. In 1842, Grove published a paper “On gaseous voltaic battery” [16] which included a drawing of 4 cells connected in series—thus the first working fuel cell. In 1845, Grove published another paper describing a fuel cell system with supply of air and supply of hydrogen from zinc sulfuric acid reaction [17]. Thus, it may be concluded that Sch¨onbein discovered the fuel cell effect, but Grove invented a practical fuel cell [15]. Many other scientists and engineers throughout the rest of the 19th century contributed to improvements and development of fuel cells, such as Ludwig Mond and Carl Langer [18,19], to whom the name “fuel cell” is attributed, Charles R. Alder Wright and Charles Thompson [19], and Louis Paul Cailleteton and Louis Joseph Colardeau [20]. They could make a unit that worked in the laboratory and would give a small amount of current, but would cost too much to be practical [18,20]. Friedrich Wilhelm Ostwald, founder of “physical chemistry” (and Nobel Prize winner in 1909), realized the potential of electrochemical devices as superior to the steam engine not only in terms of efficiency, but also in terms of environmental impact. In his lecture in 1894, he envisioned a technical revolution [21]: Could we, hence, convert the chemical energy of the fuel material into mechanical energy in such a way that heat is not involved, then we would not be bound to the inconvenient high temperatures and could gain the whole amount without putting up with that inconvenience. The way how to solve the

8

Fuel Cells for Transportation

biggest of all technical problems, the procurement of cheap energy, this way has to be found through electrochemistry. If we had a galvanic cell, which supplies immediate electrical energy from coal and the oxygen of the air, in an amount, which, to a certain extent, is in relation to the theoretical value, in that case, we stand before a technical revolution, against that of the invention of the steam engine would vanish.

Improvements in fuel cell technology and a quest for a practical device continued in the 20th century. Reid developed and patented the perforated metallic electrodes (1902), Jungner developed and patented the first porous hydrophobic air electrode (1919), Schmid developed the first double-layer hydrophilic hydrogen electrode (1923), and Niedereiter not only developed and patented perforated metallic electrodes, but also applied pressure to the entire cell (1932) [18]. The most successful fuel cell pioneer was Francis T. Bacon, who started working on practical fuel cells in 1937 in the United Kingdom, following the work of Grove and Mond [18,22]. By the end of the 1950s, he developed a 6-kW fuel cell stack and system, which was demonstrated in a forklift and used for welding in 1959 [18]. The American company Pratt and Whitney obtained licenses for Bacon’s fuel cell, which was the basis for the fuel cell system developed for the Apollo Space Program, where the fuel cells were used to generate electricity for life support, guidance, and communications. The 1959 Allis-Chalmers farm tractor, developed by Harry Karl Ihrig using modified Bacon’s fuel cell that ran off propane, was demonstrated in Milwaukee as the first fuel cell vehicle in history [18,22]. Grubb and Niedrach working in General Electric developed a fuel cell with a solid ion-exchange membrane electrolyte in 1960 [23], which was successfully applied in several missions of the Gemini space program completed during 1965
66. General Electric Company registered the Solid Polymer Electrolyte as a trademark name and then transferred it to the Hamilton-Standard Division of United Technologies Corporation. The technology was subsequently moved to International Fuel Cells, which later became UTC Fuel Cells, a unit of United Technologies Corporation. Today, this type of fuel cell is generically called the solid polymer fuel cell (SPFC) or the proton-exchange membrane (PEM) fuel cell. In the 1960s, Walter Grot working for DuPont developed Nafion perfluorosulfonic acid polymer membrane [24], which has become a standard for PEM fuel cells (PEMFCs) ever since. The 1966 General Motors Electrovan is the first hydrogen fuel cell car of record. It was powered by a 5 kW Union Carbide fuel cell. All of the fuel cell parts and hydrogen storage tanks are in the back of the van. The vehicle had a range of 120 miles, though it was only driven on company property [18,19]. In 1970, Karl Kordesch, who earlier made significant contributions in developing carbon electrodes for alkaline fuel cells, then working for Union

Fuel cells for transportation—an overview Chapter | 1

9

Carbide fitted his own Austin A40 with a hydrogen fuel cell in the trunk, and six hydrogen tanks on the roof of the vehicle, providing a range of about 300 km (180 miles), and he used the adapted vehicle as his personal transportation for over 3 years [18,22]. In 1984 under contract to the Canadian Department of National Defense, Ballard Power Systems began the development of SPFC stack technology. The strategy to lower fuel cell costs was to use less expensive materials and to use materials and fabrication techniques that could achieve higher performance. Graphite replaced niobium, which had been used in the Gemini fuel cells, as the material for the flow field plates. In 1987, Ballard tested a new membrane produced by Dow Chemical Company that allowed the cell to produce four times the current compared to that allowed by Nafion at the same cell voltage. In 1993, Ballard Power Systems demonstrated the first PEM fuel-cell-powered bus and became the first commercially successful fuel cell company [25]. In 1989, Perry Energy Systems, a division of Perry Technologies, using a fuel cell stack by Ballard (allegedly the first fuel cell stack Ballard ever sold), successfully demonstrated a PEM fuel-cell-powered submarine. Energy Partners, a successor of Perry Energy Systems, demonstrated the first passenger car running on PEMFCs, also in 1993 [26] but using its own fuel cell stack based on technology acquired from Treadwell Corp. The same company also developed an extended golf cart—people mover called Genesis, powered by hydrogen/oxygen fuel cell, which was demonstrated at the Atlanta Olympic Games for carrying journalists around the Olympic Village. Also, by 1999 the same company refurbished three John Deere Gators, off-road vehicles for lawn maintenance, and made them run on hydrogen with a new generation self-made fuel cell stack [26]. In collaboration with John Deere Co., these vehicles were demonstrated at the Palm Springs airport, where they were used for handling luggage [27]. By the turn of the century, almost every car manufacturer had built and demonstrated at least one prototype fuel-cell-powered vehicle. Some went through several iterations/improvements, and some even started manufacturing small series. It seemed that a new industry was being born. A strong push for the development of hydrogen fuel cell vehicles came from American President George W. Bush, who, in his 2003 State of the Union address [28], announced a $1.2 billion research investment to help the country “lead the world in developing clean, hydrogen-powered automobiles.” He declared: “With a new national commitment, our scientists and engineers will overcome obstacles to taking these cars from laboratory to showroom so that the first car driven by a child born today could be powered by hydrogen, and pollution-free.” After that, the development of fuel cell vehicles accelerated, not only in the United States, but also in Europe, Japan, and South Korea. Under the U.

10

Fuel Cells for Transportation

S. Department of Energy Hydrogen Program, significant improvements were made toward reaching the performance, durability, and cost goals. The Honda FCX Clarity concept car was introduced in 2008 for leasing to customers in Japan and Southern California. From 2008 to 2014, Honda leased a total of 45 FCX units in the United States [29]. Over 20 other FCEV prototypes and demonstration cars were released in that period, including the GM HydroGen4 and Mercedes-Benz F-Cell [29]. The Hyundai ix35 FCEV Fuel Cell vehicle was available for lease from 2014 to 2018, when 54 units were leased [29]. In 2018, Hyundai introduced the new fuel cell car—NEXO. Toyota Mirai is one of the first mass-produced and commercially sold fuel cell vehicles. Its sales started in December 2014 to government and corporate customers, first in Japan, and then expanded to the general public in the United States (2015) and Europe (2016). As of December 2021, global sales totalled 17,940 Mirais, mostly in United States and Japan. The secondgeneration improved Mirai was announced in December 2019 and launched in April 2021. The Honda Clarity Fuel Cell was produced from 2016 to 2021. However, in 2019, Honda announced that its focus is on hybrid and electric vehicles now. Maybe hydrogen fuel cell cars will come, but that is a technology for the next era. Honda announced it would discontinue its Honda Clarity Fuel Cell series in August 2021. However, on November 30, 2022, Honda announced that it will begin production of FCEVs in its Ohio plant in 2024, starting with the CR-V compact SUV. This CR-V will include both FCEV and BEV technologies (or plug-in FECVs) so that it combines the fastrefueling capabilities of an FCEV for long-distance travel with the “convenience” of a BEV for the short-distance urban drive. Even earlier, in 2017, Daimler phased out its FCEV development, citing declining battery costs and increasing range of EVs, and most of the automobile companies developing hydrogen cars had switched their focus to BEVs. Instead, fuel cells have been focused for heavy-duty vehicles, such as trucks and buses.

1.4

Fuel cells for transportation

Hydrogen’s rapid refueling time and higher energy density give fuel cells an advantage over battery-only solutions in applications such as warehouse logistics, public transportation, long-haul trucking, ships and ferries, and aerospace. Although fuel cells for transportation include a wide range of use cases, passenger and commercial vehicle applications of hydrogen fuel cells show the highest signs of promise for widespread adoption. For instance, passenger light-duty vehicles, which accounted for three-quarters of FCEV stock at the end of 2020, with buses and commercial vehicles are making up about 15% and 10%, respectively [30], while

Fuel cells for transportation—an overview Chapter | 1

11

applications of fuel cells in trains, unmanned aerial vehicles, e-scooters, ships, and ferries are still quite early in development with limited deployments to date. A variety of fuel cells are currently available in the market for various applications, but they all share the same basic configuration—an anode and a cathode sandwiched around an electrolyte. The electrolyte in a fuel cell determines the steps of chemical reactions that take place in the anode and cathode electrodes, operating temperature, and other factors that determine its most suitable applications. Among the various fuel cells, polymer electrolyte fuel cells (PEFCs) are best suited for transportation due to their high power density, low operating temperature, quick start-up, and fast dynamic response. PEFCs use a solid polymer as their electrolyte (either PEM or anion-exchange membrane) and porous carbon electrodes often containing a platinum or platinum alloy catalyst and operate on hydrogen and oxygen from the air. They are fueled with pure hydrogen supplied from storage tanks or reformers. Based on the use of electrolytes, PEFCs can be subdivided into two categories: PEMFC or anion-exchange membrane fuel cell (AEMFC). While PEMFCs are already deployed for passenger and public vehicles, AEMFCs are getting attention lately due to their ability to operate with nonPGM catalysts. Several major automobile manufacturers are currently selling FCEVs based on the PEMFC stack, as stated in the previous section. Though PEMFC is a mature technology and has shown promising performance and durability improvement over the past three decades, PEMFC-based FCEVs are not yet commercially competitive with ICE vehicles. Several technical challenges, including cost and durability, still need to be overcome for PEMFC, which could be achieved via decreasing expensive platinum catalyst loading in the cathode CL [31] and reducing cathode water flooding through better water management [32
34]. Due to these technical challenges for PEMFCs, a growing interest has been developed over the past years in advancing AEMFCs for transportation applications, as they offer advantages over PEMFCs in dealing with the expensive platinum catalysts and water flooding [35
38]. As AEMFC allows faster oxygen reduction, it does not require expensive platinum catalysts, thus, reducing cost. Moreover, water is produced on the anode side, which allows easier water management for AEMFCs. Both the PEMFCs and the AEMFCs have identical cell components configuration, except for their electrolytes. PEMFCs utilize proton-conducting polymer as an electrolyte, while AEMFCs utilize hydroxide-ion-conducting polymer as an electrolyte. The other differences between them are the electrode reactions and the transports of protons and hydroxide ions, as highlighted in Fig. 1.4. For PEMFC, the electrode reactions are: Anode Reaction: 2H2 -4H1 1 4e2

ð1:1Þ

12

Fuel Cells for Transportation

FIGURE 1.4 Schematic diagrams of polymer electrolyte fuel cells. The proton-exchange membrane fuel cell is on the left side, and the anion-exchange membrane fuel cell is on the right side.

Cathode Reaction: 4H1 1 4e2 1 O2 -2H2 O

ð1:2Þ

For AEMFC, the electrode reactions are: Anode Reaction: 2H2 1 4OH2 -4H2 O 1 4e2

ð1:3Þ

Cathode Reaction: O2 12H2 O 1 4e2 -4OH2

ð1:4Þ

The above reactions are single-step reactions, but multiple elementary reaction pathways are usually involved at each electrode as explained in [8,39]. For both PEMFC and AEMFC, the net electrochemical reaction is the same and can be represented as: Net Fuel Cell Reaction: 2H2 1 O2 -2H2 O 1 Heat 1 Electrical Energy ð1:5Þ In general, PEMFC provides higher power output at a given voltage than an AEMFC [8,35
38,40]. However, recent studies highlight some promising performances for AEMFCs [36]. Architecture of fuel cell powertrain: Similar to BEVs, FCEVs use an electric motor powered by onboard electricity to drive the vehicle. In FCEVs, onboard electricity is produced by a fuel cell stack running on hydrogen, rather than drawing electricity from only a battery like BEVs. However, all FCEVs have a battery pack for recapturing braking energy as well as providing extra power when needed, such as during short acceleration events. The onboard battery pack also ensures smoother power delivery from the fuel cell stack with the option to idle or turn off the fuel cell stack during low power needs. Unlike BEVs, the energy stored in the battery pack is determined by the size of the hydrogen fuel tank rather than the size of the battery pack.

Fuel cells for transportation—an overview Chapter | 1

13

The key components of a hydrogen FCEV are shown in Fig. 1.5. An FCEV includes 1. Auxiliary battery—acts like an ICE vehicle battery, and it powers the vehicle during startup and provides power to vehicle accessories. 2. Onboard traction battery pack—it stores excess energy generated by the fuel cell stack and regenerative braking and provides supplemental power to the drive shaft. 3. DC/DC converter—converts higher-voltage DC power from the traction battery pack to the lower-voltage DC power needed to recharge the auxiliary battery and to run vehicle accessories. 4. Electric traction motor—like BEVs, this motor drives the vehicle’s wheels but uses the power from the fuel cell stack as well as from the onboard traction battery pack whenever needed. 5. Fuel cell stack—the heart of an FCEV that converts hydrogen into electricity through the electrochemical reactions between stored hydrogen and atmospheric oxygen and powers the vehicle. 6. Air supply system—supplies atmospheric oxygen into the FCEV and keeps the fuel cell stack from pollutants. 7. Fuel supply system—responsible for safely supplying hydrogen to the fuel cell stack from the high-pressure hydrogen tank (pressure in the tank is as high as about 700 bar, while supply pressure is about 2 bar). 8. Fuel filler—like an ICE vehicle, this allows filling the fuel tank (in FCEVs it is a hydrogen tank). It is a nozzle from a fuel dispenser that attaches to the receptacle on the vehicle to fill the tank.

FIGURE 1.5 A hydrogen fuel cell electric car with its key components. Credit: U.S. Department of Energy (afdc.energy.gov).

14

Fuel Cells for Transportation

9. Hydrogen tank—like an ICE vehicle, it is the fuel tank for an FCEV and stores hydrogen gas on board the vehicle until it is needed by the fuel cell stack. 10. Power electronics controller—it manages the flow of electrical current delivered by the fuel cell stack and the traction battery and controls the speed and torque of the electric traction motor. 11. Cooling/thermal management system—as the fuel cell produces excess heat, this system rejects excess heat like the radiator of an ICE vehicle and manages the temperature of the fuel cell stack. 12. Transmission—unlike ICE vehicle transmission, it converts electrical power to mechanical power through the electric traction motor and drives the vehicles. Operation principle of fuel cell powertrain: A schematic diagram of an FCEV powertrain’s operation principle is shown in Fig. 1.6. Apart from the fuel cell stack, a fuel cell powertrain has four subsystems for hydrogen, oxygen, coolant, and high-voltage components. These subsystems are fuel supply system for hydrogen, air supply system for oxygen, water and thermal management system for coolant and product water, and power electronics and controller for high-voltage and related components. The fuel supply system conveys hydrogen to the anode side of the stack from a high-pressure hydrogen tank. An anode recirculation blower within the fuel supply system promotes the circulation of hydrogen in the anode path. An air supply system includes an air filter, an air compressor, and humidifiers to convey clean oxygen to the cathode side of the stack. Air is drawn directly from the atmosphere and then passed through the air filter to remove particulates and other unwanted matter. The humidifiers ensure the relative humidity of supplied air. Water and thermal management system with separate water and coolant loops removes waste heat and product water. Some of the waste heat can be

FIGURE 1.6 Schematic diagram of a fuel cell vehicle operation principle.

Fuel cells for transportation—an overview Chapter | 1

15

recirculated to the vehicle cabin during the winter days, which would improve the vehicle’s overall efficiency. The water and thermal management system also controls the operating temperature of the fuel cell stack as well as temperatures for the electric motor, power electronics, and other vehicle components. The electricity generated by the stack then goes through a power electronics and controller to the electric motor, which transmits power to the vehicle wheels. The excess electricity from the stack is stored in a battery pack to supplemental power to the drive shaft whenever needed.

1.5

Present status of fuel cells for transportation

Currently, only two car makers (Toyota and Hyundai) are still manufacturing or had active manufacturing programs for hydrogen fuel cell cars. There are two passenger car FCEVs on the market in the world right now: Toyota Mirai and Hyundai NEXO, while there are currently 12 fuel cell bus models and five truck models available on the market, according to IEA. By unit count, most of them are in China, which accounts for 94% of global fuel cell buses and 99% of fuel cell trucks [41,42]. There are several reasons why fuel cell vehicles are not (yet) on the road in significant numbers. 1. The first one is unrealistic expectations. The proponents of the hydrogen economy have created the expectation that in such an economy, hydrogen will be produced from every energy source, even including fossil fuels, and that hydrogen will be used in all applications where fossil fuels are being used, eventually enabling their complete replacement. However, it became obvious that hydrogen from fossil fuels cannot be used to replace fossil fuels, and it was not politically opportune, particularly not in the United States, until recently, to advocate for hydrogen exclusively from renewable energy. 2. The second one is the lack of hydrogen infrastructure and renewable hydrogen. The sales of hydrogen fuel cell vehicles were limited only to the areas that have hydrogen refueling infrastructure (California, Japan, Korea, and parts of Europe). There was, and still is, a classical chickenand-egg problem—no one will build a hydrogen refueling station unless there are hundreds of vehicles that will refuel there, and no one will buy a hydrogen-powered vehicle unless there are hydrogen refueling stations already in place where he/she lives and/or works. Even bigger problem was that many of the hydrogen refueling stations did not sell hydrogen from renewable energy, but hydrogen produced from natural gas. Hydrogen from natural gas is currently several times cheaper than hydrogen from renewable energy. Such hydrogen did not contribute to reducing CO2 emissions and was immediately labeled as greenwashing. True,

16

Fuel Cells for Transportation

the same applies to the BEVs that use grid electricity for recharging, so the CO2 emissions reduction depends on the electricity mix in the grid. 3. The third reason is that battery technology developed faster and conveniently piggybacked on the existing infrastructure—the electrical grid. It has been clear from the very beginning that BEVs have an advantage over FCEVs for small power, light vehicles, and short distances, that is, for city driving. However, the development of battery technology and BEVs has pushed this breakeven point further toward heavier vehicles and longer distances. Today’s BEVs already have a comparable range to FCEVs, that is, 400
500 km. The sales of BEVs worldwide have greatly outnumbered the sales of FCEVs. Although the infrastructure for BEVs also yet needs to be established, for initial penetration of BEVs required only the chargers which used the existing electric grid infrastructure. It is clear that this infrastructure will not be sufficient if the majority of the cars in the future are to be electric, but it was more than adequate for initial penetration. FCEVs still have two major advantages over BEVs—the higher energy density (in Wh/kg) and shorter duration of refueling. The energy density of the hydrogen tank plus the fuel cell is still about twice that of the best available batteries. Refueling a hydrogen tank takes a few minutes, while recharging the batteries takes an order of magnitude or longer. This advantage is still significant for heavy-duty transport such as buses, coaches, and trucks, but also for any other transportation mode that cannot afford hours of recharging (such as taxis and delivery vehicles). For city buses, a detailed analysis of a route (i.e., configuration of terrain, number of stops, number of passengers, possibility of charging at the stops, allowed time at the last stop(s), etc.) is required to determine the best mode of electrification, usual criterion being the total cost of ownership. Requirements for cars, buses, and trucks—For passenger cars, fuel cell power systems must be as durable and reliable as current automotive engines. The U.S. DOE targets durability of 5000 hours with 150,000 miles of driving life and less than 10% loss of performance. However, the ultimate target is to have 8000 hours of durability with 150,000 miles of driving life on a lower average-speed drive cycle. Also, fuel cell power systems should be able to function over a broad range of external environmental conditions (240 C to 140 C) and respond to the rapid variations in power demand required in automotive applications. For buses and heavy-duty vehicles, the requirement of cost, weight, and volume is less stringent, which makes the implementation of fuel cell powertrains less challenging in transit buses and heavy-duty trucks than in other transportation applications. For transit buses, the current requirement of a fuel cell power plant is 18,000 hours of lifetime. In addition, the voltage degradation should be less than 20%, as it leads to a loss of power performance and the ability to complete routes, including hills,

Fuel cells for transportation—an overview Chapter | 1

17

and leads to declining fuel economy and higher operating costs. The bus lifetime should also be at least 12 years, or 500,000 miles, and the cost should be less than $1 million. For heavy-duty trucks, a longer lifetime and increased efficiency of fuel cell power systems are required to compete with existing Class 8, long-haul trucks and trailers. It is expected that the fuel cell system’s lifetime to be 25,000 hours for heavy-duty trucks. The ultimate target is to have 30,000 hours of durability with over 1.0 million miles of driving life (assuming an average driving speed of 40 miles/hour). The interim and ultimate targets for peak efficiencies of fuel cell power systems are 68% and 72%, respectively. A summary of FCEV’s targets and lifetimes is given in Table 1.1. Emerging transportation sectors for FCEVs—While the use of fuel cells for cars, buses, and trucks is growing, other transportation markets are emerging. These include rail transport, aviation, and marine technology. Diesel engines are still the most efficient freight locomotives, as they consume a significant amount of fuel at high-notch levels. However, fuel cells will be efficient for yard switchers and suburban passenger trains, as they operate with frequent start
stop and acceleration events and at low-notch levels. Moreover, there are tightening standards for GHG emissions from rail engines that encourage rail operators to switch to environmentally friendly options for their future fleets. However, the durability of the fuel cell powertrain for rail transport could be a major challenge, as the requirements are 10
15 years of lifetime and 35,000 hours of stack durability with 3.6 million miles of driving life [43,44]. The International Maritime Organization (IMO) is driving aggressive emissions reduction in the shipping industry by adopting hydrogen fuel cell propulsion systems as well as fuel cell auxiliary power systems for ships. The IMO aims to cut down CO2 emissions by 50% by 2050, and hydrogen fuel cells can enable them to achieve their target. However, the potential use of fuel cells in marine applications will be limited to small ships and ferries due to the high-power demand for medium and large container ships. As the emission advantages of fuel cells are much greater than marine diesel engines, it is expected that fuel cells could become a viable option for auxiliary power for medium and large container ships. For instance, adding a 3 MW hydrogen fuel cell system to an existing large-scale container ship would deliver 43% of auxiliary power demand and reduce 2300 tonnes of CO2 emissions annually [45]. Like fuel cells for rail transport, the durability of the fuel cell systems will be a major challenge for marine applications. It is expected that the requirements for marine applications would be 25 years of lifetime and 75,000
100,000 hours of stack durability [44]. Hydrogen fuel cells are also being explored as a power source for aviation. Airbus is pursuing hydrogen-fueled airliners and has projected that its first hydrogen airliners may enter service in 2035. However, it is too ambitious to use a fuel cell stack for long-haul commercial flights. The possibility

TABLE 1.1 Key target metrics for performance, cost, and durability. Cost (US$/kW)

Peak efficiency (%)

System lifetime (h)

Durability (thousand miles)

Interim target

Ultimate target

Interim target

Ultimate target

Interim target

Ultimate target

Interim target

Ultimate target

Car

40

30

65

70

5000

8000

150

150

Bus

40

30

65

70

25,000

30,000

500

500

Truck

80

60

68

72

25,000

30,000

1000

1200

Source: U.S. Department of Energy. DOE. DOE Hydrogen and Fuel Cells Program Record. ,https://www.hydrogen.energy.gov/program_records.html. (accessed 23.01.22).

Fuel cells for transportation—an overview Chapter | 1

19

of using a fuel cell stack will be limited to short-range and small commercial aircraft. Like fuel cells in large-scale container ships, CO2 emissions from passenger aircraft can be significantly reduced by using a fuel cell stack for auxiliary power, reducing overall jet fuel consumption. The key requirements for the fuel cell stack for rail transport, aviation, and marine technology are listed in Table 1.2.

1.6

Future of hydrogen and fuel cells

The rollout of hydrogen vehicles must be accompanied by a suitable hydrogen supply infrastructure, which is to date available, at least to some extent, only in a few countries. In addition, green hydrogen, that is, hydrogen produced from renewable energy sources, must be available in sufficient quantities and at an affordable price. It appears that green hydrogen now has political support at the highest levels, at least in Europe, and necessary strategies have been put in place. The new Hydrogen Roadmap Europe [46] shifts the attention from automotive applications to hard-to-decarbonize sectors such as energy-intensive industries, trucks, aviation, shipping, and heating applications in buildings and industry, where hydrogen is the best or only choice for large-scale decarbonization. Specifically: 1. In transport, hydrogen is the most promising decarbonization option for trucks, buses, ships, trains, large cars, and commercial vehicles, where the lower energy density (hence lower range) and slow recharging performance of batteries are major disadvantages. A study by Research Center J¨ulich [47] shows that at high penetration of electric vehicles (20 million) the cost of both battery charging and hydrogen refueling infrastructures would be comparable, and concludes that a smart and complementary combination of the electric charging and the hydrogen refueling infrastructure may be optimal. In aviation, hydrogen and synthetic fuels based on hydrogen are the only at-scale option for direct decarbonization. 2. Industry can burn hydrogen to produce high-grade heat and use the fuel in several processes as feedstock, either directly or together with CO2 as TABLE 1.2 Key requirements for fuel cell stack for rail transport, aviation, and marine technology [44]. Average mileage (miles per day)

Power output (MW)

Rails

800
1000

1.5
3

Aircraft

1000
1200

0.1
8

Ships

800
1000

4
60

20

Fuel Cells for Transportation

synfuel/electrofuel. In steelmaking, for example, hydrogen can work as a reductant, substituting coal-based blast furnaces. When used as a feedstock for ammonia production and hydrotreating in refineries, it should be produced from renewable energy. 3. The decarbonization of natural gas, primarily used for heating, could be achieved with hydrogen produced from renewable energy injected into the existing natural gas grid without the need for major upgrades. Biogas, while an important lever, is unlikely available at the required scale. Electrification with heat pumps can replace natural gas to heat new buildings but requires costly or even impossible retrofits in old buildings. Full direct electrification would also lead to major seasonal imbalances in power demand that would, in turn, require a power storage mechanism at a large scale. Ultimately, energy suppliers can convert grids to run on pure hydrogen. Alternatively, natural gas can be replaced with synthetic natural gas produced from hydrogen and CO2. All gas-based heating systems can increase energy efficiency through the use of fuel cell-based combined heat and power technology. The EU Hydrogen Strategy gives a boost to clean hydrogen production in Europe [48]. Hydrogen can be used as a feedstock, a fuel, or an energy carrier and storage, as shown in Fig. 1.7, and has many possible applications which would reduce GHG emissions across industry, transport, power, and buildings sectors. In the first phase, from 2020 up to 2024, the strategic objective is to install at least 6 GW of electrolyzers in the EU and the

FIGURE 1.7 Role of hydrogen in the decarbonized energy system [49]. Credit: Hydrogen Council (https://hydrogencouncil.com) and U.S. Department of Energy (https://www.energy.gov).

Fuel cells for transportation—an overview Chapter | 1

21

production of up to 1 million tonnes of renewable hydrogen, to decarbonize existing hydrogen production, for example, in the chemical sector and facilitating take-up of hydrogen consumption in new end-use applications such as other industrial processes and possibly in heavy-duty transport. In a second phase, from 2025 to 2030, hydrogen needs to become an intrinsic part of an integrated energy system with a strategic objective to install at least 40 GW of electrolyzers by 2030 and the production of up to 10 million tonnes of renewable hydrogen in the EU. In the third phase, from 2030 onward and toward 2050, renewable hydrogen technologies should reach maturity and be deployed at a large scale to reach all hard-to-decarbonize sectors where other alternatives might not be feasible or have higher costs. In this phase, renewable electricity production needs to massively increase as about a quarter of renewable electricity might be used for renewable hydrogen production by 2050. European Hydrogen Roadmap [46] envisions that by 2040 one in sevenpassenger vehicles sold will be FCEVs, there will be 450,000 electric buses and trucks with fuel cell range extenders on the road, 2000 diesel trains replaced with hydrogen, and 2.5 million small fuel cell cogeneration units. By 2050, 39% of large fleet cars and taxis, 22% of trucks and buses, 30% of light commercial vehicles, and 9% of trains will run on hydrogen, and 23% of high-grade heat in industry will be supplied by hydrogen. This represents a significant step up from the present status of the deployment of hydrogen technologies. Although there are numerous demonstration, precommercial, and commercial projects already in place, they are concentrated in a few countries with strong industries supporting those projects. Similar plans exist in other countries such as China, Japan, and South Korea but not in the United States, as shown in Table 1.3. The capital costs of FCEVs are still higher than for BEVs and ICE vehicles. However, fuel cell capacity and hydrogen tanks come increasingly at a much lower cost than adding battery capacity and BEVs. FCEVs also have comparable sun-to-wheel efficiencies in a systemic view, as grid electricity for charging a BEV needs to be produced relatively close to its location of usage, while hydrogen can be shipped across long distances. For instance, a BEV charged using a local PV panel (1 kW) with an 11% load factor and an FCEV powered with imported renewable hydrogen from a PV panel (1 kW) in sunny regions (25% load factor) results in a comparable total output even with lower tank-to-wheel efficiency of FCEV [50]. Moreover, the cost of hydrogen is decreasing due to more initiatives for providing green hydrogen, and the new hydrogen infrastructure will be built as there is a worldwide push for the hydrogen economy. Given these scenarios, the Hydrogen Council expects that the market share of FCEVs will grow in the coming decades, and by 2050, hydrogen fuel cells can power a global fleet of more than 400 million cars, 15
20 million trucks, and around 5 million buses [51]. Hydrogen-powered trains could also replace about 20% of diesel trains

22

Fuel Cells for Transportation

TABLE 1.3 Sales versus national hydrogen strategy targets [41]. Sales (thousand units)

2017

2018

2019

2020

Through June 2021

50,000 FCEVs by 2025




1.79

6.18

8.44

8.44

G

3.7 million FCEVs, 500,000 light-duty commercial vehicles, 45,000 trucks and buses by 2030

1.19

1.42

2.18

2.67

3.08

G

200,000 fuel cell electric vehicles (FCEVs) (2025) 800,000 FCEVs (2030) 1200 fuel cell (FC) buses (2030) 10,000 FC forklifts (2030)

2.30

2.93

3.63

4.20

5.60

2.9 million FC cars domestic, 3.3 million FC cars exported (2040), 100,000 units by 2025, 81,000 units by 2022 80,000 FC taxis (2040) 40,000 FC buses (2040) 30,000 FC trucks (2040)




0.90

5.08

10.09

14.56

No strategy in place

3.53

5.90

8.04

9.25

11.12










0.15

0.34

Country

Transport targets in national strategy

China

G

Europe

Japan

G

G

G

South Korea

G

G

G

G

United States

G

Rest of world

N/A

Source: Compiled by Transport Energy Strategies citing data from IEA, countries’ national strategies, May 2022.

Fuel cells for transportation—an overview Chapter | 1

23

as well as 5% of the world’s fuel supply to airplanes and freight ships by 2050. A summary of the Hydrogen Council’s projected stock of FCEVs in 2050 as a percentage of the total fleet by vehicle type is depicted in Fig. 1.8. According to the U.S. Department of Energy, when using hydrogen produced from natural gas, FCEVs are expected to have well-to-wheels GHG emissions of less than half that of current gasoline-powered vehicles. This was also confirmed by the case study done by the DNV GL [52]. Fig. 1.9 shows the results of the case study done by the DNV GL on the lifecycle emissions of FCEVs with those of gasoline and diesel ICE vehicles, and BEVs. Two cases of BEVs and FCEVs were considered. For BEVs, it was assumed that the BEV, including its battery, is manufactured using only renewable electricity or from an electricity mix with an equivalent carbon footprint of 345 kg-CO2e/MWh. For FCEVs, it was assumed that the hydrogen is produced with a carbon footprint of 4 or 12 kg-CO2e/kgH2, corresponding to hydrogen produced by SR with and without CCS, respectively.

FIGURE 1.8 Projected stock of FCEVs as a percentage of total fleet by vehicle type in 2050 [51]. Data are from the Hydrogen Council (https://hydrogencouncil.com).

24

Fuel Cells for Transportation

FIGURE 1.9 Lifecycle CO2 emission for internal combustion engine vehicles (gasoline and diesel), FCEV with carbon capture and storage (FCEV-CCS), FCEV with steam reforming (SR), battery electric vehicle with green electricity (BEV-Green), and battery electric vehicle manufactured and running with electricity generated from mixed sources (BEV-Mix). Data do not include emissions from recycling and disposal, and the lifetime of a vehicle is assumed to be 120,000 km. Data are from DNV GL Research Review J. Aarnes, M. Eijgelaar, E.A. Hektor, Hydrogen as an energy carrier—an evaluation of emerging hydrogen value chains, DNV GL Research Review, 2018.

It is clear from Fig. 1.9 that the per-km lifecycle emission of FCEVs running on hydrogen produced by SR with CCS (FCEV-CCS) is 89 g/km, whereas a BEV running on green electricity results in 116 g/km of CO2 emissions. Conversely, an FCEV running on hydrogen produced without CCS does not provide any significant benefit compared with gasoline and diesel ICE vehicles. Thus, FCEVs with CCS have higher decarbonization potential for the transport sector.

1.7

Conclusions

The future of fuel cells, including the fuel cells for transportation, is tightly related to the energy transition that is already taking place. Energy transition, with the goal of decarbonizing the energy sector by 2050, is unavoidable if the world is to keep the global temperature increase within 2 C. In such a transition, hydrogen produced from renewable energy sources would enable the decarbonization of otherwise hard-to-decarbonize sectors. Fuel cells are the technology that will enable hydrogen’s use in transportation. In automobiles, FCEVs will have tough competition with BEVs. The key advantage of FCEVs is the significantly shorter duration of refueling.

Fuel cells for transportation—an overview Chapter | 1

25

Therefore, they will be appealing to those who prefer to keep the present convenience of vehicles with ICEs or those who cannot afford long refuelings, such as taxis and delivery vehicles. The trucks need to store significantly more energy than automobiles, so the higher energy density of fuel cell hydrogen power trains may be a decisive advantage. Fuel cell hydrogen trucks show substantial market potential at scale as they are one of the most promising zero-emission alternatives for trucking. A study on fuel cell hydrogen trucks commissioned by Fuel Cells and Hydrogen Joint Undertaking [53] showed heavy-duty fuel cell hydrogen trucks provide the operational performance most comparable to diesel trucks regarding daily range, refueling time, and payload capacity, but at a better total cost of ownership than the incumbent technology from 2027 onward. Preconditions are (1) scaled-up production of FCH trucks and that (2) hydrogen is offered below 6 EUR/kg. At this scale, the study projects a potential FCH sales share of approximately 17% of new trucks sold in 2030 in a base scenario (about 60,000 trucks). Fuel cell buses offer the best productivity and operational flexibility compared to other zero-emission concepts. FC buses use power from a fuel cell stack and a battery and run on hydrogen which can be stored and refueled at bus depots. In terms of costs, it is expected that FC buses compare similarly with other zero-emission powertrains in the long run. However, they are superior in terms of operational performance: with ranges of 300
450 km, refueling times below 10 minutes, and no infrastructure requirements on the routes, FC buses can be operated like conventional diesel buses while offering all the above-mentioned advantages of electric vehicles. A study on the use of fuel cells and hydrogen in the railway environment, commissioned by the Fuel Cells and Hydrogen Joint Undertaking and Shift2Rail Joint Undertaking, showed a good potential for fuel cells in the railway environment for the replacement of diesel trains [54]. Some of the cases evaluated already show a positive total cost of ownership (TCO) for fuel cells, while in others this technology is recognized as the most adequate zero-emission alternative: 1. Multiple units passenger trains are seen as the application closest to the market, but projects involving few trains or low daily mileage will have problems reaching appropriate TCO due to the infrastructure costs; the study showed that for the non-electrified lines longer than 100
200 km with less than 10 runs daily the fuel cell hydrogen trains are already a viable option. 2. Shunter locomotives need further technological development, but there are circumstances where they could be already competitive with diesel. 3. Freight locomotives have a more difficult economic justification although if catenary electrification is not an option, fuel cells would be the only zero-emission option.

26

Fuel Cells for Transportation

Hydrogen and hydrogen-based fuels (such as ammonia) indeed offer tremendous potential for the decarbonization of the worldwide maritime fleet, but only providing a significant carbon tax is imposed. The analysis [55] took into consideration the size and operational profile of the ships, the costs of fuel, the costs of the required onboard equipment, and the cost associated with the loss of cargo space due to the size of fuel storage. The three options that came out as the most cost-efficient are: 1. Compressed hydrogen with PEMFC for relatively small ships with an operational profile that allows for frequent refueling; 2. Ammonia with solid oxide fuel cell for deep-sea shipping applications or smaller vessels with high-value cargo (e.g., chemical tankers); 3. Liquefied hydrogen with PEMFC for every ship in between. A fact-based study of hydrogen-powered aviation considering hydrogen technology, economics, and climate impact by 2050, commissioned by the CleanSky and Fuel Cells and Hydrogen Joint Undertakings [56], found that hydrogen—as a primary fuel for propulsion, either for fuel cells, direct burn in gas turbines, or as a building block for synthetic liquid fuels—could feasibly power aircraft with entry into service by 2035 for short-range aircraft. Overall, it took a long time for fuel cell technology to come from invention to practical applications. Technology is ready, and political decisions have been made, so in the next decades, fuel cells will be widely used in transportation—not only in automobiles, but also in delivery vehicles, trucks, buses, coaches, trains, and even ships and airplanes.

References [1] IEA, Tracking transport. https://www.iea.org/reports/tracking-transport-2020, 2020 (accessed 23.01.22). [2] IEA, Global energy review: CO2 emissions in 2020. https://www.iea.org/articles/globalenergy-review-co2-emissions-in-2020, 2020. (accessed 23.01.22). [3] IEA, Energy technology perspectives. https://www.iea.org/reports/energy-technology-perspectives-2020, 2020 (accessed 23.01.22). [4] Preliminary US greenhouse gas emissions estimates for 2021. https://rhg.com/research/preliminary-us-emissions-2021/, 2022. (accessed 23.01.22). [5] Y. Wang, et al., PEM Fuel cell and electrolysis cell technologies and hydrogen infrastructure development—a review, Energy Environ. Sci. 15 (6) (2022) 2288
2328. [6] Y. Wang, et al., Polymer electrolyte membrane fuel cell and hydrogen station networks for automobiles: status, technology, and perspectives, Adv. Appl. Energy 2 (2021) 100011. [7] A.Z. Weber, S. Balasubramanian, P.K. Das, Chapter 2—Proton exchange membrane fuel cells, in: K. Sundmacher (Ed.), Advances in Chemical Engineering, in: 41, Academic Press, 2012, pp. 65
144. [8] A.Z. Weber, et al., A critical review of modeling transport phenomena in polymerelectrolyte fuel cells, J. Electrochem. Soc. 161 (12) (2014) F1254
F1299. [9] J. Verne, The Mysterious Island, Pierre-Jules Hetzel, France, 1875.

Fuel cells for transportation—an overview Chapter | 1

27

[10] J.O.M. Bockris, Energy: The Solar-Hydrogen Alternative, Australia and New Zealand Book Co, Sydney, 1975. [11] J.O.M. Bockris, T.N. Veziroglu, A solar-hydrogen energy system for environmental compatibility, Environ. Conserv. 12 (2) (1985) 105
118. [12] D.S. Scott, Smelling Land: The Hydrogen Defence Against Climate Catastrophe— Enhanced Edition, Queen’s Printer, Victoria, BC, 2008. [13] C.F. Schœnbein, On the voltaic polarization of certain solid and fluid substances, Lond. Edinb. Dublin Philos. Mag. J. Sci. 14 (85) (1839) 43
45. [14] W.R. Grove, On voltaic series and the combination of gases by platinum, Lond. Edinb. Dublin Philos. Mag. J. Sci. 14 (86
87) (1839) 127
130. [15] U. Bossel, The Birth of the Fuel Cell, 1835
1845, European Fuel Cell Forum, Oberrohrdorf, Switzerland, 2000. [16] W.R. Grove, On a gaseous voltaic battery, Lond., Edinb., Dublin Philos. Mag. J. Sci. 21 (140) (1842) 417
420. [17] W.R. Grove, On the gas voltaic battery—voltaic action of phosphorus, sulphur and hydrocarbons, Philos. Trans. R. Soc. Lond. 135 (1845) 351
361. [18] G. Sandstede, et al., History of low temperature fuel cells, in: W. Vielstich et al. (Eds.), Handbook of Fuel Cells, 2010. [19] E. Chen, in: G. Hoogers (Ed.), Chapter 2: History, in Fuel Cell Technology Handbook, CRC Press, Boca Raton, 2003. [20] https://americanhistory.si.edu/fuelcells/origins/origins2.htm. [21] W. Ostwald, Die wissenschaftliche Elektrochemie der Gegenwart und die technische der Zukunft, Z. Phys. Chem. 15U (1) (1894) 409
421. [22] M. Kendall, K. Kendall, A.P.B. Lound, HYSTORY: The Story of Hydrogen, Adelan, London, 2021. [23] https://www.freeenergyplanet.biz/fuel-cell-technology/grubb-and-niedrach-at-general-electric.html. [24] W.G. Grot, in: U. Landau, E. Yeager, D. Kortan (Eds.), Nafion Membrane and its Applications, in Electrochemistry in Industry: New Directions, Springer US, Boston, MA, 1982, pp. 73
87. [25] https://www.freeenergyplanet.biz/fuel-cell-technology/ballard-power-systems.html. [26] M. Nadal, F. Barbir, Development of a hybrid fuel cell/battery powered electric vehicle, Int. J. Hydrog. Energy 21 (6) (1996) 497
505. [27] F. Barbir, M. Nadal, M. Fuchs, Fuel cell powered utility vehicles, in: Portable Fuel Cell Conference, Lucerne, Switzerland, 1999. [28] https://georgewbush-whitehouse.archives.gov/news/releases/2003/01/20030130-20.html. [29] https:/en.wikipedia.org/wiki/fuel_cell_vehicle. [30] IEA, Hydrogen https://www.iea.org/reports/hydrogen, 2022 (accessed 23.01.22). [31] I.V. Zenyuk, P.K. Das, A.Z. Weber, Understanding impacts of catalyst-layer thickness on fuel-cell performance via mathematical modeling, J. Electrochem. Soc. 163 (7) (2016) F691
F703. [32] P.K. Das, X. Li, Z.S. Liu, Analysis of liquid water transport in cathode catalyst layer of PEM fuel cells, Int. J. Hydrog. Energy 35 (6) (2010) 2403
2416. [33] K. Jiao, X. Li, Water transport in polymer electrolyte membrane fuel cells, Prog. Energy Combust. Sci. 37 (3) (2011) 221
291. [34] Y. Wang, K.S. Chen, S.C. Cho, PEM Fuel Cells: Thermal and Water Management Fundamentals, Momentum Press, 2013.

28

Fuel Cells for Transportation

[35] B.S. Machado, N. Chakraborty, P.K. Das, Influences of flow direction, temperature and relative humidity on the performance of a representative anion exchange membrane fuel cell: a computational analysis, Int. J. Hydrog. Energy 42 (9) (2017) 6310
6323. [36] D.R. Dekel, Review of cell performance in anion exchange membrane fuel cells, J. Power Sources 375 (2018) 158
169. [37] S. Gottesfeld, et al., Anion exchange membrane fuel cells: current status and remaining challenges, J. Power Sources 375 (2018) 170
184. [38] B.S. Machado, et al., A three-dimensional agglomerate model of an anion exchange membrane fuel cell, J. Electrochem. Energy Convers. Storage 15 (1) (2018) 011004. [39] X. Li, Principles of Fuel Cells, Taylor & Francis, New York, 2006. [40] P.K. Das, X. Li, Z.S. Liu, Analytical approach to polymer electrolyte membrane fuel cell performance and optimization, J. Electroanal. Chem. 604 (2) (2007) 72
90. [41] https://www.transportenergystrategies.com/2022/05/03/hydrogen-targets-v-actual-salestheres-a-gap. [42] DOE, DOE hydrogen and fuel cells program record, https://www.hydrogen.energy.gov/ program_records.html, 2022. (accessed 23.01.22). [43] R.K. Ahluwalia et al. Total cost of ownership for line haul, yard switchers and regional passenger locomotives—preliminary results, in: H2@Rail Workshop, 2019. [44] D.A. Cullen, et al., New roads and challenges for fuel cells in heavy-duty transportation, Nat. Energy 6 (5) (2021) 462
474. [45] C.J. McKinlay, S.R. Turnock, D.A. Hudson, Fuel cells for shipping: to meet on-board auxiliary demand and reduce emissions, Energy Rep. 7 (2021) 63
70. [46] Hydrogen Roadmap Europe: A Sustainable Pathway Toward European Energy Transition, Fuel Cells and Hydrogen 2 Joint Undertaking, Brussels, 2019. [47] M. Robinius, et al., Comparative Analysis of Infrastructures: Hydrogen Fueling and Electric Charging of Vehicles. Schriften des Forschungszentrums J¨ulich Reihe Energie & Umwelt/Energy & Environment, Vol. 408, J¨ulich: Forschungszentrum J¨ulich GmbH Zentralbibliothek, Verlag, 2018. [48] A hydrogen strategy for a climate-neutral Europe. 2020, Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COMM, 2020. [49] Hydrogen-Council-Vision-Document.pdf (hydrogencouncil.com). [50] Roadmap towards zero emissions https://hydrogencouncil.com/en/roadmap-towards-zeroemissions-bevs-and-fcevs/. 2021. [51] Hydrogen—scaling up: a sustainable pathway for the global energy transition. https:// hydrogencouncil.com/wp-content/uploads/2017/11/Hydrogen-scaling-up-HydrogenCouncil.pdf, 2017, (accessed 23.01.22). [52] J. Aarnes, M. Eijgelaar, E.A. Hektor, Hydrogen as an energy carrier—an evaluation of emerging hydrogen value chains, DNV GL Research Review, 2018. [53] Y. Ruf et al., Study on Fuel Cells Hydrogen Trucks Commissioned by Fuel Cells and Hydrogen Joint Undertaking, and Conducted by Roland Berger, 2020. [54] Y. Ruf et al., Use of fuel cells and hydrogen in the railway environment, in: A Study for the Fuel Cells and Hydrogen Joint Undertaking and Shift2Rail Joint Undertaking, 2019. [55] G. Pawelec, M. Muron, Techno-Economic Assessment of Low-Carbon Hydrogen Technologies for the Decarbonization of Shipping, 2021. [56] Hydrogen-powered aviation, A fact-based study of hydrogen technology, economics, and climate impact by 2050, in: CleanSky2 and Fuel Cells and Hydrogen Joint Undertaking 2, 2020.

Chapter 2

Fuel cell fundamentals Lei Xing1, Jin Xuan1 and Prodip K. Das2 1

School of Chemistry and Chemical Engineering, University of Surrey, Guildford, United Kingdom, 2School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom

2.1

Introduction

Fuel cells are electrochemical devices that convert the chemical energy of fuels, for example, hydrogen or methanol, directly into direct current (DC) electricity. Unlike the traditional internal combustion engine, fuel cells produce electrical energy through an electrochemical reaction, rather than through combustion. A single fuel cell unit consists of a pair of anode and cathode and an electrolyte in between. Individual fuel cells can be connected in series to form a fuel cell stack, which can generate higher power for portable and stationary applications. Depending on different electrolytes, reactants, and operating temperatures, fuel cells are mainly categorized into six groups, namely, alkaline fuel cells, phosphoric acid fuel cells, molten carbonate fuel cells, solid oxide fuel cells, direct methanol fuel cells, and polymer electrolyte fuel cells (PEFCs). Based on the use of electrolytes, PEFCs are subdivided into two categories: proton-exchangemembrane fuel cells (PEMFCs) or anion-exchange membrane fuel cells. Among these fuel cells, PEMFCs are best suited for transport and small-scale stationary applications. Thus, great attention has been given to PEMFCs operated with perfluorosulfonic acid-based membranes and hydrogen to combat the increasing global energy consumption and environmental pollution caused by fossil fuelbased internal combustion engines. The electrochemical and physical phenomena that occurred in PEMFCs, including the performance curves, profiles of the reactant and product species, velocities and pressure of reactant gases, as well as the temperature distributions, are determined by the fully coupled electrochemical kinetics and transport of mass and heat. In this chapter, the fundamentals and operating principles, along with governing equations, of PEMFCs are given.

2.2 Operation principle of proton-exchange membrane fuel cells A typical PEMFC unit consists of a membrane electrode assembly (MEA) sandwiched between the flow field plates of the anode and cathode in which Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00006-X © 2023 Elsevier Ltd. All rights reserved.

29

30

Fuel Cells for Transportation

flow channels are machined, as highlighted in Fig. 2.1. The MEA includes gas diffusion layer (GDL), microporous layer (MPL), and catalyst layer (CL) at the anode and cathode, respectively, and a proton-exchange membrane (PEM) in between. At the anode, hydrogen flows into the anode flow channel and then transports to the CL through the GDL and MPL and then split into protons and electrons in the anode CL [Eq. (2.1)]. The protons pass through the PEM and reach the cathode CL, while the electrons travel via an external circuit to the cathode. At the same time, air or oxygen flows into the cathode flow channel and then transports to the CL through the GDL and MPL at the cathode. In the cathode CL, oxygen combines with the protons and electrons, which are generated in the anode CL, to produce water [Eq. (2.2)]. Anode reaction: H2 -2H1 1 2e2 Cathode reaction:

1 O2 1 2H1 1 2e2 -H2 O 2

Net cell reaction: H2 1

1 O2 -H2 O 2

ð2:1Þ ð2:2Þ ð2:3Þ

The thermodynamic potential for the net cell is calculated by the standard chemical potential at 25 C and 1.0 atm and is 1.23 V versus standard hydrogen electrode. In the case of a fuel cell supplied with reactant gases without the closed electrical circuit, the observed practical cell potential is called the open-circuit potential (OCP). The OCP is lower than the theoretical potential due to the activation losses (especially at the cathode) and mixed electrode potential losses in the fuel cell even when no external current is generated [1,2]. The mixed electrode potential arises due to the unavoidable crossover of fuel through the electrolyte from the anode to the cathode or vice versa. The relationship between fuel cell potential

FIGURE 2.1 Schematic diagram and basic operation principle of a proton-exchange membrane fuel cell. Each electrode includes a gas diffusion layer, a microporous layer, and a catalyst layer.

Fuel cell fundamentals Chapter | 2

31

FIGURE 2.2 Voltage losses in a proton-exchange membrane fuel cell and the resulting polarization curves [2,3].

and current density is called the polarization curve, which is obtained by subtracting the activation polarization losses, ohmic losses, and concentration polarization losses from the equilibrium potential. Fig. 2.2 shows the polarization curve and different voltage losses in a typical PEMFC [2,3]. Note that a majority of the voltage losses occur at the cathode due to the sluggish oxygen reduction reaction (ORR) [4].

2.3 2.3.1

Reaction kinetics and transport processes Electrode kinetics

2.3.1.1 Butler
Volmer kinetics Fuel cell operation is based on the electrochemical reactions occurring simultaneously at the anode and cathode, which are presented in Eqs. (2.1)
(2.3). The reaction rate of an electrochemical reaction is defined as the speed of the electrochemical reaction proceeds on the electrode surface. Electrical current is generated by the electrons released and consumed in the electrochemical reaction processes. Current density is the current per unit surface area. According to Faraday’s law, the current density is proportional to the charge transferred and the consumption of reactants per unit area: i 5 nFNi

ð2:4Þ

32

Fuel Cells for Transportation

where n is the amount of substance, F (96385 C mol21) is Faraday’s constant, Ni (mol m22 s21) is the flux of reactant per unit area, and i (A m22) is the current density. An electrochemical reaction involves either oxidation or reduction of reactant species. Typically, the oxidation process produces electrons while the reduction process consumes the electrons as follows: Rd2Ox 1 ne2

ð2:5Þ

On an electrode at equilibrium conditions (no external current is generated), both the oxidation and reduction processes occur at equal rates. The net current generated is the difference between the electrons released and consumed: i 5 nFðkf cOx 2 kb cRd Þ

ð2:6Þ

where kf and kb (s21) are the forward (reduction) and backward (oxidation) reaction rate coefficients, and cOx and cRd (mol m22) are the surface concentrations of the oxidized and reduced species, respectively. The reaction rate coefficient for an electrochemical reaction is a function of the Gibbs free energy.
 kB T 2 ΔG exp kf =b 5 ð2:7Þ h Rg T (1.38 3 10
23 J K21) is Boltzmann’s constant, h where kB 234 (6.626 3 10 J s) is Planck’s constant, Rg (8.314 J mol21 K21) is the universal gas constant, T (K) is the temperature, and ΔG (J mol21) is the Gibbs free energy, which is considered to consist of both chemical and electrical terms. Consequently, for a reduction reaction: ΔG 5 ΔGch 1 αRd Fϕ

ð2:8Þ

and for an oxidation reaction: ΔG 5 ΔGch 2 αOx Fϕ 21

ð2:9Þ

where ΔGch (J mol ) is the Gibbs free energy of the chemical component, ϕ (V) is the potential, and αRd and αOx are the transfer coefficients for reduction and oxidation reactions, respectively. Note that in the case of multistep electrochemical reactions, αRd 1 αOx is generally equal to n/v, where n is the number of electrons transferred in the overall reaction, and v is the stoichiometric number. The forward and backward reaction rate coefficients in Eq. (2.6) are represented, respectively:
 2 αRd Fϕ kf 5 k0; f exp ð2:10Þ Rg T

Fuel cell fundamentals Chapter | 2

kb 5 k0;b exp


 αOx Fϕ Rg T

Therefore the net current expression is obtained: 

 2 αRd Fϕ αOx Fϕ 2 k0;b cRd exp i 5 nF k0; f cOx exp Rg T Rg T

33

ð2:11Þ

ð2:12Þ

At equilibrium, the forward and backward reactions proceed simultaneously, generating zero net currents. In this case, the current densities for both forward and backward reactions are equal.

 2 αRd Fϕeq 2 αOx Fϕeq i0 5 nFk0; f cOx exp 5 k0;b cRd exp ð2:13Þ Rg T Rg T where i0 (A m22) is the exchange current density, k0, f and k0,b (s21) are the reaction rate constants of the forward and backward reactions at standard conditions (25 C and atmospheric pressure), respectively, and ϕeq (V) is the equilibrium or reversible potential. The equilibrium potentials of hydrogen oxidation reaction (HOR) and ORR are 0 and 1.229 V at standard conditions. By combining Eqs. (2.12) and (2.13), a relationship between the current density and the overpotential is obtained, which is known as the Butler
Volmer (BV) equation. 

 2 αRd Fη αOx Fη i 5 i0 exp 2 exp ð2:14Þ Rg T Rg T where η (V) is the overpotential, defined as the difference between the electrode potential and the equilibrium potential. The BV equation is valid for both anode and cathode reactions in a PEMFC: 

 2 αRd;a Fηa αOx;a Fηa ia 5 i0;a exp 2 exp ð2:15Þ Rg T Rg T 

 2 αRd;c Fηc αOx;c Fηc ic 5 i0;c exp 2 exp ð2:16Þ Rg T Rg T The anode overpotential is positive, which makes the first term in the bracket of Eq. (2.15) negligible and results in a negative sign of the anode current density obtained. Similarly, the cathode overpotential is negative, which makes the second term in the bracket of Eq. (2.16) negligible and leads to a positive sign of the cathode current density.

2.3.1.2 Agglomerate kinetics The electrochemical reactions, in which electrons, protons, and gas are involved, only take place on the surface of the catalyst. Reactant gas

34

Fuel Cells for Transportation

transports through the micropores within the CL, electrons travel through the electrically conductive solid, and protons migrate through the electrolyte. According to the functions of the different components within the CL, the existence of triple-phase boundaries (TPBs) is essential, where the content of electrolyte (for proton transfer), void space (for gas transport), and platinum (Pt)-dispersed carbon (for catalysis and electron transfer) are interacted, as shown in Fig. 2.3. A more accurate description of the CL structure based on the spherical agglomerate model is explained below, where each agglomerate is assumed to consist of three components: platinum dispersed on carbon (Pt/C), ionomer, and void space [5]. The intra-agglomerate void space is defined as the primary pores, and the void space between the agglomerate is defined as the secondary pores. As schematically represented in Fig. 2.4, the primary pores are partially occupied by ionomer and water, and the agglomerates are linked by the ionomer films. When the volume of the produced water exceeds the entire volume of the primary pores, a liquid film is formed surrounding the agglomerates. Reactant gas must dissolve in the ionomer/water film and diffuse through the primary pores before reaching the catalyst particles, which is taken into account in the classic BV relationship. Thus agglomerate kinetics is developed to consider the mass transport resistance in addition to electrochemical kinetics. Assuming that the concentrations of the dissolved species at the outer and inner boundary of the ionomer/water film are represented by ci,out and ci, 23 in (mol m ), respectively, the concentration of the dissolved species at the outer interface is described by Henry’s law as: pi ci;out 5 ð2:17Þ Hi where pi (Pa) and Hi (Pa m3 mol21) are the partial pressure and Henry’s constant of reactant species i, respectively.

FIGURE 2.3 A simplified schematic diagram of the triple-phase boundary in a catalystelectrolyte-pores interacted electrode.

Fuel cell fundamentals Chapter | 2

35

FIGURE 2.4 A schematic illustration of the catalyst layer based on the agglomerate assumption [5]. Credit: Reprinted from Xing et al. AIChE J. 63 (11) (2017) 4895
4910, with permission from AIChE.

The diffusion of reactant through the ionomer/water film can be described by Fick’s law: Ni 5 2 Deff i

@ci @r

ð2:18Þ

where Ni (mol m22 s21) is the reactants mole flux through the ionomer/water 2 21 film, Deff i (m s ) is the effective diffusion coefficient of reactants, ci (mol 23 m ) is the reactant concentration, and r (m) is the radius. It is assumed that the ionomer/water film is uniformly coated on the agglomerate, and the film thickness is much smaller than the agglomerate. Moreover, the molar rate is conserved in the ionomer/water film; namely, the molar rate is constant at the outer boundary of the ionomer/water film.

36

Fuel Cells for Transportation

Thus, the reactants mole flux through the ionomer/water film can be derived as C 5 2 ð4πr 2 ÞDeff i

@ci @r

ð2:19Þ

with the boundary conditions: r 5 ragg , ci 5 ci;in r 5 ragg 1 δ; ci 5 ci;out

ð2:20Þ

where C (mol s21) is a constant. Integrating Eq. (2.19) and applying the boundary conditions lead to the following expression: c 2 c  i;out i;in ð2:21Þ C 5 2 Deff 4πragg ðragg 1 δÞ i δ where ragg (m) is the radius of the agglomerate, and δ (m) is the thickness of the ionomer/water film. Combining Eqs. (2.18), (2.19), and (2.21), the reactants mole flux is: ragg ci;out 2 ci;in ð2:22Þ Ni 5 Deff i ðragg 1 δÞ δ According to mass balance, at a steady state, the amount of the species consumed equals to the species diffusion to the active surface. aagg Ni 5 Ri

ð2:23Þ

21

where aagg (m ) is the specific area of the agglomerate, defined as the surface area per agglomerate volume, and Ri (mol m23 s21) is the reaction rate. Assuming the reactions of either hydrogen oxidation or oxygen reduction as the first-order kinetics with respect to the reactant concentration, we get: Ri 5 kagg ci

ð2:24Þ

where kagg (s21) is the reaction rate coefficient representing the reactions that occurred in the agglomerate. By introducing the effectiveness factor into Eq. (2.24), the overall reaction rate only depends on the reactant concentration at the outer boundary of the agglomerate, obtained as: Ri;agg 5 Eagg kagg ci;out

ð2:25Þ

where Ri,agg (mol m23 s21) is the reaction rate based on the agglomerate volume, and Eagg is the effectiveness factor of the agglomerate, which represents the geometry of the agglomerate and the reactant mass transport resistance within the agglomerate. For the spherical agglomerates, the effectiveness factor is [6]:   1 1 1 2 Eagg 5 ð2:26Þ MT;agg tanhð3MT;agg Þ 3MT;agg

Fuel cell fundamentals Chapter | 2

where MT,agg is Thiele’s modulus, a dimensionless parameter [6]. sffiffiffiffiffiffiffiffiffiffiffi ragg kagg MT;agg 5 3 Deff i;agg

37

ð2:27Þ

2 21 where Deff i;agg (m s ) is the reactant effective diffusion coefficient inside the agglomerate. According to Faraday’s law, the volumetric current density is related to the reactants consumption rate via the following equation:

ii;agg 5 nFEagg kagg ci;in

ð2:28Þ

23

where ii,agg (A m ) is the volumetric current density based on the agglomerate volume, and the subscript i refers to anode or cathode, respectively. Combining Eqs. (2.22)
(2.25), the concentration of the dissolved species at the inner boundary of the ionomer/water film is obtained as:   Eagg kagg ðragg 1δÞδ 21 ci;in 5 11 ci;out ð2:29Þ aagg ragg Deff i Substituting Eq. (2.29) into Eq. (2.28) gives   ðragg 1δÞδ 21 1 ii;agg 5 nF 1 ci;out Eagg kagg aagg ragg Deff i According to Henry’s law, the above equation becomes   ðragg 1δÞδ 21 pi 1 ii;agg 5 nF 1 Hi Eagg kagg aagg ragg Deff i

ð2:30Þ

ð2:31Þ

The current density calculated by Eq. (2.14) could transfer to volumetric current density after being corrected by the specific area of the electrode. The agglomerate volumetric current therefore can be related to the Butler
Volmer kinetics as: 

 2 αRd Fη αOx Fη ii;agg 5 aagg i0 exp 2 exp ð2:32Þ RT RT On the inner boundary of the agglomerate, the intrinsic volumetric current density is obtained as: ii;agg 5 nFkagg ci;in

ð2:33Þ

Comparing the above two expressions, the reaction rate is obtained as: 

 aagg i0 2 αRd Fη αOx Fη kagg 5 exp 2 exp ð2:34Þ RT RT nFci;in

38

Fuel Cells for Transportation

2.3.1.3 Conservation of charge Assuming a finite volume dV (m3) carries an amount of charge dGv (C), the charge density gv (C m23) is: gv 5

dGv dV

where the total charge of the volume V is: ð ð dGv 5 dVgv

ð2:35Þ

ð2:36Þ

The change of total charge in time generates current flow in and/or out of the surface dS (m2) surrounding the volume dV. The expression of conservation of charge is, therefore: ð I d dVgv 5 2 dSUi ð2:37Þ dt where i (A m22) is the current density vector. By using the Gauss divergence theorem and assuming that the volume V did not change with time, the above equation can be rewritten as ð d dVgv 5 2 rUi ð2:38Þ dt Assuming the charge density at any location does not change with time, and adding a source term Qi (A m23) leads to rUi 5 Qi

ð2:39Þ

If no external source is applied, Qi is equal to the volume current density iv (A m23). The current density vector i can be related to the electric filed E (V m21) and conductivity σ (S m21) by Ohm’s law as: i 5 σUE

ð2:40Þ

At a specific point within the electric field, the voltage is equal to the negative gradient of the potential, thus E 5 2 rϕ

ð2:41Þ

Substituting Eqs. (2.40) and (2.41) into Eq. (2.39) gives rUi 5 rð2 σrϕÞ 5 Qi

ð2:42Þ

Applying Eq. (2.42) on the solid (electrode) and electrolyte (membrane and ionomer) phases, respectively, two sub-equations are obtained [7]: rUis 5 rð2 σeff s rϕs Þ 5 Qs

ð2:43Þ

Fuel cell fundamentals Chapter | 2

rUiM 5 rð2 σeff M rϕM Þ 5 QM

39

ð2:44Þ

21

eff where σeff s and σ M (S m ) are the effective electrical conductivity for the solid phase and electrolyte conductivity for the membrane and ionomer, respectively. For any volume of the computational domain, the electronic and the ionic currents generated are equal, leading to:

rUis 1 rUiM 5 0

ð2:45Þ

In a porous media, the source terms Qs and QM can be expressed in terms of the volumetric current for the electrode and electrolyte, respectively. The total current generated in the anode must be equal to the total current consumed in the cathode. In the situation of no external resource, the conservation of charge requires that: Qs 5 2 QM ð ia dV 5 2 ic dV

ð VCL;a

ð2:46Þ ð2:47Þ

VCL;c

The average volumetric current density iavg (A m23) is expressed as: ð ð 1 1 iavg 5 ia dV 5 2 ic dV ð2:48Þ VCL;a VCL;a VCL;c VCL;c where VCL;a and VCL;c (m3) are the volume of the CL of anode and cathode, respectively. The surface overpotential, the driving force for the transfer current density in an electrochemical reaction, is defined as [8]: η 5 ϕs 2 ϕM 2 Eeq

ð2:49Þ

where Eeq (V) is the equilibrium potential, which is zero on the cathode and is equal to the theoretical cell potential at a given temperature and pressure on the cathode side calculated by the Nernst equation. The equilibrium potential for the hydrogen-based anode is zero, and the following simplified expression could be used for the calculation of the equilibrium potential of the cathode [8
10]: Eceq 5 1:482 2 8:45 3 1024 T 1 4:31 3 1025 T lnðpH2 p0:5 O2 Þ

ð2:50Þ

2.3.1.4 Exchange current density and charge transfer coefficient Exchange current density and charge transfer coefficient are the two most important parameters in Butler
Volmer to determine the intrinsic activity of the electrochemical catalysts. Exchange current density is the current densities for the anode and cathode when both the forward and backward reactions are equal. It is analogous to the rate constant in chemical reactions and

40

Fuel Cells for Transportation

is a function of the operating temperature and partial pressure of the reactant. The exchange current density can be expressed as [8]:
i0 5 iref 0 aCL

pr pref r

γ




 E T 1 2 ref exp 2 RT T

ð2:51Þ

22 where iref 0 (A m ) is the reference exchange current density per unit catalyst surface area obtained at the reference temperature of 25 C and pressure of 1.0 atm, aCL (m21) is the specific area of the CL, pr (kPa) is the reactant partial pressure, pref r (kPa) is the reference pressure, γ (0.5 for HOR and 1.0 for ORR [11
13]) is the pressure dependency coefficient, T ref (298K) is the reference temperature, and E (kJ mol21) is the activation energy; it is found to be 72.4 kJ mol21 for oxygen reduction on cathode [14] and 16.9 kJ mol21 for hydrogen oxidation on anode [12], respectively. Exchange current density is a reflection of the activity of the surface of the electrode. Higher exchange current density means a lower energy barrier that the charge must overcome in moving from the electrolyte to the catalyst surface. In a PEMFC, the exchange current density of the anode is several orders of magnitude larger than that of the cathode. In other words, more current is generated at a fixed overpotential with a higher exchange current density, and the cathode overpotential is much larger than the anode overpotential. For this reason, the polarization curve is mainly determined by the ORR at the cathode. According to the literature [4,15,16], the reference 22 exchange current density (iref 0 ) for the anode is 1.0 A cm , whereas it is much smaller for the cathode and is a function of temperature:

ð3:50724001 T Þ iref 0 5 10

ð2:52Þ

The transfer coefficients account for the electrical effects on the change of Gibbs free energy in an electrochemical reaction since the forward and backward reactions coexist in the oxidation reaction at the anode and the reduction reaction at the cathode. Sousa et al. [17] assumed that the transfer coefficient for reduction reaction, αRd, is equal to the transfer coefficient for oxidation reaction, αOx, for both anode and cathode, leading to αRd,a 5 αOx,a 5 αa and αRd,c 5 αOx,c 5 αc. According to Sun et al. [6], αRd 1 αOx is set to unity. For ORR on cathode: αRd,c 5 αc, αOx,c 5 1 2 αc; for HOR on anode: αOx,a 5 αa, αRd,a 5 1 2 αa. It is well known that the transfer coefficient has a significant influence on the current density. However, it is difficult to predict the accurate value of the transfer coefficient for a particular system as it is a function of numerous conditions, such as temperature, pressure, catalyst structure, and reactant impurity. Parthasarathy et al. [18] found two different Tafel slopes in different ranges of cell voltages, that is, lower at higher cell voltages and higher at

Fuel cell fundamentals Chapter | 2

41

lower cell voltages. Based on their experimental measurement, the cathode transfer coefficient was regressed by Sun et al. [6] as: αc 5 0:495 1 2:3 3 1023 ðT 2 300Þ

ð2:53Þ

In comparison with the cathode transfer coefficient, the anode transfer coefficient changes slightly as the operating condition changes. Therefore the value reported by Bernardi and Verbrugge [4] of αa 5 0.5 is typically used. With the development of advanced Pt-based materials, the activity of electrochemical catalysts has been significantly improved. Thus the expressions of exchange current density and charge transfer coefficient developed in the 2000s are not suitable for state-of-the-art catalysts. However, the methodologies as described in the work of Parthasarathy et al. [18] are still valid, in which the Tafel slope was derived through the simplification of the Butler
Volmer equation. For the ORR at the cathode, the second term in the bracket of the BV equation could be omitted due to its very small value. After a rearrangement, Eq. (2.16) becomes ηc 5

Rg T Rg T lni0 2 lni αc F αc F

ð2:54Þ

Substituting Eq. (2.49) into the above equation leads to ϕs 2 ϕM 2 Eeq 5

Rg T Rg T lni0 2 lni αc F αc F

ð2:55Þ

In the electrochemical control zone at high cell voltage, the electrolyte phase potential, ϕM , is negligibly small, and the solid phase potential, ϕs , is approximately equal to the cell voltage E. Thus Eq. (2.55) could be expressed as: E 2 Eeq 5

Rg T Rg T lni0 2 lni αc F αc F

ð2:56Þ

Plotting E as a function of lni gives a straight line with a slope of Rg T=αc F and an intercept of ðRg T=αc FÞlni0 ; then, the exchange current density, i0 , and charge transfer coefficient, αc , are obtained.

2.3.1.5 Electrical and ionic conductivities The electrical conductivities of GDL and CL are different owing to the different compositions of the two domains and depend on the composition, for example, the volume fractions of the components responsible for the conductivity. Therefore appropriate corrections are essential. The most widely used approximation for effective electrical conductivities is the Bruggeman approximation [19]. However, there are many correlations beyond the Bruggeman approximation available in the open literature for effective

42

Fuel Cells for Transportation

electrical conductivities [19]. For instance, the effective electronic conductivity for the CL can be obtained using the following correlation [20]: σeff s 5 σs

2 2 2εe εe 1 2

ð2:57Þ

where σs (S m21) is the electrical conductivity of the solid phase (platinum-dispersed carbon in the CL), and εe is the volume fraction of the nonconductor materials, for example, ionomer and void space in the CL. The electronic conductivity of Vulcan XC-72 carbon black is 450 S m21 [21], and the electrical conductivity of platinum is regressed from the experimental data as [22]: σPt 5 1:7209 3 109 T ð20:9259Þ

ð2:58Þ

Proton transport between agglomerates requires sufficiently thick ionomer films surrounding agglomerates because the contact between agglomerates decreases when ionomer thickness decreases, which leads to a decrease in 21 proton transport. The effective ionic conductivity of CL, σeff M (S m ), can be obtained using the following equation [23]: " # ðεagg;M 2 1Þ eff ð2:59Þ σM 5 ð1 2 εCL Þ 1 1

3 σ M 11δM =ragg 1a0 where εagg;M is the volume fraction of ionomer in agglomerate, δM (m) is the thickness of the ionomer film, and ragg (m) is the radius of agglomerate. The values of the parameters of εagg;M ,δM , and ragg are associated with the CL structure. Note that the above expression is developed by improving the equation of Jaouen et al. [24,25]. To capture the trend of the relationship eff between δM and σeff M (δ M -0,σM -0), the dimensionless parameter a0 is added, which is given by: 
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi δM 1 3 ð1 2 εagg;M Þ 2 1 a0 5 min 0; ð2:60Þ ragg The intrinsic ionic conductivity, σM (S m21), strongly depends on temperature and water content, which can be obtained from [26]. 
 1 1 2 σM 5 exp 1268 ð0:5139λ 2 0:326Þ ð2:61Þ 303 T where λ is the membrane water content.

2.3.2

Multicomponent mass transport

2.3.2.1 Conservation of momentum The velocity and pressure of reactant gases flowing through the channel and porous media could be known by solving the conservation equation of

Fuel cell fundamentals Chapter | 2

43

momentum. The momentum flow through the volume element can be described by two mechanisms: convection and diffusion. Considering the fluid flows through all faces of the volume element, the conservation of momentum can be written as [27]: @ðρuÞ 5 2 rUðρuuÞ 2 rUp 2 rUτ 1 ρg @t

ð2:62Þ

where rp (Pa m21) is the pressure gradient, τ (kg m21 s22) is the shear stress tensor, and g (m s22) is the gravitational acceleration vector. The term on the left-hand side of Eq. (2.62) means the rate of momentum increase per unit of volume, while on the right-hand side, the first term is the rate of momentum gain by convection per unit of volume, the second term is the pressure force on element per unit of volume, the third term describes the rate of momentum gain by viscous transfer per unit of volume, and the last term is the gravitational force on the element per unit of volume. This equation, which is valid for any continuous medium, is the general form of the motion equation. If the behavior of the fluids obeys Newton’s law of viscosity, in which the shear force per unit of area is proportional to the negative of the local velocity gradient, the shear stress tensor is therefore being expressed as [27]:   2 T ð2:63Þ τ 5 μ ru 1 ðruÞ 2 ðrUuÞI 3 where I is the identity matrix. To describe the momentum balance in the porous media, the Brinkman equation [28] is normally used. This equation was developed based on Darcy’s law by Brinkman in 1949. An additional term was added to Darcy’s law accounting for the viscous transport in the momentum balance. Both the pressure and flow velocity vector were treated as independent variables in the Brinkman equation shown as follows: ρ @u μeff 52u 2 rUp 2 rUτ 1 ρg ð2:64Þ ε @t K where K (m2) and ε are the permeability and porosity of the porous media, respectively, and μeff (Pa s) is the effective viscosity of the fluid. The shear stress tensor is similar to that in Eq. (2.63), but the porosity of the porous media is included:   μ 2 ru 1 ðruÞT 2 ðrUuÞI ð2:65Þ τ5 ε 3 For a multicomponent mixture, the motion equation is very similar to the equations developed P for a single fluid. The difference is that the last term, ρg, is replaced by ρi gi , which accounts for the fact that each species may

44

Fuel Cells for Transportation

be affected by different external forces per unit of mass. Simultaneously, a unit tensor δ is introduced to the pressure term shown as follows: n X @ðρuÞ 5 2 rUðρuuÞ 2 rUð pδÞ 2 rUτ 1 ρ i gi ; @t i51 n ρ @u X μeff 52u 2 rUð pδÞ 2 rUτ 1 ρ i gi ; ε @t K i51

i 5 1; 2; 3; . . .; n ð2:66Þ i 5 1; 2; 3; . . .; n

ð2:67Þ

Note that the flow velocity in equations from Eqs. (2.62) to (2.67) is the superficial averaged velocity, which is defined as the volume rate of flow through a unit cross-sectional area including both pores and solid matrix. This definition makes the velocity field continuous across the boundaries between the porous and free-flow regions. The flow therefore can be described by the same velocity in the entire domain.

2.3.2.2 Maxwell
Stefan diffusion The multicomponent mass transport of gaseous reactant mixture within the porous electrode is described by the Maxwell
Stefan equation as " # N X g g g rp g g g T rT 1 Di ð1 2 sÞDij ðrxj 2 wj Þ ρ u Urwi 2 rU 2ρ 5 Mi Sgi ð2:68Þ p T j51 where wgi ; wgj , and Mi (kg mol21) are the mass fraction, mole fraction, and molecular weight of species I, respectively. Dij (m2 s21) and DTi (m2 s21) are the binary diffusion coefficient and the thermal diffusion coefficient, respectively, T (K) is the temperature, and Sgi (mol m23 s21) is the source terms, which account for the electrochemical reactions taking place within the porous CL shown as follows: SgH2 5

ia ic ic ; SgO2 5 2 ; Sgw 5 2F 4F 2F

ð2:69Þ

Note that ρg (kg m23) is the density of the gas mixture, which is given by the ideal gas law: ρg 5

pg M n RT

ð2:70Þ

where pg (Pa) is the pressure of the gas mixture, and Mn (kg mol21) is the mean molecular weight of the gas mixture, which can be related to the mole fraction of the component as follows: Mn 5

n X i51

xi Mi

ð2:71Þ

Fuel cell fundamentals Chapter | 2

45

where n is the number of component gas in the gas mixture, and Mi (kg mol21) is the molecular weight of the species i. According to the agglomerate assumption, reactant gases must first transport through the secondary pores to the outer boundary of the ionomer films surrounding the agglomerate, then diffuse into the primary pores inside the agglomerate, and reach the active sites for electrochemical reaction. Note that ionomer and water are also the diffusive media of gas transport, but with a few orders of magnitude lower diffusivity in comparison with the void space. Neglecting the mass transport of gas species through ionomer and water, the effective diffusion coefficient of species I through the secondary pores can be approximated using the Bruggman relationship as [15,26,29,30]: 1:5 0 Deff i;s 5 εs Di2P

ð2:72Þ

where εs is the volume fraction of secondary pores, and D0i2P (m2 s21) is the equivalent diffusion coefficient of species I in void space. When reactant gases diffuse through the ionomer film, they must transport through the primary pores inside the agglomerates to the surface of the platinum particles. The agglomerates consist of four components, namely, Pt/ C, ionomer, water, and primary pores. Assuming primary pores are the gas transport media, the effective transport coefficient of gas species i within the agglomerates is: 1:5 0 Deff i;p 5 εagg;p Di2P

ð2:73Þ

in which εagg;p is the volume fractions of primary pores within the agglomerate. By taking the Knudsen diffusion into account, the equivalent diffusion coefficient of species i in void space, D0i2P (m2 s21), is: 1 1 1 5 0 1 DKn;i D0i2P Di2g

ð2:74Þ

where D0i2g (m2 s21) is the intrinsic diffusion coefficient, and DKn;i (m2 s21) is the Knudsen diffusion coefficient of species i. 1 2 xi 0 j6¼i ðxj =Dij Þ

D0i2g 5 P

ð2:75Þ

where D0ij (m2 s21) is the binary diffusion coefficient of species i and j, which is calculated by the equation developed by Bird et al. [27]: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 1 1 1 0 27 3 T 1 ð2:76Þ Dij 5 1:8583 3 10 2 Mi Mj pσij Ωij

46

Fuel Cells for Transportation

where σij 5

1 pffiffiffiffiffiffiffi ðσi 1 σj Þ; εij 5 εi εj 2

ð2:77Þ 



Ωi and Ωij are functions of the reduced temperature Ti and Tij . κ κ   Ti 5 T; Tij 5 T εi εij

ð2:78Þ

There are empirical expressions for calculating Ωi and Ωij written as: 





Ωi 5 A=ðTi ÞB 1 C=expðTi DÞ 1 E=expðTi FÞ 





ð2:79Þ 

Ωij 5 A=ðTij ÞB 1 C=expðTij DÞ 1 E=expðTij FÞ 1 G=expðTij HÞ

ð2:80Þ

The Lennard
Jones potential parameters are listed in Table 2.1. The temperature-dependent binary diffusion coefficients for all pairs of gas species at 1.0 atm are presented in Fig. 2.5.

2.3.2.3 Knudsen diffusion Knudsen diffusion is a means of diffusion that occurs when the scale length of a system is comparable to or smaller than the mean free path of the particles involved. Knudsen’s effect on reactant gas diffusion should be considered when the Knudsen number ðKn Þ is greater than 0.1 [24,25]. The Knudsen number and Knudsen diffusion coefficient of oxygen diffusion in porous media are [10,32]: kB T Kn 5 pffiffiffi 2πpσii davg rffiffiffiffiffiffiffiffiffi davg 8RT DKn ;i 5 πMi 3

ð2:81Þ ð2:82Þ

TABLE 2.1 Lennard
Jones potential parameters [27]. Species i

Mi

σi

εi =κ

H2

2.016

2.915

38.0

O2

32.000

3.433

113.0

N2

28.013

3.667

99.8

H2O

18.000

2.641

809.1

Ωi A 5 1.16145; B 5 0.14874; C 5 0.52487; D 5 0.77320; E 5 2.16178; F 5 2.43787 Ωi;j A 5 1.06036; B 5 0.15610; C 5 0.19300; D 5 0.47635; E 5 1.03587; F 5 1.52996; G 5 1.76474; H 5 3.89411

Fuel cell fundamentals Chapter | 2

47

FIGURE 2.5 Temperature-dependent binary diffusion coefficients of gas species pairs at 1.0 atm [31].

where kB (1.38065 3 10223 J K21) is the Boltzmann constant, σii (m) is the particle diameter, and davg (m) is the average pore diameter in the CL, which can be calculated as [10,32]: davg 5

4 εCL ragg 3 1 2 εCL

ð2:83Þ

where εCL and ragg (m) are the porosity of the CL and the radius of the agglomerate, respectively. Taking the effective diffusion coefficient of oxygen through the void space of porous electrode as an example, as shown in Fig. 2.6, the Knudsen effect is significant (particularly at higher temperatures). The thermal diffusion coefficient of species i in the gas mixture, DTi (m2 21 s ), can be expressed as the quotient of the thermal conductivity and specific heat capacity as: DTi 5 Mi ki =cp;i

ð2:84Þ

The thermal diffusion of gas species is normally omitted in most research owing to the limited influence on the overall diffusivity compared to that trigged by the concentration gradient.

48

Fuel Cells for Transportation

FIGURE 2.6 Effective oxygen diffusion coefficient through void space of porous electrode [31].

2.3.3

Heat transport

2.3.3.1 Conservation of energy The equation of energy for a multicomponent mixture is given by using the methodology developed by Bird et al. [27]:
 @U 1 uUrU 5 2 rUq 2 prUu 2 τrUu 1 Qe ð2:85Þ ρ @t The term on the left-hand side of the equation above represents the rate of internal energy gained per unit of volume, the first term on the right hand is the rate of internal energy input per unit of volume by conduction, the second term on the right hand is the reversible rate of internal energy increase per unit of volume by compression, the third term on the right hand is the irreversible rate of internal energy increase per unit of volume by viscous dissipation, and Qe is the heat source. Instead of internal energy, the above expression can be written in terms of the temperature and heat capacity of the fluid, which becomes more convenient for calculating the temperature profiles. At constant pressure, the internal energy can be related to the temperature and heat capacity via the following equation: @U @V @T 52p 1 cp @t @t @t

ð2:86Þ

Fuel cell fundamentals Chapter | 2

49

where V (m3 kg21) is the volume per unit of mass, and cp (J mol21 K21) is the averaged heat capacity of the multicomponent mixture at constant pressure, which can be calculated by the heat capacity and mole fraction of the species of the multicomponent mixture: cp 5

n X

cp;i xi ;

i 5 1; 2; 3; . . .; n

ð2:87Þ

i51

With the aid of the equation of continuity ρðdV=dtÞ 5 rUv, then multiplying both sides of Eq. (2.86) by the averaged density of the multicomponent mixture and combining Eq. (2.85) gives: ρcp

@T 1 ρcp uUrT 5 2 rUq 2 τrUu 1 Qe @t

ð2:88Þ

The term τrUu is only significant in a high-speed flow system in which the velocity gradient is large and negligible for PEFMCs. By substituting q 5 2 krT into the above equation, a second-order partial differential equation for temperature profiles can be written as: ρcp

@T 1 ρcp uUrT 2 rUðkrTÞ 5 Qe @t

ð2:89Þ

For a multiphase heat transfer process occurred in the porous media, Eq. (2.89) becomes: " # " # ! X X @ X ðερ cp;i ÞT 1 rU ðερi cp;i ui ÞT 2 rU ki rT 5 Qe @t i 5 g;l;s i i 5 g;l i 5 g;l;s ð2:90Þ where ε is the porosity, and the subscripts g, l, and s refer to phases of gas, liquid, and solid, respectively.

2.3.3.2 Specific heat capacity and thermal conductivity The specific heat capacity and the thermal conductivity of the gas mixture are obtained by using an empirical equation developed by Wike [33], X g X xi ki P cgp 5 xi cp;i ; kg 5 ð2:91Þ j xj Φij i i in which
 "
0:5
0:25 #2 1 Mi 20:5 ki Mj Φij 5 pffiffiffi 11 11 ; Mj kj Mi 8

Φji 5

kj M i Φij ki M j

ð2:92Þ

Where cgp (J mol21 K21) is the specific heat capacity of the gas mixture, xi is the mole fraction of species i in gas mixture, cgp;i (J mol21 K21) is the specific heat capacity of species i in gas mixture, kg (W m21 K21) is the

50

Fuel Cells for Transportation

thermal conductivity of the gas mixture, ki (W m21 K21) is the thermal conductivity of species i in gas mixture, and Mi (kg mol21) is the molecular weight of species i. The expressions for the specific heat capacities of each gas component including hydrogen, oxygen, nitrogen, and water vapor are: cgp;H2 5 1:914 3 1026 T 2 2 8:314 3 1024 T 1 28:890

ð2:93Þ

cgp;O2 5 2 4:281 3 1026 T 2 1 1:371 3 1022 T 1 25:431

ð2:94Þ

cgp;N2 5 1:788 3 1025 T 2 1 2:924 3 1023 T 1 27:848

ð2:95Þ

cgp;H2 O 5 1:180 3 1026 T 2 1 9:621 3 1023 T 1 30:326

ð2:96Þ

The specific heat capacities of solid and liquid phases change slightly with temperature [34]. Consequently, the specific heat capacities of platinum, carbon black, liquid water, and membrane/ionomer are assumed as temperature-independent constants of 1.3 3 102, 894.4, 1090.0, and 4187.0 (J kg21 K21), respectively. The expressions for the thermal conductivities of hydrogen, oxygen, nitrogen, and water vapor are: kH2 5 3:777 3 1024 T 1 7:444 3 1022

ð2:97Þ

kO2 5 6:204 3 1025 T 1 8:83 3 1023

ð2:98Þ

kN2 5 5:453 3 1025 T 1 1:088 3 1022

ð2:99Þ

kHg 2 O 5 1:188 3 1024 T 2 2:404 3 1022

ð2:100Þ

For the solid components, for example, platinum, carbon black, and liquid water, the expressions for the thermal conductivities are: kPt 5 2 5:037 3 1029 T 3 1 2:483 3 1025 T 2 2 2:282 3 1022 T 1 77:80 ð2:101Þ kC 5 1:048 3 1026 T 2 2 2:869 3 1023 T 1 2:979

ð2:102Þ

kHl 2 O 5 2 1:118 3 1025 T 2 1 8:388 3 1023 T 2 0:9004

ð2:103Þ

It is worth noting that Eqs. (2.93)
(2.103) are obtained through the best fit of experimental data [27]. The effective thermal conductivity and specific heat capacity are dependent on the volume fractions of the species within a chosen domain. Without a doubt, the cathode CL is the most complicated domain in which gas mixture, liquid water, ionomer, Pt/C, etc., are all involved. The detailed expressions for the effective thermal conductivity and specific heat capacity of GDL, CL, and membrane/ionomer are listed in Table 2.2.

TABLE 2.2 Effective specific heat capacity and thermal conductivity of GDL, CL, and membrane. GDL

CL g 1 ð1 2 sÞεGDL cp

cpeff

l ð1 2 εGDL Þcp;C 1 sεGDL cp;w

k eff

ð1 2 εGDL ÞkC 1 sεGDL kwl 1 ð1 2 sÞεGDL kp

g

l LPt cp;Pt 1 ðLC 1 LS Þcp;C 1 LM cp;M 1 sεCL cp;w

Membrane g 1 ð1 2 sÞεCL cp g

LPt kPt 1 ðLC 1 LS ÞkC 1 LM kM 1 sεCL kwl 1 ð1 2 sÞεCL kp

cp;M kM

52

Fuel Cells for Transportation

2.4 2.4.1

Electrode properties Porosity of the catalyst layer

The total volume of CL consists of five components, including ionomer (M), platinum (Pt), carbon black (C), void space (P), and solid intrusion (S), for example, the penetration of GDL into CL. Vtot 5 VM 1 VPt 1 VC 1 VP 1 VS

ð2:104Þ

where the volume of the void space consists of two parts: primary pores (VP;p ) and secondary pores (VP;s ). The primary pores are the void space within the agglomerates, and the secondary pores are the void space between the agglomerates; thus we have: VP 5 VP;p 1 VP;s

ð2:105Þ

The Pt/C is constructed by platinum and carbon, thus VPt=C 5 VPt 1 VC

ð2:106Þ

Dividing both sides of Eq. (2.104) by the total volume of the CL (Vtot ), the sum of volume fractions of all components is equal to unity, leading to 1 5 LM 1 LPt 1 LC 1 LS 1 εCL

ð2:107Þ

εCL 5 εp 1 εs

ð2:108Þ

where the terms on the right side of Eq. (2.107) represent the volume fractions of ionomer (LM), platinum (LPt), carbon (LC), solid portion (LS), primary pores (εp), and secondary pores (εs), respectively, which are written as: LM 5

VM VPt VC VS VP;p VP;s ; LPt 5 ; LC 5 ; LS 5 ; εp 5 ; εs 5 Vtot Vtot Vtot Vtot Vtot Vtot

ð2:109Þ

The volume fractions of platinum, carbon, and ionomer are related to their mass loading and densities and CL thickness (δCL ) as LPt 5

mPt mC mM ; LC 5 ; LM 5 ρPt δCL ρC δCL ρM δCL

ð2:110Þ

Normally, platinum is dispersed in carbon black to construct the catalyst particles. Therefore, the volume fraction of Pt/C is the sum of the volume fractions of platinum and carbon, which can be written as [2]:
 mPt 1 12f 1 1 LPt=C 5 ð2:111Þ f ρC δCL ρPt

Fuel cell fundamentals Chapter | 2

53

where the platinum mass ratio to that of carbon (abbreviated as platinum mass ratio) is defined as: mPt f5 ð2:112Þ mPt 1 mC Due to the clamping force of the MEA, the intrusion of GDL into CL occurs. The volume fraction of the solid portion of the CL is defined as: LS 5 LGDL ð1 2 εGDL Þ

ð2:113Þ

where LGDL is the percentage of GDL penetrating the CL, and εGDL is GDL porosity. The CL porosity is therefore being written as:
 mPt 1 12f 1 1 εCL 5 1 2 LM 2 LGDL ð1 2 εGDL Þ 2 ð2:114Þ f ρC δCL ρPt Note that two unknown variables are in the above expression, εCL and δCL . There are two approaches to using the above equation. One is to fix the CL thickness and calculate the CL porosity, and the other is to fix CL porosity and calculate CL thickness. According to the CL preparation process, when the CL is prepared on the GDL or membrane, a hot press is required to reduce the contact resistance between different layers. Therefore most of the models fixed CL thickness as a constant and study the variation of CL porosity as a function of other parameters, for example, platinum loading and platinum mass ratio [35,36]. Other studies showed that the CL thickness increases when the platinum loading increases, and the CL porosity is approximately maintained as a constant. Therefore, fixing the CL porosity as a constant is a better approach. Nevertheless, a good agreement can be achieved for the CL thickness as a function of the Pt loading when modeling results are compared with the experimental data [37], as shown in Fig. 2.7.

2.4.2

Agglomerate density

Ionomer first partially fills up the primary pores within the agglomerate and then covers the agglomerate to form a thin film. The total volume of ionomer is: VM 5 VM;agg 1 VM;δ

ð2:115Þ

3

where VM;agg (m ) is the volume of ionomer within the agglomerate, and VM;δ (m3) is the volume of ionomer that existed as the thin film surrounding the agglomerate. Since the agglomerate particles consist of Pt/C, ionomer, and primary pores, the total volume of the agglomerate is: Vagg;tot 5 VPt=C 1 VM;agg 1 VP;p

ð2:116Þ

54

Fuel Cells for Transportation

FIGURE 2.7 Catalyst layer thickness with different Pt loadings in base-case condition. Experimental data from S.-Y. Lee et al., Gradient catalyst coating for a proton exchange membrane fuel cell operation under nonhumidified conditions. Electrochem. Solid. State Lett. 10 (2007) B166-B169.

The volume fraction of ionomer within the agglomerate (εagg;M ) and the volume fraction of primary pore space within the agglomerate (εagg;p ) are introduced as: εagg;M 5

VM;agg VP;p ; εagg;p 5 Vagg;tot Vagg;tot

ð2:117Þ

The total volume of the agglomerate could be calculated through the combination of the above two expressions. Vagg;tot 5

VPt=C 1 2 εagg;M 2 εagg;p

ð2:118Þ

The volume of the individual agglomerate particle (without the ionomer film) is Vagg;i 5

4 3 πr 3 agg

ð2:119Þ

Dividing Eq. (2.118) by Eq. (2.119), the total number of the agglomerates is found as: nagg 5

3VPt=C 4ð1 2 εagg;M 2 εagg;p Þπragg 3

ð2:120Þ

Fuel cell fundamentals Chapter | 2

55

Defining the number of agglomerate particles per volume of the CL, the agglomerate density is introduced as Nagg 5

2.4.3

3LPt=C 4ð1 2 εCL Þπragg 3

ð2:121Þ

Thicknesses of the ionomer and liquid water films

The volume of agglomerate with the ionomer film is equal to the volume of Pt/C, ionomer, and primary pores, leading to: 4 nagg πðragg 1δM Þ3 5 VM 1 VPt=C 1 VP;p 3 Substituting Eq. (2.122) into Eq. (2.104) leads to 4 Vtot 5 nagg πðragg 1δM Þ3 1 VP;s 1 VS 3

ð2:122Þ

ð2:123Þ

Substituting the total number of agglomerates Eq. (2.120) into the above equation and then dividing both sides by Vtot give 15

LPt=C ðragg 1δM Þ3 1 εs 1 LS 3 ð1 2 εCL Þragg

ð2:124Þ

The volume fractions of the primary and secondary pores are: εp 5

LPt=C ðεCL 2 εagg;M Þ 1 2 εCL

εs 5 εCL 2

LPt=C ðεCL 2 εagg;M Þ 1 2 εCL

ð2:125Þ ð2:126Þ

By substituting Eq. (2.126) into Eq. (2.124), the thickness of the ionomer thin film can be calculated by the following equation: "sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi # 3 ð1 2 εCL Þð1 2 εCL 2 LS Þ 1 LPt=C ðεCL 2 εagg;M Þ δM 5 ragg 21 ð2:127Þ LPt=C Defining the volume fraction of the primary pores occupied by the ionomer as: %M 5

εagg;M εagg;M 1 εagg;p

ð2:128Þ

Then Eq. (2.128) becomes to: "sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi # 3 ð1 2 εCL Þð1 2 εCL 2 LS Þ 1 LPt=C εCL ð1 2 %MÞ δM 5 ragg 21 LPt=C

ð2:129Þ

56

Fuel Cells for Transportation

If the primary pores are completely occupied by the ionomer (%M 5 1), the above expression changes to: "sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi # 3 ð1 2 εCL Þð1 2 εCL 2 LS Þ 21 ð2:130Þ δM 5 ragg LPt=C It is important to note that the volume fraction of the ionomer with the agglomerate (εagg;M ) is impossible to be larger than the porosity of the CL (εCL ). Assuming that the ionomer is hydrophilic, any liquid water is assumed to coat the entire surface of the individual agglomerate to generate a liquid water film adjacent to the outer boundary of the ionomer film. The total volume of the liquid water generated can be obtained as: Vw 5 sεCL Vtot

ð2:131Þ

where s is the liquid water saturation, which is defined as the volume fraction of the void space occupied by liquid water. Averaging the total volume of the liquid water to each agglomerate, the volume of the liquid water surrounding each agglomerate is: sεCL Vw;i 5 ð2:132Þ Nagg Then, the liquid water film thickness is given as: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3sεCL 3 δw 5 ðragg 1δM Þ3 1 2 ðragg 1 δM Þ 4πNagg

2.4.4

ð2:133Þ

Specific area

For CLs prepared through spraying catalyst ink onto GDLs that consists of Pt catalyst dispersed in Vulcan XC72 carbon black, Nafion solution (5%), and organic solvent, for example, glycerol and isopropanol, the reaction surface area per unit platinum mass (m2 kg21) can be roughly calculated by the following empirical expression [32,38]: As 5 ð227:79f 3 2 158:57f 2 2 201:53f 1 159:5Þ 3 103

ð2:134Þ

where f is the platinum mass fraction. The specific area of the CL (m21), defined as the total active area per volume of the CL, can be written as [2,15]: mPt aCL 5 As ð2:135Þ δCL

Fuel cell fundamentals Chapter | 2

57

The specific area of the agglomerate, defined as the total active area per volume of agglomerate, could be expressed as [35,36]: aagg 5

2.4.5

mPt As ð1 2 εCL Þ LPt=C δCL

ð2:136Þ

Deformation of porous electrode

To avoid gas leakage and minimize contact resistance, a certain clamping force is applied to the bipolar plates. However, the deformation caused by clamping force may damage the microstructure of GDL. As the most compressible component in MEA, the deformation of GDL is much more significant than that of the CL and PEM [39]. In addition, the deformation of GDL under the rib is more obvious than in the area under the channel owing to the special channel-rib pattern of the flow field [40]. Consequently, nonuniform deformation of GDL results in inhomogeneous variations of the physical properties of GDL under the rib and channel, such as porosity, thickness, permeability, effective diffusivity, and effective electrical conductivity, which in turn affect the cell performance and durability. These physical properties are competitive and coupled in the transport process in GDL. Over the past decades, numerous experimental and numerical studies have been reported on GDL deformation and its associated effect on cell performance [41
43]. Imaging techniques such as synchrotron X-ray imaging [44] and X-ray computed tomography (CT) [45] have been widely used to investigate the microstructural changes of GDL under clamping force. For example, Zenyuk et al. [46] and T¨otzke et al. [47] utilized X-ray CT to explore the microstructural change, such as porosity, tortuosity, and pore size distribution (PSD) of different GDL materials under various levels of compression loads, concluding that the porosity decreased with compression and the PSD shifted from bimodal to unimodal while the larger pores were shifted to that of the smaller radii. Atkinson et al. [48] investigated the variation of the porosity and ohmic resistance of different GDL types under varying degrees of compression on the influence of the mass transport resistance via X-ray CT and found that the concentration losses increased with compression owing to the decreased porosity. The morphological properties of the entire MEA with two distinct flow field arrangements under different compression ratios (CR) were investigated by Kulkarni et al. [49] and found that the symmetrical flow field arrangement had a smaller deformation compared to the asymmetrical flow field arrangement when the CR was lower than 40%. However, a higher compression would lead to the delamination of MEA in the symmetrical flow field arrangement. The polarization curves of a fuel cell under varying clamping pressure were tested by Chang et al. [50] using a homemade test fixture with three different types of GDLs. Their

58

Fuel Cells for Transportation

results indicated the existence of optimal clamping pressure to maximize the peak power density, and the optimal value was approximately 3.0 MPa in their experiments. These experimental studies revealed the variation of the morphological properties of the MEA components under realistic compression conditions and qualitatively measured the effect of the clamping pressure on the physical properties of GDLs, such as porosity, permeability, gas diffusivity, and electrical/thermal conductivity. All these experimental studies provided a comprehensive dataset for computational modeling. With respect to the numerical studies of the influence of the GDL deformation on the transport properties and cell performance, not only the variation of physical properties of GDL but also the effect of the compression on the current density distribution and liquid water distribution in an operating fuel cell are infeasible or costly to be measured by experimental approaches. For example, Li et al. [42] developed a three-dimensional, two-phase PEMFC model to investigate the local current density distribution at various assembly pressures. They found that the local current density under the channel is higher than that under the rib, and the nonuniform was more serious with increasing compression. Jiao et al. [51] investigated the effect of compression on the water transport in GDLs using OpenFOAM, concluding that the compression increased the flow resistance for the liquid water and the vapor, which led to an uneven distribution of fluids. Mehrtash et al. [52] developed a two-dimensional, half-cell, non-isothermal, multiphase model to study the variation of the water concentration and current density profiles, as well as characterization curves, with respect to different levels of compression. Zhou et al. [53] used the finite element method to analyze the relationship between GDL deformation and cell performance. The numerical results showed that the increase in compressive deformation deteriorated the cell performance when the contact resistance was not considered, and there existed an optimal clamping pressure when the contact resistance was considered. The clamping force required to achieve different levels of electrode deformation is calculated as follows: @σij 1 bi 5 0 @xj

ð2:137Þ

@2 εij @2 εkl @2 εik @2 εjl 1 2 2 50 @xk @xl @xi @xj @xj @xl @xi @xk

ð2:138Þ

where σij (Pa) is the stress tensor, bi is the body force vector, and εij is the strain tensor. Eq. (2.137) indicates that the displacement within the electrodes is single-valued and continuous. Typically, only the deformation of the GDLs is considered because the deformation of the other components, including the membrane and CLs, is very limited owing to either the higher mechanical strength or the smaller dimensions.

Fuel cell fundamentals Chapter | 2

59

Based on Hooke’s law for linear elastic deformation, the stress and strain are related through the following constitutive relation: σij 5 λδij εkk 1 2μεij λ5

Eν E ;μ5 ð1 1 νÞð1 2 2νÞ 2ð1 1 νÞ

ð2:139Þ ð2:140Þ

where E (Pa) is Young’s modulus, ν is the Poisson ratio, δij is the Kronecker delta, and the subscripts i, j, and k are the coordinates of the computational domain. The required stress to achieve a desired CR of the GDLs could be calculated. The GDL porosity after compression and the CR can be calculated as [54] εGDL 5

ε0GDL 2 CR 1 2 CR 0

CR 5 1 2 δ =δ0

ð2:141Þ ð2:142Þ

where CR is defined as the ratio of the GDL thickness before and after compression, δ0 and δ’ are the thicknesses of the initial and compressed GDL, respectively, ε0GDL represents the initial GDL porosity, and εGDL is the GDL porosity after compression.

2.5

Water management

Water exists in three phases: vapor, liquid water, and the dissolved phase in membrane and ionomer. Due to the hydrophilic property of the Nafion ionomer, water uptake occurs when the membrane and ionomer are exposed to a humidified environment.

2.5.1

Water phase-transfer and water transport through membrane

Water exists in three different phases in different solvents [55
57], including the dissolved water in the membrane and ionomer, water vapor, and liquid water in the porous media and flow channels. The dissolved water is the membrane and ionomer absorbed water, which enters the membrane and ionomer from the water vapor during water uptake and leaves the membrane and ionomer in the liquid phase when the water content of the membrane and ionomer exceeds complete saturation. The main phase-transfer mechanisms include the following processes: the phase transfer between liquid water and water vapor via condensation and evaporation, the phase transfer between dissolved water and water vapor through membrane and ionomer absorption or water uptake, and the phase transfer between liquid water and dissolved water during membrane and ionomer desorption.

60

Fuel Cells for Transportation

Water transport in the membrane plays an important role in determining the water content in the membrane/ionomer [26,55
58]. During fuel cell operation, water transport through the membrane occurs via three mechanisms: electroosmotic drag (EOD) of water molecules carried by protons migrating from anode to cathode, back-diffusion driven by the concentration gradient of water, and convection generated by the pressure gradient. We can mathematically describe the dissolved water transport through the membrane as:
 @ iM kp;M cdw ðLM cdw Þ 1 rUðnd Þ 2 rUðDw2M rcdw Þ 2 rU rp 5 Sdw ð2:143Þ @t F μw The term on the left-hand side refers to the water accumulation, water migration by EOD, back-diffusion, and hydraulic permeation, respectively. LM is the volume fraction of ionomer (LM 5 1 in the membrane, 0 , LM , 1 in the CL, LM 5 0 in GDL and channel), cdw (mol m23) is the concentration of the dissolved water, nd is the EOD coefficient, which is expressed as 2:5λ=22 [55
59], Dw2M (m2 s) is the diffusion coefficient of water through the membrane, kp;M (m2) is the hydraulic permeability of water in the membrane, μw (Pa s) is the water viscosity, p (Pa) is the pressure, and Sdw (mol m23 s21) is the source term of dissolved water. The concentration of the dissolved water depends on the water content of the membrane and ionomer according to the following expression [58]: cdw 5

ρM λ EW 1 1 ks λ

ð2:144Þ

Where ρM (kg m23) is the density of dry membrane, EW (g mol21) is the equivalent weight of membrane, and ks is the swelling coefficient, representing the volume increase of the membrane and ionomer. The polymeric matrix of the membrane and ionomer expands leading to an increase in their volume when membrane/ionomer water absorption (water uptake) occurs. Normally, a dry Nafion membrane/ionomer swells approximately 20% when fully hydrated by water vapor [58,60]. Membrane and ionomer swelling has two effects on fuel cell performance. For the membrane, higher swelling increases the ionic conductivity and water diffusion coefficient, while for the ionomer, higher swelling increases the thickness of the ionomer film surrounding the agglomerate and decreases the void space within the CL, leading to an increase in transport resistance of reactant gases, especially the oxygen [58]. Depending on the directions of membrane swelling, there are two types of membrane deformation: the through-plane membrane thickness increasing and the in-plane membrane buckling. The through-plane membrane thickness increase is caused by the zero or low fastening force from gas flow fields to the MEA [61]. Since the membrane is fixed between the bipolar plates under

Fuel cell fundamentals Chapter | 2

61

a relatively high clamping force, the thickness of the portion of the membrane under the current collector ribs is impossible to change during membrane water absorption (water uptake). However, the in-plane buckling occurs under the channels as shown in Fig. 2.8. Compared to the throughplane thickness increasing, the in-plane buckling is more important as the inplane stress is the major stress component in the membrane [62]. The inplane buckling could have a significant impact on the channel flow as the MEA bulges into the channel. The bulged GDL into the flow channels increases the mass transport resistance and can lead to pinhole formation of the membrane under the channel [63]. The volume fraction of the ionomer within CL after swelling is calculated by the following equation: 5 LM ð1 1 ks λÞ Lswell M

ð2:145Þ

FIGURE 2.8 Sketch of the membrane and ionomer swelling and the bulged MEA into the flow channel. Credit: Reprinted from Xing et al. Appl. Energy 138 (2015) 242
257, with permission from Elsevier.

62

Fuel Cells for Transportation

Dissolved water is absorbed by the membrane and ionomer when the concentration of the dissolved water is smaller than the equilibrium concentration, which is the maximum dissolved water being carried by the membrane and ionomer. After reaching the equilibrium concentration, the dissolved water moves away from the membrane and ionomer in the liquid water phase, during the process of membrane and ionomer desorption. The source terms regarding the process above are: eq

d Svd w 5 kads cw 2 cw ;

when cdw , ceq w

ð2:146Þ

d

eq Sdl w 5 kdes cw 2 cw ;

when cdw $ ceq w

ð2:147Þ

The subscripts vd and dl in the source term represent the water phase change from vapor to dissolved water and from dissolved water to liquid water, respectively. kads and kdes (s21) are the adsorption and desorption rate coefficients, given by [59,64]: 
 1:14 3 1025 fw 1 1 2 exp 2416 kads 5 ð2:148Þ 303 T δCL 
 4:59 3 1025 fw 1 1 2 exp 2416 ð2:149Þ kdes 5 303 T δCL fw 5

λVw VM 1 λVw

ð2:150Þ

where fw is the water volume fraction of the membrane and Vw and VM (m3 mol21) are the partial molar volume of water and the dry membrane, respec23 tively. ceq w (mol m ) is the equilibrium dissolved water concentration, which is determined by the equilibrium water content according to Eq. (2.144). The equilibrium water content is given by empirical correlations based on water uptake measurements [57], given as: λeq 5 16:8s 1 14:0 ð1 2 sÞ;

when s . 0

ð2:151Þ

where s is the liquid water saturation. The actual water content as a function of water activity is λ 5 0:043 1 17:81αH2 O 2 39:85α2H2 O 1 36:0α3H2 O ;

when αH2 O # 1 ð2:152Þ

In the above equation, αH2 O can be a function of both water vapor partial pressure and liquid water saturation [56], as given below: p αH2 O 5 xw 1 2s ð2:153Þ psat

Fuel cell fundamentals Chapter | 2

63

2.5.2 Diffusion of species in Nafion ionomer with different membrane water content According to the agglomerate assumption, reactant gases must diffuse through the ionomer film surrounding the agglomerate before reaching the surface of platinum catalyst particles inside the agglomerate. Owing to the much higher solubility and diffusivity of hydrogen through Nafion than that of oxygen, the diffusion of oxygen through the ionomer film is more critical [4,15]. For oxygen diffusivity through the Nafion membrane and ionomer, Marr and Li [15] provided a temperature-dependent equation by fitting the experimental data published by Parthasarathy et al. [18]. Suzuki et al. [65] reported that the oxygen diffusion coefficient is proportional to the power of the membrane water content. Using the combination of the two expressions and the influence of temperature and membrane water content, the following expression can be obtained. 210 0:708

DO2 2M 5 1:3926 3 10

λ


 T 2 273 2 1:6461 3 10210 λ0:708 1 5:2 3 10210 exp 106:65

ð2:154Þ where λ and T (K) are membrane/ionomer water content and temperature, respectively. Conversely, Henry’s constant for oxygen solubility in the Nafion ionomer depends on the relative humidity [66]. To investigate the effect of water content on Henry’s constant, one can calculate Henry’s constant from the expression given below, which is obtained by modifying the equation of Marr and Li [15] and fitting the results of Suzuki et al. [65]:
 666 ð2:155Þ HO2 5 0:11552exp 14:1 1 0:0302λ 2 T A comparison of Eq. (2.154) with the experimental data of Takamura et al. [67] and Eq. (2.155) with the simulation results of Suzuki et al. [65] is shown in Fig. 2.9. Nafion is an ideal media for dissolved water transport, and the diffusion coefficient of water through the membrane is a piecewise function that can be determined by both temperature and membrane water content [4,10,26,32,38,55
58,60,68]: 8 < D0 ð2:563 2 0:33λ 1 0:0264λ2 2 0:000671λ3 Þ; λ.4 DH2 O2M 5 D0 ð2 1:25λ 1 6:65Þ; 3,λ#4 : D0 ð2:05λ 2 3:25Þ; 2#λ#3 ð2:156Þ

64

Fuel Cells for Transportation

FIGURE 2.9 Oxygen diffusivity through Nafion membrane/ionomer as a function of water activity (left) and Henry’s constant for oxygen solubility in Nafion membrane/ionomer as a function of water content (right).

210

D0 5 1:0 3 10


 1 1 2 exp 2416 303 T

ð2:157Þ

For the above expressions, the membrane water content (λ) can be calculated from the water activity (αw ) using the following expression: 8 αw , 1 < 0:043 1 17:81αw 2 39:85αw 2 1 36:0αw 3 ; λ 5 14:0 1 1:4ðαw 2 1Þ; ð2:158Þ 1 # αw # 3 : 16:8; αw . 3 Fig. 2.10 shows the water diffusion coefficient through Nafion membrane as a function of membrane water content and the relationship between water activity with membrane water content. As shown in Fig. 2.10, when the membrane water content increases, the diffusion coefficient initially increases and then decreases to a constant, while the membrane water content increases up to 16.8 as water activity increases. Water activity can be associated with the partial pressure of water vapor as [69]: pw αw 5 5 RH ð2:159Þ psat where pw (Pa) is the partial pressure of water vapor, psat (Pa) is the saturated water vapor pressure, which is the water vapor pressure at saturation temperature, and RH is the relative humidity. The saturated water pressure is obtained by fitting the experimental data. A polynomial equation is given as psat 5 9:531 3 1024 ðT 2237Þ4 2 3:123 3 1022 ðT 2237Þ3 1 3:451ðT 2237Þ2 1 20:96ðT 2 237Þ 1 611:0

ð2:160Þ

Water could also diffuse through the Nafion membrane under a pressure force, and the hydraulic permeability of water through Nafion is associated

Fuel cell fundamentals Chapter | 2

65

FIGURE 2.10 Water diffusion coefficient through the membrane (left) and the relationship between water activity and membrane water content (right) of Nafion 117 at 30 C.

with the membrane water content as [4]: kp;M 5 2:86 3 10220 λ

2.5.3

ð2:161Þ

Two-phase flow of gas
water mixture

The generation of water is unavoidable during the operation of PEMFCs owing to the ORR and water migration under EOD. Water is typically formed at the cathode side. However, the consumption of hydrogen at the anode side could also trigger the formation of liquid water at high relative humidity because the vapor is oversaturated. At a steady-state condition, applying the continuity equation on liquid water and gas phase, respectively, the following equations are obtained: rUðρlw ulw Þ 5 Mw Sl

ð2:162Þ

rUðρgw ugw Þ 5 Mw Sg

ð2:163Þ

where Mw (kg mol21) is the molecular weight of water, and ρ (kg m23),u (m s21), and S (mol m23 s21) are the density, velocity, and source term of liquid water and water vapor, respectively. The subscript w represents water, and the superscripts l and g represent the liquid water and gas phase, respectively. According to Darcy’s law, the velocity of the liquid water and gas phase can be related to their partial pressure as: ulw 5 2

Kl rpl μlw

ð2:164Þ

ugw 5 2

Kg g rp μgw

ð2:165Þ

where K (m2), μ (Pa s), and p (Pa) are the permeability, viscosity, and partial pressure of the liquid water and gas phase, respectively. The pressure

66

Fuel Cells for Transportation

difference between the wetting and non-wetting phase within the porous media is the capillary pressure (pc ), which is expressed as: pc 5 pg 2 pl

ð2:166Þ

Substituting Eq. (2.166) into Eq. (2.164), the liquid water velocity becomes: ulw 5

Kl Kl c rp 2 rpg μlw μlw

ð2:167Þ

Introducing the liquid water saturation (s), defined as the volume fraction of the liquid water in the porous media, into Eq. (2.167) gives ulw 5

K l dpc Kl rs 2 l rpg l μw ds μw

ð2:168Þ

Capillary pressure can be linked to liquid water saturation as [70
73]:  ε 1=2 pc 5 σcosðθc Þ JðsÞ ð2:169Þ K where JðsÞ is the Leverett function and is given by [70
73]:  1:417ð1 2 sÞ 2 2:120ð12sÞ2 1 1:263ð12sÞ3 ; when θc , 90 JðsÞ 5 1:417s 2 2:120s2 1 1:263s3 ; when θc . 90 ð2:170Þ Normally, the porous electrode is hydrophobic, therefore, the Leverett function is expressed in terms of the liquid water saturation, and the contact angle, θc , is between 90 degrees and 180 degrees. The surface tension (σ) for the liquid water
air system is 0.0625 N/m. It is also worth noting that the Leverett J-function was originally obtained from the empirical measurement of water transport in unconsolidated sand [70], which may not be appropriate for some diffusion media, for example, SGL 24 series carbon paper tailored with PTFE content varying from 5 to 20 wt.%, due to the inaccurate prediction of capillary pressure at high nonwetting phase saturation [74,75] and GDLs modified with hydrophobic fluorinated ethylene propylene or polydimethylsiloxane [76]. The significant hysteresis of capillary pressure of fresh and aged diffusion media was also observed during experiments as reported in [75]. Combining Eqs. (2.168) and (2.169), the liquid water velocity is expressed in terms of liquid water saturation as: ulw 5 σcosðθc Þ

K l ε 1=2 dJðsÞ Kl Þ rs 2 ð rpg ds μlw K μlw

ð2:171Þ

Fuel cell fundamentals Chapter | 2

67

The permeability of liquid water and gas phase can be associated with the permeability of the porous media via: K l 5 krl K;

K g 5 krg K

ð2:172Þ

where krl and krg are the relative permeability of liquid water and gas phase, which are proportional to the cube of liquid water saturation as: krl 5 s3 ; krg 5 ð12sÞ3

ð2:173Þ

Substituting Eqs. (2.165) and (2.172) into Eq. (2.171) leads to: ulw 5

σcosðθc Þkrl dJðsÞ kl μg rs 2 gr wl ugw ðεKÞ1=2 l ds μw kr μw

ð2:174Þ

Substituting Eq. (2.174) into Eq. (2.162) and taking the liquid water accumulation into account, the liquid water saturation can be calculated as:
 @ ρlw krl μgw g l l ðερw sÞ 1 rU ρw Dc rs 2 g l u 5 Mw Slw ð2:175Þ @t kr μw where Dc (m2 s21) is the capillary diffusion coefficient, which is represented by the following expression: Dc 5

σcosðθc Þkrl dJðsÞ ðεKÞ1=2 ds μlw

ð2:176Þ

At a steady state, if the velocity of the gas phase can be neglected, Eq. (2.175) becomes a second-order partial differential equation for the liquid water saturation: r2 s 5

Mw l S ρlw Dc w

ð2:177Þ

The right-hand term of the above equation is the source term, which includes the water production from reaction and water phase change. The conservation of water in different domains of PEMFCs is given in Table 2.3, in which the superscripts v, l, and d represent water in vapor, liquid, and

TABLE 2.3 Conservation of water in different domains of PEMFCs. Channel

GDL

CL

2Swvl

2Swvl

2Swvd 2 Swvl

Liquid water, Swl

Swvl

Swvl

Swdl 1 Swvl

Dissolved water, Swd

0

0

Swr 1 Swvd 2 Swdl

Water vapor,

Swv

68

Fuel Cells for Transportation

dissolved phases, and superscript r represents water generation through reaction. The expressions of water phase change due to membrane and ionomer dl water uptake and desorption, Svd w and Sw , could be found from Eqs. (2.146) and (2.147). Since the generated water in CL is assumed in the dissolved phase, the source term of water generation through reaction,Srw (mol m23 23 21 s21), is expressed as 2 ic =2F. The source term, Svl s ), is introw (mol m duced for the interfacial mass transfer rate of water by condensation and evaporation, which is defined as: 8

εð1 2 sÞxgw g g > > xw p 2 psat ; when xgw pg $ psat kcon > > < RT ð2:178Þ Svl w5

εsρlw > > > keva psat 2 xgw pg ; when xgw pg , psat > : Mw where kcon (s21) and keva (atm21 s21) are the condensation and evaporation rate coefficients, respectively, and xgw is the mole fraction of water vapor.

2.6

Summary

In this chapter, the fundamentals and principles of a typical PEMFC operated with hydrogen and air are represented, and the governing equations are derived. All the important chemical and physical processes involved in the operation of PEMFCs are described by proper differential and algebraic equations. This chapter starts with a description of electrochemical kinetics through both the Butler
Volmer equation and the agglomerate model followed by the momentum transport and multicomponent mass through the channel and porous electrodes, and heat transport within entire the fuel cell unit. A series of equations are then developed to represent the properties of the CL, in which the porosity, agglomerate density, thicknesses of ionomer film and liquid water film, and the specific area can be quantitatively obtained. The deformation of the porous electrode is also briefly discussed. As a very important process during PEMFC operation, the water phase transfer and gas
liquid two-phase flow are described by partial differential equations, and the source terms responsible for the phase change are given in detail. In addition, the swelling of membrane and ionomer, transport of proton through the membrane, and diffusion of species through the ionomer film are associated with their water content, indicating the importance of membrane hydration during PEMFC operation. The governing equations build the framework of multi-physics, nonisothermal, and two-phase flow PEMFC models, although the values or expressions of some parameters may change with the development of novel materials and more accurate characterization techniques.

Fuel cell fundamentals Chapter | 2

69

Review questions Q1: Calculate the equilibrium potential of the cathode for a fuel cell operating at 80 C and a pressure of 1 atm. Q2: Calculate the reference exchange current density of the PEMFC cathode for the cell operating at 20 C, 40 C, 60 C, and 80 C. Q3: Find the effective electronic conductivity for the CL when the ionomer content is 20% and the porosity of the CL is 0.3. Q4: How the ionic conductivity of the ionomer membrane depends on temperature and water content in the membrane? Explain.

References [1] A.Z. Weber, S. Balasubramanian, P.K. Das, Chapter 2—Proton exchange membrane fuel cells, in: K. Sundmacher (Ed.), Advances in Chemical Engineering, 41, Academic Press, 2012, pp. 65
144. [2] P.K. Das, X. Li, Z.S. Liu, Analytical approach to polymer electrolyte membrane fuel cell performance and optimization, J. Electroanal. Chem. 604 (2) (2007) 72
90. [3] A.Z. Weber, et al., A critical review of modeling transport phenomena in polymerelectrolyte fuel cells, J. Electrochem. Soc. 161 (12) (2014) F1254
F1299. [4] D.M. Bernardi, M.W. Verbrugge, A mathematical model of the solid-polymer-electrolyte fuel cell, J. Electrochem. Soc. 139 (9) (1992) 2477
2491. [5] L. Xing, et al., Inhomogeneous distribution of platinum and ionomer in the porous cathode to maximize the performance of a PEM fuel cell, AIChE J. 63 (11) (2017) 4895
4910. [6] W. Sun, B.A. Peppley, K. Karan, An improved two-dimensional agglomerate cathode model to study the influence of catalyst layer structural parameters, Electrochim. Acta 50 (16
17) (2005) 3359
3374. [7] D.F. Cheddie, N.D.H. Munroe, Three dimensional modeling of high temperature PEM fuel cells, J. Power Sources 160 (1) (2006) 215
223. [8] F. Barbir, PEM Fuel Cells: Theory and Practice, Elsevier Academic Press, 2005. [9] M.S. Ismail, et al., Effects of anisotropic permeability and electrical conductivity of gas diffusion layers on the performance of proton exchange membrane fuel cells, Appl. Energy 95 (2012) 50
63. [10] L. Xing, et al., Multi-variable optimisation of PEMFC cathodes based on surrogate modelling, Int. J. Hydrog. Energy 38 (33) (2013) 14295
14313. [11] R.B. Scharifker, The kinetics of oxygen reduction in molten phosphoric acid at high temperatures, J. Electrochem. Soc. 134 (11) (1987) 2714. [12] W. Vogel, et al., Reaction pathways and poisons—II: the rate controlling step for electrochemical oxidation of hydrogen on Pt in acid and poisoning of the reaction by CO, Electrochim. Acta 20 (1) (1975) 79
93. [13] Z. Liu, et al., Study of the oxygen reduction reaction (ORR) at Pt interfaced with phosphoric acid doped polybenzimidazole at elevated temperature and low relative humidity, Electrochim. Acta 51 (19) (2006) 3914
3923. [14] E. Birgersson, M. Noponen, M. Vynnycky, Analysis of a two-phase non-isothermal model for a PEFC, J. Electrochem. Soc. 152 (5) (2005) A1021. [15] C. Marr, X. Li, Composition and performance modelling of catalyst layer in a proton exchange membrane fuel cell, J. Power Sources 77 (1) (1999) 17
27.

70

Fuel Cells for Transportation

[16] Z. Liu, Z. Mao, C. Wang, A two dimensional partial flooding model for PEMFC, J. Power Sources 158 (2) (2006) 1229
1239. [17] T. Sousa, M. Mamlouk, K. Scott, An isothermal model of a laboratory intermediate temperature fuel cell using PBI doped phosphoric acid membranes, Chem. Eng. Sci. 65 (8) (2010) 2513
2530. [18] A. Parthasarathy, et al., Temperature dependence of the electrode kinetics of oxygen reduction at the platinum/nafion interface—a microelectrode investigation, J. Electrochem. Soc. 139 (9) (1992) 2530
2537. [19] P.K. Das, X. Li, Z.S. Liu, Effective transport coefficients in PEM fuel cell catalyst and gas diffusion layers: beyond Bruggeman approximation, Appl. Energy 87 (9) (2010) 2785
2796. [20] T. Zhou, H. Liu, Effects of the electrical resistances of the GDL in a PEM fuel cell, J. Power Sources 161 (1) (2006) 444
453. [21] D. Pantea, et al., Electrical conductivity of conductive carbon blacks: influence of surface chemistry and topology, Appl. Surf. Sci. 217 (1
4) (2003) 181
193. [22] R.W. Powell, R.P. Tye, The promise of platinum as a high temperature thermal conductivity reference material, Br. J. Appl. Phys. 14 (10) (1963) 662
666. [23] S. Kamarajugadda, S. Mazumder, Numerical investigation of the effect of cathode catalyst layer structure and composition on polymer electrolyte membrane fuel cell performance, J. Power Sources 183 (2) (2008) 629
642. [24] F. Jaouen, G. Lindbergh, G. Sundholm, Investigation of mass-transport limitations in the solid polymer fuel cell cathode—I: mathematical model, J. Electrochem. Soc. 149 (4) (2002) A437
A447. [25] F. Jaouen, G. Lindbergh, K. Wiezell, Transient techniques for investigating mass-transport limitations in gas diffusion electrodes—II: experimental characterization of the PEFC cathode, J. Electrochem. Soc. 150 (12) (2003) A1711
A1717. [26] T.E. Springer, T. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell model, J. Electrochem. Soc. 138 (1991) 2334
2342. [27] R. Bird, W. Stewart, E. Lightfoot, Transport Phenomena, John Wiley and Sons, 2002. [28] H.C. Brinkman, A calculation of the viscous force exerted by a flowing fluid on a dense swarm of particles, Appl. Sci. Res. 1 (1) (1949) 27
34. [29] K. Broka, P. Ekdunge, Modelling the PEM fuel cell cathode, J. Appl. Electrochem. 27 (3) (1997) 281
289. [30] A. Suzuki, et al., Ionomer content in the catalyst layer of polymer electrolyte membrane fuel cell (PEMFC): effects on diffusion and performance, Int. J. Hydrog. Energy 36 (3) (2011) 2221
2229. [31] L. Xing, Modelling and Simulation of the Laboratory Low Temperature Proton Exchange and Direct Methanol Fuel Cells, Newcastle University, Newcastle, 2014. [32] N. Khajeh-Hosseini-Dalasm, et al., A study of the agglomerate catalyst layer for the cathode side of a proton exchange membrane fuel cell: modeling and optimization, Electrochim. Acta 60 (2012) 55
65. [33] C.R. Wilke, A viscosity equation for gas mixtures, J. Chem. Phys. 18 (4) (1950) 517
519. [34] B. Poling, J. Prausnitz, J. O’Connell, The Properties of Gases and Liquids, McGraw Hill, 2000. [35] L. Xing, et al., A two-phase flow and non-isothermal agglomerate model for a proton exchange membrane (PEM) fuel cell, Energy 73 (2014) 618
634.

Fuel cell fundamentals Chapter | 2

71

[36] L. Xing, et al., Numerical investigation of the optimal Nafions ionomer content in cathode catalyst layer: an agglomerate two-phase flow modelling, Int. J. Hydrog. Energy 39 (17) (2014) 9087
9104. [37] S.-Y. Lee, et al., Gradient catalyst coating for a proton exchange membrane fuel cell operation under nonhumidified conditions, Electrochem. Solid. State Lett. 10 (2007) B166
B169. [38] N. Khajeh-Hosseini-Dalasm, et al., A parametric study of cathode catalyst layer structural parameters on the performance of a PEM fuel cell, Int. J. Hydrog. Energy 35 (6) (2010) 2417
2427. [39] C.W. Wu, et al., A systematic review for structure optimization and clamping load design of large proton exchange membrane fuel cell stack, J. Power Sources 476 (2020) 228724. [40] T. Hottinen, et al., Inhomogeneous compression of PEMFC gas diffusion layer, J. Power Sources 171 (1) (2007) 113
121. [41] E.M. Khetabi, et al., Effects of mechanical compression on the performance of polymer electrolyte fuel cells and analysis through in-situ characterisation techniques—a review, J. Power Sources 424 (2019) 8
26. [42] W.Z. Li, et al., Three-dimensional modeling of a PEMFC with serpentine flow field incorporating the impacts of electrode inhomogeneous compression deformation, Int. J. Hydrog. Energy 44 (39) (2019) 22194
22209. [43] L. Holzer, et al., Microstructure-property relationships in a gas diffusion layer (GDL) for Polymer Electrolyte Fuel Cells, Part I: effect of compression and anisotropy of dry GDL, Electrochim. Acta 227 (2017) 419
434. [44] U.U. Ince, et al., Effects of compression on water distribution in gas diffusion layer materials of PEMFC in a point injection device by means of synchrotron x-ray imaging, Int. J. Hydrog. Energy 43 (1) (2018) 391
406. [45] M. Bosomoiu, G. Tsotridis, T. Bednarek, Study of effective transport properties of fresh and aged gas diffusion layers, J. Power Sources 285 (2015) 568
579. [46] I.V. Zenyuk, et al., Gas-diffusion-layer structural properties under compression via x-ray tomography, J. Power Sources 328 (2016) 364
376. [47] C. T¨otzke, et al., A dedicated compression device for high resolution x-ray tomography of compressed gas diffusion layers, Rev. Sci. Instrum. 86 (4) (2015) 043702. [48] R.W. Atkinson, et al., The role of compressive stress on gas diffusion media morphology and fuel cell performance, ACS Appl. Energy Mater. 1 (1) (2018) 191
201. [49] N. Kulkarni, et al., The effect of non-uniform compression and flow-field arrangements on membrane electrode assemblies—x-ray computed tomography characterisation and effective parameter determination, J. Power Sources 426 (2019) 97
110. [50] H. Chang, M. Chang, Effects of assembly pressure on the gas diffusion layer and performance of a PEM fuel cell, Appl. Mech. Mater. (2011) 110
116. [51] D. Jiao, K. Jiao, Q. Du, Numerical investigations of vapor condensation and water transport in gas diffusion layers of PEMFC, Int. J. Heat. Mass. Transf. 177 (2021) 121543. [52] M. Mehrtash, I. Tari, S. Yesilyurt, Impacts of inhomogeneous clamping force on local performance and liquid water formation in polymer electrolyte fuel cells, Int. J. Hydrog. Energy 42 (30) (2017) 19227
19245. [53] P. Zhou, C.W. Wu, Numerical study on the compression effect of gas diffusion layer on PEMFC performance, J. Power Sources 170 (1) (2007) 93
100. [54] Q. Wang, et al., The numerical study of vanadium redox flow battery performance with different electrode morphologies and electrolyte inflow patterns, J. Energy Storage 33 (2021) 101941. [55] H. Wu, P. Berg, X. Li, Steady and unsteady 3D non-isothermal modeling of PEM fuel cells with the effect of non-equilibrium phase transfer, Appl. Energy 87 (9) (2010) 2778
2784.

72

Fuel Cells for Transportation

[56] X.-G. Yang, Q. Ye, P. Cheng, Matching of water and temperature fields in proton exchange membrane fuel cells with non-uniform distributions, Int. J. Hydrog. Energy 36 (19) (2011) 12524
12537. [57] H. Wu, P. Berg, X. Li, Modeling of PEMFC transients with finite-rate phase-transfer processes, J. Electrochem. Soc. 157 (2010) B1
B12. [58] A.A. Shah, et al., Transient non-isothermal model of a polymer electrolyte fuel cell, J. Power Sources 163 (2) (2007) 793
806. [59] A.Z. Weber, J. Newman, Transport in polymer-electrolyte membranes I. Physical model, J. Electrochem. Soc. 150 (7) (2003) A1008
A1015. [60] A.Z. Weber, J. Newman, Transport in polymer-electrolyte membranes II. Mathematical model, J. Electrochem. Soc. 151 (2) (2004) A311
A325. [61] T. Uchiyama, M. Kato, T. Yoshida, Buckling deformation of polymer electrolyte membrane and membrane electrode assembly under humidity cycles, J. Power Sources 206 (2012) 37
46. [62] A. Kusoglu, et al., Mechanical behavior of fuel cell membranes under humidity cycles and effect of swelling anisotropy on the fatigue stresses, J. Power Sources 170 (2) (2007) 345
358. [63] M.N. Silberstein, M.C. Boyce, Hygro-thermal mechanical behavior of Nafion during constrained swelling, J. Power Sources 196 (7) (2011) 3452
3460. [64] S. Ge, et al., Absorption, desorption, and transport of water in polymer electrolyte membranes for fuel cells, J. Electrochem. Soc. 152 (2005) A1149. [65] T. Suzuki, K. Kudo, Y. Morimoto, Model for investigation of oxygen transport limitation in a polymer electrolyte fuel cell, J. Power Sources 222 (2013) 379
389. [66] N. Nonoyama, et al., Analysis of oxygen-transport diffusion resistance in protonexchange-membrane fuel cells, J. Electrochem. Soc. 158 (4) (2011) B416. [67] Y. Takamura, et al., Effects of temperature and relative humidity on oxygen permeation in Nafion (R) and sulfonated poly(arylene ether sulfone), ECS Trans. (2008) 16. [68] U. Pasaogullari, C.-Y. Wang, Two-phase transport and the role of micro-porous layer in polymer electrolyte fuel cells, Electrochim. Acta 49 (25) (2004) 4359
4369. [69] D. Weng, et al., Electro-osmotic drag coefficient of water and methanol in polymer electrolytes at elevated temperatures, J. Electrochem. Soc. 143 (1996) 1260
1263. [70] M.C. Leverett, Capillary behavior in porous solids, Trans. AIME 142 (01) (1941) 152
169. [71] K. Udell, Heat transfer in porous media considering phase change and capillarity—the heat pipe effect, Int. J. Heat. Mass. Transf. 28 (1985) 485
495. [72] P.K. Das, X. Li, Z.S. Liu, Analysis of liquid water transport in cathode catalyst layer of PEM fuel cells, Int. J. Hydrog. Energy 35 (6) (2010) 2403
2416. [73] P.K. Das, et al., Effects of catalyst layer structure and wettability on liquid water transport in polymer electrolyte membrane fuel cell, Int. J. Energy Res. 35 (15) (2011) 1325
1339. [74] E.C. Kumbur, K.V. Sharp, M.M. Mench, Validated Leverett approach for multiphase flow in PEFC diffusion media—I. hydrophobicity effect, J. Electrochem. Soc. 154 (2007) B1295
B1304. [75] P.K. Das, et al., Liquid-water-droplet adhesion-force measurements on fresh and aged fuel-cell gas-diffusion layers, J. Electrochem. Soc. 159 (5) (2012) B489
B496. [76] D. Thumbarathy, et al., Fabrication and characterization of tuneable flow-channel/gas-diffusion-layer interface for polymer electrolyte fuel cells, J. Electrochem. Energy Convers. Storage 17 (1) (2020) 011010.

Chapter 3

Fuel cell modeling and optimization Lei Xing1, Xueguan Song2 and Prodip K. Das3 1

School of Chemistry and Chemical Engineering, University of Surrey, Guildford, United Kingdom, 2State Key Laboratory of High-Performance Precision Manufacturing, School of Mechanical Engineering, Dalian University of Technology, Dalian, P.R. China, 3School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom

3.1

Introduction

Modeling plays a significant role in the process of proton-exchange membrane and related fuel cell design and development [1]. Normally, fuel cell design and development processes begin with a set of requirements, including power output, operating conditions, size limitations, safety specifications, and others. Based on the knowledge of materials and processes involved in the fuel cells, modeling is performed to predict the fuel cell performance. Modeling helps the designers and developers to determine the best candidate designs or improve the existing designs that satisfy the requirement. Modeling provides a better understanding of the electrochemical reactions and mass transport that occurred within the fuel cells, for example, the reactants profiles, temperature distribution, and polarization curves [2
5]. It can give a quick prediction of the fuel cell performance under various given operating conditions, material properties, and fuel cell geometries. Modeling reduces the time, effort, and cost associated with the experimental studies and provides theoretical guidance on the development and optimization of the fuel cells. The processes that occurred inside the porous electrode of proton-exchange membrane fuel cells (PEMFCs) constitute a fully coupled reaction
diffusion process. Due to the competitive relationship between the electrochemical reaction and species transport [6], optimal cell performance requires graded distributions of the functional compositions inside the electrodes and novel flow fields, where the optimal parameters for maximized cell performance, for example, platinum (Pt) loading, ionomer loading, porosity, and hydrophobicity of the electrodes, vary according to the different operational requirement of Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00018-6 © 2023 Elsevier Ltd. All rights reserved.

73

74

Fuel Cells for Transportation

PEMFCs. The optimal distributions of different functional components and novel flow field design could be achieved through multivariable optimization approaches. Thus, modeling and optimization could accelerate the commercialization and industrialization of PEMFC technology.

3.2 Fuel cell modeling approach and key physicochemical and operating parameters A complete PEMFC model includes several sub-models, including gas flow, mass and heat transport, mechanical properties of the electrodes, water transport through the membrane, gas
liquid two-phase flow, proton and electron transport, reaction kinetics, and current distributions, as shown in Fig. 3.1. The interaction of different sub-models describes the computational process in solving a completed PEMFC model. To reduce the complexity of the model, several sub-models are typically excluded. For example, when heat transfer is omitted, a PEMFC model becomes an isothermal model. When the generation and transport of liquid water are neglected, the PEMFC model can be simplified to a single-phase flow model. In most PEMFC models, the gas flow, mass transport, and current distribution are essential. Key operating parameters include inlet gas velocity, relative humidity, operating pressure,

FIGURE 3.1 Schematic computational process in solving a multi-physics, non-isothermal, multiphase flow model of a PEMFC. Symbols have their usual meanings and are described in Chapter 2.

Fuel cell modeling and optimization Chapter | 3

75

temperature, and cell voltage. To reflect the catalytic activity and mass and ion transport resistance of the porous electrode, Pt loading, platinum and carbon mass ratio, ionomer volume fraction, conductivity, porosity, and electrode thickness are also important. The governing equations describing fluid momentum, diffusive and convective mass transport, electrochemical reaction kinetics, and heat transport are found in Chapter 2. Different from other high-temperature fuel cells, water management is critical in low-temperature fuel cells, which are typically operated below 100 C. The following sections focus on the modeling of water formation and transport of PEMFCs.

3.2.1

Water formation and transport in fuel cells

Perfluorinated membranes, such as Nafion, are typically used as the electrolyte in PEMFCs. Nafion ionomer is required in the catalyst layers (CLs) to facilitate proton transport from the anode CL, through the membrane, to the cathode CL. On the membrane
CL boundary, the membrane is closely connected with the ionomer in the CLs, at both the anode and cathode. Water in PEMFCs acts as the lubricant which makes the fuel cell system run smoothly at a relatively low ionic resistance. Water can be fed into the PEMFC system through the gas inlet and/or generated by the oxygen reduction reaction (ORR) at the cathode. Water exists as water vapor in the gas mixture, which is absorbed by the Nafion membrane/ionomer as the dissolved water. During fuel cell operation, protons migrate through the membrane and are associated with a drag of water molecules from the anode to the cathode, which is the electroosmotic drag (EOD). Together with the electrochemical water production, the water content of the membrane/ionomer becomes saturated, which then leads to liquid water formation through membrane/ionomer desorption when the water content exceeds the equilibrium value. The other reason for water formation is attributed to vapor condensation when the partial pressure of vapor exceeds the equilibrium vapor pressure at a given temperature. Due to the generated gradient of water concentration between the anode and cathode, a certain amount of water could diffuse back from the cathode to the anode, which is in the opposite direction against the EOD. In addition, the pressure difference between the anode and cathode could drive water transport through the membrane, which is called hydraulic permeation. The water transport that occurred in a PEMFC is schematically shown in Fig. 3.2. Maintaining a subtle equilibrium between membrane dehydration and liquid water flooding is the key issue to achieving maximum performance and durability for PEMFCs [7]. On the one hand, water is required to guarantee the good proton conductivity of the proton-exchange membrane and ionomer of CLs [8]. Dehydrated membrane/ionomer hinders the access of protons to the active sites within the CL, resulting in an increase in activation polarization [9]. On the other hand, excess water blocks the flow channel and the void space within the porous electrodes and then increases the mass transport

76

Fuel Cells for Transportation

FIGURE 3.2 Water transport mechanism inside a proton-exchange membrane fuel cell.

resistance, leading to a “flooding” problem. The thickness of the membrane, the water content of the membrane, the humidity of the reactant gases, and the reaction rate in CL determine the concentration gradient of water between the anode and cathode. Furthermore, membrane water content and reactant gas humidity are dependent on the gas inlet humidification and the temperature and pressure in the gas channels [10]. Besides, back-diffusion prevails over EOD at lower current density, EOD prevails over backdiffusion while higher current density is achieved, and thus, the anode (including the membrane) tends to dry out, even the cathode is well hydrated at high current density [11].

3.2.2

PEMFC modeling approaches

Numerical modeling of PEMFCs is important for a better understanding of the transport processes owing to the experimental drawbacks, such as the difficulty of performing the different experimental measurements simultaneously, unrealistic operating conditions, and the high cost of materials and testing instruments. Over the last decades, numerous models and studies have been conducted to describe water management and investigate liquid water transport in PEMFCs [12
18]. Depending on different descriptions of water formation and transport, the models can be mainly categorized into four groups: dynamic models, lumped models, flooding models, and models associated with the effect of geometrical configuration.

Fuel cell modeling and optimization Chapter | 3

77

In the early 1990s, some simple models have been developed. These models were all simplified models which all applied the assumption that the reactant and charge only transport along one direction. The numerical models developed by Springer et al. [19,20] and Bernardi and Verbrugge [21,22] are usually considered pioneering modeling works for PEMFCs. These models are essentially one-dimensional (1D) models considering the membrane, CL, and gas diffusion layer (GDL) based on solving the conservation equations by assuming homogeneous materials and using effective transport properties. After that, Nguyen and White [23] and Fuller and Newman [24] developed pseudo-two-dimensional (2D) models by further considering the flow channels, which considered the effect of water humidity inlet and temperature distributions, providing more detailed water and thermal management capability. However, the models developed in that period were too simple to simulate the very complex PEMFC systems although they laid the foundation for fuel cell modeling. More numerical models were developed in the late 1990s. For instance, Yi and Nguyen [25,26] and Gurau et al. [2] developed 2D models to explore more detailed transport phenomena in PEMFCs. These models illustrated the utility of multidimensional models in the understanding of the internal conditions of fuel cells, such as the reactant and water distribution. Gloaguen and Durand [27], Bultel et al. [28
30], and Marr and Li [31] developed the agglomerate models. These models applied simplified CL structures by assuming that the large agglomerates were formed by ionomer and platinum/ carbon particles on the level of micrometer. Compared to the models developed earlier, more detailed mass and charge transport phenomena were analyzed more accurately because these models extended the 1D or pseudo-2D to the detailed 2D models. It is important to note that all the models developed were based on the single-phase assumption, which treated the water as vapor, including the water formed in the cathode CL, and supplied it along with the humidified gas inlet. The water condensation within the porous electrodes and the flow channels was not considered. Although there was not sufficient evidence to confirm that water could be condensed in CLs and GDLs while the PEMFC system operated in normal conditions, however, condensed water in the flow channels has been observed by some instruments such as high-resolution cameras [32,33]. Generally, an ideal singlephase assumption is applied when reactant gases are oxidized or reduced at the surface of a solid catalyst (Pt-Ru and other binary or ternary alloys). In reality, the condensed water could change the single-phase flow problem to a two-phase flow problem. In the 2000s, multidimensional models have been developed by many researchers to solve a complete set of conservation equations (such as continuity and Navier
Stokes) coupled with electrochemical reactions (e.g., the ORR kinetics at the cathode). Computational fluid dynamics (CFD) code and commercial software (such as ANSYS Fluent and COMSOL Multiphysics)

78

Fuel Cells for Transportation

based on finite volume methods were adopted to develop the fuel cell models, and more complicated geometry and transport phenomena were investigated [34
44]. In the area of 3D geometry, the models developed by Dutta et al. [34,35], Zhou and Liu [36], Berning et al. [37], Mazumder and Cole [38], Lee et al. [40], Um and Wang [41], and Wang and Wang [42] were mainly considered a single flow channel with the major components of reactant gases. Large-scale simulations considering multichannel or small stacks give a more accurate and more specific analysis of the distribution of reactant gases (H2 and O2), water vapor, and pressures [45
48]. The main impediment to the widespread use of these multidimensional models is the requirement for computer hardware. A reasonable simplification of the complex multidimensional models is considered as the practical way of modeling the complicated transportation and reaction inside of the PEMFCs with the relatively low computational requirement. To simplify models and reduce computation time for conservation equations, liquid water formation can be neglected by assuming liquid water as supersaturated water vapor [23
25]. The real two-phase flow models, which give more accurate predictions than the single-phase assumptions, have also been developed [39,49
55]. These two-phase flow models solved the mass, momentum, and species transport conservation equations for the gas mixture, with an extra conservation equation for liquid water transport. However, the accurate and detailed water transport behavior cannot be studied in these models because the interface tracking between liquid water and gas is not permitted. As a result, the volume of fluid (VOF) model, based on liquid water dynamics to investigate water flow in a single serpentine flow channel, was developed [56]. The real 3D model is difficult to simulate a single cell with a large size due to the expensive computations. Therefore, many researchers simplified the sub-models of CLs and developed a “3D 1 1D” model to improve the computational efficiency by approximately one order of magnitude. This model can be employed to help develop the large-size flow fields of PEMFCs. For example, Xie et al. [57] developed a “3D 1 1D” PEMFC model to implement large-scale simulation with enhanced calculation efficiency, in which bipolar plates, gas channels, and GDLs were treated as 3D computational domain and microporous layers, CLs, and membrane are treated as 1D domain. The computation time was reduced by 20 folds for large-scale PEMFCs with 345 cm2 active area. In their later work [58], the accuracy of the developed “3D 1 1D” PEMFC model was comprehensively validated in terms of overall cell performance and local distributions of current density and temperature under different operating conditions. This approach provided some guidance for researchers and engineers in the field of PEMFC design by facilitating the application of 3D modeling and simulation to large-scale PEMFCs. In short, a large number of numerical models were developed around the world, and the models were chronologically developed from simple to

Fuel cell modeling and optimization Chapter | 3

79

complex, and from general to specific. Despite the great changes in the equation forms in the models, all the processes are described based on some basic laws, for example, the laws of conservation of energy and momentum. Developing a comprehensive and comparatively simple model for the PEMFCs under practical load conditions by reasonably simplifying the complex multidimensional models is the priority of most modeling work. The main difference between the models developed in different periods is presented in Table 3.1.

3.2.3

Modeling of water transport through the membrane

The modeling of water transport in the membrane can be classified into three types: diffusive, chemical potential, and hydraulic models. The diffusive model can be explained with dilute solution theory by considering the membrane as a solvent, while water and proton as solute. This theory assumes that the interaction between different solute species can be neglected, and only the interaction between solute (water and proton) and solvent (membrane) is considered. Then, the flux of solute species in the solvent can be described by using the Nernst
Planck equation. The Nernst
Planck equation is a conservation of mass describing the flux of ions under the influence of both an ionic concentration gradient and an electric field. The general form of the Nernst
Planck equation is:
 dci D i zi e 5 r Di rci 2 uci 1 ci rϕ ð3:1Þ kB dt where t (s) is time, Di (m2 s21) is the diffusivity of the solute species i, ci (mol m23) is the concentration of the solute species i, u (m s21) is the velocity of the fluid, zi (C) is the valence of ionic specie, kB (J K21) is the Boltzmann constant, and ϕ (V) is the potential. For water transport, Di zi eci rϕ=kB becomes zero because water is in zero valence. For proton transport, Di rci becomes zero by assuming a constant concentration of proton through the membrane. Furthermore, uci becomes zero for both the water and proton transport because the membrane does not move. Therefore, the Nernst
Planck equation for proton transport is simplified to Ohm’s law. By further considering the effect of EOD, the water transport is simplified to [59]: Jw 5 DM rcw 1 ndrag

i F

ð3:2Þ

where DM (m2 s21) is the diffusivity of water through the membrane, cw (mol m23) is water concentration, ndrag is EOD coefficient through the membrane, F (C mol21) is Faraday’s constant, and i (A m22) is the current density. The diffusive model is the most successful model for water transport through membrane since its initial application [19,20]. By further considering

TABLE 3.1 Chronological development of PEMFC modeling work.

Models

Period of development

Dimensions of reactant, products, and charge transportation

Accuracy of modeling process compared to practical fuel cell

Complexity of the models and computer hardware requirement

1D and pseudo-2D

Early 1990s

One-dimensional

Crude

Simple and low

2D

Later 1990s

Two-dimensional

Medium

Medium

3D and multidimensional

After 2000s

Three-dimensional

Precise

Complex and high

Fuel cell modeling and optimization Chapter | 3

81

the interaction between different solute species, the chemical potential model is developed. The proton and water transport through membrane, therefore, can be explained by concentrated solution theory as [60,61]: ndrag σM rLH i 5 -σM rϕð3:3Þ F Jw 5 2 αw rLw 1 ndrag

i F

ð3:4Þ

where σM (Ω21) is the membrane conductivity, LH (J mol21) and Lw (J mol21) are the chemical potential of proton and water, respectively, and αw is water transport coefficient through the membrane. Comparison of Eqs. (3.2) and (3.4) shows that the concentration (c) is replaced by chemical potential (L), the diffusion coefficient (D) is replaced by transport coefficient (α), and one more term is added in Eq. (3.3) to account for the multicomponent interaction. The biggest obstacle to the widespread use of the chemical potential model is the difficulty in obtaining the transport parameters. As a result, the chemical potential model is rarely used in comparison with the diffusive model. In the diffusive and chemical potential models, the convective transport caused by the pressure gradient across the membrane is not considered. However, convective transport could happen when water enlarges the pores of the membrane. To fill the gap, the hydraulic model is developed [21,22]. Consequently, water flux due to pressure gradient and EDO can be calculated by the Nernst
Planck equation. Generally, the hydraulic model neglects the diffusive transport, and the water flux can be represented by the following equation: Jw 5 2 cw

kp i rpw 1 ndrag F μw

ð3:5Þ

where kp (m2) and μw (Pa s) are the permeability and dynamic viscosity of water in the membrane. Due to the effect of EDO, the anode side of the membrane is often prone to dry out. Therefore, the fully hydrated membrane assumption remains questionable. In fact, water convective transport is only considerable while the pressure gradient exists between the anode and cathode. If the inlet gas pressure of the anode is as same as that of the cathode, the water convective transport could be neglected because diffusive transport and EDO have a more significant influence on the water transport.

3.2.4

Modeling of water transport through porous electrodes

Depending on the assumption of the morphology of the porous electrodes, models of the water transport through the porous electrodes can be categorized into “homogenous” and “non-homogenous” models. In the homogenous model, the porous electrodes are assumed to be constructed by homogeneous

82

Fuel Cells for Transportation

materials, while the non-homogenous model applies the real or simplified microstructure of the porous electrodes. When the porous electrodes are assumed as homogeneous, the entire computational domain shares the same properties, such as conductivity, permeability, and porosity. The overall effect of the microstructure is usually reflected by the effective coefficient. Thus, the geometry generation and the model-solving process are greatly simplified and result in easier model development and lower requirement of the computer hardware. The disadvantage of this assumption is the lack of precision about the detailed process within the porous electrodes, such as nucleation water. The modeling works before the 2000s usually adopt the homogeneous assumption [2,15,16,19
30]. The GDL is typically made up of 3D random carbon fiber, and its real structure is highly anomalous [62
64]. To accurately simulate the electrochemical reactions and transport processes that occurred in a real or simplified GDL, for example, liquid water formation and transport, numerous models have been developed, including the VOF model [10,41,42,55,65
67] and Lattice Boltzmann (LB) model [68
71]. The details of the VOF approach are found in Chapters 2 and 13.

3.2.5

Catalyst layer modeling

The CLs are the core of a PEMFC, in which electrochemical reactions occurred. The CLs are prepared by spraying the catalyst particle (such as PtRu alloy) containing ink onto the carbon paper. The difference between the CL and other layers is that the catalyst particle is surrounded by the carbon fiber in the CLs and results in much smaller pores than that of the GDLs. Therefore, the VOF model is hard to be applied to the microstructure of the CLs [62]. The general process of the electrochemical reaction includes two steps, diffusion and reaction. First of all, the reactants must transport through the porous media and ionomer films and reach the surface of the catalyst particles. Then, the reactants are absorbed on the surface of the catalyst particle where products are generated via chemical reactions. Finally, the products generated on the surface of the catalyst particle must move away. The modeling approaches applied on the catalyst layer, depending on the degree of complexity, can be categorized into three groups. In the simplest approach, the CLs are treated as reactive boundaries between the GDL and membrane [37,72]. For example, Jeng et al. [72] developed a simple 2D across-the-channel model to study the mass transport of the reactant gases through the GDLs. The effectiveness of the GDLs was evaluated under different current densities, and an optimal thickness of the cathode GDL was suggested. Berning et al. [37] developed a 3D, non-isothermal model to investigate the temperature distribution in the MEA. However, the CLs were treated as 1D boundaries in this model. It is important to notice that this kind of approach often overestimates the current density due to the ignorance of the mass transport resistance in the catalyst layer. The second approach

Fuel cell modeling and optimization Chapter | 3

83

assumes the CL as a thin film fully flooded with liquid water [31] or the existence of a flooded interface between the GDL and CL [73]. The “thin film fully flooded” approach is adopted in the model developed by Marr and Li [31], in which the void space within the cathode CL was fully occupied by liquid water. The “flooded interface” approach is adopted in the analytical model developed by Das et al. [73]. These approaches led to reasonable simulation results of the cell at higher current densities. However, it underestimates the cell performance at lower current densities due to an increase in mass transport resistance within the CL. The most accurate approach to CL modeling is the agglomerate model, in which the CL is assumed as a porous 2D or 3D domain filled with Pt/C catalyst particles agglomerate surrounded by thin ionomer films [74
78]. The void spaces within the agglomerates (intra-agglomerate space) and between the agglomerates (inter-agglomerate space) are defined as the primary and secondary pores, respectively. Both the primary and secondary pores can be filled with ionomer, liquid water, and reactants. The agglomerate models can be further subcategorized into three subgroups, namely slab, cylindrical, and spherical agglomerate models [63]. To give a more accurate simulation of the diffusion
reaction process that occurred in the catalyst layer, agglomerate models are usually preferred. In the agglomerate models, the catalyst particles, ionomer, and void space are assumed to be homogeneously mixed to form the micrometer agglomerates. In the spherical agglomerate model, the diameter of the agglomerate can be less than 10 μm [64]. The agglomerate models are usually adopted to describe the relatively sluggish ORR of the cathode CL of PEMFCs.

3.3

Numerical optimization of PEMFCs

Numerical optimization has been an active research area since the 1960s. It has been used in many applications. The common principle of numerical optimization is to efficiently search for an optimal design in a coupled mathematical algorithm with the aid of a computational analysis tool. Only a few designs need to be evaluated using the optimization algorithm, so the computational time is therefore reduced. Optimal design helps the researchers to create a new design or improve an existing one. To account for the interrelationship between various parameters and optimize several objectives simultaneously in a fuel cell design, multi-objective optimization is always used. A mathematical formulation of such a problem is given by:

T Miximise or minimise JðxÞ 5 J1; J2; J3; . . .. . .; Jn W:R:T xk for k 5 1; 2; 3; . . .. . .; n ð3:6Þ Subject to :hi ðxÞ 5 0 for i 5 1; 2; 3; ??; p gi ðxÞ # 0 for i 5 1; 2; 3; ??; q xL # x # x U

84

Fuel Cells for Transportation

where J(x) is the vector of objectives. In fuel cell design, the objective can be cost, performance, durability, and others, which is represented by Ji individually. xk is known as the design variables, which are related to the objective vector, J(x). xL and xU are the lower and upper bounds of design variable xk. hi(x) and gi(x) are the design constraints. The design of fuel cells is a challenging endeavor because the multitude of physical and chemical phenomena needs to be optimized simultaneously to achieve the best performance. Normally, it requires the evaluation of a set of possible designs due to the fact that the number of possible designs increases sharply as the number of design variables increases. For example, the number of possibilities is 105 for a design with ten variables and five possibilities per variable. It is impossible to evaluate all possible designs. As a result, some design variables have to be constrained as constants to reduce the number of possible designs. This is the so-called suboptimal design. The suboptimal designs in fuel cell optimization mainly focus on the following aspects: electrode design, flow field design, fuel cell stack, and operating condition optimization. Similarly, the optimization has to mainly concentrate on limited important objectives, while ignoring other design objectives. Numerical optimization could provide insight into fuel cell design including cost reduction, performance improvement, and efficiency increment. As a relatively new research area, numerical optimization of needed fuel cells has attracted growing interest.

3.3.1

Electrode optimization

During PEMFC’s operation, the electrochemical reaction and reactant mass transport are fully coupled, and the overall rate is determined by the slowest process. The detailed mass transfer in porous electrodes is discussed in Chapter 13. While higher catalyst loading increases reaction activity and reduces activation overpotentials [79], it also reduces reactant transport and liquid water capacity inside the catalyst layer. It was shown in Ref. [79] that the catalyst loading of 0.3 mg cm22 or higher has limited or no change in activation overpotential. Thus, the functional components of the porous electrode must be optimized to achieve maximum cell performance. Holdcroft’s group is considered the pioneer in optimization of the fuel cell electrode using a numerical optimization approach [80,81]. The CL composition was optimized to achieve the maximum current density at the cell voltage of 0.6 V. The design variables were ionomer volume fraction, Pt loading, and CL’s thickness. The optimal distributions of Nafion ionomer and platinum were obtained. The optimization results indicated that the optimal ionomer loading was around 30 wt.% [80], and the electrode performance was improved by placing more ionomer and platinum near the membrane [81]. Das et al. [73] developed a single-objective analytical approach using the exact solution of activation overpotential for cathode CL’s optimization.

Fuel cell modeling and optimization Chapter | 3

85

The CL composition (ionomer and platinum loadings) and thickness were optimized to achieve maximum power density at a given cell voltage for both air and oxygen. It was found that Pt loading of about 0.2 mg cm22 can provide optimum performance at 0.8 V, which is in line with the conclusion of Song et al. [80] at 0.6 V. The optimum CL thickness with a Pt loading of 0.2 mg cm22 was found to be between 9 and 11 μm at the cell voltage of 0.8 V. However, the optimum CL thicknesses vary between 9 and 16 μm depending upon Pt loadings. Conversely, Secanell et al. optimized both Pt loading and performance of a complete MEA [82] based on the previously developed optimization framework [83]. The design variables included Pt loading, ionomer loading, GDL porosity, and platinum mass ratio. Fig. 3.3 shows the optimization of Pt loading and CL thickness (for various Pt loadings) at the cell voltage of 0.8 V of Ref. [73] and the polarization curves of the base-case design and optimal design at the cell voltage of 0.676 and 0.476 V of Ref. [83]. Fig. 3.3 shows that the cell performance was improved using the parameters obtained from the optimal design. The optimization results showed that Pt loading had to be controlled within the range of 0.1 to 0.5 mg cm22, as higher loading resulted in a waste of platinum rather than an increase in current density. For PEMFCs operated at various loads, the required activities and mass transport rates are different because the reactant and product are nonuniformly distributed inside the membrane electrode assembly. Thus, a rational design for a membrane electrode assembly (MEA) with a spatial distribution of functional components helps reduce the usage of precious components, improve cell performance, and achieve uniform distributions of current density and heat. Thus, the graded design of the functional components in the GDL, MPL, CL, and membrane along both the through-plane and in-plane directions within the MEA was reviewed to reduce the cost and improve the performance and durability of PEMFCs [84]. The simulation with a linear porosity gradient in the cathode GDL suggested remarkable improvement of the limiting current density and oxygen usage with an optimal linear porosity gradient with a porosity of 70% near the channel and 30% near the CL for the parallel and z-serpentine channel designs [85]. The results for the parallel channel are shown in Fig. 3.4. It was found that the current density increases from 1.41 to 1.66 A cm22 when the abovementioned graded porosity distribution was applied. The same trends are also observed for parallel and interdigitated channel designs. Weng et al. [86] designed a hydrophobicity-graded MPL and experimentally studied the cell performance under a variety of humidification conditions. Three MPLs with various PTFE contents, 20, 25, and 30 wt.% in the MPLs from the CL/MPL interface to the MPL/GDL interface, were sandwiched between the CL and GDL. Thanks to the relatively low PTFE loading inside the inner layer of the as-prepared MPL, product water from the ORR was retained within the CL under low humidity conditions, for example, 5%. In contrast, the hydrophobicity-graded

86

Fuel Cells for Transportation

FIGURE 3.3 Optimization of Pt loading and catalyst layer thickness at the cell voltage of 0.8 V (parts (A) and (B)) and polarization curves of the base-case design and optimal design at the cell voltage of 0.676 and 0.476 V (parts (C) and (D)). Data for parts (A) and (B) are taken from P.K. Das, X. Li, Z.-S. Liu, J. Electroanal. Chem. 604 (2007) 72
90, and parts (C) and (D) are taken from M. Secanell, K. Karan, A. Suleman, N. Djilali, Electrochim. Acta 52 (2007) 6318
6337.

FIGURE 3.4 Predicted cell performance for a porosity-graded cathode GDL with parallel channel design. Data are taken from Y.-X. Huang, C.-H. Cheng, X.-D. Wang, J.-Y. Jang, Energy 35 (2010) 4786
4794.

Fuel cell modeling and optimization Chapter | 3

87

MPL efficiently discharged liquid water from the porous electrodes at high relative humidity, for example, 50%, leading to improved cell performance, as shown in Fig. 3.5. Srinivasarao et al. [87] developed a 2D and two-phase model to optimize Pt loading, ionomer loading, weight fraction of platinum on carbon, and CL thickness of a novel design with multiple CLs. They suggested that to achieve the same cell performance as that obtained from a single CL design, the overall loading of platinum should be reduced when a multilayer CL is applied. Moreover, the optimal Pt loading of the CL close to the GDL was higher than that of the CL close to the membrane, as shown in Fig. 3.6, in which CL1 represents the MPL-CL interface and CL4

FIGURE 3.5 The cell performance of the commercial MPL (34BC) and GMPL under various relative humidity conditions: the GMPL consists of 34BA (10 wt.% PTFE) and three sublayers with various PTFE loadings from 20 to 30 wt.%. Reprinted from Weng et al., Int. J. Hydrogen Energy 36 (21) (2011) 13708
13714, with permission from Elsevier.

88

Fuel Cells for Transportation

FIGURE 3.6 Optimum distribution of platinum loading in a multilayer CL at various cell voltages: CL1 is the sublayer close to the GDL, and CL4 is the sublayer close to the membrane. Reprinted from Xing et al., Energy 177 (15) (2019) 445
464, with permission from Elsevier.

corresponds to the CL
membrane interface. In the base-case design, the Pt loading was fixed at 0.25 mg cm22 and uniformly distributed in different sublayers. At high cell voltages, the optimal platinum distribution almost does not change. At decreasing cell voltage, which corresponds to an increase in current density, the optimal Pt loading decreases from the CL
membrane interface toward the MPL
CL interface. At 0.4 V, the optimal Pt loading within the CL
membrane interface decreases to 0.15 mg cm22. This trend was confirmed by the work of Xing et al. [88]. For a function-graded PEM, a gradient density of sulfonic acid groups along the membrane thickness direction is typically used as the definition [89]. In most previous studies, such partially fluorinated sulfonic acid membranes (part-FSAs) were prepared using irradiation methods, for example, a low-energy electron beam (EB) [90,91]. The mechanism of improving the cell performance is the control of the membrane water uptake. Sato et al. [89] fabricated an FN (hybrid membrane) by mixing s-FEP (sulfonated radiation-grafted membrane) powder with a Nafion dispersion and compared the water uptake, ion exchange capacity (IEC), and cell performance achieved by using FN, s-FEP, and Nafion 112 as the membrane, respectively. The IEC and water uptake of the FN were improved compared to those of the Nafion 112 and s-FEP, resulting in the highest cell performance among the tested membranes, as shown in Fig. 3.7. The design of GDL perforation also enhances the gas supply and water removal, which is discussed in Chapter 13.

Fuel cell modeling and optimization Chapter | 3

89

FIGURE 3.7 Cell performances of MEAs based on FN (mixture of the radiation-grafted membrane with Nafion ionomer), s-FEP (sulfonated radiation-grafted membrane), and Nafion 112. Solid lines are for cell voltage, while dashed lines represent corresponding power densities. Data are taken from Y. Sato, K. Fujii, N. Mitani, A. Matsuura, T. Kakigi, F. Muto, et al., Nucl. Instrum. Methods Phys. Res. Sect. B: Beam Interact. Mater. At. 265 (2007) 213
216.

3.3.2

Flow fields optimization

The early efforts in the flow field design and optimization were published in 2004 [92]. The unique optimization objective was to maximize the current density at a cell voltage of 0.7 V with an interdigitated flow field. The design parameters included cathode pressure inlet, cathode GDL thickness, the width ratio of the gas channel, and current collector. However, all design variables reached the bounds of the optimal design. It was shown that the channel width ratio and the porosities of GDLs and CLs can be optimized using a gradient-based optimization algorithm (simplified conjugate gradient method) [93]. The optimal channel width ratio of 0.54, GDL porosity of 0.6, and CL porosity of 0.3 were obtained in the optimization results. To maintain sufficient reactant supply, excellent water removal, and low pressure drop, various novel flow field designs, in addition to traditional parallel, interdigitate, and serpentine design, have been proposed and studied [94
101]. For instance, a porous-blocked baffled flow channel was designed [94], in which porous blocks were installed between GDL and baffles. A two-phase, non-isothermal model was developed to investigate the mass transport mechanisms and optimize the porosities of the blocks at different locations along the channel for improved cell performance and reduced pumping power. In their later work [95], the baffled channel with three different leeward lengths along the flow field channels was designed, and the pumping power, net power, and power density were experimentally studied. A most recent modeling work on the combination of baffles and secondary

90

Fuel Cells for Transportation

porous layer includes the complete formulation of the Forchheimer inertial effect [96]. The porous and baffled flow field improved the cell performance by minimizing mass transport losses and enhancing water removal from the GDL. The metal foam and honeycomb flow fields were also employed to facilitate the water management and reactant transport within the porous electrode and to achieve higher uniformities of reactant and current distributions than the traditional parallel straight flow field [97,98]. To improve the water removal and cell performance, a series of sub-inlet at the parallel cathode flow channels was introduced [99,102] and was found that a suitable location of the sub-inlet along the channel and a reasonable amount of dry air fed from the sub-inlet could benefit water removal and cell performance with a small pressure along the channel. Novel confrontational design of flow fields is only important when considerable liquid water was generated at medium and high current densities, as different configurations of flow fields have very limited influence on the cell performance at low current densities [103
105]. In the past, different shapes of flow fields named by alphabet letters, such as S-shaped [100], M-like [101], and Z-shaped [106] flow fields, have been applied to promote the overall performance of fuel cells through the enhancement of mass transport and water removal. As shown in Fig. 3.8, a novel flow field was designed recently by deploying auxiliary channels inside the partially hollow ribs and drilling a series of arrayed holes on the auxiliary channels [107]. This novel design rationally utilizes the ribs of the current collector and improves the volumetric efficiency of the parallel channels, leading to improved cell performance and homogeneity of current

FIGURE 3.8 A novel flow field design with auxiliary channels and arrayed holes. Reprinted from Wang et al., AIChE J. 68 (2) (2022) e17461, with permission from John Wiley and Sons.

Fuel cell modeling and optimization Chapter | 3

91

distribution. It showed that an optimization of the flow field geometry, that is, the hole size, the area ratio of arrayed holes and auxiliary channels, nonuniform distribution of arrayed holes, could further improve the cell performance and current uniformity at an extremely lower pressure drop. Commercialized flow fields, such as the 3D fine mesh design used in Toyota Mirai [108,109] and the wave-like flow fields implemented in Honda Clarity [110], are promising solutions regarding the liquid water removal from the cell at high current density. In addition, bio-inspired flow fields, such as tree-shaped [111] and lung-shaped [112] flow fields, are commendable attempts to redistribute the reactant gas and improve the forced convection. The mass transfer in the novel flow fields is discussed in detail in Chapter 13. The modification of channel geometry and installation of obstacles in the channel is to increase the gas velocity near the outlet region and change the laminar gas flow to a turbulent pattern, with the aim of enhancing the mass transport through the porous electrodes. However, these tapered channels and baffle channels suffer from technical difficulties in the manufacture and significant pressure drop. In addition, the precious control of the pressure difference between adjacent channels is a great challenge in these configurations. Thus, a promising solution is to modify the simple parallel field to create a reasonable pressure difference between the adjacent parallel channels, for better water drainage, more efficient utilization of the rib area, and more uniform current distribution. An easily machined novel flow field with controllable pressure gradient across adjacent channels was manufactured and numerically studied by a 2D, two-phase flow model. The effect of channel-rib width ratio, GDL thickness, and pressure gradient on the profiles of oxygen concentration and water saturation within the electrode were investigated [113].

3.3.3

Fuel cell stack optimization

Fuel cell stack optimization has received little attention compared with the flow field design, electrode design, and operating condition optimization. In the fuel cell stack, the clamping load applied to a PEMFC stack is considered the most important effect on fuel cell performance due to the influence on electron transfer, mass, and thermal transport [114]. The modeling results showed that an optimal compression deformation exists when the contact resistance was considered [115]. Mohamed and Jenkins [116] optimized the number of cells in series based on a simplified zero-dimensional, isothermal fuel cell stack model, in which the optimization objective was to maximize the power output. Zhou et al. [117] developed a cold start model for PEMFCs aimed at optimizing the start-up methods. As a novel method, variable heating and load control (VHLC) was proposed. A notable feature of vehicular PEMFCs is that the fuel cell stack must undergo frequent start-up and shut-down (SU/SD) processes, which is considered one of the main mechanisms of fuel cell durability degradation

92

Fuel Cells for Transportation

[118]. Before the start of PEMFC stacks, the anode flow channel is filled with air. When hydrogen enters the anode, a process of replacing air with hydrogen is required, leading to the formation of a hydrogen/air interface at the anode side. The formation of the hydrogen/air interface results in extremely high electrode potential in the cathode catalyst layer, which leads to the oxygen evolution reaction and carbon corrosion reaction. It is named the reverse current decay mechanism, which is widely accepted as the basis for the degradation of PEMFC stacks during SU/SD processes. Since reactant gases remain in the fuel cell stack after the SD process, the potential between the electrode and proton-exchange membrane generated in the SU process would cause considerable degradation of fuel cell stack components, for example, carbon corrosion, leading to a significant reduction of PEMFC life. To optimize the operation of fuel cell stacks and mitigate the degradation in SU/SD processes, two strategies, gas purge and auxiliary load, are commonly applied [119,120]. The gas purge normally uses inert gases, for example, nitrogen, to purge reactant gases out of the stack, which could effectively reduce the residence time of reactant gases within the fuel cell stacks, achieving a relatively low potential between the CL and the protonexchange membrane [121]. Different gas purge strategies are sometimes combined, for example, using the anode reactant as the purge gas at the cathode, and vice versa. Oyarce et al. [122] compared four different gas purge strategies and found that the lifetime of PEMFCs was significantly improved. Although nitrogen purge is the most effective strategy, the availability of nitrogen limits the application of this strategy in practical PEMFCs. For the auxiliary load strategy, an extra load is applied on the fuel cell stack to consume the residual reactant gases in the porous electrodes. Thus, the potential difference between the CL and membrane is limited to a safe range to alleviate the degradation and improve the durability of the fuel cell stack. For example, the application of a higher dummy load could eliminate the high potential at the cathode more promptly during the SD process [123]. The results of Yang et al. [124] showed that the auxiliary load can effectively shorten the duration time of the hydrogen/air interface during the SD process, thereby eliminating the generation of reverse current. However, the implementation of an auxiliary load may cause local gas starvation, which could be avoided by the air purge. As a result, the hybrid strategy, which combines auxiliary load and air purge, is considered a more effective way to mitigate fuel cell stack degradation [125].

3.3.4

Operating condition optimization

The improved cell performance can be achieved with higher operating temperature, inlet pressure, and stoichiometric flow ratio due to lower activation and ohmic overpotentials [73,126
128]. Consequently, both an accurate fuel cell model and a complete fuel cell system model must be coupled in the

Fuel cell modeling and optimization Chapter | 3

93

optimization of operating conditions. Otherwise, the optimization algorithm would always choose the higher values of the operating parameters [129]. The early efforts in the optimization of the operating conditions can be found in 2000 [130]. The objective was to maximize the power density at a fixed current density. The design variables included operating temperature, pressures of anode and cathode, mole fractions of the gas inlet, stoichiometry, and relative humidity. Minimizing membrane hydration and maximizing temperature rise and cell voltage were the additional constraints. The optimization results indicated that the optimal operating conditions strongly depend on the current densities. Wu et al. [131] optimized the efficiency of the fuel cell system at low, medium, and high current densities. The design variables used were the operating temperature, pressure of cathode gas inlet, stoichiometry, and relative humidity. The optimization results suggested that, for a realistic system, the optimal cathode stoichiometry was between 1.25 and 2, cathode pressure between 1.5 and 3 atm, and cathode relative humidity between 10% and 15%. Xing et al. [132] optimized the cathode relative humidity for different Nafion ionomer contents inside the cathode CL by a 2D, isothermal, two-phase flow model. The optimal ionomer content of 10% was found for fully humidified inlet gas at the cathode. The optimal relative humidity was between 73% and 85% when the ionomer volume fraction was in the range of 10% to 50%. The operating conditions were also supervised and optimized through a data-driven surrogate model [133].

3.3.5 Multivariable optimization and data-driven surrogate modeling Since the optimal graded distributions of different components are simultaneously preceded in parallel, for example, the design of both a graded GDL and a graded MPL [134], multivariable optimization is an efficient method by which to optimize the cell performance by simultaneously optimizing a variety of variables. For example, Secanell et al. optimized both the Pt loading and the performance of a complete MEA [82,135]. The design variables included the Pt loading, ionomer loading, GDL porosity and hydrophilicity, and platinum-to-carbon mass ratio. It was shown that the cell performance was significantly improved by using the parameters obtained from the optimal design. The optimization results suggested that the cell performance can be improved by increasing the ionomer content and reducing the catalyst loadings. In addition, the Pt loading had to be controlled within the range of 0.1
0.5 mg cm22, as higher loadings resulted in the waste of platinum rather than an increase in current density. A two-objective function multivariable optimization of the cathode composition of the PEMFC was carried out by Xing et al. [136]. Five design variables, including the Pt loading, Pt/C ratio, ionomer volume fraction, CL thickness, and agglomerate size, were optimized through multiple surrogate models, and their sensitivities were

94

Fuel Cells for Transportation

analyzed by a Monte Carlo method-based approach. As a novel optimization strategy, maximizing the current density within a specific range of cell voltages was implemented for the prediction of the optimum values. In their later work, the interaction of graded Pt loading and GDL porosity [137], and Pt loading and operating temperature [138], along the in-plane direction was numerically studied to reduce the usage of Pt-based catalyst and improve the cell performance and current homogeneity. An optimization problem involving D objectives and K constraints can be formulated as:   minimize Y ðxÞ 5 y1 ðxÞ; y2 ðxÞ; . . . ; yD ðxÞ x

  subject to GðxÞ 5 g1 ðxÞ; g2 ðxÞ; . . . ; gK ðxÞ

ð3:7Þ

n

and

x 5 ðx1 ; x2 ; . . . ; xn ÞT A L ½li ; ui  i51

where Y is the objectives, and G is the constraints of the optimization design. n is the number of design variables, and 2N , li , ui , 1 N for all n i 5 1; 2; . . . ; n. Li51 ½li ; ui  means the range of the design variables, termed as the design space. The basic process of an optimization design is shown in Fig. 3.9. The data-driven optimization design is started with the definition of the problem, including the design variables and their ranges, the objectives, and the constraints. According to the formulated design problem, the design of experiment (DoE) is employed to generate the initial samples in the design space, and the response of each sample is obtained by a corresponding run of simulation or experiment. Then, the unknown mapping relationship between the input variables and their output responses is approximated using the surrogate model, which plays an important role in the optimization design and has a significant impact on the optimization results. Therefore, the accuracies

FIGURE 3.9 A typical process of optimization design.

Fuel cell modeling and optimization Chapter | 3

95

of the surrogate models are calculated to determine whether the model can be used for the following optimization design. If the model accuracy cannot satisfy the optimization requirements, new samples will be obtained by sequential sampling algorithms to improve the surrogate model until the model accuracy or computational cost met the termination conditions. Although time and resources could be significantly reduced through the use of complex physical models, for example, the multi-physics and multiphase flow models, the computational and time costs are still unaffordable when solving partial differential equations (PDEs) over complex geometry; for example, few days or weeks are required to solve a 3D steady-state small fuel cell stack model discretized into less than 100 million nodes of grid [139
141]. In this regard, a feasible approach is to train the machine learning and AI model with a certain number of data obtained from the complex mechanistic physical models, which typically refer to data-driven surrogate modeling, as shown in Fig. 3.10. Data-driven surrogate models have been applied to predict PEMFC performance in steady-state and dynamic conditions [133,142], screen catalyst materials [143,144], optimize electrode composition [145,146], and assist the development of control strategies to maximize the cell performance and mitigate material degradation [147,148]. There are numerous published works optimizing the PEMFCs performance through the implementation of data-driven surrogate models. For example, Lan et al. [149] optimized the stoichiometric ratio and flow channel geometry for high-temperature proton-exchange membrane (HT-PEM) fuel cells based on artificial neural networks (ANNs) to reduce the high computational cost of experiments or simulations. A few samples with representative performance information of fuel cells were used for approximating the complex mapping relationship between different designs with the stoichiometric ratio and flow channel geometry. The flow channel geometry was optimized for maximum current density and maximum real power under a fixed operating cell voltage. Wang et al. [150] integrated a 3D CFD fuel cell model with an AI-based data-driven surrogate model and an optimization framework to realize the multivariable global optimization of CL composition for

FIGURE 3.10 An illustrative flowchart of PEMFC unit/stack design based on data-driven surrogate modeling.

96

Fuel Cells for Transportation

improving the maximum power density. Support vector machine (SVM) learning model was trained by the data obtained from the physical model; then, a genetic algorithm (GA) was employed to search the optimal values of voltage, Pt loading, Pt percentage, IC ratio, volume fractions of ionomer and pores. Ding et al. [151] built a data-driven surrogate model to optimize the catalyst loading and ionomer content to improve Pt utilization. In their work, eight machine learning algorithms were compared, and ANN achieved the best prediction accuracy. It is widely accepted that a comprehensive physical model, after experimental validation, lays the foundation of a data-driven surrogate model because the data used for machine learning model training are the output of the physical model. Thus, the more processes that could be taken into account in a physical model, the higher reliability of the generated database for machine learning model training.

3.4

Summary

In this chapter, the chronological development of PEMFC modeling approaches is introduced, the models built at different times are compared, and key physicochemical and operating parameters are discussed. As an important process occurred during the operation of PEMFCs, particular attention is paid to the modeling approaches for water formation and transport through the membrane and the porous electrode. Different approaches to CL digitalization are also discussed and compared. In the past 30 years, a large number of numerical models have been developed, which were chronologically developed by adding complexities and details to existing simple models. These models were developed from single physics to multi-physics, from 1D to 3D, from single-phase flow to two-phase flow, and from isothermal to non-isothermal. The computational domain has also been significantly expanded from a single-cell unit to multi-cell fuel cell stacks consisting of a series of single cells. Numerical modeling is important to the diagnosis, design, optimization, and development of novel PEMFCs because the need for experimental resources and time is greatly reduced. Experimentally validated sophisticated numerical models could be used to optimize the PEMFC’s electrodes, flow fields, fuel cell stack, and operating conditions. The endeavor of PEMFC optimization has been proven to be an effective way to achieve better water management, more uniform reactants distributions, reinforced mass transport, and prolonged fuel cell life. The latest development of data-driven surrogate modeling attracts increasing attention owing to reduced computational time which is offered by mechanistic multi-physics and multiphase flow PEMFC models. A feasible approach is to train the machine learning and AI models with limited training data obtained from sophisticated mechanistic models and then use the data-driven surrogate model to predict the cell performance instead of the mechanistic model. In recent years, AI-based data-driven

Fuel cell modeling and optimization Chapter | 3

97

surrogate modeling has been successfully implemented in PEMFC optimization, which can be a promising approach in the future.

References [1] A. Weber, R. Borup, R. Darling, P.K. Das, T. Dursch, W. Gu, et al., J. Electrochem. Soc. 161 (2014) F1254
F1299. [2] V. Gurau, H. Liu, S. Kakac¸, AIChE J. 44 (1998) 2410
2422. [3] T. Hottinen, O. Himanen, Electrochem. Commun. 9 (2009) 1047
1052. [4] W.M. Yan, C. Soong, F. Chen, H.S. Chu, J. Power Sources 125 (2004) 27
39. [5] L. Xing, Y. Xu, P.K. Das, B. Mao, Q. Xu, H. Su, et al., Chem. Eng. Sci. 195 (2019) 127
140. [6] M. Kawase, K. Sato, R. Mitsui, H. Asonuma, M. Kageyama, K. Yamaguchi, et al., AIChE J. 63 (2017) 249
256. [7] T. Van Nguyen, M.W. Knobbe, J. Power Sources 114 (2003) 70
79. [8] M.A. Hickner, C.H. Fujimoto, C.J. Cornelius, Polymer 47 (2006) 4238
4244. [9] J. Stumper, M. L¨ohr, S. Hamada, J. Power Sources 143 (2005) 150
157. [10] H. Meng, C.-Y. Wang, J. Electrochem. Soc. 152 (2005) A1733
A1741. [11] G. Janssen, M.L.J. Overvelde, J. Power Sources 101 (2001) 117
125. [12] S. Knights, K. Colbow, J. St-Pierre, D. Wilkinson, J. Power Sources 127 (2004) 127
134. [13] C.-Y. Wang, P. Sinha, ECS Trans. 3 (2006) 387
396. [14] E. Kumbur, K. Sharp, M.M. Mench, J. Electrochem. Soc. 154 (2007) B1315
B1324. [15] D. Natarajan, T. Van Nguyen, J. Electrochem. Soc. 148 (2001) A1324
A1335. [16] P. Berg, K. Promislow, J. St. Pierre, J. Stumper, B. Wetton, J. Electrochem. Soc. 151 (2004) A341
A353. [17] U. Pasaogullari, C.Y. Wang, J. Electrochem. Soc. 151 (2004) A399
A406. [18] P.K. Das, X. Li, Z.-S. Liu, Int. J. Hydrog. Energy 35 (2010) 2403
2416. [19] T.E. Springer, T. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc. 138 (1991) 2334
2342. [20] T.E. Springer, M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 140 (1993) 3513
3526. [21] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 139 (1992) 2477
2491. [22] D.M. Bernardi, M.W. Verbrugge, AIChE J. 37 (1991) 1151
1163. [23] T.V. Nguyen, R.E. White, J. Electrochem. Soc. 140 (1993) 2178
2186. [24] T.F. Fuller, J. Newman, J. Electrochem. Soc. 140 (1993) 1218
1225. [25] J.S. Yi, T.V. Nguyen, J. Electrochem. Soc. 145 (1998) 1149
1159. [26] J.S. Yi, T. Van Nguyen, J. Electrochem. Soc. 146 (1999) 38
45. [27] F. Gloaguen, R. Durand, J. Appl. Electrochem. 27 (1997) 1029
1035. [28] Y. Bultel, P. Ozil, R. Durand, J. Appl. Electrochem. 28 (1998) 269
276. [29] Y. Bultel, P. Ozil, R. Durand, Electrochim. Acta 43 (1998) 1077
1087. [30] Y. Bultel, P. Ozil, R. Durand, J. Appl. Electrochem. 29 (1999) 1025
1033. [31] C. Marr, X. Li, J. Power Sources 77 (1999) 17
27. [32] D. Spernjak, A.K. Prasad, S.G. Advani, J. Power Sources 170 (2007) 334
344. [33] K. T¨uber, D. Po´cza, C. Hebling, J. Power Sources 124 (2003) 403
414. [34] S. Dutta, S. Shimpalee, J.W. Van Zee, J. Appl. Electrochem. 30 (2000) 135
146. [35] S. Dutta, S. Shimpalee, J.W. Van Zee, Int. J. Heat. Mass. Transf. 44 (2001) 2029
2042. [36] T. Zhou, H. Liu, Int. J. Transp. Phenom 3 (2001) 177
198. [37] T. Berning, D.M. Lu, N. Djilali, J. Power Sources 106 (2002) 284
294.

98

Fuel Cells for Transportation

[38] [39] [40] [41] [42] [43] [44]

S. Mazumder, J.V. Cole, J. Electrochem. Soc. 150 (2003) A1503. S. Mazumder, J.V. Cole, J. Electrochem. Soc. 150 (2003) A1510
A1517. Wk Lee, S. Shimpalee, J.W. Van Zee, J. Electrochem. Soc. 150 (2003) A341. S. Um, C.Y. Wang, J. Power Sources 125 (2004) 40
51. Y. Wang, C.-Y. Wang, Electrochim. Acta 50 (2005) 1307
1315. B.S. Machado, N. Chakraborty, P.K. Das, Int. J. Hydrog. Energy 42 (2017) 6310
6323. B.S. Machado, N. Chakraborty, M. Mamlouk, P.K. Das, J. Electrochem. Energy Convers. Storage 15 (1) (2018) 011004. Available from: https://doi.org/10.1115/1.4037942. Z. Liu, Z. Mao, C. Wang, W. Zhuge, Y. Zhang, J. Power Sources 160 (2006) 1111
1121. S. Shimpalee, S. Greenway, D. Spuckler, J.W. Van Zee, J. Power Sources 135 (2004) 79
87. H. Ju, H. Meng, C.-Y. Wang, Int. J. Heat. Mass. Transf. 48 (2005) 1303
1315. Y. Wang, C.-Y. Wang, J. Power Sources 153 (2006) 130
135. W. He, J.S. Yi, T. Van Nguyen, AIChE J. 46 (2000) 2053
2064. D. Natarajan, T. Van Nguyen, J. Power Sources 115 (2003) 66
80. T. Berning, N. Djilali, J. Electrochem. Soc. 150 (2003) A1589. G. Lin, T.V. Nguyen, J. Electrochem. Soc. 153 (2006) A372. Q. Ye, T.V. Nguyen, J. Electrochem. Soc. 154 (2007) B1242. P.C. Sui, S. Kumar, N. Djilali, J. Power Sources 180 (2008) 410
422. Y. Wang, S. Basu, C.-Y. Wang, J. Power Sources 179 (2008) 603
617. P. Quan, B. Zhou, A. Sobiesiak, Z. Liu, J. Power Sources 152 (2005) 131
145. B. Xie, G. Zhang, Y. Jiang, R. Wang, X. Sheng, F. Xi, et al., eTransportation 6 (2020) 100090. B. Xie, M. Ni, G. Zhang, X. Sheng, H. Tang, Y. Xu, et al., Int. J. Heat. Mass. Transf. 189 (2022) 122705. J.S. Newman, Electrochemical Systems, second ed., Prentice Hall, 1991. P.N. Pintauro, D.N. Bennion, Ind. Eng. Chem. Fundam. 23 (1984) 230
234. A.Z. Weber, J. Newman, J. Electrochem. Soc. 150 (2003) A1008
A1015. J.H. Nam, M. Kaviany, Int. J. Heat. Mass. Transf. 46 (2003) 4595
4611. W. Sun, B.A. Peppley, K. Karan, Electrochim. Acta 50 (2005) 3359
3374. P.K. Das, X. Li, Z.-S. Liu, J. Power Sources 179 (2008) 186
199. Y. Wang, C.-Y. Wang, Electrochim. Acta 51 (2006) 3924
3933. S. Basu, J. Li, C.-Y. Wang, J. Power Sources 187 (2009) 431
443. J.W. Park, K. Jiao, X. Li, Appl. Energy 87 (2010) 2180
2186. Y.B. Salah, Y. Tabe, T. Chikahisa, Energy Procedia 28 (2012) 125
133. L.-P. Wang, B. Afsharpoya, Math. Comp. Simul. 72 (2006) 242
248. J. Park, M. Matsubara, X. Li, J. Power Sources 173 (2007) 404
414. Y. Gao, X.X. Zhang, P. Rama, Y. Liu, R. Chen, H. Ostadi, et al., Fuel Cell 12 (2012) 365
381. K.T. Jeng, S.F. Lee, G.F. Tsai, C.H. Wang, J. Power Sources 138 (2004) 41
50. P.K. Das, X. Li, Z.-S. Liu, J. Electroanal. Chem. 604 (2007) 72
90. L. Xing, M. Mamlouk, K. Scott, Energy 61 (2013) 196
210. L. Xing, M. Mamlouk, R. Kumar, K. Scott, Int. J. Hydrog. Energy 39 (2014) 9087
9104. L. Xing, Q. Cai, C. Xu, C. Liu, K. Scott, Y. Yan, Energy 106 (2016) 631
645. L. Xing, Q. Cai, X. Liu, C. Liu, K. Scott, Y. Yan, Chem. Eng. Sci. 146 (2016) 88
103. L. Xing, Chem. Eng. Sci. 179 (2018) 198
213. P.K. Das, X. Li, Z. Xie, Z.-S. Liu, Int. J. Energy Res. 35 (2011) 1325
1339.

[45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79]

Fuel cell modeling and optimization Chapter | 3

99

[80] D. Song, Q. Wang, Z. Liu, T. Navessin, M. Eikerling, S. Holdcroft, J. Power Sources 126 (2004) 104
111. [81] D. Song, Q. Wang, Z. Liu, T. Navessin, S. Holdcroft, Electrochim. Acta 50 (2004) 731
737. [82] M. Secanell, R. Songprakorp, A. Suleman, N. Djilali, Energy Environ. Sci. 1 (2008) 378
388. [83] M. Secanell, K. Karan, A. Suleman, N. Djilali, Electrochim. Acta 52 (2007) 6318
6337. [84] L. Xing, W. Shi, H. Su, Q. Xu, P.K. Das, B. Mao, et al., Energy 177 (2019) 445
464. [85] Y.-X. Huang, C.-H. Cheng, X.-D. Wang, J.-Y. Jang, Energy 35 (2010) 4786
4794. [86] F.-B. Weng, C.-Y. Hsu, M.-C. Su, Int. J. Hydrog. Energy 36 (2011) 13708
13714. [87] M. Srinivasarao, D. Bhattacharyya, R. Rengaswamy, J. Power Sources 206 (2012) 197
203. [88] L. Xing, W. Shi, P.K. Das, K. Scott, AIChE J. 63 (2017) 4895
4910. [89] Y. Sato, K. Fujii, N. Mitani, A. Matsuura, T. Kakigi, F. Muto, et al., Nucl. Instrum. Methods Phys. Res. Sect. B: Beam Interact. Mater. At. 265 (2007) 213
216. [90] H. Fujita, F. Shiraki, Y. Oshima, T. Tatsumi, T. Yoshikawa, T. Sasaki, et al., Radiat. Phys. Chem. 80 (2011) 201
206. [91] H. Fujita, F. Shiraki, T. Yoshikawa, A. Oshima, M. Washio, J. Photopolym. Sci. Technol. 23 (2010) 387
392. [92] M. Grujicic, K.M. Chittajallu, Appl. Surf. Sci. 227 (2004) 56
72. [93] H.-H. Lin, C.-H. Cheng, C.-Y. Soong, F. Chen, W.-M. Yan, J. Power Sources 162 (2006) 246
254. [94] H. Chen, H. Guo, F. Ye, C.F. Ma, Q. Liao, X. Zhu, Int. J. Energy Res. 43 (2019) 2910
2929. [95] H. Chen, H. Guo, F. Ye, C.F. Ma, J. Power Sources 472 (2020) 228456. ˇ Penga, V. Hacker, Energies 14 (2021) 3675. [96] L. Mihanovi´c, Z. [97] Z. Bao, Z. Niu, K. Jiao, Int. J. Hydrog. Energy 44 (2019) 6229
6244. [98] S.A. Atyabi, E. Afshari, J. Clean. Prod. 214 (2019) 738
748. [99] Y. Wang, L. Yue, S. Wang, J. Power Sources 344 (2017) 32
38. [100] L. He, M. Hou, Y. Gao, X. Sun, W. Song, L. Zheng, et al., Energy Convers. Manag. 223 (2020) 113292. [101] C. Yang, Z. Wan, X. Chen, X. Kong, J. Zhang, T. Huang, et al., Energy Convers. Manag. 228 (2021) 113651. [102] Y. Wang, H. Xu, X. Wang, Y. Gao, X. Su, Y. Qin, et al., Energy Convers. Manag. 252 (2022) 115069. [103] X.-D. Wang, Y.-Y. Duan, W.-M. Yan, X.-F. Peng, Electrochim. Acta 53 (2008) 5334
5343. [104] X.-D. Wang, Y.-Y. Duan, W.-M. Yan, J. Power Sources 173 (2007) 210
221. [105] W.-M. Yan, H.-Y. Li, P.-C. Chiu, X.-D. Wang, J. Power Sources 178 (2008) 174
180. [106] M.K. Vijayakrishnan, K. Palaniswamy, J. Ramasamy, T. Kumaresan, K. Manoharan, T. K. Raj Rajagopal, et al., Int. J. Hydrog. Energy 45 (2020) 7848
7862. ˇ Penga, et al., AIChE J. 68 (2022) e17461. [107] Y. Wang, X. Wang, G. Chen, C. Chen, X. Wang, Z. [108] T. Yoshida, K. Kojima, Electrochem. Soc. Interface 24 (2015) 45
49. [109] J. Kim, G. Luo, C.-Y. Wang, J. Power Sources 365 (2017) 419
429. [110] M. Matsunaga, T. Fukushima, K. Ojima, World Electr. Veh. J. 3 (2009). [111] C.E. Damian-Ascencio, A. Saldan˜a-Robles, A. Hernandez-Guerrero, S. Cano-Andrade, Energy 133 (2017) 306
316.

100

Fuel Cells for Transportation

[112] P. Trogadas, J.I.S. Cho, T.P. Neville, J. Marquis, B. Wu, D.J.L. Brett, et al., Energy Environ. Sci. 11 (2018) 136
143. ˇ Penga, Q. Xu, H. Su, W. Shi, et al., AIChE J. 66 (2020) e16957. [113] L. Xing, Y. Xu, Z. [114] W. Zhang, C.W. Wu, J. Fuel Cell Sci. Technol. 11 (2013) 020000. [115] P. Zhou, C.W. Wu, J. Power Sources 170 (2007) 93
100. [116] I. Mohamed, N. Jenkins, J. Power Sources 131 (2004) 142
146. [117] Y. Zhou, Y. Luo, S. Yu, K. Jiao, J. Power Sources 247 (2014) 738
748. [118] K. Ettihir, L. Boulon, K. Agbossou, Appl. Energy 163 (2016) 142
153. [119] R. Lin, X. Cui, J. Shan, L. Te´cher, F. Xiong, Q. Zhang, Int. J. Hydrog. Energy 40 (2015) 14952
14962. [120] Y.-T. Mu, P. He, J. Ding, L. Chen, W.-Q. Tao, Energy Procedia 105 (2017) 1967
1973. [121] Q. Shen, M. Hou, D. Liang, Z. Zhou, X. Li, Z. Shao, et al., J. Power Sources 189 (2009) 1114
1119. [122] A. Oyarce, E. Zakrisson, M. Ivity, C. Lagergren, A.B. Ofstad, A. Bode´n, et al., J. Power Sources 254 (2014) 232
240. [123] Y. Yu, X.-Z. Yuan, H. Li, E. Gu, H. Wang, G. Wang, et al., Int. J. Hydrog. Energy 37 (2012) 15288
15300. [124] Y. Yang, W. Li, R. Lin, S. Xia, Z. Jiang, J. Power Sources 404 (2018) 126
134. [125] R. Lin, D. Liu, S. Xia, T. Ma, B. Dutruel, Int. J. Hydrog. Energy 45 (2020) 1030
1044. [126] A. Kazim, P. Forges, H.T. Liu, Int. J. Energy Res. 27 (2003) 401
414. [127] S. Kim, S. Shimpalee, J.W. Van Zee, J. Power Sources 135 (2004) 110
121. [128] M. Amirinejad, S. Rowshanzamir, M.H. Eikani, J. Power Sources 161 (2006) 872
875. [129] J. Wishart, Z. Dong, M. Secanell, J. Power Sources 161 (2006) 1041
1055. [130] A. Mawardi, F. Yang, R. Pitchumani, J. Fuel Cell Sci. Technol. 2 (2005). [131] J. Wu, Q. Liu, H. Fang, J. Power Sources 156 (2006) 388
399. [132] L. Xing, P.K. Das, X. Song, M. Mamlouk, K. Scott, Appl. Energy 138 (2015) 242
257. [133] B. Wang, G. Zhang, H. Wang, J. Xuan, K. Jiao, Energy AI 1 (2020) 100004. [134] A.M. Kannan, L. Cindrella, L. Munukutla, Electrochim. Acta 53 (2008) 2416
2422. [135] M. Secanell, R. Songprakorp, N. Djilali, A. Suleman, Struct. Multidiscip. Optim. 40 (2009) 563. [136] L. Xing, X. Song, K. Scott, V. Pickert, W. Cao, Int. J. Hydrog. Energy 38 (2013) 14295
14313. [137] L. Xing, Y. Wang, P.K. Das, K. Scott, W. Shi, Chem. Eng. Sci. 192 (2018) 699
713. ˇ Penga, Q. Xu, H. Su, F. Barbir, et al., Chem. Eng. J. 406 (2021) [138] L. Xing, Y. Xu, Z. 126889. [139] G.-H. Song, H. Meng, Acta Mech. Sin. 29 (2013) 318
334. [140] A. Demuren, R. Edwards, Modeling Proton Exchange Membrane Fuel Cells—A Review (2020) 513
547. [141] G. Zhang, H. Yuan, Y. Wang, K. Jiao, Appl. Energy 255 (2019) 113865. [142] P. Schneider, C. Sadeler, A.-C. Scherzer, N. Zamel, D. Gerteisen, J. Electrochem. Soc. 166 (2019) F322
F333. [143] G. Gu, J. Noh, I. Kim, Y. Jung, J. Mater. Chem. A 7 (2019) 17096
17117. [144] G. Fanourgakis, K. Gkagkas, E. Tylianakis, G. Froudakis, J. Am. Chem. Soc. 142 (2020) 3814
3822. [145] N. Khajeh-Hosseini-Dalasm, S. Ahadian, K. Fushinobu, K. Okazaki, Y. Kawazoe, J. Power Sources 196 (2011) 3750
3756. [146] R. Ding, Y. Ding, H. Zhang, R. Wang, Z. Xu, Y. Liu, et al., J. Mater. Chem. A 9 (2021) 6841
6850.

Fuel cell modeling and optimization Chapter | 3

101

[147] M. Moein-Jahromi, M.J. Kermani, Energy Convers. Manag. 231 (2021) 113837. [148] R. Ma, T. Yang, E. Breaz, Z. Li, P. Briois, F. Gao, Appl. Energy 231 (2018) 102
115. [149] H. Lan, L. Yang, F. Zheng, C. Zong, S. Wu, X. Song, Int. J. Hydrog. Energy 45 (2020) 12501
12513. [150] B. Wang, B. Xie, J. Xuan, K. Jiao, Energy Convers. Manag. 205 (2020) 112460. [151] R. Ding, W. Yin, G. Cheng, Y. Chen, J. Wang, R. Wang, et al., Energy AI 5 (2021) 100098.

This page intentionally left blank

Chapter 4

Lattice Boltzmann modeling and artificial intelligence Xing Li1, , Yuze Hou2, , Nada Zamel2 and Kui Jiao1 1

State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China, 2Fraunhofer Institute for Solar Energy Systems ISE, Freiburg, Germany

4.1 Overview of lattice Boltzmann method and artificial intelligence High-power density is the trend of next-generation proton-exchange membrane (PEM) fuel cells, so it is imperative to optimize the structure of porous electrodes as the site where the electrochemical reaction occurs. In order to achieve this goal, both the reactant transfer resistance and potential loss in electrochemical reactions should be minimized, which relies on a more comprehensive understanding of the transport mechanism within porous electrodes. Besides, to ensure that a fuel cell has a long lifetime, high performance, and efficient energy utilization, it is important for the PEM to maintain a consistently high level of hydration, as this is necessary for achieving high proton conductivity. However, a notable rise in water generation rate can be observed as the fuel cell operates in a high current density. Excessive liquid water will block the pores in porous electrodes and hinder the gas reactants from reaching the three-phase reaction interface, which may lead to severe water flooding in the porous electrode and thus reduce the performance of the fuel cell. Therefore, to achieve high performance and efficient water removal, it is crucial to optimize the structure of the porous electrode and carefully select appropriate electrode materials. Current research methods for studying transport phenomena in porous electrodes mainly include experimental tests and numerical simulations. Experimental methods can characterize the output performance of fuel cells, internal transport resistance and structural information of the porous electrode. Although the technical tools for experimental characterization are rapidly developing, experimental techniques are still normally time-consuming 

Equal contribution.

Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00005-8 © 2023 Elsevier Ltd. All rights reserved.

103

104

Fuel Cells for Transportation

and expensive. In addition, there are still some difficulties in characterizing the transport phenomena and the distribution of the physical quantities inside the porous electrodes. Especially in the in-situ testing of fuel cells, a large number of parameters are difficult to be accurately quantified, and the spatial and temporal resolution of experimental tests needs to be further improved. In this regard, numerical models can help to obtain a deeper insight into the parametric study and also to save experimental time and cost. Finite volume method (FVM) [1,2], pore network (PN) method [3,4], and lattice Boltzmann method (LBM) [5] are popular numerical methods and have been widely used to study the complex transport phenomena in the porous electrodes of PEM fuel cells. Since the scale and structure of each component in the porous electrode vary greatly, the appropriate numerical simulation method should be selected accordingly. The traditional computational fluid dynamics (CFD) method (such as FVM) is mainly used in macroscopic computational domains, such as single fuel cells. However, due to the difficulties in discretizing and capturing the gas
liquid boundary in porous microstructures, FVM methods usually simplify the electrode components into uniformly distributed computational domains, where the material properties are expressed as average properties. Since macroscopic fluid dynamics behavior is generated by the interaction between various fluid particles, a more microscopic perspective is required to investigate the complex flow phenomena in porous electrodes [6]. Cetinbas et al. [3] used computed tomography to reconstruct microporous layer (MPL) with realistic 3D microstructures and investigated the effect of microscopic pores on water transport in MPL using PN models. Since there is no comprehensive characterization of the microstructure of porous media, most PN methods build pore networks directly by digital images of the microstructure extracted by X-ray tomography. However, the PN reconstructed by this method simplifies the microstructure of porous media, not all morphological details are considered, and nanoscale pores cannot be reconstructed due to the limitation of resolution. In addition, the PN model is not applicable to single-phase flows and cannot consider electrochemical reactions at the Pt/ionomer interface, which limits its application in simulating reactive transport. Therefore, it is mainly used for the fast prediction of water transport in the gas diffusion layer (GDL) or the MPL. The LBM is a kinetic theory approach based on the Boltzmann equation, which retains the microscopic dynamics and also reproduces the fluid dynamics behavior at the macroscopic scale, and can solve mesoscopic-scale phenomena at a lower computational cost [7]. Tracing the motion of individual molecules at the microscopic scale is commonly used to describe fluids, while the macroscopic scale uses more macroscopic quantities such as density, fluid velocity, and temperature to describe fluids. LBM is a mesoscopic kinetic theory that lies between the macroscopic and microscopic scales, focusing on tracing a representative population of fluid molecules and

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

105

describing the distribution of fluid particles in space and time. LBM serves as a bridge between the macroscopic and microscopic scales, not only for solving Navier
Stokes conservation equations and describing macroscopic laws simply and efficiently, but also for understanding more fundamental microscopic dynamics [8]. In addition, its good kinetic properties can easily deal with complex boundary conditions in interfacial reaction dynamics and microstructures. For example, when dealing with a typical multiphase flow problem (dealing with gas
liquid boundaries), LBM can automatically track the gas
liquid boundary by introducing interaction forces, which reduces the computational expense compared to VOF in traditional fluid dynamics. Another significant advantage of the LBM is that it does not require a gridding step, which is very complicated in traditional CFD methods for gridding porous media. While traditional CFD methods require solving complex Poisson equation to obtain the pressure field, the LBM only requires solving a simple state equation that describes the relationship between pressure and density. In addition, the explicit format and local update rules of LBM are ideal for parallel computation based on a central processing unit or graphics processing unit, which makes large-scale simulations more efficient [9]. Based on the above advantages, as shown in Fig. 4.1 [10], most researchers have used LBM for pore-scale research on the porous electrodes of PEMFC, and most of the research has focused on evaluating the transport properties of the electrodes (e.g., permeability, conductivity, diffusion coefficient), investigating the gas
liquid two-phase flow mechanism, and optimizing the electrode structure design based on the research results. The LBM is a modeling approach based on physical mechanisms, which is based on the fundamental physicochemical processes of PEMFC and solves the complex coupled physicochemical processes in PEMFC showing a logarithmic relationship between accuracy and model complexity. In order to

FIGURE 4.1 Implementation processes and roles of pore-scale modeling. Reproduced from L. Chen et al., Pore-scale modeling of complex transport phenomena in porous media. Prog. Energy Combust. Sci. 88 (2022) 100968.

106

Fuel Cells for Transportation

obtain higher accuracy results, the model complexity is usually very high, which directly leads to the fact that it requires a lot of time and computational resources, which is contrary to the goal of reducing costs. In contrast, artificial intelligence (AI) approaches based on data-driven modeling focus on finding input
output relationships from experimental datasets obtained during fuel cell operation, as shown in Fig. 4.2 [11], which do not require indepth knowledge of these physicochemical processes and have high computational efficiency and more accurate results. Machine learning algorithms are often used to build data-driven models. After proper training using machine learning algorithms, data-driven models have high computational efficiency, which makes machine learning extremely important in practical applications of fuel cells, such as large multiunit systems, dynamic changes, and longtime operation. Complex physical mechanism models are often computationally and time-consumingly expensive, while simplified physical mechanism models lack high prediction accuracy. Even for a small fuel cell stack consisting of 5
10 single cells, a 3D simulation model based on the physical mechanism usually requires 100 million to 1 billion grid points, and it takes several days to weeks to predict a steady-state operating condition [12]. Data-driven models represented by machine learning have high prediction accuracy and high computational efficiency, which greatly enrich the applications of physical
mechanical models. Nowadays, researchers widely use simulation data of physical mechanism models to train machine learning models, which are often called agent models. Compared with the complex physical mechanism model, the agent model makes up for the disadvantage that its prediction accuracy is not as good as CFD methods by the advantage of its high computational efficiency [11].

FIGURE 4.2 Artificial neural network (ANN) structure between inputs. Reproduced from Y. Wang et al., Fundamentals, materials, and machine learning of polymer electrolyte membrane fuel cell technology. Energy AI 1 (2020) 100014.

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

107

It is known that the lattice Boltzmann (LB) model based on physical mechanisms can be used to study the reactant transport, two-phase flow, and electrochemical reaction processes in the porous electrodes of PEMFC. Compared with the macroscopic CFD model, the LBM can only solve a very small computational domain because it needs to consider the microstructure within the porous electrode, and the computational resources and time it consumes increase as the resolution of the model increases and the number of components to be solved increases. In contrast, the AI-based data-driven model does not depend on the physicochemical processes in porous electrodes, and after training with a large amount of data, it not only has a very high computational efficiency, but also its computational accuracy is comparable to that of physical mechanism models. Therefore, it is promising to combine LBM and AI to explore the physicochemical processes in porous electrodes in a more efficient and cost-saving way.

4.2

Application of lattice Boltzmann method in fuel cells

As the most promising energy conversion devices, physicochemical phenomena in fuel cells generally include fluid flow, heat/mass transport, and electrochemical reactions, which need to be explored at macroscopic, mesoscopic, and nanoscopic scales. However, the performance and reliability of fuel cells depend heavily on our understanding of these complex multiscale transport phenomena, which highlights the need for an accurate prediction method. The LBM is based on the description of the particle density distribution derived from Boltzmann kinetic theory and can realistically predict the fluid dynamics behavior at continuous scales, which enables to obtain the physical laws at macroscopic scales at a low computational cost, while also taking into account mesoscopic physical phenomena.

4.2.1

Current status of pore-scale research in gas diffusion layer

Hydrophilicity and hydrophobicity are important influencing parameters in studying the pore-scale model of two-phase flow in the GDL. Hao and Cheng [13] simulated liquid water transport in the GDL using the LB model and investigated the effect of hydrophobicity of carbon fibers on the dynamic behavior of liquid water. They proposed that incorporating hydrophilic transport channels in the GDL is a very effective water management strategy, and this structure can improve the liquid water distribution and thus reduce the transport resistance of reactants in the GDL. Their study also demonstrates that LBM is a powerful numerical tool for studying porous media at the pore scale. Chen et al. [14] investigated the effect of hydrophobicity of the GDL on the dynamic behavior of liquid water, and their results showed that hydrophilic gas channels lead to the accumulation of liquid water in the layer.

108

Fuel Cells for Transportation

The location of hydrophobic and hydrophilic fiber regions in the GDL had a more significant impact on the liquid water distribution compared to the polytetrafluoroethylene (PTFE) content. In addition, the influence of hydrophobicity of carbon fibers on liquid water transport also depends on the specific conditions of liquid water distribution. As liquid water is distributed in individual droplet form, the effective diffusion coefficient of reactants in hydrophilic GDLs is higher. In a later study, they [15] investigated the twophase flow process of GDL under a staggered flow field by considering the ridge width and contact angle in the flow channel. Their results indicated that a narrower ridge could accelerate the water discharge and reduce the remaining liquid water saturation. The hydrophobic carbon fiber only facilitates the shear-dominated liquid water transport, but the more hydrophobic GDL also extends the time for drainage. Garcı´a-Salaberr et al. [16] investigated the effect of partial saturation of liquid water in the GDL on the effective diffusion coefficient using the LB model, including local saturation as well as cross-plane saturation. The results indicated that the addition of PTFE decreases the gas’s effective diffusivity due to the reduction of the pore volume and the increase of the curvature of the gas transport path. Moreover, the local effective diffusivity is secondarily correlated with the local water content saturation. Based on the previous work, Deng et al. [17] investigated the effect of air or vacuum drying on the distribution of PTFE and wettability in GDLs reconstructed based on a stochastic algorithm. The model takes into account the different wettability of carbon fibers and PTFE by assigning them different contact angles. The results show that the drying method of PTFE has a significant effect on the water transport in GDLs and that higher pressure is required to enable liquid water to break through in GDLs under air-drying. Furthermore, researchers have investigated the effects of permeability and microstructure of GDL on two-phase flow in flow channels. Rama et al. [18] investigated the anisotropic permeability of a carbon cloth GDL using a single-phase LB model. They calculated the through-plane permeability and in-plane permeability separately and showed that the through-plane permeability was about four times higher than the in-plane permeability. Hou et al. [19] developed a 3D multicomponent multiphase LB model and used it to study the droplet motion in the flow channel of a PEM fuel cell. They considered the real GDL microstructure and fluid properties in the flow channel, and their results showed that the microstructure of the GDL at the bottom of the flow channel has a strong influence on the direction of liquid water movement, and that hydrophilic sidewalls and top walls can effectively remove liquid water from the GDL surface, reduce the pressure drop, and prevent uneven distribution of reactants, as shown in Fig. 4.3. The above-mentioned works explored the two-phase flow behavior in GDLs using pore-scale simulation models, and these studies facilitated the understanding of the transport mechanism of porous electrodes. Based on

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

109

FIGURE 4.3 SEM images of carbon paper GDL (A) GDL surface; (B) cross section of carbon fiber; effect of GDL on droplet in static and dynamic states (C) static contact angle test; (D) dynamic process in 3D view. Reproduced from Y. Hou et al., 3D lattice Boltzmann modeling of droplet motion in PEM fuel cell channel with realistic GDL microstructure and fluid properties. Int. J. Hydrog. Energy 45 (22) (2020) 12476
12488.

them, further work on the optimal structural design of the GDL to achieve better water management capability and higher cell performance has been performed. Shangguan et al. [20] studied the transport of liquid water in GDLs with different porosity distributions and showed that more liquid water remains in GDLs with inverted V-shaped porosity distributions compared to V-shaped porosity distributions (larger water content saturation), and the flow rate of liquid water is lower, which will lead to more severe flooding. In the same year, Wang et al. [21] applied LBM to study the permeability and kinetic behavior of liquid water inside GDLs and found that GDLs with a gradient distribution of porosity would result in higher permeability compared to constant porosity. In summary, researchers have put great efforts to study the single-phase mass transfer and two-phase flow behavior in the GDL. However, limited by the spatial-scale and time-scale issues of LBM, more computational resources are required to obtain more realistic simulation results. In addition, most of the current studies only consider two-phase flow or single-phase

110

Fuel Cells for Transportation

mass transfer processes in the GDL, and a few studies have coupled both and considered the electrochemical reactions in the catalyst layer (CL). Therefore, modeling with a focus on coupling not only single-phase or twophase flow but rather also addressing the reactive transport phenomenon of coupled two-phase flow and mass transfer in the GDL is of great importance. This will be discussed in more detail in the next section.

4.2.2 layer

Current status of pore-scale research in the microporous

The MPL is located between the GDL and CL and influences the two-phase flow and mass transfer process. Compared with the GDL with large pores and regularly shaped carbon fibers, the carbon particles and pores in the MPL are usually nanoscale, and the pore paths are very tortuous. Zamel et al. [22] used the GeoDict package solver to investigate the effects of thickness, porosity, and depth of penetration through the GDL on the thermal conductivity and diffusion coefficient of the MPL. Their results showed that the diffusion coefficient increased with decreasing thickness or increasing porosity. However, the thermal conductivity decreases with increasing porosity and is not sensitive to thickness. In addition, the penetration of MPL into GDL increases the thermal conductivity but decreases the diffusion coefficient. The large porosity and cracks in the MPL are important for the twophase flow behavior in the MPL and GDL. Kim et al. [23] investigated the transport of liquid water in the MPL and GDL using an LB model. They showed that a thicker MPL and a more hydrophobic solid surface can reduce the liquid water content in the porous transport layer (MPL 1 GDL, PTL). As the MPL becomes thicker or the solid surface becomes more hydrophobic, the liquid water distribution can reach a steady state faster, and the number of liquid water breakthrough points can be effectively reduced, thus improving water management. Deng et al. [17] investigated the role of macroscopic cracks in MPL on liquid water transport and water redistribution at the GDL/MPL interface using LBM, which considered both GDL and MPL. Their results showed that liquid water would cross the cracks of MPL to reach GDL, which is beneficial to mitigate water flooding and enhance the transport of gas reactants to the active reaction sites. However, they ignored the real microstructure of the MPL and simplified the structure of the MPL as a flat plate with cracks. In addition, researchers have also worked on optimizing the MPL structure, and Han and Meng [24] simulated the effect of perforation on the liquid water transport process in MPL and GDL using a pseudo-potential two-phase LB model. The results showed that perforation can be used as a transport pathway for liquid water, which can accelerate the discharge of liquid water from the GDL and MPL. However, due to the

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

111

excessive scale difference between GDL and MPL, they only simulated in two dimensions. However, these studies only studied the two-phase flow or mass transfer in MPL separately, and few studies coupled the two and considered their direct effects on the cell performance. Zhang et al. [25] simulated the coupled two-phase flow and mass transfer in GDL and MPL using two LB models and added a source term at the bottom of the computational domain to consider the electrochemical reactions in CL. The results showed that the presence of MPL reduced the water content in the GDL, which facilitated the transport of the gas reactant. However, they simplified the MPL to a 2D structure and could not give a true structural description. Therefore, the real transport mechanism of liquid water in MPL is not clear. In order to achieve a balance between performance and water transport, the reactant transport, electrochemical reactions and water removal processes in GDL and MPL should be coupled simultaneously, and the effect of the structure of PTL on the performance of porous electrodes needs to be investigated. As shown in Fig. 4.4, Ye et al. [26] used a stochastic algorithm to reconstruct the 3D microstructure of GDL and MPL, and built a coupled model of two-phase flow, mass transfer, and electrochemical reaction based on LBM. The model gave a more intuitive mechanism of the interaction between two-phase flow and mass transfer by adding source terms at the reaction interface as well as special bounce boundaries. Their results show that the systematic perforation design substantially reduces the gas transfer resistance by separating the transport of liquid water and oxygen. More uniform current density distribution and optimal performance were achieved under water flooding, with a 46% increase in average current density.

FIGURE 4.4 Evolution of liquid water transport and oxygen diffusion in GDLs with different porosity: (A) water transport inside GDLs; (B) oxygen diffusion inside GDLs. Reproduced from S. Ye et al., Pore-scale investigation of coupled two-phase and reactive transport in the cathode electrode of proton exchange membrane fuel cells. Trans. Tianjin Univ. (2022).

112

4.2.3

Fuel Cells for Transportation

Current status of pore-scale research in catalyst layer

As shown in Fig. 4.5 [27], the CL is composed of carbon-loaded platinum that acts as an electron transport channel, ionomer that acts as an ion transport channel, and pores that act as gas transport channels. The thickness of CL is typically less than 10 μm, and the individual platinum particles are between 2 and 5 nm in diameter. Due to its complex nanoscale microstructure, innovative mathematical models and vigorous development of computer technology are required to precisely explore the complex process of reactant transport and electrochemical reactions in CL. Chen et al. [28] used LBM to simulate the coupled physical
electrochemical processes in the CL generated by stochastic reconstructions. The Pt particles, which were neglected in previous studies, were considered in the reconstruction process, while the electrochemical reactions occurred at the Pt/ionomer interface, which more realistically described the physical
electrochemical phenomena in the CL. The presence of Pt required a higher structural resolution, which in their work was 5 nm, and this was sufficient to describe the Pt particle distribution in the CL. The results indicated that the nonuniform distribution of the ionomer increased the tortuosity of the CL, thus decreasing the effective diffusion coefficient. Their next work [29] focused on the local gas transport resistance around the carbon support based on stochastic reconstruction. The microstructure of the carbon support in the model included carbon spheres, ionomers, Pt particles, and main pores with a resolution of 0.5 nm. They investigated the local transport resistance for different Pt and ionomer contents and found that the main resistance to the mass transfer process is due to the low oxygen diffusion coefficient in the ionomer. Then, they used LBM to simulate the gas

FIGURE 4.5 (A) Schematic of catalyst layer reconstruction; (B) reconstructed catalyst layers under different Pt loadings. Reproduced from X. Li et al., Interlink among catalyst loading, transport and performance of proton exchange membrane fuel cells: a pore-scale study. Nanoscale Horiz. 7 (3) (2022) 255
266.

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

113

transfer resistance and gas concentration drop at the ionomer/pore interface, which greatly improved the model accuracy and realism [30]. Hou et al. [31] developed a pore-scale LBM to study the reactive transport process of CL at the cathode in PEMFC. They analyzed the structural properties of the CL, including pore size distribution, phase connectivity, and effective catalytic surface area (ECSA). Then, the effects of two key parameters, Pt/C ratio and I/C ratio, on the performance of the CL were investigated in terms of oxygen concentration distribution, ECSA, and reaction rate. Their results showed that for a constant Pt loading, a higher Pt/C ratio produced a thinner CL, which significantly enhanced oxygen transport and improved performance. For the same Pt/C ratio, although a higher I/C ratio resulted in more mass transfer losses, it increased ECSA and yielded better performance. In addition, they proposed an idealized CL structure design that achieved higher ECSA with low transport losses and improved the performance of CL by 50%. All the above studies coupled the gas transfer process and electrochemical reaction process of CL, but none of them considered the proton and electron conduction processes, which are crucial in the electrochemical reaction process. In order to meet the high-power density requirements of fuel cells, a comprehensive description of the transport processes of the reactants involved in the electrochemical reactions in CL is essential. Xing et al. [27] proposed a pore-scale model based on LB and used multi-relaxation time operators to ensure the model stability and solve the problem of mismatched diffusion coefficients among the reactants. The results showed that the distribution of the electron potential was not sensitive to the variation of the platinum loading compared to the distribution of the proton potential. Their work pointed out that ohmic loss was the main factor limiting the performance at high current density conditions and, therefore, suggested the use of lower platinum loadings to reduce ohmic loss. Due to the complex structure and nanoscale microscopic pores in CL, most studies on CL have focused on its mass transfer and electrochemical reaction processes, and few studies have considered the effect of two-phase flow in CL. In order to fully reveal the influence of the dynamic behavior of liquid water in the CL on the performance of PEMFC, it is necessary to combine the two-phase flow, mass transfer, and electrochemical reaction processes and to investigate the real 3D microstructure of CL and the influence of each component on the two-phase flow by LBM. Recently, as shown in Fig. 4.6, Chen et al. [32] did part of this work by developing a highresolution pore-scale LB model to study two-phase flow coupled electrochemical reaction transport processes, including oxygen diffusion, electrochemical reactions, and gas
liquid two-phase flow, in the CL microstructure. However, their computational domain was only the CL with a thickness of 0.2 μm near the MPL side and neglected that liquid water covering the Pt surface would also generate new reaction sites.

114

Fuel Cells for Transportation

FIGURE 4.6 Time evolutions of the two-phase interface at the GDL/CL interface with realistic structures of CLs with a contact angle of 120 degrees. Reproduced from L. Chen, Q. Kang, W. Tao, Pore-scale numerical study of multiphase reactive transport processes in cathode catalyst layers of proton exchange membrane fuel cells. Int. J. Hydrog. Energy 46 (24) (2021) 13283
13297.

In addition, the conventional CL structure cannot effectively solve the water flooding problem under high current density conditions, and a deeper understanding of the reactive transport process and the two-phase transport mechanism in the CL is needed to propose a more effective catalytic layer structure. Although LBM can be used for nanoscale investigation of mass transfer and two-phase flow processes in porous electrodes, LBM consumes a large number of computational resources. In order to optimize the structural parameters of porous electrodes and to select suitable electrode materials, faster and more efficient methods, such as AI methods, need to be considered.

4.3

Artificial intelligence method

Machine learning, one of the most important areas of AI methods, has been widely used in the field of fuel cells. Machine learning refers to machines summarizing laws and generalizing certain specific knowledge from a specific large amount of data, and then applying this knowledge to real-life scenarios to solve practical problems. The basic idea of machine learning is to continuously identify features and continuously model through a training set

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

115

and finally form an effective model. In the case of fuel cells, one of the most critical data is the polarization curve, which is determined by many factors, including the size of the fuel cell, material properties, operating conditions, and electrochemical
physical processes. The functions currently implemented using AI in the fuel cell field include parameter optimization, model predictive control, predictive and health management, and fault diagnosis.

4.3.1

Parameter optimization

In order to achieve higher output energy density, which is the ratio of output power to mass, and higher energy conversion efficiency, researchers need to optimize the overall or local design of the fuel cell. PEMFC consists of bipolar plate (BP), GDL, MPL, CL, and PEM. Since these components have different physical properties, researchers can optimize them in different ways and from different perspectives. For example, some researchers have focused on using more superior materials to improve the performance of PEMFCs [33,34]. Some researchers have also worked on new flow channel geometries to achieve better water and thermal management [35,36]. Others have investigated the effect of clamping forces on water transport and performance in fuel cells as well as the study of shaped materials to enhance water transport for better performance [2,37]. It can be seen that the key issue in the current research on fuel cells is the parameter optimization of fuel cells, which is the most direct path to improving the performance of fuel cells. In general, it is a process of finding the optimal structure, materials, and combinations to maximize fuel cell performance. However, it is also a process of continuous trying and finding. PEMFC is a complex system involving many different materials, structures, and coupled physical
chemical processes, and using mechanistic models to solve the problem or physical
chemical experiments to test and verify, it would be timeconsuming, expensive, and labor-intensive. Today, the advent and widespread use of AI will directly change all this. Wang et al. [38] first developed a 3D CFD fuel cell model with a CL agglomerate model, then constructed a database of the performance of PEMFC with different CL structural parameters, and then used the database to train a data-driven agent model based on support vector machine (SVM) until the accuracy of the agent model was similar to that of the physical mechanism model. The proxy model was able to calculate a polarization curve in 1 second, while a conventional CFD model could take hundreds of hours, and finally, the proxy model was combined with a genetic algorithm (GA) to obtain the optimal structural parameters of the CL. To verify the correctness of the model, as shown in Fig. 4.7, Wang et al. input the optimal structural parameters of the CL into the physical mechanism model and showed that the percentage error between the prediction of

116

Fuel Cells for Transportation

FIGURE 4.7 Surrogate model predicted and physical model simulated polarization curves under the optimal CL composition. Reproduced from B. Wang, et al., AI-based optimization of PEM fuel cell catalyst layers for maximum power density via data-driven surrogate modeling. Energy Convers. Manag. 205 (2020) 112460.

the proxy model and the maximum power density calculated by the physical mechanism model was only 1.3950%. Qiu et al. [39] developed a finite element (FE) model of a full-size single cell and related it to the pressure between the GDL and BP obtained from direct measurements. They used this model to generate a dataset and used it to train a radial basis function neural network (RBFNN), and finally input the trained RBFNN into a GA to obtain the optimal clamping pattern of the stack, and then input this clamping pattern into the FE model and found that the contact pressure distribution on the GDL was more uniform. Li et al. [40] developed a 3D steady-state CFD model of the PEMFC as the base model for the optimization target and then used ANOVA to select 6 parameters from 11 commonly used parameters that have a significant effect on the performance of the PEMFC, namely, width of the flow channel, thickness of the GDL, thickness of the PEM, inlet pressure, operating temperature, and stoichiometric ratio of the anode inlet, and then trained the three integrated learning models as data-driven agent model. Finally, the three PEMFC performance metrics, namely, power density, fuel cell system efficiency, and oxygen distribution uniformity of cathode CL, were simultaneously optimized based on non-dominated sorting genetic algorithm-II. The results showed that all three performance metrics of the PEMFC with optimal parameters were better than the model with initial structural parameters, successfully demonstrating the advantage of the method in solving timeconsuming multiple optimization problems.

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

4.3.2

117

Model predictive control

High-power density fuel cells require a good combination between oxygen supply, fuel supply, cooler, humidifier, and electronic control system. The air compressor, as one of the important auxiliary system components, helps oxygen flow to the cathode of the fuel cell, and its power consumption accounts for 80% of the auxiliary power consumption of the fuel cell system [41]. The operating conditions of fuel cells are very complex, and when the load current changes rapidly, the required mass flow rate within the fuel cell stack varies widely, which often leads to a lack of oxygen supply and thus oxygen starvation and reduced stack efficiency [42]. In particular, gas starvation can lead to many harmful consequences, such as carbon corrosion, catalyst degradation, and electrode thickness changes [43]. Oxygen starvation may shorten the life of the fuel cell and weaken the performance of the fuel cell. However, this does not mean that the more oxygen there is, the better the performance of the fuel cell. More oxygen causes the air compressor to consume more power, which reduces the energy conversion efficiency of the fuel cell. The air input also removes heat from inside the fuel cell, lowering the temperature and thus reducing the efficiency of power generation. Therefore, the adjustments of the oxygen metering ratio are a critical issue for fuel cells under actual driving conditions in fuel cell vehicles. High-order nonlinear fuel cell models are often used to analyze the nonlinear characteristics of the air supply system of a fuel cell, and their complexity leads to difficult controller design. The data-driven modeling approach, on the other hand, focuses more on experimental data and aims to describe the relationship between inputs and outputs without going deeper into the internal reaction principles of the fuel cell, making the controller design simpler. Wang et al. [41] designed a hierarchical model predictive control (MPC) by deep learning, and they used the predicted vehicle speed as the oxygen input control parameter in the fuel cell vehicle. The deep learning backpropagation neural network (BPNN) was then designed as the first-level predictor to forecast the vehicle speed by training with integrated driving cycles, and predict the fuel cell current based on its cathode flow model. Subsequently, the second-level MPC used the current disturbance prediction, and filling is introduced to regulate the oxygen mass flow. This controller avoids air starvation and air excess, increases the system output power, avoids excessive power wastage, and improves the fuel cell lifetime. The accurate control of automotive fuel cell oxygen excess ratio (OER) is necessary to improve system efficiency and service life. To this end, Chen et al. [44] improved the robustness and immunity of their controllers to prevent oxygen starvation and improve the output efficiency of the automotive fuel cell system by using a feedback linearized MPC cascade scheme for anti-disturbance control of oxygen input to the automotive fuel cell. It considers strong nonlinear coupling and disturbance injection of fuel cell oxygen

118

Fuel Cells for Transportation

supply. The feedback linearization demonstrates a robust tracking performance of nonlinear outputs, and the integral absolute error of antidisturbance control is 0.3021 lower than that of PI control under a custom test condition. Finally, the numerical validation on a hybrid driving cycle indicates that the proposed cascade control can regulate the fuel cell OER with an average absolute error of 0.02313 in the high air compressor operation efficiency zone. He et al. [45] proposed a switching MPC scheme for regulating the cyclic flow of hydrogen in a hydrogen circulation system. They designed a hydrogen circulation pump MPC scheme based on a piecewise linearization model of hydrogen circulation and a switching MPC controller. By predicting the pressure of the return manifold and the angular velocity of the pump, the proposed MPC scheme can manipulate the hydrogen circulation pump to achieve efficient and stable operation of the PEMFC. Compared with proportion integration (PI) control, as shown in Fig. 4.8, the switching MPC scheme has good control performance, fast response time, and high tracking accuracy under perturbation conditions.

4.3.3

Prognostics and health management

Prognostics and health management (PHM) is one of the most widely studied techniques for improving the durability of fuel cell devices. PHM can be used to monitor the state-of-health (SoH) and to estimate the remaining useful life (RUL) of fuel cells. It aims to improve the durability of fuel cells and save development costs. In the PHM framework, prediction is the key process to move from a “nofix” strategy to a “predictive prevention” strategy. Combining various prediction methods, the prediction process allows for estimating the RUL of a fuel cell. For fuel cells, the information provided by the prediction can be used as input to the management block, which then makes the decision to select the best system operating load [46]. The goal of making a decision is to maximize the life of the fuel cell and improve its durability. PHM technology can predict the future operation conditions of the system without having to run the system for a long time, which helps to reduce the experimental time and cost. Thus, the fuel cell development cost is reduced, and its durability is prolonged by reducing the experimental testing time in combination with other processes after making the decision. Zuo et al. [47] developed an attention-based recurrent neural network (RNN) model that more accurately predicted the degradation of the output voltage of PEMFC compared to other models. Liu et al. [48] proposed a PEMFC remaining life prediction model based on long short-term memory (LSTM) neural network for long-term aging prediction of PEMFC, which can quickly and accurately predict the remaining life of PEMFC. As shown in Fig. 4.9, the method uses regular interval sampling and locally weighted scatterplot smoothing to realize data reconstruction and data smoothing. Not only the primary trend of the original data can be preserved, but noise and spikes can be effectively removed. The LSTM RNN is adopted to estimate

FIGURE 4.8 Control results under varied step current condition: (A) start-up performance; (B) step change in stack current; (C) control voltage; (D) mass flow rate of hydrogen circulating pump. Reproduced from H. He, S. Quan, Y.-X. Wang, Hydrogen circulation system model predictive control for polymer electrolyte membrane fuel cell-based electric vehicle application. Int. J. Hydrog. Energy 45 (39) (2020) 20382
20390.

120

Fuel Cells for Transportation

FIGURE 4.9 RUL prognostic framework for PEMFC based on LSTM RNN. Reproduced from J. Liu et al., Remaining useful life prediction of PEMFC based on long short-term memory recurrent neural networks. Int. J. Hydrog. Energy 44 (11) (2019) 5470
5480.

the remaining life of test data. In total, 1154-hour experimental aging analysis of PEMFC shows that the prediction accuracy of the novel method is 99.23%, and the root mean square error (RMSE) and mean absolute error (MAE) are 0.003 and 0.0026, respectively. The comparison analysis shows that the prediction accuracy of the novel method is 28.46% higher than that of the BPNN. RMSE, relative error, and MAE are all much smaller than that of BPNN. Hua et al. [49] used echo state network to predict the remaining lifetime of PEMFC, which can accelerate the convergence speed and reduce the computational complexity compared with the traditional RNN.

4.3.4

Fault diagnosis

The durability of a fuel cell system depends heavily on the operating conditions. The life of a fuel cell system can only be extended by ensuring a

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

121

stable air supply, fuel supply, water management, and thermal management [50]. Therefore, the reliability of the balance of plant (BoP) affects the durability of the system. Since the BoP of a fuel cell system is composed of several components such as mechanical moving parts, faults are inevitable. When a fault occurs in a fuel cell system, the performance of the system will be reduced and the durability will be lowered. In addition, the fuel cell system can even be permanently damaged, which depends on the severity of the fault. Therefore, fault diagnosis techniques for fuel cell systems are essential to maintain reliability and durability of fuel cell systems. Fault diagnosis studies of fuel cell systems are mainly divided into two categories, which are model-based fault diagnosis methods and data-based fault diagnosis methods. In the model-based fault diagnosis algorithm, the fuel cell model is first built based on a real fuel cell system, which can be either a physical model or a black box model. Then, the established fuel cell model is considered as the expected normal operation state of the system, and the distance between the fuel cell system and the above-expected model is calculated. Finally, a residual analysis of the distance is performed to determine if there is a fault in the system. Although many favorable results have been achieved by physical mechanism-based methods, it is difficult to accurately establish a physical model of the system because the system is becoming more and more complex, which makes it difficult to implement fault diagnosis methods based on physical mechanism models. Therefore, data-based fault diagnosis methods are receiving increasing attention. The data-driven approach does not require a large amount of specific model knowledge, but only requires collecting a certain amount of data and constructing the key features of the system. Using machine learning algorithms, fault patterns are learned from historical information data, thus enabling reliable monitoring of faults. Park et al. [51] applied single-task learning techniques to a neural network fault diagnosis model, which combined several single-output neural networks with significantly improved diagnostic capabilities. In their study, thermal management system fault experiments are repeatedly performed under various loads and stack degradation conditions. With the degradation of the stack, waste heat from the stack increases and fault response as well as its impact on the system varies. The diagnosis model developed in this study detects faults at the component level and diagnoses the severity of the faulty components. Li et al. [52] used Fisher discriminant analysis (FDA) and SVM methods to diagnose different types of faults and different fuel cell groups, and the diagnostic accuracy could be maintained at a high level. Liu et al. [53] proposed a PEMFC fault diagnosis method using 2D image data and investigated the effectiveness of distinguishing the optimal features in various states using the K-means clustering algorithm. In the analysis, as shown in Fig. 4.10, one-dimensional (1D) voltage data from single cell is converted to a corresponding 2D image using the signal-to-image conversion technique. Various features are then extracted from the 2D image data, and optimal features are

122

Fuel Cells for Transportation

FIGURE 4.10 1D voltage signals and converted 2D images. Reproduced from Z. Liu, et al., A novel method for polymer electrolyte membrane fuel cell fault diagnosis using 2D data. J. Power Sources 482 (2021) 228894.

determined using FDA. Test data from PEMFC at different faulty states, including flooding and dehydration states, are collected for the analysis, and the effectiveness of optimal features in discriminating various states is investigated using K-means clustering method.

4.4 Combination of lattice Boltzmann method and artificial intelligence Although some researchers have coupled LBM and AI to solve some engineering problems, none have yet applied them to the field of PEM fuel cells. Meisam et al. [54] used an adaptive neuro-fuzzy inference system to predict fluid flow pattern recognition in the 3D cavity, which reduced the computational time for visualization of fluid in the 3D domain. Zhao et al. [55] proposed an improved PN extraction method to describe the throat geometry based on the watershed method. They coupled the LBM and pore network model (PNM) to simulate two-phase drainage flow in porous media at the pore scale. In order to further improve the computational efficiency, they used five artificial neural network models which relate the flow properties to

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

123

the actual shape of throat cross sections. The results showed that the trained artificial neural network model showed a better performance than the conventional PNM in predicting the flow characteristics of each throat bond. Yan et al. [56] combined Monte Carlo methods, LBM, and AI to propose a modeling framework for electrode microstructure optimization in solid oxide fuel cells (SOFCs). They used AI-assisted multi-objective optimization to simulate SOFC in the order from fabrication to electrochemical performance. They investigated the effects of initial particle parameters, particle size distribution, and porosity on the overpotential and degradation rate of the cathode of SOFC. The results showed that small particle diameter and low porosity lead to low overpotential but high degradation rate at the cathode in the studied parameter range. Since the LBM can be applied to solve the mass transfer, two-phase flow, and electrochemical reaction processes in the microstructure of a real porous electrode, it can be used in combination with AI to rapidly calculate the performance of porous electrodes. Overall, the coupled model can be used for performance prediction, structure optimization, and parameter search of porous electrodes for PEMFC.

4.5

Summary

This chapter focuses on the application of LB and AI methods in PEM fuel cells. For high-power density fuel cells to be commercialized, cost and performance are crucial factors. The microstructure of porous electrodes significantly impacts the processes of energy and mass transfer, as well as performance degradation in fuel cells. However, studying these processes at a mesoscopic scale using LB methods is computationally expensive and limited by computing time and resources. AI methods, on the other hand, can analyze large data sets to obtain relatively accurate simulation results without knowledge of the physicochemical processes. By combining LB and AI methods, the limitations of LB methods in time and space scales can be overcome, and they can be used for performance prediction, structure optimization, and parameter optimization in the porous electrodes of PEMFCs. This combination can help achieve the goal of high-power density fuel cell with ultra-low Pt loading.

References [1] R. Chen, et al., Numerical simulation of liquid water emerging and transport in the flow channel of PEMFC using the volume of fluid method, Int. J. Hydrog. Energy 45 (54) (2020) 29861
29873. [2] Z. Niu, et al., Two-phase flow and oxygen transport in the perforated gas diffusion layer of proton exchange membrane fuel cell, Int. J. Heat. Mass. Transf. 139 (2019) 58
68. [3] F.C. Cetinbas, et al., Direct simulations of pore-scale water transport through diffusion media, J. Electrochem. Soc. 166 (7) (2019) F3001
F3008.

124

Fuel Cells for Transportation

[4] J. Hinebaugh, A. Bazylak, Condensation in PEM fuel cell gas diffusion layers: a pore network modeling approach, J. Electrochem. Soc. 157 (10) (2010) B1382. [5] G.R. Molaeimanesh, M.H. Akbari, A three-dimensional pore-scale model of the cathode electrode in polymer-electrolyte membrane fuel cell by lattice Boltzmann method, J. Power Sources 258 (2014) 89
97. [6] Q. Li, et al., Lattice Boltzmann methods for multiphase flow and phase-change heat transfer, Prog. Energy Combust. Sci. 52 (2016) 62
105. [7] T. Chen, Q. Zhang, L. Cheng, Performance investigation of 2D lattice Boltzmann simulation of forces on a circular cylinder, Trans. Tianjin Univ. 16 (6) (2010) 417
423. [8] T. Kr¨uger, et al., The Lattice Boltzmann Method, Springer, Cham. XXIV, 2017, p. 694. [9] N. Delbosc, et al., Optimized implementation of the Lattice Boltzmann Method on a graphics processing unit towards real-time fluid simulation, Comp. Math. Appl. 67 (2) (2014) 462
475. [10] L. Chen, et al., Pore-scale modeling of complex transport phenomena in porous media, Prog. Energy Combust. Sci. 88 (2022) 100968. [11] Y. Wang, et al., Fundamentals, materials, and machine learning of polymer electrolyte membrane fuel cell technology, Energy AI 1 (2020) 100014. [12] A. Demuren, R.L. Edwards, Modeling proton exchange membrane fuel cells—a review, in 50 years of cfd in engineering sciences: a commemorative volume, in: A. Runchal (Ed.), Memory of D. Brian Spalding, Springer Singapore, Singapore, 2020, pp. 513
547. [13] L. Hao, P. Cheng, Lattice Boltzmann simulations of water transport in gas diffusion layer of a polymer electrolyte membrane fuel cell, J. Power Sources 195 (12) (2010) 3870
3881. [14] L. Chen, et al., Numerical investigation of liquid water transport and distribution in porous gas diffusion layer of a proton exchange membrane fuel cell using lattice Boltzmann method, Russ. J. Electrochem. 48 (7) (2012) 712
726. [15] L. Chen, et al., Pore-scale flow and mass transport in gas diffusion layer of proton exchange membrane fuel cell with interdigitated flow fields, Int. J. Therm. Sci. 51 (2012) 132
144. [16] P.A. Garcı´a-Salaberri, et al., Effective diffusivity in partially-saturated carbon-fiber gas diffusion layers: effect of through-plane saturation distribution, Int. J. Heat. Mass. Transf. 86 (2015) 319
333. [17] H. Deng, Y. Hou, K. Jiao, Lattice Boltzmann simulation of liquid water transport inside and at interface of gas diffusion and micro-porous layers of PEM fuel cells, Int. J. Heat. Mass. Transf. 140 (2019) 1074
1090. [18] P. Rama, et al., An x-ray tomography based lattice boltzmann simulation study on gas diffusion layers of polymer electrolyte fuel cells, J. Fuel Cell Sci. Technol. 7 (3) (2010). [19] Y. Hou, et al., 3D lattice Boltzmann modeling of droplet motion in PEM fuel cell channel with realistic GDL microstructure and fluid properties, Int. J. Hydrog. Energy 45 (22) (2020) 12476
12488. [20] X. Shangguan, et al., Effect of the porosity distribution on the liquid water transport in the gas diffusion layer of PEMFC, Electrochim. Acta 371 (2021) 137814. [21] H. Wang, et al., Numerical study on permeability of gas diffusion layer with porosity gradient using lattice Boltzmann method, Int. J. Hydrog. Energy 46 (42) (2021) 22107
22121. [22] N. Zamel, J. Becker, A. Wiegmann, Estimating the thermal conductivity and diffusion coefficient of the microporous layer of polymer electrolyte membrane fuel cells, J. Power Sources 207 (2012) 70
80.

Lattice Boltzmann modeling and artificial intelligence Chapter | 4

125

[23] K.N. Kim, et al., Lattice Boltzmann simulation of liquid water transport in microporous and gas diffusion layers of polymer electrolyte membrane fuel cells, J. Power Sources 278 (2015) 703
717. [24] B. Han, H. Meng, Numerical studies of interfacial phenomena in liquid water transport in polymer electrolyte membrane fuel cells using the lattice Boltzmann method, Int. J. Hydrog. Energy 38 (12) (2013) 5053
5059. [25] D. Zhang, Q. Cai, S. Gu, Three-dimensional lattice-Boltzmann model for liquid water transport and oxygen diffusion in cathode of polymer electrolyte membrane fuel cell with electrochemical reaction, Electrochim. Acta 262 (2018) 282
296. [26] S. Ye, et al., Pore-scale investigation of coupled two-phase and reactive transport in the cathode electrode of proton exchange membrane fuel cells, Trans. Tianjin Univ. (2022). [27] X. Li, et al., Interlink among catalyst loading, transport and performance of proton exchange membrane fuel cells: a pore-scale study, Nanoscale Horiz. 7 (3) (2022) 255
266. [28] L. Chen, et al., Lattice Boltzmann pore-scale investigation of coupled physicalelectrochemical processes in C/Pt and non-precious metal cathode catalyst layers in proton exchange membrane fuel cells, Electrochim. Acta 158 (2015) 175
186. [29] L. Chen, et al., Nanoscale simulation of local gas transport in catalyst layers of proton exchange membrane fuel cells, J. Power Sources 400 (2018) 114
125. [30] L. Chen, Q. Kang, W. Tao, Pore-scale study of reactive transport processes in catalyst layer agglomerates of proton exchange membrane fuel cells, Electrochim. Acta 306 (2019) 454
465. [31] Y.Z. Hou, et al., Pore-scale investigation of catalyst layer ingredient and structure effect in proton exchange membrane fuel cell, Appl. Energy (2019) 253. [32] L. Chen, Q. Kang, W. Tao, Pore-scale numerical study of multiphase reactive transport processes in cathode catalyst layers of proton exchange membrane fuel cells, Int. J. Hydrog. Energy 46 (24) (2021) 13283
13297. [33] A. Ozden, et al., A review of gas diffusion layers for proton exchange membrane fuel cells—with a focus on characteristics, characterization techniques, materials and designs, Prog. Energy Combust. Sci. 74 (2019) 50
102. [34] Y. Song, et al., Review on current research of materials, fabrication and application for bipolar plate in proton exchange membrane fuel cell, Int. J. Hydrog. Energy 45 (54) (2020) 29832
29847. [35] L. Fan, et al., Optimization design of the cathode flow channel for proton exchange membrane fuel cells, Energy Convers. Manag. 171 (2018) 1813
1821. [36] B. Wang, et al., A dot matrix and sloping baffle cathode flow field of proton exchange membrane fuel cell, J. Power Sources 434 (2019) 226741. [37] X. Zhou, et al., Effects of surface wettability on two-phase flow in the compressed gas diffusion layer microstructures, Int. J. Heat. Mass. Transf. 151 (2020) 119370. [38] B. Wang, et al., AI-based optimization of PEM fuel cell catalyst layers for maximum power density via data-driven surrogate modeling, Energy Convers. Manag. 205 (2020) 112460. [39] Y. Qiu, et al., An intelligent approach for contact pressure optimization of PEM fuel cell gas diffusion layers, Appl. Sci. 10 (12) (2020) 4194. [40] H. Li, et al., Multi-objective optimization of PEM fuel cell by coupled significant variables recognition, surrogate models and a multi-objective genetic algorithm, Energy Convers. Manag. 236 (2021) 114063.

126

Fuel Cells for Transportation

[41] X. Wang, et al., Hierarchical model predictive control via deep learning vehicle speed predictions for oxygen stoichiometry regulation of fuel cells, Appl. Energy 276 (2020) 115460. [42] H. Chen, et al., An evaluation method of gas distribution quality in dynamic process of proton exchange membrane fuel cell, Appl. Energy 232 (2018) 26
35. [43] R. Ma, et al., Data-driven proton exchange membrane fuel cell degradation predication through deep learning method, Appl. Energy 231 (2018) 102
115. [44] J. Chen, et al., Anti-disturbance control of oxygen feeding for vehicular fuel cell driven by feedback linearization model predictive control-based cascade scheme, Int. J. Hydrog. Energy 45 (58) (2020) 33925
33938. [45] H. He, S. Quan, Y.-X. Wang, Hydrogen circulation system model predictive control for polymer electrolyte membrane fuel cell-based electric vehicle application, Int. J. Hydrog. Energy 45 (39) (2020) 20382
20390. [46] N. Herr, et al., Decision process to manage useful life of multi-stacks fuel cell systems under service constraint, Renew. Energy 105 (2017) 590
600. [47] J. Zuo, et al., Deep learning based prognostic framework towards proton exchange membrane fuel cell for automotive application, Appl. Energy 281 (2021) 115937. [48] J. Liu, et al., Remaining useful life prediction of PEMFC based on long short-term memory recurrent neural networks, Int. J. Hydrog. Energy 44 (11) (2019) 5470
5480. [49] Z. Hua, et al., Remaining useful life prediction of PEMFC systems under dynamic operating conditions, Energy Convers. Manag. 231 (2021) 113825. [50] Z. Zheng, et al., A review on non-model based diagnosis methodologies for PEM fuel cell stacks and systems, Int. J. Hydrog. Energy 38 (21) (2013) 8914
8926. [51] J.Y. Park, et al., Fault diagnosis of thermal management system in a polymer electrolyte membrane fuel cell, Energy 214 (2021) 119062. [52] Z. Li, et al., Fault diagnosis for fuel cell systems: a data-driven approach using highprecise voltage sensors, Renew. Energy 135 (2019) 1435
1444. [53] Z. Liu, et al., A novel method for polymer electrolyte membrane fuel cell fault diagnosis using 2D data, J. Power Sources 482 (2021) 228894. [54] M. Babanezhad, et al., Pattern recognition of the fluid flow in a 3D domain by combination of Lattice Boltzmann and ANFIS methods, Sci. Rep. 10 (1) (2020) 15908. [55] J. Zhao, et al., Simulation of quasi-static drainage displacement in porous media on porescale: coupling lattice Boltzmann method and pore network model, J. Hydrol. 588 (2020) 125080. [56] Z. Yan, et al., Modeling of solid oxide fuel cell (SOFC) electrodes from fabrication to operation: microstructure optimization via artificial neural networks and multi-objective genetic algorithms, Energy Convers. Manag. 198 (2019) 111916.

Chapter 5

Low platinum-based electrocatalysts for fuel cells: status and prospects Huiyuan Liu and Xianguo Li Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada

5.1

Introduction

Due to the acid electrolyte and low-temperature operation, PEM fuel cells require electrocatalysts with high activity and durability to accelerate the electrochemical processes, especially the cathodic oxygen reduction reaction (ORR) that is multielectron and multistep with sluggish kinetics and the oxidizing, acidic, and high-potential environment at the cathode. At the current technology status, carbon-supported Pt-based electrocatalysts have been considered to have the highest catalytic activity and durability for ORR among the electrocatalysts available [1,2]. However, the high cost and still insufficient activity and stability associated with the present carbonsupported Pt-based electrocatalysts hamper the widespread practical application of PEM fuel cells. To reduce the usage of Pt and thus the cost, two major technical pathways on electrocatalysts for ORR have been widely employed. The first pathway is to prepare an active and stable carbonsupported Pt-based electrocatalyst by controlling the size, size distribution, composition, shape, and/or structure and developing simple low-cost synthesis methods, which can be potentially used for PEM fuel cells at a commercial level [3,4]. The second pathway is to develop nonprecious metal-based electrocatalysts with high active site density and durability [5,6]. However, although the performance of nonprecious metal-based electrocatalysts has been greatly improved recently, there is still a long way to go for practical application, compared with Pt-based electrocatalysts [7]. Therefore this chapter is primarily focused on the first pathway. The current development of the controlled synthesis of Pt-based electrocatalysts for PEM fuel cells will be covered in this chapter, including the size, Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00017-4 © 2023 Elsevier Ltd. All rights reserved.

127

128

Fuel Cells for Transportation

size distribution, functionalization of carbon supports, shape, and simple lowcost synthesis, by controlling the experimental parameters. At first, an overview of carbon supports and the effects of structure and surface properties of commercial carbon material on Pt-based electrocatalysts are presented. Then, it proceeds to the methods for loading Pt-based electrocatalysts onto carbon by one-pot synthesis and ex situ mixing methods, and the experimental studies on the size and shape of Pt-based electrocatalysts, particularly strategies to control the size and shape of Pt-based electrocatalysts, are summarized by analyzing the recent experimental studies of Pt-based electrocatalysts. Finally, different postsynthesis treatment processes and their influence on the final Pt-based electrocatalysts are reviewed, and perspectives for future research and development of Pt-based electrocatalysts are provided.

5.2

Functionalization of commercial carbon supports

To improve the activity, durability, and Pt utilization efficiency, Pt-based electrocatalysts for PEM fuel cells are usually loaded on carbon materials, which usually have high surface area, suitable porosity, excellent electrical conductivity, chemical stability, and low cost, allowing high dispersion of Pt-based nanoparticles. The commonly used commercial carbon materials include carbon black, graphene, carbon nanotubes (CNTs), and carbon nanofibers. The commercial carbon black has Vulcan XC-72R (VXC-72R), Ketjen Black EC300J (EC-300), Ketjen Black EC600JD (EC-600), and Black Pearls 2000 (BP2000). VXC-72R has a moderate surface area due to a small quantity of internal pores, usually known as solid carbon black. EC-300, EC-600, and BP-2000 have very high surface area due to abundant internal pores, usually known as porous carbon black [2,8]. The properties of the commercial carbon commonly used as Pt-based electrocatalysts support are exhibited in Table 5.1.

5.2.1 Effects of the structure and surface properties of carbon support As have been widely explored and accepted in heterogeneous catalysis, carbon support not only acts as simple physical support or electron transport pathway in fuel cells, but its structure and surface properties also have a crucial impact on the performance and durability of Pt-based electrocatalysts [16
19]. The structure properties include the surface area, pore size and size distribution, pore shape, and pore volume. The surface properties include hydrophilicity
hydrophobicity, the type and number of functional groups, and the number of defects. The effects of structure and surface properties of carbon on Pt-based electrocatalysts are as follows: 1. The effects on the size, size distribution, shape, or dispersion of Pt-based electrocatalysts will influence the electrochemical surface area (ECSA) and

TABLE 5.1 The commercial carbon commonly used as Pt-based electrocatalysts support [2,9
15]. Carbon

Type of carbon

Supplier

BET surface area (m2 g21)

Diameter (nm)

DBP adsorption (units)

VXC-72R

Furnace black

Cabot Corp.

254

30

190

EC-300

Furnace black

Ketjen Black International

800

B30

360

EC-600

Furnace black

Ketjen Black International

1270

B34

495

BP-2000

Furnace black

Cabot Corp.

1475

15

330

CNTs




Cheap Tubes; Shandong Dazhan Nano Materials; SkySpring Nanomaterials; Sigma

110
600

Outer D 1
30 nm, insider D 0.8
10 nm, length 5
30 um




Graphene




SkySpring Nanomaterials; Graphene Star; Sigma

120
700

Thickness: 1
10 nm




BET, Brunauer 2 Emmett 2 Teller method, DBP, dibutyl phthalate number (measure of carbon void volume).

130

Fuel Cells for Transportation

activity [8,20
22]. (a) The surface properties of carbon would affect the dispersion of Pt-based electrocatalysts on carbon through influencing the interaction between carbon and solvent [23]. For polar solvent, for example, water, the hydrophobic carbon has a very low affinity with it. The Pt ions aqueous solution merely contacts the external surface of the hydrophobic carbon, but cannot penetrate its pore; hence, Pt-based nanoparticles will only locate at the external surface of carbon. However, nonpolar solvent, for example, acetone, could penetrate the pore of the hydrophobic carbon. The Pt-based nanoparticles will distribute on the external surface and in the pore. The affinity of carbon with polar solvent, that is, hydrophilicity, could be improved by surface functionalization, such as introducing O-containing groups or N/S-containing groups [23]. Increasing the hydrophilicity of carbon will improve the dispersion of carbon-supported Pt-based electrocatalysts in the ink as well (commonly using the mixed solvent of water and ethanol/isopropanol, which are polar solvent) and the quality of the fabricated catalyst layer [24]. (b) The carbon’s surface properties would impact the size and dispersion of the Pt-based electrocatalysts by affecting the interaction between carbon and Pt ions [23,25,26]. If the charge of Pt ions is opposite with the groups on the carbon surface, for example, positively charged Pt complexes (Pt(acac)2 or PtCl2) and the carbon with -COOH or -SO3H, or negatively charged Pt complexes (H2PtCl4, K2PtCl4, or H2PtCl6) and the carbon with -NHx, the Pt nanoparticles often show a small size and uniform dispersion on carbon [23]. Otherwise, large Pt nanoparticles would be formed. (c) The carbon surface area will influence the size and dispersion of Pt-based nanoparticles as well. For a given Pt loading, the Pt particle size generally decreases with an increase in the carbon surface area in one-pot synthesis method. To prepare Pt/C with high Pt loadings and guarantee the even dispersion of Pt particles on carbon, the carbon with high surface area is often selected [27]. 2. The effect on the performance of Pt-based electrocatalysts in the catalyst layer. For solid carbon, for example, VXC-72R, Pt-based nanoparticles are primarily located on the external surface of carbon, leading to good local mass transport properties in the catalyst layer. However, the activity of Pt-based electrocatalysts is generally reduced because of the poison of the sulfonate groups in the ionomer [8]. For porous carbon, for example, EC-600, most Pt-based nanoparticles deposit inside the pores, commonly leading to good activity as the ionomer cannot penetrate the small pores and contact Pt-based nanoparticles located in the pore. However, deep and tortuous pores would cause poor local mass transport [8]. Moreover, the carbon’s surface properties would affect the activity of Pt-based electrocatalysts through the metal
support interaction, including geometric effects and electronic effects [17,28]. 3. The effect on the durability of Pt-based electrocatalysts [29,30]. A strong metal
support interaction or forming the chemical binding between

Low platinum-based electrocatalysts for fuel cells Chapter | 5

131

metal and support will improve the durability of Pt-based electrocatalysts due to the limited migration of Pt-based nanoparticles [29,31]. The carbon with a high graphite degree commonly shows enhanced stability and thus will improve the durability of supported Pt-based electrocatalysts due to the reduced detachment of Pt-based nanoparticles caused by carbon corrosion. 4. The effects on the ionomer distribution on Pt/C in the catalyst layer. The sulfonate groups of ionomer are negatively charged. The ionomer will repel each other with negatively charged carbon and attract with positively charged carbon due to the electrostatic interaction. Therefore the type of functional groups on the carbon surface will impact the size of aggregates and the ionomer distribution on Pt/C in the catalyst layer and further affect the local mass transport. For example, compared to the carbon-containing -SO3H
supported Pt, the ionomer would distribute more uniformly on carbon-containing -NHx-supported Pt, resulting in improved mass transport, especially at high current density region [32,33]. 5. The effects on postsynthesis treatment. In our previous experiments, it was found that the carbon surface area could impact the absorbed amount of surfactant molecules and thus influence the postsynthesis treatment. For 20 wt.% Pt/C prepared by the phase-transfer method, if using VXC72R as support, the cetyltrimethylammonium bromide (CTAB) molecules absorbed on Pt/C can be removed by washing with a copious amount of hot water [34]. However, if using EC-600 as the support of 20 wt.% Pt/C, it is difficult to completely remove the absorbed CTAB even though using much more hot water, as shown in Fig. 5.1. Compared to VXC-72R, EC-600 has a higher surface area and thus more active sites

FIGURE 5.1 The thermogravimetric analysis (TGA) curve of B 20 wt.% Pt/EC600 prepared by phase-transfer method [cetyltrimethylammonium bromide (CTAB)]. There is a clearly identifiable weight loss peak for the CTAB in the TGA curve.

132

Fuel Cells for Transportation

(i.e., nucleation sites), which strongly absorb not only the Pt nanoparticles but also CTAB molecules. Therefore the active sites on the EC-600 surface unoccupied by Pt nanoparticles will strongly absorb CTAB molecules, which would not be easy to be removed by simple hot water washing. If increasing Pt loading on EC-600 to 40 wt.% or 60 wt.%, most or even all active sites are occupied by Pt nanoparticles, and thus, there are only a little or even no strongly absorbed CTAB. The absorbed CTAB could be removed by washing with a copious amount of hot water as well [34].

5.2.2

Functionalization methods of commercial carbon supports

For commercial carbon materials, the structure cannot be modified easily. However, the surface properties of carbon can be modulated by functionalization before loading Pt-based electrocatalysts to increase the number of nucleation sites or anchoring sites of Pt-based electrocatalysts and strengthen the metal
support interaction. The commonly used functionalization methods are as follows: 1. Oxidation treatment. To increase the oxygen-containing groups, pristine commercial carbon materials are usually treated in oxidizing acid solution (e.g., HNO3 and H2SO4) at a certain temperature for a certain period [5,17,35
40], or treated under air atmosphere at an elevated temperature, or ball-milled under air atmosphere for a certain period [37]. During the treatment process, the C
C bonds on the carbon surface are attacked by the oxidants and broken, forming dangling bonds or oxygen-containing groups by bonding carbon with oxygen atoms. Based on infrared radiation (IR) spectroscopy or X-ray photoelectron spectroscopy (XPS), oxygen-containing groups can be formed on the carbon surface, such as carboxylic, hydroxyl, lactone, phenol, epoxy, carbonyl, anhydride, ether, or quinone groups. The dangling bonds and oxygen-containing groups will act as the nucleation sites or anchoring sites for the Pt-based electrocatalysts via electrostatic, coordinative, or van der Waals interactions [17,37,38,41,42]. This functionalization method is a method most frequently used as well as simple and effective. However, because of the destruction of graphite structure on the carbon surface during oxidation treatment, the conductivity and stability of carbon support are reduced [17,43,44], causing low output performance and durability for oxidized carbon-supported Pt-based electrocatalysts in PEM fuel cells. 2. Doping heteroatoms. Due to the different electron configuration, atom size, and electronegativity from carbon, the doped heteroatoms, for example, N, S, B, would modulate the surface physicochemical properties of carbon [45
49] and further influence the interaction between

Low platinum-based electrocatalysts for fuel cells Chapter | 5

133

carbon support and Pt-based electrocatalysts [50,51], which will impact the activity and durability of supported Pt-based electrocatalysts [49,52,53]. Besides, the carbon-doped heteroatoms would provide additional nonmetal active sites. Doping heteroatoms is generally performed by annealing the mixture of oxidized commercial carbon with heteroatom-containing materials or annealing the oxidized carbon in NH3/H2S/CS2 atmosphere [36]. The N-containing materials used in the literature have 1,10-phenanthroline [54], urea [50,55], or various amines [56
59] (dopamine, dicyandiamide, melamine, ethylene diamine, diethylene triamine, triethylene tetramine, tetraethylene pentamine, or pentaethylene hexamine). The S-containing materials used in the literature have thiourea [60], phenyl disulfide [38,61], benzyl disulfide [62], or sulfur (S8) [63]. The B-containing materials used in the literature have H3BO3 [49,64,65]. The doping amount of heteroatoms in carbon can be modulated by regulating the reaction conditions to some extent, for example, annealing temperature, the type or amount of the heteroatom-containing materials [36,50,57,60,66]. However, the doping amount of heteroatoms is commonly low, less than 10 at% or even less than 1 at% in some cases [36,38,48
50,55
67]. Based on the XPS measurement, the N-containing species introduced in the carbon matrix include pyrrolic N, pyridinic N, graphitic N, or pyridine-N-oxide [57]; S-containing species introduced in the carbon matrix include thiophene or -SOx [61]; B-containing species in the carbon matrix have BC3 (B-doped carbon), BC2O, or BCO2 [64] (Fig. 5.2). 3. Introducing functional groups. Carbon can also be functionalized by introducing the thiol (-SH), amine (-NHx), or carboxyl (-COOH) groups, or the groups, molecules, or polymers with -SH, -NHx, or -COOH terminal moieties. To introduce these functional groups onto carbon, two methods, that is, covalently grafting and noncovalent interaction, have been developed. Covalently grafting [21,33,68
71], that is, forming covalent binding between carbon and functional groups, can be realized by replacing oxygen-containing groups with functional groups (such as -SH or -NH2) [70], forming amide bonds (-CONH-) [72], diazonium reaction [71], or direct Friedel
Crafts reaction [21] (Fig. 5.3A
D). The molecules or polymers with functional groups could be introduced onto carbon by π-π interaction between the pyrene, phenyl moiety, or C 5 C moiety of the molecules or polymers and the graphite structure of carbon materials, as shown in Fig. 5.3E [25,26,68,73
77]. The groups, molecules, or polymers introduced act as the inter-linker between Pt-based electrocatalysts and carbon. Therefore the functional groups should be close to carbon surface as much as possible. A long and flexible chain with -SH, -NHx, or -COOH terminal moieties likely generates a large contact resistance between Pt-based electrocatalysts and

134

Fuel Cells for Transportation

FIGURE 5.2 Schematic illustration of heteroatoms-doped reduced graphene oxide (r-GO). Adapted with permission C. R. Raj, A. Samanta, S. H. Noh, S. Mondal, T. Okajima and T. Ohsaka, Emerging new generation electrocatalysts for the oxygen reduction reaction, J. Mater. Chem. A 4 (2016) 11156
11178; Copyright 2016, Royal Society of Chemistry and the atomic data of B, C, N, and S including the atomic number, outer electron configuration, covalent radius, and electronegativity. https://www.rsc.org/periodic-table.

carbon, impacting the electron transport and then decreasing the activity of electrocatalysts [21,26,78]. Furthermore, -SH or -NHx often has a strong interaction with Pt-based electrocatalysts, probably poisoning Pt-based electrocatalysts [73]. 4. Introducing metal oxides. Introducing metal oxides on carbon support will improve the performance and durability of carbon-supported Pt-based electrocatalysts, due to the high stability of metal oxides in the oxidative and high-potential environment and the enhanced metal
support interaction resulting from the strong metal
metal oxides interaction [79
81]. The metal oxide nanoparticles also act as a physical barrier, declining the opportunity to encounter each other for Pt-based nanoparticles to some degree and thus improving the durability. Besides, the CO tolerance of Pt-based electrocatalysts is commonly raised because the CO absorbed by Pt-based electrocatalysts could be easily oxidized by oxygen-containing species supplied by metal oxides [79,81,82]. The metal oxides commonly introduced include indium tin oxide [81], CeO2 [82], TiO2 [83], SnO2 [84], NbOx [85], WO3 [86], or Mn3O4 [87]. Fig. 5.4A exhibits three situations for Pt-based nanoparticles supported on carbon
metal oxide composite supports with large metal oxide particles, (1) only on carbon surface (case 1), (2) only on metal oxide surface (case 2), (3) at the junctions of carbon and metal oxide particle

Low platinum-based electrocatalysts for fuel cells Chapter | 5

135

FIGURE 5.3 The functional groups onto carbon supports could be introduced by covalently grafting, for example, forming amide bonds. (A) replacing oxygen-containing groups with
SH. (B) Friedel-Crafts reaction, (C) and diazonium reaction (E) or by π-π interaction. (A) Reprinted with permission from Ref. L. Li, H. Liu, L. Wang, S. Yue, X. Tong, T. Zaliznyak, et al., chemical strategies for enhancing activity and charge transfer in ultrathin Pt nanowires immobilized onto nanotube supports for the oxygen reduction reaction. ACS Appl. Mater. Interfaces 8 (2016) 34280
34294. Copyright 2016 American Chemical Society; (B) M.-H. Hsu, H. Chuang, F.-Y. Cheng, Y.-P. Huang, C.-C. Han, K.-C. Pao, et al., Simple and highly efficient direct thiolation of the surface of carbon nanotubes. RSC Adv. 4 (2014) 14777
14780. Copyright 2014, Royal Society of Chemistry; (C) B. Wu, C. Wang, Y. Cui, L. Mao and S. Xiong, Tailoring carbon nanotubes surface with maleic anhydride for highly dispersed PtRu nanoparticles and their electrocatalytic oxidation of methanol. RSC Adv. 5 (2015) 16986
16992. Copyright 2015 American Chemical Society; (D) L. Xin, F. Yang, S. Rasouli, Y. Qiu, Z.-F. Li, A. Uzunoglu, et al., Understanding Pt nanoparticle anchoring on graphene supports through surface functionalization. ACS Catal. 6 (2016) 2642
2653. Copyright 2016 American Chemical Society; (E) H.-S. Oh and H. Kim, Efficient synthesis of Pt nanoparticles supported on hydrophobic graphitized carbon nanofibers for electrocatalysts using noncovalent functionalization. Adv. Funct. Mater. 21 (2011) 3954
3960. Copyright 2011, Wiley-VCH.

(case 3). In case 1, metal oxide likely has no impact on the performance of Pt-based electrocatalysts; in case 2, Pt-based nanoparticles may lose the electron channel arising from the low electronic conductivity of metal oxides. The effects of metal oxides on the performance and durability of Pt-based electrocatalysts may be embodied only in case 3. To maximize the advantages of carbon
metal oxide composite supports, the amount of junctions should be boosted by increasing the dispersion of metal oxide on carbon or reducing the size of metal oxide particles to the size of Pt nanoparticles [88] (Fig. 5.4B). However, depositing small metal oxide particles on carbon is still a challenge.

136

Fuel Cells for Transportation

FIGURE 5.4 Schematic of the location of Pt-based nanoparticles loaded on carbon
metal oxide composite supports with large (A) or small (B) metal oxide particles.

TABLE 5.2 Comparison of functionalization methods. Methods

Pros

Cons

Oxidation treatment

Simple

Decreasing the conductivity and stability of carbon due to the destroyed graphite structure on carbon surface

Doping heteroatoms

Strong metal
support interaction

Low doping amount

Introducing functional groups

Covalently grafting: Stable due to covalent bondNoncovalent interaction: 1. Simple 2. Complete graphite structure

1. Long and flexible chain—a large contact resistance 2.
SH or
NHx may poison Pt 3. Covalently grafting: destroying the graphite structure 4. Noncovalent interaction: π
π stacking, a weaker interaction, unstable

Introducing metal oxides

1. Improved stability 2. Enhanced CO tolerance

Depositing small metal oxide particles is still challenging

Each functionalization method has its pros and cons, and their comprehensive comparison is summarized in Table 5.2. The selection of the carbon functionalization method ought to rely on the practical case.

5.3 Methods for loading Pt-based electrocatalysts on carbon supports To prepare the electrocatalysts with high quality, efficiently supporting Ptbased electrocatalysts on carbon is important. According to the time adding

Low platinum-based electrocatalysts for fuel cells Chapter | 5

137

carbon supports (before or after the formation of Pt-based nanostructures), the experimental methods commonly used can be categorized into one-pot synthesis method (in situ loading) and ex situ mixing method [22,42].

5.3.1

One-pot synthesis method

In the one-pot synthesis method, carbon is added to the reaction system before the formation of Pt-based electrocatalysts. Carbon-supported Pt-based electrocatalysts will be prepared only through one step, which could reduce the cost associated with the synthesis of electrocatalysts [4]. During the reaction process, Pt may directly nucleate or grow on carbon. The chemical bonds are likely formed between the Pt-based electrocatalysts and carbon support through the functional groups or dangling bonds on the carbon surface, improving the stability of electrocatalysts [89,90]. However, the morphology of Pt-based electrocatalysts may not be easily controlled because of the interference of carbon supports during the nucleation and growth process [22]. For example, if no carbon, the wormlike reverse micellar networks could be formed in the reaction solution, and Pt nanowire networks (NNWs) were prepared. If adding carbon, the carbon-supported spherical Pt nanoparticles were prepared due to only forming the spherical reverse micellar absorbed on carbon [34,91]. Besides, the organic molecules in the reaction system, for example, the organic ligands of metal precursors, organic solvent, or organic structure-capping agents, may strongly absorb carbon support, which are difficult to be eliminated, making the postsynthesis treatment hard. Therefore carbon-supported nonspherical Pt-based electrocatalysts commonly lack effective one-pot synthesis methods as their preparation often requires various organic structure-capping agents to control the shape or usually needs to be carried out in an organic solvent.

5.3.2

Ex situ mixing method

In the ex situ mixing method, Pt-based electrocatalysts are loaded onto carbon by physically mixing the preprepared Pt-based electrocatalysts with carbon. For the ex situ mixing method, the size or shape of Pt-based nanostructures will be easily controlled without the disturbance of carbon [22,92]. However, the interaction between the Pt-based electrocatalyst and carbon is usually weak because of only physical adsorption on carbon, leading to relatively inferior stability. Moreover, the Pt-based nanostructures tend to aggregate on carbon caused by the hydrogen bonds between the residual surfactants on the nanostructure surface [22]. Unlike the one-pot synthesis method, it is relatively hard for the ex situ mixing method to prepare the carbon-supported Pt-based electrocatalysts with high metal loading. For the one-pot synthesis method, the metal loading on carbon could be easily increased by increasing the ratio of Pt precursor and carbon [34]. However, for ex situ mixing method, the Pt loading amount is barely increased with

138

Fuel Cells for Transportation

TABLE 5.3 Comparison of one-pot synthesis method with ex situ mixing method. Methods

Pros

Cons

One-pot

1. Proper for high metal loading on support 2. Promoting the dispersion of metal on support 3. Good stability—relatively strong metal
carbon interaction 4. Proper for mass production

Difficult to control the shape

Ex situ mixing

Easy to control the size and shape

1. Difficult for high metal loading on support 2. Bad dispersion of Pt-based electrocatalysts on carbon 3. Weak metal
support interaction 4. Improper for mass production

increasing the ratio of Pt-based electrocatalysts and carbon when it reaches a certain critical value [22]. Compared with ex situ mixing method, the one-pot method may be more appropriate for fabricating the carbon-supported Pt-based electrocatalysts with high metal loading, high dispersion, and good stability, and more appropriate for mass production of Pt-based electrocatalysts (Table 5.3).

5.4

Synthesis of Pt-based electrocatalysts

The Pt-based electrocatalysts are formed through the nucleation and growth process. In the nucleation process, the building blocks, for example, metal atoms, metal ion dimers, or trimers (i.e., Pt(II)
Pt(I) or Pt(I)
Pt(I) dimers [93,94]), aggregate into nuclei (small clusters). The nuclei aggregate further and form seeds once over a specific size. In the growth process, seeds gradually grow into the final nanocrystals. The eventual size, size distribution, and shape of Pt-based electrocatalysts are governed by the nucleation and growth processes, experimentally depending on the types and concentration of metal precursors, reducing agents, and structure-directing agents, as well as the solvents, and reaction conditions, to name a few. During the nucleation and growth processes, the concentration of the building blocks changes with the reaction time, which is often presented in a classical Lamer’s plot [95]. As exhibited in Fig. 5.5A, when the concentration of the building blocks is larger than the nucleation threshold, nuclei or seeds are formed, and then, the concentration of the building blocks drops quickly below

Low platinum-based electrocatalysts for fuel cells Chapter | 5

139

FIGURE 5.5 (A) Plot of the concentration of building blocks against the reaction time (B) the nucleation stage overlaps with the growth stage, leading to different growth period for the seeds produced at different times; (C) the nucleation and growth stages are completely separated, causing nearly the same growth period for all the seeds [102]. (A) Adapted with permission from ref. V. K. LaMer and R. H. Dinegar, Theory, production and mechanism of formation of monodispersed hydrosols. J. Am. Chem. Soc. 72 (1950) 4847
4854. Copyright 1950 American Chemical Society; with permission from Springer.

the nucleation threshold, inhibiting additional nucleation events. Therefore short burst nucleation or a large number of seeds formed in a short time and then gradual growth can yield small and uniform nanocrystals, owing to nearly the same growth period for all seeds. In other words, the nucleation and growth process are nearly separated or only have a small overlap region, as shown in Fig. 5.5B
C. In contrast, a slow, continuous nucleation over the entire reaction time will result in a broad size distribution due to variations in the growth period for the seeds produced at different times. In the real synthesis process, a short burst nucleation is commonly realized by changing the type or improving the concentration of Pt precursors [42,96], using reducing agents with strong reducing capacity [97], improving the concentration of reducing agents [98,99], improving reaction temperature [92], varying the solvent [100], and increasing the number of nucleation sites on carbon supports for one-pot synthesis method [101]. However, burst nucleation will consume most precursors, likely leading to the formation of less shape-controlled Pt nanocrystals due to inadequate feedstock for growth [92]. The formed seeds may have a well-established shape, for example, single-crystal, singly twinned, multiply twinned, or stacking fault-lined structures, which will determine the shape of Pt-based electrocatalysts [103]. For example, a single-crystal seed could grow into octahedron, cuboctahedron, or cube; a multiply twinned seed could grow into decahedron or icosahedron [104]; a seed with stacking faults will grow into plate [105]. In addition to the seeds’ morphology, the final shape of Pt-based electrocatalysts mainly depends on the growth process. In the growth process, the deposited atom would diffuse around on the seed surface until it meets a low energy site. If the diffuse rate of atom is faster than its deposition rate, a thermodynamically stable nanocrystal would be formed. That is, the growth process occurs in thermodynamic regime (a.k.a., energy minimum principle).

140

Fuel Cells for Transportation

Oppositely, the growth process occurs in kinetic regime [42,106
109]. In thermodynamic control, the Pt-based nanocrystals would be covered by the facets with the lowest surface energy or have a minimum surface area. For Pt-based nanocrystals, the intrinsic surface energy of facets commonly follows {111} , {100} , {110} , high-index facets. The facet surface energy could be tuned by structure-directing agents as well, which will be explained later [110]. Therefore spherical nanocrystals or nanocrystals with exposed {111} facets (octahedra, decahedra, or icosahedra) are thermodynamically favorable products. In synthesis, the thermodynamically stable Pt-based nanocrystals can be formed under sufficiently high temperature or sufficiently long reaction time so that the deposited atoms could diffuse to the lowest energy sites. In the kinetic growth regime, if the growth rate (i.e., deposition rate) in the direction perpendicular to a facet is low, the facet will be retained finally; inversely, the facet will vanish. Therefore the seeds without a well-determined shape would grow into a specific shape in the kinetic growth regime [96,103]. In brief, the shape of Pt-based electrocatalyst should be a thermodynamically controlled product, while the growth process is carried out under kinetic control [111]. The various shaped Pt-based electrocatalysts, such as polyhedrons, nanocages, nanoframes, nanowires, or nanoplates, have been successfully prepared by selecting proper soft template agents or structure-capping agents [112]. Various shaped micelles or reverse micelles are formed by tuning the kind or content of surfactant or solvent, which can be used as the soft templates for the synthesis of shaped Pt-based electrocatalysts. Under the confined function of micelles or reverse micelles, Pt seed could grow into shaped nanocrystal [91,98]. Structure-capping agents include surfactant molecules, polymers, organic solvents, reducing agents, ions (halide ions), ligands of metal precursor or gas molecules (CO). They selectively adsorb on specific crystal facet since different facets have different atom arrangement and electronic structure, changing the surface energy of the facet and thus the relative growth rate along different directions [42,107,113,114]. Based on the density functional theory calculation, the surface energy (ϕ) of {100} facets is higher than {111} facets by 0.26 eV per atom for clean Pt surface. After co-adsorption of amine and CO (0.25 monolayer/0.25 monolayer), ϕ{100} is less than {111} facets by 0.02 eV per atom, indicating that {100} facets become more stable than {111} facets [113]. Therefore the shape-controlled synthesis of Pt-based electrocatalysts can be carried out by changing the metal precursors, capping agents, reducing agents, proper soft template agents, or solvents in a reaction system.

5.4.1

Synthesis of Pt-based spherical nanoparticles

Compared with nonspherical morphology, the synthesis systems of carbonsupported Pt-based spherical nanoparticles are usually less complex as some

Low platinum-based electrocatalysts for fuel cells Chapter | 5

141

structure-capping agents are commonly not needed, which favors one-pot synthesis [34,91,115,116]. The synthesis methods include chemical reduction in water [26,116
120], organic [including but not limited to ethylene glycol (EG) or benzyl ether] [115,121
126], or the mixed solution [34,90,101,127
130], electrochemical deposition [131], as well as impregnation and then H2 reduction [82,120,132
135]. In comparison with those in aqueous solution, nanoparticles synthesized in organic solvents commonly possess smaller size. However, it is usually difficult to completely remove the organic solvent molecules absorbed on carbon-supported Pt-based electrocatalysts. For example, to remove the absorbed EG, the electrocatalyst was treated at 300 C for 1 hour under Ar atmosphere [123]; to remove the absorbed oleylamine and oleic acid, the electrocatalyst was treated at 260 C in 20% O2/80% N2 for 90 minutes [97]. Heat treatment more or less causes the increase in particle size. Reactions in an aqueous solution do not have the problem, which would simplify the postsynthesis treatment process. In the reported synthesis procedures, size and dispersion on carbon of Pt-based spherical nanoparticles can be controlled by using functionalized carbon supports [26,101,117,119,125], varying the concentration of precursors or capping agent [97], using ligands (e.g., sodium citrate coordinates with Pt, Pd, or Ni ions, forming new complexes with changed reduction potential, beneficial for forming alloy with small size and good dispersion [121,136]), changing pH [119,121
123,137] or the solvent [90,129] (adding water into EG can accelerate the reduction of Pt21 by EG [138]), reducing metal confined in reverse micelles, controlling the reaction temperature [34], and varying the reducing capability of reducing agent [97].

5.4.2

Synthesis of Pt-based polyhedrons

Electrocatalytic reactions are structure-sensitive, and electrocatalytic activity commonly varies with the surface atom arrangements or the shapes of Ptbased electrocatalysts. For example, in HClO4 aq. (nonadsorbing electrolyte), for pure Pt, the order of ORR activity of low-index crystal facets follows Pt {110} . Pt{111} . Pt{100}, and for Pt3Ni, the order follows Pt3Ni{111} . Pt3Ni{110} . Pt3Ni{100}; in H2SO4 aq. (adsorbing electrolyte), the order follows Pt{110} . Pt{100} . Pt{111} due to the stronger adsorption of SO422 or HSO42 on Pt{111} facets than other facets; in KOH (adsorbing electrolyte), the order follows Pt{111} . Pt{110} . Pt{100} due to the different adsorption of OH2 on the different crystal facets [137,139]. Therefore the Pt-based polyhedrons often show higher activity and durability than the spherical nanoparticles [115,140,141], which have no precise facets on the surface. The Pt-based polyhedrons include {100} crystal facets closed nanocubes (NCs), {111} crystal facets closed nanotetrahedrons (NTs), nanooctahedrons (NOs), and nanoicosahedrons (NIs), or {100} and {111} crystal facets closed nanocuboctahedrons (a.k.a., nanotruncated octahedrons, NTOs). Moreover, from a geometric point of view, Pt-based polyhedrons would

142

Fuel Cells for Transportation

provide a larger surface area than spherical nanoparticles and thus higher ECSA when the mass and particle size is similar. Therefore the effects of synthesis, size, shape, and composition on electrocatalytic activity and durability of Pt-based polyhedrons have been widely studied [111,112]. In the synthesis process of Pt-based polyhedrons, there are several key experimental parameters responsible for the well-defined shape, high shape selectivity, good dispersion, small size as well as narrow size distribution by affecting the nucleation and growth process. These parameters include solvents, pH of the reaction solution [142,143], reductants, structure-directing agents, metal precursors, supports, addition sequence of reactants [140], reaction temperature, temperature ramp rate, and secondary species generated during the reaction process. In several reaction systems, a specific parameter can be ascribed to be the key factor for affecting size and shape. However, for most cases, the synergetic effect of various parameters in a reaction system is responsible for the well-controlled preparation of the polyhedrons [144]. The effects of experimental parameters on the formation of Pt-based polyhedrons in practice are as follows: 1. Solvents: The solvents commonly used to prepare Pt-based polyhedrons include H2O, benzyl alcohol, EG, 1-octadecene, oleic acid, oleylamine, benzyl ether, diphenyl ether, N,N-dimethylformamide, or the mixture of two or more (Fig. 5.6A). The solvents could affect the selection of structure-capping agents, metal precursors or reducing agents, and reaction conditions, due to the different boiling point, polarity, or viscosity, which will influence the physicochemical properties of solute in them or reaction condition, for example, solubility [110] or reaction temperature. Moreover, solvents may act as the reducing agent, assistant structuredirecting agent, or stabilizer as well. For example, oleylamine, N,Ndimethylformamide, benzyl alcohol, or EG usually act as the reducing agents [3,144
147]; oleylamine could help stabilize Pt3Ni {111} facets by lowering their surface energy [145] or absorb on the surface of Pt3Ni NCs to hinder their aggregation [110]. In addition, the solvents could coordinate with Pt ions, for example, oleylamine, changing and impacting the size of Pt-based polyhedrons [100,148]. For example, the size of Pt polyhedrons can be controlled readily by simply adjusting the amount of oleylamine in oleylamine/oleic acid systems. 2. Reductants (kind and concentration): The formation of Pt-based polyhedrons commonly need a low reducing rate, and thus, the weak reductants are often chosen, which are different from the synthesis of Pt-based spherical electrocatalysts that can choose some relatively strong reductants, for example, NaBH4 or ascorbic acid [149]. The reductants usually used to prepare Pt-based polyhedrons include some organic solvents as mentioned previously, formaldehyde (HCHO) [142], glucosamine [150], or poly(vinylpyrrolidone) (PVP, hydroxyl end group contributes to its

Low platinum-based electrocatalysts for fuel cells Chapter | 5

143

FIGURE 5.6 The molecule structure of several organic solvent or capping agents commonly used. (A) The organic solvents frequently used; (B) the organic capping agents commonly used for {100}-bound faceted Pt-based NCs; (C) the organic capping agents commonly used for {111}-bound faceted Pt-based polyhedrons (also includes TEAB or CTAB); (D) PVP usually used as the stabilizer.

reducing capability) [149]. Their type or amount will impact the shape of Pt-based polyhedrons. For example, if increasing HCHO solution (40 wt.%) from 0 to 0.005 and 0.010 mL, the shape selectivity of PtPd NIs will increase to above 80% due to the increased reducing rate; if further increasing the amount of HCHO solution, the formed nanoparticles would gradually change from multiple twinned particles to singlecrystalline ones with much smaller sizes; when employing 0.20 mL of

144

Fuel Cells for Transportation

HCHO solution (40 wt.%), about 5 nm PtPd nanotubes (NTs) will be achieved with a shape selectivity of above 80% [142]. 3. Structure-directing agents (a.k.a., structure-capping agents or capping agents; kind, molecular weight, alkyl chain lengths or concentration [110,149]): The structure-capping agents commonly used to synthesize Ptbased polyhedrons have one or more of CO gas, metal carbonyls (e.g., Fe (CO)5, Co2(CO)8, W(CO)6), halide (Cl2, Br2, I2), surfactants, polymers, as well as other organic molecules with oxygen-, nitrogen-, or sulfurcontaining groups (Fig. 5.6B
D). Among them, the capping agents probably used to prepare {100}-bound faceted Pt-based NCs include CO gas, CO or W0 (supplied by metal carbonyl), Ag species (supplied by AgNO3), Br2 (supplied by KBr), tetraethylammonium bromide (TEAB), CTAB, I2 (supplied by KI), cysteamine, sodium polyacrylate, or PVP (Fig. 5.6B and D) [92,96,110,113,140,144
147,151
156]. The capping agents commonly used to prepare {111}-bound faceted Pt-based NOs, NTs, or NIs include CO gas, metal carbonyls, Ag species (supplied by AgNO3), TEAB, Cl2 (supplied by NaCl), benzoic acid, CTAB, citric acid, poly(diallyldimethylammonium chloride) (PDDA), C2O422 (supplied by NaC2O4), decyltrimethylammonium bromide (DTAB), trisodium citrate, or PVP (Fig. 5.6C and D) [3,22,100,104,142,144,147
149,153,157
171]. Although some capping agents could stabilize both {100} facets and {111} facets, they tend to stabilize one kind of crystal facets in a specific reaction system or at a certain concentration. For example, when the concentration of Br2 supplied by TEAB is low, Br2 tends to coordinate with metal precursors, decreasing the reduction rate and forming the {111}-bound faceted NOs; when the concentration is high, Br2 would prefer to selectively adsorb on the {100} facets, forming NCs shape; when the concentration is medium, {111}- and {100}bound faceted NTOs are obtained [110]. 4. Metal precursors (kind or concentration): The valence state or ligands of Pt precursors [96,172], or the addition of non-Pt precursors [142,144,148], will also influence the size or shape of Pt-based polyhedrons. If (NH4)2PtCl6 is partially replaced with (NH4)2PtCl4 which is reduced more easily, the size of Pt-based electrocatalyst will be reduced due to much more nuclei formed [96]. Non-Pt metals commonly possess different standard reduction potentials, reduction kinetics, or adsorption properties of structure-capping agent from Pt, resulting in the formation of Pt-based alloys with different shapes from pure Pt. In the presence of Pd, well-defined PtPd NCs or NOs could be formed due to the coreduction of Pt and Pd which is induced by Pd nuclei formed in the very initial stage; otherwise, only uncontrollable Pt nanocrystals are obtained [144]. Although the adsorption of CO on Pt{100} facets is much stronger than on Pt{111} facets, PtNi NOs would be formed instead of NCs in the presence of sufficient Ni CO, because of the stronger adsorption of CO on Ni{111} facets than on Ni{100} facets [148].

Low platinum-based electrocatalysts for fuel cells Chapter | 5

145

5. Support: Although carbon may disturb the shape control as mentioned above [34,91], carbon supports could promote the formation of Pt-based polyhedrons sometimes. The functional groups on the carbon surface act as the binding sites for anchoring nanoparticles, thereby inducing anisotropic growth [146,166,173,174]. 6. Reaction temperature [92] and temperature ramp rate [170]: Reaction temperature or ramp rate during the heating process could affect the size or shape of Pt-based electrocatalysts by impacting the nucleation and growth rate. For example, if adding Fe(CO)5 at 180 C, small Pt nanoparticles (B3 nm) without special morphology were formed, as burst nucleation consumes most of the Pt precursors, resulting in insufficient feedstock in growth stage; if injecting Fe(CO)5 at 120 C, Pt NCs were formed in the presence of capping agents due to the sufficient feedstock; if adding Fe(CO)5 at 160 C, truncated Pt cubes were obtained [92]. 7. Secondary species: Secondary species, for example, new Pt complex [100,142,148,156] or capping agents [174], are the matters generated during the reaction instead of the matters added. The solvent or capping agent molecules may replace the initial ligands of Pt precursor due to the stronger interaction with Pt ions, which will change the redox potential of Pt and further impact the nucleation/growth processes. Through deliberately controlling the experimental parameters, plenty of well-defined Pt-based polyhedrons with improved performance and durability have been prepared successfully. However, most Pt-based polyhedrons have large sizes, sometimes larger than 10 nm [142,150,158,175], resulting in a low Pt utilization efficiency. Synthesizing 2
5 nm well-defined Pt-based polyhedrons is still a challenge. The mass production is another challenge for carbon-supported Pt-based polyhedrons face.

5.4.3

Synthesis of Pt-based open nanostructures

The solid Pt-based nanoparticles have a high proportion of Pt atoms in interior, especially Pt-based polyhedrons usually having a large size. However, the interior Pt atoms are blocked by surface atoms and cannot catalyze the reactions, resulting in the wastage of Pt atoms. Therefore the utilization efficiency of Pt atoms is commonly low (the utilization efficiency of Pt atoms is the fraction of Pt atoms catalyzing the reaction relative to total Pt atoms, which is related to ECSA) [176]. Compared to solid nanoparticles, open nanostructures could significantly decrease the consumption of Pt. Moreover, the open Pt-based nanostructures commonly show improved electrocatalytic activity, because of the simultaneous exposure of the external surface and internal surface [177,178]. The open nanostructures include nanocages with porous walls or nanoframes composed of only ridges. Unlike solid nanoparticles, if the wall thickness of nanocages or the ridge diameter of nanoframes

146

Fuel Cells for Transportation

is ultrathin, the open nanostructures could have a large size without sacrificing ECSA, which may improve the stability. The synthesis of open nanostructures primarily involves two stages: forming the solid nanoparticles (i.e., sacrificial template) and hollowing out interior atoms or both interior atoms and side face atoms. The second stage could be carried out through galvanic replacement reaction or chemical etching accompanied by Kirkendall effect. Kirkendall effect is a vacancy-mediated mechanism based on the observation of the out-diffusion of atoms inside the solid particle with vacancies diffusing inward and condensing in the interior of the particle, forming a hollow structure [179].

5.4.3.1 Galvanic replacement reaction For galvanic replacement reaction, sacrificial metallic template should have a lower reduction potential than Pt, which will drive the replacement of Pt ions with metal atoms [179
181]. As the surface template atoms are dissolved and Pt atoms are deposited, the interior template atoms would diffuse outward and vacancies would diffuse inward; that is, Kirkendall process takes place. The sacrificial templates could be pre-prepared first and then added into reaction solution containing Pt ions, for example, Pd, Ag, or Cu nanoparticles [182
187], or in situ formed in reaction system solution [188
200]. The former is relatively simple; however, the later requires carefully modulating the experimental parameters to make non-Pt metal deposit first. In the same coordination environment, Pd, Ni, or Cu would be reduced prior to Pt due to the more negative redox potential (Pd21/Pd 0.951 V vs SHE, Ni21/Ni 20.257 V vs SHE, Cu21/ Cu 0.3419 V vs SHE, Pt21/Pt 1.18 V vs SHE) [201]. Therefore the reduction kinetics or reaction potential of the metal precursors needs to be modulated, which could be realized by adding a sufficient amount of proper ligands, for example, halide ions or some N-containing organic molecules, because of the different stability of the newly formed metal complexes [189,191,193]. For example, in the presence of I2, Pd(acac)2 turns to PdI422 which is more favorable to be reduced by N,N-dimethylformamide than Pt(acac)2 [193]. Based on the cyclic voltammetry (CV) test, Pd21 species shows a more positive reduction potential than Pt21 species in the presence of I2 (KI) or Br2 (CTAB), leading to the preferential reduction of Pd21, as shown in Fig. 5.7 [191]. Similar results can be found in Lou’s work, and their study suggests that Br2 (CTAB) makes Cu21 deposit ahead of Pt21 likely due to the different reduction rates [189]. In addition to ligands, the metal precursor, reductant, or reaction temperature is also important for the preferential reduction of non-Pt ions [191
193]. For example, if using PtCl2 and PdCl2 as precursors, only solid PtPd alloy cubes were obtained even though in the presence of I2; however, if using Pt (acac)2 and Pd(acac)2 as precursors, the NCs would be achieved [193]. When using a strong reductant (e.g., ascorbic acid instead of N,N-dimethylformamide), the solid Pt-based alloy would be formed due to the co-reduction [193].

Low platinum-based electrocatalysts for fuel cells Chapter | 5

147

FIGURE 5.7 Cyclic voltammetry curves of the Ag electrode in H2PdCl4 (black) aq. or H2PtCl6 (red) aq. without any additives (A) (1), or with cetyltrimethylammonium chloride (CTAC) (A) (2), CTAB (A) (3), KI (A) (4), CTAC and CTAB (B) (1), CTAC and KI (B) (2), CTAB and KI (B) (3), and CTAC and CTAB and KI (B) (4) [191]. Reprinted by permission from Springer.

Moreover, the reaction temperature should be proper to form open structures, as a low diffusion rate of atoms at low reaction temperature would lead to a long reaction time or the co-reduction would take place at a relatively high reaction temperature caused by the enhanced reducing rate for all metal ions [193]. Therefore the combined influence of ligands, metal precursors, reducing agents as well as an appropriate temperature window will contribute to forming the hollow nanostructure instead of solid particles. In addition, accompanied by the galvanic replacement reaction, oxidative etching commonly occurs simultaneously during the formation of the open nanostructure due to the existence of O2 [180,199]. The reaction systems for preparing nanocages or nanoframes are commonly exposure to air without the protection of inert gas, such as N2 or Ar [186,187,191
193], which is often different from the case of the solid Pt alloys [4,110,145,158,202]. The presence of coordination ligands, such as halide ions or N-containing organic molecules introduced by metal precursor, solvent, or shape-capping agent, combined with O2, often results in the etching of non-Pt atoms. The sacrificial templates for Pt-based nanocages could be pre-prepared or in situ formed. However, the sacrificial templates for preparing nanoframes are primarily non-Pt metal polyhedrons in situ formed [194
197,200]. To form Pt-based nanoframes, Pt atoms should selectively deposit on vertexes and edges of Pt-based polyhedrons rather than the whole external surface [194,200]. Therefore the diffusion of Pt atoms on the side facets needs to be hindered.

148

Fuel Cells for Transportation

5.4.3.2 Chemical etching To prepare Pt-based open nanostructures, Pt-based alloys or core@Pt
shell (the core includes non-Pt core, Pt-poor core, or SiO2), as well as Pt-based polyhedrons with Pt-rich frameworks filled with Pt-poor alloy or non-Pt metal could be synthesized first and then chemically etching the non-Pt metals or the solid templates [16,177,186,203
210]. For example, if using PtNi alloy as precursors, the surface Ni atoms are etched, and the interior Pt and Ni atoms diffuse outward, finally forming Pt-based nanocages [16,210]. Chemical etching mainly refers to oxidative etching. Etchants for non-Pt metal include H1 from HNO3 or H2SO4 [16,186,203,210], Fe31 [206,207], CO [211], or dissolved O2 [178]. To accelerate the oxidative etching of nonPt metal, proper coordination ligands, such as halide ions or N-containing organic molecules, are usually added into the etching solution to stabilize dissolved metal ions or reduce the reduction potentials of metal ions. For example, when using Fe31 to etch Pd core, adding Br2 to stabilize Pd21 by Δ

21 forming PdBr422 (Pd 1 2Fe31 1 4Br2 - PdBr22 4 1 2Fe ) could facilitate the etching rate of Pd due to the decreased reduction potential (PdBr422/Pd 0.49 V vs SHE, Pd21/Pd 0.915 V vs SHE) [207]. In addition, if Fe31 is used as an etchant, acid, for example, HCl, is commonly required to prevent the hydrolysis of Fe31 and Fe21 ions [207]. In the presence of dissolved O2, Ni atoms are easier to be oxidized than Pt atoms. The oxidized Ni will turn into the soluble Ni-oleylamine complexes, increasing the dissolution rate of Ni [178]. Based on the Nernst equation, the pH of the reaction solution or the type of ligands will impact the etching process through influencing the redox potential. For example, the etching rate of Pd in the presence of Br2 is faster than the case of Pd in the presence of Cl2 due to the lower reduction potential (PdCl422/Pd 0.59 V vs SHE) [180]. Metals are more easily oxidized by O2 in acidic medium than in alkaline medium, because of the higher redox potential of O2 in acidic medium (1.229 V vs SHE) than in alkaline medium (0.401 V vs SHE). To synthesize Pt-based nanoframes by chemical etching, the phase segregation needs to occur to form the Pt-based polyhedrons with Pt-rich frameworks filled with Pt-poor alloy or non-Pt metal. In other words, Pt atoms need to be selectively deposited at vertexes and edges of Pt-poor or non-Pt metal polyhedrons or selectively migrate Pt atoms to vertexes and edges of Pt-based polyhedrons through modulating the experimental parameters [178,211
221]. Furthermore, there is a question, why could Pt selectively deposit at the edges instead of epitaxial growth? Two reasons might account for this. (1) The side faces are occupied by some species so that Pt has no choice but to deposit at edges. For example, Pt atoms could selectively deposit at edges of Pd-rich NCs in the presence of I2 at a relatively low reaction temperature, due to the deposition of Cu21 on side faces of Pd NCs [215]. (2) To minimize the total Gibbs energy or relieve internal strain arising from different atom size,

Low platinum-based electrocatalysts for fuel cells Chapter | 5

149

the element with lower surface energy or larger diameter would segregate to surface [213]. Taking {110}-closed rhombic dodecahedral PtNi with Pt-rich framework as an example, Ni preferentially segregates to faces due to the lower surface energies (Pt{110} and Ni{110} are 2.819 and 2.368 J m22, ˚ (Pt) and 1.97 A ˚ (Ni) respectively). The interior and larger Pt atoms (2.13 A [222]) tend to migrate to the vertex and edge sites of dodecahedron to release the internal strain [213]. As a large degree of lattice mismatch will be benefi˚ ), or Cu (1.96 A ˚ ) instead of Pd cial to phase segregation, Ni, Co (2.00 A ˚ (2.10 A) are commonly selected as the sacrificial metal for preparing Pt-based nanoframes [222]. In addition, to accelerate atoms migration, a relatively high reaction temperature or surface binding moieties are necessary to overcome the energy barrier. The former could increase the mobility of atoms [221]. Some surface binding moieties, for example, CO [212], could decrease the migration barrier energy of the inside Pt due to strong interaction with Pt.

5.4.4

Synthesis of 1D Pt-based nanostructures

Recently, 1D Pt-based electrocatalysts have attracted extensive attention due to the improved durability caused by the inherently anisotropic structure. For zero-dimensional (0D) Pt-based nanoparticles, the particle size would gradually increase due to the Oswald ripening process, decreasing the ECSA. However, for 1D Pt-based electrocatalysts, the dissolved Pt atoms would preferentially redeposit on the negative curvature sites to reduce the total energy, which will prevent the breakup of 1D structure and thus improve the stability (Fig. 5.8A) [223]. Furthermore, compared to the single point contact of Pt-based nanoparticles with carbon, multiple anchoring points of 1D Ptbased electrocatalysts nanostructure with carbon will strengthen the interaction between the metal and carbon support, not only improving electron transport but also suppressing the agglomeration and detachment of electrocatalysts (Fig. 5.8B and C) [141,224,225]. The 1D nanostructures include Pt-based nanowires (NWs), nanorods (NRs), nanofibers (NFs), NTs, or NNWs. The 1D Pt-based electrocatalysts, whose advantages and disadvantages are summarized in Table 5.4, could be prepared by the seed-mediated growth, oriented attachment of small nanocrystals, or 1D hard template, as described below. 1. Seed-mediated growth: Pt-based seeds could gradually grow into 1D nanostructure by confining growth in soft templates or under the control of structure-directing agents (such as HCOOH [226,227]). Soft templates mainly include the 1D reverse micelles formed by N-containing surfactants (such as dimethyldioctadecylammonium chloride [DDAC], hexadecyldimethylbenzyl ammonium chloride [HDBAC] [155], CTAB [228,229], didecyldimethylammonium bromide [DDAB] [141], CTAC, or octadecyltrimethylammonium chloride [STACB] [230]) in organic solvents (e.g.,

150

Fuel Cells for Transportation

FIGURE 5.8 (A) Ostwald ripening process of 0D Pt nanoparticles and 1D Pt NWs, the detachment or agglomeration of (B) 0D Pt nanoparticles, and (C) 1D Pt NWs supported on carbon.

TABLE 5.4 Comparison of three synthesis methods for the 1D Pt-based electrocatalysts. Methods

Pros

Cons

Seed-mediated growth

1. No necking area, more stable 2. 1D zig-zag-like nanostructures—abundant high-index facets—high activity




Oriented attachment of small nanocrystals




1. Fragile during ultrasonic dispersion due to the presence of neck between both attached particles 2. Easily forming network— adverse to supporting carbon

Hard template

Forming metal tube simultaneously exposing the internal and external surfaces

Removing the hard template commonly needs hash conditions

Low platinum-based electrocatalysts for fuel cells Chapter | 5

151

FIGURE 5.9 (A) Schematic of Pt seeds growing into NWs confined in the soft template formed with CTAB and oleylamine [242]; (B) schematic of PtRu NWs and PtRu NRs by using DDAC and HDBAC, respectively [155]; (C) HAADF-STEM image of PtNi NWs prepared by oriented attachment formation mechanism [242]; (D) high-resolution TEM image of an individual AgPt NWs in NNWs, inset: low-resolution TEM image of AgPt NNWs [236]. Adapted by permission from Springer.

oleylamine [229] or chloroform [91]) (Fig. 5.9A). The length and aspect ratio of 1D Pt-based electrocatalysts could be controlled by varying the concentration of metal precursors [91], the type or concentration of surfactant (e.g., surfactants with a different carbon number of alkyl chain), solvent, or reaction temperature [155,230]. For example, if HDBAC is used with only one alkyl chain, PtRu NRs were prepared, and the length of NRs decreases with an increase in the amount of HDBAC due to the shortened micelles; if DDAC is used with a double alkyl chain, PtRu NWs were formed due to the formation of long and stable micelles (Fig. 5.9B) [155]. The aspect ratio of PtSn NFs could change from 13.4 to 22.5 by modulating the type of surfactant (STAC or CTAC) and solvent (oleylamine or oleylamine/1-octadecene) [230]. In addition, zigzag-like PtM (M 5 Ni, Zn, Cu, Fe, or Co) NWs were prepared recently with high electrocatalytic activity, which commonly have abundant high-index facets or coordinatively unsaturated sites [224,231
235]. During the formation process of zigzag-like NWs, the smooth Pt NWs are formed prior to the deposition of non-Pt metal due to the relatively high reduction potentials of Pt, followed by the diffusion of non-Pt metal atoms into Pt lattice. The zigzag-like nanostructures are formed due to the different diffusion rates of Pt and non-Pt metal. The size of the humps can be controlled by varying the amount of non-Pt metal precursor. 2. Oriented attachment of small nanocrystals: The 1D Pt-based electrocatalysts could be obtained by continuous oriented attachment and coalescence of the Pt-based nanoparticles as well. Oriented attachment likely

152

Fuel Cells for Transportation

results from the confine function of the soft templates or different adsorption properties of shape-capping agents (such as H2, amine-terminated poly(N-isopropyl acrylamide) (PNIPAM-NH2), PVP, or Br2) on the different facets [38,225,236
238]. For example, PNIPAM-NH2 molecules more preferentially absorb on the {100} and {110} facets of Pt or PtAg nanoparticles than on {111} facets; therefore attachment preferably takes place on {111} surfaces, and 1D nanostructures will grow along ,111. orientation [236]. The 1D nanostructures formed by this method commonly show a rough surface or a wavy structure and uneven diameter due to the existence of necking area between nanoparticles and always exhibit a network-like structure composed of interwoven NWs (Fig. 5.9C and D) [236,238
240]. 3. Hard template: Pt-based NTs can be prepared by depositing Pt onto 1D templates and then removing the templates by chemical etching [241].

5.4.5

Synthesis of 2D Pt-based nanostructures

The 2D nanostructure, another anisotropic structure, has attracted much attention as well due to its unique physicochemical properties arising from the large lateral-to-thickness aspect ratio [243
245]. The 2D Pt-based electrocatalysts, for example, nanosheets (NSs) or nanoplates (NPs), often exhibit improved electrocatalytic activity and durability [243,246,247]. However, unlike intrinsically layered materials with strong covalent bond in a layer and weak van der Waals force between layers, for example, graphene, hexagonal boron nitride, or MoS2, Pt is face-centered cubic (fcc) structure, which is a highly symmetric crystal lattice and has strong nondirectional metallic bonding in 3D directions [248
250]. Therefore the formation of 2D-shaped Pt is thermodynamically unfavorable since a symmetry-breaking event takes place for anisotropic growth. To realize anisotropic growth of Pt-based nanomaterials, many strategies have been investigated, mainly including topdown approach and bottom-up approach. For the top-down approach, 2D Pt could be synthesized only by mechanical compression (e.g., heat-pressing process [251]), but not through exfoliation like intrinsically layered materials. For the bottom-up approach, 2D Pt-based nanostructures could be prepared by suppressing growth in the thickness direction and aiding lateral growth. The main bottom-up methods are summarized as follows: 1. 2D soft or hard templates: The 2D Pt-based nanostructures could be prepared by confining nucleation and growth in 2D soft templates, for example, 2D micelles or reverse micelles, or in/on 2D hard templates, for example, in the interlayer spaces of layered materials [245,252,253], or on the surface of 2D materials with surface function groups [254], or epitaxial growth on 2D metal [255,256]. For example, a gel-like material prepared with PVP and tris(hydroxymethyl)aminomethane in HCHO is

Low platinum-based electrocatalysts for fuel cells Chapter | 5

153

used as a soft template to form PtCu or PtAg NSs [105,257]. CTAB and sodium perfluorooctanoate (FC7) can assemble into bi-cellar-like micelles in water, which is used as the 2D soft template to prepare Pt nanodisks or nanowheels [252]. The hydrophilic quaternary ammonium head docosyltrimethylammonium chloride (C22TAC) could bind with PdCl422 or PtCl622 by electrostatic interaction, and then, long-chain hydrophobic C22 tail would self-assemble into a stabilized lamellar mesophase, leading to the formation of ultrathin 2D Pt-based nanostructures [247]. Due to immobilization function of GO with abundant oxygen-containing functional groups, Pt NSs could be formed by the self-assembly of Pt nanoparticles [173,254]. 2. Capping agents mediated growth: As mentioned above, capping agents could tune shape by specifical adsorption. It has been found that CO could suppress the growth in the thickness direction by strong absorption on basal {111} planes. Therefore Pt-based NSs could be successfully synthesized by introducing a second metal (such as Cu or Pd) to alter the preferential adsorption of CO on {100} facets to {111} facets [248,256,258,259]. 3. Hexagonal close-packed (hcp) Pt-based alloys: In addition to the effect of templates or capping agents, the final crystal shape also relies on its own crystal structure. For example, the fcc structured crystals often form the polyhedrons closed by {111} or {100} facets; the hcp structured crystals could form 2D nanostructure due to the uniaxial crystal structural character. Therefore 2D Pt-based alloys might be prepared by introducing the hcp structured metal, for example, Bi. PtBi NPs have been successfully synthesized by choosing Br2 as a capping agent, which preferably adsorbs on the (101) plane [246,260].

5.5

Postsynthesis treatments of Pt-based electrocatalysts

As mentioned in the last section, various ions or inorganic/organic molecules (hereafter called as residues) may absorb on or even bind to the surface of Pt-based electrocatalysts to control the size or shape during the preparation process. They might still exist on the surface of Pt-based electrocatalysts after the reaction ends [261], which could be demonstrated by IR spectroscopy, Raman spectroscopy, TGA, XPS, CV, and linear sweep voltammetry (LSV) [261,262]. These residues will physically (e.g., cover the active sites or block the contact of reactants with Pt-based electrocatalysts) or chemically (e.g., detrimentally reacting with reactants or intermediates) hinder the electrocatalytic reaction [263,264]. Hence, the prepared Pt-based electrocatalysts commonly need to be cleaned to remove the residues in order to expose the active sites before being applied in PEM fuel cell [264]. In general, the synthesized Pt-based electrocatalysts were first purified with a copious amount of solvent. The solvents frequently used include water, ethanol, acetone, toluene, isopropanol, hexane, cyclohexane, or some

154

Fuel Cells for Transportation

mixed solvents [161,166,235,242,248]. Suitable solvent should be chosen to remove specific residues due to the different solubility of various materials [265]. However, only the weakly attached residues can dissolve in solvent from the surface of Pt-based electrocatalysts. Some strongly adsorbed residues cannot be removed merely by washing with solvent [262]. Therefore to completely clean the surface of Pt-based electrocatalysts, further purification is commonly required. Generally used methods include but are not limited to acid washing, thermal annealing, UV
ozone irradiation, and electrochemical cleaning, as described below: 1. Acid washing: The amine-containing residues, for example, oleylamine or PVP, could be removed by washing with acid solution based on the acid
base reaction [263]. The electrocatalyst could be treated in pure acetic acid or its aqueous solution at about 60 C
70 C for at least 30 minutes to remove the covered amine-containing species [100,141,149,153,160,228]. It is worth being pointed out that a Pt or Pt-rich shell may be formed during the acid washing process as the surface and subsurface transition metal are leached out, which will improve the stability of Pt-based alloys in fuel cells [266]. 2. Thermal annealing: The organic residues could be degraded by annealing in air at 180 C
350 C for at least 1 hour [97,157,267
270], or evaporated by annealing under an inert atmosphere, for example, annealing under Ar atmosphere at 300 C for 1 hour to remove EG [123]. However, the high temperature frequently causes the size increase and shape modification to occur. Moreover, the heat treatment in the air would lead to the surface oxidation of Pt-based electrocatalysts. To recover the metallic surface, subsequent reduction is necessary before application, for example, heat treatment in a mixed gas of H2 and N2 [97,267,270]. 3. UV
ozone irradiation: UV
ozone irradiation cleaning is a photosensitized oxidation process [271]. The organic residues absorbed on electrocatalysts could be removed under the irradiation with B185 and B254 nm UV light at room temperature for a given period in air [92,272]. O2 molecules in air could absorb the B185 nm UV light and then dissociate, forming atomic oxygen or ozone; organic material could be excited and/or dissociated after absorbing B254 nm UV light. The excited organic species would be decomposed to simpler volatile molecules by atomic oxygen or ozone [271]. Compared with heat treatment, UV
ozone irradiation has less influence on the size and shape due to operation at room temperatures. Nevertheless, as light travels in straight lines, the shielded residues may not be eliminated completely. Similar to the thermal annealing method, the surface of Pt-based electrocatalysts would be partially oxidized by atomic oxygen or ozone; for example, the amount of Pdδ1 increases from 12% to 54% according to the XPS measurement after 4 hours of irradiation [271]. Moreover, some byproducts

Low platinum-based electrocatalysts for fuel cells Chapter | 5

155

produced by the decomposition of organic substance, for example, CO, a product of the decomposition of PVP during UV
ozone irradiation, still remain on the surface of electrocatalysts [271]. 4. Electrochemical cleaning [261,273]: During electrochemical cycles, Pt surface will be subsequently oxidized and reduction, that is, subsequently forming Pt-O bond at high potentials and Pt-H bond at low potentials (hydrogen evolution reaction occurs on Pt surface). The robust Pt-O or Pt-H bond has a high binding energy, which would displace the Pt 2 residue coordination bond (Pt-O: 82.3 kcal mol21, Pt-H: 75.1 kcal mol21, Pt-oleylamine: 23.1 kcal mol21, Pt-dodecanethiol: 17.7 kcal mol21, or Pt-triphenylphosphine: 38.1 kcal mol21) [273]. Besides, the oxidative decomposition of the organic residues occurs at high potentials simultaneously. Therefore the active sites of electrocatalysts would be gradually exposed as the electrochemical cycle number is increased. 5. Other methods: The organic residues on Pt3Ni NCs could be removed through Ar plasma treatment [145]. When Pt nanoparticles suspension solution is mixed with H2O2/H2SO4 aq. and subsequent centrifugation, PVP is removed from the Pt surface by O2 bubbles produced from the decomposition of H2O2 on Pt surface [274]. In addition, PVP can also be removed from PtPd NCs by NaBH4/tert-butylamine (TBA) treatment [264]. The hydride generated by the hydrolysis of NaBH4 would displace PVP, resulting in PVP desorption. The desorbed PVP dissolves in TBA, an organic solvent with high solubility for PVP. In brief, it is difficult to develop a universal cleaning procedure, since the different residues have different interactions with Pt-based electrocatalysts, which is determined by the chemical nature of ligands, the nature of metal, and atomic arrangement on the surface [275]. Therefore the Pt-based electrocatalysts synthesized in various reaction systems need to be cleaned by different methods based on the practical case. Besides, the postsynthesis treatment may impact the size, size distribution, shape, composition, or construction of surface atoms, which will decrease or improve the performance and durability of electrocatalyst; the former needs to be avoided. The postsynthesis treatment method should be simple, timesaving, inexpensive, and environment-friendly in terms of scale-up production.

5.6

Future direction and prospects

To improve the activity and durability of Pt-based electrocatalysts, significant progress has been achieved in the controlled synthesis of carbonsupported Pt-based electrocatalysts. However, most synthesis is limited at the laboratory level. Much more effort toward experimental research and development is still needed to commercialize the carbon-supported Pt-based electrocatalysts with high activity and durability.

156

5.6.1

Fuel Cells for Transportation

Studies on functionalization of commercial carbon supports

Functionalization of commercial carbon supports not only could significantly increase the number of nucleation sites or anchoring sites to improve the dispersion of Pt-based electrocatalysts on them, make size control easier, or strengthen metal
support interaction but also supply additional nonmetal active sites. Therefore Pt-based electrocatalysts supported on the functionalized carbon commonly show higher activity and durability than those supported on pristine carbon. Some perspectives on the existing challenges and future research directions of the functionalization of commercial carbon supports are described here: 1. To increase the nucleation sites and improve the hydrophilicity of carbon, the positively or negatively charged groups could be introduced onto its surface. No matter which charged groups are introduced, the number of nucleation sites and hydrophilicity of carbon would be enhanced, and the size or dispersion of the Pt-based electrocatalysts could be optimized by selecting the Pt precursor with the opposite charge. However, given that the electrostatic interaction between the groups on carbon and sulfonate groups in ionomer, the presence of positively charged groups on carbon surface could facilitate the uniform distribution of ionomer on the carbon-supported Pt-based electrocatalysts when making catalyst layers. Further, for the prepared Pt/C (e.g., commercial Pt/C), introducing positively charged groups onto carbon of Pt/C might be an effective way to make ionomer distributed uniformly on Pt/C. However, it is worth noting that the process introducing positively charged groups onto Pt/C should not influence the size, shape, and activity of Pt nanoparticles. 2. The amount of heteroatoms doped in commercial carbon supports is still low, commonly less than 10 at% and even less than 1 at% in some cases, limiting the effect of heteroatoms on the performance improvement of Pt-based electrocatalysts to some extent. Therefore it still needs to be investigated how to increase the doping amount of heteroatoms. 3. To maximize the potential advantages of carbon
metal oxide composite supports and decrease the effect on the electron transport between the Ptbased electrocatalysts and carbon support, the size of the metal oxide nanoparticles introduced should be sufficiently small, which is still a challenge. Therefore it is necessary developing new methods to decrease the size of metal oxide nanoparticles deposited on carbon support. In addition to introducing metal oxides, introducing other noncarbon materials (e.g., nitrides, carbides) might be an effective way to functionalize carbon. Forming a porous ultrathin metal oxide layer outside carbon supports instead of nanoparticles probably maximizes the potential advantages of composite supports.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

5.6.2

157

Production of Pt-based electrocatalysts

Although plenty of nonspherical Pt-based electrocatalysts with improved activity and durability have been synthesized, commercial electrocatalysts are still carbon-supported spherical Pt-based electrocatalysts until now. More efforts need to be devoted to the mass production of well-defined carbonsupported nonspherical Pt-based electrocatalysts. Here, some probable reasons influencing mass production capability are presented, which are expected to help accelerate this process: 1. If the production scale is increased only through scaling up the reaction system used in the laboratory, the obtained nonspherical Pt-based electrocatalysts commonly lose the control of size and shape and thus show a worse performance and durability than those prepared in a small batch. Compared with spherical nanoparticles, the reaction systems for preparing nonspherical Pt-based electrocatalysts are generally more complex, which heightens the difficulty of keeping everywhere in the scale-up reaction system uniform and consistent with the lab-scale system. Besides, for nonspherical Pt-based electrocatalysts, it is commonly necessary using various shape-capping agents to control the shape, improving the difficulty of postsynthesis treatment. The mechanisms of how the experimental parameters affect the size and shape of nonspherical Ptbased electrocatalysts are still not well understood and indistinct, and thus, it is impossible to effectively control the size and shape when scaling up for mass production. Hence, it is necessary for the mass production of nonspherical Pt-based electrocatalysts to develop simple reaction systems and understand the influencing mechanisms of every experimental parameter on the quality of the electrocatalysts synthesized. 2. Although the size and shape could be relatively easily controlled for ex situ mixing method, the one-pot synthesis method is more suitable for cost-effective mass production. In comparison with carbon-supported spherical Pt-based electrocatalysts, it is harder to prepare well-defined carbon-supported nonspherical Pt-based electrocatalysts by the one-pot method since carbon would disturb shape control. Therefore for mass production, significantly more efforts should be made to synthesize carbon-supported nonspherical Pt-based electrocatalysts by one-pot synthesis method instead of ex situ mixing method. 3. It is worth mentioning that most reported nonspherical Pt-based electrocatalysts are prepared in organic solution likely because the size or shape is more easily controlled than in aqueous solution. However, most organic solvents are expensive, toxic, and environmentally unfriendly. Further, the reactions in organic solvents often require a relatively high reaction temperature, which increases the production cost of electrocatalysts and makes it more difficult to maintain a uniform reaction

158

Fuel Cells for Transportation

temperature in a scale-up reaction system. Organic solvent molecules tend to strongly absorb on the surface of Pt-based electrocatalysts or carbon, which makes postsynthesis treatment more challenging. Taking economic and environmental considerations into account, aqueous reaction systems may be more suitable for mass production. Therefore future experimental research should be focused on how to effectively control the size and shape of Pt-based electrocatalysts in aqueous solutions.

5.6.3

Postsynthesis treatment

Although various structure-directing agents are commonly unavoidable to control the size and shape of Pt-based electrocatalysts, to their removal from the electrocatalysts surface should be paid more attention for electrocatalytic application. It has been found that not only the synthesis process is crucial for the performance of Pt-based electrocatalysts, but the postsynthesis treatments have an important influence as well. Therefore simple, inexpensive, environment-friendly, reliable, and controlled postsynthesis treatment methods need to be developed, and during the synthesis process, the amount of used ligand should be minimized or trying to adopt the ligands easily removed. Taking the above challenges into consideration, the size and shape of carbon-supported Pt-based electrocatalysts in aqueous systems are suggested to be effectively controlled by (1) using functionalized carbon supports instead of pristine carbon, (2) modifying the pH, reaction temperature, or reduction capability of reductants to control the nucleation and growth process, (3) using some structure-directing agents with small molecular, for example, halide ion and CO, which are easy to be removed, or (4) selecting the structure-directing agents with the same terminal group of -SO3H as perflorosulfonate acid ionomer, for example, sodium dodecyl sulfonate, the residues of which on electrocatalyst surface probably could contribute to the proton transport in PEM fuel cells.

References [1] J.K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J.R. Kitchin, T. Bligaard, et al., Origin of the overpotential for oxygen reduction at a fuel-cell cathode, J. Phys. Chem. B 108 (2004) 17886
17892. [2] J. Hou, M. Yang, C. Ke, G. Wei, C. Priest, Z. Qiao, et al., Platinum-group-metal catalysts for proton exchange membrane fuel cells: from catalyst design to electrode structure optimization, EnergyChem 2 (2020) 100023. [3] X. Huang, Z. Zhao, Y. Chen, E. Zhu, M. Li, X. Duan, et al., A rational design of carbonsupported dispersive Pt-based octahedra as efficient oxygen reduction reaction catalysts, Energy Environ. Sci. 7 (2014) 2957
2962. [4] R.B. Moghaddam, S. Shahgaldi, X. Li, A facile synthesis of high activity cube-like Pt/carbon composites for fuel cell application, Front. Energy 11 (2017) 245
253.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

159

[5] J. Li, Y. Song, G. Zhang, H. Liu, Y. Wang, S. Sun, et al., Pyrolysis of self-assembled iron porphyrin on carbon black as core/shell structured electrocatalysts for highly efficient oxygen reduction in both alkaline and acidic medium, Adv. Funct. Mater. 27 (2017) 1604356. [6] Y. Xie, Q. Zhong, Y. Lv, J. Li, Z. Hao, C. Tang, et al., Interwoven molecular chains obtained by ionic self-assembly of two iron(iii) porphyrins with opposite and mismatched charges, ACS Appl. Mater. Interfaces 11 (2019) 34203
34211. [7] J. Wang, H. Kong, J. Zhang, Y. Hao, Z. Shao, F. Ciucci, Carbon-based electrocatalysts for sustainable energy applications, Prog. Mater. Sci. 116 (2021) 100717. [8] V. Yarlagadda, M.K. Carpenter, T.E. Moylan, R.S. Kukreja, R. Koestner, W. Gu, et al., Boosting fuel cell performance with accessible carbon mesopores, ACS Energy Lett. 3 (2018) 618
621. [9] X. Yu, S. Ye, Recent advances in activity and durability enhancement of Pt/C catalytic cathode in PEMFC: part I. Physico-chemical and electronic interaction between Pt and carbon support, and activity enhancement of Pt/C catalyst, J. Power Sources 172 (2007) 133
144. [10] https://www.cheaptubes.com/. [11] http://www.carbon-nano.com/goldtube-265/goldtube-265. [12] https://www.ssnano.com/nanotubes___graphenes. [13] https://www.sigmaaldrich.com/catalog/search? term 5 carbon 1 nanotube&interface 5 All&N 5 0&mode 5 match% 20partialmax&lang 5 en®ion 5 CA&focus 5 product. [14] https://graphene-star.com/products. [15] https://www.sigmaaldrich.com/catalog/substance/graphenenanoplatelets1201778242511? lang 5 en®ion 5 CA. [16] T. Asset, N. Job, Y. Busby, A. Crisci, V. Martin, V. Stergiopoulos, et al., Porous hollow PtNi/C electrocatalysts: carbon support considerations to meet performance and stability requirements, ACS Catal. 8 (2018) 893
903. [17] V. Comignani, J.M. Sieben, M.D. Sanchez, M.M. Duarte, Influence of carbon support properties on the electrocatalytic activity of PtRuCu nanoparticles for methanol and ethanol oxidation, Int. J. Hydrog. Energy 42 (2017) 24785
24796. [18] L.I. Sanlı, ¸ V. Bayram, S. Ghobadi, N. D¨uzen, S.A. G¨ursel, Engineered catalyst layer design with graphene-carbon black hybrid supports for enhanced platinum utilization in PEM fuel cell, Int. J. Hydrog. Energy 42 (2017) 1085
1092. [19] A. Bharti, G. Cheruvally, Influence of various carbon nano-forms as supports for Pt catalyst on proton exchange membrane fuel cell performance, J. Power Sources 360 (2017) 196
205. [20] T. Ito, U. Matsuwaki, Y. Otsuka, M. Hatta, K. Hayakawa, K. Matsutani, et al., Threedimensional spatial distributions of Pt catalyst nanoparticles on carbon substrates in polymer electrolyte fuel cells, Electrochemistry 79 (2011) 374
376. [21] B. Wu, C. Wang, Y. Cui, L. Mao, S. Xiong, Tailoring carbon nanotubes surface with maleic anhydride for highly dispersed PtRu nanoparticles and their electrocatalytic oxidation of methanol, RSC Adv. 5 (2015) 16986
16992. [22] K.Y. Cho, Y.S. Yeom, H.Y. Seo, P. Kumar, A.S. Lee, K.-Y. Baek, et al., Molybdenumdoped PdPt@ Pt core
shell octahedra supported by ionic block copolymer-functionalized graphene as a highly active and durable oxygen reduction electrocatalyst, ACS Appl. Mater. Interfaces 9 (2017) 1524
1535. [23] F. Rodriguez-Reinoso, The role of carbon materials in heterogeneous catalysis, Carbon 36 (1998) 159
175.

160

Fuel Cells for Transportation

[24] T. Saito, K. Matsushige, K. Tanaka, Chemical treatment and modification of multi-walled carbon nanotubes, Phys. B Condens. Matter 323 (2002) 280
283. [25] Y. Cheng, C. Xu, P.K. Shen, Effect of nitrogen-containing functionalization on the electrocatalytic activity of PtRu nanoparticles supported on carbon nanotubes for direct methanol fuel cells, Appl. Catal. B: Environ. 158 (2014) 140
149. [26] B. Wu, J. Zhu, X. Li, X. Wang, J. Chu, S. Xiong, PtRu nanoparticles supported on p-phenylenediamine-functionalized multiwalled carbon nanotubes: enhanced activity and stability for methanol oxidation, Ionics 25 (2019) 181
189. [27] B.T. Sneed, D.A. Cullen, K.S. Reeves, O.E. Dyck, D.A. Langlois, R. Mukundan, et al., 3D analysis of fuel cell electrocatalyst degradation on alternate carbon supports, ACS Appl. Mater. Interfaces 9 (2017) 29839
29848. [28] X. Zhao, T. Gunji, T. Kaneko, S. Takao, T. Sakata, K. Higashi, et al., Evidence for interfacial geometric interactions at metal
support interfaces and their influence on the electroactivity and stability of Pt nanoparticles, J. Mater. Chem. A 8 (2020) 1368
1377. [29] S. Samad, K.S. Loh, W.Y. Wong, T.K. Lee, J. Sunarso, S.T. Chong, et al., Carbon and non-carbon support materials for platinum-based catalysts in fuel cells, Int. J. Hydrog. Energy 43 (2018) 7823
7854. [30] Y. Shao, G. Yin, Y. Gao, P. Shi, Durability study of Pt/C and Pt/CNTs catalysts under simulated PEM fuel cell conditions, J. Electrochem. Soc. 153 (2006) A1093. [31] M. Chen, S. Hwang, J. Li, S. Karakalos, K. Chen, Y. He, et al., Pt alloy nanoparticles decorated on large-size nitrogen-doped graphene tubes for highly stable oxygen-reduction catalysts, Nanoscale 10 (2018) 17318
17326. [32] F. Yang, L. Xin, A. Uzunoglu, Y. Qiu, L. Stanciu, J. Ilavsky, et al., Investigation of the interaction between nafion ionomer and surface functionalized carbon black using both ultrasmall angle X-ray scattering and cryo-TEM, ACS Appl. Mater. Interfaces 9 (2017) 6530
6538. [33] A. Orfanidi, P. Madkikar, H.A. El-Sayed, G.S. Harzer, T. Kratky, H.A. Gasteiger, The key to high performance low Pt loaded electrodes, J. Electrochem. Soc. 164 (2017) F418
F426. [34] S. Li, H. Liu, Y. Wang, W. Xu, J. Li, Y. Liu, et al., Controlled synthesis of high metalloading, Pt-based electrocatalysts with enhanced activity and durability toward oxygen reduction reaction, RSC Adv. 5 (2015) 8787
8792. [35] J. Li, J.-X. Liu, X. Gao, B.R. Goldsmith, Y. Cong, Z. Zhai, et al., Nitrogen-doped graphene layers for electrochemical oxygen reduction reaction boosted by lattice strain, J. Catal. 378 (2019) 113
120. [36] S. Ott, A. Orfanidi, H. Schmies, B. Anke, H.N. Nong, J. H¨ubner, et al., Ionomer distribution control in porous carbon-supported catalyst layers for high-power and low Pt-loaded proton exchange membrane fuel cells, Nat. Mater. 19 (2020) 77
85. [37] A. Solhy, B. Machado, J. Beausoleil, Y. Kihn, F. Gonc¸alves, M. Pereira, et al., MWCNT activation and its influence on the catalytic performance of Pt/MWCNT catalysts for selective hydrogenation, Carbon 46 (2008) 1194
1207. [38] M.A. Hoque, F.M. Hassan, A.M. Jauhar, G. Jiang, M. Pritzker, J.-Y. Choi, et al., Weblike 3D architecture of Pt nanowires and sulfur-doped carbon nanotube with superior electrocatalytic performance, ACS Sustain. Chem. Eng. 6 (2018) 93
98. [39] R. Rizo, D. Sebastia´n, J.L. Rodrı´guez, M.J. La´zaro, E. Pastor, Influence of the nature of the carbon support on the activity of Pt/C catalysts for ethanol and carbon monoxide oxidation, J. Catal. 348 (2017) 22
28.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

161

[40] S. Dai, J.-P. Chou, K.-W. Wang, Y.-Y. Hsu, A. Hu, X. Pan, et al., Platinum-trimer decorated cobalt-palladium core-shell nanocatalyst with promising performance for oxygen reduction reaction, Nat. Commun. 10 (2019) 440. [41] M. Laurent-Brocq, N. Job, D. Eskenazi, J.-J. Pireaux, Pt/C catalyst for PEM fuel cells: control of Pt nanoparticles characteristics through a novel plasma deposition method, Appl. Catal. B: Environ. 147 (2014) 453
463. [42] Y.-J. Wang, N. Zhao, B. Fang, H. Li, X.T. Bi, H. Wang, Carbon-supported Pt-based alloy electrocatalysts for the oxygen reduction reaction in polymer electrolyte membrane fuel cells: particle size, shape, and composition manipulation and their impact to activity, Chem. Rev. 115 (2015) 3433
3467. [43] K. Artyushkova, S. Pylypenko, M. Dowlapalli, P. Atanassov, Structure-to-property relationships in fuel cell catalyst supports: correlation of surface chemistry and morphology with oxidation resistance of carbon blacks, J. Power Sources 214 (2012) 303
313. [44] L. Castanheira, W.O. Silva, F.H.B. Lima, A. Crisci, L. Dubau, F. Maillard, Carbon corrosion in proton-exchange membrane fuel cells: effect of the carbon structure, the degradation protocol, and the gas atmosphere, ACS Catal. 5 (2015) 2184
2194. [45] Y. Tian, Y.-j Liu, J.-x Zhao, Y.-h Ding, High stability and superior catalytic reactivity of nitrogen-doped graphene supporting Pt nanoparticles as a catalyst for the oxygen reduction reaction: a density functional theory study, RSC Adv. 5 (2015) 34070
34077. [46] C.R. Raj, A. Samanta, S.H. Noh, S. Mondal, T. Okajima, T. Ohsaka, Emerging new generation electrocatalysts for the oxygen reduction reaction, J. Mater. Chem. A 4 (2016) 11156
11178. [47] K. Gao, B. Wang, L. Tao, B.V. Cunning, Z. Zhang, S. Wang, et al., Efficient metal-free electrocatalysts from n-doped carbon nanomaterials: mono-doping and co-doping, Adv. Mater. 31 (2019) 1805121. [48] J.-J. Fan, Y.-J. Fan, R.-X. Wang, S. Xiang, H.-G. Tang, S.-G. Sun, A novel strategy for the synthesis of sulfur-doped carbon nanotubes as a highly efficient Pt catalyst support toward the methanol oxidation reaction, J. Mater. Chem. A 5 (2017) 19467
19475. [49] Y. Sun, C. Du, M. An, L. Du, Q. Tan, C. Liu, et al., Boron-doped graphene as promising support for platinum catalyst with superior activity towards the methanol electrooxidation reaction, J. Power Sources 300 (2015) 245
253. [50] H.L. Tan, A. Du, R. Amal, Y.H. Ng, Decorating platinum on nitrogen-doped graphene sheets: control of the platinum particle size distribution for improved photocatalytic H2 generation, Chem. Eng. Sci. 194 (2019) 85
93. [51] H.N. Yang, D.C. Lee, K.W. Park, W.J. Kim, Platinum
boron doped graphene intercalated by carbon black for cathode catalyst in proton exchange membrane fuel cell, Energy 89 (2015) 500
510. [52] S. Mao, C. Wang, Y. Wang, The chemical nature of N doping on N doped carbon supported noble metal catalysts, J. Catal. 375 (2019) 456
465. [53] O.L. Li, Z. Shi, H. Lee, T. Ishizaki, Enhanced electrocatalytic stability of platinum nanoparticles supported on sulfur-doped carbon using in-situ solution plasma, Sci. Rep. 9 (2019) 12704. -12704. [54] W. Wang, Y. Liu, W. Dong, Y. Yang, X. Zhou, Z. Lei, Pd3Ni nanoparticles combines carbonized 1,10-phenanthroline modified carbon support: a highly efficient electrocatalyst for enhanced methanol oxidation, Int. J. Hydrog. Energy 40 (2015) 3892
3899. [55] J.M. Yoo, C.-Y. Ahn, J. Lee, Y.-E. Sung, Effect of different surface functional groups on carbon supports toward methanol electro-oxidation of Pt nanoparticles, J. Electroanal. Chem. 875 (2020) 113931.

162

Fuel Cells for Transportation

[56] Y. Cong, I.T. McCrum, X. Gao, Y. Lv, S. Miao, Z. Shao, et al., Uniform Pd0.33Ir0.67 nanoparticles supported on nitrogen-doped carbon with remarkable activity toward the alkaline hydrogen oxidation reaction, J. Mater. Chem. A 7 (2019) 3161
3169. [57] M. Park, T. Lee, B.-S. Kim, Covalent functionalization based heteroatom doped graphene nanosheet as a metal-free electrocatalyst for oxygen reduction reaction, Nanoscale 5 (2013) 12255
12260. [58] H.-P. Cong, P. Wang, M. Gong, S.-H. Yu, Facile synthesis of mesoporous nitrogen-doped graphene: an efficient methanol
tolerant cathodic catalyst for oxygen reduction reaction, Nano Energy 3 (2014) 55
63. [59] K. Jukk, N. Kongi, P. Rauwel, L. Matisen, K. Tammeveski, Platinum nanoparticles supported on nitrogen-doped graphene nanosheets as electrocatalysts for oxygen reduction reaction, Electrocatalysis 7 (2016) 428
440. [60] X. Ning, X. Zhou, J. Luo, L. Ma, X. Xu, L. Zhan, Glycerol and formic acid electrooxidation over Pt on S-doped carbon nanotubes: effect of carbon support and synthesis method on the metal-support interaction, Electrochim. Acta 319 (2019) 129
137. [61] D. Higgins, M.A. Hoque, M.H. Seo, R. Wang, F. Hassan, J.-Y. Choi, et al., Development and simulation of sulfur-doped graphene supported platinum with exemplary stability and activity towards oxygen reduction, Adv. Funct. Mater. 24 (2014) 4325
4336. [62] Z. Yang, Z. Yao, G. Li, G. Fang, H. Nie, Z. Liu, et al., Sulfur-doped graphene as an efficient metal-free cathode catalyst for oxygen reduction, ACS Nano 6 (2012) 205
211. [63] I.-Y. Jeon, S. Zhang, L. Zhang, H.-J. Choi, J.-M. Seo, Z. Xia, et al., Edge-selectively sulfurized graphene nanoplatelets as efficient metal-free electrocatalysts for oxygen reduction reaction: the electron spin effect, Adv. Mater. 25 (2013) 6138
6145. [64] A. Pullamsetty, R. Sundara, Investigation of catalytic activity towards oxygen reduction reaction of Pt dispersed on boron doped graphene in acid medium, J. Colloid Interface Sci. 479 (2016) 260
270. [65] A. Pullamsetty, M. Subbiah, R. Sundara, Platinum on boron doped graphene as cathode electrocatalyst for proton exchange membrane fuel cells, Int. J. Hydrog. Energy 40 (2015) 10251
10261. [66] J.-e Park, Y.J. Jang, Y.J. Kim, M.-s Song, S. Yoon, D.H. Kim, et al., Sulfur-doped graphene as a potential alternative metal-free electrocatalyst and Pt-catalyst supporting material for oxygen reduction reaction, Phys. Chem. Chem. Phys. 16 (2014) 103
109. [67] M. Seredych, J.-C. Idrobo, T.J. Bandosz, Effect of confined space reduction of graphite oxide followed by sulfur doping on oxygen reduction reaction in neutral electrolyte, J. Mater. Chem. A 1 (2013) 7059
7067. [68] L. Li, H. Liu, L. Wang, S. Yue, X. Tong, T. Zaliznyak, et al., Chemical strategies for enhancing activity and charge transfer in ultrathin Pt nanowires immobilized onto nanotube supports for the oxygen reduction reaction, ACS Appl. Mater. Interfaces 8 (2016) 34280
34294. [69] M. Huang, J. Zhang, C. Wu, L. Guan, Networks of connected Pt nanoparticles supported on carbon nanotubes as superior catalysts for methanol electrooxidation, J. Power Sources 342 (2017) 273
278. [70] M.-H. Hsu, H. Chuang, F.-Y. Cheng, Y.-P. Huang, C.-C. Han, K.-C. Pao, et al., Simple and highly efficient direct thiolation of the surface of carbon nanotubes, RSC Adv. 4 (2014) 14777
14780. [71] L. Xin, F. Yang, S. Rasouli, Y. Qiu, Z.-F. Li, A. Uzunoglu, et al., Understanding Pt nanoparticle anchoring on graphene supports through surface functionalization, ACS Catal. 6 (2016) 2642
2653.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

163

[72] D. Park, M.S. Ahmed, S. Jeon, Covalent functionalization of graphene with 1, 5diaminonaphthalene and ultrasmall palladium nanoparticles for electrocatalytic oxygen reduction, Int. J. Hydrog. Energy 42 (2017) 2061
2070. [73] H.-S. Oh, H. Kim, Efficient synthesis of Pt nanoparticles supported on hydrophobic graphitized carbon nanofibers for electrocatalysts using noncovalent functionalization, Adv. Funct. Mater. 21 (2011) 3954
3960. [74] X.-C. Liu, G.-C. Wang, R.-P. Liang, L. Shi, J.-D. Qiu, Environment-friendly facile synthesis of Pt nanoparticles supported on polydopamine modified carbon materials, J. Mater. Chem. A 1 (2013) 3945
3953. [75] Z. Su, C. Li, Y. Cheng, Q. Gui, Y. Xiong, Y. Tan, et al., Enhanced electrocatalytic performance of platinum nanoparticles on thiolated polyaniline-multiwalled carbon nanotubes for methanol oxidation, RSC Adv. 8 (2018) 33742
33747. [76] A. De, R. Adhikary, J. Datta, Proactive role of carbon nanotube-polyaniline conjugate support for Pt nano-particles toward electro-catalysis of ethanol in fuel cell, Int. J. Hydrog. Energy 42 (2017) 25316
25325. [77] S. Hussain, N. Kongi, L. Matisen, J. Kozlova, V. Sammelselg, K. Tammeveski, Platinum nanoparticles supported on nitrobenzene-functionalised graphene nanosheets as electrocatalysts for oxygen reduction reaction in alkaline media, Electrochem. Commun. 81 (2017) 79
83. [78] S. Chen, Z. Wei, L. Guo, W. Ding, L. Dong, P. Shen, et al., Enhanced dispersion and durability of Pt nanoparticles on a thiolated CNT support, Chem. Commun. 47 (2011) 10984
10986. [79] B. Liu, H. Wang, Y. Chen, J. Wang, L. Peng, L. Li, Pt nanoparticles anchored on Nb2O5 and carbon fibers as an enhanced performance catalyst for methanol oxidation, J. Alloy Compd. 682 (2016) 584
589. [80] Y. Li, X. Zhang, S. Wang, G. Sun, Durable platinum-based electrocatalyst supported by multiwall carbon nanotubes modified with CeO2, ChemElectroChem 5 (2018) 2442
2448. [81] H. Na, H. Choi, J.-W. Oh, Y.S. Jung, Y.S. Cho, Enhanced CO oxidation and cyclic activities in three-dimensional platinum/indium tin oxide/carbon black electrocatalysts processed by cathodic arc deposition, ACS Appl. Mater. Interfaces 11 (2019) 25179
25185. [82] A.B. Yousaf, M. Imran, N. Uwitonze, A. Zeb, S.J. Zaidi, T.M. Ansari, et al., Enhanced electrocatalytic performance of Pt3Pd1 alloys supported on CeO2/C for methanol oxidation and oxygen reduction reactions, J. Phys. Chem. C 121 (2017) 2069
2079. [83] S. Hussain, H. Erikson, N. Kongi, A. Tarre, P. Ritslaid, M. Kook, et al., Improved ORR activity and long-term durability of pt nanoparticles deposited on TiO2-decorated multiwall carbon nanotubes, J. Electrochem. Soc. 166 (2019) F1284
F1291. [84] N. Garino, A. Sacco, M. Castellino, J.A. Mun˜oz-Tabares, A. Chiodoni, V. Agostino, et al., Microwave-assisted synthesis of reduced graphene oxide/SnO2 nanocomposite for oxygen reduction reaction in microbial fuel cells, ACS Appl. Mater. Interfaces 8 (2016) 4633
4643. [85] L. Chinchilla, D. Rossouw, T. Trefz, D. Susac, N. Kremliakova, G.A. Botton, Nanoscale analysis of structural and chemical changes in aged hybrid Pt/NbOx/C fuel cell catalysts, J. Power Sources 356 (2017) 140
152. [86] M. Dou, M. Hou, Z. Li, F. Wang, D. Liang, Z. Shao, et al., Pt/WO3/C nanocomposite with parallel WO3 nanorods as cathode catalyst for proton exchange membrane fuel cells, J. Energy Chem. 24 (2015) 39
44.

164

Fuel Cells for Transportation

[87] Y.-P. Xiao, W.-J. Jiang, S. Wan, X. Zhang, J.-S. Hu, Z.-D. Wei, et al., Self-deposition of Pt nanocrystals on Mn3O4 coated carbon nanotubes for enhanced oxygen reduction electrocatalysis, J. Mater. Chem. A 1 (2013) 7463
7468. [88] I. Novomlinskiy, V. Guterman, M. Danilenko, V. Volochaev, Platinum electrocatalysts deposited onto composite carbon black
metal oxide support, Russ. J. Electrochem. 55 (2019) 690
700. [89] C.-L. Sun, C.-W. Pao, H.-M. Tsai, J.-W. Chiou, S.C. Ray, H.-W. Wang, et al., Atomistic nucleation sites of Pt nanoparticles on N-doped carbon nanotubes, Nanoscale 5 (2013) 6812
6818. [90] X. Zhang, G. Xia, C. Huang, Y. Wang, Effect of the pretreatment on the performances of graphene composite-supported Pt nanoparticle electrocatalyst, Ionics 21 (2015) 449
458. [91] Y. Song, R.M. Garcia, R.M. Dorin, H. Wang, Y. Qiu, E.N. Coker, et al., Synthesis of platinum nanowire networks using a soft template, Nano Lett. 7 (2007) 3650
3655. [92] C. Wang, H. Daimon, T. Onodera, T. Koda, S. Sun, A general approach to the size- and shape-controlled synthesis of platinum nanoparticles and their catalytic reduction of oxygen, Angew. Chem. Int. Ed. 47 (2008) 3588
3591. [93] L. Colombi Ciacchi, W. Pompe, A. De Vita, Initial nucleation of platinum clusters after reduction of K2PtCl4 in aqueous solution: a first principles study, J. Am. Chem. Soc. 123 (2001) 7371
7380. [94] L.C. Ciacchi, M. Mertig, W. Pompe, S. Meriani, A. De Vita, Nucleation and growth of platinum clusters in solution and on biopolymers, Platin. Met. Rev. 47 (2003) 98
107. [95] V.K. LaMer, R.H. Dinegar, Theory, production and mechanism of formation of monodispersed hydrosols, J. Am. Chem. Soc. 72 (1950) 4847
4854. [96] C.-K. Tsung, J.N. Kuhn, W. Huang, C. Aliaga, L.-I. Hung, G.A. Somorjai, et al., Sub-10 nm platinum nanocrystals with size and shape control: catalytic study for ethylene and pyrrole hydrogenation, J. Am. Chem. Soc. 131 (2009) 5816
5822. [97] R. Loukrakpam, J. Luo, T. He, Y. Chen, Z. Xu, P.N. Njoki, et al., Nanoengineered PtCo and PtNi catalysts for oxygen reduction reaction: an assessment of the structural and electrocatalytic properties, J. Phys. Chem. C 115 (2011) 1682
1694. [98] Y. Song, Y. Yang, C.J. Medforth, E. Pereira, A.K. Singh, H. Xu, et al., Controlled synthesis of 2-D and 3-D dendritic platinum nanostructures, J. Am. Chem. Soc. 126 (2004) 635
645. [99] W. Si, J. Li, H. Li, S. Li, J. Yin, H. Xu, et al., Light-controlled synthesis of uniform platinum nanodendrites with markedly enhanced electrocatalytic activity, Nano Res. 6 (2013) 720
725. [100] S.-I. Choi, S. Xie, M. Shao, N. Lu, S. Guerrero, J.H. Odell, et al., Controlling the size and composition of nanosized Pt
Ni octahedra to optimize their catalytic activities toward the oxygen reduction reaction, ChemSusChem 7 (2014) 1476
1483. [101] L. Zhou, Y. Wang, J. Tang, J. Li, S. Wang, Y. Wang, Facile synthesis of holey graphene-supported Pt catalysts for direct methanol electro-oxidation, Microporous Mesoporous Mater. 247 (2017) 116
123. [102] E.R. Leite, C. Ribeiro, Crystallization and Growth of Colloidal Nanocrystals (E. R. Leite and C. Ribeiro, Eds.) Springer New York, New York, 2012, pp. 19
43. [103] Y. Xia, Y. Xiong, B. Lim, S.E. Skrabalak, Shape-controlled synthesis of metal nanocrystals: simple chemistry meets complex physics? Angew. Chem. Int. Ed. 48 (2009) 60
103.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

165

[104] F. Yang, J. Ye, Q. Yuan, X. Yang, Z. Xie, F. Zhao, et al., Ultrasmall Pd-Cu-Pt trimetallic twin icosahedrons boost the electrocatalytic performance of glycerol oxidation at the operating temperature of fuel cells, Adv. Funct. Mater. 30 (2020) 1908235. [105] A. Mahmood, F. Saleem, H. Lin, B. Ni, X. Wang, Crystallinity-induced shape evolution of Pt
Ag nanosheets from branched nanocrystals, Chem. Commun. 52 (2016) 10547
10550. [106] Y. Wang, J. He, C. Liu, W.H. Chong, H. Chen, Thermodynamics versus kinetics in nanosynthesis, Angew. Chem. Int. Ed. 54 (2015) 2022
2051. [107] A.R. Poerwoprajitno, L. Gloag, S. Cheong, J.J. Gooding, R.D. Tilley, Synthesis of lowand high-index faceted metal (Pt, Pd, Ru, Ir, Rh) nanoparticles for improved activity and stability in electrocatalysis, Nanoscale 11 (2019) 18995
19011. [108] Y. Xia, X. Xia, H.-C. Peng, Shape-controlled synthesis of colloidal metal nanocrystals: thermodynamic versus kinetic products, J. Am. Chem. Soc. 137 (2015) 7947
7966. [109] S. Ghosh, L. Manna, The many “Facets” of halide ions in the chemistry of colloidal inorganic nanocrystals, Chem. Rev. 118 (2018) 7804
7864. [110] J. Zhang, J. Fang, A general strategy for preparation of Pt 3d-transition metal (Co, Fe, Ni) nanocubes, J. Am. Chem. Soc. 131 (2009) 18543
18547. [111] Y.-J. Wang, W. Long, L. Wang, R. Yuan, A. Ignaszak, B. Fang, et al., Unlocking the door to highly active ORR catalysts for PEMFC applications: polyhedron-engineered Ptbased nanocrystals, Energy Environ. Sci. 11 (2018) 258
275. [112] M. Liu, Z. Zhao, X. Duan, Y. Huang, Nanoscale structure design for high-performance Pt-based ORR Catalysts, Adv. Mater. 31 (2019) 1802234. [113] B. Wu, N. Zheng, G. Fu, Small molecules control the formation of Pt nanocrystals: a key role of carbon monoxide in the synthesis of Pt nanocubes, Chem. Commun. 47 (2011) 1039
1041. [114] N. Ortiz, S.E. Skrabalak, On the dual roles of ligands in the synthesis of colloidal metal nanostructures, Langmuir 30 (2014) 6649
6659. [115] Z. Shi, X. Li, T. Li, Y. Chen, Y. Tang, Evolution of composition and structure of PtRh/C in the acidic methanol electrooxidation process, Electrochem. Commun. 113 (2020) 106690. [116] H. Li, R. Yao, D. Wang, J. He, M. Li, Y. Song, Facile synthesis of carbon supported Pd3Au@super-thin Pt core/shell electrocatalyst with a remarkable activity for oxygen reduction, J. Phys. Chem. C. 119 (2015) 4052
4061. [117] E. Teran-Salgado, D. Bahena-Uribe, P.A. Ma´rquez-Aguilar, J.L. Reyes-Rodriguez, R. Cruz-Silva, O. Solorza-Feria, Platinum nanoparticles supported on electrochemically oxidized and exfoliated graphite for the oxygen reduction reaction, Electrochim. Acta 298 (2019) 172
185. [118] J. Liu, J. Lan, L. Yang, F. Wang, J. Yin, PtM (M 5 Fe, Co, Ni) bimetallic nanoclusters as active, methanol-tolerant, and stable catalysts toward the oxygen reduction reaction, ACS Sustain. Chem. Eng. 7 (2019) 6541
6549. [119] C. Park, E. Lee, G. Lee, Y. Tak, Superior durability and stability of Pt electrocatalyst on N-doped graphene-TiO2 hybrid material for oxygen reduction reaction and polymer electrolyte membrane fuel cells, Appl. Catal. B: Environ. 268 (2020) 118414. [120] J. Choi, J. Cho, C.-W. Roh, B.-S. Kim, M.S. Choi, H. Jeong, et al., Au-doped PtCo/C catalyst preventing Co leaching for proton exchange membrane fuel cells, Appl. Catal. B: Environ. 247 (2019) 142
149. [121] W. He, J. Liu, Y. Qiao, Z. Zou, X. Zhang, D.L. Akins, et al., Simple preparation of Pd
Pt nanoalloy catalysts for methanol-tolerant oxygen reduction, J. Power Sources 195 (2010) 1046
1050.

166

Fuel Cells for Transportation

[122] Y. Zhao, Y. Wu, J. Liu, F. Wang, Dependent relationship between quantitative lattice contraction and enhanced oxygen reduction activity over Pt
Cu alloy catalysts, ACS Appl. Mater. Interfaces 9 (2017) 35740
35748. [123] J. Liu, J. Yin, B. Feng, F. Li, F. Wang, One-pot synthesis of unprotected PtPd nanoclusters with enhanced catalytic activity, durability, and methanol-tolerance for oxygen reduction reaction, Appl. Surf. Sci. 473 (2019) 318
325. [124] J. Maya-Cornejo, R. Carrera-Cerritos, D. Sebastian, J. Ledesma-Garcia, L. Arriaga, A. Arico, et al., PtCu catalyst for the electro-oxidation of ethanol in an alkaline direct alcohol fuel cell, Int. J. Hydrog. Energy 42 (2017) 27919
27928. [125] J. Liu, X. Wu, L. Yang, F. Wang, J. Yin, Unprotected Pt nanoclusters anchored on ordered mesoporous carbon as an efficient and stable catalyst for oxygen reduction reaction, Electrochim. Acta 297 (2019) 539
544. [126] J. Quinson, M. Inaba, S. Neumann, A.A. Swane, J. Bucher, S.B. Simonsen, et al., Investigating particle size effects in catalysis by applying a size-controlled and surfactant-free synthesis of colloidal nanoparticles in alkaline ethylene glycol: case study of the oxygen reduction reaction on Pt, ACS Catal. 8 (2018) 6627
6635. [127] M. Saquib, A. Halder, Dealloyed Pt3Co nanoparticles with higher geometric strain for superior hydrogen evolution reaction, J. Solid State Chem. 262 (2018) 229
236. [128] D. Ke, J. Wang, H. Zhang, Y. Li, L. Zhang, X. Zhao, et al., Fabrication of Pt
Co NPs supported on nanoporous graphene as high-efficient catalyst for hydrolytic dehydrogenation of ammonia borane, Int. J. Hydrog. Energy 42 (2017) 26617
26625. [129] E. Zagoraiou, M.K. Daletou, L. Sygellou, S. Ballomenou, S.G. Neophytides, Highly dispersed platinum supported catalysts
effect of properties on the electrocatalytic activity, Appl. Catal. B: Environ. 259 (2019) 118050. [130] E. Zagoraiou, N. Shroti, M.K. Daletou, Development of Pt-Co catalysts supported on carbon nanotubes using the polyol method—tuning the conditions for optimum properties, Mater. Today Chem. 16 (2020) 100263. [131] K.A. Kuttiyiel, K. Sasaki, G.-G. Park, M.B. Vukmirovic, L. Wu, Y. Zhu, et al., Janus structured Pt
FeNC nanoparticles as a catalyst for the oxygen reduction reaction, Chem. Commun. 53 (2017) 1660
1663. [132] Y. Yang, C. Tan, Y. Yang, L. Zhang, B.-W. Zhang, K.-H. Wu, et al., Pt3Co@Pt core@shell nanoparticles as efficient oxygen reduction electrocatalysts in direct methanol fuel cell, ChemCatChem 13 (2021) 1587
1594. [133] Y. Xiong, L. Xiao, Y. Yang, F.J. DiSalvo, H.D. Abrun˜a, High-loading intermetallic Pt3Co/C core
shell nanoparticles as enhanced activity electrocatalysts toward the oxygen reduction reaction (ORR), Chem. Mater. 30 (2018) 1532
1539. [134] D. Wang, H.L. Xin, R. Hovden, H. Wang, Y. Yu, D.A. Muller, et al., Structurally ordered intermetallic platinum
cobalt core
shell nanoparticles with enhanced activity and stability as oxygen reduction electrocatalysts, Nat. Mater. 12 (2013) 81
87. [135] M. Liu, A. Hu, Y. Ma, G. Wang, L. Zou, X. Chen, et al., Nitrogen-doped Pt3Co intermetallic compound nanoparticles: a durable oxygen reduction electrocatalyst, J. Electroanal. Chem. 871 (2020) 114267. [136] L. Zou, J. Fan, Y. Zhou, C. Wang, J. Li, Z. Zou, et al., Conversion of PtNi alloy from disordered to ordered for enhanced activity and durability in methanol-tolerant oxygen reduction reactions, Nano Res. 8 (2015) 2777
2788. [137] L. Su, W. Jia, C.M. Li, Y. Lei, Mechanisms for enhanced performance of platinum-based electrocatalysts in proton exchange membrane fuel cells, ChemSusChem 7 (2014) 361
378.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

167

[138] D. Fang, H. Zhang, L. He, J. Geng, W. Song, S. Sun, et al., Facile synthesis of nanoporous Pt-encapsulated Ir black as a bifunctional oxygen catalyst via modified polyol process at room temperature, ChemElectroChem 6 (2019) 3633
3643. [139] V.R. Stamenkovic, B. Fowler, B.S. Mun, G. Wang, P.N. Ross, C.A. Lucas, et al., Improved oxygen reduction activity on Pt3Ni(111) via increased surface site availability, Science 315 (2007) 493
497. [140] R. Antoniassi, L. Otubo, J. Vaz, A.O. Neto, E. Spinace´, Synthesis of Pt nanoparticles with preferential (1 0 0) orientation directly on the carbon support for direct ethanol fuel cell, J. Catal. 342 (2016) 67
74. [141] H. Huang, K. Li, Z. Chen, L. Luo, Y. Gu, D. Zhang, et al., Achieving remarkable activity and durability toward oxygen reduction reaction based on ultrathin Rh-doped Pt nanowires, J. Am. Chem. Soc. 139 (2017) 8152
8159. [142] A.-X. Yin, X.-Q. Min, W. Zhu, H.-S. Wu, Y.-W. Zhang, C.-H. Yan, Multiply twinned Pt
Pd nanoicosahedrons as highly active electrocatalysts for methanol oxidation, Chem. Commun. 48 (2012) 543
545. [143] R.A. Martı´nez-Rodrı´guez, F.J. Vidal-Iglesias, J. Solla-Gullo´n, C.R. Cabrera, J.M. Feliu, Synthesis of Pt nanoparticles in water-in-oil microemulsion: effect of HCl on their surface structure, J. Am. Chem. Soc. 136 (2014) 1280
1283. [144] X. Huang, Y. Li, Y. Li, H. Zhou, X. Duan, Y. Huang, Synthesis of PtPd bimetal nanocrystals with controllable shape, composition, and their tunable catalytic properties, Nano Lett. 12 (2012) 4265
4270. [145] J. Zhang, H. Yang, J. Fang, S. Zou, Synthesis and oxygen reduction activity of shapecontrolled Pt3Ni nanopolyhedra, Nano Lett. 10 (2010) 638
644. [146] G. Chen, Y. Tan, B. Wu, G. Fu, N. Zheng, Carbon monoxide-controlled synthesis of surface-clean Pt nanocubes with high electrocatalytic activity, Chem. Commun. 48 (2012) 2758
2760. [147] H. Song, F. Kim, S. Connor, G.A. Somorjai, P. Yang, Pt nanocrystals: shape control and Langmuir 2 Blodgett monolayer formation, J. Phys. Chem. B 109 (2005) 188
193. [148] S.-I. Choi, S. Xie, M. Shao, J.H. Odell, N. Lu, H.-C. Peng, et al., Synthesis and characterization of 9 nm Pt
Ni octahedra with a record high activity of 3.3 A/mgPt for the oxygen reduction reaction, Nano Lett. 13 (2013) 3420
3425. [149] C.T. Lee, X. Yang, M. Vara, K.D. Gilroy, Y. Xia, Water-based synthesis of sub-10 nm Pt octahedra and their performance towards the oxygen reduction Reaction, ChemNanoMat 3 (2017) 879
884. [150] J. Zhu, M. Xiao, K. Li, C. Liu, X. Zhao, W. Xing, Active Pt3Ni (111) surface of Pt3Ni icosahedron for oxygen reduction, ACS Appl. Mater. Interfaces 8 (2016) 30066
30071. [151] J.R. Barbosa, M.N. Leon, C.M. Fernandes, R.M. Antoniassi, O.C. Alves, E.A. Ponzio, et al., PtSnO2/C and Pt/C with preferential (100) orientation: high active electrocatalysts for ammonia electro-oxidation reaction, Appl. Catal. B: Environ. 264 (2020) 118458. [152] R. Antoniassi, J. Silva, T. Lopes, A.O. Neto, E. Spinace´, Carbon-supported Pt nanoparticles with (100) preferential orientation with enhanced electrocatalytic properties for carbon monoxide, methanol and ethanol oxidation in acidic medium, Int. J. Hydrog. Energy 42 (2017) 28786
28796. [153] Y. Wu, S. Cai, D. Wang, W. He, Y. Li, Syntheses of water-soluble octahedral, truncated octahedral, and cubic Pt
Ni nanocrystals and their structure
activity study in model hydrogenation reactions, J. Am. Chem. Soc. 134 (2012) 8975
8981. [154] Z. Peng, C. Kisielowski, A. Bell, Surfactant-free preparation of supported cubic platinum nanoparticles, Chem. Commun. 48 (2012) 1854
1856.

168

Fuel Cells for Transportation

[155] L. Huang, X. Zhang, Q. Wang, Y. Han, Y. Fang, S. Dong, Shape-control of Pt
Ru nanocrystals: tuning surface structure for enhanced electrocatalytic methanol oxidation, J. Am. Chem. Soc. 140 (2018) 1142
1147. [156] I. Safo, M. Werheid, C. Dosche, M. Oezaslan, The role of polyvinylpyrrolidone (PVP) as a capping and structure-directing agent in the formation of Pt nanocubes, Nanoscale Adv. 1 (2019) 3095
3106. [157] T. Zhang, S.-C. Li, W. Zhu, Z.-P. Zhang, J. Gu, Y.-W. Zhang, Shape-tunable Pt
Ir alloy nanocatalysts with high performance in oxygen electrode reactions, Nanoscale 9 (2017) 1154
1165. [158] B. Li, J. Wang, X. Gao, C. Qin, D. Yang, H. Lv, et al., High performance octahedral PtNi/C catalysts investigated from rotating disk electrode to membrane electrode assembly, Nano Res. 12 (2019) 281
287. [159] N. Erini, V. Beermann, M. Gocyla, M. Gliech, M. Heggen, R.E. Dunin-Borkowski, et al., The effect of surface site ensembles on the activity and selectivity of ethanol electrooxidation by octahedral PtNiRh nanoparticles, Angew. Chem. Int. Ed. 56 (2017) 6533
6538. [160] V. Beermann, M. Gocyla, E. Willinger, S. Rudi, M. Heggen, R.E. Dunin-Borkowski, et al., Rh-doped Pt
Ni octahedral nanoparticles: understanding the correlation between elemental distribution, oxygen reduction reaction, and shape stability, Nano Lett. 16 (2016) 1719
1725. [161] J. Lim, H. Shin, M. Kim, H. Lee, K.-S. Lee, Y. Kwon, et al., Ga-doped Pt-Ni octahedral nanoparticles as a highly active and durable electrocatalyst for oxygen reduction reaction, Nano Lett. 18 (2018) 2450
2458. [162] X.Q. Huang, Z.P. Zhao, L. Cao, Y. Chen, E.B. Zhu, Z.Y. Lin, et al., High-performance transition metal-doped Pt3Ni octahedra for oxygen reduction reaction, Science 348 (2015) 1230
1234. [163] J. Wang, B. Li, X. Gao, D. Yang, H. Lv, Q. Xiao, et al., From rotating disk electrode to single cell: exploration of PtNi/C octahedral nanocrystal as practical proton exchange membrane fuel cell cathode catalyst, J. Power Sources 406 (2018) 118
127. [164] J. Wang, Q. Xue, B. Li, D. Yang, H. Lv, Q. Xiao, et al., Preparation of a graphitizedcarbon-supported PtNi octahedral catalyst and application in a proton-exchange membrane fuel cell, ACS Appl. Mater. Interfaces 12 (2020) 7047
7056. [165] J. Wang, B. Li, D. Yang, H. Lv, C. Zhang, Preparation of an octahedral PtNi/CNT catalyst and its application in high durability PEMFC cathodes, RSC Adv. 8 (2018) 18381
18387. [166] Z. Zhao, M. Feng, J. Zhou, Z. Liu, M. Li, Z. Fan, et al., Composition tunable ternary Pt
Ni
Co octahedra for optimized oxygen reduction activity, Chem. Commun. 52 (2016) 11215
11218. [167] A.-X. Yin, X.-Q. Min, Y.-W. Zhang, C.-H. Yan, Shape-selective synthesis and facetdependent enhanced electrocatalytic activity and durability of monodisperse sub-10 nm Pt 2 Pd tetrahedrons and cubes, J. Am. Chem. Soc. 133 (2011) 3816
3819. [168] C. Zhang, W. Sandorf, Z. Peng, Octahedral Pt2CuNi uniform alloy nanoparticle catalyst with high activity and promising stability for oxygen reduction reaction, ACS Catal. 5 (2015) 2296
2300. [169] D. Wu, X. Shen, L.Q. Zhou, T. Nagai, Y. Pan, L. Yao, et al., A vacuum impregnation method for synthesizing octahedral Pt2CuNi nanoparticles on mesoporous carbon support and the oxygen reduction reaction electrocatalytic properties, J. Colloid Interface Sci. 564 (2020) 245
253.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

169

[170] C. Zhang, S.Y. Hwang, Z. Peng, Size-dependent oxygen reduction property of octahedral Pt
Ni nanoparticle electrocatalysts, J. Mater. Chem. A 2 (2014) 19778
19787. [171] C. Zhang, S.Y. Hwang, A. Trout, Z. Peng, Solid-state chemistry-enabled scalable production of octahedral Pt
Ni alloy electrocatalyst for oxygen reduction reaction, J. Am. Chem. Soc. 136 (2014) 7805
7808. [172] C. Cui, L. Gan, H.-H. Li, S.-H. Yu, M. Heggen, P. Strasser, Octahedral PtNi nanoparticle catalysts: exceptional oxygen reduction activity by tuning the alloy particle surface composition, Nano Lett. 12 (2012) 5885
5889. [173] P. Qiu, J. Bi, X. Zhang, S. Yang, Organics- and surfactant-free molten salt medium controlled synthesis of Pt-M (M 5 Cu and Pd) Bi- and trimetallic nanocubes and nanosheets, ACS Sustain. Chem. Eng. 5 (2017) 4205
4213. [174] M. Rana, M. Chhetri, B. Loukya, P.K. Patil, R. Datta, U.K. Gautam, High-yield synthesis of sub-10 nm Pt nanotetrahedra with bare ⟨111⟩ facets for efficient electrocatalytic applications, ACS Appl. Mater. Interfaces 7 (2015) 4998
5005. [175] J. Wu, L. Qi, H. You, A. Gross, J. Li, H. Yang, Icosahedral platinum alloy nanocrystals with enhanced electrocatalytic activities, J. Am. Chem. Soc. 134 (2012) 11880
11883. [176] Y. Xia, M. Zhao, X. Wang, D. Huo, Toward affordable and sustainable use of precious metals in catalysis and nanomedicine, MRS Bull. 43 (2018) 860
869. [177] L. Zhang, L.T. Roling, X. Wang, M. Vara, M. Chi, J. Liu, et al., Platinum-based nanocages with subnanometer-thick walls and well-defined, controllable facets, Science 349 (2015) 412
416. [178] C. Chen, Y.J. Kang, Z.Y. Huo, Z.W. Zhu, W.Y. Huang, H.L.L. Xin, et al., Highly crystalline multimetallic nanoframes with three-dimensional electrocatalytic surfaces, Science 343 (2014) 1339
1343. [179] Z. Fang, Y. Wang, C. Liu, S. Chen, W. Sang, C. Wang, et al., Rational design of metal nanoframes for catalysis and plasmonics, Small 11 (2015) 2593
2605. [180] Y. Zheng, J. Zeng, A. Ruditskiy, M. Liu, Y. Xia, Oxidative etching and its role in manipulating the nucleation and growth of noble-metal nanocrystals, Chem. Mater. 26 (2014) 22
33. [181] M. Zhao, X. Wang, X. Yang, K.D. Gilroy, D. Qin, Y. Xia, Hollow metal nanocrystals with ultrathin, porous walls and well-controlled surface structures, Adv. Mater. 30 (2018) 1801956. [182] R. Wu, Q.-C. Kong, C. Fu, S.-Q. Lai, C. Ye, J.-Y. Liu, et al., One-pot synthesis and enhanced catalytic performance of Pd and Pt nanocages via galvanic replacement reactions, RSC Adv. 3 (2013) 12577
12580. [183] W. Zhang, J. Yang, X. Lu, Tailoring galvanic replacement reaction for the preparation of Pt/Ag bimetallic hollow nanostructures with controlled number of voids, ACS Nano 6 (2012) 7397
7405. [184] X. Yang, L.T. Roling, M. Vara, A.O. Elnabawy, M. Zhao, Z.D. Hood, et al., Synthesis and characterization of Pt
Ag alloy nanocages with enhanced activity and durability toward oxygen reduction, Nano Lett. 16 (2016) 6644
6649. [185] Z. Li, R. Zeng, L. Wang, L. Jiang, S. Wang, X. Liu, A simple strategy to form hollow Pt3Co alloy nanosphere with ultrathin Pt shell with significant enhanced oxygen reduction reaction activity, Int. J. Hydrog. Energy 41 (2016) 21394
21403. [186] G. Chen, X. Yang, Z. Xie, F. Zhao, Z. Zhou, Q. Yuan, Hollow PtCu octahedral nanoalloys: efficient bifunctional electrocatalysts towards oxygen reduction reaction and methanol oxidation reaction by regulating near-surface composition, J. Colloid Interface Sci. 562 (2020) 244
251.

170

Fuel Cells for Transportation

[187] J.W. Hong, S.W. Kang, B.-S. Choi, D. Kim, S.B. Lee, S.W. Han, Controlled synthesis of Pd
Pt alloy hollow nanostructures with enhanced catalytic activities for oxygen reduction, ACS Nano 6 (2012) 2410
2419. [188] Y. Vasquez, A.K. Sra, R.E. Schaak, One-pot synthesis of hollow superparamagnetic CoPt nanospheres, J. Am. Chem. Soc. 127 (2005) 12504
12505. [189] B.Y. Xia, H.B. Wu, X. Wang, X.W. Lou, One-pot synthesis of cubic PtCu3 nanocages with enhanced electrocatalytic activity for the methanol oxidation reaction, J. Am. Chem. Soc. 134 (2012) 13934
13937. [190] Y. Zhou, J. Yang, C. Zhu, D. Du, X. Cheng, C.H. Yen, et al., Newly designed graphene cellular monolith functionalized with hollow Pt-M (M 5 Ni, Co) nanoparticles as the electrocatalyst for oxygen reduction reaction, ACS Appl. Mater. Interfaces 8 (2016) 25863
25874. [191] P. Qiu, S. Lian, G. Yang, S. Yang, Halide ion-induced formation of single crystalline mesoporous PtPd bimetallic nanoparticles with hollow interiors for electrochemical methanol and ethanol oxidation reaction, Nano Res. 10 (2017) 1064
1077. [192] C. Fang, J. Zhao, R. Jiang, J. Wang, G. Zhao, B. Geng, Engineering of hollow PdPt nanocrystals via reduction kinetic control for their superior electrocatalytic performances, ACS Appl. Mater. Interfaces 10 (2018) 29543
29551. [193] X. Huang, H. Zhang, C. Guo, Z. Zhou, N. Zheng, Simplifying the creation of hollow metallic nanostructures: one-pot synthesis of hollow palladium/platinum singlecrystalline nanocubes, Angew. Chem. Int. Ed. 48 (2009) 4808
4812. [194] Z. Zhang, Z. Luo, B. Chen, C. Wei, J. Zhao, J. Chen, et al., One-pot synthesis of highly anisotropic fivefold-twinned PtCu nanoframes used as a bifunctional electrocatalyst for oxygen reduction and methanol oxidation, Adv. Mater. 28 (2016) 8712
8717. [195] F. Nosheen, Z.-c Zhang, J. Zhuang, X. Wang, One-pot fabrication of single-crystalline octahedral Pt
Cu nanoframes and their enhanced electrocatalytic activity, Nanoscale 5 (2013) 3660
3663. [196] J. Ding, X. Zhu, L. Bu, J. Yao, J. Guo, S. Guo, et al., Highly open rhombic dodecahedral PtCu nanoframes, Chem. Commun. 51 (2015) 9722
9725. [197] H.-J. Niu, H.-Y. Chen, G.-L. Wen, J.-J. Feng, Q.-L. Zhang, A.-J. Wang, One-pot solvothermal synthesis of three-dimensional hollow PtCu alloyed dodecahedron nanoframes with excellent electrocatalytic performances for hydrogen evolution and oxygen reduction, J. Colloid Interface Sci. 539 (2019) 525
532. [198] S. Luo, P.K. Shen, Concave platinum-copper octopod nanoframes bounded with multiple high index facets for efficient electrooxidation catalysis, ACS Nano 11 (2017) 11946
11953. [199] S. Luo, M. Tang, P.K. Shen, S. Ye, Atomic-scale preparation of octopod nanoframes with high-index facets as highly active and stable catalysts, Adv. Mater. 29 (2017) 1601687. [200] Y. Qin, W. Zhang, K. Guo, X. Liu, J. Liu, X. Liang, et al., Fine-tuning intrinsic strain in penta-twinned Pt
Cu
Mn nanoframes boosts oxygen reduction catalysis, Adv. Funct. Mater. 30 (2020) 1910107. [201] P. Vany´sek, Electrochemical series, Handb. Chem. Phys. 93 (2012) 5
80. [202] C. Kim, S.S. Kim, S. Yang, J.W. Han, H. Lee, In situ shaping of Pt nanoparticles directly overgrown on carbon supports, Chem. Commun. 48 (2012) 6396
6398. [203] S. Bai, C. Wang, W. Jiang, N. Du, J. Li, J. Du, et al., Etching approach to hybrid structures of PtPd nanocages and graphene for efficient oxygen reduction reaction catalysts, Nano Res. 8 (2015) 2789
2799.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

171

[204] K. Deng, Y. Xu, Y. Li, Z. Dai, Z. Wang, X. Li, et al., Integration mesoporous surface and hollow cavity into PtPdRh nano-octahedra for enhanced oxygen reduction electrocatalysis, Nanotechnology 31 (2019) 025401. [205] D.S. He, D. He, J. Wang, Y. Lin, P. Yin, X. Hong, et al., Ultrathin icosahedral Pt-enriched nanocage with excellent oxygen reduction reaction activity, J. Am. Chem. Soc. 138 (2016) 1494
1497. [206] X. Wang, M. Luo, H. Huang, M. Chi, J. Howe, Z. Xie, et al., Facile synthesis of Pt
Pd alloy nanocages and Pt nanorings by templating with Pd nanoplates, ChemNanoMat 2 (2016) 1086
1091. [207] X. Wang, L. Figueroa-Cosme, X. Yang, M. Luo, J. Liu, Z. Xie, et al., Pt-based icosahedral nanocages: using a combination of {111} facets, twin defects, and ultrathin walls to greatly enhance their activity toward oxygen reduction, Nano Lett. 16 (2016) 1467
1471. [208] K. Deng, Y. Xu, D. Yang, X. Qian, Z. Dai, Z. Wang, et al., Pt
Ni
P nanocages with surface porosity as efficient bifunctional electrocatalysts for oxygen reduction and methanol oxidation, J. Mater. Chem. A 7 (2019) 9791
9797. [209] T. Tamaki, H. Kuroki, S. Ogura, T. Fuchigami, Y. Kitamoto, T. Yamaguchi, Connected nanoparticle catalysts possessing a porous, hollow capsule structure as carbon-free electrocatalysts for oxygen reduction in polymer electrolyte fuel cells, Energy Environ. Sci. 8 (2015) 3545
3549. [210] X. Tian, X. Zhao, Y.-Q. Su, L. Wang, H. Wang, D. Dang, et al., Engineering bunched Pt-Ni alloy nanocages for efficient oxygen reduction in practical fuel cells, Science 366 (2019) 850
856. [211] C. Wang, L. Zhang, H. Yang, J. Pan, J. Liu, C. Dotse, et al., High-indexed Pt3Ni alloy tetrahexahedral nanoframes evolved through preferential CO etching, Nano Lett. 17 (2017) 2204
2210. [212] A. Oh, H. Baik, D.S. Choi, J.Y. Cheon, B. Kim, H. Kim, et al., Skeletal octahedral nanoframe with cartesian coordinates via geometrically precise nanoscale phase segregation in a Pt@Ni core-shell nanocrystal, ACS Nano 9 (2015) 2856
2867. [213] Z. Niu, N. Becknell, Y. Yu, D. Kim, C. Chen, N. Kornienko, et al., Anisotropic phase segregation and migration of Pt in nanocrystals en route to nanoframe catalysts, Nat. Mater. 15 (2016) 1188. [214] N. Becknell, Y. Son, D. Kim, D. Li, Y. Yu, Z. Niu, et al., Control of architecture in rhombic dodecahedral Pt-Ni nanoframe electrocatalysts, J. Am. Chem. Soc. 139 (2017) 11678
11681. [215] W. Ye, S. Chen, M. Ye, C. Ren, J. Ma, R. Long, et al., Pt4PdCu0.4 alloy nanoframes as highly efficient and robust bifunctional electrocatalysts for oxygen reduction reaction and formic acid oxidation, Nano Energy 39 (2017) 532
538. [216] S. Chen, M. Li, M. Gao, J. Jin, M.A. van Spronsen, M.B. Salmeron, et al., Highperformance Pt
Co nanoframes for fuel-cell electrocatalysis, Nano Lett. 20 (2020) 1974
1979. [217] X. Zhu, L. Huang, M. Wei, P. Tsiakaras, P.K. Shen, Highly stable Pt-Co nanodendrite in nanoframe with Pt skin structured catalyst for oxygen reduction electrocatalysis, Appl. Catal. B: Environ. 281 (2021) 119460. [218] N. Becknell, C. Zheng, C. Chen, Y. Yu, P. Yang, Synthesis of PtCo3 polyhedral nanoparticles and evolution to Pt3Co nanoframes, Surf. Sci. 648 (2016) 328
332. [219] X. Sun, B. Huang, X. Cui, B. E, Y. Feng, X. Huang, Platinum
copper rhombic dodecahedral nanoframes with tunable channels as efficient bifunctional electrocatalysts for fuel-cell reactions, ChemCatChem 10 (2018) 931
935.

172

Fuel Cells for Transportation

[220] T. Kwon, M. Jun, H.Y. Kim, A. Oh, J. Park, H. Baik, et al., Vertex-reinforced PtCuCo ternary nanoframes as efficient and stable electrocatalysts for the oxygen reduction reaction and the methanol oxidation reaction, Adv. Funct. Mater. 28 (2018) 1706440. [221] H. Kwon, M.K. Kabiraz, J. Park, A. Oh, H. Baik, S.-I. Choi, et al., Dendrite-embedded platinum-nickel multiframes as highly active and durable electrocatalyst toward the oxygen reduction reaction, Nano Lett. 18 (2018) 2930
2936. [222] https://www.rsc.org/periodic-table. [223] M. Li, Z. Zhao, T. Cheng, A. Fortunelli, C.-Y. Chen, R. Yu, et al., Ultrafine jagged platinum nanowires enable ultrahigh mass activity for the oxygen reduction reaction, Science 354 (2016) 1414
1419. [224] N. Zhang, L. Bu, S. Guo, J. Guo, X. Huang, Screw thread-like platinum
copper nanowires bounded with high-index facets for efficient electrocatalysis, Nano Lett. 16 (2016) 5037
5043. [225] X. Yan, Y. Chen, S. Deng, Y. Yang, Z. Huang, C. Ge, et al., In situ integration of ultrathin PtCu nanowires with reduced graphene oxide nanosheets for efficient electrocatalytic oxygen reduction, Chem. Eur. J. 23 (2017) 16871
16876. [226] S. Sun, F. Jaouen, J.-P. Dodelet, Controlled growth of Pt nanowires on carbon nanospheres and their enhanced performance as electrocatalysts in PEM fuel cells, Adv. Mater. 20 (2008) 3900
3904. [227] H. Meng, Y. Zhan, D. Zeng, X. Zhang, G. Zhang, F. Jaouen, Factors influencing the growth of Pt nanowires via chemical self-assembly and their fuel cell performance, Small 11 (2015) 3377
3386. [228] K. Li, X. Li, H. Huang, L. Luo, X. Li, X. Yan, et al., One-nanometer-thick PtNiRh trimetallic nanowires with enhanced oxygen reduction electrocatalysis in acid media: integrating multiple advantages into one catalyst, J. Am. Chem. Soc. 140 (2018) 16159
16167. [229] Y. Zhu, L. Bu, Q. Shao, X. Huang, Subnanometer PtRh nanowire with alleviated poisoning effect and enhanced C
C bond cleavage for ethanol oxidation electrocatalysis, ACS Catal. 9 (2019) 6607
6612. [230] Y. Zhu, L. Bu, Q. Shao, X. Huang, Structurally ordered Pt3Sn nanofibers with highlighted antipoisoning property as efficient ethanol oxidation electrocatalysts, ACS Catal. 10 (2020) 3455
3461. [231] L. Bu, S. Guo, X. Zhang, X. Shen, D. Su, G. Lu, et al., Surface engineering of hierarchical platinum-cobalt nanowires for efficient electrocatalysis, Nat. Commun. 7 (2016) 11850. [232] N. Zhang, Y. Feng, X. Zhu, S. Guo, J. Guo, X. Huang, Superior bifunctional liquid fuel oxidation and oxygen reduction electrocatalysis enabled by PtNiPd core-shell nanowires, Adv. Mater. 29 (2017) 1603774. [233] M. Luo, Y. Sun, X. Zhang, Y. Qin, M. Li, Y. Li, et al., Stable high-index faceted Pt skin on Zigzag-like PtFe nanowires enhances oxygen reduction catalysis, Adv. Mater. 30 (2018) 1705515. [234] Z. Wu, Y.-Q. Su, E.J. Hensen, X. Tian, C. You, Q. Xu, Highly stable Pt3Ni nanowires tailored with trace Au for the oxygen reduction reaction, J. Mater. Chem. A 7 (2019) 26402
26409. [235] Y. Xu, X. Cui, S. Wei, Q. Zhang, L. Gu, F. Meng, et al., Highly active zigzag-like Pt-Zn alloy nanowires with high-index facets for alcohol electrooxidation, Nano Res. 12 (2019) 1173
1179.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

173

[236] X. Jiang, G. Fu, X. Wu, Y. Liu, M. Zhang, D. Sun, et al., Ultrathin AgPt alloy nanowires as a high-performance electrocatalyst for formic acid oxidation, Nano Res. 11 (2018) 499
510. [237] Y. Ma, W. Gao, H. Shan, W. Chen, W. Shang, P. Tao, et al., Platinum-based nanowires as active catalysts toward oxygen reduction reaction: in situ observation of surfacediffusion-assisted, solid-state oriented attachment, Adv. Mater. 29 (2017) 1703460. [238] P. Wang, Y. Zhang, R. Shi, Z. Wang, Trimetallic PtPdCu nanowires as an electrocatalyst for methanol and formic acid oxidation, N. J. Chem. 42 (2018) 19083
19089. [239] W. Hong, C. Shang, J. Wang, E. Wang, Bimetallic PdPt nanowire networks with enhanced electrocatalytic activity for ethylene glycol and glycerol oxidation, Energy Environ. Sci. 8 (2015) 2910
2915. [240] D. Fang, L. Wan, Q. Jiang, H. Zhang, X. Tang, X. Qin, et al., Wavy PtCu alloy nanowire networks with abundant surface defects enhanced oxygen reduction reaction, Nano Res. 12 (2019) 2766
2773. [241] T. Guo, H. Xiang, W. Li, H. Li, H. Chen, S. Liu, et al., Synthesis of ultrathin and composition-tunable PdPt porous nanowires with enhanced electrocatalytic performance, ACS Sustain. Chem. Eng. 8 (2020) 2901
2909. [242] X. Fan, S. Luo, X. Zhao, X. Wu, Z. Luo, M. Tang, et al., One-nanometer-thick platinum-based nanowires with controllable surface structures, Nano Res. 12 (2019) 1721
1726. [243] L.Z. Bu, N. Zhang, S.J. Guo, X. Zhang, J. Li, J.L. Yao, et al., Biaxially strained PtPb/Pt core/shell nanoplate boosts oxygen reduction catalysis, Science 354 (2016) 1410
1414. [244] H. Liu, J. Qin, S. Zhao, Z. Gao, Q. Fu, Y. Song, Two-dimensional circular platinum nanodendrites toward efficient oxygen reduction reaction and methanol oxidation reaction, Electrochem. Commun. 98 (2019) 53
57. [245] Z. Li, X. Zhang, H. Cheng, J. Liu, M. Shao, M. Wei, et al., Confined synthesis of 2D nanostructured materials toward electrocatalysis, Adv. Energy Mater. 10 (2020) 1900486. [246] H. Liao, J. Zhu, Y. Hou, Synthesis and electrocatalytic properties of PtBi nanoplatelets and PdBi nanowires, Nanoscale 6 (2014) 1049
1055. [247] H. Lv, L. Sun, D. Xu, S.L. Suib, B. Liu, One-pot aqueous synthesis of ultrathin trimetallic PdPtCu nanosheets for the electrooxidation of alcohols, Green. Chem. 21 (2019) 2367
2374. [248] J.W. Hong, Y. Kim, D.H. Wi, S. Lee, S.U. Lee, Y.W. Lee, et al., Ultrathin free-standing ternary-alloy nanosheets, Angew. Chem. Int. Ed. 128 (2016) 2803
2808. [249] Y. Chen, Z. Fan, Z. Zhang, W. Niu, C. Li, N. Yang, et al., Two-dimensional metal nanomaterials: synthesis, properties, and applications, Chem. Rev. 118 (2018) 6409
6455. [250] T. Wang, M. Park, Q. Yu, J. Zhang, Y. Yang, Stability and synthesis of 2D metals and alloys: a review, Mater. Today Adv. 8 (2020) 100092. [251] K. Huang, J.W. Hou, Q.Y. Zhang, G. Ou, D.D. Ning, N. Hussain, et al., Ultrathin twodimensional metals with fully exposed (111) facets, Chem. Commun. 54 (2018) 160
163. [252] Y. Song, R.M. Dorin, R.M. Garcia, Y.-B. Jiang, H. Wang, P. Li, et al., Synthesis of platinum nanowheels using a bicellar template, J. Am. Chem. Soc. 130 (2008) 12602
12603. [253] A. Mahmood, H. Lin, N. Xie, X. Wang, Surface confinement etching and polarization matter: a new approach to prepare ultrathin PtAgCo nanosheets for hydrogen-evolution reactions, Chem. Mater. 29 (2017) 6329
6335.

174

Fuel Cells for Transportation

[254] S. Yang, P. Qiu, G. Yang, Graphene induced formation of single crystal Pt nanosheets through 2-dimensional aggregation and sintering of nanoparticles in molten salt medium, Carbon 77 (2014) 1123
1131. [255] F. Lin, K. Wang, Y. Tang, J. Lai, M. Lou, M. Huang, et al., Enhanced bifunctional fuel cell catalysis via Pd/PtCu core/shell nanoplates, Chem. Commun. 54 (2018) 1315
1318. [256] W. Wang, X. Zhang, Y. Zhang, X. Chen, J. Ye, J. Chen, et al., Edge enrichment of ultrathin 2D PdPtCu trimetallic nanostructures effectuates top-ranked ethanol electrooxidation, Nano Lett. 20 (2020) 5458
5464. [257] F. Saleem, B. Xu, B. Ni, H. Liu, F. Nosheen, H. Li, et al., Atomically thick Pt-Cu nanosheets: self-assembled sandwich and nanoring-like structures, Adv. Mater. 27 (2015) 2013
2018. [258] L. Dai, Y. Zhao, Q. Qin, X. Zhao, C. Xu, N. Zheng, Carbon-monoxide-assisted synthesis of ultrathin PtCu alloy nanosheets and their enhanced catalysis, ChemNanoMat 2 (2016) 776
780. [259] Y. Feng, B. Huang, C. Yang, Q. Shao, X. Huang, Platinum porous nanosheets with high surface distortion and pt utilization for enhanced oxygen reduction catalysis, Adv. Funct. Mater. 29 (2019) 1904429. [260] C.-Y. Wang, Z.-Y. Yu, G. Li, Q.-T. Song, G. Li, C.-X. Luo, et al., Intermetallic PtBi nanoplates with high catalytic activity towards electro-oxidation of formic acid and glycerol, ChemElectroChem 7 (2020) 239
245. [261] I.A. Safo, M. Oezaslan, Electrochemical cleaning of polyvinylpyrrolidone-capped Pt nanocubes for the oxygen reduction reaction, Electrochim. Acta 241 (2017) 544
552. ¨ zaslan, Effects of capping agents on the oxygen reduction [262] I.A. Safo, C. Dosche, M. O reaction activity and shape stability of Pt nanocubes, ChemPhysChem 20 (2019) 3010
3023. [263] Z. Niu, Y. Li, Removal and utilization of capping agents in nanocatalysis, Chem. Mater. 26 (2014) 72
83. [264] M. Luo, Y. Hong, W. Yao, C. Huang, Q. Xu, Q. Wu, Facile removal of polyvinylpyrrolidone (PVP) adsorbates from Pt alloy nanoparticles, J. Mater. Chem. A 3 (2015) 2770
2775. [265] I.A. Safo, C. Dosche, M. Oezaslan, TEM, FTIR and electrochemistry study: desorption of PVP from Pt nanocubes, Z. Phys. Chem. 232 (2018) 1319
1333. [266] H. Zhu, S. Zhang, D. Su, G. Jiang, S. Sun, Surface profile control of FeNiPt/Pt core/shell nanowires for oxygen reduction reaction, Small 11 (2015) 3545
3549. [267] C. Wang, M. Chi, D. Li, D. van der Vliet, G. Wang, Q. Lin, et al., Synthesis of homogeneous pt-bimetallic nanoparticles as highly efficient electrocatalysts, ACS Catal. 1 (2011) 1355
1359. [268] C. Wang, M. Chi, G. Wang, D. van der Vliet, D. Li, K. More, et al., Correlation between surface chemistry and electrocatalytic properties of monodisperse PtxNi1-x nanoparticles, Adv. Funct. Mater. 21 (2011) 147
152. [269] D. Li, C. Wang, D. Tripkovic, S. Sun, N.M. Markovic, V.R. Stamenkovic, Surfactant removal for colloidal nanoparticles from solution synthesis: the effect on catalytic performance, ACS Catal. 2 (2012) 1358
1362. [270] V. Beermann, M. Gocyla, S. K¨uhl, E. Padgett, H. Schmies, M. Goerlin, et al., Tuning the electrocatalytic oxygen reduction reaction activity and stability of shape-controlled Pt
Ni nanoparticles by thermal annealing—elucidating the surface atomic structural and compositional changes, J. Am. Chem. Soc. 139 (2017) 16536
16547.

Low platinum-based electrocatalysts for fuel cells Chapter | 5

175

[271] M. Crespo-Quesada, J.-M. Andanson, A. Yarulin, B. Lim, Y. Xia, L. Kiwi-Minsker, UV
ozone cleaning of supported poly(vinylpyrrolidone)-stabilized palladium nanocubes: effect of stabilizer removal on morphology and catalytic behavior, Langmuir 27 (2011) 7909
7916. [272] S.-W. Chou, Y.-R. Lai, Y.Y. Yang, C.-Y. Tang, M. Hayashi, H.-C. Chen, et al., Uniform size and composition tuning of PtNi octahedra for systematic studies of oxygen reduction reactions, J. Catal. 309 (2014) 343
350. [273] L. Lu, B. Lou, S. Zou, H. Kobayashi, J. Liu, L. Xiao, et al., Robust removal of ligands from noble metal nanoparticles by electrochemical strategies, ACS Catal. 8 (2018) 8484
8492. [274] J. Monzo´, M.T.M. Koper, P. Rodriguez, Removing polyvinylpyrrolidone from catalytic Pt nanoparticles without modification of superficial order, ChemPhysChem 13 (2012) 709
715. [275] M.A. Montiel, F.J. Vidal-Iglesias, V. Montiel, J. Solla-Gullo´n, Electrocatalysis on shapecontrolled metal nanoparticles: progress in surface cleaning methodologies, Curr. Opin. Electrochem. 1 (2017) 34
39.

This page intentionally left blank

Chapter 6

Platinum group metal-free catalysts for fuel cells: status and prospects Md Aman Uddin and Ahmed Imtiaz Rais Department of Mechanical Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh

6.1

Introduction

Global warming is the biggest concern of recent climate summits, and all countries agreed to reduce fossil fuel use [1]. Polymer electrolyte fuel cell (PEFC) is a promising energy technology that can significantly reduce overall dependency on fossil fuels. Currently, all automotive manufacturers are investing in fuel cell electric vehicle (FCEV) development. Approximately, 7000 FCEVs are sold or leased in the United States up to now [2]. However, widespread commercialization of PEFC requires a reduction of cost, specifically a reduction of the total amount of platinum (Pt) catalyst loading. Although platinum loading decreased from 1 mg cm22 in 2002 down to 0.1 mg cm22 in lab-scale fuel cells in the last two decades [3], and platinum group metal (PGM) catalyst loading is still responsible for 40% of fuel cell stack cost of light-duty vehicles (5000 hours application), assuming production of 500,000 stacks per year [4]. Recently, fuel cells in heavy-duty vehicles are also getting significant attention, and the platinum requirement is much higher ( . 0.3 mg cm22) in these vehicles due to long-term applications of up to 25,000 hours [5]. Additionally, when fuel cell vehicles will go for mass production, the price of platinum can become unpredictable due to its localized reserve (90%) in Russia and South Africa alone, which is a concern for the world [6]. So, there is a need to develop cheaper, more durable, and high-performing PGM-free catalysts. PGM-free catalysts are synthesized through heat treatment of a transition metal precursor (Fe, Co, salt, etc.), a carbon precursor (carbon-containing compound or polymer), and a nitrogen precursor (M-N4 macrocycle, NH3, etc.) [7]. Although volatile opinions came up from different research groups Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00007-1 © 2023 Elsevier Ltd. All rights reserved.

177

178

Fuel Cells for Transportation

regarding the role of metal ions in the active site, through advanced characterization and density functional theory calculations, atomically dispersed four nitrogen-coordinated iron (FeN4) was found to be the active oxygen reduction reaction (ORR) sites in acidic media [8]. Such catalysts are labeled as Fe-N-C or Co-N-C and, in general, M-N-C (M 5 Fe, Co, Mn, etc.). Over the last 5 years, there has been a tremendous effort to improve PGM-free catalysts that can compete with PGM catalysts. Banham et al. reported 570 mW cm22 peak power density under air in an industrially relevant scale membrane electrode assembly (MEA) [9], and Uddin et al. reported 610 mW cm22 peak power density under automotive operating conditions [10]. Although there is an improvement in the initial performance of PGMfree catalysts, the overall stability of these catalysts remains poor (Fig. 6.1). It is reported that during stability testing, there is a rapid initial decay lasting about 10
15 hours followed by a more gradual decay [11
14]. Although several authors hypothesized various mechanisms for performance decay, no clear evidence was demonstrated [15]. This chapter summarizes state-of-the-art PGM-free catalyst development methods and the integration of these catalysts in MEA. Subsequently, an understanding of the stability and degradation challenges of PGM-free catalysts is reviewed. Finally, a current understanding of mitigation strategies is summarized.

6.2

Platinum group metal-free catalyst development

Researchers have been studying earth-abundant metals (Fe, Co, Mn, etc.) based electrocatalysts for more than 60 years [16]. Electrochemistry has witnessed several breakthroughs during this period in PGM-free catalyst

FIGURE 6.1 Key milestones of PGM-free catalysts. Reprinted with permission from Y. He, G. Wu, PGM-free oxygen-reduction catalyst development for proton-exchange membrane fuel cells: challenges, solutions, and promises. Acc. Mater. Res. 3 (2022) 224
236. Copyright (2022) American Chemical Society.

Platinum group metal-free catalysts for fuel cells Chapter | 6

179

synthesis methods, determining the active sites and a definite way to enhance the activity of the catalysts rather than indefinite trial and error methods. The first breakthrough in PGM-free ORR catalyst synthesis occurred in 1964 by Jasinski [16]. He discovered that cobalt phthalocyanine (CoPc) reduces O2 in an alkaline medium. The compound contains metal-N4 bonds [16]. Further research revealed that cobalt and iron macrocycles also reduced O2 in alkaline or acidic media [17]. Heat treatment of the macrocycles at elevated temperatures in an inert atmosphere significantly improved activity, stability, and durability [18,19], but the macrocycle compounds were expensive. Later, researchers synthesized a PGM-free catalyst using nitrogen-carbon containing polymer (polyacrylonitrile), replacing expensive macrocycles in 1989 [20]. Most of the synthesis processes lacked active site density and catalyst morphology control. Dodelet’s group put a lot of effort into increasing active sites by increasing micropores [21,22]. In 2009, they achieved a high MEA performance of 99 A cm23 at 0.8 V iR-free cell voltage (ViR-free) using their highly microporous Fe-N-C catalysts [21]. They used pore filler and iron precursors in the carbon black to increase porosity. They did two-step heat treatments of precursors in Ar at 1050 C and in NH3 at 950 C. NH3 partly gasifies the carbon and creates micropores inside the catalysts. In 2011, Dodelet’s group used a metal
organic framework (MOF) as the host for N and Fe precursors to further increase the porosity and surface area [22]. Their MOF was a Zn(II) zeolitic imidazolate framework (ZIF): ZIF-8. The synthesis method included the preparation of a catalyst precursor from iron(II) acetate (FeAc), 1,10-phenanthroline (Phen), and ZIF-8, and ballmilling precursors followed by two steps of pyrolysis [22]. They reported a record volumetric activity of 230 A cm23 at 0.8 ViR-free and a peak power density of 0.91 W cm22 in H2-O2. The high performance was due to the improved porous structure of ZIF-8. The authors attributed the high porosity to the evaporation of zinc during heat treatment resulting in an interconnected hollow structure [22]. Inspired by Dodelet’s works, several groups improved fuel cell performance through the synthesis of materials with a wide range of pore structures [23
27]. Atanassov’s group has developed a sacrificial support method [23
25]. Silica precursor is used in open-frame carbon structure, which is removed during pyrolysis, creating internal porosity [23
25]. Zelenay’s group extensively worked on polyaniline (PANI)-based Fe-N-C catalysts [26,27]. Their catalysts contain atomically dispersed and nitrogen coordinated iron in graphene-like carbon. Using 30 wt.% iron in precursors and multiple heat treatments and acid leaching techniques, they synthesized a high surface area catalyst (1600 m2 g21) and achieved high performance of 190 mA cm22 at 0.8 V in H2-O2 in 2014 [27]. To further improve morphology, they included another precursor, cyanamide (CM), along with PANI in their synthesis process, which decomposes at around 260 C during the heat

180

Fuel Cells for Transportation

treatment creating a hierarchical pore structure [26]. They reported a highpower density of 0.42 W cm22 in H2
air using CM 1 PANI-Fe-C catalyst. Following Dodelet’s works on ZIF-8 precursors [22], various synthesis methods were adopted to improve fuel cell performance using MOF materials [8,28]. In the synthesis of Fe-ZIF-8-based ORR catalyst, a guest
host chemistry strategy was adopted where iron nanocluster acted as guest and ZIF-8 as host [29]. The activity could be enhanced by pyrolysis treatment under NH3 [30]. The catalyst thus synthesized demonstrated improved activity in MEAs. But the nanostructure of the ZIF-8 catalyst remained heterogeneous. It lacked control over the synthesis process and structure. Wu’s group used metal-doped ZIF-8 as a precursor to accommodate a more homogenous nanostructure and achieved a high half-wave potential of 0.85 V versus RHE [31]. They chemically doped Fe into ZIF-8 precursor, and Fe partially replaced Zn creating FeN4 complexes (Fig. 6.2A). By tuning metal concentration in methanol solution, the authors synthesized catalyst particles from 20 to 1000 nm (Fig. 6.2B). Most importantly, one-step pyrolysis of Fe-doped ZIF-8 precursor produced atomic Fe sites dispersed in the carbon network (Fig. 6.2C
D) [31]. With the decrease in particle size, ORR activity increased, indicating the increase of accessible active sites with

FIGURE 6.2 (A) Synthesis method of Fe-doped ZIF catalysts. (B) Fe-ZIF catalysts with various particle sizes. (C) Image of well-defined 50-nm catalyst particles. (D) Electron energy loss spectroscopy analysis. (E) ORR activity for Fe-ZIF catalysts with various particle sizes and Pt/C catalysts (60 μgPt cm22) at 25 C and 900 rpm. (F) Accelerated stress test (voltage cycling from 0.6 to 1.0 V) of 50 nm catalyst in 0.5 M H2SO4. Reprinted with permission from H. Zhang, S. Hwang, M. Wang, Z. Feng, S. Karakalos, L. Luo, Z. Qiao, X. Xie, C. Wang, D. Su, Y. Shao, G. Wu, Single atomic iron catalysts for oxygen reduction in acidic media: particle size control and thermal activation, J. Am. Chem. Soc. 139 (2017) 14143
14149. Copyright (2017) American Chemical Society.

Platinum group metal-free catalysts for fuel cells Chapter | 6

181

smaller particles (Fig. 6.2E). ORR activity of the smallest particles (20 nm) was low, possibly due to the agglomeration of particles [31]. Accelerated stress test of 50 nm catalyst demonstrated enhanced stability with only 20 mV loss in half-wave potential after 10,000 cycles (Fig. 6.2F) [31]. In another study [32], Wu’s group optimized iron content in the precursor. They found that too little iron content in precursor generated few active sites, and too much iron content generated clustering of iron. ORR activity of the catalyst increased with the increase of iron percentage in precursor up to a threshold value of 1.5 at.% [32]. Further increase of iron percentage in precursor generated clustering of iron and increased primary particle size resulting in lower ORR activity [32]. With their highly active catalyst (1.5 at.% Fe in precursor), the authors reported a current density of 44 mA cm22 at 0.87 V (iR-free) in H2
O2 (Fig. 6.1) and 75 mA cm22 at 0.8 V in H2
air [32]. All the FeN4 active sites are not electrochemically accessible because a significant fraction is buried under the N-doped carbon matrix. Additionally, scaling up the catalysts is challenging. So, researchers are focusing on a simple method to scale up the catalyst synthesis. Wu’s group reported a chemical vapor deposition method that produces accessible catalysts [33]. The catalyst was synthesized using the one-pot method by the interactions between gaseous 2-methylimidazole (2-MeIM) and iron-doped zinc oxide (ZnO) substrate and subsequent thermal activation at 1000 C. They reported promising ORR activity and 27 mA cm22 of MEA performance at 0.9 ViR-free under O2 [33]. Jiao et al. reported another synthesis method using CVD where interaction between gaseous iron(III) chloride (FeCl3) and ZIF-8 precursor effectively converted the ZnN4 sites into FeN4 sites during thermal activation at 750 C to form FeNC-CVD-750 [34]. The resulting Fe-N-C catalyst exhibited high ORR activity and produced 33 mA cm22 at 0.9 ViR-free under O2. Wu and his coworkers recently reported record performance using Fe-N-C catalyst synthesized from a Fe2O3@ZIF-8 composite precursor [35]. Later, the catalyst was heat-treated at 1100 C with ammonium chloride (AC), labeled as Fe-N-C-AC. This treatment increased micropore and mesopore volumes for this catalyst, increasing MEA performance. The authors reported an initial performance of MEA is 44.2 mA cm22 at 0.9 ViR-Free under O2 and a very highpower density of 601 mW cm22 in the air [35]. The authors additionally deposited a thin layer of nitrogen
carbon species via CVD, which enhanced catalyst durability, with only 30 mV loss at 0.8 A cm22 after 30,000 squarewave voltage cycles under H2
air. End of durability test performance is higher than Fe-N-C-AC catalysts, even comparable with Pt catalysts, although the initial performance is lower than Fe-N-C-AC catalysts [35]. The kinetic performance at 0.9 ViR-Free under O2 and durability in voltage cycling meet US Department of Energy (DOE) performance metrics for PGM-free catalysts (Fig. 6.1). These performance targets were achieved

182

Fuel Cells for Transportation

using different catalysts and MEAs, but the ultimate DOE target is to achieve these goals using the same catalyst in a single MEA.

6.3 Integration of platinum group metal-free catalyst in membrane electrode assembly PGM-free cathodes’ volumetric activity (A m23) is much lower than PGMbased cathodes due to lower active sites and/or low turnover frequency. So, to get comparable performance, the cathode thickness of PGM-free cathodes should be ten times higher than PGM-based cathodes [15]. Fabrication of this thick electrode is very critical. The following factors must be considered during MEA fabrication: 1. 2. 3. 4. 5. 6.

Creating macropore and mesopore, Ionomer loading, Ionomer equivalent weight (EW), Dispersion of ionomer in solvent, Controlling catalyst primary particle size, Hydrophobicity of cathode.

The morphology of the catalyst impacts fuel cell performance, as the micropores (diameter , 2 nm) are believed to host FeN4 active sites [21]. Penetration of Nafion micelles (size range 1
5 nm) through these tiny pores is challenging [13], causing lower catalyst utilization. A wide pore size distribution, including mesopores (diameter 2
50 nm) and macropores (diameter . 50 nm), is required for better utilization of catalysts and improved transport of reactants and water [9]. Shui et al. emphasized the presence of macropores for better transport of reactants and products in and out of the thick catalyst layer [36]. Researchers worked rigorously to control the morphology of catalysts, as discussed in the previous section. In a typical PGM-free electrode preparation [10,37
39], first, the catalyst and ionomer slurry are prepared. Ionomer is dispersed in water/n-propanol solvent. The mixture is then sonicated in a bath sonicator. Sonication time varies from a couple of minutes to hours, depending on catalyst materials. The catalyst ink is then coated onto a gas diffusion layer (GDL) using a manual doctor blade technique [10,37
39], brush painting [8], or electrospinning [40]. The catalyst-coated GDL is used as the cathode in the MEA. A commercial Pt-catalyzed gas diffusion electrode (GDE) or Pt-coated membrane is used for the anode. The cathode and the anode are hot-pressed (500 psi) onto a membrane at 130 C for 4 minutes [10]. GDE is prepared instead of a catalyst-coated membrane (CCM) for better water management by reducing large interfacial voids between microporous layer (MPL) and the catalyst layer [41]. In the case of CCM, there will be an interface between MPL and the catalyst layer with large voids and rough surfaces, which may cause flooding [41].

Platinum group metal-free catalysts for fuel cells Chapter | 6

183

6.3.1 Effect of ionomer loading, equivalent weight of ionomer, and dispersion of ionomer in solvent Babu et al. used comprehensive characterization and modeling of Zelenay’s [26] best-in-class CM-PANI-Fe catalyst to evaluate electrode-level losses and showed that thick cathode challenges persist even with a dramatic 40X increase in active site activity or density [42,43]. For example, at 0.7 V, the optimum thickness with a 40X increase in active site density still requires a 5
10X thicker cathode than that for Pt [43]. Parameter sensitivity studies for power density at 0.7 V indicate that the leading factors limiting power density are the hydrophobicity and proton conductivity of the electrode and the tortuosity of the ionomer (Fig. 6.3A) [43]. So, there is a need to engineer electrode design. Ionomer distribution in electrode is very critical for better fuel cell performance [9,10,42,44,45]. The ionomer distribution depends on ionomer loading, ionomer dispersion in the solvent used in catalyst preparation,

FIGURE 6.3 (A) Impact of different parameters for the high-performing MEA at 0.7 V. (B) Change of Nafion volume fraction and film thickness with the increase of Nafion loading. (C) MEA performance of CM-PANI-Fe-C cathodes with the change of Nafion content from 30 wt.% to 60 wt.%. (D) Change of cathode catalyst layer (CCL) ionic conductance with EW of ionomer for 40 wt.% (dashed line) or 30 wt.% (dot dashed line) ionomer loading. (E) Change of cell performance at 0.5 A cm22 and CCL ionic conductance with relative humidity. (F) Change of electrode transport resistance with electrode proton resistance with varying ink compositions, ionomer contents, fabrication methods, and relative humidity. (A
C) Reprinted with permission from S.K. Babu, H.T. Chung, P. Zelenay, S. Litster, Modeling Electrochemical performance of the hierarchical morphology of precious group metal-free cathode for polymer electrolyte fuel cell. J. Electrochem. Soc. 164 (2017) F1037
F1049; (D
E) D. Banham, T. Kishimoto, T. Sato, Y. Kobayashi, K. Narizuka, J. ichi Ozaki, Y. Zhou, E. Marquez, K. Bai, S. Ye, New insights into non-precious metal catalyst layer designs for proton exchange membrane fuel cells: improving performance and stability. J. Power Sources. 344 (2017) 39
45. Copyright (2017) Elsevier. (F) L. Osmieri, J. Park, D.A. Cullen, P. Zelenay, D.J. Myers, K.C. Neyerlin, Status and challenges for the application of platinum group metal-free catalysts in proton-exchange membrane fuel cells. Curr. Opinion. Electrochem. 25 (2021) 100627. Copyright (2021) Elsevier.

184

Fuel Cells for Transportation

surface chemistry, and electrode morphology [46]. Babu et al. [42] reported a significant change in fuel cell performance due to ionomer loading changes. At low ionomer loading (35 wt.%), low activity was observed, which was attributed to insufficient ionomer presence inside the catalyst layer. Insufficient ionomer presence reduces the ionic conductivity of the electrode [44]. At high ionomer loading (60 wt.%), excess ionomer created hydrophilic zones inside the catalyst layer resulting in flooding (Fig. 6.3B). Optimum performance was observed at 50 wt.% ionomer loading (Fig. 6.3C) [42]. Authors used 1100 EW ionomer in their MEAs. Banham et al. [9] reported a very high performance of the Fe-N-C electrode at an industrial scale (MEA of 50 cm2) with a peak power density of 570 mW cm22 under H2
air. They used 700 EW ionomer, and the optimum ionomer content of their catalyst was typically higher than 40 wt.%. So, optimum ionomer loading can be different depending on catalyst materials and EW of the ionomer. Ionomer is used to conduct protons into a triple-phase boundary [43]. Proton conductivity will increase with the decrease of EW of ionomer, but there is a critical limit of conductance (100 S) above which performance is independent of catalyst layer conductance (Fig. 6.3D
E) [44]. On the other hand, low EW ionomer causes severe flooding, which can be mitigated by lowering the ionomer to carbon (I/C) ratio and using an IPA-rich solvent [47]. Shawn et al. reported lower (50%) proton transport resistance with 725 EW ionomer than 1100 EW ionomer, both at I/C of 0.6. With a combination of I/C and IPA-rich ink solvent, they reported a 50% increase in limiting current density [47]. Osmieri et al. [48] extensively studied the dispersion of ionomer in solvent (water/n-propanol content) on gas transport and fuel cell performance using various in situ and ex situ techniques. They reported a change in ionomer distribution around catalyst agglomerates depending on the ink water to n-propanol ratio. Electrode prepared using a water-rich solvent (82 wt.% H2O) produced high performance at high RH, whereas performance was higher for electrodes prepared using 50 wt.% H2O solvent at low RH (Fig. 6.3F). They used 1100 EW ionomer and catalyst produced by Pajarito Powder LLC. Depending on the EW of ionomer and catalyst properties, the solvent composition of the best-performing electrode can be different, as reported by Banham et al. [9] and Litster et al. [47].

6.3.2

Effect of primary particle size

Uddin et al. [10] established an understanding of the effect of primary particle size, agglomerates, and ionomer integration on the fuel cell performance of PGM-free catalysts. It was found that particle size and ionomer distribution significantly affected fuel cell performance. Catalyst with a smaller primary particle (40 nm) showed the highest activity in the rotating disk

Platinum group metal-free catalysts for fuel cells Chapter | 6

185

electrode (RDE) experiment, but in the fuel cell environment, the smallersize catalyst was not the best-performing catalyst, and the fuel cell performance was not even near the best-performing one [10]. The lower performance of smaller-size catalysts was due to particle agglomeration and the formation of fused aggregates, where the ionomer could not infiltrate. The lack of ionomer within the catalyst aggregates hindered the utilization of the catalyst and resulted in lower fuel cell performance. Significantly greater catalyst utilization (i.e., higher cathode activity) was achieved through improved ionomer integration with the catalyst aggregates (Fig. 6.4A). Small primary particles with better ionomer infiltration showed the best fuel cell performance (Fig. 6.4B). In total, 100-nm-size primary particles showed better ionomer infiltration and better fuel cell performance and achieved 610 mW cm22 peak power density under automotive operating conditions (Fig. 6.4C).

6.3.3

Engineering cathode to improve water management

The typical catalyst loading of PGM-free cathodes is 4 mg cm22, and the thickness of the electrode is around 80
100 μm depending on the catalysts. This thick electrode causes mass transport losses in the high current density

FIGURE 6.4 (A) Optimization of ionomer integration. (B) MEA performance of Fe-MOF catalysts with primary particles ranging from 40 to 600 nm at 80 C, 100% RH, Nafion 212 membrane, and 1 atm H2
air partial pressure. (C) High MEA performance in air and O2 for 100 nm size catalyst at 94 C, 100% RH, Nafion 211 membrane, and 1.7 atm H2
air partial pressure. Reprinted with permission from A. Uddin, L. Dunsmore, H. Zhang, L. Hu, G. Wu, S. Litster, High power density platinum group metal-free cathodes for polymer electrolyte fuel cells. ACS Appl. Mater. Interfaces 12 (2020) 2216
2224. Copyright (2020) American Chemical Society.

186

Fuel Cells for Transportation

FIGURE 6.5 (A) Interfaces of electrodes prepared using CCM, GDE-CCM, and GDE. (B) MEA performance of three electrodes at 45 C and no back pressure. (C) Micro-CT images of the MEA with perforation spacing of 500 μm. (D) MEA performances of perforated GDLs in air at 80 C and 100% RH. (A, B) Reprinted with permission from J. Liu, M.R. Talarposhti, T. Asset, D.C. Sabarirajan, D.Y. Parkinson, P. Atanassov, I. V. Zenyuk, Understanding the role of interfaces for water management in platinum group metal-free electrodes in polymer electrolyte fuel cells. ACS Appl. Energy Mater. 2 (2019) 3542
3553. Copyright (2019) American Chemical Society; (C, D) L. Dunsmore, A. Uddin, H. Zhang, G. Wu, S. Litster, Non-planar platinum group metal-free fuel cell cathodes for enhanced oxygen transport and water rejection. J. Power Sources 506 (2021) 230188. Copyright (2021) Elsevier.

region [9]. Low catalyst-loaded cathode (1 mg cm22) shows a low mass transport effect, but performance is very low in the kinetic region due to fewer active sites [9]. In thick electrodes, water accumulation happens, known as electrode flooding [41]. Liu et al. [41] studied flooding phenomena using in situ and operando X-ray computed tomography and reported that interfacial inhomogeneity (Fig. 6.5A) plays a vital role in water management and the performance of fuel cells (Fig. 6.5B). The primary purpose of the MPL is to prevent water accumulation in electrodes and to facilitate oxygen transport [49]. The MPL is made of carbon powder bound with polytetrafluoroethylene (PTFE) and consists of tiny hydrophobic pores (  60 nm) [49]. These pores limit the amount of liquid water that cannot be removed from the cathode due to significant capillary forces, and it is often impossible without cracks in MPL [50,51]. Additionally, in the PGM cathode, due to high current density, heat generation is high enough to vaporize water, and the vapor transports through the

Platinum group metal-free catalysts for fuel cells Chapter | 6

187

pores [51,52]. However, in the PGM-free cathode, due to low volumetric current density, low heat generation occurs, and liquid water saturation increases. Dunsmore et al. used laser-perforated MPL to incorporate a low capillary pressure pathway in the electrode [39]. They prepared GDE by coating catalyst ink onto the perforated GDL, where the catalyst filled the perforation. PGM-free catalyst penetrated inside the perforation is hydrophilic, and these hydrophilic pathways create lower capillary pressure for water removal. This creates a separate transport path for water through perforation and oxygen through other regions (Fig. 6.5C). This reduces saturation in the thick CCL, eliminating one of the biggest performance obstacles for PGM-free catalysts. The reduction of saturation reduces oxygen transport resistance and improves performance (Fig. 6.5D). Additionally, to increase hydrophobicity in the electrode, Dunsmore et al. adopted several strategies to improve the hydrophobicity of the electrode by incorporating PTFE particles into the catalyst slurries, infiltrating catalyst slurries into ePTFE and hydrophobic carbon paper, and initiating CVD of hydrophobic polymers onto the catalyst layer, but due to low conductivity, performance did not improve compared to baseline [53]. Slack et al. [40] incorporated hydrophobic polyvinylidene fluoride (PVDF) binder in PGMfree based electrospun nanofiber cathode and found better performance than fibrous cathode without PVDF binder, but overall performance was lower than the conventional cathode. A graded ionomer concept is also proposed to optimize proton transport and oxygen transport [7]. The primary reaction zone is close to the membrane [44]. Based on this concept, high ionomer loading is used close to the membrane for better proton transport, and low ionomer loading is close to the MPL to facilitate oxygen transport. The performance of MEA with gradient ionomer content is higher than the one with uniform ionomer content [7]. Spendellow et al. proposed separate channels for proton transport and oxygen transport [54]. In their proton channel electrode, they incorporated a separate, non-tortuous proton transport pathway which reduced proton resistance, enabling a very high performance of 478 mA cm22 at 0.7 V compared to 361 mA cm22 for the baseline electrode, a 32% increase in current density at 100% RH.

6.4 Stability and durability of platinum group metal-free cathode So far, researchers have put all their efforts into developing high-performing catalysts, and their initial performance is comparable to PGM catalysts. Still, the overall stability of these catalysts remains poor. During stability testing, it is observed that there is a rapid initial decay lasting about 10
15 hours followed by a more gradual decay (Fig. 6.6A
B) [10,12,15]. Four primary

188

Fuel Cells for Transportation

(B)

400

H2-Air, 80°C

HFR-free Voltage (V)

Current Density (mA/cm2) @0.7 V

(A)

300 200 1 mA/cm 2 /h

100 0

0

10

20

30

1

H2-Air, 80°C

0.8 0.6

0h 12 h 25 h 40 h

0.4 0.2 0

40

0

0.2

0.4

0.6

0.8

1

1.2

Current Density (A/cm2)

Time (h)

(C)

(D)

1

0.5

s

te

de bi ar C

et

de xi

lic

ol

al

rr Py

lfa

200

O

150

Su

100

Time /h

Su lfi de s

50

om

0

Po

rp

lic al et M

0

ic

0

an

0.1

1.5

rg

0.6 V

MEA 0.6 V 200 h (1.65 wt% Fe) MEA 0.4 V 200 h (1.20 wt% Fe) Fresh MEA (7.40 wt% Fe)

2

O

0.2

2.5

ra zi n Py rid in ic

0.4 V

0.3

hy

Fe in indicated coordination (wt%)

Current density / A cm-2

0.4

FIGURE 6.6 (A) MEA stability test of 100-nm particle size Fe-MOF catalyst at 0.7. (B) MEA performances obtained during the stability test at 80 C, 100% RH, and 1 atm H2
air partial pressure. (C) Stability test of a PANI
Fe
C catalyst-based MEA at 0.4 and 0.6 V in air at 80 C. (D) The weight percent of Fe in fresh MEA with a PANI
Fe
C cathode and MEAs after stability test at 0.4 V and 0.6 V. (A
B) Reprinted with permission from A. Uddin, L. Dunsmore, H. Zhang, L. Hu, G. Wu, S. Litster, High power density platinum group metal-free cathodes for polymer electrolyte fuel cells. ACS Appl. Mater. Interfaces. 12 (2020) 2216
2224. Copyright (2020) American Chemical Society). (C, D) M. Ferrandon, X. Wang, A.J. Kropf, D.J. Myers, G. Wu, C.M. Johnston, P. Zelenay, Stability of iron species in heat-treated polyaniline-iron-carbon polymer electrolyte fuel cell cathode catalysts. Electrochim. Acta. 110 (2013) 282
291. Copyright (2013) Elsevier.

hypotheses are proposed for the instability of the performance:[9] (1) micropore flooding; (2) active site protonation or anion binding; (3) demetallation; and (4) attack by H2O2 or radicals.

6.4.1

Micropore flooding

Dodelet’s group proposed micropore flooding for rapid initial decay of performance [13,55]. It is believed that micropores are hydrophobic at the beginning of fuel cell operation, and later, carbon oxidation leads to micropores becoming hydrophilic. These hydrophilic micropores are flooded by liquid water resulting in rapid and irreversible performance decay [13]. Later, Banham’s group did a systematic analysis of stability testing with a combination of cyclic voltammetry and double-layer capacitance measurement to monitor micropore flooding and concluded that most pores are

Platinum group metal-free catalysts for fuel cells Chapter | 6

189

wetted at the beginning of testing, and loss of the performance is primarily kinetic [12]. Litster’s group also drew a similar conclusion [38]. In a recent publication [56], Dodelet’s group agreed with Banham’s group that micropore flooding might not be the primary reason for rapid initial decay.

6.4.2

Active site protonation

Several authors reported active site protonation or anion binding could be a reason for performance decay [57
59]. Liu et al. [57] first proposed that the protonation of pyridinic-N was the primary mechanism for performance decay and tested the stability of two different catalysts, one rich with pyridinic-N and quaternary-N and the other one rich with only quaternary-N, and found that the catalyst with quaternary-N remained stable. In contrast, there was a performance decay in the catalyst with pyridinic-N and quaternary-N. But the initial performance of the catalyst with pyridinic-N and quaternary-N was two times higher than the catalyst with quaternary-N, and performance remained higher throughout the 100 hours stability test. The high performance throughout the stability test suggests that the protonation of pyridinic-N may not be the main reason for initial performance decay [15]. Herranz et al. proposed another protonation method that active sites like FeN4. . .N protonated and formed FeN4. . .NH1 and subsequently neutralized by SO32 anion resulting in lower ORR activity [59]. The same group failed to prove the hypothesis in a later publication [13].

6.4.3 Demetallation, carbon oxidation, and attack of peroxide and associated radicals Fe demetallation has been reported in several publications [60
62]. Ferrandon et al. [60] extensively tested the stability of iron of PANI
Fe
C catalyst in MEAs for 200 hours at 0.4 and 0.6 V (Fig. 6.6C) and reported iron loss from the catalyst (Fig. 6.6D). Choi et al. [61] concluded that inactive Fe leached at a potential lower than 0.7 V and had no impact on ORR activity, but at a potential higher than 0.9 V, Fe leached out due to oxidation of carbon support destructing FeN4 active sites resulting in significant ORR decay. Several authors talked about degradation by H2O2 and/or associated radical oxygen species generated via a Fenton reaction (Fe21 1 H2O2 1 H1 5 Fe31 1  OH 1 H2O) [13,63
65]. Schulenburg et al. [64] and Goellner et al. [65] attributed the degradation of steady-state performance to the chemical attack on Fe-N-C by H2O2 and radicals. Dodelet’s group [63] demonstrated that even a low level of peroxide decomposed active catalytic sites releasing iron ions in the sulfuric acid solution and losing ORR activity. They attributed performance loss during the stability of MEA to the peroxide

190

Fuel Cells for Transportation

attack and subsequent decomposition of active sites and iron release. Their later publication [13] attributed electrooxidation followed by micropore flooding for rapid initial performance decay. After the criticism from Banham’s group [12] presented in the previous section, they tested the stability of their catalyst at potentials ranging from 0.2 to 0.8 V [56]. They abandoned micropore flooding as a possible cause of initial decay, but they did not agree with Banham group’s hypothesis of oxidation of active sites by H2O2 or radicals. Using M¨ossbauer spectrum and neutron activation analysis, they found a decrease in FeN4-like catalytic sites during stability tests. They hypothesized that FeN4-like catalytic sites were not thermodynamically stable in micropores due to the fast transport of water. They attributed the demetallation of FeN4 active sites inside the micropores for rapid initial decay of performance [56].

6.5

Mitigation strategies

From the previous section, it can be said that the attack of H2O2 and associated radicals are the main reason for the performance degradation of Fe-N-C catalysts. These radicals can damage active sites in two ways [66]: 1. direct oxidation of carbon to CO2, which leads to demetallation of active sites, and 2. formation of oxygen species on the carbon surface. Choi et al. [67] reported that FeN4-like active sites in Fe-N-C catalyst become unstable in acidic media of fuel cell environment due to oxidation of carbon surface by reactive oxygen species. The carbon surface oxidation weakens O2 binding to active sites and eventually deactivates the sites. These sites can be activated by the electrochemical reduction of carbon surfaces [67]. Following Choi et al. [67], Boldrin et al. reported that water and oxygen species are chemisorbed on catalyst surfaces and impact performance during ORR by blocking the carbon surface surrounding active sites or altering local electronic structure [68]. These sites can be activated by heat treatment in Ar at 600 C or electrochemical reduction at 20.3 V versus RHE [68]. Litster’s group reported a similar performance recovery method [47] before Boldrin et al. [68]. In their recent publication [38], the authors proposed a recovery method by applying a low potential (0.05 V) without supplying oxygen in the cathode (Fig. 6.7A). The authors reported two types of active sites in Fe-N-C catalysts: type 1 active sites degrade during stability test by forming Fe21 to Fe31, and type 2 remains unmodified. The recovery method facilitates the cleaning of oxygen from the carbon surfaces and possibly restoring type 1 active sites from Fe31 to Fe21 (Fig. 6.7B) [38]. This work was supported by the recent publication of Li et al. [69]. Using M¨ossbauer, they reported that type 1 active sites are high spin Fe(III)Nx, and

Platinum group metal-free catalysts for fuel cells Chapter | 6

191

FIGURE 6.7 (A) MEA performance at 0.7 V hold in air with recovery voltage applied at 0.05 V in the absence of oxygen for 5 min after every 75 min of 0.7 V hold. (B) Percent increase of site density during recovery. Reprinted with permission from D.E. Beltra´n, A. Uddin, X. Xu, L. Dunsmore, S. Ding, H. Xu, H. Zhang, S. Liu, G. Wu, S. Litster, Elucidation of performance recovery for fe-based catalyst cathodes in fuel cells. Adv. Energy Sustainability Res. 2 (2021) 2100123.

type 2 active sites are low or intermediate spin Fe(II)Nx. So, for durable FeN-C catalysts, type 2 active sites are preferable and/or stabilizing type 1 in an acidic environment. One of the mitigation strategies to improve stability is to scavenge H2O2 and radicals. Ceria nanoparticles (CeO2) are widely used as a peroxide scavenger in Pt-based PEFC [70,71]. Wei et al. reported that the addition of CeO2 during catalyst synthesis could improve the stability of Fe-N-C catalysts [72]. Litster et al. [73] mixed CeO2 nanoparticles with catalysts but did not observe improvement in durability, possibly due to nonuniform mixing and agglomeration of nanoparticles. Xie et al. used tantalum
titanium oxide (Ta
TiOx) nanoparticles as a radical scavenger and reported that these particles suppressed H2O2 yield by 51% at 0.7 V in the RDE test [66]. After the

192

Fuel Cells for Transportation

accelerated stress test, a fuel cell with Ta
TiOx showed only 3% performance decay at 0.9 ViR-free, whereas a fuel cell without Ta
TiOx showed a 33% performance loss [66]. So, Fe-N-C catalyst with Ta
TiOx provided a promising mitigation strategy. Another strategy to reduce the formation of hydroperoxyl radicals is to develop ORR cathode catalysts that are Fe-free, such as Co-N-C catalysts. However, Co-N-C catalyst performance is lower than Fe-based catalysts [37]. So, to combine the high activity of Fe-N-C catalyst and durability of Co-N-C catalyst, binary FeCo-N-C catalysts are also synthesized [47].

6.6

Summary

Low-cost PGM-free catalysts can reduce the cost of fuel cell electric vehicles. The initial performance of these catalysts is competitive with their Pt counterpart, but the stability of these catalysts has not yet reached the commercial level. The current reported kinetic performance at 0.9ViR-Free under O2 and durability in voltage cycling meet US DOE 2025 performance metrics for PGM-free catalysts. These performance targets were achieved using different catalysts and MEAs, but the ultimate DOE target is to achieve these goals using the same catalyst in a single MEA. Researchers are putting a lot of effort into improving performance, understanding the mechanism behind the degradation, and developing mitigation strategies. The hydrophobicity and proton conductivity of the electrode and tortuosity of the ionomer are the leading factors affecting the utilization of catalysts. Oxidation of carbon support and the chemical attack of H2O2 and associated radicals are the primary reasons for the performance degradation of Fe-N-C catalysts. Several mitigation strategies are proposed, including the reduction of oxidized carbon and the addition of radical scavengers.

Acknowledgments The authors would like to thank Dr. Prodip Das for his invitation to write this chapter. The authors also acknowledge the financial support through the Basic Research Grant (R-60/ Re-5336, 30/6/2021) from Bangladesh University of Engineering and Technology. Finally, the authors thank the editors for their helpful guidance and input.

References [1] UNFCCC, COP 26 Glasgow Climate Pact, Cop26, 2021. ,https://unfccc.int/sites/default/ files/resource/cop26_auv_2f_cover_decision.pdf . . [2] S. Satyapal, U.S. Department of Energy Hydrogen and Fuel Cell Technology Overview, FC EXPO 2020, Tokyo, Japan. ,https://www.energy.gov/sites/default/files/2020/03/f72/ fcto-satyapal-fcexpo-feb20.pdf . .

Platinum group metal-free catalysts for fuel cells Chapter | 6

193

[3] R.L. Borup, A. Kusoglu, K.C. Neyerlin, R. Mukundan, R.K. Ahluwalia, D.A. Cullen, et al., Recent developments in catalyst-related PEM fuel cell durability, Curr. Opin. Electrochem. 21 (2020) 192
200. [4] S.T. Thompson, D. Papageorgopoulos, Platinum group metal-free catalysts boost cost competitiveness of fuel cell vehicles, Nat. Catal. 2 (2019) 558
561. [5] D.A. Cullen, K.C. Neyerlin, R.K. Ahluwalia, R. Mukundan, K.L. More, R.L. Borup, et al., New roads and challenges for fuel cells in heavy-duty transportation, Nat. Energy 6 (2021) 462
474. [6] M. Lefe`vre, J.P. Dodelet, Recent advances in non-precious metal electrocatalysts for oxygen reduction in PEM fuel cells, ECS Trans. 45 (2012) 35
44. [7] Y. He, S. Liu, C. Priest, Q. Shi, G. Wu, Atomically dispersed metal-nitrogen-carbon catalysts for fuel cells: advances in catalyst design, electrode performance, and durability improvement, Chem. Soc. Rev. 49 (2020) 3484
3524. [8] H.T. Chung, D.A. Cullen, D. Higgins, B.T. Sneed, E.F. Holby, K.L. More, et al., Direct atomic-level insight into the active sites of a high-performance PGM-free ORR catalyst, Science 357 (2017) 479
484. [9] D. Banham, T. Kishimoto, Y. Zhou, T. Sato, K. Bai, J. Ozaki, et al., Critical advancements in achieving high power and stable nonprecious metal catalyst-based MEAs for real-world proton exchange membrane fuel cell applications, Sci. Adv. 4 (3) (2018) 1
8. [10] A. Uddin, L. Dunsmore, H. Zhang, L. Hu, G. Wu, S. Litster, High power density platinum group metal-free cathodes for polymer electrolyte fuel cells, ACS Appl. Mater. Interfaces 12 (2020) 2216
2224. [11] Y. He, G. Wu, PGM-free oxygen-reduction catalyst development for proton-exchange membrane fuel cells: challenges, solutions, and promises, ACC. Mater. Res. 3 (2022) 224
236. [12] J. Choi, L. Yang, T. Kishimoto, X. Fu, S. Ye, Z. Chen, et al., Is the rapid initial performance loss of Fe/N/C non precious metal catalysts due to micropore flooding? Energy Environ. Sci. 10 (2017) 296
305. [13] G. Zhang, R. Chenitz, M. Lefe`vre, S. Sun, J.P. Dodelet, Is iron involved in the lack of stability of Fe/N/ C electrocatalysts used to reduce oxygen at the cathode of PEM fuel cells? Nano Energy 29 (2016) 111
125. [14] G. Liu, X. Li, J.W. Lee, B.N. Popov, A review of the development of nitrogen-modified carbon-based catalysts for oxygen reduction at USC, Catal. Sci. Technol. 1 (2) (2011) 207
217. [15] D. Banham, S. Ye, K. Pei, J.I. Ozaki, T. Kishimoto, Y. Imashiro, A review of the stability and durability of non-precious metal catalysts for the oxygen reduction reaction in proton exchange membrane fuel cells, J. Power Sources 285 (2015) 334
348. [16] R. Jasinski, A new fuel cell cathode catalyst, Nature 201 (1964) 1212
1213. [17] H. Jahnke, M. Sch¨onborn, G. Zimmermann, Organic dyestuffs as catalysts for fuel cells, Top. Curr. Chem. 61 (1976) 133
181. [18] V.S. Bagotzky, M.R. Tarasevich, K.A. Radyushkina, O.A. Levina, S.I. Andrusyova, Electrocatalysis of the oxygen reduction process on metal chelates in acid electrolyte, J. Power Sources 2 (1978) 233
240. [19] J.A.R. van Veen, H.A. Colijn, Oxygen reduction on transition-metal porphyrins in acid electrolyte II. Stability, Berichte Der Bunsenges. F¨ur Phys. Chem. 85 (1981) 700
704. [20] S. Gupta, D. Tryk, I. Bae, W. Aldred, E. Yeager, Heat-treated polyacrylonitrile-based catalysts for oxygen electroreduction, J. Appl. Electrochem. 19 (1989) 19
27.

194

Fuel Cells for Transportation

[21] M. Lefe`vre, E. Proietti, F. Jaouen, J.P. Dodelet, Iron-based catalysts with improved oxygen reduction activity in polymer electrolyte fuel cells, Science 324 (2009) 71
74. [22] E. Proietti, F. Jaouen, M. Lefe`vre, N. Larouche, J. Tian, J. Herranz, et al., Iron-based cathode catalyst with enhanced power density in polymer electrolyte membrane fuel cells, Nat. Commun 2 (2011). [23] A. Serov, M.H. Robson, M. Smolnik, P. Atanassov, Templated bi-metallic non-PGM catalysts for oxygen reduction, Electrochim. Acta 80 (2012) 213
218. [24] A. Serov, K. Artyushkova, P. Atanassov, Fe-N-C oxygen reduction fuel cell catalyst derived from carbendazim: synthesis, structure, and reactivity, Adv. Energy Mater. 4 (2014) 1301735. [25] M.J. Workman, A. Serov, L. Tsui, P. Atanassov, K. Artyushkova, Fe
N
C catalyst graphitic layer structure and fuel cell performance, ACS Energy Lett. 2 (2017) 1489
1493. [26] G. Wu, K.L. More, C.M. Johnston, P. Zelenay, High-performance electrocatalysts for oxygen reduction derived from polyaniline, iron, and cobalt, Science 332 (2011) 443
447. [27] Q. Li, G. Wu, D.A. Cullen, K.L. More, N.H. Mack, H.T. Chung, et al., Phosphate-tolerant oxygen reduction catalysts, ACS Catal. 4 (2014) 3193
3200. [28] S. Ma, G.A. Goenaga, A.V. Call, D.J. Liu, Cobalt imidazolate framework as precursor for oxygen reduction reaction electrocatalysts, Chem. Eur. J. 17 (2011) 2063
2067. [29] Q. Lai, L. Zheng, Y. Liang, J. He, J. Zhao, J. Chen, et al., Metal-organic-frameworkderived Fe-N/C electrocatalyst with five-coordinated Fe-Nx sites for advanced oxygen reduction in acid media, ACS Catal. 7 (2017) 1655
1663. [30] J. Herranz, F. Jaouen, M. Lefe`vre, U.I. Kramm, E. Proietti, J.P. Dodelet, et al., Unveiling N-protonation and anion-binding effects on Fe/N/C catalysts for O2 reduction in protonexchange-membrane fuel cells, J. Phys. Chem. C. 115 (2011) 16087
16097. [31] H. Zhang, S. Hwang, M. Wang, Z. Feng, S. Karakalos, L. Luo, et al., Single atomic iron catalysts for oxygen reduction in acidic media: particle size control and thermal activation, J. Am. Chem. Soc. 139 (2017) 14143
14149. [32] H. Zhang, H.T. Chung, D.A. Cullen, S. Wagner, U.I. Kramm, K.L. More, et al., Highperformance fuel cell cathodes exclusively containing atomically dispersed iron active sites, Energy Environ. Sci. 12 (2019) 2548
2558. [33] S. Liu, M. Wang, X. Yang, Q. Shi, Z. Qiao, M. Lucero, et al., Chemical vapor deposition for atomically dispersed and nitrogen coordinated single metal site catalysts, Angew. Chem. Int. Ed. 59 (2020) 21698
21705. [34] L. Jiao, J. Li, L.L.R. Richard, Q. Sun, T. Stracensky, E. Liu, et al., Chemical vapour deposition of Fe
N
C oxygen reduction catalysts with full utilization of dense Fe
N4 sites, Nat. Mater. 20 (2021) 1385
1391. [35] S. Liu, C. Li, M.J. Zachman, Y. Zeng, H. Yu, J. Kropf, et al., Durable and high-power iron-based cathodes in competition with platinum for proton-exchange membrane fuel cells. ChemRxiv. 2021. [36] J. Shui, C. Chen, L. Grabstanowicz, D. Zhao, D.J. Liu, Highly efficient nonprecious metal catalyst prepared with metal
organic framework in a continuous carbon nanofibrous network, Proc. Natl. Acad. Sci 112 (2015) 10629
10634. Available from: https://doi.org/ 10.1073/pnas.1507159112. [37] Y. He, S. Hwang, D.A. Cullen, M.A. Uddin, L. Langhorst, et al., Highly active atomically dispersed CoN4 fuel cell cathode catalysts derived from surfactant-assisted MOFs: carbonshell confinement strategy, Energy Environ. Sci. 12 (2019) 250
260. [38] D.E. Beltra´n, A. Uddin, X. Xu, L. Dunsmore, S. Ding, H. Xu, et al., Elucidation of performance recovery for Fe-based catalyst cathodes in fuel cells, Adv. Energy Sustainability Res. 2 (2021) 2100123.

Platinum group metal-free catalysts for fuel cells Chapter | 6

195

[39] L. Dunsmore, A. Uddin, H. Zhang, G. Wu, S. Litster, Non-planar platinum group metalfree fuel cell cathodes for enhanced oxygen transport and water rejection, J. Power Sources 506 (2021) 230188. [40] J. Slack, B. Halevi, G. McCool, J. Li, R. Pavlicek, R. Wycisk, et al., Electrospun fiber mat cathode with platinum-group-metal-free catalyst powder and Nafion/PVDF binder, ChemElectroChem 5 (2018) 1537
1542. [41] J. Liu, M.R. Talarposhti, T. Asset, D.C. Sabarirajan, D.Y. Parkinson, P. Atanassov, et al., Understanding the role of interfaces for water management in platinum group metal-free electrodes in polymer electrolyte fuel cells, ACS Appl. Energy Mater. 2 (2019) 3542
3553. [42] S.K. Babu, H.T. Chung, P. Zelenay, S. Litster, Resolving electrode morphology’s Impact on Platinum group metal-free cathode performance using nano-CT of 3D hierarchical pore and ionomer distribution, ACS Appl. Mater. Interfaces 8 (2016) 32764
32777. [43] S.K. Babu, H.T. Chung, P. Zelenay, S. Litster, Modeling electrochemical performance of the hierarchical morphology of precious group metal-free cathode for polymer electrolyte fuel cell, J. Electrochem. Soc. 164 (2017) F1037
F1049. [44] D. Banham, T. Kishimoto, T. Sato, Y. Kobayashi, K. Narizuka, J.I. Ozaki, et al., New insights into non-precious metal catalyst layer designs for proton exchange membrane fuel cells: improving performance and stability, J. Power Sources 344 (2017) 39
45. [45] L. Osmieri, J. Park, D.A. Cullen, P. Zelenay, D.J. Myers, K.C. Neyerlin, Status and challenges for the application of platinum group metal-free catalysts in proton-exchange membrane fuel cells, Curr. Opin. Electrochem. 25 (2021) 100627. [46] K. Artyushkova, M.J. Workman, I. Matanovic, M.J. Dzara, C. Ngo, S. Pylypenko, et al., Role of surface chemistry on catalyst/ionomer interactions for transition metal-nitrogencarbon electrocatalysts, ACS Appl. Energy Mater. 1 (1) (2018) 68
77. [47] S. Litster, Advanced PGM-free cathode engineering for high power density and durability. US DOE Annual Merit Review Meeting, 2020. [48] L. Osmieri, G. Wang, F.C. Cetinbas, S. Khandavalli, J. Park, S. Medina, et al., Utilizing ink composition to tune bulk-electrode gas transport, performance, and operational robustness for a Fe
N
C catalyst in polymer electrolyte fuel cell, Nano Energy 75 (2020) 104943. [49] J.P. Owejan, J.E. Owejan, W. Gu, T.A. Trabold, T.W. Tighe, M.F. Mathias, Water transport mechanisms in PEMFC gas diffusion layers, J. Electrochem. Soc. 157 (2010) B1456. [50] I.V. Zenyuk, E.C. Kumbur, S. Litster, Deterministic contact mechanics model applied to electrode interfaces in polymer electrolyte fuel cells and interfacial water accumulation, J. Power Sources 241 (2013) 379
387. [51] I.V. Zenyuk, R. Taspinar, A.R. Kalidindi, E.C. Kumbur, S. Litster, Computational and experimental analysis of water transport at component interfaces in polymer electrolyte fuel cells, J. Electrochem. Soc. 161 (2014) F3091
F3103. [52] J. Lee, R. Yip, P. Antonacci, N. Ge, T. Kotaka, Y. Tabuchi, et al., Synchrotron investigation of microporous layer thickness on liquid water distribution in a PEM fuel cell, J. Electrochem. Soc. 162 (2015) F669
F676. [53] L. Dunsmore, Tackling Challenges Related to Thick Electrodes in Platinum Group Metal Free Polymer Electrolyte Fuel Cells, Carnegie Mellon University, 2021. [54] J.S. Spendelow, Accessible PGM-Free Catalysts and Electrodes, US DOE Annual Merit Review Meeting, 2021. [55] L. Yang, N. Larouche, R. Chenitz, G. Zhang, M. Lefe`vre, J.P. Dodelet, Activity, performance, and durability for the reduction of oxygen in PEM fuel cells, of Fe/ N/C electrocatalysts obtained from the pyrolysis of metal-organic-framework and iron porphyrin precursors, Electrochim. Acta 159 (2015) 184
197.

196

Fuel Cells for Transportation

[56] R. Chenitz, U.I. Kramm, M. Lefevre, V. Glibin, G. Zhang, S. Sun, et al., A specific demetalation of Fe-N4 catalytic sites in the micropores of NC_Ar 1 NH3 is at the origin of the initial activity loss of the highly active Fe/N/C catalyst used for the reduction of oxygen in PEM fuel cells, Energy Environ. Sci. 11 (2018) 365
382. Available from: https://doi. org/10.1039/C7EE02302B. [57] G. Liu, X. Li, P. Ganesan, B.N. Popov, Studies of oxygen reduction reaction active sites and stability of nitrogen-modified carbon composite catalysts for PEM fuel cells, Electrochim. Acta 55 (2010) 2853
2858. [58] G. Liu, X. Li, B.N. Popov, Stability study of nitrogen-modified carbon composite catalysts for oxygen reduction reaction in polymer electrolyte membrane fuel cells, ECS Trans. 25 (2009) 1251
1259. [59] J. Herranz, F. Jaouen, M. Lef, U.I. Kramm, E. Proietti, J.P. Dodelet, P. Bogdano, et al., Unveiling N-protonation and anion-binding effects on Fe/N/C catalysts for O2 reduction in proton-exchange-membrane fuel cells, J. Phys. Chem. C 115 (2011) 16087
16097. Available from: https://doi.org/10.1021/jp2042526. [60] M. Ferrandon, X. Wang, A.J. Kropf, D.J. Myers, G. Wu, C.M. Johnston, et al., Stability of iron species in heat-treated polyaniline-iron-carbon polymer electrolyte fuel cell cathode catalysts, Electrochim. Acta 110 (2013) 282
291. [61] C.H. Choi, C. Baldizzone, J. Grote, A.K. Schuppert, F. Jaouen, K.J.J. Mayrhofer, Stability of Fe-N-C catalysts in acidic medium studied by operando spectroscopy, Angew. Chem. Int. Ed. 127 (2015) 12944
12948. Available from: https://doi.org/10.1002/anie.201504903. [62] C.H. Choi, C. Baldizzone, G. Polymeros, E. Pizzutilo, O. Kasian, A.K. Schuppert, et al., Minimizing operando demetallation of Fe-N-C electrocatalysts in acidic medium, ACS Catal. 6 (2016) 3136
3146. [63] J.P. Dodelet, M. Lefe`vre, Fe-based catalysts for the reduction of oxygen in polymer electrolyte membrane fuel cell conditions: determination of the amount of peroxide released during electroreduction and its influence on the stability of the catalysts, Electrochim. Acta 48 (19) (2003) 2749
2760. Available from: https://doi.org/10.1016/S0013-4686(03) 00393-1. [64] H. Schulenburg, S. Stankov, V. Schu, I. Dorbandt, S. Fiechter, P. Bogdanoff, et al., Catalysts for the oxygen reduction from heat-treated iron (III) tetramethoxyphenylporphyrin chloride: structure and stability of active sites, J. Phys. Chem. B 107 (34) (2003) 9034
9041. Available from: https://doi.org/10.1021/jp030349j. [65] V. Goellner, V. Armel, A. Zitolo, E. Fonda, Degradation by hydrogen peroxide of metalnitrogen-carbon catalysts for oxygen reduction, J. Electrochem. Soc. 162 (6) (2015) H403
H414. Available from: https://doi.org/10.1149/2.1091506jes. [66] H. Xie, X. Xie, G. Hu, V. Prabhakaran, S. Saha, L. Gonzalez-Lopez, et al., Ta
TiOx nanoparticles as radical scavengers to improve the durability of Fe
N
C oxygen reduction catalysts, Nat. Energy 7 (2022) 281
289. [67] C.H. Choi, H.K. Lim, M.W. Chung, G. Chon, N.R. Sahraie, A. Altin, et al., The Achilles’ heel of iron-based catalysts during oxygen reduction in an acidic medium, Energy Environ. Sci. 11 (2018) 3176
3182. [68] P. Boldrin, D. Malko, A. Mehmood, U.I. Kramm, S. Wagner, S. Paul, et al., Deactivation, reactivation and super-activation of Fe-N/C oxygen reduction electrocatalysts: gas sorption, physical and electrochemical investigation using NO and O2, Appl. Catal. B: Environ. 292 (2021) 120169.

Platinum group metal-free catalysts for fuel cells Chapter | 6

197

[69] J. Li, M.T. Sougrati, A. Zitolo, J.M. Ablett, I.C. O˘guz, T. Mineva, et al., Identification of durable and non-durable FeNx sites in Fe
N
C materials for proton exchange membrane fuel cells, Nat. Catal. 4 (2021) 10
19. [70] V. Prabhakaran, C.G. Arges, V. Ramani, Investigation of polymer electrolyte membrane chemical degradation and degradation mitigation using in situ fluorescence spectroscopy, Proc. Natl Acad. Sci. 109 (4) (2012) 1029
1034. Available from: https://doi.org/10.1073/ pnas.1114672109. [71] P. Trogadas, J. Parrondo, V. Ramani, Degradation mitigation in polymer electrolyte membranes using cerium oxide as a regenerative free-radical scavenger, Electrochem. SolidState Lett. 11 (7) (2008) B113
B116. Available from: https://doi.org/10.1149/1.2916443. [72] H. Wei, X. Su, J. Liu, J. Tian, Z. Wang, K. Sun, et al., modified phenylenediamine-based Fe/N/C with enhanced durability/stability as non-precious metal catalyst for oxygen reduction reaction, Electrochem. Commun. 88 (2018) 19
23. [73] S. Litster, Advanced PGM-Free Cathode Engineering for High Power Density and Durability. US DOE Annual Merit Review Meeting, 2019.

This page intentionally left blank

Chapter 7

Effective transport properties for fuel cells: modeling and experimental characterization Pablo A. Garcı´a-Salaberri1 and Prodip K. Das2 1

Departamento de Ingenierı´a Te´rmica y de Fluidos, Universidad Carlos III de Madrid, Madrid, Spain, 2School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom

7.1

Introduction

The demand to reduce the use of hydrocarbons in the automotive sector has intensified in the last decades to mitigate climate change [1,2]. Currently, battery electric vehicles (BEVs) sales exceed 1 million annually, while fuel cell electric vehicle (FCEV) sales barely exceed 15,000 units around the world (B1%). This sharp difference is explained by the lower level of deployment of hydrogen infrastructure compared to electric vehicle charging stations (about a decade late), together with the current higher capital and operational cost of FCEVs [3]. However, FCEVs present several advantages over BEVs, such as fast refueling times, higher range (B600 km between refueling), larger longevity (above 200,000 km), and better driver experience (similar operation to conventional internal combustion engines with lower anxiety related to lack of range). Thus the FCEV market is expected to grow gradually in the coming years as production costs on a larger scale are reduced and the availability of hydrogen fueling stations is increased. Similar costs to BEVs may be achieved by 2030. Currently, most major automotive companies have started to commercialize FCEVs, such as Toyota Mirai, Hyundai Nexo and Honda Clarity, or have included the development of FCEVs in their R&D plans (e.g., Ford Motor Company and Mercedes) [4]. The main barriers hindering widespread commercialization of FCEVs (apart from the scarce development of hydrogen infrastructure) are: (1) high cost associated with the use of platinum-based catalysts (and other uncommon materials), (2) insufficient performance and durability, and (3) lack of economies of scale in FCEV production. In fact, all these aspects are interrelated since high cost and low performance and durability prevent the creation Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00011-3 © 2023 Elsevier Ltd. All rights reserved.

199

200

Fuel Cells for Transportation

of economies of scale. Changing this situation requires optimization of components (higher durability and performance at lower cost) used in polymer electrolyte fuel cells (PEFCs). Key aspects to be addressed are the reduction of Pt loading and the development of durable components with improved effective transport properties and electrochemical activity (especially for oxygen reduction at the cathode) [5]. As shown in Fig. 7.1, catalyst represents 40% of the cost of a PEFC system, bipolar plates around 30%, and 20% comes from other components (membrane, backing layers and gaskets). The balance of plant (pumps, sensors, compressors, recirculation blowers, etc.) amounts to approximately 10%. One of the main issues that complicates the development of highperformance, durable components is the small thickness and multifunctional character of thin porous layers in PEFCs. For example, backing layers must fulfill several critical functions, such as providing a transport pathway for reactants and products through its pore volume and ensuring charge and heat conduction through its solid matrix. Recently, Toyota Motor Corporation has shown that PEFC performance can be significantly improved by the combined optimization of components and operation [6]. For instance, cathode catalysts experience voltage cycles, variable multiphase conditions, and repeated start-up/shutdown during operation, which lower performance with time. The design of next-generation PEFCs and porous materials therein must be conceived in an integral form, combining electrochemistry, transport in porous media, and multiscale fabrication methods. The aim of this chapter is to analyze the main characteristics and effective transport properties of porous layers present in PEFCs. Modeling and experimental techniques widely used for fuel cell characterization are also discussed. The content is not intended to provide an exhaustive literary review but to provide the reader with an overview of porous layers in fuel cells. The chapter is divided into three sections. Section 7.2 is devoted to the

FIGURE 7.1 Evolution of the cost of an 80-kWnet PEFC system and component cost distribution based on projection to high-volume manufacturing (500,000 units/year).

Effective transport properties for fuel cells Chapter | 7

201

structure and composition of porous components. Section 7.3 is devoted to effective transport properties. Section 7.4 is devoted to modeling and experimental techniques. The concluding remarks are presented in Section 7.5.

7.2 Structure and composition of porous transport layers in fuel cells There are several porous layers in PEFCs. These include gas diffusion layers (GDLs), microporous layers (MPLs), and catalyst layers (CLs). Every PEFC has one layer of each on both sides (anode and cathode sides) of the polymer electrolyte (or membrane). CLs, where the electrochemical reactions take place, are in direct contact with the polymer electrolyte and MPL. MPLs are situated in between CLs and GDLs. The purpose of MPLs is to improve water management by providing effective water transport. It also minimizes the contact resistance between the GDL and CL, limits the loss of catalyst to the GDL interior, and prevents dry-out of the membrane at low current densities or low humidity. GDLs are adjacent to the gas flow channel on both sides of a PEFC, and they allow reactant gases to diffuse to the CL. Additionally, GDLs provide mechanical support to the cell and allow product water to transport from the CL to the gas flow channel, thus preventing water flooding inside the cathode CL. Each of these layers has unique composition and structure due to their distinct purposes in a PEFC, as schematically highlighted in Fig. 7.2. The microstructure of widely used CL for PEFC consists of a matrix of platinum (Pt) catalyst particles supported on carbon particles, electrolyte membrane (also known as ionomer), and void space (pores) with a pathway for electrons and protons to reach the reaction sites. This type of CL is known as the Pt/C CL, as Pt nanoparticles are dispersed onto the surfaces of larger carbon black particles. Both liquid phase and gas phase coexist inside CLs. Thus, pores can be filled with either liquid water or reactant gases, as shown in Fig. 7.2. An alternative to the Pt/C CL is the nanostructured thin film (NSTF) CL. The key features of the NSTF CL are low Pt loading,

FIGURE 7.2 Schematic illustrations of the structure and composition of CL, MPL and GDL for PEFCs [7].

202

Fuel Cells for Transportation

thinner than Pt/C CL, and it does not have carbon support or any additional ionomer, as the Pt catalyst is directly deposited to form an electronically conductive and electrochemically active layer. The MPL consists of carbon nanoparticles mixed with hydrophobic polytetrafluoroethylene (PTFE) and void space with a pathway for electrons to reach the CL. Like CL, MPL pores can also be filled with liquid water and reactant gases during fuel cell operation, and they provide a passage for reactant gases to reach the CL and liquid water to reach the flow channel via the GDL. The GDLs are made by weaving carbon fibers into carbon cloth or by pressing carbon fibers together into carbon paper. They are often rendered wetproof by saturating pores with PTFE emulsions, followed by drying and sintering to affix PTFE particles to carbon fibers to improve liquid transport. The scanning electron microscope (SEM) images of typical carbon paper and carbon cloth GDLs with and without PTFE treatment are shown in Fig. 7.3, along with the pore size distribution of SGL 39BA GDL [8]. Comparing Fig. 7.3B with Figs. 7.3C and 7.3D, one can easily distinguish the PTFE loading in the GDL. As the microstructure and composition of CL, MPL and GDL vary significantly, the key parameters for reactants, water, and electron transport are porosity, pore size and wettability, which also vary significantly. A discussion of these parameters is given below.

7.2.1

Porosity and pore size

The porosity of porous layers (GDL, CL and MPL) in fuel cells is the measure of the void space. It is represented by the fraction of the volume of voids over the total volume and is often denoted by the symbol φ or ε (varies between 0 and 1). For GDLs that are often made by weaving carbon fibers into a carbon cloth or by pressing carbon fibers together into a carbon paper, the porosity can be determined by subtracting the volume occupied by carbon fibers from 1. However, not all pores are connected, and there can be PTFE loading in GDL. Thus, experiments (such as mercury porosimetry, helium pycnometry, or imbibition method) are often used to measure the true porosity of porous transport layers. Typically, the uncompressed porosity of GDLs can be between 0.7 and 0.9, while the thickness varies between 100 and 400 μm. Several commonly used GDLs are listed in Table 7.1. The diameter of carbon fibers varies between 5 and 10 μm, while the pore sizes of GDL can vary between 10 and 100 μm. However, most of the pores are between 10 and 40 μm for SGL 39BA GDL, as shown in Fig. 7.3E. MPLs in PEFCs act as a transition layer between the GDL and CL. Thus, the porosity and pore size of MPLs are always lower than GDLs but higher than CLs. Typical fuel cell MPLs (such as SIGRACET C-type) are often based on carbon nanoparticles mixed with PTFE (roughly 77 wt.% carbon black and 23 wt.% PTFE). Fig. 7.4 shows a SEM

Effective transport properties for fuel cells Chapter | 7

203

FIGURE 7.3 SEM images of carbon cloth and carbon paper GDLs and pore size distribution of carbon paper GDL. (A) Carbon cloth GDL, (B) Toray-TGP-H-120 (0 wt.% PTFE loading), (C) SGL 39BA (5 wt.% PTFE loading), (D) SGL 39BA (5 wt.% PTFE loading), and (E) pore size distribution of SGL 39BA. Pore size distribution data are from Ref. D. Thumbarathy et al., Fabrication and characterization of tuneable flow-channel/gas-diffusion-layer interface for polymer electrolyte fuel cells. J. Electrochem. Energy Convers. Storage 17 (2020) 011010, https:// doi.org/10.1115/1.4044814. Credit: Dr. Deepashree Thumbarathy and Mr. Xiao Liu (Sustainable Energy Systems Lab, Newcastle University, UK), and Weber Lab (Lawrence Berkeley National Laboratory, USA).

204

Fuel Cells for Transportation

TABLE 7.1 Commonly used Sigracet and Toray GDLs and their porosity values. Name

Thickness (μm)

PTFE-treated

MPL

Porosity

SGL 10 BA

400

Yes

No

0.88

SGL 10 BC

420

Yes

Yes

0.80

SGL 24 BA

190

Yes

No

0.84

SGL 24 BC

230

Yes

Yes

0.76

SGL 25 BA

190

Yes

No

0.88

SGL 25 BC

235

Yes

Yes

0.80

SGL 34 BA

280

Yes

No

0.83

SGL 34 BC

315

Yes

Yes

0.75

SGL 39 BA

280

Yes

No

0.80

SGL 39 BC

325

Yes

Yes

0.80

TGP-H-030

110

No

No

0.80

TGP-H-060

190

No

No

0.78

TGP-H-090

280

No

No

0.78

TGP-H-120

370

No

No

0.78

FIGURE 7.4 (A) SEM image of a typical fuel cell MPL and (B) pore size distribution. Credit: SEM image is reprinted from P.K. Das, D. Thumbarathy, Chapter 16—Heat and fluid flow in porous media for polymer-electrolyte fuel cells, in: Convective Heat Transfer in Porous Media, CRC Press, Boca Raton, USA, 2019, pp. 341
360, with permission from Taylor & Francis; J. Becker et al., A multi-scale approach to material modeling of fuel cell diffusion media, Int. J. Heat. Mass. Transf. 54 (2011) 1360
1368, https://doi.org/10.1016/j.ijheatmasstransfer.2010.12.003.

Effective transport properties for fuel cells Chapter | 7

205

image of a MPL. The typical pore size of a MPL can be between 10 and 200 nm, porosity can be in the 0.3
0.5 range, and the thickness varies between 10 and 50 μm. A SEM image and a pore size distribution of a typical MPL are shown in Fig. 7.4 [9]. The pore size distribution data presented in Fig. 7.4 is taken from Ref. [10]. Both GDL and MPL have a relatively simpler structure and composition compared with the CL. As the typical Pt/C CL is made of Pt and carbon nanoparticles mixed with electrolyte, the pore size and porosity of CL are significantly smaller than that of MPL and GDL due to the formation of agglomerates of Pt/C particles covered with electrolyte. This leads to two types of pores in a Pt/C CL: primary pores and secondary pores. Primary pores are between Pt/C particles inside agglomerates. These pores typically range from 20 to 60 nm. The secondary pores are in between agglomerates and void space. These pores are usually larger than primary pores and vary between a few hundred nanometers and 50 nm. The porosity of a Pt/C CL can be between 0.2 and 0.3, while the thickness of CLs depends on the amount of catalyst loading and varies between 10 and 50 μm. On the other hand, 3M’s NSTF CL is an extended surface catalyst that includes singlecrystalline whiskers of an organic compound coated with Pt alloy. NSTF CLs are 20
30 times thinner than conventional Pt/C CLs (typically less than 1 μm). Thus, it has significantly lower Pt loading (in the order of 0.05
0.15 mg-Pt/cm2). SEM images of conventional Pt/C and 3 M’s NSTF CLs and the pore size distribution of a Pt/C CL are shown in Fig. 7.5 [11,12]. The pore size distribution for Pt/C CL is taken from Ref. [13].

7.2.2

Wettability

As GDLs are available with a wide range of porosity values and thicknesses, they are also available with a wide range of surface wettabilities. The surface wettability is always represented by the static contact angle for smooth surfaces. For GDLs, both static and dynamic contact angles are required for properly analyzing the water
GDL surface interaction due to the inherent surface roughness of GDL surfaces. Conversely, the wettability of GDL pores is often represented through the capillary pressure
saturation relationship rather than using a contact angle measurement. The surface wettability of a GDL is dictated by the amount of PTFE loading on it as well as its surface roughness. Experimental data show that the static contact angle for GDL without PTFE loading (SGL 24AA) can be 141 6 3 degrees [14], while GDL with 5 wt.% PTFE loading (SGL 39BA) can be 145 6 1 degrees [8]. It is considered that GDL with 5 wt.% PTFE loading would be sufficient for obtaining a pronounced hydrophobicity for typical PEFC operation. However, a higher PTFE loading may be required for low-temperature operation and faster water removal from GDL surface. Experimental data of the static contact angle for a GDL with 10, 20, and 30 wt.% (SGL 24CA, SGL

206

Fuel Cells for Transportation

FIGURE 7.5 SEM images of CLs and pore size distribution: (A) conventional Pt/C CL, (B) PtCoMn alloy-based nanostructured thin film (NSTF) CL, and (C) pore size distribution of a conventional Pt/C CL (20 nm increments). Credit: Adapted with permission from P.K. Das, A.Z. Weber, Proceedings of the ASME 11th fuel cell science, engineering, and technology conference, Paper No. Fuel Cell 2013-18010, 2013 and M.K. Debe, J. Electrochem. Soc. 160 (6), F522
F534, 2013; [13] H. Schulenburg et al., 3D imaging of catalyst support corrosion in polymer electrolyte fuel cells, J. Phys. Chem. C. 115 (2011) 14236
14243, https://doi.org/ 10.1021/jp203016u.

24DA, and SGL 24EA) PTFE loadings are reported as 158 6 2 degrees, 159 6 2 degrees, and 156 6 3 degrees, respectively, [14]. Higher PTFE loadings ( . 10 wt.%) do not provide a higher contact angle as the surface of GDL is statured with PTFE after a certain PTFE loading. At higher PTFE loading, the pore size, however, will be smaller as well as GDL porosity, leading to lower effective transport properties. To overcome the issues with effective transport properties at high PTFE loading, there are initiatives to selectively modify GDL surfaces with hydrophobic materials instead of the full coverage of PTFE on GDL surface. For instance, a recent study shows that the GDL surface can be selectively treated using a pattern (such as polka dot, stripes, or checkered) of hydrophobic monomers, such as polydimethylsiloxane (PDMS) matrix containing fumed silica particles or fluorinated ethylene propylene (FEP) [8]. The SEM images of GDLs with selectively

Effective transport properties for fuel cells Chapter | 7

207

modified surface wettability using stripes PDMS-Si and FEP are shown in Fig. 7.6. It has been shown that GDL patterned with FEP can have pore sizes between 5 and 40 μm like the base GDL (SGL 39BA GDL), while GDL patterned with PDMS-Si exhibits slightly smaller pores (5
30 μm). However, both FEP-coated and PDMS-Si-coated exhibit significantly higher contact angles, 162 and 159 degrees, respectively, compared with a base GDL of

FIGURE 7.6 SEM images of GDLs with selectively modified surface wettability. (A) Base GDL (SGL 39BA), (B) and (C) FEP-coated GDLs, and (D) and (E) PDMS-Si-coated GDLs at various magnifications showing the surface morphology. Credit: Reprinted from Thumbarathy et al., J. Electrochem. Energy Conv. Storage, 17 (1) (2020) 011010
1, with permission from ASME.

208

Fuel Cells for Transportation

145 degrees. Thus, one can achieve significantly higher surface wettability with the least amount of compromise in GDL pore size and porosity. Moreover, these GDLs can perform better in PEFCs and provide higher limiting currents as compared to base GDLs [8]. This may pave the way to design novel and tunable GDLs for high-power PEFCs.

7.3

Effective transport properties

As discussed above, porous layers must provide several critical functions, such as providing a transport pathway for reactants and products through the pore volume and ensuring charge and heat conduction through their solid matrix. CLs have the added functionality of providing an electrochemically active surface area. As shown in Fig. 7.7, relevant effective transport properties include

FIGURE 7.7 Concentration, velocity magnitude and electronic potential/temperature fields, corresponding to calculations of through-plane (TP) and in-plane (IP) effective diffusivity, permeability, and effective electrical/thermal conductivity, respectively, on a carbon paper GDL (10 wt.% PTFE-treated TGP-H-120). Credit: Reprinted from P.A. Garcı´a-Salaberri et al. Int. J. Heat Mass Transfer 127 (2018) 687
703, with permission from Elsevier.

Effective transport properties for fuel cells Chapter | 7

209

permeability used in Darcy’s law, tortuosity factor used to correct Fick’s law of diffusion, and effective electrical and thermal conductivities used in Ohm’s and Fourier’s laws, respectively. An overview of effective diffusivity, local mass transport resistance, permeability, and thermal, electrical and ionic conductivities of GDLs, MPLs and CLs is presented below.

7.3.1

Effective diffusivity

Most GDLs, especially carbon paper GDLs, feature anisotropic effective diffusivity due to preferential alignment of fibers and pores in the material plane [15]. As shown in Fig. 7.8, a higher diffusivity is found in the in-plane direction. The through-plane effective diffusivity is notably lower than predicted by most widely used models, such as Bruggeman’s effective medium theory and the random fiber model of Tomadakis and Sotirchos [16,17]. The deviation from these idealized models is caused by the complex structure formed by fibers, binder and PTFE [18]. Typical effective diffusivities in the through-plane direction (normalized with respect to the bulk value) are in the range between 0.2 and 0.4, being 1.5
2 times larger than the effective diffusivity in the material plane. Beyond pore structure, the distribution and amount of liquid water in the pore space (i.e., water saturation) have a strong impact on GDL diffusivity. The relative effective diffusivity, defined as the ratio between the effective diffusivity under wet and dry conditions, depends on water saturation as a power law of the form, g(s) 5 (1 2 s)n, where the saturation exponent n lies between 2 and 5 depending on the arrangement of water and peak saturation [20,21] (see Fig. 7.9).

FIGURE 7.8 Normalized effective diffusivity and permeability in the through-plane (TP) and in-plane (IP) directions as a function of GDL compression ratio computed with the hybrid pore network/continuum model presented in [19] as compared to previous experimental data. Credit: Reprinted from P.A. Garcı´a-Salaberri, Int. J. Heat Mass Transfer 167 (2021) 120824, with permission from Elsevier.

210

Fuel Cells for Transportation

FIGURE 7.9 Through-plane relative effective diffusivity, gTP , as a function of average saturation, savg , of carbon paper GDLs (Toray TGP-H-120) determined numerically in [20] compared to previous experimental data. Credit: Reprinted from P.A. Garcı´a-Salaberri et al., Int. J. Heat Mass Transfer 86 (2015) 319
333, with permission from Elsevier.

Microporous layer’s effective diffusivity has been less examined, usually considering the whole bilayer (GDL 1 MPL) due to the exceedingly small thickness of MPL coating [22]. The normalized dry effective diffusivity of MPLs is rather isotropic due to the spherical shape of carbon nanoparticles, being around 0.1
0.2 (3
5 times lower than GDLs). The lower diffusivity is explained by the dominant role of Knudsen diffusion in pores smaller than 1 μm due to frequent collision of molecules with pore walls [23,24]. The effect of water on MPL effective diffusivity is still a source of investigation. A saturation exponent around n 5 1.5 is assumed in most macroscopic models [25,26]. A growing number of works have been devoted in the last years to CL effective diffusivity due to its critical role in PEFC performance and durability. CL effective diffusivity is rather isotropic, and its normalized value is somewhat lower than that of a MPL (0.1
0.2) because of Knudsen effect in primary and secondary pores (6
100 nm). CL effective diffusivity can be increased using highly porous open structures, such as those obtained by freeze-drying and electrospraying [27,28]. Achieving high CL diffusivity is critical to increase performance when the number of active sites is reduced at low Pt loading [29].

7.3.2

Local mass transport resistance

In addition to CL diffusivity, the local mass transport resistance from the pore space toward Pt nanoparticles plays a critical role at low cathode Pt

Effective transport properties for fuel cells Chapter | 7

211

loading. As shown in Fig. 7.10, the local resistance can be decomposed into five components: (1) the interfacial resistance at pore/water interface, (2) the diffusive resistance across water film, (3) the interfacial resistance at water/ ionomer interface, (4) the diffusive resistance across thin ionomer film, and (5) the interfacial resistance at ionomer/Pt surface [30]. Recent experimental works have shown that the diffusive resistance across thin ionomer films (1
10 nm in thickness) is dominant, while the interfacial resistances only sum up to one-third of the ionomer resistance [31]. The low oxygen permeability in CL ionomer films is explained by finite-size substrate interactions, which lead to the formation of a dense ionomer region with extremely low diffusivity near Pt interface. Strategies to reduce the local mass transport resistance at low Pt loading include the design of CLs with an increased electrochemically active surface area, high permeation coefficient of oxygen in ionomer, homogeneous ionomer thickness, and nanostructures with superhydrophobic properties. Besides, low ionic conductivity in thin ionomer films must be avoided through a careful design of the multiscale ionomer network.

FIGURE 7.10 (A) Schematic of the 1D multiscale model used in [30] to examine local oxygen transport resistance in the cathode CL. (B) Sketch of CL microstructure, indicating transport pathways of oxygen, electrons, and protons. (C) Close-up view of oxygen transport resistances from the pore space toward Pt nanoparticles. Credit: Reprinted from A. Sa´nchez-Ramos et al., J. Electrochem. Soc. 168 (2021) 124514 (open access).

212

7.3.3

Fuel Cells for Transportation

Permeability

Permeability of fibrous GDLs is also anisotropic, with higher values in the inplane direction (see Fig. 7.8). Permeability of uncompressed GDLs ranges between 10212 and 10211 m2, decreasing strongly upon compression owing to the decrease of both porosity and pore size [32]. The dependence of GDL permeability on porosity can be well correlated through the Carman
Kozeny equation [33], being the Carman
Kozeny constant in the range kck B1 2 100. Carbon cloth GDLs typically have a higher permeability (lower kck ) and more isotropic character compared with carbon paper GDLs due to the preferential transport routes provided by large pore sizes between yarns. Permeabilities of MPLs and CLs are significantly lower than that of GDLs, being in the order of KB10214 210213 m2 [34
36]. Lower values prevail in CLs due to the smaller pore size of this layer.

7.3.4

Effective thermal conductivity

Effective transport properties that depend on the solid phase usually show a higher degree of anisotropy, reaching values even one order of magnitude larger in the in-plane direction than in the through-plane direction [37]. Such difference is caused by the high interconnectivity of fibers in the material plane unlike infrequent contacts between fibers (and binder) in the throughplane direction. This situation does not hold for effective transport properties that rely on the fluid phase, which typically show a lower degree of anisotropy. GDL effective thermal conductivity varies significantly between fabrics, from 0.2 to 0.4 Wm21 K21 (e.g., SGL Carbon’s SIGRACET and Mitsubishi Rayon Corporation) up to around 1 Wm21 K21 (Toray TGP-H series) [38,39]. Key factors affecting the effective thermal conductivity of GDLs include PTFE and binder contents and assembly compression. The effective thermal conductivity of MPLs and CLs is lower than that of GDLs, ranging between 0.1 and 0.15 Wm21 K21 [40]. The effect of compression on MPL and CL properties is lower due to their smaller pore sizes and stiffness. Further work is still needed to examine the interplay between PEFC operation and effective thermal conductivity, including interfacial contact resistances, phase change phenomena, and the impact of water saturation on effective thermal conductivity (especially in MPL and CL) [41].

7.3.5

Effective electrical conductivity

The effective electrical conductivity of GDL, MPL and CL is high, ranging between 102 and 103 S m21, with a higher anisotropy in GDLs. The fine pore structure of MPL coating allows a reduction of interfacial contact resistances caused by the macroporous GDL [42]. The design of carbonunsupported CLs, despite their lower use in PEFCs, requires control of

Effective transport properties for fuel cells Chapter | 7

213

ionomer volume fraction and distribution to avoid bottlenecks for electron transport [43].

7.3.6

Effective ionic conductivity

The effective ionic conductivity of CL and membrane plays a key role in cell performance due to the higher size of protons and thus larger difficulty of proton conduction. For similar volume fractions of the conductive phase (ionomer for proton transport and carbon for electron transport), membrane ionic conductivity (10 S m21) is around two orders of magnitude lower than the electrical conductivity of porous layers (102 2 103 S m21) [30]. Moreover, CL’s effective ionic conductivity is ten-fold lower than that of the bulk membrane, being in the range of 0.05
5 S m21 under well-humidified conditions. CL effective ionic conductivity increases with ionomer volume fraction but remains constant for exceedingly high ionomer volume fractions. This is explained by the increase of the tortuosity factor of the ionomer network at high ionomer volume fractions [44]. The tailored design of the multiscale CL ionomer structure is crucial to optimize the coupling between proton conduction, oxygen diffusion and water transport in low Pt-loading CLs.

7.4

Modeling and experimental techniques

A combination of experimental and modeling work is necessary to examine multiphysics phenomena and guide the design of novel components and cell architectures. The variety of transport phenomena that take place in PEFCs at different scales makes it necessary the use of models with different levels of sophistication according to their dimensionality and resolution of porous media microstructure. Moreover, experimental work involves different disciplines, such as the characterization of single- and two-phase effective transport properties, ohmic resistances, and electrochemical performance. An overview of modeling and experimental techniques used in PEFCs (and related electrochemical devices) is presented below.

7.4.1

Modeling

As shown in Fig. 7.11, three main modeling techniques are used to examine mass, charge and heat transport in porous layers: (1) macroscopic modeling, (2) pore-scale modeling, and (3) hybrid modeling (i.e., a combination of both macroscopic continuum and pore-scale formulations). Macroscopic modeling relies on a volume average description of conservation equations in porous components, including mass, momentum, energy, species (H2, O2, N2, and H2O), electronic and protonic charge, membrane water content, and liquid water. The formulation is closed through constitutive relationships that define effective transport properties of porous components as a

214

Fuel Cells for Transportation

FIGURE 7.11 Flow chart of the three main modeling approaches used to examine transport in PEFCs according to their microstructural resolution. Credit: Reprinted from P.A. Garcı´aSalaberri, General aspects in the modeling of fuel cells: from conventional fuel cells to nano fuel cells, in: Nanotechnology in Fuel Cells, Micro and Nano Technologies, 2022, pp. 77
121.

function of macroscopic properties (e.g., porosity, water saturation, ionomer volume fraction, etc.). Macroscopic continuum modeling is the most extended approach to analyzing transport at the cell/stack scale due to moderate computational cost and availability of multiphysics models in (commercial) CFD codes [26,45]. Two modeling approaches are used at the pore scale: pore network modeling (PNM) and direct numerical simulation (DNS). PNM idealizes the pore space of porous layers as a network of pore bodies interconnected by throats. The size, shape and coordination number of the pore/throat assembly are determined according to the porous media microstructure. Multiple transport processes can be simulated on the network, such as capillarity, diffusion and convection [46]. The development of dual networks incorporating both fluid and solid phases has also become increasingly common to analyze, for

Effective transport properties for fuel cells Chapter | 7

215

example, two-phase transport in GDLs [47]. On the other hand, DNS solves conservation equations directly on microstructures taken from segmented tomography images or virtual reproductions generated through mathematical algorithms [48]. The latter method is generally used to overcome limitations of tomography images, such as the impossibility to differentiating binder and PTFE in GDLs, and ionomer and Pt nanoparticles in CLs. Comparatively, PNM offers a significantly lower computational cost than DNS, although the accuracy of results largely depends on the pore network calibration. Hence, PNM is better suited to perform parametric analysis at the microstructural level in engineering applications, while DNS is recommended for the extraction of detailed information during porous material design. Both modeling techniques can also be used to determine effective transport properties for macroscopic models [37]. The use of hybrid models, which combine macroscopic continuum and pore-scale modeling techniques, has recently increased. The aim of hybrid models is to combine in a single framework the strengths of both modeling techniques, that is, the ease of implementation of continuum approaches in CFD codes and explicit microstructural information from the pore scale [19]. This type of modeling is particularly useful to improve the predictions of two-phase transport through the multiscale pore structure of MEAs, while accounting for variation of operating conditions and heterogeneities at the cell scale [49]. The combination of PNM and continuum modeling is preferred in engineering applications due to its lower computational cost. In particular, the computational cost arising from the coupling of DNS and continuum modeling can be prohibitive under two-phase conditions due to the wide range of spatial and temporal scales [42]. In addition to the previous three modeling approaches, a large body of work is also devoted to analytical and semianalytical modeling of effective properties (see Refs. [50,51]). Prediction of effective transport properties based on a simplified representation of the geometry of representative elementary volumes is useful for fundamental understanding of transport processes and examination of multicomponent materials. Moreover, analytical and semianalytical models can be incorporated into macroscopic, pore-scale, and hybrid models when all spatial and temporal scales cannot be resolved due to excessive computational cost [30]. This practice is common, for example, for the CL, where transport in agglomerates or around carbon nanoparticles is described by an analytical or semianalytical sub-model. The large-scale model and the nanoscale sub-model can be coupled by continuity of fluxes or any other physical condition [52].

7.4.2

Experimental

A wide variety of in-situ characterization, diagnostic and visualization techniques are available to examine PEFC operation depending on physics and

216

Fuel Cells for Transportation

materials, and spatial and temporal scales involved. Techniques include traditional methods, such as electrochemical impedance spectroscopy (EIS) or limiting-current method, and more novel technologies, such as in-operando X-ray computed tomography and neutron imaging (e.g., for visualization of water transport in GDLs) and segmented cells to track the evolution of local current density, temperature and membrane resistance [53
57]. These in-situ techniques are complemented with information from the ex-situ characterization of multifunctional components (see, e.g., [58,59]). Fig. 7.12 shows a flowchart of widely used methods for the characterization of porous layers (GDL, MPL, and CL) effective properties, including (1) structural and morphological properties, (2) effective transport properties, (3) mechanical properties, and (4) electrochemical properties. For completeness, techniques used to characterize some relevant properties of polymeric materials (i.e., membrane) are also included. Common structural and morphological properties of porous layers characterized experimentally include porosity, PTFE content by weight, specific surface area, and wettability. Porosity can be simply measured using Archimedes’ principle by weighting porous samples both dry and submerged in a wetting fluid and subsequently determining the solid volume. This technique has been successfully applied to thin porous media (GDL-MPLs and CLs) with high precision [60,61]. Pore size distribution is usually characterized by mercury intrusion porosimetry (MIP) [62]. Modern equipment allows invasion pressures up to 60,000 bar with a resolution down to 3.6 nm. However, nanometric pore size distributions can also be measured by gas physisorption (using the same equipment to determine Brunauer
Emmett
Teller (BET) surface area),

FIGURE 7.12 Flowchart of experimental techniques used for characterization of structural and morphological properties, effective transport properties, mechanical properties, and electrochemical properties in PEFCs.

Effective transport properties for fuel cells Chapter | 7

217

thereby avoiding alterations caused bya high working pressure [63]. Alternatively, porosity and pore size can be determined from tomography images. Porosity is directly calculated by voxel counting, while pore size is calculated by determining the size of the largest inscribed sphere containing a voxel [20,21]. Unlike MIP, working with tomography images removes any influence caused by the invasion history of mercury from the exterior surface of a sample. Wettability is frequently characterized through contact angle measured by the sessile drop method [14,64]. In this technique, a water droplet is placed on the sample surface, and the contact angle is measured by fitting a tangent to the three-phase point where the liquid surface touches the porous surface. The visualization of material microstructure is carried out by SEM. Quantification of internal contact angles in porous layers is more complicated but can be accomplished by analysis of tomography images from water invasion experiments [65]. Other common techniques used to examine the composition of porous layers are thermogravimetric analysis (e.g., PTFE volume fraction in GDL), elemental analysis with energy-dispersive X-ray spectroscopy, EDX, and X-ray photoelectron spectroscopy, XPS (e.g., the composition of ionomer, carbon, and Pt in CL), and positron annihilation lifetime spectroscopy, PALS (e.g., free volume fraction in ionomer) [60,66,67]. Permeability can be measured with a permeability tester based on Darcy’s law. In this apparatus, gas is fed through a sample at a prescribed flow rate (measured with a flow meter), and the resultant pressure drop is recorded (measured with a differential pressure sensor) [33]. Permeability in different directions can be determined by changing the transport direction in experiments. Different methods have been used to measure the effective diffusivity of thin GDL-MPLs and CLs, including diffusion bridge, Loschmidt cell and electrochemical limiting-current method, as well as transient methods in which concentration evolution is fitted against an analytical solution of Ficks’ law [68,69]. Effective thermal conductivity has been mostly characterized using in-house setups by measuring the temperature gradient across samples subjected to a heat flux. The contribution of other thermal resistances in the setup (e.g., interfacial resistances) is subtracted to isolate the sample resistance and determine its effective thermal conductivity [70]. Another technique used for the characterization of effective thermal conductivity is the laser flash method. This technique is based on the detection of the transient temperature rise on the backside of a sample when it is heated with an energy pulse [71]. Effective electrical conductivity of porous layers is commonly measured by the four-point probe method using separate pairs for current-carrying and voltage-sensing electrodes. The setup with four probes introduces almost negligible contact and spreading resistances associated with voltage probes, providing high accuracy [72]. Effective ionic conductivity is characterized by either direct current (DC) or alternating current (AC) methods using a four-point probe setup and EIS, respectively, [44,73]. Similar results have been historically obtained with both methods, even

218

Fuel Cells for Transportation

though the use of EIS is more extended due to widespread availability of electrochemical cells. Moreover, EIS facilitates conductivity measurements in ultrathin layers, such as CLs. The main concern of EIS is the need to select an appropriate equivalent circuit for the correct interpretation of AC Nyquist plots. Analysis of two-phase transport in PEFCs requires the characterization of several additional transport properties, such as water retention curves, breakthrough pressure, and droplet adhesion forces at the GDL/channel interface. Quasi-static transport of water (or other invading fluids) during drainage and imbibition can be characterized by the method of standard porosimetry or volume displacement method [74]. The purpose of these experiments is to measure the amount of water in a sample at fixed capillary pressure. Special care must be taken with evaporation in thin porous media due to the small water volumes used in experiments. Droplet adhesion force can be measured, for example, by using a rotating-stage goniometer while injecting water from the bottom or by placing droplets with a syringe on the sample surface [14,75]. Other relevant transport properties in PEFCs are electroosmotic drag coefficient and diffusivity of water in the membrane, permeability of gas species across the membrane, water sorption isotherm of membrane and CL, and electrical and thermal contact resistances. For brevity, the reader is referred to focused reviews [53,54,76]. Mechanical properties of porous layers include anisotropic Young’s modulus, shear modulus, and Poisson’s ratio. Young’s modulus in the throughplane direction is typically measured in compressive tests with universal test machines adapted to thin porous layers, such as those used for packaging applications [77]. Hysteretic behavior at compression can be evaluated in loading
unloading cyclic tests. Young’s modulus in the material plane can be determined using different loading conditions, such as tension, compression, or bending. Conditions in fuel cells (a mixture of tension and compression) are better reproduced with 2-point, 3-point, or 4-point bending tests. Measurement of shear modulus and Poisson’s ratio is carried out in ad hoc apparatuses due to the difficulty to determine these properties in thin porous layers. Viscoelastic properties of polymeric materials are also measured in cyclic and relaxation experiments as a function of temperature and relative humidity [78]. Among electrochemical properties, electrochemical surface area (ECSA) and exchange current density are key parameters to evaluate CL performance. ECSA of Pt-based electrocatalysts is calculated in electrochemical cells from the hydrogen adsorption/desorption region of cyclic voltammetry curves after correcting for double-layer charging current. The rotating disk electrode is a useful technique for determining exchange current density and symmetry factors of electrochemical kinetics [79]. However, experiments in hydrogen pump cells and PEFCs are also a common practice to determine

Effective transport properties for fuel cells Chapter | 7

219

electrochemical parameters in conditions closer to the final application [80,81]. The voltage drop in the electrochemical cell is broken down into activation and ohmic losses to isolate the contribution of the former (mass transport losses can be neglected). Electrochemical parameters are then determined by fitting results to a kinetics expression, usually Tafel or Butler
Volmer kinetics. Additional electrochemical techniques widely used to determine electrocatalytic information are linear sweep voltammetry and CO-stripping voltammetry [53].

7.5

Summary

Polymer electrolyte fuel cells are promising candidates as clean power sources for the transportation sector (e.g., submarines, trains, and light-duty and heavy-duty vehicles). However, their market penetration is still hindered by excessive cost and insufficient performance and durability, preventing economies of scale. In the last decades, leading automotive companies have devoted large investments to optimize and commercialize PEFC technology. However, the progress achieved is not yet sufficient to compete with traditional (e.g., internal combustion engine) and alternative (e.g., battery electric vehicle) technologies. Apart from the need for a hydrogen infrastructure, one key drawback is the large amount of Pt needed to maintain acceptable performance and durability. To overcome this situation and reduce Pt loading, while increasing performance and durability, optimal design of multifunctional porous layers with improved effective transport properties and electrochemical activity is mandatory. This task is complicated by the variety of multiphysics, multiphase and multiscale transport processes that take place in porous layers. Future PEFC designs shall focus not only on the optimization of cathode catalyst layer microstructure but the integral optimization of the full membrane electrode assembly and flow field (e.g., using 3D printing and related techniques).

References [1] G. Claeys, et al., How to make the European Green Deal work, Policy Contribution 13 (2019) 1
21. Available from: https://www.jstor.org/stable/resrep28626. [2] M.M. Whiston, et al., Expert elicitation on paths to advance fuel cell electric vehicles, Energy Policy 160 (2022) 112671. Available from: https://doi.org/10.1016/j.enpol.2021.112671. [3] M.M. Whiston, et al., Expert assessments of the cost and expected future performance of proton exchange membrane fuel cells for vehicles, Proc. Natl. Acad. Sci 116 (2019) 4899
4904. Available from: https://doi.org/10.1073/pnas.1804221116. [4] C.E. Thomas, Fuel cell and battery electric vehicles compared, Int. J. Hydrog. Energy 34 (2009) 6005
6020. Available from: https://doi.org/10.1016/j.ijhydene.2009.06.003. [5] A. Kongkanand, M.F. Mathias, The priority and challenge of high-power performance of low-platinum proton-exchange membrane fuel cells, J. Phys. Chem. Lett 7 (2016) 1127
1137. Available from: https://doi.org/10.1021/acs.jpclett.6b00216.

220

Fuel Cells for Transportation

[6] T. Yoshida, K. Kojima, Toyota MIRAI fuel cell vehicle and progress toward a future hydrogen society, Electrochem. Soc. Interface 24 (2015) 45. Available from: https://doi. org/10.1149/2.F03152if. [7] P.K. Das, Transport Phenomena in Cathode Catalyst Layer of PEM Fuel Cells, University of Waterloo, Waterloo, 2010, http://hdl.handle.net/10012/5282 [8] D. Thumbarathy, et al., Fabrication and characterization of tuneable flow-channel/gas-diffusion-layer interface for polymer electrolyte fuel cells, J. Electrochem. Energy Convers. Storage 17 (2020) 011010. Available from: https://doi.org/10.1115/1.4044814. [9] P.K. Das, D. Thumbarathy, Heat and fluid flow in porous media for polymer-electrolyte fuel cells, in: Y. Mahmoudi, K. Hooman, K. Vafai (Eds.), Convective Heat Transfer in Porous Media, CRC Press, 2019, pp. 341
360. Available from: https://doi.org/10.1201/ 9780429020261-16. [10] J. Becker, et al., A multi-scale approach to material modeling of fuel cell diffusion media, Int. J. Heat. Mass. Transf 54 (2011) 1360
1368. Available from: https://doi.org/10.1016/ j.ijheatmasstransfer.2010.12.003. [11] P.K. Das, A.Z. Weber, Water management in PEMFC with ultra-thin catalyst-layers Available from: https://doi.org/10.1115/FuelCell2013-18010. [12] M.K. Debe, Tutorial on the fundamental characteristics and practical properties of nanostructured thin film (NSTF) catalysts, J. Electrochem. Soc 160 (2013) F522
F534. Available from: https://doi.org/10.1149/2.049306jes. [13] H. Schulenburg, et al., 3D imaging of catalyst support corrosion in polymer electrolyte fuel cells, J. Phys. Chem. C. 115 (2011) 14236
14243. Available from: https://doi.org/ 10.1021/jp203016u. [14] P.K. Das, et al., Liquid-water-droplet adhesion-force measurements on fresh and aged fuel-cell gas-diffusion layers, J. Electrochem. Soc 159 (2012) B489
B496. Available from: https://doi.org/10.1149/2.052205jes. [15] M.F. Mathias, J. Roth, J. Fleming, W. Lehnert, Diffusion media materials and characterization, Handb. Fuel Cell 3 (2010) 517
537. [16] M.M. Tomadakis, S.V. Sotirchos, Ordinary, transition, and Knudsen regime diffusion in random capillary structures, Chem. Eng. Sci. 48 (1993) 3323
3333. Available from: https://doi.org/10.1016/0009-2509(93)80149-K. [17] B. Tjaden, et al., On the origin and application of the Bruggeman correlation for analysing transport phenomena in electrochemical systems, Curr. Opin. Electrochem 12 (2016) 44
51. Available from: https://doi.org/10.1016/j.coche.2016.02.006. [18] R. Fl¨uckiger, et al., Anisotropic, effective diffusivity of porous gas diffusion layer materials for PEFC, Electrochim. Acta 54 (2008) 551
559. Available from: https://doi.org/ 10.1016/j.electacta.2008.07.034. [19] P.A. Garcı´a-Salaberri, Modeling diffusion and convection in thin porous transport layers using a composite continuum-network model: application to gas diffusion layers in polymer electrolyte fuel cells, Int. J. Heat. Mass. Transf. 167 (2021) 120824. Available from: https://doi.org/10.1016/j.ijheatmasstransfer.2020.120824. [20] P.A. Garcı´a-Salaberri, et al., Effective diffusivity in partially-saturated carbon-fiber gas diffusion layers: effect of through-plane saturation distribution, Int. J. Heat. Mass. Transf. 86 (2015) 319
333. Available from: https://doi.org/10.1016/j.ijheatmasstransfer.2015. 02.073. [21] P.A. Garcı´a-Salaberri, et al., Effective diffusivity in partially-saturated carbon-fiber gas diffusion layers: effect of local saturation and application to macroscopic continuum

Effective transport properties for fuel cells Chapter | 7

[22]

[23]

[24] [25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

221

models, J. Power Sources 296 (2015) 440
453. Available from: https://doi.org/10.1016/j. jpowsour.2015.07.034. M.S. Ismail, et al., Effective diffusivity of polymer electrolyte fuel cell gas diffusion layers: an overview and numerical study, Int. J. Hydrog. Energy 40 (2015) 10994
11010. Available from: https://doi.org/10.1016/j.ijhydene.2015.06.073. C. Chan, et al., Experimental measurement of effective diffusion coefficient of gas diffusion layer/microporous layer in PEM fuel cells, Electrochim. Acta 65 (2012) 13
21. Available from: https://doi.org/10.1016/j.electacta.2011.12.110. M. G¨obel, et al., Multi-scale structural analysis of gas diffusion layers, J. Power Sources 355 (2017) 8
17. Available from: https://doi.org/10.1016/j.jpowsour.2017.03.086. P.A. Garcı´a-Salaberri, et al., Hydration and dehydration cycles in polymer electrolyte fuel cells operated with wet anode and dry cathode feed: a neutron imaging and modeling study, J. Power Sources 359 (2017) 634
655. Available from: https://doi.org/10.1016/j. jpowsour.2017.03.155. A.Z. Weber, et al., A critical review of modeling transport phenomena in polymerelectrolyte fuel cells, J. Electrochem. Soc. 161 (2014) F1254. Available from: https://doi. org/10.1149/2.0751412jes. K. Talukdar, et al., Minimizing mass-transport loss in proton exchange membrane fuel cell by freeze-drying of cathode catalyst layers, J. Power Sources 427 (2019) 309
317. Available from: https://doi.org/10.1016/j.jpowsour.2019.04.094. Y. Kim, et al., Fabrication of platinum group metal-free catalyst layer with enhanced mass transport characteristics via an electrospraying technique, Mater. Today Energy 20 (2021) 10064. Available from: https://doi.org/10.1016/j.mtener.2021.100641. A.Z. Weber, A. Kusoglu, Unexplained transport resistances for low-loaded fuel-cell catalyst layers, J. Mater. Chem. A 41 (2014) 7207
17211. Available from: https://doi.org/ 10.1039/C4TA02952F. A. Sa´nchez-Ramos, et al., Modeling the effect of low pt loading cathode catalyst layer in polymer electrolyte fuel cells: part I. model formulation validation, J. Electrochem. Soc. 168 (2021) 124514. Available from: https://doi.org/10.1149/1945-7111/ac4456. T. Schuler, et al., Fuel-cell catalyst-layer resistance via hydrogen limiting-current measurements, J. Electrochem. Soc. 166 (2019) F3020. Available from: https://doi.org/ 10.1149/2.0031907jes. L. Holzer, et al., Microstructure-property relationships in a gas diffusion layer (GDL) for polymer electrolyte fuel cells, part I: effect of compression and anisotropy of dry GDL, Electrochim. Acta 227 (2017) 419
434. Available from: https://doi.org/10.1016/j. electacta.2017.01.030. J.T. Gostick, et al., In-plane and through-plane gas permeability of carbon fiber electrode backing layers, J. Power Sources 162 (2006) 228
238. Available from: https://doi.org/ 10.1016/j.jpowsour.2006.06.096. M.S. Ismail, et al., Effect of polytetrafluoroethylene-treatment and microporous layercoating on the in-plane permeability of gas diffusion layers used in proton exchange membrane fuel cells, J. Power Sources 195 (2010) 6619
6628. Available from: https://doi.org/ 10.1016/j.jpowsour.2010.04.036. O.M. Orogbemi, et al., The effects of the composition of microporous layers on the permeability of gas diffusion layers used in polymer electrolyte fuel cells, Int. J. Hydrog. Energy 41 (2016) 21345
21351. Available from: https://doi.org/10.1016/j.ijhydene. 2016.09.160.

222

Fuel Cells for Transportation

[36] J. Zhao, et al., Gas permeability of catalyzed electrodes in polymer electrolyte membrane fuel cells, Appl. Energy 209 (2018) 203
210. Available from: https://doi.org/10.1016/j. apenergy.2017.10.087. [37] P.A. Garcı´a-Salaberri, et al., Analysis of representative elementary volume and throughplane regional characteristics of carbon-fiber papers: diffusivity, permeability and electrical/thermal conductivity, Int. J. Heat Mass. Transf. 127 (2018) 687
703. Available from: https://doi.org/10.1016/j.ijheatmasstransfer.2018.07.030. [38] H. Sadeghifar, et al., Effect of Polytetrafluoroethylene (PTFE) and micro porous layer (MPL) on thermal conductivity of fuel cell gas diffusion layers: modeling and experiments, J. Power Sources 248 (2014) 632
641. Available from: https://doi.org/10.1016/j. jpowsour.2013.09.136. [39] N. Zamel, Measurement of the through-plane thermal conductivity of carbon paper diffusion media for the temperature range from 2 50 to 1 120 C, Int. J. Hydrog. Energy 36 (2011) 12618
12625. Available from: https://doi.org/10.1016/j.ijhydene.2011.06.097. [40] J. Zhang, et al., A review of the microporous layer in proton exchange membrane fuel cells: materials and structural designs based on water transport mechanism, Renew. Sust. Energ. Rev. 156 (2022) 111998. Available from: https://doi.org/10.1016/j.rser.2021.111998. [41] A.D. Shum, et al., Investigating phase-change-induced flow in gas diffusion layers in fuel cells with x-ray computed tomography, Electrochim. Acta 256 (2017) 279
290. Available from: https://doi.org/10.1016/j.electacta.2017.10.012. [42] P.A. Garcı´a-Salaberri, et al., Implications of inherent inhomogeneities in thin carbon fiber-based gas diffusion layers: a comparative modeling study, Electrochim. Acta 295 (2019) 861
874. Available from: https://doi.org/10.1016/j.electacta.2018.09.089. [43] P.A. Garcı´a-Salaberri, 1D two-phase, non-isothermal modeling of a proton exchange membrane water electrolyzer: an optimization perspective, J. Power Sources 521 (2022) 230915. Available from: https://doi.org/10.1016/j.jpowsour.2021.230915. [44] D.C. Sabarirajan, et al., Determining proton transport in pseudo catalyst layers using hydrogen pump DC and AC techniques, J. Electrochem. Soc 167 (2020) 084521. Available from: https://doi.org/10.1149/1945-7111/ab927d. [45] P.A. Garcı´a-Salaberri, General aspects in the modeling of fuel cells: from conventional fuel cells to nano fuel cells, in: Nanotechnology in Fuel Cells, Elsevier, 2022, pp. 77
121, https://doi.org/10.1016/B978-0-323-85727-7.00015-1 [46] J.T. Gostick, et al., OpenPNM: a pore network modeling package, Comput. Sci. Eng. 18 (2016) 60
74. Available from: https://doi.org/10.1109/MCSE.2016.49. [47] M. Aghighi, J.T. Gostick, Pore network modeling of phase change in PEM fuel cell fibrous cathode, J. Appl. Electrochem. 47 (2017) 1323
1338. Available from: https://doi. org/10.1007/s10800-017-1126-6. [48] J. Hack, et al., X-ray micro-computed tomography of polymer electrolyte fuel cells: what is the representative elementary area? J. Electrochem. Soc. 167 (2020) 013545. Available from: https://doi.org/10.1149/1945-7111/ab6983. [49] D. Zapardiel, P.A. Garcı´a-Salaberri, Modeling the interplay between water capillary transport and species diffusion in gas diffusion layers of proton exchange fuel cells using a hybrid computational fluid dynamics formulation, J. Power Sources 520 (2022) 230735. Available from: https://doi.org/10.1016/j.jpowsour.2021.230735. [50] P.K. Das, et al., Effective transport coefficients in PEM fuel cell catalyst and gas diffusion layers: beyond Bruggeman approximation, Appl. Energy 87 (2020) 2785
2796. Available from: https://doi.org/10.1016/j.apenergy.2009.05.006.

Effective transport properties for fuel cells Chapter | 7

223

[51] M. Ahadi, et al., Thermal conductivity of catalyst layer of polymer electrolyte membrane fuel cells: part 2
analytical modeling, J. Power Sources 354 (2017) 215
228. Available from: https://doi.org/10.1016/j.jpowsour.2017.03.100. [52] I.V. Zenyuk, et al., Understanding impacts of catalyst-layer thickness on fuel-cell performance via mathematical modeling, J. Electrochem. Soc. 163 (2016) F691. Available from: https://doi.org/10.1149/2.1161607jes. [53] J. Wu, et al., Diagnostic tools in PEM fuel cell research: part I electrochemical techniques, Int. J. Hydrog. Energy 33 (2008) 1735
1746. Available from: https://doi.org/ 10.1016/j.ijhydene.2008.01.013. [54] J. Wu, et al., Diagnostic tools in PEM fuel cell research: part II: physical/chemical methods, Int. J. Hydrog. Energy 33 (2008) 1747
1757. Available from: https://doi.org/ 10.1016/j.ijhydene.2008.01.020. [55] M. Schulze, et al., Segmented cells as tool for development of fuel cells and error prevention/prediagnostic in fuel cell stacks, J. Power Sources 173 (2007) 19
27. Available from: https://doi.org/10.1016/j.jpowsour.2007.03.055. [56] P. Boillat, et al., Neutron imaging of fuel cells
recent trends and future prospects, Curr. Opin. Electrochem. 5 (2017) 3
10. Available from: https://doi.org/10.1016/j.coelec. 2017.07.012. [57] I.V. Zenyuk, Bridging x-ray computed tomography and computational modeling for electrochemical energy-conversion and
storage, Curr. Opin. Electrochem. 13 (2019) 78
85. Available from: https://doi.org/10.1016/j.coelec.2018.10.016. [58] A. Ozden, et al., A review of gas diffusion layers for proton exchange membrane fuel cells—with a focus on characteristics, characterization techniques, materials and designs, Prog. Energy Combust. Sci. 74 (2019) 50
102. Available from: https://doi.org/10.1016/j. pecs.2019.05.002. [59] N. Zamel, X. Li, Effective transport properties for polymer electrolyte membrane fuel cells
with a focus on the gas diffusion layer, Prog. Energy Combust. Sci. 39 (2013) 111
146. Available from: https://doi.org/10.1016/j.pecs.2012.07.002. [60] R.R. Rashapov, et al., Characterization of PEMFC gas diffusion layer porosity, J. Electrochem. Soc. 162 (2015) F603. Available from: https://doi.org/10.1149/2.0921506jes. [61] S. Shukla, et al., Determination of PEFC gas diffusion layer and catalyst layer porosity utilizing Archimedes principle, J. Electrochem. Soc 166 (2019) F1142. Available from: https://doi.org/10.1149/2.0251915jes. [62] J.T. Gostick, et al., Capillary pressure and hydrophilic porosity in gas diffusion layers for polymer electrolyte fuel cells, J. Power Sources 156 (2006) 375
387. Available from: https://doi.org/10.1016/j.jpowsour.2005.05.086. [63] T. Suzuki, et al., Investigation of porous structure formation of catalyst layers for proton exchange membrane fuel cells and their effect on cell performance, Int. J. Hydrog. Energy 41 (2016) 20326
20335. Available from: https://doi.org/10.1016/j.ijhydene.2016.09.078. [64] D.L. Wood, et al., Surface properties of PEMFC gas diffusion layers, J. Electrochem. Soc. 157 (2009) B195. Available from: https://doi.org/10.1149/1.3261850. [65] C.P. Liu, et al., Measurement of contact angles at carbon fiber
water
air triple-phase boundaries inside gas diffusion layers using x-ray computed tomography, ACS Appl. Mater. Interfaces 13 (2021) 20002
20013. Available from: https://doi.org/10.1021/ acsami.1c00849. [66] J.J. Conde, et al., Mass-transport properties of electrosprayed Pt/C catalyst layers for polymer-electrolyte fuel cells, J. Power Sources 427 (2019) 250
259. Available from: https://doi.org/10.1016/j.jpowsour.2019.04.079.

224

Fuel Cells for Transportation

[67] M.M. Gomaa, et al., Crosslinked PVA/SSA proton exchange membranes: correlation between physiochemical properties and free volume determined by positron annihilation spectroscopy, Phys. Chem. Chem. Phys. 20 (2018) 28287
28299. Available from: https:// doi.org/10.1039/C8CP05301D. [68] R.R. Rashapov, J.T. Gostick, In-plane effective diffusivity in PEMFC gas diffusion layers, Transp. Porous Media 115 (2016) 411
433. Available from: https://doi.org/10.1007/ s11242-016-0648-4. [69] P. Mangal, et al., Experimental study of mass transport in PEMFCs: through plane permeability and molecular diffusivity in GDLs, Electrochim. Acta 167 (2015) 160
171. Available from: https://doi.org/10.1016/j.electacta.2015.03.100. [70] E. Sadeghi, et al., A novel approach to determine the in-plane thermal conductivity of gas diffusion layers in proton exchange membrane fuel cells, J. Power Sources 196 (2011) 3565
3571. Available from: https://doi.org/10.1016/j.jpowsour.2010.11.151. [71] A. Drobnik, et al., Some aspects of the application of the laser flash method for investigations on the thermal diffusivity of porous materials, Laser Technol. III 1391 (1991) 361
369. Available from: https://doi.org/10.1117/12.57190. [72] X.Z. Yuan, et al., Development of a 3-in-1 device to simultaneously measure properties of gas diffusion layer for the quality control of proton exchange membrane fuel cell components, J. Power Sources 477 (2020) 229009. Available from: https://doi.org/10.1016/j. jpowsour.2020.229009. [73] N. Uren˜a, et al., On the conductivity of proton-exchange membranes based on multiblock copolymers of sulfonated polysulfone and polyphenylsulfone: an experimental and modeling study, Polymers 13 (2021) 363. Available from: https://doi.org/10.3390/polym13030363. [74] J.T. Gostick, et al., Direct measurement of the capillary pressure characteristics of water
air
gas diffusion layer systems for PEM fuel cells, Electrochem. Commun. 10 (2008) 1520
1523. Available from: https://doi.org/10.1016/j.elecom.2008.08.008. [75] A.D. Santamaria, et al., Liquid-water interactions with gas-diffusion-layer surfaces, J. Electrochem. Soc 161 (2014) F1184. Available from: https://doi.org/10.1149/2. 0321412jes. [76] A. Kusoglu, A.Z. Weber, New insights into perfluorinated sulfonic-acid ionomers, Chem. Rev. 117 (2017) 987
1104. Available from: https://doi.org/10.1021/acs.chemrev.6b00159. [77] J. Kleemann, et al., Characterisation of mechanical behaviour and coupled electrical properties of polymer electrolyte membrane fuel cell gas diffusion layers, J. Power Sources 190 (2009) 92
102. Available from: https://doi.org/10.1016/j.jpowsour.2008.09.026. [78] M.N. Silberstein, M.C. Boyce, Constitutive modeling of the rate, temperature, and hydration dependent deformation response of Nafion to monotonic and cyclic loading, J. Power Sources 195 (2010) 5692
5706. Available from: https://doi.org/10.1016/j.jpowsour. 2010.03.047. [79] F. Jaouen, et al., Oxygen reduction activities compared in rotating-disk electrode and proton exchange membrane fuel cells for highly active FeNC catalysts, Electrochim. Acta 87 (2013) 619
628. Available from: https://doi.org/10.1016/j.electacta.2012.09.057. [80] K.C. Neyerlin, et al., Study of the exchange current density for the hydrogen oxidation and evolution reactions, J. Electrochem. Soc. 154 (2007) B631. Available from: https:// doi.org/10.1149/1.2733987. [81] K.C. Neyerlin, et al., Determination of catalyst unique parameters for the oxygen reduction reaction in a PEMFC, J. Electrochem. Soc. 153 (2006) A1955. Available from: https://doi.org/10.1149/1.2266294.

Chapter 8

Liquid water transport and management for fuel cells Anthony D. Santamaria1 and Prodip K. Das2 1

General Motors, Global Propulsion Systems Pontiac Engineering Center, Pontiac, MI, United States, 2School of Engineering, The University of Edinburgh, Edinburgh, United Kingdom

8.1

Water production

Water in proton exchange membrane fuel cell (PEMFC) is produced in the cathode catalyst layer (cCL) as a result of the oxygen reduction reaction with hydrogen and then transported through the diffusion media [gas diffusion layers (GDLs) and microporous layers (MPLs)] via two key mechanisms: convection and diffusion. Pressure differences drive convective fluid flow, while concentration gradients drive the diffusion process. For nonisothermal conditions, temperature-gradient-driven flow, also known as phase-changeinduced (PCI) flow, also plays an important role in water transport in PEMFC [1,2]. Understanding these processes in conjunction with the structure of each PEMFC layer, and the interfaces between them, is critical to developing an overall water balance picture. The rate of water produced is directly proportional to the current density which can be expressed as: i m_ H2 O 5 MH2 O ð8:1Þ 2F where m_ H2 O is the mass flow rate of water, i is current, F is Faraday’s constant, and MH2 O is the molecular weight of water. The water content of cCL is dominated by the membrane and MPL uptake rates whose conditions can be controlled via cathode and anode reactant gas pressure, temperature, humidity, flow rate, and CL’s wetting properties [3,4]. When reactant gases become saturated, which is often the case under normal PEMFC operating conditions, condensation leads to liquid accumulation. The ratio of water contained in the gas phase can be determined by calculating the specific humidity, ω:
 MH2 O Pv ω5 ð8:2Þ MAir P 2 Pv

Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00002-2 © 2023 Elsevier Ltd. All rights reserved.

225

226

Fuel Cells for Transportation

where MAir is the molecular weight of the nonwater gas species (air for this example), P is the total gas pressure, and Pv is the saturation vapor pressure. The saturation pressure varies with temperature and can be obtained by looking them up in the reference tables in the ASHRAE Handbook. Condensation occurs when the partial pressure of water in the gas stream exceeds the saturation vapor pressure; the temperature at which this occurs is known as the dew point. Water, therefore, exists as a two-phase system, both a vapor and a liquid, throughout the membrane electrode assembly (MEA) and gas flow channels (GC). The MEA in a fuel cell includes polymer electrolyte membrane (PEM), CLs, MPLs, and GDLs. Two-phase flow can lead to liquid saturation of critical porous electrode regions which deleteriously affects oxygen diffusion rates and may lead to flooding as well as chemical and mechanical degradation. Conversely, proton conduction in the membrane depends on water; the electrolyte must remain well hydrated to minimize ionic resistance. Overly dry conditions can lead to membrane dehydration which also negatively impacts the performance and lifetime of PEMFC. Therefore, developing strategies for PEMFC’s design and operation aimed at maintaining proper hydration while mitigating flooding has been, and continues to be, of prime importance. In this chapter, the processes associated with the transport of liquid water from CL to GC are discussed for each layer of the MEA and reactant channels which are outlined in Fig. 8.1.

8.2

Two-phase flow basics

Due to the small length scales of the MEA, capillary and viscous forces govern two-phase flow; the dimensionless parameters that quantify them are the capillary number, Ca, and viscosity ratio, M, defined as: uμ ð8:3Þ Ca 5 nw γ and M5

μnw μwet

ð8:4Þ

respectively, where u is the superficial velocity of the nonwetting phase, γ is the surface tension, and μwet and μnw are the wetting and nonwetting phase viscosities, respectively. For PEMFC, liquid transport occurs predominately in the capillary fingering (CF) regime for hydrophobic pores [6]. In the CF regime, the intruding fluid has a viscosity greater than the viscosity of the displaced fluid, and the flow rate of the intruding fluid is low, as shown in the drainage phase diagram in Fig. 8.2. Considering the in-plane percolation of the intruding fluid, the CF flow regime features the formation of a network of irregular conduits, or fingers, within the porous media. The presence of capillary-driven transport in these systems means that properties related to

Liquid water transport and management for fuel cells Chapter | 8

227

FIGURE 8.1 Schematic diagram of water transport through PE (Nafion), CL, MPL, GDL, and GC porous layers highlighting capillary and electroosmotic liquid water flows, water production by the electrochemical reaction, and percolation throughout all the layers. Environmental SEM images of water on PE, MPL, and GDL surfaces are also shown [5]. Reprinted from Hwang et al., J. Electrochem. Soc. 156 (10) (2009) B1192
B1200, with permission from the Electrochemical Society.

FIGURE 8.2 (A) Drainage phase diagram showing key transport regimes. In the MEA of PEMFC, Ca between 1028 to 1025 and M around 17.5 is typical, indicating liquid water undergoes capillary fingering within the porous media. (B) Example of typical liquid saturation versus capillary pressure curves for two different polytetrafluoroethylene (PTFE) loadings of GDL [7]. Reprinted from Das and Weber, Proc. of the ASME 11th Fuel Cell Science, Engineering and Technology Conference, Paper No. FuelCell2013-18010 (2013), pp. V001T01A002, with permission from the ASME.

228

Fuel Cells for Transportation

material wetting strongly influence transport behavior and can have large implications for electrochemical performance. Liquid imbibition in pores is governed by the Young
Laplace equation: Pc 5 2

2γcosθs r

ð8:5Þ

where Pc is the liquid capillary pressure, θs is the static liquid contact angle with the electrode surface, and r is the liquid surface radius. Capillarydriven flow is always affected by pore structure and its hydrophobicity. Smaller radius droplets, for example, have a higher liquid pressure; therefore, they tend to be absorbed into larger droplets they encounter. Since GDL uptake of water results in liquid-filled pores unavailable to gas diffusion, it is useful to determine the relationship between capillary pressure and liquid saturation. Experimentally measured capillary pressure
water saturation curves (Fig. 8.2B) show hysteresis between liquid uptake and withdrawal processes due to changing water pathways resulting from the dynamic nature of liquid interfaces [7,8]. Another technique used to assess the barrier for liquid transport through a GDL is to measure breakthrough pressure [9,10]. Breakthrough pressure is the maximum capillary pressure observed when injecting liquid through a GDL. During this process, pressure has been observed to build steadily until reaching a breakthrough point where pressure is relieved by droplet formation on the opposite surface [11,12]. This method can capture the drainage process associated with fully saturated conditions where the liquid is the primary phase present in the GDL. Liquid water that emerges from the GDL surface builds up in the flow channels as droplets. The liquid
gas two-phase flow patterns that occur in PEMFC channels typically fall under the categories of slug flow and annular flow [13]. Fig. 8.3A shows closeups of these and other common flow patterns associated with PEMFC operation. Two-phase flow in PEMFCs is

FIGURE 8.3 (A) Overview of common two-phase flow patterns observed in PEMFC. (B) Two-phase flow regime trends relative to superficial gas and liquid velocities. Reprinted from Hussaini et al., J. Power Sources 187 (2) (2009) 444
451, with permission from Elsevier.

Liquid water transport and management for fuel cells Chapter | 8

229

mainly dominated by the gas phase with mass flow quality, x, greater than 0.9. The mass flow quality, x, is defined as: x5

Gg Gl 2 Gg

ð8:6Þ

where Gg and Gl are gas-phase mass flux and liquid-phase mass flux, respectively [14,15]. The mass flux is the product of phase density and superficial velocity. The two-phase flow regime observed in channels is directly linked to both air and liquid flow velocities. Fig. 8.3B presents an overview of flow regime trends. Typically, higher gas velocities and lower liquid velocities trend toward single-phase conditions, while lower gas velocities and higher liquid flow rates lead to droplets and slugs. Droplet formation is very common under a wide range of PEMFC operating conditions. Significant effort has been devoted toward understanding liquid droplet growth and removal from GDL surfaces as they can be a precursor to slug formation and channel flooding. Several models of droplet detachment have been proposed using a force balance between inertial, drag, and surface adhesion for simplified droplet geometries and aerodynamic drag correlations [16]. Their development was motivated in part by the need to approximate the reactant channel gas speeds required for liquid removal. In general, droplet detachment versus gas velocity curves have been shown to follow a power-law relationship; smaller droplets require significantly higher gas velocities for removal than larger ones. Adhesion force, sometimes referred to as the surface tension force, may be inferred by considering the line of contact between the droplet and GDL (assumed to be a circle where the sphere meets the GDL) and the dynamic angle, through the relationship: 1 Fγx 5 πdγsin2 θs sin ðθa 2 θr Þ 2

ð8:7Þ

where Fγx is the surface tension force, d is the wetted diameter, and θa and θr are the advancing and receding angles, respectively [17]. Experimental validation usually relies on ex situ transparent test channels where droplets are grown on the GDL and removed via flow gases while images of the droplet dynamics are captured for analysis. These models are useful for predicting droplet instability over a host of operational, design, and GDL surface treatment conditions [18]. While the general agreement with experimental data is achieved, this adhesion approximation may be limiting as it relies on a single dynamic angle rather than a variable one and necessitates a complicated experimental setup. Numerical volume-of-fluid (VOF) models have also explored the incorporation of experimentally measured static and dynamic droplet angles which allowed for reasonable estimation of droplet dynamics and detachment [19]. The wetting of GDL surfaces is complex due to their inherent roughness and chemically heterogeneous surface

230

Fuel Cells for Transportation

which influence adhesion measurements. An alternative and simpler method for calculating liquid adhesion force to the GDL surface is to measure a droplet’s sliding angle, which has been demonstrated by various techniques [8,20]. Sliding angle, θs , refers to the angle at which a tilted droplet falls off due to gravity. The maximum adhesion, fa , between the liquid and the surface can then be calculated as a ratio of gravity forces to approximated wetting perimeter [8]: fa 5

ρVgsinθs πdw

ð8:8Þ

where ρ is the liquid density, V is the liquid volume, g is gravity, and dw is the wetted diameter. Since adhesion is calculated on a force-per-length basis, the total adhesion force can be calculated by Fγ 5 fa μdw . This method can be employed by either directly pipetting droplets onto the GDL surface or by injecting water through the GDL to simulate in situ liquid droplet formation. Fig. 8.4A diagrams the technique, and Fig. 8.4B shows the typical trend in sliding angle with respect to droplet volume for the two droplet formation methods. Usually, the GDL injection method requires a higher sliding angle for detachment corresponding to larger adhesion; further details are discussed in Refs. [8,12]. Early work has shown the potential for using sliding-anglebased adhesion measurements to approximate droplet detachment and capture effects due to surface wetproofing, liquid flow rate variation, vibration, and aging [12,21
23]. Several direct and indirect methods exist to investigate liquid
gas twophase flow in fuel cell reactant channels. The former includes transparent cell analysis [24], X-ray tomography [25], Neutron imaging (including 2D

FIGURE 8.4 (A) Schematic of a droplet undergoing the sliding angle test. (B) Typical trends observed for sliding angle with respect to droplet volume as well as both pipetted (top injection) and through GDL injected droplets (bottom injection) [8]. Reprinted from Das et al., J. Electrochem. Soc. 159 (5) (2012) B489
B496, with permission from the Electrochemical Society.

Liquid water transport and management for fuel cells Chapter | 8

231

radiography and 3D tomography) [26
28], and gas chromatography [29]. While direct imaging offers far more certainty as to the distribution of liquid water, as depicted in Fig. 8.5A, the limited beamline facilities, complicated setups, and high costs associated with these methods mean they are usually reserved for the highest priority studies. Indirect diagnostics, such as pressure drop, high-frequency resistance, and water mass accounting, offer simplified methods to study water’s effects and may be combined with direct imaging [30]. A major drawback to indirect methods is that they presently do not resolve the specific liquid distribution as well as imaging. One indirect method commonly employed relies on the two-phase flow pressure drop signal which is an immediate result of water accumulation in the flow channels. Studies may define a pressure amplification coefficient, ϕTP : ϕTP 5

ΔPTP ΔPSP

ð8:9Þ

which is the ratio of the actual measured two-phase pressure drop, ΔPTP , divided by the single-phase pressure drop, ΔPSP , for the gas phase [31]. Overall, the two-phase pressure drop is considered a reliable in situ diagnostic tool for monitoring the overall state of liquid water in PEMFC [32]. A comprehensive review of two-phase pressure drop in PEMFC channels is available in Ref. [33]. Despite significant efforts [34
36], an accurate prediction of two-phase flow pressure drop is still challenging. Theoretically,

FIGURE 8.5 (A) 3D Neutron tomography of a PEMFC. (B) Ex situ two-phase flow data for steady-state airflow and increasing water flow rate (WFR) through the GDL. Reprinted from Tang et al., J. Power Sources 195 (19) (2010) 6774
6781 and Niknam et al., Results Eng. 5 (2020) 100071, with permission from Elsevier.

232

Fuel Cells for Transportation

pressure drop (ΔPTP ) for two-phase flow is a function of frictional (ΔPF ), gravitational (ΔPG ), and acceleration (ΔPA ) pressure drops: ΔPTP 5 ΔPF 1 ΔPG 1 ΔPA

ð8:10Þ

The acceleration pressure loss is negligible due to the relatively low superficial gas velocities in PEMFC’s flow channels. Gravitational pressure loss is also insignificant in channels due to the dominant influence of surface tension. Therefore, the frictional two-phase flow pressure drop is adequate to approximately explain the total pressure loss in PEMFC channels [37]. Experimentally, the two-phase pressure drop may be calculated, in situ or ex situ, by ΔPTP 5 Pin 2 Pout which is the pressure difference between the inlet and outlet of a channel, flow field, or stack manifolds depending on testing scale. Fig. 8.5B demonstrates time-series ex situ single-channel, two-phase pressure trends at various water injection rates and at a fixed air flow rate. Reasonable predictions for the frictional two-phase flow pressure drop can be achieved through homogeneous equilibrium or separated flow models. The homogenous equilibrium model works well for high mass qualities; however, the model is not as useful for PEMFC applications [36]. The separated flow model is based on the summation of pressure gradients in the gas phase, liquid phase, and the interaction of gas
liquid, as shown in the equation below [38]: "



 
 #1=2 dp dp dp dp dp 2 5 2 1 2 1C 2 2 ð8:11Þ dz TP dz f dz g dz f dz g where p, z, and C are the pressure, streamwise coordinate, and Chisholm parameter, respectively. The subscripts TP, f , and g represent two-phase, saturated liquid, and saturated vapor, respectively. The prediction accuracy depends significantly on Chisholm parameter, C [39]. This parameter is a function of multiple factors such as flow regime and capillary geometry [40]. This empirical model has been shown to reasonably predict two-phase signatures for certain ranges of steady-state operation [41]. Other numerical-based two-phase modeling efforts utilize VOF and lattice Boltzmann methods, among others, to simulate gas
liquid
solid interfaces and interactions. Studies have investigated liquid droplet formation and detachment predicted by VOF schemes and have demonstrated reasonable agreement with experimental data on the local level [19,42]. Larger fullscale flow-field simulations are a challenge due to the increased computational domain size as well as complex boundary conditions, such as the rough GDL surface, which has a significant influence on liquid transport. The emergence of water from the GDL’s randomly distributed fibrous and chemically heterogeneous surface leads to droplet pinning effects that are difficult to handle computationally and are of current interest to the research community [43]. Two-phase modeling is necessary as single-phase models

Liquid water transport and management for fuel cells Chapter | 8

233

fail to capture the complex behavior resulting from liquid water physics in porous media and channels [44,45]. The need for effective two-phase models is paramount to future PEMFC engineering where significant cost and time savings may be achieved during the design iteration process. In assessing overall system performance for a PEMFC, the power consumption due to a two-phase pressure drop must be accounted for. The net system power (not accounting for humidification and other subsystems) can be calculated by subtracting the pumping power from the cell or stack power output [46]. The pumping power is calculated using the inlet and outlet pressures of a PEMFC system. The following demonstrates the process applied to the cathode side. For the case of a standard air blower, the equations [47]:
 iAξc MO2 m_ Air 5 ð8:12Þ 4F 0:21 and m_ Air Cp T E_ 5 η




Pin Pout

k21 k

! 21

ð8:13Þ

_ where A is the active cell area, maybe used to estimate power consumption, E, ξ c , is the cathode stoichiometry, MO2 is the molecular weight of oxygen, Cp is the specific heat of air, T is the blower inlet air temperature, η is the pump efficiency, Pin is the air pressure at the blower inlet (usually set to atmosphere), Pout is the blower outlet pressure, and k is the specific heat ratio. The outlet pressure for the blower may be related to the inlet pressure of the fuel cell system which is directly impacted by the two-phase pressure drop ΔPTP of the cell. The parasitic effects of liquid buildup can be isolated by comparing the power required using the two-phase pressure signal to that of single-phase flow. For the total cell pumping power, the anode losses would also need to be considered in cases where a hydrogen pumping system is present.

8.3

PEMFC architecture

Traditional MEAs are multilayer architectures consisting of PEM, CLs, MPLs, and GDLs as highlighted in earlier chapters. For final assembly, each layer is laid up, and the MEA is formed via heat pressing. In a cell, gaskets are utilized to seal the MEA which is compressed between a set of flow fields or bipolar plates. The flow field contains the channels responsible for reactant delivery and product removal. Overall, the MPL and GDL are composed of randomly distributed carbon powder particles and fibers. Due to the current manufacturing process and hydrophobic treatment methods, MPLs and GDLs are chemically and structurally heterogeneous [8]. Allowing effective gas diffusion while providing liquid pathways is a primary function of these layers. Studies have identified specific pore structures and distributions

234

Fuel Cells for Transportation

that could enhance liquid drainage and improve reactant delivery to catalyst sites; however, these designs are beyond standard manufacturing capabilities [48]. The inability to control structure at these scales using current manufacturing methods and materials significantly limits technologies such as PEMFC from reaching full potential. MEA design and manufacturing remain large engineering challenges due to the small scales, complex interfaces, and specific material requirements. Nonlinearities due to highly coupled physical phenomena including thermal, fluid, and charge transport, highlight the need for MEAs to be studied in a holistic manner [49]. The following sections discuss liquid transport on a layer-by-layer basis.

8.3.1

Membrane

Proton transport occurs through the ionic conducting PEM. While there are many variants, the commonly used persulfonated PTFE, a trade named Nafion, consists of a PTFE backbone with sulfonic acid (SO32H1) functional groups to facilitate charge transport. The ionic conductivity of Nafion is strongly dependent upon water content and increases with rising humidity. Therefore, maintaining a properly hydrated membrane is critical to electrochemical performance as well as preventing degradation. Liquid generated at the cathode CL tends to back-diffuse through the membrane toward the drier anode. This is opposed by the protons migrating from the anode to the cathode through a phenomenon known as electroosmotic drag as highlighted in the previous chapters. The balance of water within the membrane, as well as the net flux of water through it, and is sensitive to and can be controlled by cathode and anode reactant conditions, electrode structure, and water generation rate (current density). Recently, a shift to thinner membranes has been pursued which promotes higher water back-diffusion from the cathode to the anode. Thinner membranes offer lower ohmic losses, while some drawbacks may be higher gas crossover rates and susceptibility to punctures.

8.3.2

Catalyst layer

The CL, which is sandwiched between the membrane surface and GDL, is a thin film of carbon-supported catalyst. Traditionally, CL material consists of several nanometer-sized particles of platinum deposited on carbon powder (such as Vulcan XC72) and then bound with an ionomer. The catalyst reduces the reaction activation energy, while the carbon support and ionomer binder provide electron and proton conduction pathways, respectively. Product water forms in the cCL, and therefore, sufficient pore space must be available for liquid water to effectively permeate away from reaction regions to avoid gas blockages. The anode catalyst layer also impacts cCL water migration by affecting back-diffusion through the PEM. At higher temperatures, PCI water transport from the CL occurs in vapor form; however, at

Liquid water transport and management for fuel cells Chapter | 8

235

lower temperatures such as during startup, liquid formation may be unavoidable. A key goal of commercializing PEMFC technology is reducing platinum group metal (PGM) loadings while achieving increased power density. This can be partially accomplished by using thinner CLs. While typically a CL can be as thick as 30 μm, current research into ultrathin CL, targeting ,1 μm, poses large challenges for water removal due to flooding sensitivity [49].

8.3.3

Microporous layer

The MPL serves a multifaceted role including providing a protective buffer between the GDL and membrane, enhancing electrical and thermal conductivity, and reducing interface liquid buildup. It is deposited directly onto the GDL surface and is usually composed of fine carbon powder as well as PTFE for hydrophobicity. This layer is typically less than 50 μm thick and has a porosity of B50% with pore diameters ranging between 0.05 and 1 μm [50]. Without an MPL, liquid tends to indiscriminately fill GDL pores at the CL interface which reduces oxygen and hydrogen diffusion rates. The addition of an MPL has been shown to prevent such buildup by providing liquidspecific pathways, usually in the form of randomly distributed micro-cracks, that facilitate water transport while allowing smaller pores to remain open for gas diffusion. Traditionally, manufactured MPLs can be modified to improve performance. Methods usually entail providing effective pathways for gas and liquid transport. For example, small perforations in the MPL have been attributed to higher performance in PEMFC [51]. The addition of larger micron-sized holes has been observed to favor liquid flow which prevents water from indiscriminately filling all pores at the MPL/GDL interface. Other recent work, diagramed in Fig. 8.6A, has demonstrated the use of pore-forming agents which open randomly distributed micro-holes (B10 μm diameter) in the MPL [52]. These modified MPLs were compared to an untreated MPL using fast operando X-ray tomographic microscopy where it was confirmed the MPL with micro-holes resulted in enhanced liquid drainage and oxygen transport. Such results agree with a growing consensus that larger holes better facilitate liquid flow while small pores enable gas diffusion. Similar improvements have also been achieved using perforated GDL [53]. As these methods improve, controlling hole location and pattern may be of significance since MPL/GDL drainage may be affected by flow-field channel and rib locations.

8.3.4

Gas diffusion layer

The carbon paper, or sometimes carbon cloth, based on GDL provides larger pores for liquid uptake from the CL or MPL and is usually coated with a

236

Fuel Cells for Transportation

FIGURE 8.6 (A) MPL holes for liquid drainage and (B) porous gradient in electro-spun bilayer GDL. Reprinted from Nagai et al. J. Power Sources 435 (2019) 226809, with permission from Elsevier (Part A) and Balakrishnan et al. ACS Appl. Energy Mater. 3 (2020) 2695
2707 (Part B).

hydrophobic treatment such as PTFE at 5%
10% wt. loadings. Loadings higher than 10%, while available, have been shown to have minimal impact on hydrophobicity and may reduce performance as PTFE fills void space where liquid percolation or oxygen diffusion would occur [12]. Typical pore sizes of GDL range from 10 to 100 μm. At these scales, capillary and viscous effects dominate two-phase flow. The CF flow regime associated with low liquid velocities observed in GDLs is characterized by the invasion of pores by the nonwetting fluid (in this case liquid water) by displacing the wetting fluid (reactant gases). Under saturated vapor conditions, liquid builds up at the MPL/GDL interface. As the liquid pressure rises, it forces the liquid front to fill pores following a path of least resistance; larger pores may be filled first as well as those with lower hydrophobicity. This results in a pattern of irregular conduits and branches propagating along with the GDL thickness that look like “fingers” for which this flow regime is named. As GDLs are 100 to 400 μm thick, liquid water may traverse dozens of pores before reaching the GDL/ channel interface. As the water begins to penetrate the GDL, it encounters resistance due to capillary effects which cause the liquid pressure to increase. At some point, the pressure is enough to overcome the maximum resistance, and a breakthrough occurs whereby water reaches the GDL/channel interface [12]. The maximum pressure, referred to as breakthrough pressure, is a commonly used measurement of a GDLs propensity for through-plane liquid transport. Lower breakthrough pressures may be desirable to facilitate liquid drainage. Thinner GDLs and reduced PTFE loadings have been shown to reduce breakthrough pressure [12]. The amount and location of water for both in-plane and through-plane GDL directions are challenging to map due to the small feature sizes and opaque materials. Experimental methods

Liquid water transport and management for fuel cells Chapter | 8

237

utilizing neutron and X-ray imaging techniques have, more recently, uncovered capillary versus saturation trends with such detail [54]. Additionally, numerical pore network models, which apply interface physics to GDL structures to simulate water pathways, have been useful in identifying the impact of key GDL morphologies and properties on liquid saturation [55
57]. GDL porosity is also affected by the flow-field compression which leads to reduced porosity beneath land or rib areas. Compression is necessary for good electrical contact between the plates and porous media; however, overcompression leads to GDL damage, higher pressure drop, and reduced performance [58]. A compressed GDL will intrude into the channel which reduces the cross-sectional flow area while also reducing the angle between the GDL and channel wall leading to increased flow resistance for both gas and liquid phases. The interaction between the flow field and the GDL surface is also sensitive to temperature effects. For example, the location of coolant channels may cause temperature gradients which could cause liquid condensation in the local GDL. In addition to adding perforations for liquid-specific pathways, as discussed in the MPL section, GDL designers have proposed pore gradientstructured GDL to take advantage of the capillary-driven flow. Recently, groups have explored the use of transverse pore gradients for enhanced liquid drainage using electrospinning methods to create bilayer GDL [59]. As shown in Fig. 8.6, the catalyst interfacing layer consisted of average pore diameter of about 175 nm while the channel interface layer’s average pore diameter was about 687 nm. Using X-ray radiography, liquid water content was found to be reduced in the bilayer GDL compared to a traditional GDL. Further discussion on advanced porous media is touched upon in Section 8.11.

8.4

Channels and flow fields

Channel geometry, including land width, channel width, and channel depth, significantly influences two-phase flow behavior and impacts PEMFC performance. Extensive in situ experimental testing over a wide range of channel dimension combinations with depths and widths ranging from 0.25 to 1 mm has shown that optimal performance, accounting for pumping power at higher pressures, is achieved at a hydraulic diameter of B0.4 mm for certain ranges of stoichiometry [60,61]. Other results focusing on channel and rib widths (in parallel designs) have shown that narrower dimensions benefit high current density operation while wider benefit low current density conditions; overall, widths between 0.7 and 1 mm achieve a balance of high power and reduced manufacturing effort [62,63]. A 3D numerical simulation of interdigitated and parallel flow-field designs has demonstrated that sub-0.5mm characteristic length channels were found to have the best performance which was attributed to improved water removal due to higher pressures

238

Fuel Cells for Transportation

[64]. At B0.5 mm, peak performance is also achieved for square channel dimensions in interdigitated, serpentine, and spiral interdigitated designs [65]. While they are increasingly powerful tools and excellent at singlephase studies, larger-scale two-phase simulations should be supported by experimental verification when possible. Flow-field engineering usually requires extensive iteration between modeling, development, and testing to understand a new design’s behavior. An overview of flow-field patterns, which play a significant role in liquid water build and the operating parameters, is presented in Fig. 8.7. Traditional flow fields, which have been studied extensively, include serpentine, parallel, and interdigitated designs. Over the past few decades, several other layouts have been proposed which are briefly introduced in the following. Serpentine flow field: Serpentine flow patterns, consisting of either a long single path (Fig. 8.7A) or multiple paths (Fig. 8.7B), are effective at removing liquid water due to the higher pressure drop and gas velocities across the cell. However, a larger pressure drop across the cell can result in a significant concentration gradient between the inlet and outlet, and high pumping losses. Pressure differences between adjacent channels induce crossflow through the GDL which can reduce diffusion lengths beneath land areas improving performance and reducing overall pressure drop [66]. In these designs, the water content can be reduced at higher current densities, for stoichiometric-based flow, due to the elevated pressure drop forcing droplets from channels. Parallel flow field: Parallel flow fields consist of a series of channels connected in parallel by an inlet and outlet manifold (Fig. 8.7B). Generally, these have the lowest pressure drop due to a large effective cross section for flow. Since the pressure drop across each channel is similar, convective crossflow between adjacent channels is minimal. This makes diffusion the dominant gas transport mechanism. In these designs, typically, narrower land widths and shallower channel depths are beneficial to performance [40,67]. The lower pressure and multiple routes for reactant gases can lead to significant water accumulation and reactant maldistribution, especially under land areas where flooding can occur. Some of these issues may be overcome by using high aspect ratio designs (i.e., longer length channels) which results in higher gas velocities and pressure drop to force liquid removal. One benefit to parallel flow fields is that they can provide reduced pressure drop compared to more complicated designs [44,45,68]. Thus, they are better suited for lower current density operation, and hybrid designs that shift between interdigitated (at high current density) have been demonstrated [46]. Parallel flow fields may be improved by adding channel wave patterns which can enhance convective transport with minimal pressure drop [69]. Interdigitated flow field: While resembling parallel patterns, interdigitated inlet channels are not directly connected to the outlet channels (Fig. 8.7D). This results in higher-pressure inlet channels driving reactant

Liquid water transport and management for fuel cells Chapter | 8

239

FIGURE 8.7 Overview of different flow-field designs for PEMFCs. Part (E) is adapted from Guo et al., Int. J. Hydrog. Energy 39 (36)(2014) 21185
21195, with permission from Elsevier, and part (F) is adapted from Fly et al., Energies 12 (7) (2019) 1186.

flow through the GDL to adjacent lower-pressure outlet channels. While convective transport in the porous media has been shown to significantly boost electrochemical performance, these gains require higher pumping power. Interdigitated designs are especially sensitive to rib width, GDL properties, and compression. Studies have also shown a propensity for inlet channel flooding in larger aspect ratio designs [45]. Thus, further efforts are needed to improve the performance of interdigitated flow fields.

240

Fuel Cells for Transportation

Biomimetic/fractal flow field: Biological-inspired and fractal-based designs use patterns derived from nature to maximize gas delivery to the CL (Fig. 8.7E) [70]. While a growing interest has motivated computational fluid dynamic modeling and theoretical studies focusing on optimization, understating these systems’ performance is in the early stages [71,72]. Given the complex geometry sometimes involved, a broader understanding of how these designs handle liquid water and perform in stacks is needed. Metal foams and meshes: These offer alternatives to traditional channel and rib architectures (Fig. 8.7F) [73]. Metal foam’s high porosity, compressibility, and electrical conductivity make them potentially well suited for PEMFC applications such as gas distributors and coolant channels [74,75]. Metal foams have been demonstrated to provide even delivery of reactant gases and avoid the GDL liquid buildup associated with land areas [76]. In several cases, foam-based flow fields have outperformed channel-based ones [77]. One challenge related to liquid management is that a lack of channels to direct water may lead to water buildup in the foam pores. Recent work has focused on understanding liquid transport in foams as well as targeting solutions such as hydrophobic coatings to improve liquid drainage [78]. Unlike foams, which have random pore structures, metal meshes can be manufactured with specific porous features. Toyota used 3D metal meshes for the cathode plates in the first-generation Mirai cells [79]. Features such as baffling may be included to improve convective transport while providing pathways for liquid uptake from the GDL. Single cells are combined to form stack assemblies, whereby individual flow-field plates are connected via larger manifolds. These manifolds are responsible for bulk reactant delivery and exhaust and so must handle the net liquid water removed from all cells. Manifolds usually connect single cells in parallel; therefore, blockages in any single cell can lead to overall stack instability. For some systems, humid exhaust air is passed through a condenser to remove water which may be recycled for humidification. In vehicles such as the Toyota Mirai, excess liquid water is drained periodically (some models include a manual drain button). The behavior of liquid water in PEMFC flow channels depends on the liquid-to-gas ratio, the superficial velocities of each phase, the surface characteristics of the channel and GDL, and the channel geometry. Pressure drop measured across the length of a channel is important as engineers seek to remove water with minimal parasitic energy consumption. Depending on the channel size, different forces can affect the transport mechanism of twophase flow [37]. Channels may be classified based on channel hydraulic diameter, Dh: (i) conventional channels Dh . 3 mm, (ii) minichannels 3 mm . Dh . 200 μm, and (iii) microchannels 200 μm . Dh . 10 μm [80]. Modern PEMFC flow channels usually lie in the minichannel range, while some channel features may be on the microchannel scale. Liquid transport in PEMFC flow channels is capillary dominated as the dimensionless parameter

Liquid water transport and management for fuel cells Chapter | 8

241

Bond number is less than 1 [13]. The Bond number, which is a ratio of gravitational acceleration effects to surface tension effects on the liquid
vapor interface, is defined as: Bo 5

ρgl2 σ

ð8:14Þ

where ρ is the liquid density, g is gravity, l is the characteristic length, and σ is the surface tension. Due to the small characteristic length scale associated with the flow-field channels, the gravitational impact on the liquid
gas twophase flow is insignificant (it may, however, have an impact in larger manifold regions). Another important parameter is the dimensionless Weber number: We 5

ρv2 l σ

ð8:15Þ

where v is the gas velocity, and We is the ratio of aerodynamic drag to capillary force. Liquid removal relies on aerodynamic drag; viscous and pressure forces must exceed liquid adhesion to channel and GDL surfaces for removal. Increasing gas velocity, associated with a higher Weber number, is a very effective method for managing liquid buildup; however, it leads to higher pressure and, therefore, higher parasitic power consumption by the blower or compressor. Cross-sectional channel geometry impacts liquid uptake from the GDL surface to flow channels. Capillary wetting of channel/GDL corner regions is predicted by the Concus
Finn condition which depends on the hydrophobicity of the channel walls and GDL surface, as well as the draft angle between them [81]. Rath et al. defined this limit for PEMFC applications as [82]: θwall 5 ð2α 1 πÞ2θGDL

ð8:16Þ

where θwall is the limiting wall contact angle, θGDL is the GDL contact angle, and 2α is the open angle between the channel wall and the GDL surface. This relationship can be used to determine the propensity for a channel design to wick water from GDL surfaces for improved water management [83]. Results have shown that for typical GDL and hydrophilic flow channels, corner filling will not occur when α is 52 degrees or smaller.

8.5

Liquid management concerns and strategies

To prevent porous media and channel flooding, liquid buildup must be continuously monitored and controlled. As discussed in prior sections, flooding can lead to reactant gas redistribution which causes performance instability and degradation. During operation, water accumulation may be monitored indirectly by sensors and controlled by adjusting gas flow rates, back pressure, cell temperature, gas humidity, and/or current density. Additionally,

242

Fuel Cells for Transportation

there are many MEA and flow-field design strategies that have been explored to better accommodate liquid while also enhancing its removal. The following sections discuss liquid management as it relates to several relevant PEMFC topics.

8.6

Pressure and flow control

Liquid removal from reactant channels is facilitated primarily by aerodynamic drag forces. Higher gas flow rates reduce liquid buildup; however, they require greater pumping power. Ideally, designers would seek to operate a PEMFC at the lowest stoichiometry that maintains stable desired performance. For the cathode, the air compressor is adjusted to supply necessary flow rates and can be varied based on sensor feedback. For example, since liquid buildup leads to higher pressure drop, pressure transducer signals may be used to guide compressor speeds. On the anode, where hydrogen may be supplied from a pressurized tank and fuel distribution and utilization are a large concern, higher flow rates may be achieved using a circulating pump [84]. For stoichiometric-based gas control, higher flow channel velocities at elevated current densities have been shown to improve water management. In some cases, flooding may be more prevalent at lower current densities, despite lower water production, as gas pressure drop is not sufficient to push water out. Back pressure refers to the elevated flow-field pressure attained by restricting outflow gas using controllable valves. In some PEMFCs, higher back pressures are used to improve electrochemical performance by increasing reactant concentration and or by controlling gas routing in the flow field [85,86]. Higher pressures mean greater gas density which increases the drag force on droplets. Some early work has shown that generating acoustic sound waves around 80 Hz in the reactant gases can enhance droplet detachment by inducing droplet vibration [87,88]. An overview of general PEMFC control schemes is available in Ref. [89]. Recently, the use of machine learning (ML) techniques has been explored for control, as well as to guide complex design decisions in PEMFC. ML uses various algorithms, such as artificial neural networks and decision trees, which may be trained using large data sets to develop an analytical model with minimal human intervention. These models can be updated automatically as new information becomes available. Early work has used ML for the state of health estimation, flow-field design, two-phase flow analysis, among others [90
92]. Two-phase flow prediction, such as anticipating the onset of flooding, would be beneficial to PEMFC operation as instability and degradation events could be avoided. Recent studies have demonstrated the ability to map images of ex situ PEMFC liquid water distributions with two-phase pressure drop [93
95]. Eventually, complex dynamics between liquid, pressure drop, and overvoltage may be used to train ML

Liquid water transport and management for fuel cells Chapter | 8

243

systems for in situ monitoring and control. Long short-term memory (LSTM) is a recurrent neural network architecture used in the field of deep learning [96]. Unlike standard feedforward neural networks, LSTM has feedback connections that allow it to process not only single data points but also entire sequences such as time-series data. LSTMs have recently been used to examine the use of time-series two-phase pressure data to predict channel water buildup [97]. Overall, ML implementation is in its early stages and may prove useful in developing future control and design schemes.

8.7

Thermal regulation and humidification

Heat is generated as a byproduct of PEMFC operation. Typically, steadystate PEMFC operation occurs between temperatures of 50 C and 80 C. Stacks usually require active cooling systems such as coolant channels between cells for thermal regulation. Reactant gas vapor pressure and evaporation rates are very sensitive to temperature. PEMFC systems commonly set the reactant gas humidity using dew point temperature control. Humid gases benefit membrane hydration; however, under humid conditions liquid may be prone to condense in diffusion media and channels. Dry gases have higher water uptake; however, membrane dehydration is a greater risk, especially in inlet regions and at higher flow rates. In a typical external humidification loop, water is evaporated into inlet gases from water condensed out of downstream gases. Some systems, such as the Toyota Mirai, rely on internal circulation (self-humidification) to humidify gases which greatly reduces system complexity and cost. This is accomplished by running the anode gas stream counter to the cathode gas stream, which humidifies upstream airflow [84]. Researchers have recently demonstrated evaporative cooling using specially modified GDLs, where water flows through dedicated anode flow-field channels, parallel to the gas channels, and is distributed over the cell area thanks to a modified GDL [98].

8.8

Startup/shutdown

Startup (or cold start when temperatures are subfreezing) refers to the transient process by which a dormant PEMFC is brought to its nominal operating conditions and power output. Usually, the goal is to achieve startup with minimal time and energy while avoiding cell degradation. During this period, cell or stack temperatures are below targets and must be ramped up to steady state. Water production is related to the current which, during startup, may be ramped up continuously or in steps [99]. Lower temperature gases have reduced capacity for water uptake and so water management during startup may be enhanced by supplying higher flow rates. Studies have shown improved startup performance at higher stoichiometry [100]. The startup

244

Fuel Cells for Transportation

routine also depends on the way the cell was shut down. Various purge protocols exist to prepare a PEMFC for dormancy. The amount of water allowed to remain in the MEA and flow channels may depend on the duty cycle and environmental conditions. For normal operation, minimal purging may be required to clear channels while the membrane is left humid for efficient restart. The purge duration and intensity for a PEMFC being prepared for subfreezing conditions may be greater to prevent residual water. Water expands as it freezes which can lead to delamination and cracking of membrane and diffusion media [101]. Degradation may be reduced via purging or by freeze prevention methods such as coolant loop antifreeze injection and thermal insulating. Cold start is a challenge as ice, water, and vapor may exist simultaneously in the MEA. Understanding the dynamics of this multiphase process is necessary as many PEMFC applications encounter extreme cold temperature environments. Freezing temperatures can also affect other PEMFC subsystems containing nozzles and valves where ice formation leads to blockages. Cold start can be separated into two categories, assisted and unassisted. Assisted cold start refers to the use of resistance heating (powered by batteries) or heating of coolant loop fluids to warm the cell above freezing [102]. While effective, these methods rely on other subsystems and can be energy-intensive. Unassisted cold start refers to the use of reaction waste heat to warm key cell components; therefore, the reaction takes place initially at cold temperatures. Cold startup protocols may rely on hybrid strategies using both assisted and unassisted methods. During unassisted cold start, water produced in the cCL condenses and begins to immediately fill the pore void space [103]. If gas temperature and flow rate do not reach conditions where water uptake keeps up with production, the cCL floods and power output ceases. Water is also removed via back-diffusion through the membrane, and studies have shown cold start performance is tied to initial membrane hydration [104]. While a drier membrane can improve cold start performance, the drying process can degrade the membrane; therefore, some level of residual water, even during freezing, may be beneficial. Cold start capability is also linked to MEA and flow-field design [100]. Extensive numerical simulation of cold start under varying conditions has shown that while low current densities during startup allow for higher pore space utilization, higher current densities produce waste heat at a higher rate per water production which may be beneficial to a cold start in certain scenarios [105,106]. Due to land/wall heat conduction, ice was observed to appear first under land areas. Cold start cycling of an improperly purged MEA can lead to significant performance decline as well as hydrogen starvation causing irreversible electrode degradation [107]. Overall, purging bulk water from channels and GDL while leaving some residual in the membrane and then using higher stoichiometries while initiating a current rampup are the basis for the successful unassisted startup.

Liquid water transport and management for fuel cells Chapter | 8

8.9

245

Surface coatings

Another strategy to influence liquid distribution is to alter the wetting properties of the channel walls and GDL through various hydrophobic and hydrophilic surface treatments. Droplet motion, characterized by contact angle hysteresis, is restrained by the pinned wetted perimeter of a droplet to a GDL and/or channel surface [19]. The pinning of a contact line results in a change in the shape of the liquid surface to accommodate pressure, gravitational, and shear forces without any bulk motion of the droplet. The difference between advancing and receding contact angles at the onset of motion is referred to as the maximum contact angle hysteresis [18]. These dynamics allow a droplet to distort on the surface of the GDL without moving. Hydrophobic coatings have been shown to lower the maximum hysteresis and cause water to “bead up” into smaller slugs which may be beneficial to purging processes and high current density operation. The influence of superhydrophobic coatings with static contact angled of .150 degrees on PEMFC two-phase pressure drop and liquid buildup is also being investigated [108]. Some drawbacks are that hydrophobic and superhydrophobic channel walls may cause water to build up within GDL under land areas and cause droplets to have larger cross sections normal to flow; therefore, these coatings should be applied strategically [109]. Hydrophilic and superhydrophilic channel surfaces can improve liquid uptake from the GDL and under certain conditions result in film flow [110]. This may benefit gas diffusion by keeping the GDL surface clear of water droplets which improves performance stability. The impact of long-term degradation on these coatings is not completely understood as thermal/mechanical stresses may reduce their effectiveness over time [111].

8.10 Ultrathin electrodes Ultrathin low PGM loading electrode architectures are sought after to achieve high power density operation at a reduced cost. A major hurdle to their adoption is sensitivity to liquid water as their thinness provides less void space to accommodate simultaneous water uptake and gas diffusion [112]. Two types, traditional Pt/C CL and nanostructured thin film (NSTF) CL, have received significant attention in recent years. Due to structural differences between the two, several strategies have been reported to mitigate flooding effects within them. For Pt/C designs, a thinner cathode MPL with higher air stoichiometry operation may enhance water removal [7]. In NSTF CL, flooding was minimized by moving water out via the anode during lowtemperature operation. This requires a thinner membrane or higher membrane permeability, removal of the anode MPL, and reduced anode pressure (i.e., Pa , Pc) [7]. Steinbach et al. performed systematic testing, including in situ neutron imaging, of NSTF layers highlighting that, surprisingly, the

246

Fuel Cells for Transportation

anode side GDL had a large impact on performance [49]. Further work improving GDL morphology for liquid-specific transport would benefit these scenarios where water back-diffusion is utilized to relieve cCL flooding. Another potential challenge to ultrathin architectures is increased gas crossover, especially oxygen moving from the cathode to the anode, and is the focus of some recent efforts.

8.11 Patterned and structured porous media Structured porous media refers to media having a prescribed rather than random structure. Recently, a theoretical analysis of structured GDLs of various lattice configurations concluded that significant performance improvements may be possible [48]. Several groups have achieved enhancements by modifying traditionally manufactured MPLs and GDLs [52,59,113]. A common goal among many efforts is to provide separate pathways for gas diffusion and liquid transport. For example, small perforations added to the MPL and GDL have been observed to improve liquid drainage leading to higher performance [51,53]. Thumbarathy et al. [113,114] recently demonstrated a spray coating technique where they used a mask with the desired pattern (stripes or polka dots) and a spray system to pattern GDL surfaces. They have designed GDLs with alternate stripes of two hydrophobic surfaces of different contact angles and with polka dots of a hydrophobic polymer of a higher contact angle than the base GDL. Their goal was to bias water pathways and provide favorable pathways to reactant gases to mitigate reactant starvation at high current densities. Fig. 8.8A highlights the alternate stripes of two hydrophobic surfaces by spraying hydrophobic polymer using a patterned mask (such as fluorinated ethylene propylene or a mixture of

FIGURE 8.8 (A) Additive manufacturing of alternate stripes of two hydrophobic surfaces of different contact angles [113] and (B) DIW-printed carbon aerogel electrode—simple cubic structured lattice [116]. Part (A) is adapted from Thumbarathy et al. J. Electrochem. Energy Conv. Storage 17 (1) (2020) 011010, with permission from ASME, and Part (B) is from Chandrasekaran et al. J. Mater. Res. 32 (22) (2017) 4166
4185.

Liquid water transport and management for fuel cells Chapter | 8

247

polydimethylsiloxane and fumed silica) on a conventional GDL surface [113]. As the areas with higher contact angles will be difficult to penetrate by water, it will provide favorable pathways for reactant gases. The results shown by Thumbarathy et al. indicate that one can achieve over a 20% increase in limiting current with modified GDLs as compared with the unmodified GDLs [113]. Groups have also demonstrated the inclusion of water pathways using radiation-grafted hydrophilic patterning of traditional GDL materials [115]. These approaches, while promising, are limited by the underlying GDL structure. Advanced manufacturing techniques, such as microscale resolution 3D printing, may offer promising future solutions to these difficult problems. Two common printing methods are direct ink writing (DIW) and stereolithography (SL). DIW printing utilizes a position-controlled nozzle to extrude carbon ink filaments used to construct 3D structures. Inks for these systems must be thixotropic; that is, they flow under applied shear stress while returning to a self-supporting solid when the stress is removed [116]. In SL printing, a UV light source is used to form 3D objects by curing a resin in a bath layer by layer. Projected light, in the shape of the object’s cross section at a specific position, is focused on each corresponding layer by adjusting the height of the resin which is contained on a movable platform. Once all the features are cured, the remaining liquid resin is drained from the bath leaving only the desired structure and/or support material. The process allows for intricate geometries and feature sizes ,1 μm. Lawrence Livermore National Laboratory has recently demonstrated the use of DIW 3D-printed carbon aerogels (CAs) in electrochemical systems such as electrolyzers and is currently probing PEMFC applications [117]. CAs are highly porous ( . 95%) solid materials known for their very low density, high surface area, high electrical conductivity, thermal and chemical robustness, ultrafine open pore structure, and good mechanical properties [116]. Fig. 8.8B shows a simple cubic DIW-printed CA structure; many other structures, including body-centered cubic lattices, face-centered cubic lattices, as well as gradient lattices and monoliths, are being investigated [116]. Pore gradient lattices could allow for liquid pathways utilizing differences in Laplace pressure to transport water away from flood-sensitive regions [118]. In these electrodes since the filaments are highly porous, aerogels are created using a supercritical drying or freeze-drying process; reactant gas can diffuse through them, unlike the solid fibers of conventional GDL.

8.12 Summary PEMFCs have the potential to play a significant role in helping industries such as heavy-duty trucking, shipping, rail, mobile power generation, and aerospace sectors reach low carbon emissions targets. To promote PEMFC adoption in these competitive markets, research and development must

248

Fuel Cells for Transportation

continue to seek solutions that reduce costs while increasing system lifetimes. Toward these pursuits, the management of liquid water should be considered a primary engineering concern. This chapter highlights liquid transport mechanisms, two-phase flow behavior, operation strategies, and novel materials and manufacturing methods, which will offer new opportunities and challenges in this area. Knowledge of basic two-phase flow physics in porous media and channels offers engineers a practical starting point for understanding current architectures and for designing next-generation PEMFC systems.

References [1] [2] [3] [4] [5] [6] [7]

[8] [9] [10] [11] [12] [13] [14] [15] [16]

[17] [18] [19] [20] [21] [22]

A.Z. Weber, J. Newman, J. Electrochem. Soc. 153 (2006) A2205. A.Z. Weber, et al., J. Electrochem. Soc. 161 (2014) F1254
F1299. G. Lin, W. He, T. Van Nguyen, J. Electrochem. Soc. 151 (2004) A1999. P.K. Das, X. Li, Z.-S. Liu, Int. J. Hydrog. Energy 35 (2010) 2403
2416. G.S. Hwang, M. Kaviany, J.H. Nam, M.H. Kim, S.Y. Son, J. Electrochem. Soc. 156 (2009) B1192. E.F. Medici, J.S. Allen, J. Power Sources 191 (2009) 417
427. P.K. Das, A.Z. Weber, in: ASME 2013 11th International Conference on Fuel Cell Science, Engineering and Technology (American Society of Mechanical Engineers), Minneapolis, MN, Paper No: FuelCell2013-18010, 2013, p. V001T01A002. https://doi. org/10.1115/FuelCell2013-18010. P.K. Das, A. Grippin, A. Kwong, A.Z. Weber, J. Electrochem. Soc. 159 (2012) B489
B496. Z. Lu, M.M. Daino, C. Rath, S.G. Kandlikar, Int. J. Hydrog. Energy 35 (2010) 4222
4233. J. Benziger, J. Nehlsen, D. Blackwell, T. Brennan, J. Itescu, J. Membr. Sci 261 (2005) 98
106. M. Mortazavi, K. Tajiri, Int. J. Hydrog. Energy 39 (2014) 9409
9419. A.D. Santamaria, P.K. Das, J.C. MacDonald, A.Z. Weber, J. Electrochem. Soc. 161 (2014) F1184
F1193. F.Y. Zhang, X.G. Yang, C.Y. Wang, J. Electrochem. Soc. 153 (2006) A225. K.A. Triplett, S.M. Ghiaasiaan, S.I. Abdel-Khalik, A. LeMouel, B.N. McCord, Int. J. Multiph. Flow 25 (1999) 395
410. K.A. Triplett, S.M. Ghiaasiaan, S.I. Abdel-Khalik, D.L. Sadowski, Int. J. Multiph. Flow 25 (1999) 377
394. C.H. Schillberg, S.G. Kandlikar, in: ASME 5th International Conference on Nanochannels, Microchannels, and Minichannels (ASMEDC), Puebla, Mexico, ,https:// asmedigitalcollection.asme.org/ICNMM/proceedings/ICNMM2007/4272X/299/318387., 2007, pp. 299
310. S.C. Cho, Y. Wang, K.S. Chen, J. Power Sources 206 (2012) 119
128. E.C. Kumbur, K.V. Sharp, M.M. Mench, J. Power Sources 161 (2006) 333
345. A. Theodorakakos, et al., J. Colloid Interface Sci 300 (2006) 673
687. P.K. Das, A.D. Santamaria, A.Z. Weber, Procedia Eng. 105 (2015) 751
756. Available from: https://doi.org/10.1016/j.proeng.2015.05.066. A.D. Santamaria, J. Haydu, D. Senft, J. Riofrio, IMECE2015 (Volume 6B: Energy, 2015; https://doi.org/10.1115/IMECE2015-50651). A.D. Santamaria, M. Mortazavi, J. Electrochem. Soc. 167 (2020) 104507.

Liquid water transport and management for fuel cells Chapter | 8

249

[23] J.T. Sellman, A.D. Santamaria, Int. J. Hydrog. Energy 42 (2017) 12551
12558. [24] D. Spernjak, A.K. Prasad, S.G. Advani, J. Power Sources 170 (2007) 334
344. [25] I.V. Zenyuk, D.Y. Parkinson, G. Hwang, A.Z. Weber, Electrochem. Commun 53 (2015) 24
28. [26] N.J. Cooper, A.D. Santamaria, M.K. Becton, J.W. Park, Int. J. Hydrog. Energy 42 (2017) 16269
16278. [27] A.D. Santamaria, M.K. Becton, N.J. Cooper, A.Z. Weber, J.W. Park, J. Power Sources 293 (2015) 162
169. [28] H.-Y. Tang, et al., J. Power Sources 195 (2010) 6774
6781. [29] X.-G. Yang, N. Burke, C.-Y. Wang, K. Tajiri, K. Shinohara, J. Electrochem. Soc. 152 (2005) A759. [30] Y. Shao, et al., Energy Convers. Manag 185 (2019) 169
182. [31] I.S. Hussaini, C.-Y. Wang, J. Power Sources 187 (2009) 444
451. [32] S.A. Niknam, M. Mortazavi, A.D. Santamaria, Results Eng. 5 (2020) 100071. [33] M. Mortazavi, K. Tajiri, Renew. Sustain. Energy Rev 45 (2015) 296
317. [34] R. Banerjee, S.G. Kandlikar, J. Power Sources 247 (2014) 9
19. [35] R. Banerjee, S.G. Kandlikar, J. Power Sources 268 (2014) 194
203. [36] M. Grimm, E.J. See, S.G. Kandlikar, Int. J. Hydrog. Energy 37 (2012) 12489
12503. [37] M. Mortazavi, et al., J. Electrochem. Energy Convers. Storage 18 (2021) 031013. Available from: https://doi.org/10.1115/1.4049572. [38] R.W. Lockhart, R.C. Martinelli, Chem. Eng. 45 (1949) 39
48. [39] D. Chisholm, Int. J. Heat. Mass. Transf 10 (1967) 1767
1778. [40] S. Saisorn, S. Wongwises, Exp. Therm. Fluid Sci 34 (2010) 454
462. [41] M. Mortazavi, M. Heidari, S.A. Niknam, Heat Transf. Eng. 41 (2020) 1784
1799. [42] J. Benner, M. Mortazavi, A.D. Santamaria, in: IMECE2018 (Volume 6A: Energy), 2018, ,https://doi.org/10.1115/IMECE2018-86579.. [43] I.V. Zenyuk, E. Medici, J. Allen, A.Z. Weber, Int. J. Hydrog. Energy 40 (2015) 16831
16845. [44] N.J. Cooper, A.D. Santamaria, M.K. Becton, J.W. Park, Energy Convers. Manag 136 (2017) 307
317. [45] A.D. Santamaria, N.J. Cooper, M.K. Becton, J.W. Park, Int. J. Hydrog. Energy 38 (2013) 16253
16263. [46] A.D. Santamaria, J. Bachman, J.W. Park, Int. J. Hydrog. Energy 38 (2013) 5807
5812. [47] J. Larminie, A. Dicks, M.S. McDonald, Fuel Cell Systems Explained, Wiley, 2003, vol. 2. [48] D. Niblett, V. Niasar, S. Holmes, J. Electrochem. Soc. 167 (2020) 013520. [49] A.J. Steinbach, et al., Joule 2 (2018) 1297
1312. [50] J. Becker, C. Wieser, S. Fell, K. Steiner, Int. J. Heat. Mass. Transf 54 (2011) 1360
1368. [51] A. Mohseninia, et al., ChemSusChem. 13 (2020) 2931
2934. [52] Y. Nagai, et al., J. Power Sources 435 (2019) 226809. [53] D. Gerteisen, T. Heilmann, C. Ziegler, J. Power Sources 177 (2008) 348
354. [54] A. Lamibrac, et al., J. Electrochem. Soc. 163 (2016) F202
F209. [55] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, J. Power Sources 173 (2007) 277
290. [56] K.-J. Lee, J.H. Nam, C.-J. Kim, J. Power Sources 195 (2010) 130
141. [57] P.K. Sinha, C.-Y. Wang, Electrochim. Acta 52 (2007) 7936
7945. [58] P. Chippar, O. Kyeongmin, K. Kang, H. Ju, Int. J. Hydrog. Energy 37 (2012) 6326
6338. [59] M. Balakrishnan, et al., ACS Appl. Energy Mater 3 (2020) 2695
2707.

250

Fuel Cells for Transportation

[60] N.J. Cooper, T.L. Smith, A.D. Santamaria, J. Wan Park, J. Electrochem. Energy Convers. Storage 15 (2018) 041006. [61] N.J. Cooper, T. Smith, A.D. Santamaria, J.W. Park, Int. J. Hydrog. Energy 41 (2016) 1213
1223. [62] J. Scholta, et al., J. Power Sources 155 (2006) 66
71. [63] T.L. Smith, A.D. Santamaria, J.W. Park, K. Yamazaki, Procedia CIRP 14 (2014) 275
280. [64] S.-W. Cha, R. O’Hayre, S.J. Lee, Y. Saito, F.B. Prinz, J. Electrochem. Soc. 151 (2004) A1856. [65] S.W. Cha, R. O’Hayre, Y. Saito, F.B. Prinz, J. Power Sources 134 (2004) 57
71. [66] J. Park, X. Li, J. Power Sources 163 (2007) 853
863. [67] N. Akhtar, et al., Int. J. Hydrog. Energy 34 (2009) 3104
3111. [68] J. Bachman, et al., Int. J. Hydrog. Energy 37 (2012) 17172
17179. [69] Y. Sando, ECS Trans 25 (2009) 211
224. [70] N. Guo, et al., Int. J. Hydrog. Energy 39 (2014) 21185
21195. [71] A. Iranzo, C.H. Arredondo, A.M. Kannan, F. Rosa, Energy 190 (2020) 116435. [72] N. Kizilova, M. Sauermoser, S. Kjelstrup, B.G. Pollet, Entropy 22 (2020) 176. [73] A. Fly, et al., Energies 12 (2019) 1186. [74] C.-J. Tseng, et al., Energy Convers. Manag 62 (2012) 14
21. [75] A. Santamaria, J. Zhang, in: FEDSM2017 (Volume 1B, Symposia: Fluid Measurement and Instrumentation; Fluid Dynamics of Wind Energy; Renewable and Sustainable Energy Conversion; Energy and Process Engineering; Microfluidics and Nanofluidics; Development and Applications in Computational Fluid Dynamics; DNS/LES and Hybrid RANS/LES Methods), 2017, https://doi.org/10.1115/FEDSM2017-69358. [76] M. Kim, C. Kim, Y. Sohn, Fuel Cell 18 (2018) 123
128. [77] J.E. Park, et al., Int. J. Hydrog. Energy 44 (2019) 22074
22084. [78] Z. Bao, Y. Wang, K. Jiao, J. Power Sources 480 (2020) 229150. [79] T. Yoshida, K. Kojima, Interface Mag 24 (2015) 45
49. [80] S.G. Kandlikar, W.J. Grande, Heat. Transf. Eng 24 (2003) 3
17. [81] P. Concus, R. Finn, PNAS 63 (1969) 292
299. [82] C.D. Rath, S.G. Kandlikar, Colloids Surf. A: Physicochem. Eng. Asp 384 (2011) 653
660. [83] S.H. Han, N.H. Choi, Y.D. Choi, Renew. Energy 44 (2012) 88
98. [84] H. Lohse-Busch, K. Stutenberg, M. Duoba, S. Iliev, Technology assessment of a fuel cell vehicle: 2017 Toyota Mirai, United States, 2018, doi:10.2172/1463251. [85] J. Bachman, A. Santamaria, H.-Y. Tang, J.W. Park, J. Power Sources 198 (2012) 143
148. [86] S. Tong, J.C. Bachman, A. Santamaria, J.W. Park, J. Power Sources 242 (2013) 195
201. [87] M. Mortazavi, A.D. Santamaria, J.Z. Benner, V. Chauhan, J. Electrochem. Soc. 166 (2019) F3143
F3153. [88] A.M. Schafer, J.S. Allen, ECS Trans 41 (2011) 1887
1896. [89] W.R.W. Daud, et al., Renew. Energy 113 (2017) 620
638. [90] V. Chauhan, M. Mortazavi, J.Z. Benner, A.D. Santamaria, Energy Rep 6 (2020) 2713
2719. [91] S. Laribi, K. Mammar, Y. Sahli, K. Koussa, Sustain. Energy Technol. Assess 34 (2019) 35
42. [92] H. Pourrahmani, M. Moghimi, M. Siavashi, M. Shirbani, Appl. Therm. Eng 150 (2019) 433
444. [93] A.D. Santamaria, M. Mortazavi, V. Chauhan, J. Benner, O. Philbrick, H. Jia, C. Ling, J. Electrochem. Soc. 168 (2021) 054505. [94] A.D. Santamaria, et al., ECS Trans 98 (2020) 279.

Liquid water transport and management for fuel cells Chapter | 8

251

[95] A.D. Santamaria, M. Mortazavi, V. Chauhan, J. Benner, ECS Meeting Abstracts, IOP Publishing, 2019, p. 1538. [96] S. Hochreiter, J. Schmidhuber, Neural Comput 9 (1997) 1735
1780. [97] J. Benner, M. Mortazavi, V. Chauhan, O.K. Philbrick, A.D. Santamaria, in: ASTFE Proceedings, Las Vegas, NV, 2022. [98] M. Cochet, et al., Fuel Cell 18 (2018) 619
626. [99] K. Tajiri, Y. Tabuchi, C.-Y. Wang, J. Electrochem. Soc. 154 (2006) B147. [100] A.D. Santamaria, J. Bachman, J.W. Park, Electrochim. Acta 107 (2013) 327
338. [101] X.G. Yang, Y. Tabuchi, F. Kagami, C.-Y. Wang, J. Electrochem. Soc 155 (2008) B752. [102] Q. Guo, Y. Luo, K. Jiao, Int. J. Hydrog. Energy 38 (2013) 1004
1015. [103] L. Mao, C.-Y. Wang, J. Electrochem. Soc. 154 (2007) B139. [104] J. Ko, H. Ju, Appl. Energy 94 (2012) 364
374. [105] K. Jiao, X. Li, Int. J. Hydrog. Energy 35 (2010) 5077
5094. [106] K. Jiao, X. Li, Electrochim. Acta 54 (2009) 6876
6891. [107] H. Chen, X. Zhao, T. Zhang, P. Pei, Energy Convers. Manag 182 (2019) 282
298. [108] A.D. Santamaria, et al., ECS Trans 104 (2021) 221. [109] A. Turhan, S. Kim, M. Hatzell, M.M. Mench, Electrochim. Acta 55 (2010) 2734
2745. [110] H.-Y. Tang, A. Santamaria, J.W. Park, C. Lee, W. Hwang, J. Power Sources 196 (2011) 9373
9381. [111] F.A. De Bruijn, V.A.T. Dam, G.J.M. Janssen, Fuel Cell 8 (2008) 3
22. [112] I.V. Zenyuk, P.K. Das, A.Z. Weber, J. Electrochem. Soc. 163 (2016) F691. [113] D. Thumbarathy, G. Gupta, M. Mamlouk, P.K. Das, J. Electrochem. Energy Convers. Storage 17 (2020) 011010. Available from: https://doi.org/10.1115/1.4044814. [114] D. Thumbarathy, P.K. Das, Electrochemical Conference on Energy and the Environment: Bioelectrochemistry and Energy Storage, Electrochemical Society, Glasgow, 2019, ,https://doi.org/10.1149/MA2019-04/3/169.. [115] A. Forner-Cuenca, et al., Adv. Mater 27 (2015) 6317
6322. [116] S. Chandrasekaran, P.G. Campbell, T.F. Baumann, M.A. Worsley, J. Mater. Res. 32 (2017) 4166
4185. [117] V.A. Beck, A.N. Ivanovskaya, S. Chandrasekaran, J.-B. Forien, S.E. Baker, E.B. Duoss, M.A. Worsley, PNAS 118 (2021). e2025562118. [118] A.D. Santamaria, S. Chandrasekaran, O.K. Philbrick, M. Worsley, ECS Transactions, The Electrochemical Society, Vancouver, BC, 2022.

This page intentionally left blank

Chapter 9

Fuel cell short stack testing Graham Smith1 and Katie McCay2 1

National Physical Laboratory, Teddington, United Kingdom, 2SINTEF Industry, Sustainable Energy Technology, Trondheim, Norway

9.1

Introduction

Fuel cell systems using low-temperature proton exchange membrane fuel cells (PEMFCs) fed with “clean” (low carbon) hydrogen are being deployed to power a range of low-emission vehicles from drones and small passenger vehicles to light aircraft and ships. The power requirements for vehicles vary widely, with systems of B100 kW net typically required for road-worthy passenger vehicles while systems at MW scale are required to power even modest-sized ships and aircraft. The need to transport reactant air and product water limits the practical size of individual fuel cells to only a few hundred cm2. With 2030 power targets of 2 W cm22 at 0.66 V (representing B50% fuel efficiency) [1], many individual fuel cells are required to power most transportation systems. Conveniently, PEMFCs are thin and planar and can thus be stacked in electrical series to produce “stacks” able to produce useful amounts of power. Stacks of hundreds of cells are typical, and powers in the range of 100
250 kW per stack are common, with multiple stacks used where required. Such stacks therefore represent how fuel cells are operated in the real world. Academic fuel cell testing, particularly for material development, is often performed in single cells 5
50 cm2 in size, but testing at the scale this does not always adequately reproduce all the phenomena occurring in full-sized cells and stacks. There is a regular need to perform testing on devices highly representative of real-world use including testing to evaluate the performance and lifetime of new materials under real conditions, to validate new flow field designs, to optimize operating conditions and balance of plant design, to design control algorithms, and to assess the impact of different operating conditions and a range of other activities. The production and routine testing of full-sized stacks are both challenging and costly. Many advanced diagnostic techniques are not readily applicable to systems of such size, and precise control over operating parameters is also an issue. Testing is therefore routinely Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00010-1 © 2023 Elsevier Ltd. All rights reserved.

253

254

Fuel Cells for Transportation

FIGURE 9.1 A short stack is constructed using the same cells, manifolding, and endplates as used in a full stack but with a reduced number of cells. The components of a single cell: membrane, catalyst layers, gas diffusion and microporous layers, gasketing, and bipolar plates are shown schematically in relationship to the overall stack.

carried out on “short stacks,” built using identical cells to those used in fullsize stacks but with typically only 5
20 cells per stack, as shown in Fig. 9.1. Though short stack testing is of importance for the continued development and deployment of fuel cells for transportation applications, there is little information in the open literature on how to perform testing at this scale [2]. This chapter aims to provide a practical guide to characterizing fuel cell short stacks in a laboratory environment with dedicated test equipment. It will outline the principles, facilities, considerations, operating conditions, and available diagnostic techniques. Though the focus is on PEMFC shortstack testing, readers will also find much that is directly applicable to smaller single cell testing.

9.2

Principles of fuel cell operation and testing

This section will briefly introduce the principles of PEMFC operation with a focus on areas where testing differs between single cell and short stack testing. Full details of the operating principle, reaction kinetics, and transport

Fuel cell short stack testing Chapter | 9

255

processes involved in a PEMFC are described in Chapter 2, Fuel cell fundamentals. A key outcome of all types of PEMFC testing is usually the relationship between voltage and current, colloquially referred to as performance, and how this evolves over time as the stack is used, often called durability. Testing of short stacks is normally performed galvanostatically, with a load drawing a fixed current while the stack and cell voltages are the measurands. The stack voltage can be considered to represent the thermodynamic efficiency of the cell, that is, the ratio of energy in the fuel converted into electricity versus that turned into heat, Eq. (9.1). Similarly, the current represents the rate at which reactions are occurring in the cell, while the product of current and voltage is the stack electrical power, Eq. (9.2). Efficiency 

Measured stack voltage ½V 1:23 ½V 3 Number of cells

PowerElectrical ½W 5 Stackcurrent ½A 3 Voltage ½V

ð9:1Þ ð9:2Þ

For any PEMFC to function, it must be supplied with conditioned air and hydrogen and have a load to draw power and a sink for the heat generated. Typical flows through a stack are illustrated in Fig. 9.2. The properties of the gas feeds and the temperature of the stack all strongly influence both performance and durability. To carry out any meaningful measurements, these properties must be defined with a set of operating conditions, at which measurements on the stack are performed. These operating conditions usually aim to approximate those experienced during use. Philosophically, small single cells are homogeneous, and operating conditions may be chosen to simulate the conditions experienced across the area of a full-size cell, for example, with dryer colder conditions replicating those at the inlet and hotter wetter ones mimicking those at the outlet, or to simulate conditions experienced in real systems [3]. In contrast, short stacks innately experience this spatial inhomogeneity, and instead, the scope of operating conditions is chosen solely to reflect the conditions experienced in real systems. In both cases, it is typical for measurements to be performed at a range of conditions to provide a more robust data set. There are a large number of published testing conditions, test methods, and accelerated stress tests defined for single cells and individual components, including those intended to be standards [3
6]. There are, however, fewer standardized testing protocols that are widely adopted for short-stack testing, though there is a range of standardized vehicle drive cycles [7,8] and systems level tests that may be adapted to the short stack level [9], some detailed public test procedures [10,11] and standards relating to safety that contain tests that may be applicable [12]. A typical set of parameters required to specify one set of operating conditions for a fuel cell is given in Table 9.1. This is adapted from the test conditions used for experiments in the European Union Fuel Cells & Hydrogen Joint Undertaking HYDRAITE

256

Fuel Cells for Transportation

FIGURE 9.2 Schematic of a short stack showing flows into and through the stack.

project [13], but nominal maximum and minimum parameters would usually be defined by the hardware and component supplier. To those experienced with single cell testing, there will be a number of parameters that differ in the way they are defined. In contrast to small single cells, short stacks generate an appreciable amount of heat, and the stack is non-isothermal at the macro-scale both across the plane of the cells (coolant inlet to outlet) and also frequently with differences between cells. In watercooled short stacks, the main temperature specification is the inlet temperature

Fuel cell short stack testing Chapter | 9

257

TABLE 9.1 Typical operating conditions for PEMFC short stack.

Stack

Anode

Parameters

Unit

Property

Coolant inlet temperature



C

70

Coolant flow rate



C

To achieve ΔT , 10

Maximum current

A cm22

1.2

Fuel gas inlet temperature



C

80

Fuel gas inlet dew point at max currenta



C

64

Fuel gas inlet pressure

kPaa

130

Fuel gas composition

n/a

Certified to ISO 14687:2019: Part B

Fuel stoichiometry recirculation

n/a

1.01

Fuel recirculation stoichiometry

1.5

Fuel stoichiometry open-end Cathode

1.3

Oxidant inlet temperature



C

80

Oxidant gas inlet humidity



C

64

Oxidant gas outlet pressure

kPaa

120

Oxidant

n/a

Grade 2.1.1 according to ISO 8573
1:2010

Cathode stoichiometryb

n/a

2.0

Minimum current density for stoichiometry

A cm22

0.3

a

Property impacted by running with recirculation. Often specified in NLPM, N being normal—values referenced to IUPAC standard conditions 273.15K and 100 kPa.

b

of the coolant, as it is supplied at the stack inlet and the flow rate of the coolant. Flow rates can be held constant or else varied to achieve a maximum target difference between coolant temperature at the inlet and outlet. The latter option would therefore increase the flow rate with a higher current to remove the extra heat being generated. In real systems, this has the benefit of reducing the energy required for pumping water unnecessarily. For gasses supplied, the parameters of interest are the flow rate, temperature, humidity, and pressure. For the cathode side, parameters are defined in the same way as they are for single cells. Flow rates are usually defined in

258

Fuel Cells for Transportation

terms of stoichiometry—the ratio of gas supplied to that theoretically required to maintain a given current as predicted by Faraday’s law. For example, when running with a stoichiometry of 1 all reactant is consumed in one pass through the cell, while with a stoichiometry of 2 only 50% of the reactant gas is consumed. A minimum stoichiometry is usually specified to ensure a flow when the current is very low or zero. Pressures of a few bar gauge are typical; increased pressures on the cathode typically improve performance by increasing the partial pressure of reactants at the catalyst layer and enhancing water vapor removal. In systems, both the cathode stoichiometry and pressure are practically limited by the parasitic energy demand for supplying compressed air so that the range is similar to that used for single-cell testing. The humidity of the cathode is a delicate balancing act in all PEMFCs, but due to the variety of conditions experienced by real stacks in operation a range of values are normally studied. Cathode inlet temperatures are normally simply controlled to several degrees above the stack coolant inlet temperature to prevent condensation at the inlet of the stack. On the anode side, the situation is more complex. Single cells typically operate in an open-end mode where the anode operates much like the cathode, operating at a fixed stoichiometry. Short stacks may operate like this too, but when considering their use in systems, it is wasteful of hydrogen. Short stacks therefore rarely operate with anode stoichiometries c1.0; instead, the anode operates in dead-end mode [14], as highlighted in Fig. 9.3. In this configuration, the exhaust is closed, and hydrogen is supplied to replace that is consumed by the hydrogen oxidation reaction. In practice, this is usually implemented as forward pressure control where gas is supplied to the anode to maintain a target pressure. In this configuration, water and nitrogen will gradually build up at the anode as they cross over from the cathode and must be removed to prevent flooding and dilution of hydrogen. Typically, a valve downstream of the

FIGURE 9.3 Schematic of a dead end, purged, anode recirculation system.

Fuel cell short stack testing Chapter | 9

259

stack anode exhaust with a high flow rate is actuated quickly to purge water droplets from the cell and create a net gas flow through the stack which limits nitrogen accumulation. The length and frequency of the purge can be adjusted to balance water removal with the hydrogen wasted. A hydrogen utilization is often defined for this configuration—the percentage of the hydrogen supplied by the stack consumed producing electricity. Automotive systems typically target hydrogen utilizations . 99%. On its own, a purge is often insufficient to prevent variations in hydrogen concentration and water accumulation in the anode. To improve homogeneity, recirculation is usually also implemented. Here, a pump, impinging jet, or similar is implemented between the anode outlet and inlet; the hydrogen flow rate through the stack is higher than that required to sustain the anode reaction as in the open-end case, but the excess hydrogen from the outlet is sent back to the inlet, so it is not wasted. A separator is implemented to remove some fraction of water from the recirculated hydrogen so that the anode inlet is composed of dry hydrogen from a tank and humidified hydrogen from recirculation. Even in a laboratory, it is challenging to measure the flow rate of a wet gas, especially when its composition and humidity vary. A convenient way to specify or control the recirculation rate is via the humidity at the anode inlet; if fresh gas is assumed to be dry and recirculated assuming to be at 100% relative humidity, then the inlet humidity can be converted into a pseudo-flow rate. As with cathodes, inlet temperatures for the anode are typically set above coolant inlet temperatures to prevent condensation; when operating with recirculation, this usually necessitates heating of the recirculation loop too, which may in turn influence the relative humidity of the recirculated gas.

9.3

Testing requirements

The above section outlines the conditions to which short stacks must be subjected in order to be as representative as possible of real operating systems. This section addresses some practical considerations for performing such testing in a controlled manner in a laboratory. The first consideration is safety [12]. Hydrogen is a flammable gas which burns in the air with an invisible flame. In mixtures with air, hydrogen concentrations above the lower flammability limit (B4%) form combustible or explosive mixtures that may deflagrate or detonate depending on conditions. Even though the autoignition temperature of hydrogen is .500 C, these explosive mixtures are very easy to ignite with ignition energy of only 0.02 mJ sufficient. This is lower than a weak electrostatic discharge (B10 mJ) which can be produced from a gas flow or a person. The priority therefore is usually to prevent hydrogen leaks and avoid the formation of hydrogen concentrations above the lower flammability limit if they can occur. The flow rates involved in short-stack testing are typically two

260

Fuel Cells for Transportation

orders of magnitude higher than those used for single-cell testing so additional caution must be exercised. Regulations vary, and risk assessments must be performed for specific cases, but three mitigation measures are usually the minimum implemented. First, leak checking of connections and the fuel cell before use are critical, with an outline procedure for checking leaks from a stack discussed below. Second, hydrogen alarms are implemented to detect any leaks that do occur from the stack or hydrogen supply infrastructure; these alarms are often interlocked with the hydrogen supply (electrolyzer or cylinder shutoff) to prevent significant concentrations from building up. Third, testing is normally carried out in environments with forced ventilation and extraction of the exhaust; the large volumes of air passing through these systems rapidly dilute any leaked or exhausted hydrogen to well below the flammability limits. Beyond hydrogen, typical hazards include high temperatures and the use of pressurized gasses. The electrical power from short stacks must also be considered as it is in the order of kW (vs W in small single cells). Each cell in a fuel cell contributes a maximum B 1 V to the stack; for a grounded short stack of fewer than 20 cells, the maximum voltage under any fault condition (not including another power supply) is generally less than 20 V and therefore typically below the extra low-voltage threshold. Electricity below this threshold is usually considered to present a low risk of electrocution though it may be harmful under certain conditions. Full-size cells can easily produce currents over 500 A during normal operation and higher if shortcircuited. Any external electrical shorts are liable to experience very high temperatures, and high-resistance electrical connections may melt or burn if not sized sufficiently. Laboratory testing is usually carried out using a test station. These systems are usually automated and at a minimum provide control over gas temperature, humidity, pressure and flow rates, coolant temperature and flow rate, and the current, while recording operating conditions and cell/stack voltages. There are several manufacturers who sell test stations commercially, including Greenlight Innovation (Canada), Horiba FuelCon (Germany), and Scribner Associates (USA), and many institutes and companies carrying out testing produce their own bespoke systems. Ancillary equipment such as potentiostats, frequency response analyzers, and gas analysis instruments are also regularly used to perform diagnostic measurements on short stacks and may be integrated with the test stations. Test stations are complex, and a full discussion of their design or set of specifications is beyond the scope of this book; however, there are several practical issues relating to their use which are of particular interest when performing high-quality short-stack testing. In short stacks, voltage logging is typically carried out both across the entire stack but also at an individual cell level. Differential cell voltage meters are typically connected to the lowest potential anode, every bipolar plate, and the highest potential cathode so that they log every cell voltage.

Fuel cell short stack testing Chapter | 9

261

As with single cells, inlet and recirculation connections must also be adequately insulated and temperature-controlled to avoid cold spots that may cause condensation of water; the introduction of water droplets frequently leads to water management issues and corresponding voltage fluctuations. Gas connections to stack hardware are often more complicated than the simple tubing connectors found on single-cell hardware. One priority is to ensure that all wetted materials in the connection will not introduce impurities. The best materials to use for connections are therefore stainless steel and high-quality perfluorocarbons such as polytetrafluoroethylene or other fluoro-elastomers. Other plastics, especially silicones and sulfur-cured polymers, should be avoided as they may off-gas or leach impurities. The quality of the fluids supplied to the test system is of high importance, as the performance and durability of PEMFCs are susceptible to even trace levels of impurities. The use of recirculation systems may exacerbate this by enriching certain impurities that break through the stack. Common gas specifications such as 5.0 N (denoting 99.999% pure gas) are insufficient to determine if harmful impurities are present. For the anode side, the detailed gas specification must be interrogated to determine if the hydrogen meets the ISO 14687:2019-D [15] purity requirements of hydrogen for automotive use. This standard sets threshold limits for a range of impurities and is widely used when assessing hydrogen purity supplied by real-world infrastructure. On the cathode side, there is far less standardization around air quality, primarily because it is challenging to control during real-world use. At the least, the air should be filtered for particles and oil, typically as class 1:-:1 in ISO 8573
1:2010 [16]. Note that this, and other standards, does not specify concentration limits for potentially damaging chemical species such as NOx and SOx. Water may be supplied to the humidification system of the test station; this should also be free of impurities, so ASTM Type I water [17] or similar is preferred. Note that even water meeting this strict standard may have an impact on some experiments; for instance, dissolved oxygen in the water may be supplied at an appreciable concentration to the anode. Water or another coolant is usually also recirculated through the stack and topped up intermittently; this coolant must be low conductivity to avoid the risk of short-circuiting between cells so specific low conductivity coolants or Type I water is again preferred. As the results of testing are strongly dependent upon the operating conditions, the test station’s control over these parameters must be good. It is often necessary when testing a new stack to tune the controllers to ensure that they provide a stable operating condition and respond rapidly to changes in stack conditions or currents. When operating the stack dynamically, with large changes in current, it is common that the operating conditions experienced by the stack will deviate significantly from the nominal ones specified for testing. This is likely to lead to hysteresis or aging phenomena that are poorly reproducible. Regular logging of the actual operating conditions is therefore essential; it is common to log all relevant parameters at 10 Hz.

262

Fuel Cells for Transportation

A key issue when operating dynamically is the change in the flow rate of gasses. Gas flow is often controlled by digital mass flow controllers; these regularly take between 1 and 5 seconds to respond to a step change in flow, which may lead to starvation if the flow and current are not modified in unison. When operating the anode in forward pressure control mode, the flow rate to the anode is usually able to respond to step changes rapidly, though this depends on the hardware. Evaluation of uncertainty is an often ignored but essential part of highquality stack testing. With the typical uncertainty on cell voltage monitoring hardware, only B 6 1 mV (k 5 1) drift and fluctuations in the cell voltage and deviations from the nominal operating condition are the predominant sources of uncertainty. The random fluctuations around the setpoint value caused by imperfect control and systematic variations that occur during dynamic cycles also contribute to uncertainty. Even with perfect control, uncertainties on the measurements of operating conditions, such as temperature, are sometimes surprisingly high ( . 6 1K (k 5 1) is common for even trivial temperature measurements with thermocouples). Ideally, all measurement equipment should be calibrated regularly to traceable SI standards. When reporting results, it is important that the range of operating conditions experienced and the uncertainty of the measurement of each operating condition are noted. Seemingly, trivial issues about equipment layout may also contribute to measurement uncertainty and hinder reproducibility between testing organizations. One example of this is the location of measurement devices, such as thermocouples, relative to the stack. These are usually defined as “at the inlet,” but this is a loose definition, and in practice, a location difference of centimeters is possible, a significant enough distance that measurable pressure and temperature changes are experienced.

9.4

Measurement techniques

Fuel cell characterization usually involves basic operation, measurement of the device performance to determine power and efficiency, diagnostic measurements to determine the underlying performance limitations or a specific property of interest, and aging to understand how these evolve with use. Fuel cell performance measurements and diagnostics typically take a few hours, while aging is often performed for hundreds to thousands of hours with performance and diagnostic testing performed regularly throughout. This section will briefly outline some key principles of operation, common performance, and diagnostic techniques and discuss aging measurements.

9.4.1

Characterization of test station

To ensure reproducible and high-quality results from fuel cell testing, it is crucial to have a well-characterized test station. At the most basic level, this

Fuel cell short stack testing Chapter | 9

263

usually involves ensuring that all measurement systems are calibrated regularly and that the uncertainty on measurements and control systems is known. More complex measurements specific to short stacks, such as online gas analysis, often also require specific properties related to the recirculation system such as recirculation rate, recirculation loop volume, and hydrogen utilization to be measured.

9.4.1.1 Recirculation loop volume The volume of the recirculation loop must be known when computing concentrations to absolute units, for example, from a measured nitrogen concentration to a rate of nitrogen flux from cathode to anode. The volume can be calculated analytically by summing up the volume of each part, though it is more easily accessed by a simple experiment using the ideal gas law. Simply, the outlet of the recirculation loop is blocked, and dry purge gas is supplied at a known flow rate (nitrogen at 1 NLPM is convenient for most systems) while the pressure and temperature are monitored. Application of the ideal gas law yields the normal volume. This has the benefit of including the volume of the stack’s anode in the calculation. 9.4.1.2 Recirculation rate The rate of recirculation is challenging to measure directly as the composition of the anode loop varies over time from almost pure hydrogen to a mixture with nitrogen and is also highly humidified. The high degree of humidification prevents the use of most commercial mass flow meters, which are usually water intolerant. There are several methods to overcome this measurement challenge; the easiest is to measure the pressure drop over the anode of the stack, with this approximately, linearly proportional to the flow rate over the narrow range of rates used. If an absolute flow rate is required, a calibration curve may be prepared to back-calculate the flow rate. This method does not readily consider the impact of composition changes, humidity, or water accumulation in the stack without measurements in the recirculation loop to assess these. Alternatively, if one is used, the recirculation pump may be characterized under relevant conditions offline, and an expected flow rate is extracted; again, a calibration curve and additional measurements to account for humidity and composition may be required. A final method to estimate the flow rate is to measure the humidity at the anode inlet and outlet and, with knowledge of the dry hydrogen supply rate, perform a mass balance on the water [18]. 9.4.1.3 Hydrogen utilization As discussed above, hydrogen utilization is an important parameter for understanding the fuel efficiency of the system, and the effective utilization in a test system will impact water removal and anode gas composition.

264

Fuel Cells for Transportation

Usually, any small losses from hydrogen crossover in the stack (measurements of which are discussed below) or hydrogen consumption due to crossover oxygen are ignored, and a study of utilization focuses instead on measuring the hydrogen lost during the intermittent purge as well as any from analytical equipment, such as mass spectrometers, in use on the test bench which may consume gas. The gas consumption from analytical equipment is usually well defined or easily measured, with, for example, a rotameter after drying the sample gas. It is much less easy to measure the purge as this is typically performed at a very high flow rate for , 1 seconds; such dynamic flows do not lend themselves to direct measurements with mass flow meters, while introducing volumetric measurement techniques would necessarily influence the purge flow and hence the result. Different techniques for measuring utilization have been compared [19]. The most robust method is to monitor impurity enrichment (and therefore utilization) directly by introducing an inert impurity such as methane into the anode at a known concentration and measuring its concentration in the recirculation loop. As this impurity does not react, it is assumed to be lost only through anode purges so that the ratio of the measured concentration versus the inlet concentration provides the enrichment factor from which the effective purge rate and utilization may be calculated.

9.4.2

Stack operation

Basic operation of the stack is usually required to perform any meaningful testing. The haphazard application of gasses, coolant, temperature, pressure, or load is likely to cause damage to the stack or accelerate degradation. It is therefore important that well-established procedures are followed. The basic principles these procedures follow to prevent rapid degradation or damage to the stack are: G

G

G

To avoid anode starvation: if there is insufficient hydrogen to sustain a current, then the local potential at the anode increases causing water splitting or catalyst support corrosion which can be very rapidly damaging. If gross anode starvation occurs to a cell, then it may “reverse;” that is, the cell voltage becomes negative as these reactions occur to sustain the load. To prevent this case, it is common to automatically disconnect the load when a minimum cell voltage lower than a threshold value, typically 0.3 V, is detected for more than 100 ms. To prevent the formation of hydrogen
air boundaries during start-up and shutdown processes when gasses may have crossed through the cell; these boundaries can cause local cell reversal and result in rapid degradation [13]. To avoid flooding: liquid water droplets at the inlet may cause local starvation at the anode or cathode, leading to an unstable cell voltage and increasing the length of time taken for the stack to equilibrate.

Fuel cell short stack testing Chapter | 9 G

265

To avoid prolonged time at potentials greater than B0.85 V as this has been shown to accelerate cathode catalyst degradation by support corrosion, dissolution of catalyst alloying components, and catalyst degradation via oxidation.

Additional concerns may occur in specialized tests, such as when starting and stopping the cell in freezing conditions when ice formation may cause damage to the stack components.

9.4.2.1 Leak testing The first activity undertaken after connecting a stack to the test bench is a leak test. This usually checks for leaks from the anode and cathode to the atmosphere, as well as leaks in the stack connections and the test bench, and internally between the anode, cathode, and coolant. It stops potentially dangerous hydrogen leaks from occurring and can identify internal leaks which may indicate damage or conditions and reduce performance. Various detailed leak test protocols are available from the JRC [3], the STACKTEST project [20], and various IEC standards. In a typical test, the exhaust of the test station is sealed, and an inert gas is supplied to generate a pressure differential (e.g., anode environment or anode
cathode); once a pressure equal to the highest expected operating pressure of the stack is achieved, the flow is stopped and the pressure is monitored. The pressure drop over time is monitored with a faster decay equivalent to a bigger leak. In general, an acceptable leak rate is under 0.3 kPa min21, but stack and test station manufacturers often specify their own maximum permissible leak rates. At all times, care should be taken not to exceed the maximum pressure differentials specified by the stack manufacturer, which can cause damage to the membrane or gasketing. 9.4.2.2 Start-up Various start-up protocols have been defined [20,21] as these generally seek to avoid starvation, flooding, and fuel
air fronts; they can be adapted for most uses, but additional steps for implementing a dead-end anode, recirculation loop, or following a manufacturer’s operational thresholds may need to be added. The typical steps of a laboratory start-up protocol are: 1. Purging the cathode and then anode with nitrogen. 2. Pressurizing, heating, and humidifying the stack under nitrogen flow. 3. Introducing hydrogen to the anode side at which point the cell voltage usually rises and then falls as the anode potential drops while the cathode potential remains high, and then the cathode potential also falls as crossover hydrogen dominates its potential. 4. Introducing air to the cathode side at which point the cell voltage increases rapidly to its open-circuit voltage, typically . 0.9 V. 5. Applying a load.

266

Fuel Cells for Transportation

Due to the lack of inert purging gas in a real-world setting, hydrogen is sometimes very quickly flooded into the anode before the load is connected, and the fuel cell then heats up under operation. The high flow rates required for this step make it difficult to implement properly on a test bench using mass flow controllers. Where a purge gas is available, this protocol is generally recommended to completely avoid the formation of fuel
air fronts and for safety reasons. Some important considerations to be made when starting up a PEMFC stack include: G

G

G

G

The temperature of the gasses at the inlet should always be around 5 C above the coolant temperature, and above the gas dew point, to prevent condensation. Possible pressure spikes associated with a transition to dead-end mode should be accounted for and kept within the pressure limits of the stack and loop components, for example, the recirculation pump. When transitioning to dead-end mode, the stoichiometry will be set equal to 1; however, this can lead to starvation toward the outlet so the anode recirculation should be started before or immediately as the dead end is implemented. The open-circuit voltage and time at open circuit should be limited, so current should be drawn as soon as the cell voltages start rising.

9.4.2.3 Operation, break-in, and conditioning After a stack is started up, it is usually possible to operate it across a range of conditions; however, its performance even at steady operating conditions will fluctuate and evolve with time. On the short-time scale, around a second, noise on the scale of B1 mV is often observed in the cell voltage as local operating conditions fluctuate a small amount; on the tens of second scale to minutes, fluctuations are normally a result of cell equilibration or as a result of coarse fluctuation in the operating conditions (e.g., an oscillating temperature due to poor PID control). Equilibration is often observed after a change in operating conditions, which is particularly obvious after a large change in current; when the current increases rapidly, there is a corresponding fall in the cell voltage which then tends to recover slightly as applied gas flows stabilize, temperature control adjusts, and then as the various components inside each cell equilibrate with changed temperatures, humidity, etc. Often, to prevent starvation, gas flows are changed slightly before any change in current. When performing diagnostic measurements, it is common to specify a preconditioning step to ensure that the cell is equilibrated before performing the measurement; this is often steady-state operation until the cell voltages are stable to within some threshold. During the first few tens of hours of operation, new stacks are often seen to break in as the cell voltage increases slightly. The processes behind break-in are not yet completely elucidated and deserve a discussion beyond the scope of this

Fuel cell short stack testing Chapter | 9

267

work, yet there are various procedures applied to accelerate and standardize the process. Exact conditioning protocols for fuel cell stacks are dependent on the manufacturer or fuel cell testing lab, but generally include high current density holds, current/voltage/reactant cycling, hydrogen pumping, recovery cycles, or a combination of these. Additionally, the operating conditions such as reactant flow rate/stoichiometry, temperature, and relative humidity can be modulated during the conditioning protocol [22]. Fuel cell manufacturers have for many years tried to develop rapid break-in protocols that do not use inert gasses or expensive test stations, with several patents filed [23,24]. A standard break-in procedure used in many European fuel cell laboratories [20] consists of stabilizing operating conditions at a set current (e.g. 0.6 A cm22) for 1 hour before alternating periods of low and high current density (e.g. 0.2 and 1.2 A cm22) held for 180 seconds each, with the break-in complete when the average cell voltage is consistent (within 5 mV) over two subsequent periods of the same current density.

9.4.2.4 Shutdown Laboratory test procedures for the shutdown of fuel cells have also been defined on the single cell [21] and stack [20] levels. As with the start-up procedures, the shutdown protocols in a laboratory environment are often more cautious than in the “real world” to avoid any excessive degradation. Thus, inert gasses are used to purge the compartments. The cathode side is purged first, and a small load may be applied during purging to consume residual oxygen and limit the amount of time spent at open circuit. As the cathode purge continues, the cathode potential shifts from being dominated by the oxygen reduction reaction to being dominated by the hydrogen oxidation reaction due to hydrogen crossover through the membrane. This causes the stack voltage to fall very close to zero. In this condition, the stack is producing no power and is safe from starvation or other adverse processes, and thus, stacks are sometimes left in this state if only a short shutdown is required. In a full shutdown, the anode side is subsequently purged with nitrogen while the stack temperature, pressure, and humidity set points are gradually lowered to ambient conditions. It is important not to condense liquid water in the stack, which means that the dew point should always be lower than the stack temperature and gas temperature. If the stack is expected to be stored, transported, or exposed to freezing temperatures, it is often necessary to add additional drying steps. During a shutdown, at no point should the cell potential become negative, this indicates the formation of air
fuel mixes, often caused by incomplete purging of humidification systems or similar. 9.4.3

Diagnostics

Once a fuel cell is operating, it is typical to perform diagnostic measurements to characterize the performance and elucidate performance limitations

268

Fuel Cells for Transportation

or to understand a specific property of interest. Understanding why a stack is behaving in a certain way is crucial for optimizing both performance and lifetime. One key concern with diagnostics is the extent to which they themselves influence the performance and durability of the stack; many diagnostic techniques exacerbate degradation phenomena so their use must be carefully considered. Practically, diagnostics on stacks is more complex than on the single-cell level. As well as the more complex electrical connections and higher current/voltage requirements for equipment, the fact that individual cells are connected electrically in series means that controlling the potential of one single cell is more challenging. The techniques introduced here are those commonly or readily applied to short stacks. Many diagnostic techniques have been developed which are highly applicable to stack design but either monitor only one single repeat unit in a stack, for example, current mapping, or else require advanced instrumentation and often specially designed cells, for example, neutron imaging.

9.4.3.1 Polarization curves Polarization curves are the most common diagnostic tool as they are simple to perform and powerful. Simply described, the current is varied galvanostatically at otherwise constant operating conditions, and the voltage response is monitored. Practically, at each current step a dwell time is required, for the stack to equilibrate, before the voltage is averaged over 30 seconds. For short stacks, it is common to monitor every cell voltage individually and at least report the mean, minimum, and maximum voltages recorded. The dwell time for the stack to equilibrate within a few mV is typically on the order of a few minutes, but this needs to be evaluated on a case-by-case basis; to reduce the time at cell voltages above 0.85 V and therefore avoid rapid degradation [25], shorter dwell times B 60 seconds are usually used at low current densities. Typically, curves are recorded stepwise from a high current density, usually based on a stack design limit or minimum cell voltage, to the open circuit and back again. Even with long equilibration times at each step, there is often hysteresis between the forward and backward current sweeps, which may be attributed to liquid water in gas diffusion layers [26]. A typical fuel cell polarization curve displays four distinct regions indicative of which process is most limiting performance, as illustrated in Fig. 9.4A. At open-circuit potential, the cell voltage is dominated by crossover and equilibrium conditions; at low current densities by the activation overpotential required to drive the electrochemical reactions; at intermediate current densities by ohmic resistance; and at high current densities by mass transport resistance. In a well-designed stack, most cells will have very similar polarization curves and can achieve B 2 A cm22 with cell voltages . 0.6 V. Polarization curves are suitable for fitting to simple models to

Fuel cell short stack testing Chapter | 9

269

FIGURE 9.4 Illustrations of (A) polarization curve; (B) Nyquist plot of electrochemical impedance spectroscopy with an equivalent circuit.

separate activation, ohmic and mass transport losses, which then allows for accurate diagnosis of performance limitations. To aid in modeling and extracting various physical properties of the stack, polarization curves are often carried out at a range of different conditions, for instance, using cathode gasses with different concentrations of oxygen (e.g., pure oxygen or helium
oxygen mixtures) to better quantify mass transport limitations. These measurements operate in the same way as single cells but are less common due to the larger amounts of pure gas mixtures required. More simply, it is common to perform polarization curves at a range of operating conditions to identify the impact these have on achievable performance.

9.4.3.2 Electrochemical impedance spectroscopy Electrochemical impedance spectroscopy (EIS) is a powerful diagnostic technique that provides additional information on the causes of performance limitations, differentiating them by the time scale on which they occur. While a direct current (DC) load is applied during normal operation, in EIS an additional sinusoidal alternating current (AC) is superimposed on the DC, and the resulting sinusoidal voltage perturbation is monitored. To ensure a linear response, the AC perturbation is usually of small amplitude; 5%
10% of the applied DC load is typical. In a normal experiment, the AC frequency is swept between B 10 kHz and B 100 mHz, recording several points at each decade of frequency and averaging over a number of periods to ensure a good signal-to-noise response. Practically, EIS is performed using a frequency response analyzer, often as part of a laboratory potentiostat/galvanostat with a booster. The AC signal is applied over the entire stack, and usually, only the voltage response of the stack is measured, though multichannel instruments can be connected to each cell in a stack to record individual cell responses. A Nyquist plot of an idealized EIS response from a fuel cell is shown in Fig. 9.4B. The high-frequency intercept is characteristic of the ohmic resistance of cell components such as the membrane, catalyst layer, backing, and

270

Fuel Cells for Transportation

end plate, and the contact resistances between each of them. The contribution from the charge transfer (activation) is shown in the mid-frequency range. It is caused by the interfacial kinetics of the ORR process occurring at the cathode, and the double-layer capacitance within the catalyst layer. The contributions from the mass transport typically occur at the low-frequency range of the spectra. These mass transport limitations arise from the nonuniform distribution of reactants. In real data, individual features are rarely well resolved, and EIS responses often include other features such as inductive loops, extra or fewer semicircles, or a Warburg impedance. Quantitative analysis of EIS data requires fitting it into a model. A major issue with impedance analysis is that no model is unique in describing a measured spectrum; as such validated physical models of the system are preferred to empirical models and overfitted empirical models must be avoided.

9.4.3.3 Voltammetry Many electrochemical techniques used to characterize electrochemical systems and single-cell fuel cells are voltammetric, where the potential of an electrode (the working electrode) is controlled against another (a combined reference and counter electrode), and the current passed between the two is measured. These are often applied to fuel cells to characterize the surfaces of catalysts, measure crossover, and detect short circuits. In fuel cell measurements, the electrode of interest or working electrode (either anode or cathode) is supplied with inert gas, while the combined counter-reference electrode is supplied with hydrogen, to provide a facile counter electrode and a stable reference potential. A potentiostat is used to control the potential and monitor the current. With a stack, it is not possible to apply a voltammetric measurement across the entire stack as the current observed will be limited by the worst performing cell, and certain cells or parts of cells may experience extreme, and therefore corrosive potentials. Instead, connections must be made to pairs of adjacent bipolar plates to perform voltammetry on one cell at a time. This process can therefore be time-consuming. Note that for repetitive measurements, especially fast cycling, the application of high potentials may degrade the working electrode, so it is important to limit the amount of voltammetry, as well as the higher potential limit, and consider the impact of this diagnostic technique on the measurement. In linear sweep voltammetry (LSV), the potential of the electrode is swept in one direction at a constant scan rate. It is commonly used to quantify the hydrogen crossover, or the amount of hydrogen gas crossing over from the anode to the cathode. During such experiments, nitrogen is supplied to the cathode, and the anode is kept under 100% hydrogen; an applied cathode potential positive of B 0 V versus the anode under hydrogen oxidizes any hydrogen that has crossed over, producing a current. In a typical measurement, the potential of one cell is swept at a scan rate of 1 mV s21

Fuel cell short stack testing Chapter | 9

271

starting from a potential of 0.1 V until a potential of 0.6 V is reached. As the potential is swept, a limiting current will be reached where all the hydrogen molecules that have made their way across the membrane will be oxidized by the working electrode catalyst layer; this limiting current may then be back-calculated to a hydrogen flux with corrections for double-layer charging if necessary. LSV can also be used to identify electrical short circuits through the membrane between the anode and cathode electrodes with a current that is linearly dependent on voltage indicative of a short and easily used to estimate the resistance. In cyclic voltammetry (CV), the experimental setup is essentially the same, though diluted hydrogen (typically 5% in nitrogen) may be used at the reference electrode to minimize the influence of hydrogen crossover. For CVs, the potentiostat is used to cycle the potential of the working electrode between two limits often B0.05 and 0.8 V versus the anode under hydrogen at a constant scan rate, typically 20 mV s21.

9.4.3.4 Electrochemically active surface area The electrochemically active surface area (ECSA) is nominally defined as the area of the catalyst in an electrode that is available to perform electrochemical reactions, and it is strongly correlated with the performance of cells. Its loss over time is a major degradation mechanism so quantification of this parameter is useful. In addition, it is a state-of-health monitor for the catalyst and will have a large impact on the stack performance and impurity tolerance. There are several methods for estimating the ECSA of platinum and platinum alloy catalysts in a fuel cell stack. CVs may be used to estimate the ECSA of the platinum catalysts in fuel cell electrodes. At a potential of B 0 V versus RHE in aqueous conditions, a monolayer of Habs is present on the surface of a platinum catalyst. By cycling the potential between B 0 and 0.35 V versus RHE, the charge associated with the absorption and desorption can be measured. This charge is proportional to the ECSA with a charge of 210 μC cm22 usually used for polycrystalline platinum. Typically, ECSA measurements are performed by depressurizing the anode and cathode compartments, then flowing nitrogen on the side where the ECSA will be measured, and diluted hydrogen on the counter-reference electrode. Diluted hydrogen is preferred in order to avoid high hydrogen crossover rates influencing the shape of the voltammogram [27]. A high relative humidity is recommended, and flow rates should be low, approximately equal to 1 cmstp3 cm22 active area [28]. In general, calculating the ECSA from the desorption peak is less reliable for studying large electrodes like a PEMFC stack because of the problems associated with proper baseline correction in an often-skewed anodic scan [29]. Therefore, it is recommended to use the adsorption peak. Carbon monoxide (CO)-stripping voltammetry has been shown to be more accurate than the more commonly used hydrogen adsorption/desorption

272

Fuel Cells for Transportation

voltammetry, especially for small platinum nanoparticle catalysts and platinum-alloy nanoparticle catalysts [27]. In this technique, the charge associated with the formation of a CO monolayer is used to estimate the ECSA. Diluted hydrogen (e.g., 5% hydrogen in nitrogen) is typically used on the anode side of the cell, while diluted CO (e.g., 1% CO in nitrogen) is purged over the cathode side to create an adsorbed CO monolayer on the cathode catalyst, followed by purging with an inert gas to remove any remaining CO. The potential is swept in a similar method to the traditional ECSA measurement using CV, with the first cycle producing a distinct CO oxidation peak and the second cycle acting as a baseline. The area of the CO oxidation peak is proportional to the ECSA with a charge density of 420 μC cm22 often used, as CO oxidation is a two-electron process compared to hydrogen adsorption/desorption which only requires one electron. The measurement of ECSA by this method assumes that each molecule of CO occupies one site on the available platinum surface and that all sites that are active, accessible, and occupied during the measurement. To avoid the need to perform voltammetry on every cell in a stack, a potentiometry technique was developed by Brightman et al. [30] to estimate cathode catalyst ECSA. Briefly, the fuel cell is conditioned with hydrogen on the anode and nitrogen on the cathode. A small amount of air is introduced to the cathode to increase the cell voltage to open circuit; then, nitrogen is again introduced to allow the cell voltages to drop. Once the cell voltages have dropped to around 0.8 V, a small constant current B 1 to 8 mA cm22 is drawn from the stack, to force the potential to drop further. The time taken for the voltage of each cell to fall from 0.4 to 0.1 V is measured, and the total charge passed during this time is assumed to be equivalent to the charge required to oxidize Habs on the platinum. There are numerous systematic errors introduced by this technique including the influence of hydrogen crossover, and the choice of applied current density is found to be important [30].

9.4.3.5 Current interrupt Current interrupt (CI) is a technique commonly used to measure the ohmic resistance inside an operating fuel cell on the single cell and stack level. In this technique, the load is periodically disconnected, and the transient voltage response is monitored. The difference in the voltage immediately before and after the current removal can be used to extract the ohmic losses, which occur on the smallest timescale (as observed in the highest frequency region of EIS spectra) [31]. For stack testing, CI measurements provide data on the individual cell level if the voltage of each cell is being monitored individually, assuming that the data collection rate is quick enough [2]. The CI technique is a simpler and more accessible way of monitoring individual cell performance via ohmic losses than EIS and needs much less specialized equipment.

Fuel cell short stack testing Chapter | 9

273

9.4.3.6 Online gas and water analysis Analysis of the gas composition of the anode or cathode outlets and in the anode recirculation loop can provide a further understanding of stack behavior and details of testing conditions. For example, quantification of gas composition on the anode side is especially relevant when employing a recirculation loop due to the buildup of crossover gasses from the cathode, which will have an impact on performance, while at the outlets, gasses from the oxidation of the carbon support may be used to understand corrosion processes. Techniques such as gas chromatography, Fourier transform infrared spectroscopy, optical feedback cavity enhanced absorption spectroscopy, or selected ion flow tube mass spectrometry have all been used for online gas analysis, and any instrument that can effectively quantify trace gasses in a hydrogen matrix with sufficient sensitivity and time resolution for the required measurement can be used. During experiment design and technique selection, the impact of any analytical equipment on the hydrogen utilization should be considered along with the need for drying or other pretreatments of the gas [32]. Analysis of effluent water from a stack can detect the presence of soluble degradation products, for example, metal ions from bipolar plates or fluoride ions from membrane and ionomer decomposition. Several relevant techniques have been applied including fluoride ion selective electrode, or ion chromatography, which has been used to identify membrane degradation in PEM electrolyzer short stacks through fluoride ion release [33]. However, the use of such techniques only provides information about the whole stack, and quantification remains imprecise as impurities may remain adsorbed into the membrane or deposit on other components in the fuel cell stack. 9.4.3.7 Impurity analysis The tolerance of the fuel cell stack to certain impurities is sometimes investigated, particularly when it may be operated on hydrogen which is anticipated to contain trace contaminants or to evaluate recovery or mitigation strategies. Impurities such as CO, hydrogen sulphide, ammonia, and others may be present based on the source of the hydrogen fuel, as well as cathode contaminants such as salts, NOx, and SOx from the operating environment. Such impurities can impact different parts of the fuel cell [34]; for example, CO acts primarily on the catalyst by blocking active sites [35], whereas ammonia will both reduce the conductivity of the membrane and block the catalyst sites [36]. The deliberate application of impurities can also be used as a tool to understand processes inside a fuel cell stack. For example, the tolerance to CO can give an indication of the state of health of the catalyst layer, as the drop in cell voltage associated with CO poisoning should be dependent on the coverage of CO on the platinum catalyst. A more heavily degraded (i.e., lower surface area) catalyst will therefore poison more quickly. Similarly, isotopically labeled CO has been used to investigate oxygen crossover [37].

274

9.4.4

Fuel Cells for Transportation

Durability testing

To understand the degradation phenomena and evaluate degradation rates, durability testing of fuel cell stacks is performed. Testing the durability of stacks is both very important and very time-consuming. As stacks are designed to last for many thousands of hours of operation, with heavy-duty targets of up to 30,000 hours operation with a loss in nominal performance of 10%, the degradation rates are very low under normal operation. The degradation rate, often reported in μV h21, is highly variable based on the load set point [2], but will also vary over the stack lifetime, with higher degradation rates observed at the beginning and end of life [38]. Degradation or aging protocols are designed to emulate the typical operation of fuel cell stacks under various conditions and applications. They can be defined as either natural or accelerated. Natural aging protocols consist of applying normalized drive cycles (also known as dynamic load cycles) to the stack to simulate “real-world” operations for various applications. There are many examples of drive cycles that have been developed worldwide [7], and drive cycle selection is based on the type of application, hybridization strategy, and sometimes location. Natural aging experiments can take tens of thousands of hours, dependent on the load profile, stack degradation rate, and end-of-life criteria, so there are very few who decide to test a fuel cell stack over the entire projected lifetime. Instead, degradation rates can be measured over several hundred hours and extrapolated with the help of mathematical models. Natural aging protocols are combined with diagnostic measurements at shorter intervals, to understand the specific degradation mechanisms. Alternatively, accelerated stress tests (ASTs), are designed to increase the observed stack degradation rate by accelerating the rate of one or several degradation mechanisms by applying stressors such as harsher operating conditions, mechanical effects, impurities in the inlet gas streams, external environmental conditions, or load cycling. These accelerated protocols can make a fuel cell stack reach its end-of-life performance in 500 hours or less. However, it is difficult to precisely correlate an accelerated and a natural degradation rate, as the current/voltage profile and thus degradation phenomena are often vastly different. The overall voltage loss observed after a period of testing at constant or dynamic load may be recovered after shutdown and restart. The part of the cell voltage that is recovered is called the reversible degradation, and the part that is not recovered is the irreversible degradation. Possible mechanisms for reversible degradation include platinum surface oxidation, catalyst contamination, ionomer and membrane structural changes or contamination, catalyst support degradation, or poor water management. Some of the degradation methods can be completely recovered by a shutdown step including cooling of the stack (e.g., water management); however, degradation due to

Fuel cell short stack testing Chapter | 9

275

poisoning may be recovered through potential cycling, with low potentials (around 0.1 V) particularly beneficial for impurity desorption. A thorough review of reversible degradation mechanisms has been published by Mitzel. et al. [39].

9.4.4.1 Constant load operation Applying a constant load for a set amount of time is a subset of natural aging durability testing that allows understanding of specific phenomena that may occur at different load conditions. This is of relevance to end-use applications where operational profiles may be different from those typically tested; for example, the high efficiency and long lifetime required by fuel cells for heavy-duty shipping may necessitate operation at lower current densities and more stable operational profiles, than in automotive applications where higher powers are required. Long-term degradation rates at various current set points can be used as input into degradation models that try to understand the impact of drive cycle behavior on the fuel cell stack. In theory, the degradation rate after a single drive cycle should be the average of the degradation rates at each current density relative to the amount of time spent at each. However, degradation rates are often higher for dynamic load cycles because the dynamic behavior puts extra stress on the fuel cell stack due to large changes in heat and water production. Quantifying this difference is an important part of understanding stack degradation in real-world conditions. 9.4.4.2 Start/stop durability It has been shown that often the most severe degradation during fuel cell operation occurs during the shutdown and start-up procedures [40]. Therefore, quantifying the degradation rate during start-up or shutdown is an important part of understanding stack behavior. Start-up/shutdown cycles can be performed where the stack is sequentially started up and shut down, with some performance characterization after a set number of cycles to determine the performance degradation; thus, a start-up/shutdown penalty can be generated. Since frequent start-ups and shutdowns are more relevant to real-world operation, protocols that do not use inert gasses are the most realistic and should be applied during start/stop durability analysis.

9.5

Summary

New materials, cell designs, operating conditions, manufacturing processes, etc., are often validated by testing short stacks, typically of 5 to 20 cells, with cell designs and manifolds identical to those in real devices. Testing at this scale therefore bridges as highly experimental and industrially useful and so is therefore highly relevant. Nevertheless, there are limited practical

276

Fuel Cells for Transportation

resources available to those wishing to perform testing on short stacks. This chapter has provided an introductory guide for characterizing fuel cell short stacks, outlining the principles, facilities, operating conditions, techniques, and considerations necessary to operate a short stack. Techniques that can be used for performance measurement and diagnosis of performance limitations were also introduced. When combined, a clear picture of the stack performance, the limitations on this performance, and degradation mechanisms at different stages of the stack lifetime and under different operating conditions can be obtained.

Acknowledgments This work was produced with funding from “Metrology for Hydrogen Vehicles 2” (MetroHyVe 2) which received funding from the EMPIR program cofinanced by the Participating States and from the European Union’s Horizon 2020 research and innovation program under Grant Agreement No. 19ENG04.

References [1] Key Performance Indicators (KPIs), ,https://www.clean-hydrogen.europa.eu/knowledgemanagement/key-performance-indicators-kpis_en., n.d. (accessed May 5, 2022). [2] M. Miller, A. Bazylak, A review of polymer electrolyte membrane fuel cell stack testing, J. Power Sources 196 (2011) 601
613. Available from: https://doi.org/10.1016/J. JPOWSOUR.2010.07.072. [3] G. Tsotridis, A. Pilenga, G. De Marco, T. Malkow, EU harmonised test protocols for PEMFC MEA testing in single cell configuration for automotive applications, JRC Science for Policy Report, 2015. https://doi.org/10.2790/54653. [4] X.Z. Yuan, H. Li, S. Zhang, J. Martin, H. Wang, A review of polymer electrolyte membrane fuel cell durability test protocols, J. Power Sources 196 (2011) 9107
9116. Available from: https://doi.org/10.1016/J.JPOWSOUR.2011.07.082. [5] S. Stariha, N. Macauley, B.T. Sneed, D. Langlois, K.L. More, R. Mukundan, et al., Recent advances in catalyst accelerated stress tests for polymer electrolyte membrane fuel cells, J. Electrochem. Soc. 165 (2018) F492
F501. Available from: https://doi.org/ 10.1149/2.0881807JES/XML. [6] Fuel cell technologies—Part 7-1: test methods—single cell performance tests for polymer electrolyte fuel cells (PEFC) IEC TS 62282-7-1:2017, 2017. [7] H. Chen, Z. Song, X. Zhao, T. Zhang, P. Pei, C. Liang, A review of durability test protocols of the proton exchange membrane fuel cells for vehicle, Appl. Energy. 224 (2018) 289
299. Available from: https://doi.org/10.1016/J.APENERGY.2018.04.050. [8] I. Bloom, L.K. Walker, J.K. Basco, T. Malkow, A. Saturnio, G. De Marco, et al., A comparison of Fuel Cell Testing protocols—a case study: protocols used by the U.S. Department of Energy, European Union, International Electrotechnical Commission/Fuel Cell Testing and Standardization Network, and Fuel Cell Technical Team, J. Power Sources. (2013). Available from: https://doi.org/10.1016/j.jpowsour.2013.06.026. [9] Fuel cell technologies—Part 4-102: fuel cell power systems for industrial electric trucks— Performance test methods, IEC 62282-4-102:2017, 2017.

Fuel cell short stack testing Chapter | 9

277

[10] Stack-Test Test Modules, ,http://stacktest.zsw-bw.de/media-centre/test-modules.html., 2015 (accessed May 5, 2022). [11] C. Harms, F. K¨ohrmann, A. Dyck, Study of the influence of key test parameters on the performance of a PEMFC stack, Solid. State Ion. 275 (2015) 75
79. Available from: https://doi.org/10.1016/J.SSI.2015.03.023. [12] Fuel cell technologies—part 2-100: fuel cell modules—safety, IEC 62282-2-100:2020, 2020. [13] Hydraite Project Website, www.hyrdaite.eu, n.d. (accessed May 5, 2022). [14] A. Klymovskyi, A. Kabza, J. Ihonen, J. Viitakangas, D2.3 First recommendation for short stack test methods for studying hydrogen contaminants in automotive PEMFC systems, (2019) 11. [15] Hydrogen fuel quality—product specification, ISO 14687:2019, 2019. [16] Compressed air—Part 1: contaminants and purity classes, ISO 8573-1:2010, 2010. [17] Standard Specification for Reagent Water, ASTM D1193-06(2018), 2018. [18] P. Koski, L.C. Pe´rez, J. Ihonen, Comparing anode gas recirculation with hydrogen purge and bleed in a novel PEMFC laboratory test cell configuration, Fuel Cell 15 (2015) 494
504. Available from: https://doi.org/10.1002/FUCE.201400102. [19] P. Koski, J. Viitakangas, J. Ihonen, Determination of fuel utilisation and recirculated gas composition in dead-ended PEMFC systems, Int. J. Hydrog. Energy. 45 (2020) 23201
23226. Available from: https://doi.org/10.1016/J.IJHYDENE.2020.04.252. [20] A. Kabza, L. Topal, P. Piela, B. Guicherd, J. Mitzel, J. Hunger, et al., Stack-Test Master Doc. (2015). Available from: http://stacktest.zsw-bw.de/. [21] G. De Marco, T. Malkow, G. Tsotridis, A. Pilenga, EU harmonised test protocols for PEMFC MEA testing in single cell configuration for automotive applications, 2015. https://doi.org/10.2790/54653. [22] S.S. Kocha, B.G. Pollet, Advances in rapid and effective break-in/conditioning/recovery of automotive PEMFC stacks, Curr. Opin. Electrochem. 31 (2022) 100843. Available from: https://doi.org/10.1016/J.COELEC.2021.100843. [23] J. Zhang, N. Ramaswamy, B. Lakshmanan, S.P. Kumaraguru, Fuel cell stack break-in procedures and break-in conditioning systems, US20180261858A1, 2017. [24] H.S. Choo, J.H. Lee, H.S. Shin, S.K. Lee, Pre-activation method for fuel cell stack, US20140045086A1, 2012. [25] N. Macauley, D.D. Papadias, J. Fairweather, D. Spernjak, D. Langlois, R. Ahluwalia, et al., Carbon corrosion in PEM fuel cells and the development of accelerated stress tests, J. Electrochem. Soc. 165 (2018) F3148
F3160. Available from: https://doi.org/10.1149/ 2.0061806jes. [26] V. Gurau, F. Barbir, H. Liu, An analytical solution of a half-cell model for PEM fuel cells, J. Electrochem. Soc. (2000) 2468. Available from: https://doi.org/10.1149/1.1393555. [27] T.R. Garrick, T.E. Moylan, M.K. Carpenter, A. Kongkanand, Electrochemically active surface area measurement of aged Pt alloy catalysts in PEM fuel cells by CO stripping, J. Electrochem. Soc. 164 (2017) F55
F59. Available from: https://doi.org/10.1149/2.0381702jes. [28] K.R. Cooper, In situ pem fuel cell electrochemical surface area and catalyst utilization measurement, Fuel Cell Mag. (2009) 1
3. Available from: https://doi.org/10.1016/S14642859(00)80060-7. [29] P. Piela, J. Mitzel, S. Rosini, W. Tokarz, F. Valle, A. Pilenga, et al., Looking inside polymer electrolyte membrane fuel cell stack using tailored electrochemical methods, J. Electrochem. Energy Convers. Storage. 17 (2020). Available from: https://doi.org/ 10.1115/1.4046106.

278

Fuel Cells for Transportation

[30] E. Brightman, G. Hinds, R. O’Malley, In situ measurement of active catalyst surface area in fuel cell stacks, J. Power Sources 242 (2013) 244
254. Available from: https://doi.org/ 10.1016/j.jpowsour.2013.05.046. [31] T. Mennola, M. Mikkola, M. Noponen, T. Hottinen, P. Lund, Measurement of ohmic voltage losses in individual cells of a PEMFC stack, J. Power Sources 112 (2002) 261
272. Available from: https://doi.org/10.1016/S0378-7753(02)00391-9. [32] HYDRAITE Public Reports, ,https://hydraite.eu/public-reports/., 2021 (accessed May 5, 2022). [33] P. Marocco, K. Sundseth, T. Aarhaug, A. Lanzini, M. Santarelli, A.O. Barnett, et al., Online measurements of fluoride ions in proton exchange membrane water electrolysis through ion chromatography, J. Power Sources 483 (2021) 229179. Available from: https://doi.org/10.1016/J.JPOWSOUR.2020.229179. [34] B. Shabani, M. Hafttananian, S. Khamani, A. Ramiar, A.A. Ranjbar, Poisoning of proton exchange membrane fuel cells by contaminants and impurities: review of mechanisms, effects, and mitigation strategies, J. Power Sources. 427 (2019) 21
48. [35] Y. Matsuda, T. Shimizu, Y. Hashimasa, Effect of carbon monoxide on polymer electrolyte fuel cell performance with a hydrogen circulation system, J. Electrochem. Soc. 167 (2020) 0044509. [36] Y.A. Gomez, A. Oyarce, G. Lindbergh, C. Lagergren, Ammonia contamination of a proton exchange membrane fuel cell, J. Electrochem. Soc. 165 (2018) F189
F197. [37] H. Becker, T. Bacquart, M. Perkins, N. Moore, J. Ihonen, G. Hinds, et al., Operando characterisation of the impact of carbon monoxide on PEMFC performance using isotopic labelling and gas analysis, J. Power Sources Adv. 6 (2020) 100036. Available from: https://doi.org/10.1016/j.powera.2020.100036. [38] T. Matsuura, J. Chen, J.B. Siegel, A.G. Stefanopoulou, Degradation phenomena in PEM fuel cell with dead-ended anode, Int. J. Hydrog. Energy. 38 (2013) 11346
11356. Available from: https://doi.org/10.1016/J.IJHYDENE.2013.06.096. [39] J. Mitzel, Q. Zhang, P. Gazdzicki, K.A. Friedrich, Review on mechanisms and recovery procedures for reversible performance losses in polymer electrolyte membrane fuel cells, J. Power Sources 488 (2021) 229375. Available from: https://doi.org/10.1016/J. JPOWSOUR.2020.229375. [40] T. Fletcher, R. Thring, M. Watkinson, An energy management strategy to concurrently optimise fuel consumption & PEM fuel cell lifetime in a hybrid vehicle, Int. J. Hydrog. Energy. 41 (2016) 21503
21515.

Chapter 10

Power demand for fuel cell system in hybrid vehicles Rui Ma1, Elena Breaz2 and Fei Gao2 1

School of Automation, Northwestern Polytechnical University, Xi’an, P.R. China, 2University of Technology of Belfort-Monte´bliard, Belfort, France

10.1 Introduction to hybrid fuel cell powertrain To cope with climate change and air pollution, different targets have been set up worldwide, one of them being the cutoff of at least 40% in greenhouse gas (GHG) emissions by 2030. As the transport sector is one of the main contributors to GHG emissions and the main cause of air pollution in cities, it is directly concerned with these emission reduction targets. To reach this goal, the transport sector aims to replace conventional vehicles with environment-friendly vehicles. Among environment-friendly vehicles, one of the potential solutions is fuel cell hybrid electric vehicle (FCHEV). The fuel cell, more precisely the proton exchange membrane fuel cell (PEMFC), is an electrochemical energy conversion device that can directly convert hydrogen energy into electricity. PEMFC has received extensive research attention, and it can have many advantages to be used for transport sector, such as high-power density, silent operation, low operating temperature (which gives them a relatively short start-up time), no electrolyte leakages risk due to the solid polymer membrane, and zero GHG emission. All these advantages make hydrogen fuel cell to be naturally considered in various transportation applications such as light-duty vehicles, trains, trucks, buses, and even aircraft. However, the power dynamic performance of the fuel cell is usually limited due to the internal electrical
fluidic
thermal characteristics, which may fail to meet the full requirements of the vehicles under high dynamic operating conditions. Therefore, a fuel cell powertrain is usually a hybrid one. By hybridization, the high dynamics which cannot be delivered by the fuel cell are taken by the additional power source. Moreover, to further reduce the hydrogen consumption of the fuel cell vehicle, it is important to recover the brake energy, which is impossible with a standalone fuel cell Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00016-2 © 2023 Elsevier Ltd. All rights reserved.

279

280

Fuel Cells for Transportation

powertrain since the fuel cell system is not reversible. Therefore, the fuel cell powertrain must be hybridized with other energy sources/storage systems to increase its efficiency. By hybridization, the complementary advantages of each power source in terms of energy and power are explored, thus leading to better overall powertrain performances. A hybrid fuel cell powertrain will at least include a fuel cell power generation system and an additional power source/energy storage device, such as lithium-ion batteries, supercapacitors, flow batteries, and other power sources with good power density. As shown in Fig. 10.1, FCHEV is based on a series powertrain hybrid architecture, and the hybrid powertrain is composed of three different energy sources: fuel cell (as primary energy source), battery, and supercapacitor. The fuel cell is connected to the DC bus through a unidirectional DC/DC converter, while the battery and supercapacitor could be connected through bidirectional DC/DC converters. Bidirectional DC/DC converters allow energy flow in both directions to recover the brake energy. The DC bus provides power to the electric motor via a DC/AC inverter. Obviously, other combinations are possible such as fuel cell 1 battery, fuel cell 1 supercapacitor. As it has been already mentioned previously, hybrid powertrain can reach a higher dynamic performance as the instantaneous power is being provided by the battery or supercapacitor (depending on the combination). In these hybrid systems, PEMFC provides the main power with relatively slow dynamics of the load demand. The lithium-ion battery and/or supercapacitor as additional power sources/energy storage devices can help to provide the instantaneous power of the load demand. Based on the hybrid powertrain configuration, if only two power sources are included, the fuel cell power supply will be complemented by the other power source to meet the load demand. If three or more power sources are included, the load power demand distribution method needs to be designed to decide how to match power among them.

FIGURE 10.1 FCHEV architecture.

Power demand for fuel cell system in hybrid vehicles Chapter | 10

281

The algorithm for planning the power matching mode of each power source to the load demand in the system is defined as an energy management strategy (EMS). For the complex nonlinear system shown in Fig. 10.1, a reasonably designed EMS is the guarantee for the safe and stable operation of the system. A carefully designed EMS first needs to be able to match the output power of the hybrid fuel cell powertrain with the load demand. When the load fluctuates, the hybrid fuel cell powertrain should adjust the output power in time to ensure that the DC bus voltage remains stable. In addition, the hybrid fuel cell powertrain also needs to be able to control the energy storage level of the energy storage devices. When the energy storage level is higher than the reference value, the fuel cell should reduce the proportion of its output power to increase the proportion of the output power of the energy storage device, so that the energy storage could be discharged to meet again the reference value. When the energy storage level is lower than the reference value, the fuel cell should increase its output power to reduce the proportion of the output power of the energy storage device, so that the energy storage device could be charged to reach the reference value. For the hybrid powertrain in FCHEV, the EMS needs to guarantee a flexible load demand power distribution. The EMS is the core of power demand control in FCHEV.

10.2 Fuel cell hybrid electric vehicle road testing profiles A driving cycle represents a typical vehicle driving pattern developed either by different countries and organizations or by researchers to assess the vehicle performance in terms of fuel consumption, emissions estimation, driving range, and so on. In other words, a driving cycle is a series of data points representing the speed of a vehicle versus time. Driving cycles are largely employed by industry and academia for different purposes. First, their main use is in vehicle emission certification procedure for newly manufactured vehicles. Second, driving cycles are extensively used by researchers for simulations in the vehicle’s design phase to compare the efficiency, functionality, and performance of the powertrain. Moreover, in the case of fuel cell vehicles, the driving cycles are used as well in powertrain modeling and fuel cell power sizing. Several standard driving cycles have been deployed during past decades. Based on their characteristics, driving cycles can be split into two categories, more specifically modal driving cycles and transient driving cycles [1]. Modal driving cycles, such as New European Driving Cycle (NEDC) or Japanese 10 mode (J10), are combining constant driving phases including cruising, acceleration/deceleration, and idling [2]. These driving cycles are considered unrealistic today leading to an underestimation of emissions and fuel consumption. On the other hand, transient driving cycles, such as the Worldwide Harmonized Light Vehicle Test Cycle (WLTC) or Federal Test

282

Fuel Cells for Transportation

Procedure (FTP-75), cover a wider range and variation of speed and acceleration [2]. Usually, these cycles are based on real driving data, being more realistic than modal cycles. For many years, NEDC has been largely used as a reference driving cycle for the vehicle’s certification. However, it cannot accurately represent real driving behavior, as it contains many unrealistic constant driving phases. In real-world driving, the transitions in speed and acceleration are much more aggressive, leading to higher fuel consumption and emissions than NEDC. Therefore, starting with 2017, the NEDC has been replaced with the WLTC [3] to predict more accurately the emissions and fuel consumption under real-world driving conditions. The WLTC has been developed based on real-world driving data collected from different regions of the world, including Europe, the United States, Japan, Korea, and India. Three different driving cycle classes have been developed due to the limited driving capabilities of certain vehicles. These classes have been designed with respect to power-to-mass ratio and maximum vehicle speed [3]. For most passenger cars, the concerned class is class 3
2 which is presented in Fig. 10.2. As shown in Fig. 10.2, the duration of this driving cycle is 30 minutes and comprises four different speed regions: low speed, medium speed, high speed, and extra high speed. Specifically, the low-speed range lasts for 589 seconds, the medium speed for 433 seconds, the high speed for 455 seconds, and the extra high speed for 323 seconds. The covered distance is almost 23.3 km, which is more than double compared with NEDC. This implies a lower influence of cold-start emissions. The maximum speed of WLTC is 131.3 km h21, while

FIGURE 10.2 WLTC class 3
2.

Power demand for fuel cell system in hybrid vehicles Chapter | 10

283

the average speed is 46.5 km h21. Idling represents 12.6% and is reflected in start
stop system effects, being relatively lower compared with NEDC. The cruising and the transient time represent 3.7% and 83.7%, respectively, while the maximum acceleration is 1.67 m s22. All these characteristics place the WLTC in the transient cycle type, with effects on a more realistic estimation of the emissions and fuel consumption. There are several other standard driving cycles such as Urban Dynamometer Driving Schedule, U.S. Federal Highway HFET, California LA-92, New York City cycle, Japanese JC08, which are used for the various analysis of vehicles in different regions of the world [4]. However, all these cycles have similar drawbacks. Either were developed for a given city or region or were not able to represent harmonized real-world driving patterns. In consequence, their use in vehicle analysis will lead to similar unrealistic results as NEDC. Therefore, in the automotive industry as well as in research, people are using today’s other driving cycles, such as Artemis driving cycles (urban, rural road, and motorway driving cycles) [5], which can better approach the real driving patterns. For exemplification, the Artemis motorway driving cycle is presented in Fig. 10.3. This driving cycle has two variants with maximum speeds of 130 and 150 km h21. Fig. 10.3 depicts the version with a maximum speed of 150 km h21. The total duration of the Artemis motorway 150 driving cycle is 1068 seconds, while the driving distance is 29.5 km. To conclude, different driving cycles available in the literature are very useful in vehicle performance assessment. These driving cycles can approach

FIGURE 10.3 Artemis motorway driving cycle.

284

Fuel Cells for Transportation

more or less real-world driving patterns. Usually, driving cycles are used as input parameters, no matter is for vehicle emissions estimation, fuel consumption, powertrain modeling, etc. In the next section, we will demonstrate how to use a driving cycle as well as other input parameters to determine the total power demand for the fuel cell hybrid powertrain. The chosen driving cycle is WLTC class 3
2, as it is largely used by vehicle manufacturers and researchers as well.

10.3 Fuel cell hybrid electric vehicle body modeling As it has been mentioned in the introduction of this chapter, an FCHEV is naturally a hybrid one, due to the relatively slow dynamics of a fuel cell. Therefore based on the configuration chosen, the power needed to advance the vehicle could be provided either by the fuel cell or by the batteries, supercapacitors, etc. It is necessary to determine the total power needed by the fuel cell hybrid powertrain, to be able to size its different power energy sources. A common way to determine the total power profile of the powertrain is to use a given driving cycle. Indeed, any driving cycle can be used to convert vehicle speed profile into electrical power demand profile. More specifically, to determine the electrical power profile of the powertrain, the vehicle speed in the driving cycle is first converted into mechanical power profile; then, the mechanical power profile is finally converted into the electrical power profile. The driving cycle data are used as an input of vehicle body model to determine the power profile at each time point of the cycle. For any vehicle, no matter if it is powered by an internal combustion engine or an electric engine, its movement behavior along its movement direction is determined by all the forces acting on it in the driving direction. As shown in Fig. 10.4, the main forces acting on a vehicle moving are wheels traction force (Ftrac ), aerodynamic drag force (Faero ), rolling resistance force (Froll ), and grade resistance force (Fgrad ). The wheels traction

FIGURE 10.4 Forces acting on a vehicle moving uphill.

Power demand for fuel cell system in hybrid vehicles Chapter | 10

285

force (Ftrac ) moves the vehicle forward while encountering the opposition to its motion of all the other resistance forces (Faero ; Froll and Fgrad ). The vehicle model can be described based on Newton’s second law: mv UδU

dv 5 Ftrac 2 Faero 2 Froll 2 Fgrad dt

ð10:1Þ

where mv is vehicle mass in kg, v is vehicle ground speed in m s21, δ is mass factor, equivalent translational mass of rotational inertias of rotating components (for ordinary vehicle, δ is comprised between 1.08 and 1.1), Ftrac is the wheels traction force in N, Faero is the aerodynamic drag force in N, Froll is the rolling resistance force in N, and Fgrad is the grade resistance force in N. Then, the wheels traction power (Ptrac ) in W can be obtained by: Ptrac 5 Ftrac Uv

ð10:2Þ

where the wheels traction force, extracted from Eq. (10.1), is: Ftrac 5 mv UδU

dv 1 Faero 1 Froll 1 Fgrad dt

ð10:3Þ

and it depends on vehicle speed and acceleration, which can be easily obtained from the driving cycle data. The aerodynamic drag force, acting on a moving vehicle, is caused first by the viscous friction of the surrounding air on the vehicle surface. Second, the losses are also caused by the pressure difference between the front and the rear of the vehicle, generated by a separation of the airflow. For a passenger car, the aerodynamic resistance is mostly due to the car body (around 65%). The rest is due to the wheel housings (around 20%), the exterior mirrors, window housings, antennas, etc., (around 10%), and the engine ventilation (around 5%) [6]. Usually, the aerodynamic resistance force is approximated by simplifying the vehicle to be a prismatic body with a front surface Af , and then, it can be modeled by: 1 Faero 5 ρair UAf UCD Uv2 2

ð10:4Þ

where ρair is the air density in kg m23 (its typical value is 1.293 kg m23), Af is the vehicle equivalent front surface in m2 (typical value for a passenger car is between 1.5 m2 and 2 m2), CD is the drag coefficient (typical value for a passenger car is between 0.24 and 0.32), and v is the vehicle ground speed in m s21 (assuming wind speed is zero). The rolling resistance force can be modeled as: Froll 5 mv UgUCr Ucos ðαÞ

ð10:5Þ

286

Fuel Cells for Transportation

with: Cr 5 C0 1 C1 Uv

ð10:6Þ

where mv is vehicle mass in kg, g is the gravitational acceleration in m s22 (its value is 9.8 m s22), Cr is the rolling resistance coefficient (typical value for a passenger car is between 0.009 and 0.025), α is the slope rate of road in degree or radian, C0 and C1 are two empirical coefficients, and V is vehicle ground speed in m s21. The rolling resistance coefficient (Cr ) depends on many variables. The most important influencing quantities are vehicle speed ðvÞ, tire pressure ( p), and road surface conditions. The influence of the tire pffiffiffi pressure is approximately proportional to 1= p. A wet road can increase the rolling coefficient by 20%, and driving in extreme conditions can easily double that value [6]. In vehicle performance calculation, it is sufficient to consider the rolling resistance coefficient as a linear function of speed as described by Eq. (10.6). The grade resistance force, when driving on a road with a certain slope, opposes the forward motion (grade climbing) or helps the forward motion (grade descending). In vehicle performance analysis, this grade resistance force can be modeled by: Fgrad 5 mv UgUsin ðαÞ

ð10:7Þ

where mv is the vehicle mass in kg, g is the gravitational acceleration in m s22 (its value is 9.8 m s22), and α is the slope rate of road in degree or radian. As it has been mentioned previously, once the power traction (mechanical power) of the FCV is determined, then it could be converted into electric power required from various energy sources, in order to be able to design an adapted EMS of the hybrid powertrain. To be able to convert mechanical power into electric power, the overall powertrain efficiency needs to be known. The FCHEV electric power then can be expressed as: Pelec 5

Ptrac ηpowertrain

ð10:8Þ

where ηpowertrain represents the powertrain’s overall efficiency. The overall powertrain efficiency is given by: ηpowertrain 5 ηmech Uηinv Uηmotor

ð10:9Þ

where ηmech is the mechanical drivetrain (transmission) efficiency, ηinv is the DC/AC inverter efficiency, and ηmotor is the motor efficiency. Once the total electric power is determined, how to distribute this power need among energy sources in the hybrid powertrain will be decided by the EMS control strategy. If we take an example of a powertrain with two

Power demand for fuel cell system in hybrid vehicles Chapter | 10

287

energy sources (a fuel cell as main energy source and a battery pack), the electric power can be written as: Pelec ðtÞ 5 PFC ðtÞ 1 Pbatt ðtÞ

ð10:10Þ

where PFC is the power of fuel cell system, and Pbatt is the power of battery pack.

10.4 Fuel cell power demand from hybrid powertrain 10.4.1 The importance of energy management strategy When the total electrical power required for the FCHEV is deduced, the challenge will be how to allocate the power between the different energy sources. It is necessary to control the power flows between the different energy sources to guarantee the stable and high-efficiency operation of the system. This control implies the development of a suitable EMS. It should be noticed that a well-designed hybrid powertrain should take high energy density, high-power density, and optimized system performance into consideration, and only a reasonably designed EMS could be capable to deal with the high nonlinearity of the hybrid system and the strong coupling between multiple power sources. Usually, EMS can be regarded as a top-level control strategy, and its basic idea is to balance the electrical power demand between different power sources. More specifically, EMS can split the output power of the main power source and other power sources in the powertrain by designing power distribution rules or algorithms. It should be underlined that, for the same total electrical power requirement, based on the adopted EMS, the individual power profile of each energy source can be different. Therefore depending on the control objectives, the EMS can influence various parameters such as fuel consumption, the sizing of the energy sources, and the lifetime of the energy sources. The main objectives of the EMS of the hybrid powertrain are the optimization of the following criteria: hydrogen consumption, the mass and/or volume of the hybrid energy storage system (HESS), the dynamic performance of HESS, fuel cell lifetime, and the overall cost of the system. Based on the developed strategy, it is possible to optimize either one or several of these criteria simultaneously. Therefore a reasonably designed EMS can further optimize system performance based on the accurate matching of load power. Issues like high cost or short lifetime which are often considered as the core factors hindering fuel cell development can be optimized by designing the corresponding optimization objective function in the EMS. For example, the fuel cell can be vulnerable under high dynamic operating conditions, and the fast variation of the output power demand may lead to irreversible

288

Fuel Cells for Transportation

internal damage. A well-designed EMS can not only keep fuel cell output power within its high-efficiency range to reduce hydrogen consumption but also guarantee the fuel cell operation under lower stress to slow down its aging and degradation.

10.4.2 Major influential factors of fuel cell operation in vehicle applications Due to the high randomness and complexity of the vehicle’s operating environment and conditions, the design of a fuel cell hybrid powertrain needs to fully consider the load characteristics. Usually, the operating conditions of FCHEV include starting, steering, accelerating, deceleration, uphill, reversing, and normal driving. In the above-mentioned conditions, the load power demand may face instantaneous variation. For example, in the starting, accelerating, and uphill conditions, the load power will increase rapidly. If fuel cell is used as the main power source in the hybrid power supply system to provide the slow dynamic power, then the instantaneous component of the load power demand needs to be provided by the other power sources/energy storage devices. The main power source and other power sources can have a reference power output through the EMS, which guarantees a stable operation of the powertrain. Besides the instantaneous power, when the vehicle is operating in accelerating or uphill conditions, the load power demand presents the characteristic of a long-term high-power level. Therefore under both operating conditions, the hybrid power supply system needs to output the load power demand within the safe power range of the main power source. Meanwhile, other power sources/energy storage devices of the powertrain should be able to respond to the high dynamic load variation. Once the load power demand exceeds the maximum output power of the main power source, the EMS should be able to match the remaining part of the load power demand to other power sources. In addition, when the vehicle operates under regenerative brake mode, the EMS needs to rapidly distribute the energy absorption through the energy storage devices in the system and ensure the stability of the bus voltage. Finally, when FCHEV is operating under any special condition, EMS should also be able to match the load power and optimize system performance according to the specific situation. For example, when driving in a congested city or a rugged mountain road, frequent start
stop and steering conditions will occur. The EMS needs to be able to reasonably distribute the operating stress, caused by frequent changes from the load demand, to the power sources to avoid their rapid degradation, thus increasing the system lifetime. To summarize, Table 10.1 lists the requirements of FCHEV for EMS under different operating conditions. Besides the above-mentioned common passenger vehicle operating conditions, the EMS can also contribute to improving the performance of the system

Power demand for fuel cell system in hybrid vehicles Chapter | 10

289

TABLE 10.1 EMS for different FCHEV operating conditions. Load power demand characteristics

EMS demand

Instantaneous power

Utilize the fast response characteristics of other power sources to match up and reduce the operating stress of PEMFC

Long-term and/or highpower level

Matching load power with fuel cell and other power sources to improve system energy efficiency

Regenerative brake

Energy storage devices respond quickly to recover braking energy

for certain special use cases. For example, the system will not have significant load fluctuations under some operating conditions such as fixed-route buses. Therefore under such operating conditions, EMS can be designed to further improve overall efficiency and reduce energy consumption. When operating at different power levels, the efficiency of the power devices in the hybrid power supply system is significantly different. Since the power loss of the hybrid power supply system mainly includes power source energy conversion efficiency and power converter efficiency, the overall efficiency of the system has a complex and highly nonlinear feature. An appropriate EMS needs to be designed for energy consumption optimization. Another function of EMS is to optimize the degradation of the fuel cell. Research results have shown that the aging of fuel cells is significantly affected by the amplitude and frequency of variations of its output power. Since the operating conditions and driving conditions of the vehicle have strong random characteristics, a poor EMS design could add unnecessary operating stress to fuel cell. When the fuel cell operates under such undesirable conditions for a long time, the degradation will accelerate. Except for the above-mentioned overall system efficiency and system lifetime, the state of charge (SOC) of the energy storage devices should also be considered when designing the EMS. Thanks to the recent development of artificial intelligence technology, online optimization solutions can be realized for more complex and highly nonlinear cost functions. EMS with multiobjective optimization of hybrid power supply system would have more applications in the future.

10.4.3 State of the art of fuel cell hybrid electric vehicles energy management strategies As mentioned before, EMS is a core element of the fuel cell hybrid powertrain control. It is used to distribute the load power demand between the fuel

290

Fuel Cells for Transportation

cell and other power sources/energy storage devices. Based on previous studies, countless control strategies have emerged, such as initial linear programming, PID control, and state machine control. Besides, the novel dynamic programming technology, fuzzy logic control, model predictive control, and optimal control theory also show good performances in the applications. According to the principle, the EMSs can be divided into three categories as shown in Fig. 10.5: rule-based control, optimization-based method, and learning-based method.

10.4.3.1 Rule-based strategy The rule-based strategies mainly set the working state of the hybrid system by the operating rules of the vehicle. According to the rules used in the EMS, this type of strategy can furtherly be divided into deterministic rulebased and fuzzy logic control methods. The strategies based on deterministic rules are simple and easy to implement, while the strategies based on fuzzy logic have stronger robustness and higher adaptability. Generally speaking, rule-based strategies are easy to be applied with lighter calculation, whereas the determination of control parameters is usually based on experience. The dependence makes it difficult to achieve optimal control and is not suitable for dynamic conditions, which may lead to lower fuel economy and insufficient control effects. State machine control, as a classic deterministic rule-based strategy, divides problems into finite states with transitions between them according to the driving conditions of the FCHEV. The increasing number of states can improve the control effect and complexity simultaneously. Li et al. [7] designed a state machine strategy based on droop control to coordinate multiple power sources. Five states are defined in the strategy to distribute the power between fuel cell, battery, and supercapacitor through regulating unidirectional and bidirectional DC/DC converters. By evaluating a real driving cycle, the strategy could satisfy the power demand with fast response and improve the efficiency of the system. Meanwhile, the hydrogen consumption of the hybrid system is also decreased. Wang et al. [8] proposed EMS based on finite state machine control for fuel cell hybrid systems. The strategy fully considered the key parameters, such as the power capability and the battery SOC, and showed great performance of fuel economy and dynamic properties. Thermostat control, as another typical deterministic rule-based strategy, is suitable for real-time control because of its simple structure and fewer control parameters. The basic idea of the method is keeping the fuel cell operating at a constant point according to the battery SOC. Song et al. [9] proposed an optimized thermostat control strategy with driving pattern recognition. With the parameter optimization of genetic algorithm under specific driving conditions, the strategy can automatically switch the power

Power demand for fuel cell system in hybrid vehicles Chapter | 10

291

FIGURE 10.5 Different fuel cell hybrid powertrain EMS categories.

distribution mode under corresponding conditions. The simulation results show a better economic performance in dynamic conditions while maintaining the advantage of thermostat control. Meng et al. [10] designed a

292

Fuel Cells for Transportation

multimode EMS containing thermostat control and power follow method. By choosing the strategy according to the driving states, the vehicle achieves better efficiency. Meanwhile, the lifetime of fuel cell is improved by avoiding unnecessary start and stop. Since the hybrid system is often characterized by strong coupling and nonlinearity, state machine control usually becomes more complex when multipower sources are engaged. Therefore fuzzy logic control, as an abstraction of state machine control, has drawn increasing attention for lower model dependence. The classic fuzzy control is based on “If-Then” rules and membership functions. Nowadays, it is even extended to predictive fuzzy control, adaptive fuzzy control, and wavelet transform fuzzy control. Shen et al. [11] formulated a fuzzy logic EMS to protect the fuel cell and ensure efficiency. The strategy makes the fuel cell to be operated within highefficiency range and can generate an incremental power output within the affordable power slope. The experimental results indicated that the incremental fuzzy logic EMS smoothed the fuel cell power efficiently. Optimizing the parameter by backtracking search algorithm and sequential dynamic programming, Dao et al. [12] presented an EMS based on fuzzy logic control for a fuel cell hybrid excavator system. The strategy not only improves hydrogen economy and vehicle performance but also enhances the chargesustaining capability of auxiliary power units. To improve the fuel economy, Li et al. [13] optimized the fuzzy membership function by particle swarm algorithm. Compared with traditional fuzzy logic-based strategies, the proposed method reduces hydrogen consumption significantly. To furtherly study the effects of different rule-based strategies, researchers have conducted more comparative studies. In these works, PI-based EMS is often mentioned as another classical rule-based strategy. PI strategies are usually realized by controlling key parameters, such as battery SOC, and do not rely on expert experience as deterministic rule-based strategies. Soumeur et al. [14] compared the classical PI, fuzzy logic, state machine, and frequency decoupling schemes for FCHEV and found that each strategy has a different effect. Fig. 10.6 describes the evolution of the required power of the hybrid system and each power resource, including fuel cell, battery, and supercapacitor. As can be seen from the figure, fuel cell always provides the main power, and the rest power can be supplied by energy storage devices. Because frequency decoupling strategy can minimize the power overshoot of fuel cells, it is better for fuel cell autonomy than fuzzy control. Li et al. [15] divided the fuzzy subset into seven parts in the EMS. Compared with power following control, the strategy based on fuzzy logic is better in both fuel economy and operating efficiency.

10.4.3.2 Optimization-based strategy For fuel cell hybrid systems, optimization-based strategies aim to find the global or local optimal solution of the system by solving the minimum value

Power demand for fuel cell system in hybrid vehicles Chapter | 10

293

FIGURE 10.6 Electric vehicle power evolution during all trajectory. (A) Classical PI strategy. (B) Fuzzy logic strategy. (C) Frequency decoupling strategy. (D) State machine strategy. Credit: Comparative study of energy management strategies for hybrid proton exchange membrane fuel cell four wheel drive electric vehicle, J. Power Sources 462 (2020) 228167. https://doi.org/ 10.1016/j.jpowsour.2020.228167.

of the objective function under constrained conditions. It generally focuses on calculating the optimal power point of the fuel cell, thereby greatly reducing the hydrogen consumption of the hybrid system. As a typical global optimization-based strategy, dynamic programming is the most commonly used technique for optimal power allocation in FCHEV. It usually divides the optimization problem into a series of subproblems and then calculates the cost function at each discrete time step. Finally, the strategy will get the path with the minimum cost for each step, and global optimization will be obtained. To allocate the power for a fuel cell hybrid system, Hou et al. [16] proposed a dynamically efficient EMS based on dynamic programming. The EMS is quantitatively analyzed, and the simulation results indicated that the strategy can meet the power demand and reduce the calculation time simultaneously. Then, aiming to optimize the hydrogen economy and system efficiency of an FCHEV, Hou et al. [17] proposed another EMS based on a bi-loop dynamic programming strategy, which analyzed the influence of the different discrete steps of state variables. A discrete step, which can guarantee accuracy and reduce computational time, is selected. Compared with approaches based on convex programming, the strategy based on dynamic programming could get a better result. Wang et al. [18] compared some rule-based energy EMSs with the dynamic programming algorithm and found that rule-based strategies could achieve better

294

Fuel Cells for Transportation

economy and operating efficiency when compared with the dynamic programming algorithm. As another global optimization-based strategy, Pontryagin minimum principle (PMP) can determine an optimal allocation strategy under a known driving cycle for the hybrid powertrain as well. The algorithm is suitable for dynamic control systems and especially applicable to constrained systems. Compared with the strategy based on dynamic programming, the strategy based on PMP transforms a global optimization problem into an instantaneous optimization problem by introducing a costate. Li et al. [19] proposed an EMS based on PMP with the appropriate capacity configuration results obtained by the particle swarm optimization algorithm. The simulation results indicated that the strategy can reduce the cost efficiently. However, the initial value of the costate is related to prior knowledge of the driving cycles, which limits the online application of the method. Song et al. [20] provide an online costate updating method for uncertain driving cycles to realize real-time control. By incorporating a fuel cell power variation limiting factor with a weight coefficient into the PMP to suppress power changes, the durability of the fuel cell can be improved. Furthermore, the simulation results indicate that the operating cost can be effectively reduced. The equivalent consumption minimization strategy (ECMS) is representative of instantaneous optimization. It can transform one global optimization problem into instantaneous one and then distribute power for the fuel cell hybrid system instantaneously. For a fuel cell/battery/supercapacitor hybrid system, Fu et al. [21] proposed an EMS based on ECMS to improve the fuel cell lifetime and the fuel economy. Using ECMS and the adaptive low-pass filter method, the system could guarantee the fuel cell operation within a highly efficient range, and the hydrogen consumption can thus be reduced. Torreglosa et al. [22] proposed the ECMS based on limiting parameters, such as battery SOC, bus voltage, and fuel cell power. The experimental result shows that ECMS can effectively reduce hydrogen consumption when compared with fuzzy control-based strategies, and can achieve better control performance than state machine control. Considering the degradation of fuel cell, battery, and supercapacitor, Li et al. [23] proposed an adaptive strategy based on ECMS for the FCHEV. Compared with rule-based methods and traditional ECMS, the proposed strategy improved by sequential-based programming approach could reduce hydrogen consumption and enhance the durability of fuel cell. Hegazy et al. [24] made another comparative study between nine different EMSs. Ranked according to the performance of the strategies, ECMS based on mine blast algorithm contributes the most to the fuel cell hybrid system than others in hydrogen consumption and dynamic response. The strategies mentioned above have been widely applied in fuel cell hybrid powertrain systems. While the rule-based methods are easy to implement and have less computational burden, the rules rely on expert experience

Power demand for fuel cell system in hybrid vehicles Chapter | 10

295

too much and cannot provide an optimal solution for energy allocation. The methods based on dynamic programming are suitable for global optimization problems, while the algorithm requires predefined profiles and is not suitable for real-time conditions. Aiming at instantaneous optimization problems, the strategies based on theories, such as ECMS and model prediction control, show great performance. The ECMS could transfer the global optimization problem into a local problem. Thus, the system is allowed to provide real-time solutions to allocate energy between different sources. However, because these instantaneous optimization-based EMSs consider more about the instantaneous condition, the results are generally based on a local optimum. The global optimization of hybrid systems is sometimes hard to be obtained. To address the problems of the classic strategies mentioned above, innovative methods based on learning algorithms have received great attention in recent years.

10.4.3.3 Learning-based strategy Since most of the current EMSs are based on predictive algorithms or rules relying on expert experience, they have the disadvantage of poor adaptability to driving conditions. The learning-based strategy could overcome the weakness by using a large data set of real-time and historical information to obtain optimal control. Besides, it could also provide model-free control for the hybrid system. Neural network is one of the most practical models for adaptive optimization. Pedro et al. [25] used neural network to manage the power fluxes and minimize the equivalent energy consumption. The simulation result shows that the proposed strategy could save energy up to 18%. Neural network can also be used to predict the power demand for the hybrid vehicle. Compared with the method based on Markov chain, neural network proposed by Li et al. [26] provides better overall performance. It is not only suitable for power requirements prediction but also suitable for driving pattern reorganization. For further optimization of the method, neural network could also be combined with other algorithms. Genetic algorithm is a programming technique which searches for the solution to the problem by mimicking the evolution process. Since the algorithm has great global search performance and low computing complexity, it is suitable for energy management optimization. Combining genetic algorithm and neural network could also gain great performance in energy management. In order to improve the durability of the fuel cell, which is mainly reduced by frequent start
stop operation, Min et al. [27] proposed the strategy based on the neural network which has been optimized by genetic algorithm. The method uses the simulation results of dynamic programming as the training set, and the genetic algorithm can reselect the training data for the neural network. This novel energy management is confirmed to have a lower hydrogen consumption and extended fuel cell durability due to reduced frequent start
stop.

296

Fuel Cells for Transportation

As a popular learning-based method in artificial intelligence, EMS based on reinforcement learning could tackle partial disadvantages of conventional EMS and gains increasing attention in the field. The algorithm is developed for the problems built on the Markov decision process which assuming the future states are only affected by the current state. Reddy et al. [28] developed an intelligent strategy based on reinforcement learning for an FCHEV and used NEDC as the driving cycle. The simulation results show that the novel method not only improves the system efficiency but also extends the lifetime of the battery. Sun et al. [29] also employed reinforcement learning algorithm for a fuel cell hybrid system which uses battery and supercapacitor as auxiliary power units. The proposed strategy used fuzzy filter to improve the reinforcement learning algorithm and combined ECMS to optimize the fuel cell lifetime and hydrogen economy. Compared with other conventional strategies, these learning-based strategies offer several advantages, such as less computational time, lower model complexity, and higher accuracy. However, problems can include long learning time and the requirements of data preparation, which hinder their penetration into large-scale commercial applications. For this type of strategy, several algorithms, such as temporal Monte Carlo, Q-learning, and some others, could be implemented with different advantages and disadvantages. For example, deep Q-learning, which combines the Q-learning algorithm and deep learning, uses a neural network instead of the lookup table to overcome the instability of traditional Q-learning. Implementing deep Qlearning in fuel cell hybrid systems is possible to improve fuel economy [30]. In summary, the EMS is vital to developing eco-friendly and economic powertrain. Thus, in addition to realizing the accurate load matching, the EMS also needs to keep the system away from undesirable conditions. The presented different EMSs in this section have their own advantages and disadvantages. For rule-based EMS, it may be difficult to optimize system efficiency and lifetime by designing state rules. Although optimization-based EMS is more effective than rule-based EMS in improving system performance, complex optimization target models and constraints will cause a computational burden which affects system online performance. Learning-based strategies have great performance in calculating time, optimization effect, etc. However, this type of method still has problems such as large data demand and computational complexity. Considering that it is difficult to have a positive impact on all critical performance indicators in the system with a single EMS structure, the EMS of the hybrid power supply system can refer to a hierarchical method to achieve a more stable and more efficient operation of the system with long lifetime.

10.5 A case study for the fuel cell hybrid electric vehicles energy management strategy To illustrate how EMS works in the fuel cell hybrid power system, here we implement a simple EMS for FCHEV based on fuzzy logic control.

Power demand for fuel cell system in hybrid vehicles Chapter | 10

297

The considered model in this case study is based on a series architecture of FCHEV (the same as the vehicle model built in Section 10.3). Specifically, a fuel cell and a battery are employed to provide power for the vehicle, and the EMS is designed to allocate the power between the fuel cell and the battery. As shown in Fig. 10.7, to calculate the total power demand of the powertrain, the WLTC driving cycle is used as input parameter of the FCHEV. The driving cycle contains four different regions: low speed, medium speed, high speed, and extra high speed. More detailed information about the cycle is found in Section 10.2. From the driving cycle and the vehicle body model, the electrical power demand can be deduced, and then, it will be used as the input for the EMS. Table 10.2 summarizes the parameters used in the model. To satisfy the power demand, the fuel cell provides the main power (PFC ) for the hybrid system, while the battery provides supplemental power (PB )

FIGURE 10.7 Screenshot of system module in Simulink.

TABLE 10.2 FCHEV model parameters. Parameters

Numerical value

Vehicle total weight, mv

1755 kg

Vehicle front surface, Af

2.214 m2

Vehicle drag coefficient, CD

0.2567

Fuel cell system rated output power, PFC

30 kW

H2 initial mass onboard

4 kg

Battery charging/discharging efficiency

0.91

Battery initial SOC

0.8

Battery capacity

20 kWh

Air density

1.293 kg m23

Empirical coefficient, C0

0.01

Empirical coefficient, C1

3.6 3 1024

298

Fuel Cells for Transportation

during large load variation and recovers the regenerative power during the braking phase. A simple fuzzy logic control is implemented here in MATLABs/Simulink. The total power demand and the SOC of the battery are chosen as two input variables for fuzzy logic controller, and the output variable is the output power distribution coefficient of the fuel cell system. According to the formula PDemand 5 PB 1 PFC , the power distribution coefficient α could be calculated by the following formula: α 5 PFC =PDemand 1 2 α 5 PB =PDemand

ð10:11Þ

The membership functions and the rules of the fuzzy logic controller then need to be designed. The fuzzy subset of the power demand is set to {EL, L, ML, MH, H, EH}, representing extremely low, low, medium-low, mediumhigh, high, and extremely high, respectively. The fuzzy subset of battery SOC is set as {L, M, H}, representing low, medium, and high, respectively. The fuzzy subset of the power distribution coefficient α is set as {VL, L, M, H, VH}, which represent extremely low, low, medium, high, and extremely high, respectively. The fuzzy logic controller adopts the IF-THEN rule to map the inputs and the output. The detailed rules are listed in Table 10.3. We can adjust the default curve parameters according to the demand interval of SOC within [0.750.8]. In this case, the parameter values of curves L, M, and H are [0.425 58.48 0.28], [0.044 5.503 0.751], and [0.222 31.73 1.02], respectively. In the membership function of output, the range of the required power is normalized to [0, 1] according to the maximum and minimum values, and the parameters of each curve are [0.225 2.5 0], [0.225 2.5 0.45], [0.299 2.5 0.9652], [0.298 2.5 1.473], and [0.225 2.5 1.8], respectively. Besides, to make the fuel cell provide more output power to charge the battery when the SOC is too low, we can adjust the output power distribution coefficient α as:  αfinal 5 α; SOC . 0:75 ð10:12Þ αfinal 5 α 1 5; SOC # 0:75

TABLE 10.3 Detailed rules for the fuzzy EMS. α

SOC

Power demand EL

L

ML

MH

H

EH

L

H

H

VH

VH

VH

VH

M

H

H

H

H

H

VH

H

VL

L

M

M

M

M

Power demand for fuel cell system in hybrid vehicles Chapter | 10

299

Based on the EMS design above, the MATLAB/Simulink diagram of the three-dimensional surface plot of the controller output is shown in Fig. 10.8. The simulation results for the vehicle under WLTC driving cycle are shown in Fig. 10.9. Specifically, the initial SOC is set to 0.7 to test the performance when the SOC is out of the desired range [0.75, 0.8]. It can be seen from Fig. 10.9 that the EMS can allocate effectively the power demand during the whole cycle from 0 to 1800 seconds. In the beginning, the power demand is relatively low, while the fuel cell provides the power demand for

1.6

FCPower

1.4 1.2 1 0.8 0.6 0 1

0.8

0.6

0.4

0.5 0.2

DemandPower

0

1

SOC

FIGURE 10.8 3D diagram of fuzzy logic strategy.

FIGURE 10.9 Simulation results for the vehicle under WLTC driving cycle.

300

Fuel Cells for Transportation

the vehicle during most of the time and also the power needed to charge the battery. From 0 to 4208 seconds, the SOC increases from 0.7 to 0.75, and also from 420 to 1800 seconds, while it increases with a smaller slope until 1500 seconds when the power demand becomes higher. When the vehicle power demand is higher than the nominal output power of the fuel cell, the battery starts to provide additional power to meet the demand. It can also be noticed that the battery can be recharged by the regeneration braking when the vehicle decelerates. From the enlarged medium part [589, 1022 seconds] shown in Fig. 10.10, we can notice that the fuel cell has a relatively low dynamic power profile. When the power demand increases suddenly, the battery can output the instantaneous power to compensate the fuel cell output. When the power demand decreases suddenly, the battery can be recharged by the extra power provided by the fuel cell. In this way, the fuel cell and battery lifespan can be optimized in addition to improved system efficiency. In summary, the adopted strategy could integrate the power demand and SOC to realize the power distribution of the system according to the functional characteristics of fuel cell and battery. In terms of application, the control process is simple, and the results are satisfactory. However, the design of the fuzzy controller has a great influence on the results. Therefore exploring the precise membership function and establishing reasonable fuzzy rules are the keys to this control strategy.

FIGURE 10.10 Simulation results of the enlarged medium part [589, 1022 second].

Power demand for fuel cell system in hybrid vehicles Chapter | 10

301

10.6 Conclusion FCHEV represents a viable solution for transport sector decarbonization. In past decades, intensive research has been conducted in the field. Today, commercial FCHEVs are already available on the market. However, some challenges remain to be solved in order to make FCHEV competitive with conventional vehicles in terms of cost and durability. A fuel cell hybrid powertrain implies more than one energy source; thus the control of power flow between different energy sources (fuel cell, batteries, supercapacitors, etc.) and the drive train is quite challenging. This task could be completed by an EMS that must engage the right source at the right time to meet the load power demand. Recently, a lot of research has been done regarding the energy management strategies of hybrid powertrains. Each EMS has its own advantages and drawbacks, and no general or best EMS algorithm exists. The EMS of FCHEV must be carefully designed in function of the optimization objectives. For the same power profile, based on the designed EMS, the power allocation between different energy sources of the hybrid powertrain can be different, which has a direct impact on hydrogen consumption and energy sources’ lifetime and cost. To conclude, this chapter gives a general idea of how the total power demand profile of a FCHEV can be determined and how this power profile can be distributed between different power sources. For a better understanding of the readers, a case study of a simple EMS based on fuzzy logic control for FCHEV is demonstrated.

References [1] Y. Cui, H. Xu, F. Zou, et al., Optimization based method to develop representative driving cycle for real-world fuel consumption estimation, Energy 235 (2021) 121434. Available from: https://doi.org/10.1016/j.energy.2021.121434. [2] A.H.U. Bhatti, S.A.A. Kazmi, A. Tariq, et al., Development and analysis of electric vehicle driving cycle for hilly urban areas, Transp. Res. Part. D: Transp. Environ. 99 (2021) 103025. Available from: https://doi.org/10.1016/j.trd.2021.103025. [3] M. Tutuianu, P. Bonnel, B. Ciuffo, et al., Development of the World-wide harmonized Light duty Test Cycle (WLTC) and a possible pathway for its introduction in the European legislation, Transp. Res. Part. D: Transp. Environ. 40 (2015) 61
75. Available from: https://doi.org/10.1016/j.trd.2015.07.011. [4] E.G. Giakoumis, A.T. Zachiotis, Comparative evaluation of eight legislated driving schedules in terms of cycle metrics and emissions from a diesel-powered turbocharged van, Transp. Res. Part. D: Transp. Environ. 58 (2018) 139
154. Available from: https://doi.org/ 10.1016/j.trd.2017.11.002. [5] M. Andre´, The ARTEMIS European driving cycles for measuring car pollutant emissions, Sci. Total Environ. 334
335 (2004) 73
84. Available from: https://doi.org/10.1016/j. scitotenv.2004. [6] L. Guzzella, A. Sciarretta, Vehicle propulsion systems: introduction to modeling and optimization, 2007. https://doi.org/10.1007/978-3-642-35913-2.

302

Fuel Cells for Transportation

[7] Q. Li, H. Yang, Y. Han, et al., A state machine strategy based on droop control for an energy management system of PEMFC-battery-supercapacitor hybrid tramway, Int. J. Hydrog. Energy 41 (36) (2016) 16148
16159. Available from: https://doi.org/10.1016/j. ijhydene.2016.04.254. [8] Y. Wang, Z. Sun, Z. Chen, Energy management strategy for battery/supercapacitor/fuel cell hybrid source vehicles based on finite state machine, Appl. Energy 254 (2019) 113707. Available from: https://doi.org/10.1016/j.apenergy.2019.113707. [9] K. Song, F. Li, X. Hu, L. He, W. Niu, S. Lu, et al., Multi-mode energy management strategy for fuel cell electric vehicles based on driving pattern identification using learning vector quantization neural network algorithm, J. Power Sources 389 (2018) 230
239. Available from: https://doi.org/10.1016/j.jpowsour.2018.04.024. [10] X. Meng, D. Hao, R. Wang et al. Research on the energy management strategy of range extended fuel cell electric vehicle, in: Proceedings of the IEEE Fourth Information Technology, Networking, Electronic and Automation Control Conference (ITNEC), 2020, pp. 503
507. Available from: https://doi.org/10.1109/ITNEC48623.2020.9084694. [11] Y. Shen, P. Cui, X. Wang, et al., Variable structure battery-based fuel cell hybrid power system and its incremental fuzzy logic energy management strategy, Int. J. Hydrog. Energy 45 (21) (2020) 12130
12142. Available from: https://doi.org/10.1016/j.ijhydene. 2020.02.083. [12] H.V. Dao, X.D. To, H.V.A. Truong, T.C. Do, C.M. Ho, T.D. Dang, et al., Optimizationbased fuzzy energy management strategy for PEM fuel cell/battery/supercapacitor hybrid construction excavator, Int. J. Precis. Eng. Manuf.-Green Technol. 8 (2021) 1267
1285. Available from: https://doi.org/10.1007/s40684-020-00262-y. [13] W. Li, G. Feng, S. Jia, Research on multi-energy management system of fuel cell vehicle based on fuzzy control, J. Intell. Fuzzy Syst. (2020) 1
13. Available from: https://doi.org/ 10.3233/jifs-189458. [14] M.A. Soumeur, B. Gasbaoui, O. Abdelkhalek, J. Ghouili, T. Toumi, A. Chakar, Comparative study of energy management strategies for hybrid proton exchange membrane fuel cell four wheel drive electric vehicle, J. Power Sources 462 (2020) 228167. Available from: https://doi.org/10.1016/j.jpowsour.2020.228167. [15] D. Li, B. Xu, J. Tian, Z. Ma, Energy management strategy for fuel cell and battery hybrid vehicle based on fuzzy logic, Processes 8 (8) (2020) 882. Available from: https://doi.org/ 10.3390/pr8080882. [16] S. Hou, B. Dong, Y. Zhang, et al. Dynamic programming algorithm for energy management strategy of the fuel cell vehicle, in: Proceedings of the Third Conference on Vehicle Control and Intelligence (CVCI), 2019. Available from: https://doi.org/10.1109/ cvci47823.2019.8951723. [17] S. Hou, J. Gao, Y. Zhang, et al., A comparison study of battery size optimization and an energy management strategy for FCHEVs based on dynamic programming and convex programming, Int. J. Hydrog. Energy 45 (41) (2020) 21858
21872. Available from: https://doi.org/10.1016/j.ijhydene.2020.05.248. [18] Y. Wang, Z. Sun, X. Li, et al., A comparative study of power allocation strategies used in fuel cell and ultracapacitor hybrid systems, Energy 189 (2019) 116142. Available from: https://doi.org/10.1016/j.energy.2019.116142. [19] F. Li, Z. Yang, Y. Wang, et al., Energy management strategy of tram with hybrid energy storage system based on Pontryagin’s minimum principle, Trans. China Electrotech. Soc. 34 (S2) (2019) 752
759. Available from: https://doi.org/10.19595/j.cnki.1000-6753.tces. L80191.

Power demand for fuel cell system in hybrid vehicles Chapter | 10

303

[20] K. Song, X. Wang, F. Li, M. Sorrentino, B. Zheng, Pontryagin’s minimum principlebased real-time energy management strategy for fuel cell hybrid electric vehicle considering both fuel economy and power source durability, Energy 205 (2020) 118064. Available from: https://doi.org/10.1016/j.energy.2020.118064. [21] Z. Fu, Z. Li, P. Si, F. Tao, A hierarchical energy management strategy for fuel cell/battery/supercapacitor hybrid electric vehicles, Int. J. Hydrog. Energy 44 (39) (2019) 22146
22159. Available from: https://doi.org/10.1016/j.ijhydene.2019.06.158. [22] J. Torreglosa, Francisco Jurado, P. Garcia-Trivin˜o, Luis Ferna´ndez-Ramı´rez, Hybrid fuel cell and battery tramway control based on an equivalent consumption minimization strategy, Control. Eng. Pract. 19 (2011) 1182
1194. Available from: https://doi.org/10.1016/j. conengprac.2011.06.008. [23] H. Li, A. Ravey, A. N’Diaye, A. Djerdir, Online adaptive equivalent consumption minimization strategy for fuel cell hybrid electric vehicle considering power sources degradation, Energy Convers. Manag. 192 (2019) 133
149. Available from: https://doi.org/10.1016/j. enconman.2019.03.090. [24] H. Rezk, A.M. Nassef, M.A. Abdelkareem, A.H. Alami, A. Fathy, Comparison among various energy management strategies for reducing hydrogen consumption in a hybrid fuel cell/supercapacitor/battery system, Int. J. Hydrog. Energy 46 (8) (2021) 6110
6126. Available from: https://doi.org/10.1016/j.ijhydene.2019.11.195. [25] P.M. Mun˜oz, G. Correa, M.E. Gaudiano, D. Ferna´ndez, Energy management control design for fuel cell hybrid electric vehicles using neural networks, Int. J. Hydrog. Energy 42 (48) (2017) 28932
28944. Available from: https://doi.org/10.1016/j.ijhydene.2017.09.169. [26] T. Li, H. Liu, D. Ding, Predictive energy management of fuel cell supercapacitor hybrid construction equipment, Energy 149 (2018) 718
729. Available from: https://doi.org/ 10.1016/j.energy.2018.02.101. [27] D. Min, Z. Song, H. Chen, et al., Genetic algorithm optimized neural network based fuel cell hybrid electric vehicle energy management strategy under start-stop condition, Appl. Energy 306 (2022) 118036. Available from: https://doi.org/10.1016/j.apenergy.2021.118036. [28] N.P. Reddy, D. Pasdeloup, M.K. Zadeh, et al. An intelligent power and energy management system for fuel cell/battery hybrid electric vehicle using reinforcement learning, in: Proceedings of the IEEE Transportation Electrification Conference and Expo (ITEC), 2019, pp. 1
6. Available from: https://doi.org/10.1109/itec.2019.8790451. [29] H. Sun, Z. Fu, F. Tao, et al., Data-driven reinforcement-learning-based hierarchical energy management strategy for fuel cell/battery/ultracapacitor hybrid electric vehicles, J. Power Sources 455 (2020) 227964. Available from: https://doi.org/10.1016/j.jpowsour.2020.227964. [30] X. Ma, Y. Xie, C. Chigan, Meta-deep Q-learning for eco-routing, in: Proceedings of the IEEE Second Connected and Automated Vehicles Symposium (CAVS), 2019, pp. 1
5. https://doi.org/10.1109/cavs.2019.8887764.

This page intentionally left blank

Chapter 11

Bipolar plates and flow field design Xianguo Li Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada

11.1 Introduction Proton exchange membrane (PEM) fuel cell has been under intensive development in the past several decades and has reached the early stage of commercial deployment. It is now an integral part of key strategies for climate change mitigation and postpandemic economic recovery, especially for the hard-to-decarbonize sectors like transport, including ground, water-surface and air transports. For a single fuel cell unit, hydrogen is supplied to the anode side of a PEM fuel cell via an anode flow plate, while air is supplied to the cathode side via a cathode flow plate. Both flow plates have flow fields (or channels) guiding the flow and distribution of the reactant gases. The electrochemical oxidation reaction of hydrogen at the anode electrode and reduction reaction of oxygen at the cathode electrode provide electric energy output along with the production of waste heat and edible quality of water. The anode and cathode electrodes sandwich a proton exchange (or conducting) membrane as the electrolyte, forming the so-called membrane electrode assembly (MEA), as shown in Fig. 11.1. The anode and cathode electrodes are porous, composed of typically three different layers, commonly referred to as gas diffusion layer (GDL), microporous layer, and catalyst layer (CL), respectively. The output electrical potential between the cathode and anode electrode of a single-cell unit is typically in the range of 0.6
0.85 V, while the output electrical current can be as much as over 2 A cm22 of the flat electrode area. Thus, the electrical power output from each single-cell unit is limited. For practical applications, often several hundred single-cell units are connected in series to boost the overall output voltage and hence the electrical power. In order to provide reactant gases (hydrogen and air) to each MEA and connect MEAs in electrically series arrangement, bipolar plates are used Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00003-4 © 2023 Elsevier Ltd. All rights reserved.

305

306

Fuel Cells for Transportation

FIGURE 11.1 Schematic of a single fuel cell unit including an anode flow plate, a cathode flow plate, and a membrane electrode assembly (MEA). AE, Anode electrode; AFP, anode flow plate; CFP, cathode flow plate; CE, cathode electrode; I, electrical current (in the direction opposite to the direction of electron motion), Q, waste heat; We, electrical energy.

between the MEAs, such that one side of an individual bipolar plate has a pattern of flow channels providing hydrogen gas flow for distribution over the anode electrode surface of the neighboring MEA, while the other side of the same bipolar plate has a similar or different pattern of flow channels providing air flow for distribution over the cathode electrode surface of another neighboring MEA. That is, a bipolar plate is simply the combined version of an anode flow plate and a cathode flow plate, hence its name. Also, a cooling flow pattern might be integrated with the bipolar plate for thermal management. MEAs connected together with the bipolar plate arrangement form a stack of fuel cells providing output electrical power required for practical applications. A schematic of the stack arrangement with bipolar plates and MEAs is shown in Fig. 11.2 along with the cooling arrangement. Fig. 11.2 also indicates that hydrogen and oxygen gases in their respective flow channels transport to their respective porous electrode area both facing the flow channels and the land between the channels. Therefore, a large flow channel size is beneficial for better reactant gas transport to the electrode and onto the CL for electrochemical reaction. On the other hand, electrical current, I, transports from the electrode adjacent to both the flow channels and the land between the channels of one MEA to the land between the channels and onto the neighboring MEA surface that is both facing the land and the channel. For this purpose of the current transport, a large land size is beneficial to minimize ohmic resistance losses in the components.

Bipolar plates and flow field design Chapter | 11

307

FIGURE 11.2 Schematic of a cross-sectional view of bipolar plate with cooling flow channel sandwiched between the anode and cathode flow fields (A), a face view of the flow channel (B), and fuel cell stack building block, or repeating unit (C).

Thus, a practical bipolar plate must balance the needs for both reactant gas supply and current transport. Liquid water produced in the cathode reaction is often in the liquid form because the PEM fuel cell operation temperature is low, less than 100 C. Product water is removed via the reactant gas flow in the channels. However, liquid water is often trapped in the porous electrode region adjacent to the land area for two reasons: the porous electrode is compressed more facing the land area than the flow channel area, and the compressed electrode has smaller pore sizes, trapping water there due to increased capillary effect; further, electrode facing the land area has a weaker or no reactant gas flow to take the liquid water away. Although heat can be transported through the solid portion via conduction and reactant flow via convection, and be removed by the coolant flow in the cooling channels, the solid land has high thermal conductivity and is the main path for heat transport. Again, it is clear that optimization of bipolar plate design is essential to achieve good and reliable performance for fuel cell stacks. A major part of the bipolar plate design includes the design of flow fields, proper selection of materials with high electrical and thermal conductivity, along with other attributes, and cost-effective methods for their volume manufacturing to meet the largescale commercial requirement. This will form the main content of the present chapter.

11.2 Bipolar plates As shown in Fig. 11.2, the bipolar plate arrangement for current collection results in the interspersion of each MEA between two fluid-impermeable, electrically and thermally conductive plates, often also referred to as the anode and the cathode plates, respectively. These plates often have reactant flow channels, hence the name reactant flow field plates. The flow channels are often formed on both sides of the same plate: one side serves as the

308

Fuel Cells for Transportation

anode plate to one MEA, and the other side as the cathode plate to the adjacent MEA; thus, the plate is called bipolar (separator) plate. In this stack design, waste heat is removed through coolant flow in a separate cooling flow channel either built into the same bipolar plate or through a separate cooling plate, placed at every or every few bipolar plates. These plates collectively keep the fuel (hydrogen), oxidant (air), and cooling fluid apart, preventing them from mixing with each other and thereby ensuring safe operation and maintenance. In a fuel cell stack, the individual MEA is very thin, typically less than 0.5 mm thick, while bipolar plates are relatively much thicker, in the range from approximately tenths of one mm to several mm thick. Thus, bipolar plates can account up to 80% by weight, 50% by volume, and 40% by cost of PEM fuel cell stacks [1]; hence, a key component in the stacks that play an important role in achieving high performance, low cost, and long lifetime for fuel cells operating at high power densities, and significant reduction in the thickness of the bipolar plate is highly desired. This section describes the basic functionality and requirement for a bipolar plate in PEM fuel cell stacks.

11.2.1 Functions As pointed out earlier, the bipolar plate is one of the key components in fuel cell stacks and must perform a number of functions well simultaneously in order to achieve good stack performance and durability at low cost. The functionalities that a bipolar plate must provide include G

G G

G

G

Providing a current path normal to the plate’s major surface, or current collector; Providing reactant gas supply to the MEA surface, or reactant supplier; Providing separation of reactants between adjacent MEAs, or gas separator; Providing a means for the removal of reaction products, water and heat, or product removal; Providing structural support to the mechanically weak MEAs, or structural support.

To collect the electric current produced in each of the MEAs in a fuel cell stack, bipolar plates connect each MEA electrically to yield the electrical power output from the stack. The flow channels on each side of the plates provide hydrogen and air gas to the surface of each MEA, so that hydrogen and air gas can transport to the anode and cathode CL, respectively, for proper electrochemical reaction and electrical power generation. On the other hand, the bipolar plates separate the reactant gases in the flow channels on both of the plate surfaces from mixing, avoiding safety concerns and hazardous situations. The reaction byproducts, water and waste heat, are transported to the bipolar plates, and the reactant gas streams in the flow channels for

Bipolar plates and flow field design Chapter | 11

309

removal from the MEA surfaces, so that MEA surfaces can be regenerated for continual electrochemical reaction and uninterrupted power generation, or preventing reaction byproducts from accumulating inside the MEA structure. Otherwise, waterflooding and stack overheating will occur, accelerating the degradation in stack performance and durability. The thin MEAs with brittle carbon-based electrodes in the stack are supported by the bipolar plates for their structural integrity. The functionalities of the bipolar plates are met through an appropriate selection of plate materials and optimal designs of flow fields. Plate designs, or plate topologies, include G

G

G

the configuration of the plate, such as the shape, size, and thickness of the plate; the configuration of the flow fields, such as the orientation, shape, size, path layout, and coverage of the reactant flow fields; the similarity or difference between the anode and cathode flow fields; the configuration of the coolant flow fields, such as the orientation, shape, size, path layout, and coverage of the cooling flow channels.

The plate materials influence the plate designs and methods of fabrication and have a significant impact on the performance, durability, and cost of the plates. Therefore, a proper selection of materials along with optimal design must be developed for the bipolar plates, because the above plate functions have conflicting requirements on the bipolar plate design.

11.2.2 Requirements To make bipolar plates function properly in a stack environment, it is essential that the properties of the materials, the attributes of the designs, and the manufacturing processing for the bipolar plates must be all considered for optimization or balance against specific but conflicting requirements. For current collection, the bipolar plates must have high electronic conductivity, short path length and large cross-sectional area for electron transport, and low interfacial contact resistance, since there are many interfacial contacts, from hundreds to thousands of them between the bipolar plates and the neighboring MEAs, and within the bipolar plate structures, depending on the designs. For reactant supply to each MEA surface and further to the CL, the flow fields should be designed in such a manner that the reactant concentration over the MEA surface is uniform, or as uniform as possible. This may not be easy to accomplish because the reactant is consumed in the MEA (or CL); often, the reactant concentration is higher in the inlet region of the flow field, decreases along the flow channel due to the consumption of the reactant along the MEA surface, and becomes the lowest near the outlet region of the flow channels. Another challenge is that the reactant stream flows

310

Fuel Cells for Transportation

along the direction of the flow channel layout (convection), while the transport of reactant from the gas stream in the flow channel to the MEA surface and ultimately to the CL is transverse to this dominant flow direction, or by molecular diffusion. Mass transfer via diffusion is slow, because of the low numerical values of the diffusion coefficients. The flow field design should provide a convective motion toward the MEA surface in order to increase the rate of mass transfer for the reactant gas transport to the CL, and this is crucially important for stack performance at high current densities, or for developing high power density fuel cell stacks. Another aspect of the designs for the bipolar plates and flow fields is that the MEA surface next to the land region between the neighboring flow channels is not directly exposed to the reactant gas flow, or in practical operation, devoid of reactant gas supply. Further, this land region can be up to 50% of the active surface area of the MEAs. Therefore, it is imperative to develop flow field and bipolar plate designs in such a manner that proper reactant supply, similar to the region facing the flow channels, be developed for the uniform and enhanced utilization of the MEA surface area for maximum access by the reactant gas. To ensure that the reactant gases, hydrogen and air, are prevented from mixing, or crossover from the anode flow channel to the cathode flow channel through the bipolar plate, the plate must be solid and dense without gas permeation. Therefore, the plates must have sufficient thickness to avoid gas permeation, even for hydrogen having a small molecular size. This is required for safe operation and to avoid potentially hazardous situations from occurrence. However, in order to have a high power density on a per unit mass and volume basis, lightweight materials would be preferred for bipolar plates. Water in the operating fuel cell stack can be partly in the vapor form and partly in the liquid form, depending on the stack operating condition and the stack design. Liquid water has a high value of surface tension, or capillary force, and tends to stick or accumulate in the pore structure of the MEAs, resulting in water flooding and blocking the reactant gas from reaching the reaction sites in the CL. For effective product water removal, the flow channels must be designed such that water can be easily removed via the reactant gas stream in the flow channels. The removal in the form of water vapor tends to be more effective than in the liquid form, and liquid water removal needs to consider the surface wettability of the flow channels. This is because the configuration, shape, and transport of the liquid water in the flow channels can change substantially with the surface wettability condition of the flow channel walls. Designs should be implemented to remove liquid water entrapped in the electrode region adjacent to the lands between the channels. For waste heat removal or heat transfer, the bipolar plates should have high thermal conductivity, short path length and large cross-sectional area for heat transfer, and low interfacial contact resistance to heat transfer at the

Bipolar plates and flow field design Chapter | 11

311

many interfaces within a stack. However, heat transfer not only occurs through the solid contacts of the bipolar plates with the MEAs, heat is also transported via the reactant gas streams, or through the flow channel regions as well. Thermal management is critical to the stack operation and performance because the fuel cell is a 1 T heat engine thermodynamically; it requires a uniform temperature distribution throughout to achieve the best performance. Any nonuniformity in the stack temperature distribution tends to lower the stack performance. Further, local hot spots tend to increase the degradation of the stack components, including membrane and catalyst nanoparticles. In addition, bipolar plates must have sufficient ability for heat and electrical transport in both in-plane and through-plane directions such that each plate can be maintained for uniform temperature and electrical potential. This requires not only plate materials to have high thermal and electrical conductivity, but also sufficient thickness as well as the flow field design for efficient and effective heat and electrical transport. MEAs are very thin and flat, and electrode structures are very brittle and made of porous carbon materials. Bipolar plates provide support for structural integrity. Therefore, bipolar plates must be sufficiently thick or strong in mechanical strength to resist mechanical deformation, and its surfaces contacting the MEAs must be also flat and smooth to minimize the interfacial surface contact resistances and to provide uniform compression required for the stack assembly. Fabricating bipolar plate surfaces with high flatness and smoothness will result in a high cost of manufacturing these plates. Further, bipolar plates are exposed to the reactant gas streams of hydrogen and air with water vapor at a high temperature of nearly 100 C; this requires the bipolar plates to be chemically stable in the reducing and oxidizing environment and to resist hydrogen embrittlement as well. In addition, bipolar plates are also subject to different electrical potentials in both through-plane and in-plane directions; therefore, they should have sufficient resistance to corrosion and erosion as well for durability and performance degradation, as impurities and ions from the bipolar plate corrosion may contaminate the MEA operation and reduce its performance and lifetime.

11.3 Flow field design Flow field is the most important feature of a bipolar plate, and it has a significant impact on the performance, durability, and cost of a fuel cell stack. There are many flow field designs that have been proposed, patented, and investigated [1
4], and they can be classified in general as dead-ended and open-ended for the flow fields. A dead-ended design implies that the flow channel has an inlet, but without an outlet so that all the reactant gas coming into the flow channel will be consumed completely in the stack or individual cells. The advantage of this design includes a simplified flow supply and control, and minimal parasitic losses associated with the reactant gas supply

312

Fuel Cells for Transportation

during operation. The downside is that impurities in the reactant gas will accumulate in the stack, which will result in performance degradation; due to the impurity accumulation, a concentration gradient can occur from the flow channel inlet region to the region farthest away from the inlet, leading to additional concentration polarization. Therefore, except for special applications where pure reactant gases are used, dead-ended designs are rarely adopted. Most of the practical applications employ open-ended designs, and they are rich in literature. However, commonly they can be classified into two categories: flow field without and with guided flow path. The flow field without a guided flow path provides the least flow resistance, or the reactant gas will flow through a path that has minimal flow resistance. However, the reactant gas may not flow to the regions with high flow resistance, leading to nonuniform reactant gas distribution over the entire MEA surface. As such, this type of flow field is not used as often. Most of the flow field found in practice belongs to the type of design with guided flow path. Further, oxygen flow rate supplied to a fuel cell stack or individual cells in a typical PEM fuel cell operation is about twice as much as it is needed for stoichiometric electrochemical reaction in the stack or individual cells. This is equivalent to the oxygen utilization of 50% or often referred to as the stoichiometry of 2 for oxygen in PEM fuel cell literature. Since oxygen is about 21% in the air, the air flow rate into a stack is about 9.52 times the amount of oxygen needed for the stoichiometric electrochemical reaction; this results in a very high air flow rate, hence, excessive high parasitic power loss associated with air supply, which could be as high as over 30% of the stack output power. It is imperative to develop cathode flow field designs that would balance the optimal stack performance and minimized air supply power. Since today’s PEM fuel cells use high-purity hydrogen gas for the anode, and hydrogen gas has high molecular diffusivity for mass transport through the porous anode electrode, the anode flow field design is not as critical as the cathode. Therefore, without loss of generality the following flow field description would be intended for the cathode, while the anode flow field could be taken as identical to the cathode or further simplified without causing notable impacts on the stack operation and performance. However, the relative arrangement of anode and cathode flow would be important for stack operation, and this would be an important part of the stack design.

11.3.1 Flow field without guided flow path A representative flow field without guided flow path is illustrated in Fig. 11.3. It is seen that the flow network could have many flow paths formed by a regularly patterned array of pins, and these pins can be arranged in line or staggered. The pins can have any shape, but cubical and circular pins are more common. The spaces between the pins form the flow network,

Bipolar plates and flow field design Chapter | 11

313

FIGURE 11.3 A schematic of flow field without guided flow path. Adapted from X. Li, I. Sabir, Int. J. Hydrog. Energy 30 (2005) 359
371.

while the actual flow path will be complex, as shown in Fig. 11.3. The main flow path from the inlet to the exit will follow the path of least resistance, while the recirculation zone on either side of the main flow path will be developed. The recirculation zone will end up with lower reactant concentration due to reactant transport to the MEA for reaction, thus creating a large region of concentration disparity. Further, liquid water tends to accumulate in the region of the recirculation zone as well, further deteriorating the stack performance. Similar phenomena can occur for a smaller recirculation zone that may exist behind each pin. As a result, this type of flow field design without guided flow path tends to have a low reactant pressure drop from the inlet to the exit, hence a small parasitic loss associated with the reactant supply. However, stack performance is usually poor and may be plagued by waterflooding issues.

11.3.2 Flow field with guided flow path Flow field with guided flow path is the most commonly used design for distributing reactant gas to the MEA surface, as illustrated in Fig. 11.4. It typically involves flow channels either connected in parallel as shown in Fig. 11.4A or in series as shown in Fig. 11.4B; the latter is often referred to as serpentine flow channels.

FIGURE 11.4 A schematic of flow field with guided flow path. (A) Flow field with guided flow path in parallel connection (or parallel flow channel). (B) Flow field with guided flow path in series connection (serpentine flow channel). (C) Cross-sectional view of flow channels and lands; the shape can be rectangular, square, trapezoidal, semicircular, or others [5]. Adapted from A. Pollegri, P. M. Spaziante, US Patent No. 4,197,178 (1980).

Bipolar plates and flow field design Chapter | 11

315

11.3.2.1 Parallel flow channels The parallel flow channel design usually has an inlet manifold and an exhaust manifold to provide the distribution of the reactant gas into each parallel flow channel, and then to collect the reactant gas exhausted from the fuel cell. This design often results in a nonuniform mass flow distribution among the channels, higher for the channels in the regions near the flow inlet and far end from the flow inlet (or flow outlet). This nonuniform reaction concentration distribution would impact unfavorably on the cell/stack performance. More serious issues occur when a fuel cell is operated for an extended period of time, and product water formed at the cathode tends to accumulate on the cathode side of the flow channels. Liquid water has a high capillary effect and tends to stick to the walls of the flow channels, difficult to be removed or carried away by the reactant flow in the channels. Water tends to further accumulate and coalesce to form large droplets, which can increase the resistance to the reactant gas flow through the channels with large droplets. This is because the pressure differential across each parallel channel is identical, and this pressure differential is the driving force for fluid flow through the channels. As a result, reactant gas flows preferentially through the channels with the least water droplet obstruction (or resistance). This in turn will allow the channels with the least reactant gas flow to accumulate more water droplets, leading to channels with stagnant gas (no flow through). Experimentally, poor and unstable cell performance would be observed for parallel flow channel design. Fundamentally, if the pressure losses (drops) in the inlet and exhaust manifolds are far less than the pressure losses through the parallel channels, a uniform flow among the parallel channels can be achieved. In practice, either complex manifold or channel design modifications or both are used for this purpose, as will be detailed later. Another issue with the parallel flow channel design is that the MEA surface in contact with the land portion of the flow channel is poorly fed with the reactant gas. This is exacerbated by the compression force exerted by the channel land on the MEA, leading to lower porosity of the MEA in these regions for reactant gas transport. Thus, up to half of the MEA surface area may be underutilized for power generation due to a lack of reactant gas. Therefore, parallel flow channels have not been favored in practical fuel cell stacks. It might be mentioned that there are varieties of parallel flow channel designs that have been developed in the literature [2], including radial channels, center-to-edge channels, inclined channels creating variable channel cross sections, curved channels (such as in-phase and out-of-phase wavy channels), but their characteristics remain the same. 11.3.2.2 Serpentine flow channels To tackle the challenges observed for parallel flow channel design, channels connected in series have been developed [6], which are commonly referred

316

Fuel Cells for Transportation

to as serpentine flow channels, as shown in Fig. 11.4B. This flow channel design has one continuous flow path from the inlet to the outlet so that any liquid water reaching the flow channel can be forced out by the reactant gas flow that traverses through the entire active area of the MEA surface. No stagnant or recirculating flow exists. Experimentally good and stable cell performance can be observed. Serpentine flow channel design also provides the possibility of water removal from the cell reaction completely in the form of water vapor, avoiding liquid water blocking the flow channels and porous electrode structures as well. Considering ideal gas behavior for the reactant gas
water vapor mixture in the flow channel, the following relation exists: Pvap Pvap N_ vap 5 5 Pgas PT 2 Pvap N_ gas

ð11:1Þ

where N_ vap and N_ gas stand for, respectively, the molar flow rate of the water vapor and reactant gas in the reactant gas
water vapor mixture in the flow channel; similarly, Pvap and Pgas are the corresponding partial pressures for the water vapor and reactant gas, respectively. Because PT 5 Pvap 1 Pgas

ð11:2Þ

where PT represents the total pressure for the reactant gas
water vapor mixture in the flow channel. The total pressure along the flow channel is reduced due to the reactant gas transported into the MEA for electrochemical reaction, ΔPgas , as well as due to flow frictional losses, Pf , and minor losses due to the sudden changes in the local flow channel geometry, ΔPminor , or PT 5 PT;inlet 2 ΔPgas 2 ΔPf 2 ΔPminor

ð11:3Þ

where PT;inlet is the total pressure at the inlet of the flow channel. Substituting Eqs. (11.2) and (11.3) into Eq. (11.1) yields N_ vap Pvap 5 PT;inlet 2 ΔPgas 2 ΔPf 2 ΔPminor N_ gas

ð11:4Þ

Therefore, Eq. (11.4) indicates that total pressure loss along the flow channel enhances the water vapor flow rate for the same reactant gas stream, or pressure losses enhance the gas stream’s ability to take away more water vapor. Consequently, if a serpentine flow channel is designed properly, with appropriate frictional and minor losses along the flow channel, water in the MEA structure and flow channel can be removed entirely in the vapor form. That is, proper total pressure loss for a serpentine flow channel is desirable and beneficial for water removal. It might be pointed out that the pressure losses arising from the cell consumption of the reactant gas and frictional loss might be limited to the operation condition and the size and shape of the active area, while the minor losses can be easily instituted by sudden changes in the local geometry of the flow channel.

Bipolar plates and flow field design Chapter | 11

317

It might be mentioned that the reactant gas flow rate in the flow channel can be expressed as N_ gas 5 N_ gas;inlet 2 ΔN_ gas

ð11:5Þ

where ΔN_ gas stands for the rate of reactant gas that is consumed by the electrochemical reaction in the MEA from the flow channel inlet up to the location in the flow channel, and reactant gas consumption can lower the capability of the reactant gas stream in the flow channel to remove water in the vapor form. Since as mentioned earlier, the oxygen stoichiometry for the PEM fuel cell is typically large, in the order of 2, while 79% of the air is nitrogen which is not consumed in the fuel cell reaction; the air flow rate at the channel inlet is about 9.52 times the amount of oxygen needed for the stoichiometric electrochemical reaction in the MEA, or Stair ðJAÞ N_ gas;inlet 5 0:21ð4F Þ

ð11:6Þ

where Stair represents the stoichiometry for the reactant air, J the current density per unit MEA surface area, A the MEA surface area, and F 5 96,487 Coulomb/mole is the Faraday constant. It might be pointed out that the stoichiometry for air is the same as that for oxygen because only oxygen serves as oxidant for the electrochemical reaction in the cathode. That is, the reactant gas consumed on the cathode side is the oxygen in the air, or ΔN_ gas 5 ΔN_ oxygen 5

JA 4F

ð11:7Þ

Substitution of Eqs. (11.5)
(11.7) into Eq. (11.4) yields the rate of water vapor removal by the reactant gas stream in the flow channel:
 Pvap JA N_ vap 5 ð4:76Stair 2 1Þ ð11:8Þ 4F PT;inlet 2 ΔPgas 2 ΔPf 2 ΔPminor Therefore, the change in the reactant gas molar flow rate along the cathode flow channel is small due to the oxygen consumption in the MEA. However, this would be substantial for the reactant gas flow along the anode flow channel, because typically pure hydrogen is used as the anode gas, and the stoichiometry for hydrogen gas is normally small, in the order of 1.1
1.2, or up to 90% of the hydrogen would be consumed in the MEA reaction. This will make the design of the anode flow channel with complete removal of liquid water in the vapor form more challenging if a substantial amount of water is transported from the cathode to the anode via backdiffusion. This is however not a problem for PEM fuel cells using pure hydrogen, since hydrogen has much higher mass transport capability, and is much less impacted by waterflooding phenomena. A method of flow channel design for effective water removal through the flow channel is given in [7].

318

Fuel Cells for Transportation

When channels are connected in series, the reactant gas has a dominant flow along the flow channel, resulting in a pressure drop between the neighboring channels as shown in Fig. 11.5. This pressure differential between the corresponding locations in the neighboring channels will also drive the reactant gas to flow through the porous electrode in the MEA that is adjacent to the flow channels, as illustrated in Fig. 11.5B. This secondary flow creates a convection flow bringing the reactant gas toward the CL, such that the mass transport from the flow channel to the CL is now by the combined effect of convection and diffusion, enhancing the rate of mass transport. This secondary flow further transports the reactant gas to the region of the MEA in contact with the channel land, making the entire MEA with sufficient access to the reactant gas, compared with the parallel channel design described earlier. In addition, this secondary flow facilitates the removal of water entrapped in the porous MEA structure in contact with the channel land. Therefore, this induced cross-flow between the neighboring channels improves the fuel cell performance considerably [8
10]. On the other hand, the serpentine flow channel design has a very long reactant flow path for practical MEAs of rather large sizes, leading to significant pressure drops for the reactant gas flow from the channel inlet to the outlet. This has several negative effects. First, large pressure drop results in significant parasitic losses associated with the reactant gas supply, as much

FIGURE 11.5 Illustration of cross-channel leakage flow for a serpentine flow channel design. Adapted from T. Kanezaki, X. Li, J.J. Baschuk, J. Power Sources 162 (2006) 415
425.

Bipolar plates and flow field design Chapter | 11

319

as over 30% of the fuel cell stack output power can be consumed in the supply of air, for example. Second, significant reactant gas concentration changes occur along the flow channel from the inlet to the outlet. Further, water collection by the reactant gas stream along the single channel has the tendency of causing MEA dehydration in the inlet region of the channel and waterflooding near the exit region of the channel. These can reduce the MEA performance and durability considerably. Therefore, a single serpentine flow channel design is now commonly not employed in practice.

11.3.2.3 Interdigitated flow channels As discussed in the previous subsection, the advantage of the serpentine flow field design includes higher pressure drop aiding the removal of water in the form of vapor, creating cross-channel flow through the porous electrode adjacent to the land area between the channels, promoting mass transfer to the CL through a combined effect of convection and diffusion, and enhancing the removal of water in the compressed porous electrode structure next to the channel land. All these provide a more uniform and higher reactant concentration in the CL throughout the entire MEA. This effect of the serpentine flow field design can be further enhanced significantly by blocking each channel at the alternate end, as shown in Fig. 11.6, forming what is referred to as the interdigitated flow field design [11]. Essentially, the flow channels are dead-ended so that the main flow direction is through the porous electrode in contact with the flow channel, and the secondary crosschannel flow illustrated in Fig. 11.5 for the serpentine flow field design becomes the dominant flow for this design. Strong convection normal to the MEA surface enhances significantly the mass transfer to the CL, and the removal of water in the electrode structure. As a result, the “best” fuel cell performance can be achieved for this flow field design. However, the pressure loss, hence the associated parasitic power loss, for the reactant supply becomes prohibitively large for practical applications. This flow field design is usually used in laboratory testing to achieve the maximum MEA performance or the maximum potential of MEAs. 11.3.2.4 Strategies for improvement and hybridization of flow channel designs As described in the previous subsections, the three fundamental designs of flow field: parallel, series, and interdigitated, have respective advantages and disadvantages. To achieve optimal flow field designs, the above three fundamental designs can be modified in a variety of ways into various hybridized flow field designs to improve fuel cell performance and durability. It is to be emphasized that for practical applications, the flow field designs should be as simple as possible and easy for manufacturing for cost reduction while yielding optimal performance and durability.

320

Fuel Cells for Transportation

FIGURE 11.6 Schematic of interdigitated flow field design. Adapted from A. Kazim, H.T. Liu, P. Forges, J. Appl. Electrochem. 29 (1999) 1409
1416.

Parallel flow field design, as described earlier, has several shortcomings, including short channel length leading to low-pressure drop, ΔPchannel , in each individual channel from the inlet manifold to the exhaust manifold, compared to the pressure drop across the inlet and exhaust manifolds, ΔPmanifold . This low ΔPchannel results in nonuniform flow distribution of reactants among the parallel channels. Several approaches have been considered to address this shortcoming. For example, instead of a square active area, elongated or rectangular active area is now commonly adopted, so that the channels are arranged in the longer direction of the plate to have a longer flow path for increased pressure drop arising from the reactant gas consumption and frictional losses. A further approach is to use curved instead of straight flow channels, so that the flow path length can be further increased, with additional pressure drop associated with the increased flow path length and minor losses arising from the secondary flow induced by the curved flow path. Examples of curved flow channels include wavy flow channels and zigzag type of flow channels, as shown in Fig. 11.7, among many other possible curved channels. Furthermore, the curved channels can be arranged in phase, such as those shown in Fig. 11.7, or out of phase, as shown in Fig. 11.8, or any other possible combinations. It might be pointed out that the curved flow channels not only provide additional pressure drop for the reactant gas flow in the channels, ΔPchannel ,

Bipolar plates and flow field design Chapter | 11

321

FIGURE 11.7 Schematic of curved flow field designs. (A) Wavy flow channel. (B) Zigzag flow channel.

FIGURE 11.8 Schematic of wavy flow channels in out-of-phase arrangement. Adapted from M.C. Johnson, D.P. Wilkinson, J. Kenna, O.R. Vanderleeden, J. Zimmerman, M. Tabatabaian, US Patent No. 6,586,128 (2003).

322

Fuel Cells for Transportation

as described earlier, it also induces the cross-channel flow due to flow inertial for the curved channels arranged in phase, as shown in Fig. 11.8. For the curved channels arranged out of phase [12] as shown in Fig. 11.8, the channel cross-sectional area changes along the flow direction, and the changes are out of phase for the neighboring channels. A larger cross-sectional area for the channel flow results in a lower flow velocity by the conservation of mass; then, the Bernoulli equation yields a higher fluid pressure, whereas a smaller flow area results in a higher flow velocity, hence a lower fluid pressure. This then leads to a pressure differential across the neighboring channels in a periodic manner, leading to oscillating cross-channel leakage flows. Although these phenomena improve the mass transfer, especially for the electrode area adjacent to the land of the channels, the oscillating nature of the crossover flow would minimize the benefits attainable. More complex curved flow channels have also been investigated (e.g., [13]) but at the increased cost of manufacturing. A further approach that has been investigated in many studies is to deploy obstacle or obstruction in the flow channels in order to increase the pressure loss ΔPchannel in the flow channels. The obstruction can be shaped in any configuration, such as a circular cylinder (needle), a flat plate, or a trapezoid of different windward and leeward angles with different surface wettabilities that are positioned in the middle of the flow channels and fixed at the channel bottom [14,15]. They can also be in any shape and size that are attached to any of the side or bottom surfaces of the flow channels [12,16]. The obstruction can not only increase the pressure drop for the channel to provide more uniform reactant flow distribution among the channels, but also it can facilitate the liquid water transport and removal in the channels. The obstruction can also be placed at the channel inlet or the channel outlet as well, leading to the same effect of increasing the pressure drop along the flow channel [17]. To facilitate the removal of liquid water in the flow channel, the surface wettability of the flow channel can be modified [18
21]. For example, if the channel surface is made hydrophobic, liquid water tends to coalesce into droplet form, which can be carried out of the channel by the reactant gas flow. However, if the droplet size is large, over about 50% of the channel size, droplet can block the reactant flow substantially, leading to less reactant flow in the channel for parallel channel design, or no flow at all if the droplet blocks the entire flow channel. This results in a stagnant reactant gas region, and the reactant gas can be consumed quickly, such that the cell performance will be drastically reduced. Therefore, for high liquid water content in the gas stream, such as stack operation at high current densities, hydrophilic flow channels tend to function much better. This is because liquid water will be spread out on the channel wall surface into thin film, and liquid water flow in the thin film region near the channel surface avoiding blockage of the entire channel. In summary, a hydrophobic channel surface

Bipolar plates and flow field design Chapter | 11

323

is beneficial for fuel cell performance if liquid water content is low, otherwise hydrophilic channel surface will function better. It might be mentioned that, in practice, waterflooding tends to occur more frequently at low current density operation if fully humidified reactant gas streams are supplied. This is because the corresponding low reactant gas velocity in the flow channels cannot overcome the water capillary effect, such that water tends to accumulate in the flow channels and the porous structure of the electrode. The above discussion applies when the channel surface wettability is uniform everywhere. However, surface wettability can be made to change along the flow direction such that the wettability gradient can induce additional capillary force that can drive the transport of liquid water in the channel. This approach is desirable to facilitate liquid water removal and transport in the flow channel. But the coating of the many flow channels for practical large-size bipolar plates is not easy to implement to achieve the wettability gradient. Similarly, serpentine flow field, as discussion earlier, has a number of disadvantageous characteristics that can impact the performance of the fuel cell stack and power system negatively. This mainly arises from the single, long flow channel traversing the entire active area sequentially, not only leading to large pressure drops for the reactant gas, creating large concentration nonuniformity, and a dry region near the inlet region and flooded region near the channel exit region. To address these issues, various improved designs have been developed, as shown in Fig. 11.9. Basically, the conventional serpentine flow field design shown in Figs. 11.4B and 11.5 has a single flow field traversing through the entire active area from the channel inlet to the outlet in a sequential manner. In contrast, the single flow channel can be arranged such that a higher pressure channel can be positioned next to a lower pressure channel to promote the cross-channel leakage flow, to improve the reactant gas distribution more uniformly over the entire active surface with better water distribution and removal through enhanced crosschannel leakage flow. Fig. 11.9 illustrates the single flow field traversing through the entire active area twice (Fig. 11.9A) and three times (Fig. 11.9B) by creating a large span between the high-pressure channels interspersed with the low-pressure channels. It can be demonstrated [22] that the interchannel pressure differential can be regulated by the channel design, hence enhancing the degree of interchannel cross-leakage flow, or the effect of convection through the porous electrode for mass transfer to the CL, the degree of water removal from the MEA adjacent to the land area, and thereby the overall cell performance. Other variations of the design have also been developed, such as radially, or from the central to the periphery of the bipolar plates, with a similar effect as those shown in Fig. 11.9. It should be pointed out that designs like that shown in Fig. 11.9 have high-pressure channels (or upstream portion of the flow) next to the lowpressure channels (downstream portion of the flow), leading to a large

324

Fuel Cells for Transportation

FIGURE 11.9 Schematic of serpentine flow field design traversing the active area twice and thrice. (A) A single serpentine flow channel traversing through the active area twice; red color represents the high pressure (first traverse) and blue low-pressure portion of the channel (second traverse). (B) A single serpentine flow channel traversing through the active area three times; red color represents the high pressure (first traverse), blue intermediate pressure (second traverse), and pink low-pressure portion of the channel (third traverse). Adapted from I. Sabir, MASc thesis, University of Waterloo (2005).

pressure differential between the neighboring channels. This large pressure differential may result in excessive cross-channel flow, a phenomenon called reactant gas short-circuiting between the upstream and downstream portion of the flow. This phenomenon may reduce the reactant gas flowing through the entire length of the serpentine channel, thus reducing the fuel cell performance. This could also potentially accelerate the fatigue and damage to the porous electrode structure adjacent to the channel land, especially when the fuel cell operation is at high current densities, and dynamic with loading cycling.

Bipolar plates and flow field design Chapter | 11

325

Interdigitated flow channel has an unreasonably high-pressure drop due to the dead-end channel design. An easy improvement is to open up the dead end for each channel slightly by using porous gas block or even a small hole or slit to allow some fluid to be escaped into the manifold area, instead of completely through the porous electrode adjacent to the flow channels. This would become more like parallel flow channels with flow restrictions at the channel inlets and/or outlets, as described earlier. Another modification is to have flow channel converging (narrowing in) along the flow direction to compensate for decreasing concentration along the channel due to reactant transporting into the MEA for electrochemical consumption. Similarly, a flow channel depth is decreasing gradually along the flow direction. These last two methods are the same in their effect by reducing the cross-sectional area of the flow channels to accelerate the fluid flow to make up for the loss of reactant to the reaction in the MEA. It is noticed that the modifications discussed might blur the boundary between the type of flow channels we have discussed so far. From the preceding discussion, it becomes evident that each of the fundamental flow field designs themselves has pros and cons, leading to not optimal for fuel cell performance and durability. Therefore, hybridization of flow channel designs, or some form of the combinations of the fundamental designs, becomes imperative in order to achieve optimal flow field design. One of the widely used hybrid flow field designs is a combination of parallel
series connections in much the same way as in the electrical circuit. Many such designs have been patented and extensively investigated as shown in the review article [2]. A simple version of the parallel
series combination [12] is to use several parallel channels traversing through the active area in much the same way as the serpentine channels, such that the total reactant flow is divided into the number of parallel channels, and each individual channel length is significantly reduced. As a result, the amount of mass flow in the individual channels, thereby the pressure loss ΔPchannel through the channels, is substantially reduced. However, the channel is maintained sufficient length so that ΔPchannel is still much larger than ΔPmanifold . Further, the number of parallel channels is limited in this design, such that the issue of nonuniform reactant gas flow distribution among the parallel channels can be virtually eliminated. All the design modifications mentioned earlier for the different types of flow field designs can also be incorporated, such as curved flow channels like wavy and zigzag channels, obstructions placed in the various locations along the flow channels to induce the crosschannel leakage flow, though the effect is much reduced compared to the interdigitated flow field designs. Modification of the flow channel surface wettability has also been investigated by either surface coating or using different bipolar plate materials. Literature is rich in this area of research, and many flow field design patents are available as well. Finally, it might be deserving to mention that except for the three fundamental flow field designs discussed early, there are many other possible

326

Fuel Cells for Transportation

designs that might or might not have been extensively investigated. Some examples include biomimetic, fractal, and constructal-law-based flow field designs. Although these designs may have been more deeply rooted in nature and theory, the nonstructured, nonregularly patterned flow channels with cascadingly decreasing channel sizes are too complex for cost-effective manufacturing. Hence, they are not employed in practical fuel cell stacks, and they are not elaborated on further in this chapter.

11.3.2.5 Flow field design with metallic bipolar plates Metals have high electrical and thermal conductivity, as well as high mechanical strength. Therefore, they can be used in a variety of ways for flow fields. For example, metal foams can be used as a porous flow field. With proper designs of foam structures, a convective flow of reactant gas is produced normal to the MEA surface to facilitate convective mass transport of reactant gas to the CL. Further, the entire MEA surface is in close contact with the porous metal foam so that electrical and thermal transport along with mass transport occurs throughout the entire MEA surface. This is in contrast with the parallel, serpentine, and interdigitated flow channel designs as well as their various hybridizations in that only the channel land is available for the electrical transfer, while mass transfer is only through the channel region, although heat transfer is a bit complicated, but mainly through the land region as well. Clearly, porous metal foams provide a clear advantage in utilizing the entire MEA surface for all the transport phenomena occurring in fuel cells. Similarly, metal wire meshes function essentially in the same way as metal foams. Metals can be made into a very thin sheet of 1 mm thick or less, due to their high ductility and formability. Metal sheets can be easily bent in such a way that forms flow channel on both sides of the sheet. Flow channel on one side of the sheet can serve as flow field for one reactant gas, and the channel on the other side of the sheet for another reactant gas. More often, two such bent metal sheets are fused together in a structure as shown in Fig. 11.10 [23]. This integrated assembly provides reactant gas flow fields on the two outside surfaces of the sheets, while the coolant flow field is in the space between the two metal sheets. This design is practically advantageous since it is very efficient in the use of available space and materials, leading to high stack power density, which is also due to the high electrical and thermal conductivity of the metal materials giving rise to superior stack performance. 11.3.2.6 Latest developments in practical flow field design Several commercial fuel cell vehicles have been available in the market place, including the first-generation Toyota Mirai (Mirai 1) introduced in December 2014, second-generation Mirai (Mirai 2) introduced in late 2020, and Hyundai Nexo available since early 2018, Hyundai’s second-generation

Bipolar plates and flow field design Chapter | 11

327

commercial fuel cell vehicle after its ix35 fuel cell vehicle. Nexo fuel cell design uses wire screen as the porous flow field. The flow field used in Mirai 1 is a sintered metal foam, as shown in Fig. 11.11. The threedimensional porous structured flow field is made of fine metal meshes, and it creates a flow component normal to the MEA surface, enhancing the air flow toward the MEA surface and thereby facilitating air transport to the cathode CL by a strong convective flow. The metal mesh surfaces are treated for hydrophilic surface conditions to absorb liquid water into thin films on the mesh surfaces, preventing the formation of large water droplets that might block the pores of the flow field. Therefore, product water generated in the cathode is removed quickly through the flow field, avoiding waterflooding at high current density operation. The contact area between the metal foam flow field and the MEA surface is small and distributed over the MEA surface so that air transport to the CL is not impeded by the channel land in the conventional flow field designs, while nearly uniform distribution of the contact area improves the transport of electrons as well. Further, near the cathode inlet region the metal foam structure is designed in such a way that less air flow is directed toward the MEA surface, mitigating potential cathode drying, because in Mirai design, humidification is achieved inside the stack, and the conventional external humidifier is eliminated. In addition, the unit cell in Mirai 1 is in rectangular shape with the long side for the anode hydrogen flow, as indicated in Fig. 11.11C. The anode flow plate has a channel-based fine-pitch flow field, while a 3D fine metal mesh flow field is used for the cathode side. Product water in the cathode CL near the air outlet region is transported to the anode side to humidify the inlet hydrogen stream via back-diffusion through the thin membrane, whereas the inlet air is humidified by the water transported back from the anode through the membrane, thus forming a water transport cycle within the MEA. Further through strategic design and control of coolant flow (such as larger cooling channels and higher coolant flow) in the air inlet region, lower temperature is maintained in the air inlet region to prevent water evaporation from the membrane and lower the amount of water required for the humidification of the air stream. Therefore, water management is achieved

FIGURE 11.10 Schematic of integrated flow field design made of thin metal sheets, incorporating reactant gas flow channels and coolant flow channels into one piece. Adapted from F. Issacci, T.J. Rehg, US Patent No. 6,686,084 (2004).

FIGURE 11.11 Schematic of a section of a unit fuel cell and 3D fine metal mesh flow field used in the first-generation Toyota Mirai fuel cell vehicle (Mirai 1). (A) Face view, (B) side view, and (C) flow arrangement for the anode and cathode gas streams as well as coolant [24,25]. (A) and (C) adapted from N. Konno, S. Mizuno, H. Nakaji, SAE Paper Number 201501-1175; https://doi.org/10.4271/2015-01-1175; (B) adapted from T. Yoshizumi, H. Kubo, M. Okumura, SAE Technical Paper 2021-01-0740. https://doi.org/10.4271/2021-01-0740.

Bipolar plates and flow field design Chapter | 11

329

through both the flow field designs and coolant flow control. It might be pointed out that a much longer hydrogen flow path is beneficial to capture an adequate amount of water from the air outlet region of the cathode via back-diffusion to humidify the inlet hydrogen, as well as inlet air. Water transport through the membrane is facilitated by a very thin membrane electrolyte layer, about 1/3 thinner than the membrane used in Toyota’s previous FCHV-adv vehicle leased in 2008. The 3D fine mesh-based flow field used in Mirai 1 has been recognized to have a higher number of parts and components that increased the cost, and reactant flow through the porous fine mesh flow field also has a higher pressure loss than the pressure loss through a conventional flow channel. As a result, Toyota Mirai 2 uses a straight parallel flow channel design for the cathode and a zigzag (or wavy) flow channel design for the anode, as shown in Fig. 11.12A and B. This counterflow arrangement for the anode and cathode streams is not only good for effective mass

FIGURE 11.12 Schematic of the anode and cathode flow fields used in the second-generation Toyota Mirai fuel cell design (Mirai 2). (A) Schematic of the anode, cathode, and coolant flow arrangement. (B) A combined face view of the anode and cathode flow field. (C) A magnified view of the squared area shown in (B) above for the cathode flow field illustrating the periodic partially narrowed straight flow channel. (D) Cross-channel leakage flow for the cathode induced by the periodically partially narrowed flow channels shown in (C). (A) and (D) adapted from T. Yoshizumi, H. Kubo, M. Okumura, SAE Technical Paper 2021-01-0740. https://doi.org/ 10.4271/2021-01-0740; (B) and (C) Adapted from T Takahashi, T Ikeda, K Murata, O Hotaka, and eight more co-authors, J. Electrochem. Soc. 169, 044523 (2022).

330

Fuel Cells for Transportation

transfer, but also reduces the manifold arrangement from four sides in Mirai 1 (Fig. 11.11C) to 2 sides (Fig. 11.12A and B). On the other hand, the manifolds for anode and cathode flow are relatively simple for the cross-flow arrangement in Mirai 1, while they are much more elaborate and occupy much more space (the inactive area) for the counterflow arrangement in Mirai 2 as shown in Fig. 11.12A and B. This again illustrates that engineering design is an optimization process to balance the various conflicting requirements for an optimal solution. Further, the cathode parallel flow channels have periodic partially narrowed channel points as shown in Fig. 11.12C; this induces a pressure differential between the neighboring channels, forcing a cross-channel secondary flow, as illustrated in Fig. 11.12D. Therefore, this cathode channel design enhances water removal in the flow channel, as well as the compressed electrode structure adjacent to the land area, and increases oxygen transport to the CL. It is reported [25,26] that the oxygen concentration in the GDL is about 2.3 times higher than the conventional straight channel without periodic narrowed channel points and is as high as the concentration when using the 3D fine metal mesh flow field used in Mirai 1.

11.4 Materials and manufacturing Many materials can be and have been used to make bipolar plates, which largely determine the method of manufacturing employed. To satisfy the functions and requirements of the bipolar plates as described earlier, bipolar plate materials need to have specific physical and electrochemical properties such as excellent electrical and thermal conductivity, high corrosion resistance and gas impermeability, good mechanical strength, and low cost. The criteria or considerations for bipolar plate material selection may include the following: G

G

G

G

Chemical stability/compatibility: The candidate material should not produce disruptive hydride layer when it is in contact with hydrogen gas, should not passivate, and become nonconductive both electrically and thermally when it is in contact with air (oxygen). Corrosion-resistance: Required for durability and performance, especially low release rate of contaminating cations, that could potentially poison catalyst, ionomer, and membrane. Dissolution-resistance: Required for durability and performance, this may influence the selection of manufacturing method, for example, 3D printing of carbon composite for flow channels may not hold together strong enough that erosion may occur when subjected to humidified air or hydrogen gas stream that may also carry liquid water. Low cost: Since the method of manufacturing is largely determined by the materials selected, cost includes both the cost of raw materials and

Bipolar plates and flow field design Chapter | 11

G

G

G

G

G

G

G

331

the cost of manufacturing the bipolar plates. Therefore, the candidate materials must allow easy and high-volume manufacturing. Lightweight: Low-density materials for easy shipping, handling, and onboard mobile applications, and allow high power density operation. Electrical conductivity: Electron transports only through the portion of the bipolar plate surface that is in contact with the MEA, and high electronic conductivity is necessary. Gas diffusivity/impermeability: Bipolar plate separates hydrogen and air in fuel cells; it must prevent gas from crossover for safety requirement Recyclability: The candidate materials should be recyclable and reusable to facilitate end-of-life disposal processing and a circular economy. Mechanical strength: Must be strong mechanically to support the MEA and the stack for structural integrity under rough vibrating or shock operating conditions, and high resistance to plastic deformation under the start/stop working condition. Thermal conductivity: Waste heat generated in the fuel cell stack must be removed effectively to maintain uniform stack temperature. This is especially important since the temperature difference between the fuel cell stack and the ambient (or coolant) is small. Surface finish and flatness: The bipolar plate surface must be sufficiently smooth and flat to minimize interfacial contact resistance.

In November 2017, the United States Department of Energy published a report about the technical targets for fuel cell systems and components, and the technical targets for bipolar plates are shown in Table 11.1 [27]. Although the thermal conductivity value is not included in the table, other literature, for example, [1], states that thermal conductivity for the bipolar plates should be larger than 20 W (cm K)21.

11.4.1 Typical materials and classification Titanium and niobium (with gold coating) were the base materials for bipolar plates used by General Electric in the 1960s for fuel cells used in space exploration, and graphite became popular for terrestrial commercial application since the early 1970s, mainly because of its low cost and high corrosion resistance [28,29]. Further, graphite has low bulk resistivity and low electrical contact resistance with electrode backing materials. On the other hand, graphite is hard and brittle, manufacturing the flow field is difficult and expensive, and graphite bipolar plate is heavy and bulky as well. As a result, many other materials have been explored, and they generally fall into three categories: G G G

Graphite; Carbon composites; and Metals.

332

Fuel Cells for Transportation

TABLE 11.1 Technical targets for bipolar plate materials [27]. Characteristic Plate cost Plate weight

Unit 21

$ kW

21

kg kW

3 21

22

21



Status

2025 target

5.40

2

, 0.4

0.18

,2 3 1026

2 3 1026

Plate H2 permeation

Std cm s cm Pa 3 atm 100% RH

Corrosion (anode)

μA cm22

No active peak

, 1 and no active peak

Corrosion (cathode)

μA cm22

, 0.1

,1

Electrical conductivity

S cm21

. 100

. 100

Areal specific resistance

Ohm cm2

0.006

, 0.01

Flexural strength

MPa

. 34 (carbon plate)

. 40

Forming elongation

%

20
40

40

@ 80 C,

11.4.2 Graphite Graphite is the most commonly used material for bipolar plates in PEM fuel cell stacks, often referred to as conventional plate material, or serving as a “standard” for comparison with new alternative materials being developed. Graphite is pure carbon, with carbon atoms forming strong hexagonal bonds. The individual sheets of these hexagonal rings of carbon atoms are called graphene layers, and graphene layers stack together to form graphite. In industry, flexible graphite, also known as graphite foil, or expanded graphite, is used. It is manufactured by exposing graphite flake into acids and oxidizing agents with intensive heat treatment so that the bonds between each of the layers are weakened, and the graphite is expanded. The expanded graphite is then compressed to a porous mat structure impregnated with resins to improve mechanical strength and fluid impermeability. Therefore, flexible graphite is nonisotropic with directional dependence for its thermal and electrical conductivities. Naturally, the conductivity value parallel to its surface is much larger than the corresponding value normal to its surface, and the difference could be as high as by the order of magnitude. This property is beneficial to maintain thin but large bipolar plates used in practical

Bipolar plates and flow field design Chapter | 11

333

applications to have a uniform distribution of temperature and electrical potential in the in-plane direction. On the other hand, graphite is quite brittle, and mechanical handling requires care. Machining flow fields is time-consuming and expensive. It also has a relatively low electrical and thermal conductivity compared to metals. Further, graphite is quite porous, and thick plates are required to make gas impervious, thus limiting the power density achievable for fuel cell stacks. However, for some particular applications, this might be an advantage; such as, for forklifts and lift trucks a heavy power system is needed to balance the weight to be lifted. Due to its excellent resistance to corrosion that lends to durable operations, graphite bipolar plates are extensively used in PEM fuel cell stacks for heavy-duty applications such as buses, trucks, trains, and surface ships. Graphite bipolar plates are commonly machined or molded with flow fields. Compression molding is another method, and graphite mixture with additives and/or binders is compression molded and preferably subjected to a heat treatment in the absence of oxygen [30]. Suitable additives may include aluminum oxide, zircon dioxide, silicon dioxide, titanium dioxide, silicon carbide, and powdered coke. Suitable binders are heat curable from 300 C to 800 C and may include carbohydrates such as fructose, glucose, galactose, and mannose and oligosaccharides like sucrose, maltose, and lactose. Flexible graphite can also be embossed into a flow field plate as well. Properties and structure of natural graphite sheets are available in [31].

11.4.3 Carbon composite Carbon composites typically consist of 80%
90% carbon and graphite, and the specific formulation of natural and synthetic graphite particles, carbon fibers, and amorphous carbon particles is proprietary for specifically desired properties. Resins used to bind the ingredients together include epoxy-based or vinyl ester-based. Carbon composite plates can be categorized as carbonbased or metal-based, depending on whether metals are included in the formulation. Metal particles contribute to better structural integrity, electrical and thermal conductivity, and gas impermeability, while graphite provides corrosion resistance. However, carbon-based composites are the norm and are typically used in practice. It is evident that the amount and dispersion of the binder resins are important for the resulting plate performance. If they are inadequate, permeable regions could develop, while too much would result in losses in electrical and thermal conductivity. Manufacturing of carbon composite bipolar plates can be compression molding, transfer molding, or injection modeling; however, compression molding is favored because of the high fill content, and it is suitable for mass production. However, injection molding has a higher rate of production than compression molding, since the cycle time is on the order of tens of

334

Fuel Cells for Transportation

seconds for injection molding as compared to tens of minutes for thermoplastic compression molding. Composite composition also influences manufacturability (hence cost) and functional properties of the resulting plates, requiring optimization for specific applications. After molding, the plate surface is often covered by resin-rich skin that needs to be removed to minimize interfacial contact resistance to electrical and thermal conduction. Highspeed molding is also favorable for volume production, and the plates produced can have good dimensional accuracy and repeatability. A recent review article provides a detailed summary of composite bipolar plates [32].

11.4.4 Metallic bipolar plates Metals have high ductility and malleability, mechanical formability, easy for machining and forming, hence for mass production. They also have high electrical and thermal conductivity, low gas permeability, and high mechanical strength for structural integrity, thus the possibility of thin sheets being used as the bipolar plates, reducing the size and weight of the resulting fuel cell stacks. Metal bipolar plates also have high survivability for mobile applications when subjected to sudden impacts and crashes such as automotive accidents. Metal materials that have been considered include stainless steel, titanium, aluminum, nickel alloys; commercial metallic bipolar plates often use stainless steel (typically SS 316, SS 316L, etc.) and titanium. However, in the acidic environment of PEM fuel cells, and exposing to humidified oxidizing and reducing conditions, metals tend to form oxides, passive layers, and metal ions quickly. Metal oxide layers are usually not good conductors for electron and heat, and metal ions are contaminants for the ionomer and membrane electrolyte materials and reduce the fuel cell performance and durability. This poor chemical stability of metals in PEM fuel cells poses challenges to fuel cell performance and durability. As a result, metal bipolar plates require a protective surface coating to improve its chemical stability. The challenge is that the surface coating film or layer must be conductive electrically and thermally and durable over the fuel cell lifetime. This limits the choice of suitable coating materials. Toyota had used gold plating for its bipolar plates prior to Mirai technology, but gold plating is not economical and has not been continued for commercial applications. Many surface coating techniques have been considered, such as electrodeposition, ion implantation (II), physical vapor deposition, chemical vapor deposition, magnetron sputtering, arc ion plating, dip coating, pack cementation, chemical passivation, plasma surface diffusion alloying, and micro-arc alloying [4,33]. They can be categorized as metallic, carbon-based, and composite coatings. The metallic coating includes pure metallic, metallic nitrides, and metallic carbides [34], and pure metallic coating uses expensive noble metals such as gold and platinum that can meet performance and durability requirements, but they are too costly and usually not used in commercial

Bipolar plates and flow field design Chapter | 11

335

applications. Carbon-based coating includes graphite, self-assembled graphene film, and diamond-like coatings, and these coatings need to improve the bonding with the substrate surface for durability and for avoiding surface damages like scratches during the shipping, handling, and assembly processes [35]. Composite coating consists of conductive carbon and polymers. Currently, commercial stainless steel and titanium bipolar plates are often coated by titanium nitride (TiN) and titanium carbide (TiC) as well as carbon-based coating. Both TiN and TiC are hard ceramic materials that have good chemical stability, corrosion resistance, low cost, and high mechanical hardness with good wear resistance; they have superior properties to pure noble metals. Carbon-based coatings have excellent performance and cost advantages. However, one of the challenges for bipolar plate coatings is to achieve a uniformly coated surface for the metallic bipolar plates. The metal plate used to make bipolar plate is very thin, in the order of 0.1 mm or thinner. Therefore, the manufacturing of the flow field and other structural features such as sealing on its surface is accomplished by plastic forming methods, such as hydroforming, stamping, roll forming, hydraulic pressure, and in-die mechanical joining processes. These methods allow high-volume production for further cost reduction as well. Metallic bipolar plates are very thin with high electric and thermal conductivity and mechanical strength; they result in high power density fuel cell stacks. Therefore, it is believed that they are more suitable for light-duty vehicles and aerospace applications where lightweight and compact designs are required. One challenge is whether to have plate surface coating before making the field or afterward. The coated surface layer may be damaged by the plastic forming processing; but after the forming, the plate surface is no longer flat with flow fields and other structural features and may not be coated uniformly. One consideration is to coat the plate twice, once before the forming procedure, and second time after the forming, to ensure the bipolar plate surface is properly protected, and this would also potentially resolve or at least compensate for the nonuniform coating each time. However, double coating increases the step of manufacturing and slows down the rate of production, hence the cost of the plates manufactured. In general, flexible graphite and compression-molded carbon composite bipolar plates are more durable and have lower cost (roughly 50% less) than metallic bipolar plates, and the latter also results in higher stack power density than the flexible graphite plates. The best stack power density of 4 kW L21 (Ballard Power Systems Inc.) has been achieved for carbon-based plate stacks as compared with 5.4 kW L21 (Toyota Mirai 2) for metallic bipolar plate stacks (note end plates are not included in these power density calculations). Carbon-based plates have higher durability and have reached more than 30,000 hours of operation in fuel cell transit bus operations, while metal plates in today’s technology have a shorter lifetime, limited to lightduty vehicle applications of 5000 hours of operation. Another important

336

Fuel Cells for Transportation

consideration is the recyclability of carbon plates. At the end of the stack lifetime, carbon-based bipolar plates can be reused with fresh MEAs, and such refurbished fuel cell stacks can still achieve the original product specifications, whereas the catalyst in the used MEAs taken from the used stacks can be retrieved for reuse as well. On the other hand, metallic bipolar plates cannot be reused. This would substantially lower the cost of refurbished stacks using carbon plates. Therefore, care and due consideration must be given when developing bipolar plates for a targeted application.

11.5 Summary PEM fuel cells have reached the early stage of commercial applications, becoming an integral part of strategies for global climate change mitigation and postpandemic economic recovery. Bipolar plates contribute up to 80% by weight, 50% by volume, and 40% by cost of PEM fuel cell stacks, hence a key component in the stacks that play an important role in achieving high performance, low cost, and long lifetime for fuel cells operating at high power densities. This chapter provides an overview of the basic functionalities and requirements for bipolar plates, various basic flow field designs and associated various design improvement features, proper materials, and associated manufacturing methods that have been developed to achieve the functions and requirements for the bipolar plates. State-of-the-art knowledge and technology are highlighted, and technical challenges and future directions of the R&D are outlined. This chapter offers a practical starting point for the understanding of the state of the art in bipolar plate architecture and for the innovative development of next-generation bipolar plate technology. For commercial applications, mass production in high volumes via an automated production line will be a key factor in the cost reduction of bipolar plates that can play a significant part in lowering the total cost of the fuel cell stacks and systems.

Acknowledgments The author is grateful for the continued financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC) via Discovery Grant, as well as other NSERC grants, and many industrial partners over the past several decades.

References [1] [2] [3] [4] [5] [6]

A. Tang, L. Crisci, L. Bonville, J. Jankovic, J. Renew. Sustain. Energy 13 (2021) 022701. X. Li, I. Sabir, Int. J. Hydrog. Energy 30 (2005) 359
371. M. Sauermoser, N. Kizilova, B.G. Pollet, S. Kjelstrup, Front. Energy Res. 8 (2020) 13. K. Xiong, W. Wu, S. Wang, L. Zhang, Appl. Energy 301 (2021) 117443. A. Pollegri, P.M. Spaziante, US Patent No. 4,197,178 (1980). D.S. Watkins, K.W. Dircks, D.G. Epp, US Patent No. 4,988,583 (1991).

Bipolar plates and flow field design Chapter | 11 [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35]

337

X. Li, I. Sabir, J. Park, J. Power Sources 163 (2007) 933
942. T. Kanezaki, X. Li, J.J. Baschuk, J. Power Sources 162 (2006) 415
425. J. Park, X. Li, J. Power Sources 163 (2007) 853
863. K. Jiao, J. Bachman, Y. Zhou, J.W. Park, Appl. Energy 115 (2014) 75
82. A. Kazim, H.T. Liu, P. Forges, J. App Electrochem. 29 (1999) 1409
1416. M.C. Johnson, D.P. Wilkinson, J. Kenna, O.R. Vanderleeden, J. Zimmerman, M. Tabatabaian, US Patent No. 6,586,128 (2003). C. Yin, Y. Song, M. Liu, Y. Gao, K. Li, Z. Qiao, et al., Appl. Energy 305 (2022) 117893. Y. Qin, X. Li, K. Jiao, Q. Du, Y. Yin, Appl. Energy 113 (2013) 116
126. Y. Qin, Q. Du, Y. Yin, K. Jiao, X. Li, J. Power Sources 222 (2013) 150
160. Z.R. Guo, H. Chen, H. Guo, F. Ye, Energy Convers. Manag. 274 (2022) 116467. Y. Wang, Z.Y. Sun, L. Yang, Energy Convers. Manag. 252 (2022) 115077. B. Mondal, K. Jiao, X. Li, Int. J. Energy Res. 35 (2011) 1200
1212. Y. Qin, X. Li, Q. Du, Y. Yin, K. Jiao, Int. J. Hydrog. Energy 38 (2013) 12879
12885. Y. Wang, S.A. Shakhshir, X. Li, P. Chen, Int. J. Low-Carbon Technol. 9 (2014) 225
236. Y. Qin, X. Li, Y. Yin, Int. J. Energy Res. 42 (2018) 3315
3327. I. Sabir, MASc thesis, University of Waterloo (2005). F. Issacci, T.J. Rehg, US Patent No. 6,686,084 (2004). N. Konno, S. Mizuno, H. Nakaji, SAE Paper Number 2015-01-1175; Available from: https://doi.org/10.4271/2015-01-1175. T. Yoshizumi, H. Kubo, M. Okumura, SAE Technical Paper 2021-01-0740. Available from: https://doi.org/10.4271/2021-01-0740. T. Takahashi, T. Ikeda, K. Murata, O. Hotaka, et al., J. Electrochem. Soc. 169 (2022) 044523. US Drive, Fuel Cell Technical Team Roadmap. ,https://www.energy.gov/sites/default/ files/2017/11/f46/FCTT_Roadmap_Nov_2017_FINAL.pdf., November 2017. K. Prater, J. Power Sources 29 (1990) 239
250. C. Paola, S. Supramaniam, J. Power Sources 102 (2001) 253
269. R. Meissner, M. Irgang, K. Eger, P. Weidlich, H. Dreyer, US Patent No. 5,736,076 (1998). M. Cermak, N. Perez, M. Collins, M. Bahrami, Sci. Rep. 10 (2020) 18672. K.I. Jeong, J. Oh, S.A. Song, D. Lee, D.G. Lee, S.S. Kim, Compos. Struct. 262 (2021) 113617. Y. Wang, D.F.R. Diaz, K.S. Chen, Z. Wang, X.C. Adroher, Mater. Today 32 (2020) 178
203. M. Viral, C.J. Smith, J. Power Sources 114 (2003) 32
53. I. Alaefour, S. Shahgaldi, J. Zhao, X. Li, Int. J. Hydrog. Energy 46 (2021) 11059
11070.

This page intentionally left blank

Chapter 12

Heat transport and thermal management Siyuan Wu1, Kui Jiao2 and Jae Wan Park1 1

Department of Mechanical and Aerospace Engineering, University of California, Davis, CA, United States, 2State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China

12.1 Introduction The energy efficiency of a proton exchange membrane fuel cell (PEMFC) is limited to about 60% due to the heat generated during its operation. The amount of waste heat produced is nearly equal to the output power of the PEMFC, which means the heat production rate of a PEMFC stack with 100 kW output power is about 70 kW [1]. From the electrochemical reaction kinetics and proton conductivity perspectives, a higher working temperature is favored. Higher temperatures can also facilitate the diffusion of reaction gases and product water. On the contrary, when the operating temperature is lower than 60 C, flooding is prone to occur at the electrodes. Flooding will hinder the transport of reactant gases, cause local oxygen starvation at the electrochemical reaction sites, and result in aggravated concentration overpotential. The nonuniform temperature distribution also has negative impacts on the durability and stability of PEMFC [2]. The operating temperature range of PEMFC is dictated by the material properties of the polymer electrolyte membrane. The most well-known and well-established membrane is Nafion, a perfluorosulfonic acid membrane developed by Dupont. To achieve better proton conductivity, the Nafion membrane needs to be well hydrated. A higher temperature will cause the membrane to dehydrate. As a result of membrane dehydration, the membrane water content will drop, and the conduction of proton and electron will be interrupted. Consequently, severely irreversible degradation may happen. Hence, the proper operation temperature of PEMFC ranges between 60 C and 80 C [3]. So, it is important to keep the temperature constant and uniformly distributed within a PEMFC stack, which requires the assistance of a thermal management subsystem. Moreover, the thermal management subsystem approximately accounts for 8% of the total cost of a PEMFC system [4]. The thermal management of Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00001-0 © 2023 Elsevier Ltd. All rights reserved.

339

340

Fuel Cells for Transportation

PEMFC stack also includes cold start and waste heat recycle strategies [5]. Therefore, performing thermal management more effectively and efficiently is of paramount importance.

12.2 The heat in proton exchange membrane fuel cell 12.2.1 Heat generation The heat sources of PEMFC emanate from the reversible heat of the electrochemical reaction, the irreversible heat of the electrochemical reaction, the Joule heat, and the latent heat of phase change [6], as shown in Fig. 12.1. According to the second law of thermodynamics, the concomitant inefficiency of the reversible electrochemical reaction releases heat, namely, reversible heat or entropic heat, which represents the reaction rate and the change in entropy of the electrochemical reactions. When current flows, to keep the temperature steady, reversible heat needs to be supplied or removed from the electrode region in due course. The reversible heat accounts for 30% of the total heat generated approximately [5]. Its expression is given by: qrev 5 jUT

dEr dT

ð12:1Þ

–

∆h nF

–

∆g nF

qrev

Open Circuit Voltage

Voltage, V

qirrev + qohmic + qOCV

1 Activation loss 2 Ohmic loss 3 Concentration loss

Power Output

1

2

3

Current density, A cm-2 FIGURE 12.1 Illustration of a typical PEMFC polarization curve with heat sources.

Heat transport and thermal management Chapter | 12

341

where j is the net current density calculated from the Butler
Volmer equation, T the temperature, and Er the reversible voltage as a function of temperature. The irreversibility associated with the electrochemical reaction, for instance, the activation overpotentials of cathode and anode, also contributes a considerable amount of waste heat. The irreversible heat accounts for about 60% of the total waste heat generation [1]. The electrochemical reaction that happens in PEMFC can be divided into two half-reactions at the cathode and anode. The changes in entropy and overpotential associated with each halfreaction are different. Since the oxygen reduction reaction takes place at the cathode, there will be more heat released at the cathode, imposing challenges to the temperature management of the membrane electrode assembly (MEA) [7]. The expression of irreversible heat is given by: qirrev 5 jUηact

ð12:2Þ

where ηact is the activation overpotential. The Joule heat is emanated from ionic and electronic resistances. This waste heat generation mechanism includes the heat generated from the conduction of protons within the membrane and the electrolyte in the catalyst layer (CL); the heat generated from the conduction of electrons in porous electrodes, bipolar plates (BPs), and current collectors; and the heat generated from the interfacial contact resistances between each component. Its expression is given by: qohm;H1 5

qohm;e2 5

2 ~ i ion

κeff ion 2 ~ ie

κeff e

ð12:3Þ

ð12:4Þ

where the currents ~ i ion and ~ ie are functions of ionic conductivity and electronic conductivity and phase potential ϕion and ϕe , respectively.

κeff ion

~ i ion 5 2 κeff ion rϕion

ð12:5Þ

~ ie 5 2 κeff e rϕe

ð12:6Þ

is the effective ionic conductivity, and κeff where e the effective electronic conductivity. When the proton exchange membrane has not been adequately hydrated, its ionic resistance will be relatively large, and there will be an appreciable amount of Joule heat caused by proton conduction, whose amount dominates the Joule heat. Under certain circumstances, where local electronic resistance and interfacial contact resistance are noticeable, local hot spots will probably occur. During the operation of PEMFC, phase change of water often happens and accompanies a significant amount of heat releasing or absorbing. During the process of membrane hydration, heat will also be released or absorbed

342

Fuel Cells for Transportation

when water changes from a certain phase to membrane water. Take the phase change from vapor to liquid water, for example, the heat generated from this process can be expressed: qphase 5 Sv2l h

ð12:7Þ

where h is the latent heat of phase change and Sv2l is the source term of phase change. The expression of the energy conservation equation in PEMFC is given by: i h i @h εsρl cp;l T 1 εð1 2 sÞρg cp;g T 1 rU εsρl cp;l ul T 1 εð1 2 sÞρg cp;g ug T @t

5 rU keff rT 1 ST ð12:8Þ where ST is the source term of heat. The source terms of heat for each component in a PEMFC can be expressed as the summation of irreversible heat, reversible heat, Joule heat, and latent heat of phase change. Fig. 12.2 illustrates a representative temperature profile across a PEMFC with heat transfer modes and heat sources for each component. The heat source in the proton exchange membrane (MEM) is from Joule heat of ionic resistance: 2

:rϕion : κeff ion

ð12:9Þ

The heat source in the anode CL is the summation of irreversible heat, Joule heat of ionic and electronic resistances, latent heat of phase change,

FIGURE 12.2 A representative temperature profile across a PEMFC with heat transfer modes and heat sources for each component; not to scale.

Heat transport and thermal management Chapter | 12

343

and entropic heat:   ΔSa T 2 2 eff 1 ðSv2l 2 Sd2l Þh ð12:10Þ Ja ηaact  1 :rϕe : κeff e 1 :rϕion : κion 1 Ja 2F Analogously, the expression of heat source in the cathode CL is given by:   ΔSc T 2 2 eff 1 ðSv2l 2 Sd2l Þh Jc ηcact  1 :rϕe : κeff e 1 :rϕion : κion 1 Jc 4F

ð12:11Þ

The heat sources in the gas diffusion layer (GDL) and the microporous layer consist of Joule heat from electronic resistance and latent heat of water phase change: 2

:rϕe : κeff e 1 Sv2l h

ð12:12Þ

The heat source in the BPs is simply the Joule heat from electronic resistance: 2

:rϕe : κeff e

ð12:13Þ

Lastly, the heat source in the flow channels is the latent heat of water phase change: ð12:14Þ

Sv2l h

where ΔS is the change in entropy, and Sd2l is the source term of membrane water phase change. As mentioned previously, the combination of irreversible heat and entropic heat accounts for over 90% of the total waste heat generated in PEMFC. It is critical to develop an effective way to dissipate this heat [8].

12.2.2 Heat transport Heat transport is thermal energy in transit due to a spatial temperature difference. Wherever a temperature difference exists either in an object or between objects, there will be heat transport [9]. The heat transport mode within an object is defined as conduction. According to Fourier’s law, the one-dimensional conduction heat transfer rate qcond of an object with temperature distribution T ðxÞ is given by: qcond 5 2 kA

dT ðxÞ dx

ð12:15Þ

where k is the thermal conductivity and is a characteristic of a material, and A is the area that is perpendicular to the direction of heat transport. Applying the energy balance, the heat diffusion equation, namely, the energy conservation equation, is given by:


 @ @T @ @T @ @T k @T k k k ð12:16Þ 1 1 1 q_ 5 @x @x @y @y @z @z α @t

344

Fuel Cells for Transportation

where q_ is the rate of energy generated per unit volume, α is the thermal diffusivity, and it is expressed: α5

k ρcp

ð12:17Þ

where ρ is the density, and cp the specific heat. The energy conservation equation indicates that the rate of change of thermal energy stored in the system is equal to the summation of the net rate of energy transferred into the system by conduction and the rate of energy generated per unit volume [9]. In PEMFC, heat conduction takes place in all components. The heat transport mode between a surface and fluid in motion at different temperatures is defined as convection. According to Newton’s law of cooling, the expression of convection heat transfer rate is given by: qconv 5 hAðTs 2 TN Þ

ð12:18Þ

where h is the convection heat transfer coefficient, Ts is the temperature of the surface, and TN is the temperature of the fluid. The heat transfer coefficient h is a function of flow conditions in the thermal boundary layer, the surface morphology, and the fluid. Analogous to the development of the velocity boundary layer, a thermal boundary layer develops for a free stream flow over an isothermal surface when their temperature differs, as shown in Fig. 12.3. The part of the moving fluid that is in direct contact with the isothermal surface rapidly reaches thermal equilibrium with the surface. Then, via heat transport, the temperature of the adjacent layer of moving fluid is influenced. This process continues until a distance is far enough, namely, the thickness of the thermal boundary layer δt ðxÞ, that the thermal effect of the surface can be neglected and the temperature of the fluid keeps at the free stream temperature TN and no longer changes. The thickness of the thermal boundary layer is defined as the distance from the isothermal surface where T 2 Ts 5 0:99ðTN 2 Ts Þ, and it is proportional to x.

FIGURE 12.3 The development of the thermal boundary layer.

Heat transport and thermal management Chapter | 12

345

Combining Fourier’s law and Newton’s law of cooling, we have:   2k @T @y y50 h5 ð12:19Þ Ts 2 TN This expression suggests defining a dimensionless convection heat transfer coefficient termed the Nusselt number: Nu 

hx k

ð12:20Þ

The Nusselt number is equal to the dimensional temperature gradient at the surface, and it provides a measure of the convection occurring at the isothermal surface [9,10]. For a given geometry, the Nusselt number for a laminar flow is a function of the Reynolds number and the Prandtl number: Nu 5 0:664 3 Re1=2 Pr 1=3

ð12:21Þ

The last mode of heat transport is radiation, electromagnetic radiation generated by the thermal motion of particles in objects. Any object with a finite temperature greater than absolute zero emits thermal radiation. According to the Stefan
Boltzmann law, the expression of radiation heat transport rate is given by:

4 qrad 5 εσA Ts4 2 Tsurr ð12:22Þ where σ is the Stefan
Boltzmann constant (σ 5 5:67 3 1028 W=m2 UK4 ), ε is the radiative property of the surface termed the emissivity that ranges between 0 and 1, and Tsurr is the absolute temperature of the surrounding. Under PEMFC working temperature, which has no significant difference compared with the surrounding temperature, the radiation can be neglected [5]. The waste heat generated in PEMFC is usually dissipated from the surface via heat conduction and convection. The thermal dissipation rate is dictated by the thermal properties of each component. Since the major portion of waste heat is generated from the cathode electrode and dissipated from the BPs, the heat transport along the through-plane direction plays an important role in waste heat removal and the formation of heat spots. There will be a temperature gradient between MEA and BPs, which has the highest and the lowest temperature, respectively. Taking the CL, for instance, a onedimensional through-plane temperature variation can be expressed:

δCL I 2 Δh δ2CL ST 2F 2 Vcell ΔTCL B eff 5 ð12:23Þ eff kCL kCL where 2 Δh 2F represents the electromotive force, h is the enthalpy, δCL is the thickness of the CL, ST is heat production rate, I is the current density, Vcell eff is the output voltage, kCL is the effective through-plane heat conductivity, and F is Faraday’s constant. Similarly, the temperature variations in MPLs,

346

Fuel Cells for Transportation

GDLs, and BPs can also be calculated using their corresponding thermal properties [1]. Some important thermal properties are listed in Table 12.1. These numbers are measured from ex situ experiments, and more accurate numbers need to be observed from in situ experiments. Eq. (12.23) does not consider the morphology of lands and channels and neglects the lateral heat transport along the in-plane direction. In PEMFC, the reactant gases under normal stoichiometric only take away a minor portion of waste heat. The expression of the percentage of waste heat removed by reactant gases is given by:     ρg cp;g ΔTAch u 1 ρg cp;g ΔTAch u Heat removal rate by reactant gas a c

β5 5 Heat production rate I 2 Δh 2 V A cell m 2F ð12:24Þ where ΔTa and ΔTc are the temperature difference between the inlet and outlet of anode and cathode respectively, ρ the density of reactant gases, u the velocity, Ach the cross-sectional area of the flow channel, and Am is the electrochemical active surface area. Assuming the change in temperature at the anode equals to cathode: ρg;a cp;g;a C2ξHa 1 ρg;c cp;g;c C2ξOc 2

2 ΔT β 4F 2 Δh 2 V cell 2F

ð12:25Þ

where C is the species concentration, and ξ is the stoichiometric. For stoichiometric equals to 1.5, even a 10 C temperature between inlet and outlet, the percentage of waste removed by reactant gases is less than 5%, which is negligible [5]. Since the heat transport mode in the bipolar plate is only conduction, neglecting the convection from the flow channel, the heat in the part of GDL that underlies flow channels will initially transfer laterally to the part in contact with the lands; later, transfer to the bipolar plate along the through-plane direction by conduction; and then dissipate to ambient. The equivalent circuit of this process is comprised of two thermal resistors in series, namely, the thermal resistor of in-plane and through-plane directions. The maximum change in temperature along these two directions is given by: through2plane ΔTmax

1

in2plane ΔTmax 2

5

1 2I



2 Δh 2F 2 Vcell



eff kGDL;through δGDL



2 I 2 Δh 2F 2 Vcell Wch eff 2δGDL kGDL;in

ð12:26Þ

ð12:27Þ

where δGDL is the thickness of GDL, Wch is the width of the flow channel, eff eff kGDL;through and kGDL;in is the effective thermal conductivities of through-plane

TABLE 12.1 Thermal properties of individual components in a typical seven-layer PEMFC [11
16]. Components

Chemical composition

Thermal conductivity (W mK21)

Heat capacity (J C21m23)

Thickness (mm)

Polymer electrolyte

Perfluorosulfonic acid membranes

Nafion 0.2
0.95

1.65 3 106

0.01
0.13

0.2
1.5

1.69 3 10

0.001
0.01

Catalyst layer

Mixture of Pt, C, and ionomer

6

Gas diffusion layer

Carbon fiber substrate

Toray paper 0.2
2.8

5.68 3 10

0.1
0.3

Microporous layer

Carbon powder and polytetrafluoroethylene

0.05
0.3

-

B0.02

Bipolar plate

Graphite or its composite, or metal

Graphite 30 Titanium 17 SS316L 16.3

Graphite 1.57 3 106 Titanium 2.35 3 106 SS316L 4.11 3 106

0.3
2

5

348

Fuel Cells for Transportation

and in-plane directions, respectively. The total temperature variation is simply the summation of these two.

12.3 Proton exchange membrane fuel cell thermal management The thermal management of a PEMFC includes maintaining the operating temperature of the stack within a desirable range and creating a uniform temperature distribution across the stack and its individual cells. The cooling method applied depends on the power of the fuel cell. Commonly used cooling methods include air cooling, liquid cooling, heat spreader cooling, phasechange cooling, nanofluid cooling, etc. The devices employed in the thermal management subsystem and its control strategies are briefly introduced.

12.3.1 The cooling of proton exchange membrane fuel cell There are two determining factors when designing the cooling system of a PEMFC stack. First, the optimal operating temperature of a PEMFC stack is restricted to 80 C. Compared to the local peak temperature, about 1800 C, in a cylinder of an internal combustion engine during the expansion stroke [17], the impetus to remove the heat of a PEMFC is significantly weakened. Second, the contribution of exhaust gas to waste heat removal is too small. Although increasing the stoichiometric of the reactant gases seems a proper way to alleviate the difficulty of removing heat, however, it also put the membrane at risk of dehydrating. Hence, nearly the entire waste heat produced by a PEMFC has to be dissipated with the assistance of the cooling system. Because of these two factors, a PEMFC stack needs to be equipped with a relatively large cooling system, which brings challenges to the arrangement of cooling equipment in a vehicle. The waste heat must be removed from the fuel cell stack in time. Otherwise, the increased temperature will promote the electrochemical reaction rate; as a result, a higher local current density will generate more heat. This vicious cycle will form local hot spots and eventually burn out the membrane. For different applications, there are different cooling strategies, for instance, air cooling, liquid cooling, and phase-change cooling.

12.3.1.1 Air cooling Compared with other cooling methods, air cooling has the simplest configuration and better system integration since it does not require the components like water pumps, piping, valves [18] Air cooling has a much lower heat removal capability due to the low heat capacity of air. Hence, air cooling is usually adopted for a PEMFC stack with a power of # 5 kW. For a PEMFC stack of less than 5 W, the natural convection cooling via BPs is sufficient. For the ones greater than 5 W, forced convection cooling by auxiliary

Heat transport and thermal management Chapter | 12

349

equipment is required, like external fans or blowers. The natural convection heat transfer coefficient of air is 2.5
25 Wm22K21. The forced convection heat transfer coefficient of air can reach 10
500 Wm22K21[19]. The air cooling PEMFC stack has two typical configurations. First, the cooling air and reactant air are transported using separate channels. The cooling air can either be supplied from the fans, the blowers, or share the compressor with reactant air, as shown in Fig. 12.4. Second, the cooling air uses the flow channels of reactant air. During the operation of PEMFC, only a small portion of the air that passes by the cathode reaction sites is involved in the electrochemical reaction. Others are expelled as exhaust gas. However, as calculated previously, the heat removal capability of air is low. If you want cooling air to remove more than half of the waste heat, the stoichiometric of air needs to be increased to about 20 [1], which will increase the potential of membrane drying out. Other drawbacks of air cooling include the relatively low convection heat transfer coefficient of air; as a result, a larger heat exchange surface area is required to provide adequate cooling. What’s more, the temperature variation of cooling air between inlet and outlet is relatively large, which will result in nonuniform temperature and current density distributions. In addition, when cooling air is supplied by fans, different locations on the windward side have different airflow velocities. In certain circumstances, this temperature variation can be up to 10 C [20].

FIGURE 12.4 Air cooling with separate channels.

350

Fuel Cells for Transportation

12.3.1.2 Heat spreader cooling Heat spreader cooling, namely, edge cooling, is another typical passive cooling method employed in the PEMFC cooling system. The common materials of heat spreaders are graphite, pyrolytic graphite, copper, and other metals with high thermal and electric conductivities [21,22]. The heat will be transferred from the center region of the heat spreader to its edges via conduction and dissipate to the ambient. In practice, forced convection of air or other cooling liquid will be used to help remove the heat from the heat spreader. The heat spreader will be sandwiched between adjacent cells, but their interval depends on special needs. According to the assembling compression pressure, there may be large interfacial resistance arising between BPs and heat spreaders, having the potential of hot spot formation. Due to the mechanism of the heat spreader, there must be a temperature gradient between the central and edge regions, which is greatly dictated by the thermal properties of the material and the geometry of the fuel cell and the heat spreaders. This temperature gradient will result in different electrochemical reaction rates spatially. So, the design of the configuration should be carefully optimized before integrating heat spreaders in a PEMFC stack [23]. 12.3.1.3 Liquid cooling For the application of a PEMFC stack in a high-power ( . 5 kW) system which has stricter requirements for power density and temperature adaption, liquid cooling is the most proper method. Compared to air, liquid coolant has a higher heat capacity. Compared to air cooling, liquid cooling has better heat transfer capability, better environmental adaption, and more uniform temperature distribution [24]. Up to now, all fuel cell vehicles (FCVs) are equipped with liquid cooling systems [1]. As shown in Fig. 12.5, the flow channels for liquid coolant are usually manufactured between the cathode and anode BPs of two adjacent cells in a PEMFC stack. The coolant flow channels provide a heat exchange place for liquid coolant to remove the waste heat from fuel cells via forced convection. One of the major drawbacks of liquid cooling is the extra components included in its system, like coolant pipes, radiator, valves, deionizing filter, and coolant pump, which increases the cost and difficulty of system integration. Moreover, limited to the optimal working temperature of PEMFC, the temperature of the exhausted liquid coolant is generally lower than 75 C, which has a limited temperature difference with ambient. Under identical output power, the volume of the radiator employed in a PEMFC system is much larger than that in an internal combustion engine system [25]. Additionally, to keep a uniform current distribution, the change in temperature between coolant inlets and outlets needs to be minimized. In order to have adequate cooling, less change in temperature means a higher flow rate of liquid coolant. In most cases, the cross-sectional area of the coolant

Heat transport and thermal management Chapter | 12

351

FIGURE 12.5 Illustration of a design of coolant flow channels in a PEMFC with baffled cathode flow channels.

channel is relatively small. As a result, more power input to the pump is necessary for the liquid coolant with a higher flow rate to overcome the resistance created by the small coolant channel. The most important is the requirement of low ionic concentration in the coolant, which has a negative impact on the safe operation of PEMFC. The most adopted liquid coolants used in PEMFC are water, engine oil, the mixture of water, and ethylene glycol. If water is chosen to be the coolant, deionized water must be used. During the operation of PEMFC, the ionic contamination from metallic BPs will increase the ionic concentration in the coolant. When a certain amount of ionic concentration is achieved, the liquid coolant will become electric conductive, which will lower the energy conversion efficiency of the fuel cell stack. The electrolytic reaction will take place in coolant, producing oxygen and hydrogen and arising safety issues during operation. Picking the mixture of deionized water and ethylene glycol as coolant can lower the freezing point temperature to 235.6 C and increase the electronic resistance. However, the thermal conductivity is decreased as well. Also, the oxidation reaction of glycol will produce ions. Once the

352

Fuel Cells for Transportation

mixture of deionized water and ethylene glycol is contaminated by ions, its deionization process will be more difficult than that of water.

12.3.1.4 Nanofluids cooling Thermal conductivity is an important factor in judging heat transfer capability. For instance, the thermal conductivities of single-wall and multiwall carbon nanotubes are 3500 Wm21K21 and 3180 Wm21K21, respectively, [26]. As shown in Table 12.2, compared to the common liquid coolant, some metals and their oxide have higher thermal conductivities. Under room temperature, the thermal conductivities of carbon nanotube and copper are about 5188 and 654 times greater than that of water and are about 12619 and 1591 times greater than that of ethylene glycol, respectively. Therefore, adding ultrafine nanosized particles into the base fluid is expected to have boosted thermal conductivity. The application of nanotechnology in the thermal management of PEMFC is a novel concept. The definition of nanofluids is a suspension of nanoparticles in a base fluid [26]. The base fluid is basically chosen from the liquid coolants used in PEMFC cooling, namely, deionized water, the mixture of deionized water and ethylene glycol solution, and engine oil. The nanoparticles can be metallic, nonmetallic, or carbon nanotubes. Due to the characteristics of nanoparticles, like small sizes and large specific surface areas, dispersing nanofluids in liquid coolants can significantly increase the thermal conductivity (15% to 70% depending on the volumetric fraction of nanofluids in the suspension), fluid viscosity, convection heat transfer coefficient, and thermal diffusivity [27
29]. The enhanced thermal properties can reduce the size of the radiator, the weight of the system, and the parasitic loss from the coolant pump. Although the mixture of deionized water and ethylene glycol has lowered the freezing point to an ideal value, its heat conductivity has also been reduced. However, by adding nanofluid, the freezing point of the mixture of deionized water and ethylene glycol can be further decreased to about 240 C with its heat conductivity increased. Another advantage of nanofluids is their low electric conductivity. Nanoparticles can capture the free ions generated from metallic BPs and immobilize them to make them neutralized. Immobilized ions will no longer contribute to the electric conductivity. Both cations and anions can be attracted and captured by nanoparticles. This process keeps running until the saturation of ions is reached. The main drawback is the difficulty of preparing a stable homogeneous suspension of nanofluids and their storage [30,31]. 12.3.1.5 Phase-change cooling Single-phase cooling method, like liquid coolant cooling and air cooling, is a way to remove the waste heat by using the sensible heat of the coolant. Phase-change cooling utilizes the latent heat of phase change of the coolant to remove heat. The energy associated with the latent heat of phase change

TABLE 12.2 Thermophysical properties of selected materials at 300K [9]. Material

Thermal conductivity (Wm  K21 )

Specific heat (Jkg  K21)

Density (kg  m23 Þ

Silver

429

235

10500

Copper

401

385

8933

Aluminum

237

903

2702

Silicon

148

712

2330

Aluminum oxide

46

765

3970

Titanium oxide

8.4

710

4157

Silicon oxide

1.38

745

2220

Water (liquid)

0.613

4179

997

Ethylene glycol

0.252

2415

1114.4

Engine oil

0.145

1909

884.1

354

Fuel Cells for Transportation

is orders of magnitude greater than that of sensible heat. For instance, under standard atmospheric pressure, the water’s latent heat of vaporization is 2257 kJ kg21, which is more than 1000 times greater than the sensible heat of rising the temperature of water vapor by 1 C, which is 2.02 kJ kg21 [32]. Hence, phase-change cooling is capable of providing a considerable amount and efficient heat removal, and only a small amount of coolant is enough to provide adequate cooling. But it may have a concern of liquid coolant leakage, two-phase flow instability, and corrosion. The representative applications of phase-change cooling in the PEMFC stack include evaporative cooling, flow boiling cooling, heat pipe cooling, and phase-change materials cooling, etc., [33]. The definition of evaporative cooling is to remove waste heat through the evaporation of liquid coolant. The liquid coolant used in evaporative cooling should have a boiling temperature higher than the operating temperature of PEMFC, B80 C. Deionized water is often chosen. Generally, by introducing deionized water into the cathode airflow channels, reactant gas will be humidified, and the membrane water content will increase. Most of the deionized water injected into the flow channels will be brought to the reaction sites by air and then evaporates, absorbing the heat generated by the electrochemical reaction and keeping the temperature of the reaction region in a desirable range. The evaporated water vapor, along with some of the product water, is then expelled from fuel cells, condensed in an external condenser, and stored in a water reservoir for further use. Since humidification of the reactant gases and cooling happens simultaneously within the fuel cell, an external humidifier can be eliminated from the system [33,34]. Flow boiling cooling is another kind of phase-change cooling, which has been extensively applied in aerospace engineering. Different from evaporative cooling, the coolant liquid selected for flow boiling cooling has a boiling point lower than the operating temperature of PEMFC but higher than the ambient temperature. During the flow boiling cooling process, the temperature of the coolant maintains its boiling point and removes waste heat by utilizing the energy absorbed from vaporization. Hence, it can fulfill the requirement of uniform temperature distribution at the electrochemical reaction region. Similar to traditional liquid coolant cooling, flow boiling cooling needs separate coolant channels. According to the strict requirements of the thermal properties, the coolant candidates should possess the characteristics including saturation temperature to be 10 C
20 C less than 80 C and high latent heat of vaporization [35]. HFE-7100, a segregated hydrofluoroether, is considered a qualified candidate used in flow boiling cooling for PEMFC. HFE-7100 is advantageous in its thermal properties (at 1 atm, boiling point 61 C, and freezing point 2135 C), chemical stability, electric insulation, substantial latent heat of vaporization (112 kJ kg21 at 1 atm), and environmentally friendly [36]. Other selections of coolant include Forane 365 HX, a mixture of HFC-365mfc and HFC-4310mee, etc., [37].

Heat transport and thermal management Chapter | 12

355

BP MEA

Current collector End plate

Co nd sec ense tio r n

H ou eat tpu t

Ev ap sec orato tio r n

Envelope Wick

He inp at ut

Vapor flow Liquid flow

FIGURE 12.6 Illustration of a design of heat pipes cooling in PEMFC and its working principle.

Heat pipe cooling is an effective and robust phase-change cooling method. As shown in Fig. 12.6, heat pipes made of copper or aluminum alloy, filled with an adequate amount of liquid coolant, are embedded into the BPs of PEMFC [38]. The macroscopic working principle of heat pipe cooling is that waste heat generated from PEMFC will be absorbed in the evaporator section of the heat pipe, then be transported by vaporized coolant to the condenser section, and be rejected to ambient. Microscopically, inside the heat pipes, the liquid coolant evaporates at the evaporator section, absorbing heat from outside; then, the vaporized coolant migrates to the end of the heat pipes via the hollow part in the center of the heat pipe and condenses at the condenser section, releasing heat. Later, the condensed liquid coolant will move back to the evaporator section via the wick by utilizing capillarity, gravity, or forced movement by an external pump, forming a circulation. According to the requirement of working fluid coolant circulation, it can be designed as active or passive cooling. For the passive one, the ancillary cooling system is no longer needed, and parasitic losses will be reduced. Mini- or micro-passive heat pipe cooling is suitable for PEMFC stacks less than 100 W. Sorption and loop-pulsating heat pipes are usually implemented in PEMFC stacks greater than 100 W. Theoretically, phase-changing materials (PCMs), as energy storage materials, can perfectly control the temperature of PEMFC and recycle the waste

356

Fuel Cells for Transportation

heat, optimizing energy utilization. The working principle of PCMs is employing the latent heat of phase change to achieve the storage and reuse of energy, mitigating the spatially and temporally mismatching of energy between the supplier and demander. According to the types of phase change, PCMs can be classified into solid
solid, solid
liquid, solid
vapor, and liquid
vapor PCMs. Since the changes in volume and pressure of vaporization are relatively large, the PCMs used for PEMFC cooling are generally solid
solid and solid
liquid. According to the chemical components, PCMs can be classified into organic and inorganic PCMs. According to the temperature of phase change, PCMs can be classified into high-temperature ( . 200 C), moderate-temperature (100 C
200 C), and low-temperature (,100 C) PCMs [39]. The criteria associated with the selection of PCMs for PEMFC cooling are proper melting temperature, high latent heat of phase change, high thermal conductivity, high specific heat, less change in density during phase changing, no supercooled phenomenon, and chemically stable. Low-temperature organic PCMs, like paraffin wax, is appropriate for deployment in PEMFC cooling. The application of PCMs in the cooling of fuel cells is still at a preliminary stage. PCMs are advantageous in uniformity of temperature distribution, energy conservation, simple configuration, and low maintenance cost [40]. Due to the limited and narrow spaces in the PEMFC stack, entire waste heat can barely be removed by PCMs solely. However, the utilization of PCMs for waste heat recovery and partial removal is promising.

12.3.2 Thermal management subsystem The objectives of the thermal management subsystem of PEMFC are to regulate and optimize the heat transfer process, mitigate waste heat reduction, increase energy utilization, and, ultimately, enhance the performance of PEMFC. Besides maintaining a stable working temperature, reducing the temperature difference between single cells is another important objective for the thermal management subsystem as well. As shown in Fig. 12.7, the components in a representative liquid cooling system include coolant pump, coolant tank, thermostat, fan, positive temperature coefficient (PTC), deionization filter, and coolant pipeline. There are usually three heat exchangers in the system, which are the intercooler for the cathode, the heat exchanger for the anode, and the main radiator to dissipate heat [5,6]. The function of the coolant pump is to drive liquid coolant to circulate in the pipeline. The response of the coolant pump to the change in heat production rate from the PEMFC stack is adjusting the coolant flow rate. Commonly used types of pumps include centrifugal, rotary, and rotary volume. The prevalently used is the centrifugal coolant pump, which possesses characteristics like simple structure, lightweight, high lift and head.

Heat transport and thermal management Chapter | 12

357

Fan

Radiator

Thermostat PTC Deionization filter

Intercooler

Coolant tank

PEMFC stack

Heat exchanger

Coolant pump

FIGURE 12.7 The layout of a representative PEMFC thermal management subsystem.

In most cases, the air is compressed and supplied to the cathode of PEMFC. This compression process lifts the temperature of reactant air. The intercooler, a device that is used to cool down the temperature of the air to a range that is favorable by PEMFC before it enters the humidifier, is placed downstream to the compressor. The principle of the intercooler is to exchange heat between air and coolant. The role of the cooling fan is to provide adequate ventilation for the forced convective heat removal of the radiator. Under subfreezing temperatures, PEMFC encounters challenges from the cold start. During the cold start, PTC endeavors to warm up the coolant to a desirable temperature as quickly as possible, in order to shorten the time of cold start. Hydrogen is stored in the pressurized gaseous hydrogen tank. A regulating valve is generally used to control the flow of hydrogen. Assuming adiabatic expansion, the temperature of hydrogen will drop dramatically. Then, the anode heat exchanger is employed to heat up the hydrogen before it enters the humidifier. As mentioned, during the operation of PEMFC, metallic BPs and some devices from the thermal management subsystem will release ions and contaminate the coolant. As a result of increased ionic concentration, the electric conductivity of coolant increases, performance degradation happens, and safety issues arise. The deionization filter is a device used to alleviate this phenomenon. Deionization is a chemical process of removing cations and

358

Fuel Cells for Transportation

anions by reaction with cation exchange resin and anion exchange resin, respectively. The cations and anions in the coolant will be replaced by the H1 and OH2 from the ion exchange resin. A deionization filter can maintain the coolant at a high level of electric insulation. The thermostat can control the amount of coolant that flows through the radiator, according to the temperature of the coolant. The commonly used thermostats include traditional paraffin wax, electronic, and motor. They all share similar working principles. Take the paraffin wax thermostat as an example, when the temperature of the coolant is low, in order to reach the desired temperature, the thermostat will lead the coolant not to pass through the radiator, forming a small circulation loop. When the load of the PEMFC stack increases, as a result, the heat production rate and the temperature of coolant increase. The paraffin inside the thermostat absorbs heat and melts. The thermostat will open due to the squeeze from the expanding volume of paraffin wax. Part of the coolant goes through the radiator and exchanges heat. When the temperature of the coolant further increases, the thermostat will wide open, leading all the coolant to the radiator, forming a big circulation loop. When the temperature decreases, paraffin wax solidifies and its volume decreases. The thermostat will close. The final stage of the heat removal process is the heat dissipation from the radiator to the environment. It is important to select a radiator with ample capability of heat exchange. A typical radiator consists of water inlet chamber, water outlet chamber, and core. The coolant exits from the PEMFC stack, enters the water inlet chamber in the radiator, and then separately enters different heat dissipation pipes. The heat exchange inside the radiator comprises two processes: the heat exchange between coolant and radiator, and heat exchange between radiator and ambient. The heat carried by hightemperature coolant, first, is convectively transferred from the coolant to the inner walls of the radiator; then, conductively transferred from inner walls to outer walls and fins; and, consequently, convectively transferred to the ambient. The above components are connected by the pipeline, which has strict requirements of cleanness and electric non-conductivity, forming a whole coolant circulation loop. The effect of cooling is affected by the coolant flow rate, structural design of the radiator, and speed of the fan. A greater coolant flow rate means less change in temperature of the coolant, accordingly, less temperature difference of single cells. However, the concomitantly increased energy consumption of the coolant pump will influence the overall efficiency of the thermal management subsystem.

12.3.3 Control strategy The system of FCVs is a typical multiinput, multioutput, nonlinear, coupled, and complex system. Many variables need to be controlled simultaneously.

Heat transport and thermal management Chapter | 12

359

The common control strategies applied in the thermal management of FCVs include Proportional
integral
derivative (PID) control, Model predictive control (MPC) control, adaptive control, fuzzy control, robust control, and artificial neural network (ANN) control.

12.3.3.1 Proportional
integral
derivative control PID control is the most common control algorithm. Due to its robust performance in a wide range of operating conditions and its simple algorithm, PID control has been universally adopted in industrial control. The working principle of PID control is straightforward. The reading of the process variable from a sensor is the input; then, by adjusting the proportional coefficient, integral coefficient, and derivative coefficient, a satisfactory actuator output can be calculated. The proportional coefficient is dictated solely by the error, which is the difference between the process variable, for instance, PEMFC stack real-time temperature, and its set point, 80 C. The ratio of output response to the error is determined by the proportional gain. An increased proportional gain will increase the speed of response, but the controlled variable is also prone to oscillate. The function of the integral coefficient is to eliminate the steady-state error. The derivative coefficient is proportional to the rate of change of the process variable, which can reduce the response time and the trend of overshoot [41]. PID control is robust, simple, and flexible. However, the effect of PID control applied in PEMFC thermal management, whose system is nonlinear, time-varying, and coupled, may not be good enough. 12.3.3.2 Model predictive control MPC, namely, predictive control, is one of the optimal control algorithms that have been successfully used in the industry. The control process consists of a predictive model, rolling optimization, and feedback correction. MPC uses a predictive model of the system to forecast the future output based on past information, showing the future behavior of the system. It emphasizes the function of the model, not the structure of the model. MPC is suitable for complex systems with dynamic changes and uncertainties, which has difficulty building a precise model [42]. A PEMFC system possesses hysteresis, nonlinearity, and uncertainty. MPC can deal with multiinput, multioutput systems that may have interactions between the inputs and outputs. Compared to PID control, MPC is advantageous in reduced time to steady state, less oscillation, and stronger robustness. In practice, the time spent on online or onthe-fly calculations is relatively long. 12.3.3.3 Adaptive control Adaptive control can automatically compensate for variations in system dynamics by adjusting the controller characteristics so that the overall system

360

Fuel Cells for Transportation

performance remains the same or rather maintained at the optimum level. The adaptive control system includes elements to measure the process dynamics and other elements to alter the controller characteristics accordingly [43]. Based on the fundamentals of adaptive control, the adaptive inverse control can automatically trace the dynamic change in PEMFC and achieve optimal control of stack temperature according to the time-varying parameters of the PEMFC thermal management system. Adaptive control simplifies the computation and does not need complex nonlinear modeling. Compared to PID control, adaptive control is more favorable to the stable operation of PEMFC because of the less voltage perturbation. It can achieve the optimal control of temperature under various loads.

12.3.3.4 Fuzzy control Fuzzy control, or fuzzy logic control, usually decomposes a complex system into several simple subsystems based on the intuition or understanding of the designer regarding the controlled system. Fuzzy control does not need an accurate mathematical model of the controlled subsystem, the input can be vague, and it is more robust to disturbance than other traditional nonlinear controllers. A set of straightforward control rules are designed for each subsystem. Combining all the local control actions through fuzzy membership functions, a global control rule is built. The working principle of fuzzy control includes the fuzzification of inputs, execution of control rules, and defuzzification of the outputs. Fuzzy control usually outperforms other controllers in complex, nonlinear, and undefined systems. When controlling the temperature of the PEMFC stack by adjusting the coolant flow rate and speed of the cooling fan, since an accurate mathematical model is unnecessary for fuzzy control, it can usually better execute the control of temperature [44,45]. 12.3.3.5 Robust control Robustness is the capability of keeping the performance and parameters unchanged under external disturbances and uncertainties. Robust control is meant to focus on the robustness of the control algorithm. The design of a robust control system is typically based on the worst-case scenario. Robust control methods are well suited to applications where system stability and reliability are top priorities, where process dynamics are known, and variation ranges for uncertainties can be estimated [42]. 12.3.3.6 Artificial neural network An ANN is an information processing system developed to imitate the structures and features of the human brain. This makes it able to deal with complex logical operations. ANN establishes corresponding mathematical models based on numerous experimental data, which are subsequently adopted for

Heat transport and thermal management Chapter | 12

361

control purposes. ANN is highly favored owing to multiparallel processing and strong adaptiveness, which is capable of approximating complex nonlinear mapping with good precision [6].

12.3.4 Cold start The cold start of PEMFC, namely, the capability of successful starting up from a subfreezing temperature, is one of the most important issues that need to be addressed before the commercialization of FCVs. Under subfreezing temperatures, water will freeze. The electrochemical reaction sites and porous electrodes will be covered by ice, decreasing the electrochemical active surface area and hindering the transport of reactant gases. The microporous structures in electrodes are prone to be destroyed by the expansion of the volume of water during the phase change. In PEMFC, the ice formation mechanism is complicated and affected by many factors, including the volume fraction of ionomer, startup temperature, porosity, and wettability of the CL. When the temperature reaches the freezing point, part of the membrane water will still be unfrozen. Nafion still possesses a certain extent of ionic conductivity; however, it is much lower than that under normal temperature. In order to improve the cold-start capability, several techniques have been massively researched and tested at the single-cell level, for instance, purging after shutdown, novel fuel cell designs, and novel cold-start strategies. At the system level, the improvement of cold-start capability is usually achieved by means of thermal management, including reactant gases preheating, stack heating, coolant heating, and external hydrogen burner. Although, in practice, some FCVs are already capable of successfully starting up from 230 C, the time of cold start and the number of auxiliary heating devices still need to be further reduced. The mechanism and location of ice formation, the mechanism of supercooled water formation, and the impact of ice on oxygen reduction reaction all need to be deeper understood.

12.4 Summary The higher operating temperature can facilitate the electrochemical reaction kinetics, the proton conductivity, and the diffusion of reactants, which is favorable for a fuel cell. However, high temperature, low relative humidity, and/or high air stoichiometric ratio can also make the proton exchange membrane prone to dehydrate, which will consequently increase the proton conduction resistance. Another potential reason that may lead to membrane dehydration is the electroosmotic drag, which is worsened at high current densities. When the operating temperature is lower than 60 C, flooding is more likely to happen at the electrodes. Flooding is caused by liquid water accumulation in the pores of electrodes. As a result, reactant gases transport will be hindered. The flooding is dominated by the phase change of liquid

362

Fuel Cells for Transportation

water and water vapor, which is significantly affected by the temperature. Moreover, the through-plane temperature variance also has an appreciable impact on the membrane water content and current density uniformity. Hence, it is important to maintain the operating temperature between 60 C and 80 C and uniformly distributed within a fuel cell. The heat sources in a PEMFC include the reversible and irreversible heat of electrochemical reaction, the Joule heat, and the latent heat of phase change. The reversible heat, namely, entropic heat, represents the reaction rate and the change in entropy of the electrochemical reactions, which accounts for nearly 30% of the total heat. The irreversible heat emanates from the activation overpotential, which contributes about 60% of the total heat. The Joule heat is generated from ionic and electronic resistances. Lastly, the latent heat is associated with the phase change of water. Since radiation heat transfer can be neglected, the major heat transfer modes in a PEMFC are conduction and convection. Conduction heat transfer takes place in all PEMFC components, and convection heat transfer occurs in the porous electrode materials but mainly in the flow channel. The waste heat transports from the highest temperature region, namely, the MEA to BPs. Since only a few amounts of waste heat are removed by reactant gases, which is negligible, the major portion of waste heat will be dissipated from the surface of BPs, where the assistance of the thermal management subsystem is needed. Due to the low heat capacity of air, passive cooling methods, for instance, air cooling and heat spreader cooling, usually can only deal with a PEMFC stack with a power of less than 5 kW. For the application of PEMFC stack in a high-power ( . 5 kW) system which has stricter requirements for power density and temperature adaption, liquid cooling is the most proper method. Other cooling methods, like nanofluids cooling, heat pipe cooling, and PCM cooling, are all promising and under development. The components of a thermal management subsystem are illustrated, and their functions are elucidated. Different control strategies applied in the thermal management subsystem, including PID control, MPC control, fuzzy control, and ANN control, are introduced. Their corresponding algorithm, pros, and cons are well explained. The cold start of PEMFC, which is the capability of successful starting up from a subfreezing temperature, is one of the most important issues that need to be addressed before the commercialization of FCVs. Several thermal management techniques have been widely studied and tested to ensure a successful cold start. These techniques include reactant gas preheating, coolant heating, stack heating, and external hydrogen burner. The mechanism and location of ice formation, the mechanism of supercooled water formation, and the impact of ice on oxygen reduction reaction all need to be further studied. In summary, the thermal management subsystem not only affects the system’s performance but also significantly influences the durability and lifespan of the system. Keeping the stack and system components at proper operating temperature, mitigating both the in-plane and through-plane

Heat transport and thermal management Chapter | 12

363

temperature variance inside the stack, and enhancing cold-start capability are all critical issues that the thermal management subsystem needs to deal with. Therefore, performing thermal management more effectively and efficiently is of paramount importance.

Nomenclature A α C cp δ Er ε ε F h h h I ~i ion ~ ie j k f κef ion ef f κe Nu Pr ρ q qirrev qrev qohm;H1 qohm;e2 qphase Re S Sd2l Sv2l ST T u W σ ηact ϕ ξ

area thermal diffusivity concentration specific heat thickness reversible voltage emissivity porosity Faraday’s constant enthalpy latent heat of phase change convection heat transfer coefficient current density ionic conductivity electronic conductivity net current density thermal conductivity effective ionic conductivity effective electronic conductivity Nusselt number Prandtl number density heat transfer rate irreversible heat reversible heat ionic Joule heat electronic Joule heat heat from phase change Reynolds number entropy source term of membrane water source term of phase change source term of heat temperature velocity width Stefan
Boltzmann constant activation overpotential phase potential stoichiometric

364

Fuel Cells for Transportation

References [1] Q. Chen, et al., Thermal management of polymer electrolyte membrane fuel cells: A review of cooling methods, material properties, and durability, Appl. Energy 286 (2021) 116496. [2] M. Ghasemi, et al., A numerical study on thermal analysis and cooling flow fields effect on PEMFC performance, Int. J. Hydrog. Energy 42 (38) (2017) 24319
24337. [3] T.J.P. Freire, E.R. Gonzalez, Effect of membrane characteristics and humidification conditions on the impedance response of polymer electrolyte fuel cells, J. Electroanal. Chem (2001). [4] J. Wang, H. Wang, Y. Fan, Techno-economic challenges of fuel cell commercialization, Engineering 4 (3) (2018) 352
360. [5] Y. Wang, K.S. Chen, PEM Fuel Cells: Thermal and Water Management Fundamentals, Momentum Press, 2013. [6] K. Jiao, et al., Water and Thermal Management of Proton Exchange Membrane Fuel Cells, Science Press, Beijing, 2021. [7] S.G. Kandlikar, Z. Lu, Thermal management issues in a PEMFC stack—a brief review of current status, Appl. Therm. Eng. 29 (7) (2009) 1276
1280. [8] S.G. Goebel, Evaporative cooled fuel cell, Google Patents, 2005. [9] T.L. Bergman, et al., Fundamentals of Heat and Mass Transfer, John Wiley & Sons, 2011. [10] X. Li, Principles of Fuel Cells, CRC Press, 2005. [11] Y. Wang, et al., Materials, technological status, and fundamentals of PEM fuel cells
a review, Mater. Today 32 (2020) 178
203. [12] Y. Wang, et al., A review of polymer electrolyte membrane fuel cells: Technology, applications, and needs on fundamental research, Appl. Energy 88 (4) (2011) 981
1007. [13] A. Ozden, et al., A review of gas diffusion layers for proton exchange membrane fuel cells—with a focus on characteristics, characterization techniques, materials and designs, Prog. Energy Combust. Sci. 74 (2019) 50
102. [14] J.-P. Maes, S. Lievens, Methods for fuel cell coolant systems, Google Patents, 2007. [15] M.C.L. de Oliveira, G. Ett, R.A. Antunes, Materials selection for bipolar plates for polymer electrolyte membrane fuel cells using the Ashby approach, J. Power Sources 206 (2012) 3
13. [16] B. Andreaus, A. McEvoy, G. Scherer, Analysis of performance losses in polymer electrolyte fuel cells at high current densities by impedance spectroscopy, Electrochim. Acta 47 (13
14) (2002) 2223
2229. [17] W. Li, et al., Theoretical and experimental analysis of the lithium-ion battery thermal runaway process based on the internal combustion engine combustion theory, Energy Convers. Manag. 185 (2019) 211
222. [18] M. Ramezanizadeh, et al., A review on the approaches applied for cooling fuel cells, Int. J. Heat. Mass. Transf. 139 (2019) 517
525. [19] M. Assad, M.A. Rosen, Design and Performance Optimization of Renewable Energy Systems, Academic Press, 2021. [20] Bu Qingyuan, et al., Experimental and simulation of cathode fan system of air-cooling PEMFC, Ciesc J. 66 (10) (2015) 4211
4217. [21] R.H. Horng, et al., Cup-shaped copper heat spreader in multi-chip high-power LEDs application, Opt. Express 20 (105) (2012) A597
A605. [22] H.Q. Nguyen, B. Shabani, Proton exchange membrane fuel cells heat recovery opportunities for combined heating/cooling and power applications, Energy Convers. Manag. 204 (2020) 112328.

Heat transport and thermal management Chapter | 12

365

[23] C.-Y. Wen, et al., Thermal management of a proton exchange membrane fuel cell stack with pyrolytic graphite sheets and fans combined, Int. J. Hydrog. Energy 36 (10) (2011) 6082
6089. [24] E.J. Choi, et al., Parametric analysis of simultaneous humidification and cooling for PEMFCs using direct water injection method, Int. J. Hydrog. Energy 42 (17) (2017) 12531
12542. [25] S. Rogg, et al., Cooling modules for vehicles with a fuel cell drive, Fuel Cell 3 (3) (2003) 153
158. [26] M.R. Islam, et al., The potential of using nanofluids in PEM fuel cell cooling systems: a review, Renew. Sustain. Energy Rev. 48 (2015) 523
539. [27] L. Godson, et al., Enhancement of heat transfer using nanofluids—an overview, Renew. Sustain. energy Rev. 14 (2) (2010) 629
641. [28] Choi, S.U.-S., Nanofluid technology: current status and future research, Argonne National Lab.(ANL), Argonne, IL (United States), 1998. [29] S. Murshed, K. Leong, C. Yang, Thermophysical and electrokinetic properties of nanofluids
a critical review, Appl. Therm. Eng. 28 (17
18) (2008) 2109
2125. [30] I. Zakaria, et al., A review of nanofluid adoption in polymer electrolyte membrane (PEM) fuel cells as an alternative coolant, J. Mech. Eng. Sci. 8 (2015) 1351. [31] M. Hemmat Esfe, M. Afrand, A review on fuel cell types and the application of nanofluid in their cooling, J. Therm. Anal. Calorim. 140 (4) (2020) 1633
1654. [32] M.J. Moran, et al., Fundamentals of Engineering Thermodynamics, John Wiley & Sons, 2010. [33] A. Fly, R. Thring, A comparison of evaporative and liquid cooling methods for fuel cell vehicles, Int. J. Hydrog. Energy 41 (32) (2016) 14217
14229. [34] A. Fly, R. Thring, Temperature regulation in an evaporatively cooled proton exchange membrane fuel cell stack, Int. J. Hydrog. Energy 40 (35) (2015) 11976
11982. [35] E.J. Choi, J.Y. Park, M.S. Kim, Two-phase cooling using HFE-7100 for polymer electrolyte membrane fuel cell application, Appl. Therm. Eng. 148 (2019) 868
877. [36] E.J. Choi, J.Y. Park, M.S. Kim, A comparison of temperature distribution in PEMFC with single-phase water cooling and two-phase HFE-7100 cooling methods by numerical study, Int. J. Hydrog. Energy 43 (29) (2018) 13406
13419. [37] U. Soupremanien, et al., Tools for designing the cooling system of a proton exchange membrane fuel cell, Appl. Therm. Eng. 40 (2012) 161
173. [38] M.V. Oro, E. Bazzo, Flat heat pipes for potential application in fuel cell cooling, Appl. Therm. Eng. 90 (2015) 848
857. [39] Y. Zhao, et al., Research progress of phase change cold storage materials used in cold chain transportation and their different cold storage packaging structures, J. Mol. Liq. 319 (2020) 114360. [40] A.P. Sasmito, T. Shamim, A.S. Mujumdar, Passive thermal management for PEM fuel cell stack under cold weather condition using phase change materials (PCM), Appl. Therm. Eng. 58 (1
2) (2013) 615
625. [41] National_Instruments. PID theory explained. ,https://www.ni.com/zh-cn/innovations/ white-papers/06/pid-theory-explained.html., 2020. [42] P. Zhang, Advanced Industrial Control Technology, William Andrew, 2010. [43] A. Gupta, D.S. Yan, Mineral Processing Design and Operations: An Introduction, Elsevier, 2016. [44] M. Ahmadian, Active control of vehicle vibration, 2001. [45] J.F. Silva, S.F. Pinto, Linear and nonlinear control of switching power converters,, Power Electronics Handbook, Elsevier, 2018, pp. 1141
1220.

This page intentionally left blank

Chapter 13

Mass transport in the cathode Linhao Fan1, , Zhiming Bao1, , Daniela Fernanda Ruiz Diaz2, Yun Wang2 and Kui Jiao1 1

State Key Laboratory of Engines, Tianjin University, Tianjin, P.R. China, 2Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States

13.1 Mass transfer in cathode gas flow fields The flow field, also called the flow distributor or flow channel, is typically embedded in the bipolar plates. In the cathode of a proton-exchange membrane fuel cell (PEMFC), the inlet humidified air/oxygen is distributed in the flow field, diffuses into the gas diffusion layer (GDL), and finally reaches the reaction site in the catalyst layer (CL). Meanwhile, the liquid water produced by an oxygen reduction reaction (ORR), condensed from water vapor, and dragged by protons is transported from CL and GDL to the flow field and then removed by the gas flow. The cathode flow field must have a strong ability of air/oxygen distribution and water removal. Thus, it is quite important to understand and enhance the mass transfer in the cathode flow field.

13.1.1 Characterization of oxygen distribution and water removal Oxygen distribution and water removal are the two most concerning mass transfer processes in cathode flow fields. It is observed that the flow rate and gas pressure are nonuniformly distributed in the whole cathode flow channel [1]. Flow maldistribution of the gas reactant is the main reason for the mass transfer loss, which limits the operating current density of PEMFCs [2]. The mass transfer limitation is exacerbated for the PEMFC with a large active area and complex flow field plate, which is a common configuration for commercial fuel cell stacks [3]. Thus, the design of the flow channel/field has great significance for commercial stacks to achieve ultrahigh power density operation. 

Authors are equally contributed.

Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00014-9 © 2023 Elsevier Ltd. All rights reserved.

367

368

Fuel Cells for Transportation

On the other hand, the liquid water expelled from the GDL forms droplets and films at multiple positions in the flow channel/field, as shown in Fig. 13.1A. The accumulated liquid water could block the transport pathway of gas flow and damage reactant distribution. Fig. 13.1B illustrates the forces on a liquid droplet under gas flow. It is the gas pressure gradient, gas drag force, and wall adhesion force that determine the liquid dynamics [4]. The surface adhesion must be overcome so that the gas flow can blow the liquid drops away from the flow field. However, it is difficult to remove all the liquid water that is retained in the flow field because gas flow is nonuniformly distributed in the flow field. In addition, with liquid water constantly discharged from the GDL, most of the time the gas reactant is transported together with liquid water. The blockage of liquid water could induce a weak flow area in the flow field and even reactant starvation of PEMFC [5]. Hence, liquid removal is of great importance for the mass transfer of flow field. Since the liquid water dynamics in the flow field are influenced by gas flow rate and wall adhesion, the water removal is not only related to the structure of the flow channel/field, but also to the surface wettability, and operation conditions of PEMFC such as relative humidity and stoichiometry ratio.

13.1.2 Mass transfer in conventional flow channels Over the years, much effort has been put into the design of various flow fields. The most conventional ones are the parallel channel, serpentine channel, and interdigitated channel, as shown in Fig. 13.2. The parallel flow channel (Fig. 13.2A) has the simplest structure and the lowest pressure drop. The disadvantages of heterogeneous gas flow in different channels and weak

FIGURE 13.1 (A) Liquid distribution and (B) attachment on the GDL surface in a gas channel [6]. Reproduced from I.S. Hussaini, C. Wang, Visualization and quantification of cathode channel flooding in PEM fuel cells, J. Power Sources 187 (2) (2009) 444
451; S.C. Cho, Y. Wang, K. S. Chen, Droplet dynamics in a polymer electrolyte fuel cell gas flow channel: forces, deformation, and detachment. I: theoretical and numerical analyses, J. Power Sources 206 (2012) 119
128.

Mass transport in the cathode Chapter | 13

369

FIGURE 13.2 Flow characteristics in flow field and GDL with (A) parallel channel, (B) serpentine channel, and (C) interdigitated channel. Reproduced from K. Jiao, X. Li, Water transport in polymer electrolyte membrane fuel cells, Prog. Energy Combust. Sci. 37 (3) (2011) 221
291.

convective transport in the through-plane direction make it inefficient in both oxygen transfer and water removal. But on the other hand, the simple structure of parallel flow channels is easily manufactured and attractive to mass production. Thus, modifying the channel geometry of parallel flow channels is a promising approach to promoting mass transfer performance under the limit of manufacturing costs [7]. Further details about bipolar plates and flow field design for a PEMFC are available in Chapter 11. The serpentine flow channel (Fig. 13.2B) has better utilization of reactants compared with the parallel flow channel due to a longer flow path. The high-pressure drop along the mainstream helps to create convective flow across the channels, which contributes to oxygen transport to the GDL. However, introducing turns and corners to the flow field leads to a risk of waterflooding. Liquid water tends to accumulate at the corners of channels where the local gas flow rate is low and constantly covers the surface of GDL. Furthermore, the longer removal pathway of liquid could result in a frequent fluctuation of gas pressure because the liquid droplets could merge and block the channels at the outlet of flow channels. Thus, the parallel serpentine channel, rather than the single serpentine channel, is often used to maintain water management capacity. In addition, the shape of turns in some serpentine channels could be modified to avoid severe liquid accumulation at the corners of channels [8].

370

Fuel Cells for Transportation

Interdigitated flow channel (Fig. 13.2C) generally consists of two unconnected dead-ended channels. With the huge pressure difference between the channels, the gas reactant in the inlet-side channel is forced to flow through the GDL to the outlet-side channel by convection. This design can greatly promote the reactant transport to the CL and the liquid removal from the GDL to the outlet-side channel. However, the major issues that hinder the application of interdigitated channels are the high-pressure drop and strong water removal ability. The pressure drop of interdigitated channels is much higher than in other flow field designs. Even if the performance of PEMFCs is promoted, the increase in pumping loss may reduce the overall efficiency of the power system. On the other hand, the strong water removal ability could lead to membrane dehydration, especially at low humidity operation, which can cause severe degradation of cell durability and performance. Thus, it is better to operate PEMFCs with interdigitated channels at high relative humidity conditions [8]. In addition to flow field designs, the channel geometry parameters, such as width, height, and length, also play a role in oxygen distribution and water removal. Though controversial ideas exist about the effect of geometry parameters, it is acknowledged that a narrower channel/rib could promote the uniformity of gas flow and contribute to water removal due to the increased channel number and pressure drop [9]. A lower channel height is preferred for high-performance designs to improve gas flow rate and pressure drop, but it could also lead to maldistribution of oxygen if the height is much lower than the width. The channel length is reported to have different influences on different flow fields regarding water management. It is suggested to reduce the channel length of serpentine channels but improve that of parallel channels to avoid liquid accumulation. Furthermore, the channel cross section could also affect the mass transport of PEMFC. Rectangular cross sections are commonly used in conventional flow field designs which maintain a constant height throughout the active area. But investigations indicate that a trapezoidal cross section can enhance the mass transport between the flow field and GDL due to the larger contact area at the flow field/GDL interface [8]. The modification of the flow field cross section has recently raised attention for high-performance PEMFC design, which inspires innovations in convectional flow fields.

13.1.3 Mass transfer in novel flow fields To further enhance mass transfer in PEMFCs, quite a few studies have been conducted on the design of flow fields with complex structures. But only a small part of the proposed flow field is put into use in fuel cell devices. Considering the mass transfer performance as well as manufacturing ability, three types of new flow fields that have the potential for high-performance

Mass transport in the cathode Chapter | 13

371

PEMFC design are introduced, including modified parallel flow channel, baffle flow field, and porous material flow field, as displayed in Fig. 13.3. As aforementioned, the parallel channel has the advantages of lowpressure drop, little liquid accumulation, and high manufacturing ability. It will be a promising choice if the maldistribution of gas reactant in parallel flow channels could be overcome. To this end, the main target of the modified parallel channel is to improve the flow resistance and crossflow. Until now, two approaches of modification have been proven practical for automotive applications. The first one is changing the straight channels to wavy channels. The wavy channel shown in Fig. 13.3A has been proven effective in generating crossflow. With a trapezoidal cross section, the crossflow could be further guided to the under-rib area. Not only the distribution of the gas reactant but also the water discharge from GDL could be improved. The liquid accumulation at the turns of the flow channel, which is commonly observed in the serpentine channel, is also avoided since the angle of the turn is designed lower than 90 degrees. This design is put into use by Honda and Toyota in their fuel cell vehicles and proved superior to conventional designs under high-power density operation [10,11]. The other approach that could also improve the mass transfer of the parallel channel is changing the cross section of straight channels. The fact that the narrower channel has a

FIGURE 13.3 Novel flow fields for high-performance PEMFC design: (A) wavy flow channel, (B) partially narrowed straight flow channel, (C) 3D fine mesh flow field, and (D) metal foam flow field. Reproduced from Refs. [10
13].

372

Fuel Cells for Transportation

larger pressure drop and stronger liquid removal ability inspires this modification on the parallel flow channel. Fig. 13.3B exhibits a partially narrowed parallel flow field with varied channel cross sections, which is proposed by Toyota. Both the width and height of the channel are changed at the narrowed point of the flow channel, which increases local flow resistance and pushes gas into GDL. It is worth noting that the narrowed points are staggered in the flow field to enlarge the pressure difference between adjacent channels and to generate under-rib crossflow [11]. Except for changing cross sections of channels, it is also feasible to improve the flow resistance by adding baffles in the flow field. Attempts have been made to insert baffles in the flow channels, but the designed flow fields failed to achieve the target mass transfer capacity since the maldistribution of reactant among channels is not changed. However, the idea that the baffles with featured structures can promote mass transfer between the flow field and GDL is proven effective [14]. Thus, totally replacing the ribs with baffles will be a potential option as long as the baffles can conduct heat and electrons like the ribs. To this end, Toyota designed a 3D fine mesh flow field with all the baffles manufactured on a single titanium plate (Fig. 13.3C) [12]. The connected baffles manage to distribute gas flow, remove liquid, and conduct heat and electrons. The gas flow uniformity is largely improved due to the staggered arrangement of baffles, and liquid
gas separation is achieved by the special flow pattern in the flow field. Even though the mass transport target is achieved, this design is not applied in their next generation of fuel cell vehicles due to the high manufacturing cost and problems with corrosion resistance. For all the above flow field designs, one of the principles is to promote reactant distribution by setting a level of flow resistance between the flow field inlet and outlet. Hence, the gas reactant cannot flow directly to the outlet but is distributed by the featured structures to the whole active area. If this structure that provides flow resistance is not a rib, baffle, or block but a 3D skeleton that fills the whole space of the flow field, the gas distribution will be further promoted compared with any previous flow field. Fortunately, there is a special material with this desired property—foam material (Fig. 13.3D). Research on foam material flow field shows that it can well distribute the gas flow with a controllable flow resistance [15]. The stiffness and conductivity of foam material are considerable, and the manufacturing cost is acceptable. However, the major concern of the foam material flow field is water management since the porous structure can retain liquid droplets. Therefore, both high porosity and large pore size of the foam material are required. Compression and hydrophobic treatment are two approaches that could improve its mass transfer capacity [13,16]. Furthermore, the porous nature of foam material makes it possible to replace the GDL, in which scenario the foam material flow field is directly in contact with the CL. With this design, the transport resistance at the interface between the flow field and GDL could be eliminated, and the PEMFC stack could be

Mass transport in the cathode Chapter | 13

373

more compact [2]. But there remain problems that hinder the usage of foam material flow field, such as the variation of pore size from flow field to gas diffusion region and the poor contact between foam material and CL. In a word, the design of a novel complex flow field may provide a promising approach to the ultrahigh power density PEMFC.

13.1.4 Effects of gas diffusion layer and operation conditions In addition to the flow field structure, a few features unique to PEMFC also need to be considered, including GDL surface properties, GDL intrusion into the flow field, and operating conditions such as stoichiometry ratio, relative humidity. The wettability and roughness of the GDL surface mainly influence the growth and detachment of liquid droplets. The droplet is flatter with a hydrophilic and smoother surface, which needs a higher gas flow rate or larger droplet diameter to be removed [1]. The surface properties of GDL are determined by both the carbon paper/cloth and PTFE treatment. The intrusion of GDL into the flow field, caused by the compact pressure, may induce a maldistribution of gas flow among flow channels [1]. Furthermore, the mass transfer in the flow field is also related to the stoichiometry ratio which determines the total gas flux in the flow channels/fields. A high stoichiometry ratio means excessive reactant, high flow rate, large pressure drop, and strong liquid removal capacity in the flow field. The relative humidity affects both the mass fraction of oxygen and the evaporation of liquid in the flow field. High relative humidity of inlet gas can slightly reduce the oxygen diffusion from the flow field to GDL and evaporation of liquid, but largely promote membrane hydration. Therefore, it is also important to find the optimized operating condition for different types of flow field.

13.2 Mass transfer in cathode gas diffusion layer and microporous layer The GDL is a highly porous layer that bridges the flow field and CL. It functions as the transport pathway for oxygen, liquid water, electrons, and heat and provides mechanical support to the membrane and CL. To achieve the expected functions, two types of carbon fiber-based materials are adopted: carbon paper and carbon cloth. Carbon cloth is a woven fabric, while carbon paper is a nonwoven carbon composite [17]. However, carbon paper is more commonly used because it is easy to be applied with a microporous layer (MPL) to obtain better water management, oxygen distribution, and physical contact [18]. Thus, this section is presented with a focus on the GDL made of carbon paper and applied by an MPL. Carbon paper is a carbon
carbon composite that consists of carbon fibers and binders. During the manufacturing processes, the fibers are graphitized while the resin-based portion, that is, binder, remains as amorphous carbon. Thus, the fibers orient almost perpendicular to the thickness of carbon paper

374

Fuel Cells for Transportation

and result in the highly anisotropic structure of carbon paper. The anisotropic mass transfer properties are usually identified through the planes perpendicular and parallel to the thickness of carbon paper, which are also called “through-plane” and “in-plane,” respectively, for convenience. The throughplane transport is weaker than in-plane transport because of the larger transport resistance by fibers, for which reason the GDL can further distribute oxygen, water, and electrons [18]. In addition, the carbon paper is generally treated hydrophobic with PTFE to facilitate liquid water removal and compressed for better contact with the flow field and CL. The PTFE and compression both influence the anisotropic properties of GDL. In summary, the understanding of the mass transfer characteristics of GDL is essential for the design and modification of PEMFCs. The MPL, composed of carbon agglomerates, is usually integrated between GDL and CL to achieve a good contact and smooth interface between GDL and CL [19]. The MPL has a finer pore size than GDL and is believed to enhance the overall performance of PEMFC. However, the exact contribution of MPL to mass transfer is hard to estimate because it cannot be regarded as a stand-alone layer without GDL [18]. Thus, the effect of MPL is mostly demonstrated by a comparison between the bare carbon paper and carbon paper/MPL assembly.

13.2.1 Characterization of oxygen transport In operation, oxygen is constantly consumed in the cathode CL by the ORR, which causes a difference in oxygen concentration between the flow field and CL. Since the gas flow in the flow field is perpendicular to the surface of GDL which has a considerable through-plane flow resistance, the oxygen transport in GDL is dominated by the diffusion which is driven by the concentration gradient. Only a small amount of convection exists in GDL, such as the under-rib crossflow. The oxygen diffusion rate is influenced not only by the concentration gradient but also by the GDL structure and the existence of liquid water. To quantify the oxygen transport capacity of a GDL, the effective diffusivity is usually used which could be calculated from oxygen flux, concentration difference, and diffusion distance. In a dry GDL, the effective diffusivity is merely related to the structural parameters of GDL, including porosity, fiber diameter, and tortuosity. Specifically, the porosity of GDL is considered a key parameter that determines the oxygen diffusivity, while the other parameters have a minor influence [20]. Attempts have been made to investigate the correlation of effective diffusivity with structural parameters. Based on the effective medium approximation, Bruggeman’s correlation is widely used to calculate the effective gas diffusivity of porous material [21], written as: Deff 5 ε1:5 Dbulk

ð13:1Þ

Mass transport in the cathode Chapter | 13

375

where Deff is the effective diffusivity, Dbulk is the bulk diffusivity, and ε is the porosity. In the meantime, the general percolation theory was also used to obtain the correlations of effective diffusivity for porous material with a random distribution of fibers [22], such as:
 Deff ε2εp α 5 ð13:2Þ Dbulk 12εp where εp is the percolation threshold. However, these correlations were derived only for the transport of homogeneous porous material which cannot be applied to an isotropic material such as carbon paper. Recently, efforts have been devoted to understanding the isotropic properties of GDL. The fiber orientation of GDL has been considered in both experimental and numerical research, which yields more precise correlations of the throughplane and in-plane effective diffusivities [20], written as: 8   3ð1 2 εÞ > > > 1 2 2:76ε coshð3ε 2 1:92Þ through 2 plane > < 32ε Deff   ð13:3Þ 5 3ð1 2 εÞ Dbulk > > > 1 2 1:72ε coshð2:07ε 2 2:11Þ in 2 plane > : 32ε In addition, the compression of GDL has a negative influence on the effective diffusivity in both through- and in-plane directions. The in-plane diffusivity exhibits a larger decrease with compression because the distance between fibers is lowered along the through-plane direction which leads to a smaller in-plane pathway of gas [23]. Again, liquid water also influences the oxygen transport in GDL. The void space of GDL is partly occupied by liquid water, which further reduces the available transport pathway for oxygen. Since the oxygen diffusivity in liquid water is very small and neglectable, the space blocked by liquid water could just be regarded as a solid region. The occupation of liquid is different from solid fibers, but it is reported to have a comparable effect on the effective transport diffusivity of GDL. Thus, the effective diffusivity of a partially saturated GDL, called relative effective diffusivity, could be estimated by the correlation of a dry GDL that is corrected by liquid water saturation. This correction could be made on either the porosity to represent the nonoccupied space or the effective diffusivity by introducing a correction factor [24], for example,
 Deff ε20:11 0:785 εwet 5 εð1 2 sÞ or 5ε ð12sÞ2 ð13:4Þ 120:11 Dbulk where εwet is the equivalent porosity of a partially saturated GDL, and s is the liquid water saturation. Though the MPL can contribute to the overall performance of PEMFC, its effect on mass transport is negative. The

376

Fuel Cells for Transportation

through-plane effective diffusivity is reported to decrease by 39% in the presence of MPL [25]. Thus, it is necessary to limit the thickness of MPL to reduce the transport resistance of MEA.

13.2.2 Characterization of liquid water transport Excessive water generated by an ORR and electroosmosis drag in the cathode CL needs to be discharged. Most of the water is removed in the liquid phase because PEMFCs are operated at a relatively low temperature. Thus, the liquid removal in GDL is of great importance for oxygen transport as well as membrane hydration. The liquid removal in GDL can be divided into several stages. The liquid water emerges from the MPL surface, forms liquid clusters, and passes through the GDL, and then, the liquid is constantly discharged through the clusters in form of permeation, as shown in Fig. 13.4. The liquid removal and saturation in GDL are governed by both permeability and capillary pressure, which indicate the convective and diffusive contribution to overall liquid transport, respectively [18]. The permeability of a dry GDL, usually called intrinsic permeability, is independent of fluid but governed by the structure of the GDL. In general, the intrinsic permeability of a carbon paper GDL ranges from 10210 to 10213 m2. It is mainly related to the porosity and fiber diameter of the GDL. Efforts have been made to the correlation between intrinsic permeability and structural parameters, such as the anisotropic permeability of randomly overlapping fibrous structure proposed by Tomadakis and Robertson based on the Kozeny
Carman equation [26], expressed as: K 5 R2

εðε2εp Þα12 8ðln εÞ2 ð12εÞα ½ðα11Þε2εp 2

ð13:5Þ

FIGURE 13.4 Liquid water dynamics in a GDL, which is digitally reconstructed using X-ray ¨ F. Marone, M. Stampanoni, A. tomographic microscopy [28]. Reproduced from R. Fluckiger, ¨ Wokaun, F. N. Buchi, Investigation of liquid water in gas diffusion layers of polymer electrolyte fuel cells using x-ray tomographic microscopy, Electrochim. Acta 56 (5) (2011) 2254
2262.

Mass transport in the cathode Chapter | 13

377

where K is the intrinsic permeability, R is the fiber radius, and α is the constant that depends on transport directions. The relative permeability is related to the capillary pressure and liquid water saturation, and many equations are available to estimate relative permeability, but the most commonly used equation for relative permeability is still the power-law function which is simple in form and easily fitted with experiments [27]: krwp 5 s4 or knrwp 5 ð12s2 Þ4

ð13:6Þ

where krwp and knrwp are the relative permeability for the wetting and nonwetting phase, respectively. In addition, the compression also influences the intrinsic permeability of a GDL. Both the through-plane and in-plane permeability decrease with compression, while the variation of in-plane permeability is much bigger due to the elimination of in-plane flow pathways [23]. Capillary pressure is generated under the interactions of forces within and between two immiscible fluids and their bounding solids, including surface tension, interfacial tension, and solid adhesion. It can be calculated as the difference in pressure across the two-phase interface of gas and liquid. To evaluate the capillary pressure in GDL, an empirical function was first proposed based on experiments conducted by Leverett [29], written as:  ε 0:5  1:417ð1 2 sÞ 2 2:120ð12sÞ2 1 1:263ð12sÞ3 0 , θc , 90 Pc 5 σ cos θc 1:417s 2 2:120s2 1 1:263s3 90 , θc , 180 K ð13:7Þ where Pc is the capillary pressure, σ is the surface tension coefficient, and θc is the contact angle. It is observed that this equation describes the capillary pressure as a function of fluid property (surface tension coefficient), wettability (contact angle), structure parameter (porosity), intrinsic permeability, and liquid water saturation of GDL. However, this equation is not accurate since the experiments were set for packed beds, which have a different structure from the fibrous GDL. Further investigations have been carried out to find more precise correlations of capillary pressure, such as the empirical equations proposed by Ye and Nguyen [30], which are derived for TORAY carbon paper TGP-H-060 with 10% PTFE loading, written as: Pc 5 2:09e22:2ð0:3212sÞ 2 e44:9ð0:3212sÞ 1 35:6

ð13:8Þ

It should be mentioned that the PTFE treated on GDL determines the contact angle of solid in GDL and thus influences both the capillary pressure and relative permeability. With the treatment of PTFE, the contact angle of a GDL generally ranges from 110 to 130 degrees to maintain membrane hydration as well as liquid removal capacity. It is worth noting the liquid removal in GDL is largely changed by applying MPL. As aforementioned, the liquid water has to penetrate GDL through the cracks on MPL. The MPL is reported to have two orders of magnitude

378

Fuel Cells for Transportation

lower permeability than GDL due to the small pores, which hinder most of the liquid transport from CL to GDL [31]. In this scenario, the cracks on MPL function as the main pathway of liquid water because they are much larger than the pores of MPL. The liquid water is removed within a limited region, that is, liquid clusters, in GDL, and the flooding by liquid water is effectively reduced [32]. Thus, the transport resistance of oxygen caused by liquid saturation also drops. On the other hand, membrane hydration is enhanced by MPL because a higher level of liquid water could be retained in CL and membrane. The proton conductivity of the membrane is promoted, and ohmic loss of PEMFC is thus reduced.

13.2.3 Characterization of electron transport The electron conduction of GDL plays an important role in the overall electrical circuit because GDL is one of the components that mainly contributes to the overall ohmic resistance of a PEMFC. The electron transport in solid fibers and the matrix of GDL, driven by the electric potential gradient, is similar to the diffusion of oxygen diffusion in the void space which is driven by the concentration gradient [18]. Thus, the effective medium approximation could also be applied to the effective electrical conductivity of GDL. However, the correlations for a homogeneous porous material, such as Bruggeman approximation-based equations [21], cannot identify the anisotropic properties of GDL. To this end, further investigations have been made with a focus on the through-plane and in-plane effective electrical conductivities, which are evaluated by experimental measurements and numerical models that consider the 3D fibrous structure of GDL. Correlations are proposed for anisotropic conductivities based on the effective medium approximation, but they are corrected by structural parameters, such as [33]: 8   3ε > 20:007 > > 1 2 0:962ð12εÞ through 2 plane exp½0:889ð1 2 εÞ > 3 2 ð1 2 εÞ σeff <   5 3ε > σs 20:016 > > 1 2 0:962ð12εÞ in 2 plane exp½0:367ð1 2 εÞ > : 3 2 ð1 2 εÞ ð13:9Þ where σeff is the effective conductivity, and σs is the conductivity of solid fibers. For the carbon paper, the reported through-plane and in-plane conductivities range from 300 to 1400 S m21 and 5000 to 23000 S m21, respectively. It is worth noting that the PTFE treatment and compression also influence the effective diffusivity of GDL. It is found that PTFE has little effect on the in-plane conductivity, but significantly reduces the throughplane conductivity due to the electrical resistance created by PTFE between the carbon fibers [34]. The compression of GDL is reported to improve both

Mass transport in the cathode Chapter | 13

379

the through-plane and in-plane conductivities, while the effect on throughplane conductivity is more significant because of better contact between the fibers along the through-plane direction [23]. The presence of MPL largely eliminates the contact resistance between GDL and CL, which further reduces the ohmic loss of PEMFC in addition to its effect on the proton conductivity of the membrane. It is reported that the through-plane electrical conductivity of a porous electrode could be increased by 40% by introducing MPL. The in-plane conductivity is less influenced, which only exhibits a 4% change in the presence of MPL [31].

13.2.4 Innovations in gas diffusion layer structure and material The innovations on GDL are made to further improve the oxygen transport capacity by governing the water removal pathway in GDL. Some of them are made directly on the carbon paper/cloth because it is commercially available and cost-effective, while others are made to replace the carbon paper/ cloth with other porous materials. For the innovations made on carbon paper, perforation and mixed wettability are reported to have the potential to further enhance mass transfer. The perforation of GDL, as shown in Fig. 13.5A, can concentrate liquid water in the perforated pores and reduce the liquid occupation in other places of GDL. It is found that the perforation located near the breakthrough point has the largest effect on the liquid distribution in a GDL. A hydrophilic perforation perimeter is proven to facilitate liquid removal [35]. However, the perforation may cause problems of corrosion resistance and mechanical strength which must be avoided. On the other hand, the mixed wettability of carbon paper is applicable because the carbon paper is treated hydrophobic by PTFE, and the drying method can change the distribution of PTFE. Vacuum and air-drying (Fig. 13.5B) can result in homogeneous and heterogeneous PTFE distribution, respectively, the latter of which can improve the capillary pressure near CL and retain more water in CL and membrane [36]. Though carbon paper has long been used as the material of GDL, the carbon substrates require complex manufacturing processes and have lower electrical conductivity than metals. In this scenario, attempts have been made to use metallic material as an alternative. Since the metals have a larger mechanical strength and electrical conductivity, the GDL can be made thinner, which reduces the thickness of MEA. However, corrosion of the metal substrate will occur when no effective coating is applied, and thus, extra cost of coating is inevitable. Except for metallic GDL, a GDL-less design with a foam material flow field, as aforementioned (Fig. 13.3D), is proved feasible. The replacement of GDL with foam material that is integrated with the flow field may eliminate the interfacial transport of mass and electrons between the flow field and GDL, and thus avoid the interfacial transport resistance [2]. Though similar problems of corrosion to metallic

380

Fuel Cells for Transportation

FIGURE 13.5 Modifications on carbon fibers: (A) GDL perforation and positions of the perforated holes, (B) mixed PTFE distribution by the vacuum and air-drying methods [37]. Reproduced from D. Gerteisen, T. Heilmann, C. Ziegler, Enhancing liquid water transport by laser perforation of a GDL in a PEM fuel cell, J. Power Sources 177 (2) (2008) 348
354. Niu, Z. Bao, J. Wu, Y. Wang, K. Jiao, Two-phase flow in the mixed-wettability gas diffusion layer of proton exchange membrane fuel cells, Appl. Energy 232 (2018) 443
450.

GDL should be addressed for its practical use, it still provides a promising approach for the further improvement of PEMFC compactness and power density. Efforts have also been made to modify the GDL surface by 3D printing highly hydrophobic patterns (using fluorinated ethylene propylene [FEP] or a mixture of polydimethylsiloxane [PDMS] and fumed silica) on it or direct 3D printing of GDL with a controlled structure to improve water transport in the cathode [38
40]. Further details on patterned and structured GDLs are available in Chapter 8.

13.3 Mass transfer in cathode catalyst layer As the key component of PEMFC, the CL provides the sites wherein the hydrogen oxidation reaction and ORR occur. Generally, CL consists of catalysts, catalyst supports, and ionomer electrolyte. Platinum (Pt) is commonly used as a catalyst owing to its good catalyst activity and durability in an electrochemical environment. However, the usage of Pt leads to an

Mass transport in the cathode Chapter | 13

381

extremely high cost of PEMFC. Therefore, the Pt alloy (e.g., PtNi and PtCo) and Pt-free (e.g., Fe-N-C and Co-N-C) catalysts are considerably developed to reduce the Pt loading in recent years [41]. Moreover, novel catalyst architectures (e.g., core
shell, nanocage, nanowire, nanoframe, etc.) are also developed to improve the catalyst activity [42
47]. Carbon is commonly employed as the catalyst support, and the continuous carbon aggregates ensure the electron conduction. An accessible porous carbon with a high surface area significantly improves the CL performance and durability in comparison with the solid carbon [48,49]. In addition, carbon nanotube (CNT), carbon nanofiber (CNF), and graphene are also the choices for catalyst support [50]. The electrolyte material is an ionomer such as perfluorosulfonic acid (PFSA) that contains the hydrophobic backbone and hydrophilic side chain with a sulfonic acid group in the terminal. Ionomer contact with catalyst particles provides continuous proton transport paths. Carbon-supported catalyst aggregates are mixed with an ionomer and self-organize into an extremely complex porous structure. Based on the structural properties, the transport phenomena in CLs can be divided into two levels, that is, the transport phenomena in the local region near catalysts and in the porous structures of bulk CLs as shown in Fig. 13.6. In the local region near catalysts, the gas molecules need first to be dissolved into the ionomer film and then diffused in the ionomer film toward the catalysts. Meanwhile, the protons migrate in the ionomer, and the produced water will exclude from the ionomer film to the pores of CLs. The bulk CL can be considered a porous medium. The liquid water transport and gas diffusion processes occur in the pores of CLs. Therefore, the transport phenomena in CLs are complex and have a significant effect on the performance of PEMFC. It is essential to understand the mass transfer in CLs and provide theoretical guidance for the design of novel high-performance CLs.

FIGURE 13.6 Schematic of mass transfer in CLs.

382

Fuel Cells for Transportation

13.3.1 Mass transfer in local region near catalysts The electrochemical reactions occur on the catalyst surface. Therefore, the mass transfer in the local region near catalysts directly affects the electrochemical reactions. The local transport phenomena are complex due to the complex structures near catalysts. The material of ionomer electrolyte is generally PFSA same as the membrane material. However, the properties of the ionomer electrolyte are different from the membrane because the ionomer electrolyte is a film with a thickness of approximately 2
10 nm on catalysts [51,52]. The structure of ionomer film can be investigated using some experimental technologies, such as neutron scattering (NR) and surface X-ray scattering (SXS). The hydrophobic fluorocarbon groups of PFSA were observed to be the main species adsorbed on the Pt surface using the NR and SXS technologies [53,54]. However, the ionomer film becomes more hydrophilic when the Pt atoms are oxidized to be PtO, and PtO also leads to a longrange restructuring of PFSA chains [53]. To conduct the protons, the ionomer film needs to be hydrated well. Therefore, the swelling phenomena will occur when the water content in the ionomer film changes. The swelling degree decreases with decreasing ionomer film thickness when the film’s thickness is over B20 nm [55]. The water uptake and swelling are controlled by the underlying chemical
mechanical energy balance between the substrates and ionomer film. Generally, the ionomer film’s coverage and thickness on the carbon-supported catalyst aggregates are heterogeneous, which highly affects the local mass transfer and the electrochemical reactions in CLs [56]. To achieve the commercialization of PEMFC, the Pt loading should be reduced further to a level of , 0.1 g kW21 [2]. However, the transport resistances, especially the oxygen transport resistance, increase significantly at a low Pt loading, severely decreasing the electrochemical performance of PEMFCs [57
59]. Moreover, the local oxygen transport resistance accounts for approximately 77% of the total oxygen transport resistance in CLs when the Pt loading is approximately 0.05 mg cm22 [60]. The increased local oxygen transport resistance mainly results from the decreased surface area of catalysts, which leads to the increased transport flux per surface area [61]. Therefore, the electrochemical surface area of CLs should be increased to reduce the local oxygen transport resistance. Although the advanced experimental technologies can observe the structural properties of ionomer film, it is still difficult for in situ observation of the transport phenomena in the ionomer film with an ultrasmall size. Molecular dynamics (MD) simulation is a powerful tool to explore the nanoscale transport phenomena. Many researchers have used MD simulations to investigate the transport phenomena in the ionomer film. The ionomer film on the Pt surface can be divided into three regions: ionomer
Pt interface, bulk-like region, and ionomer
gas interface according to the MD

Mass transport in the cathode Chapter | 13

383

simulation results [62
64]. The ionomer
Pt interface has a dense structure with a tight arrangement of PFSA chains, which is the primary cause for the observed high O2 transport resistance as shown in Fig. 13.7. The oxygen transport paths are different in different regions. Especially, the oxygen molecules permeate via the water sites in the ionomer
Pt interface due to the high transport resistance through the tightly arranged PFSA chains. However, in the bulk ionomer region, the oxygen molecules mainly permeate via the small cavities formed by PFSA chains and the interface between water clusters and PFSA backbone at low water contents and high water contents, respectively. The structure of the ionomer
gas interface is less dense, resulting in high oxygen solubility and low oxygen transport resistance. Furthermore, the oxygen transport phenomena near the polyhedral Pt particles are more complex than that on the Pt surface [65]. The same as that on the Pt surface, a dense ultrathin sublayer with a tight arrangement of PFSA is present on Pt facets. In the ionomer near Pt edges and corners, the structure is less dense due to the weaker Pt attraction, resulting in a higher O2 density than that on Pt facets. As shown in Fig. 13.7, 90% of

FIGURE 13.7 (A) Density of ionomer along the thickness direction at different water contents (λ); (B) oxygen transport resistance in three regions; (C) snapshots of the ionomer
Pt interface (the gray, yellow, and blue beads represent the fluorocarbon groups of PFSA, sulfur/oxygen atoms of PFSA, and water molecules/hydroniums); (D) numbers of the oxygen molecules that reach the different parts of Pt nanoparticles at different simulation times; (E) portions of the oxygen fluxes of different parts; (F) oxygen transport routes near Pt nanoparticles. Reproduced from L. Fan, Y. Wang, K. Jiao, Oxygen permeation resistances and routes in nanoscale ionomer thin film on platinum surface, J. Electrochem. Soc. 168 (2021) 014511; L. Fan, Y. Wang, K. Jiao, Oxygen transport routes in ionomer film on polyhedral platinum nanoparticles, ACS Nano 14 (2020) 17487
17495.

384

Fuel Cells for Transportation

oxygen molecules reach the Pt cube and tetrahedron particles via their upper corner and edge regions. Therefore, the area in the vicinity of Pt particles can be divided into two areas. One is the high oxygen transport resistance area near the Pt facets, and another is the low oxygen transport resistance area near the Pt edges and corners. Oxygen molecules tend to permeate through the low oxygen transport resistance area reaching Pt edges and corners for the electrochemical reaction. To decrease the local oxygen transport resistance, many researchers have modified the molecular structures of ionomer materials, the interactions between the electrode and electrolyte, and the morphologies of catalysts and catalyst support. The side chain length has a significant effect on the properties of PFSA chains. Rolfi et al. [66] shortened the side chain length and added an MDO monomer into the PFSA chains, which increased the oxygen permeability by 20% compared with Nafion 212 and decreased the oxygen transport resistance. As mentioned above, the ionomer film’s thickness on the carbon-supported Pt aggregates is heterogeneous. The oxygen transport resistance is lower, and the electrochemical reaction rate is higher when the ionomer film is thinner, while the electrochemical reaction is difficult to occur due to the lack of oxygen when the ionomer film is thicker. Therefore, the electrochemical reaction in CLs is heterogeneous, and the Pt utilization is also lower. To solve the problem, Ott et al. [67] nitrided the carbon support surface and thus altered the interactions between the electrode and electrolyte. The ionomer films can more evenly cover the carbon-supported Pt aggregates, thereby decreasing the local oxygen transport resistance and increasing the utilization of Pt. Furthermore, adding ionic liquids into the electrolyte is another method to modify the interactions between the electrode and electrolyte. The ionic liquids can inhibit the Pt oxidation and the adsorption of nonreactive species and improve the durability of catalysts by suppressing the Pt particle growth and dissolution. Additionally, Fan et al. [68] found that the ionic liquid additives significantly alter the ultrathin sublayer structure by inhibiting the tight arrangement of PFSA chains. The IL addition provides a larger free space for O2 dissolution in the ionomer ultrathin sublayer. Consequently, the O2 density in the ultrathin sublayer is improved by an order of magnitude, and the O2 transport flux across the ionomer film is increased by up to eight times due to IL molecules’ presence. Moreover, to further reduce the oxygen transport resistance due to the coverage of ionomer films on catalysts, an accessible porous carbon with nanosized pores was proposed by Yarlagadda et al. [48]. The PFSA chains cannot enter the nanosized pores, thereby avoiding covering the catalysts in the pores. However, the protons and oxygen molecules can permeate through the pores toward catalysts for the reaction. The protons may migrate via the water channels in the pores due to the absence of an ionomer. But as far, the detailed transport mechanisms of oxygen and protons in the nanosized pores are unclear.

Mass transport in the cathode Chapter | 13

385

13.3.2 Mass transfer in bulk catalyst layers With increasing current density, more oxygen is needed, and more liquid water is produced in CLs. Therefore, the porous structure of CLs should be designed well to enhance oxygen and liquid water transport. Understanding the transport phenomena is fundamental to the novel structure design. Generally, the liquid water transport phenomena in channels and GDLs of PEMFC can be observed by the X-ray scattering and neutron scattering [69
71]. However, the pore size of CLs is so small that it is still difficult for in situ observation of the liquid water transport phenomena in CLs. In this case, the simulation models are a good tool to investigate the transport phenomena in CLs. The liquid water saturation can be obtained by solving the Leverett-J empirical correlation based on the Darcy law [72
74]. The transport properties can be represented by the liquid water saturation to some extent. However, the structure of CLs is complex, and the pore size is small. Therefore, the accuracy of the Leverett-J empirical correlation used in CLs is questioned. Lattice Boltzmann (LB) method has been developed to be an effective method in the last few decades, which is a method in mesoscopic scale between the MD simulation method in microscopic scale and the method based on the continuity assumption in macroscopic scale. Therefore, the LB method is very suitable to investigate the multiphase flow in the porous structure of CLs. The liquid water is produced on the catalyst surface, flows into the small pores first, and is then transferred into the big pores. Meanwhile, the morphology of liquid water changes from the small drops to the connected water channels. The hydrophobic design of CLs benefits the breakthrough of water and thus avoids the coverage of water on catalysts, thereby decreasing the oxygen transport resistance [75]. At present, the commercial CLs are formed by random stacking of carbonsupported Pt aggregates (Fig. 13.8). Therefore, the porous structures of CLs are unordered, and the pore tortuosity is large, which results in a high mass transfer resistance. Especially at high current density, the electrochemical reactions need

FIGURE 13.8 CLs with (A) the random stacking of carbon-supported catalyst aggregates, (B) the layered structure formed by fibers, and (C) the array columnar structure.

386

Fuel Cells for Transportation

more oxygen and produce more water, and thus, the high mass transfer resistance will lead to a high voltage loss. The ordered CLs are developed to decrease the transport resistance. The CLs with fiber arrangement can be prepared using the electrospinning technique (Fig. 13.8). This type of CL has a layered structure formed by many fibers, which is beneficial to the proton and electron transport [76]. Moreover, the CLs with array architectures have highly ordered transport channels of protons, electrons, gases, and liquid water (Fig. 13.8). This type of CL consists of array columnar catalysts (such as the NSTF catalysts developed by the 3M company) or array columnar catalyst supports with the catalyst particles (such as the CNT as catalyst support) [77,78]. Further details on the structure and composition of CL, GDL, and MPL are available in Chapter 7.

13.4 Summary Mass transfer and two-phase flow play an important role in PEMFC operation, especially in the high current density region. Further improvement of power density for commercialization demands a more effective mass transfer capacity for the whole PEMFC. The utilization of modified parallel channels, such as wavy and partially narrowed channels, has become a choice for commercial PEMFC stacks due to their enhancement in mass transfer. This design modification plays an important role in the further design of highpower density PEMFCs. On the other hand, the baffle and porous material flow field are two potential approaches to high-power density PEMFC due to their strong mass transfer capacity, but issues associated with manufacturing cost and corrosion resistance need to be addressed. The modification of carbon paper GDL (such as pattering the surfaces with highly hydrophobic materials) will be a cost-effective way to further enhance the mass transfer of GDL. In addition, metallic GDLs are also promising if an effective coating is applied. The metallic material GDL, if integrated with the porous material flow field, will produce an integrated BP-MEA design which eliminates the interface resistance between the flow field and MEA. To better organize the mass transfer, the order-structured CLs are a promising approach, which can achieve highly efficient mass transfer pathways, uniform reaction sites, and high utilization of Pt. Moreover, novel architectures of catalysts with high activity should be developed to decrease the Pt loading. All these designs are essential for ultrahigh power density (9 kW L21) operation for PEMFCs.

Nomenclature CL Dbulk Deff

catalyst layer bulk diffusivity effective diffusivity

Mass transport in the cathode Chapter | 13 ε εp εwet s K R α krwp knrwp Pc σ θc σ eff σs λ GDL LB MPL MD PEMFC

387

porosity porosity for percolation threshold equivalent porosity of partially saturated GDL liquid water saturation intrinsic permeability fiber radius constant depending on permeation direction relative permeability for wetting phase relative permeability for nonwetting phase capillary pressure surface tension coefficient contact angle effective electrical conductivity conductivity of solid fibers water content in ionomer film gas diffusion layer lattice Boltzmann microporous layer molecular dynamics proton-exchange membrane fuel cell

References [1] Y. Wang, D.F.R. Diaz, K.S. Chen, Z. Wang, X.C. Adroher, Materials, technological status, and fundamentals of PEM fuel cells—a review, Mater. Today 32 (2020) 178
203. [2] K. Jiao, J. Xuan, Q. Du, Z. Bao, B. Xie, B. Wang, et al., Designing the next generation of proton-exchange membrane fuel cells, Nature 595 (7867) (2021) 361
369. [3] T. Tsukamoto, T. Aoki, H. Kanesaka, T. Taniguchi, T. Takayama, H. Motegi, et al., Three-dimensional numerical simulation of full-scale proton exchange membrane fuel cells at high current densities, J. Power Sources 488 (2021) 229412. [4] S.C. Cho, Y. Wang, K.S. Chen, Droplet dynamics in a polymer electrolyte fuel cell gas flow channel: forces, deformation, and detachment. I: theoretical and numerical analyses, J. Power Sources 206 (2012) 119
128. [5] K. Jiao, X. Li, Water transport in polymer electrolyte membrane fuel cells, Prog. Energy Combust. Sci. 37 (3) (2011) 221
291. [6] I.S. Hussaini, C. Wang, Visualization and quantification of cathode channel flooding in PEM fuel cells, J. Power Sources 187 (2) (2009) 444
451. [7] B.H. Lim, E.H. Majlan, W.R.W. Daud, T. Husaini, M.I. Rosli, Effects of flow field design on water management and reactant distribution in PEMFC: a review, Ionics 22 (3) (2016) 301
316. [8] H.-W. Wu, A review of recent development: transport and performance modeling of PEM fuel cells, Appl. Energy 165 (2016) 81
106. [9] A.P. Manso, F.F. Marzo, J. Barranco, X. Garikano, M.G. Mujika, Influence of geometric parameters of the flow fields on the performance of a PEM fuel cell. A review, Int. J. Hydrog. Energy 37 (20) (2012) 15256
15287. [10] N. Saito, H. Kikuchi, Y. Nakao, New fuel cell stack for FCX clarity, Honda R&D Technical Review 21, 2009. 1.

388

Fuel Cells for Transportation

[11] T. Yoshizumi, H. Kubo, M. Okumura. Development of high-performance FC stack for the new MIRAI. No. 2021-01-0740. SAE Technical Paper, 2021. [12] T. Yoshida, K. Kojima, Toyota MIRAI fuel cell vehicle and progress toward a future hydrogen society, Electrochem. Soc. Interface 24 (2) (2015) 45. [13] Z. Bao, Z. Niu, K. Jiao, Gas distribution and droplet removal of metal foam flow field for proton exchange membrane fuel cells, Appl. Energy 280 (2020) 116011. [14] J. Kim, G. Luo, C.-Y. Wang, Modeling two-phase flow in three-dimensional complex flow-fields of proton exchange membrane fuel cells, J. Power Sources 365 (2017) 419
429. [15] Z. Bao, Z. Niu, K. Jiao, Numerical simulation for metal foam two-phase flow field of proton exchange membrane fuel cell, Int. J. Hydrog. Energy 44 (2019) 6229
6244. [16] Z. Bao, Y. Wang, K. Jiao, Liquid droplet detachment and dispersion in metal foam flow field of polymer electrolyte membrane fuel cell, J. Power Sources 480 (2020) 229150. [17] A. El-Kharouf, T.J. Mason, D.J. Brett, B.G. Pollet, Ex-situ characterisation of gas diffusion layers for proton exchange membrane fuel cells, J. Power Sources 218 (2012) 393
404. [18] N. Zamel, X. Li, Effective transport properties for polymer electrolyte membrane fuel cells
with a focus on the gas diffusion layer, Prog. Energy Comb. Sci 39 (1) (2013) 111
146. [19] J.H. Chun, K.T. Park, D.H. Jo, J.Y. Lee, S.G. Kim, E.S. Lee, Determination of the pore size distribution of micro porous layer in PEMFC using pore forming agents under various drying conditions, Int. J. Hydrog. Energy 35 (20) (2010) 11148
11153. [20] N. Zamel, X. Li, J. Shen, Correlation for the effective gas diffusion coefficient in carbon paper diffusion media, Energy Fuels 23 (12) (2009) 6070
6078. [21] P.K. Das, X. Li, Z.-S. Liu, Effective transport coefficients in PEM fuel cell catalyst and gas diffusion layers: beyond Bruggeman approximation, Appl. Energy 87 (9) (2010) 2785
2796. [22] J.H. Nam, M. Kaviany, Effective diffusivity and water-saturation distribution in singleand two-layer PEMFC diffusion medium, Int. J. Heat. Mass. Transf. 46 (24) (2003) 4595
4611. [23] Z. Bao, Y. Li, X. Zhou, F. Gao, Q. Du, K. Jiao, Transport properties of gas diffusion layer of proton exchange membrane fuel cells: effects of compression, Int. J. Heat. Mass. Transf 178 (2021) 121608. [24] N. Zamel, X. Li, J. Becker, A. Wiegmann, Effect of liquid water on transport properties of the gas diffusion layer of polymer electrolyte membrane fuel cells, Int. J. Hydrog. Energy 36 (9) (2011) 5466
5478. [25] J.M. LaManna, S.G. Kandlikar, Determination of effective water vapor diffusion coefficient in PEMFC gas diffusion layers, Int. J. Hydrog. Energy 36 (8) (2011) 5021
5029. [26] M.M. Tomadakis, T.J. Robertson, Viscous permeability of random fiber structures: comparison of electrical and diffusional estimates with experimental and analytical results, J. Compos. Mater. 39 (2) (2005) 163
188. [27] I.S. Hussaini, C.Y. Wang, Measurement of relative permeability of fuel cell diffusion media, J. Power Sources 195 (12) (2010) 3830
3840. [28] R. Fl¨uckiger, F. Marone, M. Stampanoni, A. Wokaun, F.N. B¨uchi, Investigation of liquid water in gas diffusion layers of polymer electrolyte fuel cells using x-ray tomographic microscopy, Electrochim. Acta 56 (5) (2011) 2254
2262. [29] K.S. Udell, Heat transfer in porous media considering phase change and capillarity—the heat pipe effect, Int. J. Heat. Mass. Transf. 28 (2) (1985) 485
495.

Mass transport in the cathode Chapter | 13

389

[30] Q. Ye, T.V. Nguyen, Three-dimensional simulation of liquid water distribution in a PEMFC with experimentally measured capillary functions, J. Electrochem. Soc 154 (12) (2007) B1242. [31] M.V. Williams, E. Begg, L. Bonville, H.R. Kunz, J.M. Fenton, Characterization of gas diffusion layers for PEMFC, J. Electrochem. Soc. 151 (8) (2004) A1173. [32] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, On the role of the microporous layer in PEMFC operation, Electrochem. Commun 11 (3) (2009) 576
579. [33] N. Zamel, X. Li, J. Shen, Numerical estimation of the effective electrical conductivity in carbon paper diffusion media, Appl. Energy 93 (2012) 39
44. [34] M.S. Ismail, T. Damjanovic, D.B. Ingham, M. Pourkashanian, A. Westwood, Effect of polytetrafluoroethylene-treatment and microporous layer-coating on the electrical conductivity of gas diffusion layers used in proton exchange membrane fuel cells, J. Power Sources 195 (9) (2010) 2700
2708. [35] Z. Niu, J. Wu, Z. Bao, Y. Wang, Y. Yin, K. Jiao, Two-phase flow and oxygen transport in the perforated gas diffusion layer of proton exchange membrane fuel cell, Int. J. Heat. Mass. Transf 139 (2019) 58
68. [36] Z. Niu, Z. Bao, J. Wu, Y. Wang, K. Jiao, Two-phase flow in the mixed-wettability gas diffusion layer of proton exchange membrane fuel cells, Appl. Energy 232 (2018) 443
450. [37] D. Gerteisen, T. Heilmann, C. Ziegler, Enhancing liquid water transport by laser perforation of a GDL in a PEM fuel cell, J. Power Sources 177 (2) (2008) 348
354. [38] D. Thumbarathy, G. Gupta, M. Mamlouk, P.K. Das, Fabrication and characterization of tuneable flow-channel/gas-diffusion-layer interface for polymer electrolyte fuel cells, J. Electrochem. Energy Convers. Storage 17 (1) (2020) 011010. [39] S. Chandrasekaran, P.G. Campbell, T.F. Baumann, M.A. Worsley, Carbon aerogel evolution: allotrope, graphene-inspired, and 3D-printed aerogels, J. Mater. Res. 32 (22) (2017) 4166
4185. [40] V.A. Beck, A.N. Ivanovskaya, S. Chandrasekaran, J.B. Forien, S.E. Baker, E.B. Duoss, et al., Inertially enhanced mass transport using 3D-printed porous flow-through electrodes with periodic lattice structures, Proc. Natl Acad. Sci. 118 (32) (2021) e2025562118. [41] L. Du, V. Prabhakaran, X. Xie, S. Park, Y. Wang, Y. Shao, Low-PGM and PGM-free catalysts for proton exchange membrane fuel cells: stability challenges and material solutions, Adv. Mater. 33 (2020) 1908232. [42] J. Fan, M. Chen, Z. Zhao, Z. Zhang, S. Ye, S. Xu, et al., Bridging the gap between highly active oxygen reduction reaction catalysts and effective catalyst layers for proton exchange membrane fuel cells, Nat. Energy 6 (2021) 475
486. [43] M. Li, Z. Zhao, T. Cheng, A. Fortunelli, C.Y. Chen, R. Yu, et al., Ultrafine jagged platinum nanowires enable ultrahigh mass activity for the oxygen reduction reaction, Science 354 (2016) 1414
1419. [44] X. Huang, Z. Zhao, C. Liang, Y. Chen, E. Zhu, Z. Lin, et al., High-performance transition metal-doped Pt3Ni octahedra for oxygen reduction reaction, Science 348 (2015) 1230
1234. [45] C. Chen, Y. Kang, Z. Huo, Z. Zhu, W. Huang, H.L. Xin, et al., Highly crystalline multimetallic nanoframes with three-dimensional electrocatalytic surfaces, Science 343 (2014) 1339
1343. [46] X. Tian, X. Zhao, Y.Q. Su, L. Wang, H. Wang, D. Dang, et al., Engineering bunched PtNi alloy nanocages for efficient oxygen reduction in practical fuel cells, Science 366 (2019) 850
856.

390

Fuel Cells for Transportation

[47] J. Li, S. Sharma, X. Liu, Y.T. Pan, J.S. Spendelow, M. Chi, et al., Hard-magnet L10-CoPt nanoparticles advance fuel cell catalysis, Joule 3 (2019) 124
135. [48] V. Yarlagadda, M.K. Carpenter, T.E. Moylan, R.S. Kukreja, R. Kukreja, W. Gu, et al., Boosting fuel cell performance with accessible carbon mesopores, ACS Energy Lett. 3 (2018) 618
621. [49] E. Padgett, V. Yarlagadda, M.E. Holtz, M. Ko, B.D.A. Levin, R.S. Kukreja, et al., Mitigation of PEM fuel cell catalyst degradation with porous carbon supports, J. Electrochem. Soc 166 (2019) F198. [50] N. Jung, D.Y. Chung, J. Ryu, S.J. Yoo, Y.E. Sung, Pt-based nanoarchitecture and catalyst design for fuel cell applications, Nano Today 9 (2014) 433
456. [51] Y. Kurihara, T. Mabuchi, T. Tokumasu, Molecular dynamics study of oxygen transport resistance through ionomer thin film on Pt surface, J. Power Sources 414 (2019) 263
271. [52] L. Chen, R. Zhang, Q. Kang, W.Q. Tao, Pore-scale study of pore-ionomer interfacial reactive transport processes in proton exchange membrane fuel cell catalyst layer, Chem. Eng. J. 391 (2020) 123590. [53] D.L. Wood III, J. Chlistunoff, J. Majewski, R.L. Borup, Nafion structural phenomena at platinum and carbon interfaces, J. Am. Chem. Soc. 131 (2019) 18096
18104. [54] T. Masuda, H. Fukumitsu, T. Kondo, H. Naohara, K. Tamura, O. Sakata, et al., Structure of Pt(111)/ionomer membrane interface and its bias-induced change in membrane electrode assembly, J. Phys. Chem. C. 117 (2013) 12168
12171. [55] A. Kusoglu, D. Kushner, D.K. Paul, K. Karan, M.A. Hickner, A.Z. Weber, Impact of substrate and processing on confinement of Nafion thin films, Adv. Funct. Mater. 24 (2014) 4763
4774. [56] N. Goswami, A.N. Mistry, J.B. Grunewald, T.F. Fuller, P.P. Mukherjee, Corrosioninduced microstructural variability affects transport-kinetics interaction in PEM fuel cell catalyst layers, J. Electrochem. Soc 167 (2020) 084519. [57] N. Nonoyama, S. Okazaki, A.Z. Weber, Y. Ikogi, T. Yoshida, Analysis of oxygentransport diffusion resistance in proton-exchange-membrane fuel cells, J. Electrochem. Soc 158 (2011) B416. [58] A. Ohma, T. MashioT, K. Sato, H. Iden, Y. Ono, K. Sakai, et al., Analysis of proton exchange membrane fuel cell catalyst layers for reduction of platinum loading at Nissan, Electrochim. Acta 56 (2011) 10832
10841. [59] J.P. Owejan, J.E. Owejan, W. Gu, Impact of platinum loading and catalyst layer structure on PEMFC performance, J. Electrochem. Soc. 160 (2013) F824. [60] C. Wang, X. Cheng, J. Lu, S. Shen, X. Yan, J. Yin, et al., The experimental measurement of local and bulk oxygen transport resistances in the catalyst layer of proton exchange membrane fuel cells, J. Phys. Chem. Lett. 8 (2017) 5848
5852. [61] A.Z. Weber, A. Kusoglu, Unexplained transport resistances for low-loaded fuel-cell catalyst layers, J. Mater. Chem. A 2 (2014) 17207
17211. [62] R. Jinnouchi, K. Kudo, N. Kitano, Y. Morimoto, Molecular dynamics simulations on O2 permeation through Nafion ionomer on platinum surface, Electrochim. Acta 188 (2016) 767
776. [63] Y. Kurihara, T. Mabuchi, T. Tokumasu, Molecular analysis of structural effect of ionomer on oxygen permeation properties in PEFC, J. Electrochem. Soc. 164 (2017) F628
F637. [64] L. Fan, Y. Wang, K. Jiao, Oxygen permeation resistances and routes in nanoscale ionomer thin film on platinum surface, J. Electrochem. Soc 168 (2021) 014511.

Mass transport in the cathode Chapter | 13

391

[65] L. Fan, Y. Wang, K. Jiao, Oxygen transport routes in ionomer film on polyhedral platinum nanoparticles, ACS Nano 14 (2020) 17487
17495. [66] A. Rolfi, C. Oldani, L. Merlo, D. Facchi, R. Ruffo, New perfluorinated ionomer with improved oxygen permeability for application in cathode polymeric electrolyte membrane fuel cell, J. Power Sources 396 (2018) 95
101. [67] S. Ott, A. Orfanidi, H. Schmies, B. Anke, H.N. Nong, J. H¨ubner, et al., Ionomer distribution control in porous carbon-supported catalyst layers for high-power and low Pt-loaded proton exchange membrane fuel cells, Nat. Mater. 19 (2020) 77
85. [68] L. Fan, Y. Wang, K. Jiao, Enhancing oxygen transport in the ionomer film on platinum catalyst using ionic liquid additives, Fundam. Res. (2021). Available from: https://doi.org/ 10.1016/j.fmre.2021.09.004. [69] P. Rahimian, L. Battrell, R. Anderson, N. Zhu, E. Johnson, L. Zhang, Investigation of time dependent water droplet dynamics on porous fuel cell material via synchrotron based X-ray imaging technique, Exp. Therm. Fluid Sci. 97 (2018) 237
245. [70] J. Mishler, Y. Wang, P. Mukundan, J. Spendelow, D.S. Hussey, D.L. Jacobson, et al., Probing the water content in polymer electrolyte fuel cells using neutron radiography, Electrochim. Acta 75 (2012) 1
10. [71] J. Mishler, Y. Wang, R. Mukundan, R.L. Borup, D.S. Hussey, D. Jacobson, In situ investigation of water distribution in polymer electrolyte fuel cell using neutron radiography, ECS Trans. 33 (2010) 1443. [72] G. Zhang, L. Wu, Z. Qin, J. Wu, F. Xi, G. Mou, et al., A comprehensive threedimensional model coupling channel multi-phase flow and electrochemical reactions in proton exchange membrane fuel cell, Adv. Appl. Energy 2 (2021) 100033. [73] G. Zhang, J. Wu, Y. Wang, Y. Yin, K. Jiao, Investigation of current density spatial distribution in PEM fuel cells using a comprehensively validated multi-phase non-isothermal model, Int. J. Heat. Mass. Transf 150 (2020) 119294. [74] H. Chen, H. Guo, F. Ye, C.F. Ma, Modification of the two-fluid model and experimental study of proton exchange membrane fuel cells with baffled flow channels, Energy Convers. Manag. 195 (2019) 972
988. [75] L. Chen, Q. Kang, W.Q. Tao, Pore-scale numerical study of multiphase reactive transport processes in cathode catalyst layers of proton exchange membrane fuel cells, Int. J. Hydrog. Energy 46 (2021) 13283
13297. [76] H. Liu, D. Si, H. Ding, S. Wang, J. Zhang, Y. Liu, Cold start capability and durability of electrospun catalyst layer for proton exchange membrane fuel cell, Int. J. Hydrog. Energy 46 (2020) 11140
11149. [77] S. Du, Recent advances in electrode design based on one-dimensional nanostructure arrays for proton exchange membrane fuel cell applications, Engineering 7 (2021) 33
49. [78] S. Murata, M. Imanishi, S. Hasegawa, R. Namba, Vertically aligned carbon nanotube electrodes for high current density operating proton exchange membrane fuel cells, J. Power Sources 253 (2014) 104
113.

This page intentionally left blank

Chapter 14

Control-oriented computational fluid dynamics models for polymer electrolyte membrane fuel cells Jian Zhao1, Xianguo Li1, Chris Shum2 and John McPhee2 1

Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada, 2Department of Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada

14.1 Introduction The operation of polymer electrolyte membrane (PEM) fuel cells in automotive applications involves complicated transport phenomena of various reactant and product species, momentum, electrons, protons, and energy as well as electrochemical behaviors [1]. These phenomena, determining the performance of PEM fuel cells, can be affected by various operational variables, such as temperature, pressure, humidity, and flow rates. Therefore how to control these variables precisely and accurately is of significant importance to the stability and robustness of fuel cell operation [2,3]. Many control strategies have been reported to be adopted in fuel cell control, for example, proportional
integral
derivative (PID) control, adaptive control, and model predictive control [2,4
6]. The PID control is commonly employed to control the operational variables of PEM fuel cells, such as temperature, pressure, and flow rates, through feedback and feed-forward strategies. The PID implementation is simple and less resource-demanding; however, the PID control can be slow and less accurate to control nonlinear fuel cell behaviors. Adaptive control, often integrated into a PID controller, enables the modifications to operational variables based on sensing data to achieve optimized performance. The model predictive control can analyze and predict the nonlinear fuel cell behaviors and then optimize the control strategies to maintain high performance and prevent unsafe operations [2], utilizing a fuel cell performance prediction model. Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00015-0 © 2023 Elsevier Ltd. All rights reserved.

393

394

Fuel Cells for Transportation

The fuel cell performance prediction model can be empirical, analytical, data-driven, and numerical models [7
25]. Empirical models are established to estimate the current
voltage relation under given conditions through a simple and computationally light mathematical equation [9,10,18,20
22]. Empirical models are usually validated within a narrow range of operating conditions due to the limited experimental data availability, and it is difficult to estimate the impact of temperature, pressure, and flow rates on the fuel cell performance, which are important to control the coolant, reactant supply, and humidification subsystems [17]. Analytical models are built based on simple conservation laws in which spatial dimensions are not necessary and unknown empirical coefficients are not required. The analytical model is usually computationally fast and used in model predictive control of fuel cells but unable to predict the dynamic behaviors [26
30]. Data-driven models, such as neural networks, do not require any physical knowledge of fuel cells as long as sufficient data are available for model training. Once a data-driven model is well trained, it can be applied to predict the current
voltage relation with good accuracy and fast computational speed. However, poorly selected training data can introduce system-specific model bias, and the degree to which the training data represent the full range of operating conditions can affect the ability of data-driven models to generalize. In contrast, numerical computational fluid dynamics (CFD) models can be used to model the complicated transport phenomena and electrochemical behaviors within multidimensional components of fuel cells over a wide range of operational conditions. The numerical CFD models take into account detailed transport phenomena and electrochemical behaviors in various PEM fuel cell components, such as membranes, catalyst layers (CLs), gas diffusion layers (GDLs), flow channels, and bipolar plates [31,32], as presented in Fig. 14.1. The fuel cell operation involves reactant transport via diffusion and convection, two-phase water flow, phase change, energy transfer, electron and proton conduction, and electrochemical kinetics [33
36]. Many 3D CFD models were developed to simulate the PEM fuel cell operation processes with high fidelity [34,37
39]. However, 3D CFD models are usually computationally slow due to a large number of iterations and meshes, which is not suitable for fuel cell control. Many reduced-dimensional CFD models have been established with reasonable assumptions to accelerate the computing speed. Among these reduced-dimensional models, 1D and pseudo-2D models demonstrated the potential to be used to control fuel cell operation due to the fast computational speed in comparison with 3D models, high fidelity in comparison with empirical and analytical models, and less dependence on a large quantity of experimental data in comparison with data-driven models [7,40
52]. Therefore this chapter focuses on the review and recent progress of control-oriented CFD models for PEM fuel cells. The development and comparison of various 1D CFD fuel cell models are reviewed in Section 14.2, while the details of pseudo-2D models are examined and compared in

Computational fluid dynamics models Chapter | 14

395

FIGURE 14.1 PEM fuel cell components, accessories, and operational principles [53]. PEM, Polymer electrolyte membrane. Reprinted from Y. Zhao, Y. Liu, G. Liu, Q. Yang, L. Li, Z. Gao, Air and hydrogen supply systems and equipment for PEM fuel cells: a review, Int. J. Green. Energy 00 (2021) 1
18. doi:10.1080/15435075.2021.1946812.

Section 14.3. Finally, the modeling accuracy and computational speed reported in various control-oriented models are summarized in Section 14.4, and a summary of this chapter is given in Section 14.5.

14.2 1D computational fluid dynamics model The 1D CFD model addresses the nonlinear transport phenomena and electrochemical behaviors within fuel cell components in one direction—along-theflow-channel or through-the-membrane directions. Most 1D fuel cell models focus on the through-the-membrane direction, which is the focus of this section. Table 14.1 summarizes various 1D through-the-membrane CFD fuel cell models developed in the past three decades. The majority of the models summarized in Table 14.1 are established by inputting the operational variables of reactant conditions (composition, flow rate, temperature, pressure, and RH), cell temperature, and current density, which are the key variables that determine the overall fuel cell performance. The most frequently reported output variable is the cell voltage, while based on the nature of CFD models, other performance indicators, such as cell resistance, water flow rates, and temperature, can also be calculated.

TABLE 14.1 1D computational fluid dynamics models for polymer electrolyte membrane fuel cells Model features

Springer et al.

Bernardi and Verbrugge

Wohr ¨ et al.

Baschuk and Li

Rowe and Li

Bao et al.

Falca˜o et al.

Gao et al.

Li et al.

Abdin et al.

Jiang et al.

Year

1991

1992

1998

2000

2001

2006

2009

2010

2015
17

2016

2018

Membrane

O

O

O

O

O

O

O

O

O

O

O

CL

O

O

O

O

O

O

O

O

O

O

O

GDL

O

O

O

O

O

O

O

O

O

O

O

Flow channel

O

3

O

O

O

3

O

O

O

O

O

Bipolar plate

3

3

3

O

3

3

O

O

O

O

3

Cooling channel

3

3

3

3

3

3

3

O

3

3

3

Reactant RH

O

O

O

O

O

O

O

O

O

O

O

Reactant temperature

O

O

O

O

O

O

O

O

O

O

O

Reactant pressure

O

O

O

O

O

O

O

O

O

O

O

Cell temperature

O

O

O

O

O

O

O

O

O

O

O

Stoichiometry ratio/ flow rate

O

O

O

O

O

O

O

O

O

O

O

Current density

O

O

O

O

O

O

O

O

O

O

O

Domain

Key input variables

Physics included Activation loss

O

O

O

O

O

O

O

O

O

O

O

Ohmic loss

O

O

O

O

O

O

O

O

O

O

O

Multispecies transport

O

O

O

O

O

O

O

O

O

O

O

Membrane water

O

O

O

O

O

O

O

O

O

O

O

Two-phase flow

3

O

O

O

3

O

3

3

O

O

O

Electron transport

O

O

O

O

O

3

3

3

O

O

O

Proton transport

O

O

O

O

O

O

O

O

O

O

O

Nonisothermal

3

3

O

3

O

3

O

O

O

3

O

Convection

3

3

O

3

3

3

3

3

3

3

3

Steady-state behavior

O

O

O

O

O

O

O

3

O

O

O

Transient behavior

3

3

O

3

3

3

3

O

O

3

3

Aging effects

3

3

3

3

3

3

3

3

O

3

3

Voltage

O

O

O

O

O

O

O

O

O

O

O

Outlet pressure

3

3

3

3

3

O

3

3

3

3

3

Outlet species composition

O

3

3

3

3

3

3

3

3

3

3

Outputs reported

(Continued )

TABLE 14.1 (Continued) Model features

Springer et al.

Bernardi and Verbrugge

Wohr ¨ et al.

Baschuk and Li

Rowe and Li

Bao et al.

Falca˜o et al.

Gao et al.

Li et al.

Abdin et al.

Jiang et al.

Outlet species flow rate

O

3

3

3

3

3

3

3

3

3

3

Temperature

3

3

3

3

3

3

3

O

3

3

3

Resistance

O

3

3

3

3

3

3

3

3

3

3

ECSA

3

3

3

3

3

3

3

3

O

3

3

Steady-state I
V curve

3

3

3

O

O

O

O

3

3

O

O

Transient voltage variation

3

3

3

3

3

3

3

O

O

3

3

Other physical variables

O resistance

3

3

3

3

3

3

O temp.

3

3

3

Computation speed

3

3

3

3

3

3

3

O

3

3

3

References

[7]

[49]

[52]

[51]

[42]

[46]

[41]

[43]

[47,48]

[44]

[54]

Validation

Note: O and 3 denote the features that are included and excluded in the corresponding studies, respectively. Source: Adapted from J. Zhao, X. Li, C. Shum, J. McPhee, A review of physics-based and data-driven models for real-time control of polymer electrolyte membrane fuel cells, Energy AI 6 (2021) 100114. doi:10.1016/j.egyai.2021.100114.

Computational fluid dynamics models Chapter | 14

399

Springer et al. [7] established a 1D mathematical PEM fuel cell model in the 1990s to predict fuel cell performance. Their models incorporated the steady-state diffusion of reactant species, water vapor, and dissolved water in membranes in the computational domain of membrane, CLs, GDLs, and flow channels. If the partial pressure of water vapor is higher than its saturation value at a given temperature, excessive vapor will be condensed. In Springer et al.’s work, the liquid water is assumed to exist in pores as uniformly distributed water drops with negligible volume. The properties of the droplets are assumed to be identical to those of water vapor. Bernardi and Verbrugge [49] further developed a 1D steady-state fuel cell model based on an MEA, treating the CL as a macro-homogeneous layer and assuming the membrane as fully saturated. Therefore the transport properties of dissolved water in membranes were assumed constant. The transport of gas species and liquid water in the pores of GDLs and CLs was simulated. Weisbrod et al. [50] developed a detailed mathematical model for membranes taking into account the effect of water content. The relation between the water content in membranes and the transport coefficients and chemical properties of membranes was established based on the experimental data reported in [7]. Baschuk and Li [51] numerically investigated the effect of liquid water in electrodes on the performance of PEM fuel cells. The output voltage of the fuel cells is obtained by solving the species transport, Nernst equation, Butler
Volmer equations, and electron and proton conduction equations, while energy transport is omitted. Rowe and Li [42] further integrated the nonisothermal effect in fuel cell components, while the two-phase flow was not considered. W¨ohr et al. [52] incorporated both liquid water and thermal heat in their 1D CFD model. This enabled the prediction of dynamic behaviors of PEM fuel cells when external load, reactant flow rate, and humidity are changed. Falca˜o et al. [41] developed a steady-state heat and water management model for PEM fuel cells. In GDLs, the thermal energy transfer was assumed to be dominated by conduction, and in CLs, the heat generation due to the reaction was simulated. The mechanisms of water transport were assumed to be governed by electroosmotic drag and diffusion. Abdin et al. [44] developed a simplified 1D fuel cell model aiming to increase the computing speed. Four sub-models were included in their simulation, namely, anode, cathode, membrane, and voltage. Simple mass conservation laws were imposed in the anode, cathode, and membrane models such that the effect of operating condition changes on voltages was modeled. The current and temperature are assumed to be uniform, and water is assumed to be in the vapor phase at the interface between membranes and CLs. Diffusion is assumed to be dominated in electrodes as no pressure gradient is considered in their model. The model was further implemented in the MATLAB/Simulink with fast computing speed, and the simulation results were found in good agreement with the experimentally determined polarization curve.

400

Fuel Cells for Transportation

Gao et al. [43] developed a 1D dynamic model with three major multiphysics mechanisms—electrical, fluids, and energy. These multi-physics phenomena are solved in various components of fuel cells, including membranes, electrodes, flow distribution plates, and coolant plates. The activation overpotential in the cathodic electrode was empirically determined from an experimentally measured polarization curve for a practical fuel cell stack. The authors reported that this model is about 30% faster than the real-world operation of PEM fuel cells with good prediction accuracy. Jiang et al. [54] established a 1D fuel cell model that considers the interior temperature variation, two-phase flow, and phase changes. They proposed a switching function to evaluate the water states in the void regions of PEM fuel cells. This switching function coupled the liquid water and vapor within the same region in a single governing equation. A sensitivity analysis demonstrated that the physical and empirical coefficients have a strong impact on the stability and accuracy of the modeling results. Recently, degradation mechanisms have been incorporated into CFD models to forecast long-term fuel cell operation. Li et al. [47] integrated catalyst degradation and electrochemical surface area (ECSA) reduction in traditional 1D models. Two mechanisms, Ostwald ripening and catalyst dissolution as shown in Fig. 14.2, were modeled, and the impact of temperature, humidity, and reactants on the availability of reaction sites and long-term

FIGURE 14.2 Schematic of 1D Pt degradation model taking Pt Ostwald ripening and re-precipitation into account. Reprinted from Y. Li, K. Moriyama, W. Gu, S. Arisetty, C.Y. Wang, A one-dimensional Pt degradation model for polymer electrolyte fuel cells, J. Electrochem. Soc. 162 (2015) F834
F842. https://doi.org/10.1149/2.0101508jes.

Computational fluid dynamics models Chapter | 14

401

performance degradation were evaluated. The authors pointed out that catalytic losses are significant when Pt loading is small, and this limited the further decrease of Pt loading. Li and Wang [48] modified the agglomerate catalyst layer model to calculate the ECSA reduction due to catalyst degradation and oxygen transport resistance in the wrapped ionomer layer on the reaction site surface. The agglomerate model is coupled with a traditional 1D model to investigate the nonuniform Pt degradation within the CLs. It will be important to incorporate the mechanical, thermal, and chemical/electrochemical degradation mechanisms in further fuel cell model development for better long-term performance estimation. It is rarely reported in the literature to use the 1D model to control PEM fuel cells, while many studies demonstrated its capabilities to control fuel cells in a real-time manner. For instance, Gao et al. [43] claimed that their 1D model, incorporated electrical, fluid, and thermal phenomena in all essential cell components, including membrane, electrodes, bipolar, and coolant plates, can be about 30% faster than the real-world process. The 1D model assumes that the properties along the flow channel direction are uniform, and this is reasonable when the flow rate is sufficiently large and the temperature distribution is marginal.

14.3 Pseudo-2D computational fluid dynamics model To address the nonuniformity along the flow channel direction, pseudo-2D (or 1 1 1) CFD model is also widely investigated [12,40,45] due to its high computational speed and relatively high fidelity. The most commonly used pseudo-2D model is a combination of 1D through-the-membrane CFD model and simplified along-the-flow-channel model based on simplified mass or energy conservation laws [55
62]. Table 14.2 summarizes the pseudo-2D CFD models for PEM fuel cells developed in the past three decades. Similar to 1D models, most pseudo-2D models input the temperature, pressure, and humidity of reactants, cell temperature, flow rates, and current and output overall or distribution of temperature, species concentration, voltage, and current, along the reactant streams. Fuller and Newman [55] established a pseudo-2D fuel cell model for an MEA in the 1990s. The anode was fed with humidified reformed methanol, which is composed of H2, CO2, and vapor, while the cathode was filled with humidified air cocurrently. In their model, the transport of reactants and heat transfer were calculated in the through-the-membrane direction as a 1D model, while the mass conservation was imposed in the flow channels which are divided into several segments. The authors compared isothermal and nonisothermal assumptions and concluded that the water and heat are closely related, which can significantly affect the overall fuel cell performance.

TABLE 14.2 Pseudo-2D computational fluid dynamics models for polymer electrolyte membrane fuel cells. Model features

Fuller and Newman

Nguyen and White

Shamardina et al.

Chupin et al.

Goshtasbi et al.

Yang et al.

Year

1993

1993

2010

2010

2016
19

2019

O

O

O

O

O

O

Domain Membrane



3



O

CL

O

O

O

3

GDL

O

O

O

O

O

O

Flow channel

O

O

O

O

O

O

Bipolar plate

3

O

3

O

O

O

Cooling channel

3

3

3

O

O

O

O

O




O

O

O

Key input variables Reactant RH Reactant temperature

O

O







O

O

Reactant pressure

O

O

O

O

O

O

Cell temperature

O

O

O

3 

O

O

Stoichiometry ratio/flow rate

O

O

O

O

O

Current density

3



O

3



3



3

O 

O

Physics included Activation loss




O

O

O

O

O

Ohmic loss

O

O

O

O

O

O

Multispecies transport

O

O

O

O

O

O

Membrane water

O

O

3

O

O

O

Two-phase flow

3

O

3

O

O

O

Ice formation

3

3

3

3

3

O

Nonisothermal

O

O

3

O

O

O

Convection

3

3

3

3

3

3

Steady-state behavior




O

O

O

O

O

Transient behavior




3

3

3

O

O

Reactant crossover

3

3

O

3

O

3

Pressure drop

3

3

3

3

O

O

Voltage distribution

3

O

3

3

3

3

Current density distribution

O

O

3

O

3

3

Pressure distribution

3

3

3

3

3

3

Reactant distribution

O

O

3

3

3

3

Water distribution

O

O

3

O

O

O

Temperature distribution

O

O

3

O

O

O

Outputs reported

(Continued )

TABLE 14.2 (Continued) Model features

Fuller and Newman

Nguyen and White

Shamardina et al.

Chupin et al.

Goshtasbi et al.

Yang et al.

Validation

3

3

O

3

O

O

Polarization curve

3

3

O

3




O

Computation speed

3

3

O

3

O

3

References

[55]

[56]

[59]

[60]

[40,63]

[61,62]



Note: O, 3 , and—denote the features that are included, excluded, and not mentioned in the corresponding studies, respectively, means the model utilizes voltage as input and current as output,   means values are given indirectly,    means coolant temperature is given, and    means catalyst layer is treated as interface. Source: Adapted from J. Zhao, X. Li, C. Shum, J. McPhee, A review of physics-based and data-driven models for real-time control of polymer electrolyte membrane fuel cells, Energy AI 6 (2021) 100114. doi:10.1016/j.egyai.2021.100114.

Computational fluid dynamics models Chapter | 14

405

Nguyen and White [56] established a pseudo-2D model to investigate the steady-state behaviors of PEM fuel cells. The liquid water was assumed to be zero-volume droplets, and no pressure gradient is considered in the flow channels. Due to the high conductivity of bipolar plates, the voltage variation along the flow channel direction was ignored. Based on the model, the authors implemented different humidity control strategies. The modeling results indicated that the cathode reactants should be humidified if air is supplied instead of pure oxygen and anode reactants can be humidified to compensate for insufficient back-diffusion of water through the membranes. Shamardina et al.’s [59] steady-state pseudo-2D model considered the reactant transport between anode and cathode and the consumption of oxygen along the flow stream. To address the gas crossover of O2 and H2 through the membrane, a pseudo-crossover current was defined. In their model, the membrane conductivity was assumed to be independent of water contents, and this may underestimate the ohmic resistance when the membrane is dry. The authors reported that it requires only several seconds to simulate the fuel cell operation when operating variables are changed based on a computer, although the details of the computing devices were not given. Chupin et al. [60] established a steady-state pseudo-2D model, which combined the two-phase flow and an agglomerate catalyst layer model. As shown in Fig. 14.3A, a 1D model was established to model the transport of reactants, heat transfer, and electrochemical kinetics between anode and cathode, while the mass conservation law was imposed in the flow channels.

(A)

(B)

FIGURE 14.3 Schematic of the pseudo-2D PEM fuel cell model by Chupin et al.: (A) flow channel and (B) cooling channel. PEM, Polymer electrolyte membrane. Reprinted from S. Chupin, T. Colinart, S. Didierjean, Y. Dube´, K. Agbossou, G. Maranzana et al., Numerical investigation of the impact of gas and cooling flow configurations on current and water distributions in a polymer membrane fuel cell through a pseudo-two-dimensional diphasic model, J. Power Sources 195 (2010) 5213
5227. https://doi.org/10.1016/j.jpowsour.2010.03.027.

406

Fuel Cells for Transportation

The authors concluded that the coolant flow direction has a marginal effect on the overall fuel cell performance but a significant effect on the current distribution along the flow channels, as shown in Fig. 14.3B. The comparison of cocurrent and counterflow of anode and cathode streams suggested that the counterflow mode is more stable than the cocurrent flow. Goshtasbi et al. [40] established a semiempirical pseudo-2D dynamic model for PEM fuel cells. Their model incorporated heat and liquid water transport within cell components with cell voltage as the input variable and current as the output variable. The membrane water content was assumed to be linearly distributed across a thin membrane, and this simplified the governing equation of membrane water as an ordinary differential equation (ODE) below instead of a traditional partial differential equation (PDE). Fig. 14.4A presents the pseudo-2D framework of the PEM fuel cell model,

FIGURE 14.4 Pseudo-2D PEM fuel cell model developed by Goshtasbi et al.: (A) pseudo-2D model and (B) numerical solution method. PEM, Polymer electrolyte membrane. Adapted from A. Goshtasbi, B.L. Pence, T. Ersal, Computationally efficient pseudo-2D non-isothermal modeling of polymer electrolyte membrane fuel cells with two-phase phenomena, J. Electrochem. Soc. 163 (2016) F1412
F1432. https://doi.org/10.1149/2.0871613jes.

Computational fluid dynamics models Chapter | 14

407

which is composed of 25 nodes along the flow streams and 43 nodes from anode to cathode. The 1D along-the-flow channel model and the flow channel model were connected by modifying the boundary conditions in the corresponding computational domain. This weak connection allowed adopting various numerical schemes to solve different governing equations. The energy governing equation was solved, enabled via a fully implicit backward Euler scheme, and the liquid water governing equation is solved using the BEAM and Warming implicit scheme. These implicit schemes applied allowed to adopt large time steps and accelerate the computing speed corresponding to the real-world operation. The overall differential-algebraic equations were solved using Newton’s method, as shown in Fig. 14.4B. The authors also noted that parallel computing can be applied to this model to accelerate the computational speed. In this study, the gas dynamics, which corresponds to 0.01 seconds, was neglected, and the time step of 0.1 seconds was proposed for optimized modeling accuracy and computing speed. Overall, the computing speed was reported to be 2
4 times higher than the real-world cell operation. Yang et al. [61] developed a dynamic system-level fuel cell model, which is composed of a pseudo-2D dynamic fuel cell stack model integrated with humidifier, H2 pump, radiator, and air compressor models, as shown in Fig. 14.5. Explicit schemes were adopted to solve the PDEs with a very small time step of 1 μs. This work demonstrated the prediction of the dynamic operation of fuel cell systems with reasonable accuracy, and their model can be adopted for controlling the whole fuel cell system if the computing speed can be accelerated. The pseudo-2D model can be potentially used for fuel cell stacks and systems control in a real-time manner with high fidelity and computing

FIGURE 14.5 Schematic of the pseudo-2D PEM fuel cell stack model by Yang et al. PEM, Polymer electrolyte membrane. Reprinted from Z Yang, Q Du, Z Jia, C Yang, J Xuan, K. Jiao, A comprehensive proton exchange membrane fuel cell system model integrating various auxiliary subsystems, Appl. Energy 256 (2019) 113959. https://doi.org/10.1016/j.apenergy.2019.113959.

408

Fuel Cells for Transportation

speed. The computing speed is determined by the total of assumptions applied, physicochemical phenomena incorporated in the model, numerical methods to solve governing equations, and time step size [40,61
63]. In comparison with 1D models, the pseudo-2D model included the nonuniformity of reactants, temperature, and other parameters along the flow channel on the basis of 1D models. This enabled the prediction of nonuniform distribution of reactant concentration, temperature, pressure, current, voltage, and other parameters, which can provide useful information for fuel cell control to avoid harmful operations.

14.4 Model accuracy and computing speed A fuel cell model with higher fidelity will be more accurate within a wide range of operational conditions, while the demand for computing resources is usually contrariwise [64]. As shown in Fig. 14.6, empirical and analytical models are based on simplified empirical and analytical equations, which are usually computationally light but ignore many details of the fuel cell operation. The 3D and 2D models can capture more details and accurately simulate the fuel cell performance, while these models often involve many iterations and meshes that require a large amount of computing time, varying from hours to months. Data-driven models are highly dependent on the quality and quantity of experimental data. When the data-driven model is well trained, it can be used to predict the fuel cell performance with high accuracy and fast computing speed. If the experimental data used for model training are sparse, noisy, or coverage of the potential range operating conditions is insufficient, the prediction accuracy will decrease. The 1D and pseudo-2D CFD models can be balanced between computing speed and prediction accuracy.

FIGURE 14.6 Comparison of various control-oriented fuel cell models. Adapted from J. Zhao, X. Li, C. Shum, J. McPhee, A review of physics-based and data-driven models for real-time control of polymer electrolyte membrane fuel cells, Energy AI 6 (2021) 100114. https://doi.org/ 10.1016/j.egyai.2021.100114.

Computational fluid dynamics models Chapter | 14

409

The computing speed of reduced-dimensional CFD models is determined by the simplification of transport and electrochemical phenomena and the prediction accuracy required by the control demand. Theoretically, a model incorporating more phenomena can lead to more prediction accuracy over a wide range of operating conditions but require more computational time. Empirical and analytical models are based on empirical correlation or simple conversation laws, which are usually computationally fast and have been reported to be implemented in model predictive control of fuel cells [2,4
6]. However, the accuracy of these models is still under debate beyond the narrow range of operating conditions in which the model is validated. In contrast, the 1D and pseudo-2D fuel cell models are able to incorporate complicated transport and electrochemical phenomena inside different components, while maintaining high computing speed. Although having not been directly implemented in fuel cell control, the 1D and pseudo-2D models are 30%
50% faster than the real-world operations [40,43]. The 1D and pseudo-2D models can be further improved by omitting insignificant phenomena, simplifying governing equations, and optimizing algorithms and numerical schemes. The prediction accuracy of CFD models over a wide range of PEM fuel cell operations is affected by the multi-scale, multiphase, and multidimensional phenomena inside the fuel cells. These phenomena include but are not limited to activation loss, ohmic loss, multispecies transport, membrane water, two-phase flow, ice formation, nonisothermal, convection, dynamic behavior, fuel crossover, and pressure drop. The 3D and 2D models demonstrate high fidelity as the governing equations can be solved iteratively with a fine-mesh domain, which is able to simulate more detailed interior fuel cell dynamics. Another challenge for CFD modeling is to accurately estimate the transport properties, such as effective diffusion coefficients, permeability, effective conductivity, ECSA, and phase change coefficients, in different cell components [65
69]. However, discrepancies exist in different literature studies to mathematically express the above transport properties. The modeling accuracy can also be determined by different mechanical, thermal, and chemical/electrochemical degradation mechanisms, and the models can be improved if the operation history is recorded and analyzed [70
74]. The implementation of CFD models on the electronic control unit may require proper simplification of different governing equations, which may reduce the prediction accuracy [75]. It is found that the modeling accuracy and computational speed are often in a trade-off relation, and pseudo-2D and 1D CFD models have demonstrated the potential to balance them in a realtime control application [40,54,63]. A more detailed summary of the advantages and disadvantages of various control-oriented fuel cell models is given in Table 14.3.

TABLE 14.3 Comparison of multidimensional polymer electrolyte membrane fuel cell performance models. Model 3D and pseudo-3D CFD

Schematic

Advantages G G

G G

2D CFD

G

G

Pseudo-2D CFD

G

G

G

1D CFD

G

G

G

Challenges

Best accuracy Detailed 3D physical and electrochemical phenomena Assistance with component design Reduced experimental efforts

G

Faster calculation than 3D models (B mins per steady-state case) Reasonable accuracy under certain conditions

G

Can be faster than the real time (reported to be 50% faster) Reasonable accuracy under specific conditions Potential direct implementation on ECU

G

Can be faster than the real time (reported to be 33% faster) Reasonable accuracy under specific conditions Potential direct implementation on ECU

G

G

G

G G G

G

G G

G

Computationally expensive (from hours to months) Accuracy relies on many transport and electrochemical properties and geometry of cell components Not suitable for implementation on the available ECU Slow computational speed for ECU implementation Transport phenomena in channels are simplified More assumptions than 3D Careful validation needed

Requires strict validation against experimental data Weak connection in the along-the-channel direction

More assumptions required Requires strict validation against experimental data May not be able to predict the performance within the entire operational range with sufficient accuracy Trade-off relation between accuracy and computing speed

0D (analytical)

G G G G

0D (empirical)

G G

G

G

Data-driven

G

G

G

G

Theoretically faster than 1D models System analysis and design Thermodynamic analysis Used for model predictive control studies

G

Theoretically faster than 1D models Good accuracy with proper curve fitting Used for model predictive control studies Direct implementation on ECU

G

Fast calculation after training (,a few seconds) Excellent accuracy once properly trained No requirement on the knowledge of the fuel cell design No requirement on transport or electrochemical coefficients

G

G

G G

G G

G

Requires strict validation against experimental data Accuracy over the entire range of operation may not be reliable

Requires well-structured experimental data for curve fitting to determine the unknown coefficient Only valid for validated cases Accuracy over the entire range of operation may not be reliable

Algorithm and model architecture should be properly selected A high volume of experimental data is required Experimental data should be representative of typical operations Uncertain accuracy when operating fuel cell in extreme conditions that are not trained

CFD, Computational fluid dynamics; ECU, electronic control unit. Source: Adapted from J. Zhao, X. Li, C. Shum, J. McPhee, A review of physics-based and data-driven models for real-time control of polymer electrolyte membrane fuel cells, Energy AI 6 (2021) 100114. doi:10.1016/j.egyai.2021.100114.

412

Fuel Cells for Transportation

14.5 Summary Control-oriented models of PEM fuel cells with good fidelity and computational speed are the core of model predictive control to optimize fuel cell operation for better performance and a longer lifetime. CFD models demonstrate a good balance between the two trade-off factors—accuracy and computing speed—and less dependence on a significantly large number of experimental data in comparison to data-driven models. In this chapter, the recent development of reduced-dimensional CFD models, 1D and pseudo2D, which can be potentially used for fuel cell control has been reviewed. Most 1D and pseudo-2D models do not take all fundamental phenomena into account; instead, various simplifications of the governing equations have been applied to the conservation and transport of mass, momentum, thermal heat, reactant species, product species, two-phase flow, dissolved water in membranes. The electrochemical kinetics is also treated as empirical correlations in some studies despite the accurate electrochemical kinetics theory is still under debate. To implement pseudo-2D and 1D CFD models in fuel cell controls, the model development still needs improvement, for example, (1) improving modeling accuracy over a wide range of operational conditions, (2) creating a comprehensive and large experimental database for structural, transport, and electrochemical properties of different fuel cell components as well as overall polarization curves, (3) taking degradation mechanisms into account for long-term performance estimation, and (4) balancing the model simplification and computing speed.

Acknowledgments This work received financial support from Toyota Motor Engineering & Manufacturing North America, Inc., Toyota Motor Manufacturing Canada, and Natural Sciences and Engineering Research Council of Canada through a Collaborative Research and Development Grant with the project number CRDPJ 543945-19.

References ¨ . Ekici, M. K¨oksal, Process modeling of an automotive PEM fuel cell system, [1] A. Akroot, O Int. J. Green. Energy 16 (2019) 778
788. Available from: https://doi.org/10.1080/ 15435075.2019.1641105. [2] W.R.W. Daud, R.E. Rosli, E.H. Majlan, S.A.A. Hamid, R. Mohamed, T. Husaini, PEM fuel cell system control: a review, Renew. Energy 113 (2017) 620
638. Available from: https://doi.org/10.1016/j.renene.2017.06.027. [3] S.S. Khan, H. Shareef, A.H. Mutlag, Dynamic temperature model for proton exchange membrane fuel cell using online variations in load current and ambient temperature, Int. J. Green. Energy 16 (2019) 361
370. Available from: https://doi.org/10.1080/15435075. 2018.1564141. [4] C.H. Woo, J.B. Benziger, PEM fuel cell current regulation by fuel feed control, Chem. Eng. Sci. 62 (2007) 957
968. Available from: https://doi.org/10.1016/J.CES.2006.10.027.

Computational fluid dynamics models Chapter | 14

413

[5] D. Wu, C. Peng, C. Yin, H. Tang, Review of system integration and control of proton exchange membrane fuel cells, Electrochem. Energy Rev. 3 (2020) 466
505. Available from: https://doi.org/10.1007/s41918-020-00068-1. [6] V. Das, S. Padmanaban, K. Venkitusamy, R. Selvamuthukumaran, F. Blaabjerg, P. Siano, Recent advances and challenges of fuel cell based power system architectures and control—a review, Renew. Sustain. Energy Rev. 73 (2017) 10
18. Available from: https://doi.org/10.1016/J.RSER.2017.01.148. [7] T. Springer, T. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell model, J. Electrochem. Soc. 138 (1991) 2334
2341. [8] M.M. Hussain, J.J. Baschuk, X. Li, I. Dincer, Thermodynamic analysis of a PEM fuel cell power system, Int. J. Therm. Sci. 44 (2005) 903
911. Available from: https://doi.org/ 10.1016/j.ijthermalsci.2005.02.009. [9] S. Srinivasan, E.A. Ticianelli, C.R. Derouin, A. Redondo, Advances in solid polymer electrolyte fuel cell technology with low platinum loading electrodes, J. Power Sources 22 (1988) 359
375. Available from: https://doi.org/10.1016/0378-7753(88)80030-2. [10] G. Squadrito, G. Maggio, E. Passalacqua, F. Lufrano, A. Patti, An empirical equation for polymer electrolyte fuel cell (PEFC) behaviour, J. Appl. Electrochem. 29 (1999) 1449
1455. Available from: https://doi.org/10.1023/A:1003890219394. [11] G. Karimi, J.J. Baschuk, X. Li, Performance analysis and optimization of PEM fuel cell stacks using flow network approach, J. Power Sources 147 (2005) 162
177. Available from: https://doi.org/10.1016/J.JPOWSOUR.2005.01.023. [12] A. Miotti, A. di Domenico, Y.G. Guezennec, S.V. Rajagopalan, Control-oriented model for an automotive PEM fuel cell system with imbedded 1 1 1D membrane water transport, in: 2005 IEEE Vehicle Power and Propulsion Conference, IEEE, 2005, pp. 611
618. Available from: https://doi.org/10.1109/VPPC.2005.1554622. [13] F. Musio, F. Tacchi, L. Omati, P. Gallo Stampino, G. Dotelli, S. Limonta, et al., PEMFC system simulation in MATLAB-Simulinks environment, Int. J. Hydrog. Energy 36 (2011) 8045
8052. Available from: https://doi.org/10.1016/j.ijhydene.2011.01.093. [14] E. Hosseinzadeh, M. Rokni, A. Rabbani, H.H. Mortensen, Thermal and water management of low temperature proton exchange membrane fuel cell in fork-lift truck power system, Appl. Energy 104 (2013) 434
444. Available from: https://doi.org/10.1016/j. apenergy.2012.11.048. [15] E. Hosseinzadeh, M. Rokni, Development and validation of a simple analytical model of the proton exchange membrane fuel cell (PEMFC) in a fork-lift truck power system, Int. J. Green. Energy 10 (2013) 523
543. Available from: https://doi.org/10.1080/ 15435075.2012.678525. [16] R. Long, B. Li, Z. Liu, W. Liu, A hybrid system using a regenerative electrochemical cycle to harvest waste heat from the proton exchange membrane fuel cell, Energy 93 (2015) 2079
2086. Available from: https://doi.org/10.1016/j.energy.2015.09.132. [17] Q. Xu, W. Zhang, J. Zhao, L. Xing, Q. Ma, L. Xu, et al., Effect of air supply on the performance of an active direct methanol fuel cell (DMFC) fed with neat methanol, Int. J. Green. Energy 15 (2018) 181
188. Available from: https://doi.org/10.1080/15435075. 2018.1431547. [18] X. Zhang, D. Yang, M. Luo, Z. Dong, Load profile based empirical model for the lifetime prediction of an automotive PEM fuel cell, Int. J. Hydrog. Energy 42 (2017) 11868
11878. Available from: https://doi.org/10.1016/J.IJHYDENE.2017.02.146. [19] N.S. Sisworahardjo, T. Yalcinoz, M.Y. El-Sharkh, M.S. Alam, Neural network model of 100 W portable PEM fuel cell and experimental verification, Int. J. Hydrog. Energy 35 (2010) 9104
9109. Available from: https://doi.org/10.1016/J.IJHYDENE.2010.05.124.

414

Fuel Cells for Transportation

[20] J. Kim, S. Lee, S. Srinivasan, C.E. Chamberlin, Modeling of proton exchange membrane fuel cell performance with an empirical equation, J. Electrochem. Soc. 142 (1995) 2670
2674. Available from: https://doi.org/10.1149/1.2050072. [21] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R. Roberge, T.J. Harris, Performance modeling of the Ballard Mark IV solid polymer electrolyte fuel cell II. Empirical model development, J. Electrochem. Soc. 142 (1995) 9
15. Available from: https://doi.org/10.1149/1.2043959. [22] D. Chu, R. Jiang, C. Walker, Analysis of PEM fuel cell stacks using an empirical current
voltage equation, J. Appl. Electrochem. 30 (2000) 365
370. Available from: https:// doi.org/10.1023/A:1003905109007. [23] R. Cownden, M. Nahon, M. Rosen, Exergy analysis of a fuel cell power system for transportation applications, Exergy, Int. J. 1 (2001) 112
121. Available from: https://doi.org/ 10.1016/s1164-0235(01)00017-6. [24] A. Kazim, Exergy analysis of a PEM fuel cell at variable operating conditions, Energy Convers. Manag. 45 (2004) 1949
1961. Available from: https://doi.org/10.1016/j.enconman. 2003.09.030. [25] M.E. Youssef, K.E. Al-Nadi, M.H. Khalil, Lumped model for proton exchange membrane fuel cell (PEMFC), Int. J. Electrochem. Sci. 5 (2010) 267
277. [26] R. Tirnovan, S. Giurgea, Efficiency improvement of a PEMFC power source by optimization of the air management, Int. J. Hydrog. Energy 37 (2012) 7745
7756. Available from: https://doi.org/10.1016/j.ijhydene.2012.02.029. [27] L. Hong, J. Chen, Z. Liu, L. Huang, Z. Wu, A nonlinear control strategy for fuel delivery in PEM fuel cells considering nitrogen permeation, Int. J. Hydrog. Energy 42 (2017) 1565
1576. Available from: https://doi.org/10.1016/j.ijhydene.2016.07.240. [28] K. Promislow, J. St-Pierre, B. Wetton, A simple, analytic model of polymer electrolyte membrane fuel cell anode recirculation at operating power including nitrogen crossover, J. Power Sources 196 (2011) 10050
10056. Available from: https://doi.org/10.1016/j. jpowsour.2011.08.070. [29] I. Matraji, S. Laghrouche, M. Wack, Pressure control in a PEM fuel cell via second order sliding mode, Int. J. Hydrog. Energy 37 (2012) 16104
16116. Available from: https://doi. org/10.1016/j.ijhydene.2012.08.007. [30] M.A. Danzer, S.J. Wittmann, E.P. Hofer, Prevention of fuel cell starvation by model predictive control of pressure, excess ratio, and current, J. Power Sources 190 (2009) 86
91. Available from: https://doi.org/10.1016/j.jpowsour.2008.12.089. [31] S.R. Badduri, G.N. Srinivasulu, S.S. Rao, Experimental analysis of PEM fuel cell performance using lung channel design bipolar plate, Int. J. Green. Energy 16 (2019) 1591
1601. Available from: https://doi.org/10.1080/15435075.2019.1677238. [32] S. Mukherjee, A. Bates, S.C. Lee, D.-H. Lee, S. Park, A review of the application of CNTs in PEM fuel cells, Int. J. Green. Energy 12 (2015) 787
809. Available from: https://doi.org/10.1080/15435075.2013.867270. [33] G. Zhang, Z. Bao, B. Xie, Y. Wang, K. Jiao, Three-dimensional multi-phase simulation of PEM fuel cell considering the full morphology of metal foam flow field, Int. J. Hydrog. Energy (2020). Available from: https://doi.org/10.1016/j.ijhydene.2020.05.263. [34] J. Zhao, X. Li, Oxygen transport in polymer electrolyte membrane fuel cells based on measured electrode pore structure and mass transport properties, Energy Convers. Manag. 186 (2019) 570
585. Available from: https://doi.org/10.1016/J.ENCONMAN. 2019.02.042.

Computational fluid dynamics models Chapter | 14

415

[35] J. Macedo-Valencia, J.M. Sierra, S.J. Figueroa-Ramı´rez, S.E. Dı´az, M. Meza, 3D CFD modeling of a PEM fuel cell stack, Int. J. Hydrog. Energy 41 (2016) 23425
23433. Available from: https://doi.org/10.1016/j.ijhydene.2016.10.065. [36] C. Robin, M. Ge´rard, M. Quinaud, J. d’Arbigny, Y. Bultel, Proton exchange membrane fuel cell model for aging predictions: simulated equivalent active surface area loss and comparisons with durability tests, J. Power Sources 326 (2016) 417
427. Available from: https://doi.org/10.1016/J.JPOWSOUR.2016.07.018. [37] D.S. Falca˜o, P.J. Gomes, V.B. Oliveira, C. Pinho, A.M.F.R. Pinto, 1D and 3D numerical simulations in PEM fuel cells, Int. J. Hydrog. Energy 36 (2011) 12486
12498. Available from: https://doi.org/10.1016/J.IJHYDENE.2011.06.133. [38] Y. Li, Z. Zhou, X. Liu, W.-T. Wu, Modeling of PEM fuel cell with thin MEA under low humidity operating condition, Appl. Energy 242 (2019) 1513
1527. Available from: https://doi.org/10.1016/J.APENERGY.2019.03.189. [39] L.M. Pant, M.R. Gerhardt, N. Macauley, R. Mukundan, R.L. Borup, A.Z. Weber, Alongthe-channel modeling and analysis of PEFCs at low stoichiometry: development of a 1 1 2D model, Electrochim. Acta 326 (2019) 134963. Available from: https://doi.org/ 10.1016/j.electacta.2019.134963. [40] A. Goshtasbi, B.L. Pence, T. Ersal, Computationally efficient pseudo-2D non-isothermal modeling of polymer electrolyte membrane fuel cells with two-phase phenomena, J. Electrochem. Soc. 163 (2016) F1412
F1432. Available from: https://doi.org/10.1149/ 2.0871613jes. [41] D.S. Falca˜o, V.B. Oliveira, C.M. Rangel, C. Pinho, A.M.F.R. Pinto, Water transport through a PEM fuel cell: a one-dimensional model with heat transfer effects, Chem. Eng. Sci. 64 (2009) 2216
2225. Available from: https://doi.org/10.1016/j.ces.2009. 01.049. Contents. [42] A. Rowe, X. Li, Mathematical modeling of proton exchange membrane fuel cells, J. Power Sources 102 (2001) 82
96. Available from: https://doi.org/10.1016/S03787753(01)00798-4. [43] F. Gao, B. Blunier, A. Miraoui, El, A. Moudni, A multiphysic dynamic 1-D model of a proton-exchange-membrane fuel-cell stack for real-time simulation, IEEE Trans. Ind. Electron. 57 (2010) 1853
1864. Available from: https://doi.org/10.1109/TIE.2009.2021177. [44] Z. Abdin, C.J. Webb, E.M.A. Gray, PEM fuel cell model and simulation in Matlab
Simulink based on physical parameters, Energy 116 (2016) 1131
1144. Available from: https://doi.org/10.1016/j.energy.2016.10.033. [45] A. Goshtasbi, B.L. Pence, T. Ersal, A real-time pseudo-2D bi-domain model of PEM fuel cells for automotive applications, in: Proc. ASME 2017 Dynamic Systems and Control Conference, 2017, pp. 1
10. [46] C. Bao, M. Ouyang, B. Yi, Modeling and control of air stream and hydrogen flow with recirculation in a PEM fuel cell system—I. Control-oriented modeling, Int. J. Hydrog. Energy 31 (2006) 1879
1896. Available from: https://doi.org/10.1016/J.IJHYDENE. 2006.02.031. [47] Y. Li, K. Moriyama, W. Gu, S. Arisetty, C.Y. Wang, A one-dimensional Pt degradation model for polymer electrolyte fuel cells, J. Electrochem. Soc. 162 (2015) F834
F842. Available from: https://doi.org/10.1149/2.0101508jes. [48] Y. Li, C.Y. Wang, Modeling of transient platinum degradation in a low Pt-loading PEFC under current cycling, J. Electrochem. Soc. 164 (2017) F171
F179. Available from: https://doi.org/10.1149/2.0081704jes.

416

Fuel Cells for Transportation

[49] D.M. Bernardi, M.W. Verbrugge, A mathematical model of the solid-polymer-electrolyte fuel cell, J. Electrochem. Soc. 139 (1992) 2477
2491. Available from: https://doi.org/ 10.1149/1.2221251. [50] K.R. Weisbrod, S.A. Grot, N.E. Vanderborgh, Through-the-electrode model of a proton exchange membrane fuel cell, ECS Proc. (1995
23)(1995) 152
166. Available from: https://doi.org/10.1149/199523.0152PV. Vol. [51] J. Baschuk, X. Li, Modelling of polymer electrolyte membrane fuel cells with variable degrees of water flooding, J. Power Sources 86 (2000) 181
196. Available from: https:// doi.org/10.1016/S0378-7753(99)00426-7. [52] M. W¨ohr, K. Bolwin, W. Schnurnberger, M. Fischer, W. Neubrand, G. Eigenberger, Dynamic modelling and simulation of a polymer membrane fuel cell including mass transport limitation, Int. J. Hydrog. Energy 23 (1998) 213
218. Available from: https://doi. org/10.1016/S0360-3199(97)00043-8. [53] Y. Zhao, Y. Liu, G. Liu, Q. Yang, L. Li, Z. Gao, Air and hydrogen supply systems and equipment for PEM fuel cells: a review, Int. J. Green. Energy 00 (2021) 1
18. Available from: https://doi.org/10.1080/15435075.2021.1946812. [54] Y. Jiang, Z. Yang, K. Jiao, Q. Du, Sensitivity analysis of uncertain parameters based on an improved proton exchange membrane fuel cell analytical model, Energy Convers. Manag. 164 (2018) 639
654. Available from: https://doi.org/10.1016/j.enconman.2018.03.002. [55] T.F. Fuller, J. Newman, Water and thermal management in solid-polymer-electrolyte fuel cells, J. Electrochem. Soc. 140 (1993) 1218
1225. Available from: https://doi.org/ 10.1149/1.2220960. [56] T.V. Nguyen, R.E. White, A water and heat management model for proton-exchangemembrane fuel cells, J. Electrochem. Soc. 140 (1993) 2178
2186. Available from: https://doi.org/10.1149/1.2220792. [57] S. Um, C. Wang, K. Chen, Computational fluid dynamics modeling of proton exchange membrane fuel cells, J. Electrochem. Soc. 147 (2000) 4485
4493. [58] C.-Y. Wang, Fundamental models for fuel cell engineering, Chem. Rev. 104 (2004) 4727
4766. Available from: https://doi.org/10.1021/CR020718S. [59] O. Shamardina, A. Chertovich, A.A. Kulikovsky, A.R. Khokhlov, A simple model of a high temperature PEM fuel cell, Int. J. Hydrog. Energy 35 (2010) 9954
9962. Available from: https://doi.org/10.1016/J.IJHYDENE.2009.11.012. [60] S. Chupin, T. Colinart, S. Didierjean, Y. Dube´, K. Agbossou, G. Maranzana, et al., Numerical investigation of the impact of gas and cooling flow configurations on current and water distributions in a polymer membrane fuel cell through a pseudo-twodimensional diphasic model, J. Power Sources 195 (2010) 5213
5227. Available from: https://doi.org/10.1016/j.jpowsour.2010.03.027. [61] Z. Yang, Q. Du, Z. Jia, C. Yang, J. Xuan, K. Jiao, A comprehensive proton exchange membrane fuel cell system model integrating various auxiliary subsystems, Appl. Energy 256 (2019) 113959. Available from: https://doi.org/10.1016/j.apenergy.2019.113959. [62] Z. Yang, Q. Du, Z. Jia, C. Yang, K. Jiao, Effects of operating conditions on water and heat management by a transient multi-dimensional PEMFC system model, Energy 183 (2019) 162
476. Available from: https://doi.org/10.1016/j.energy.2019.06.148. [63] A. Goshtasbi, J. Chen, J.R. Waldecker, S. Hirano, T. Ersal, On parameterizing PEM fuel cell models, 2019 Am. Control Conf. (2019) 903
908. [64] J. Zhao, X. Li, C. Shum, J. McPhee, A review of physics-based and data-driven models for real-time control of polymer electrolyte membrane fuel cells, Energy AI 6 (2021) 100114. Available from: https://doi.org/10.1016/j.egyai.2021.100114.

Computational fluid dynamics models Chapter | 14

417

[65] J. Zhao, S. Shahgaldi, I. Alaefour, S. Yang, X. Li, Pore structure and effective diffusion coefficient of catalyzed electrodes in polymer electrolyte membrane fuel cells, Int. J. Hydrog. Energy 43 (2018) 3776
3785. Available from: https://doi.org/10.1016/j.ijhydene. 2018.01.019. [66] J. Zhao, A. Ozden, S. Shahgaldi, I.E. Alaefour, X. Li, F. Hamdullahpur, Effect of Pt loading and catalyst type on the pore structure of porous electrodes in polymer electrolyte membrane (PEM) fuel cells, Energy 150 (2018) 69
76. Available from: https://doi.org/ 10.1016/j.energy.2018.02.134. [67] S. Shahgaldi, J. Zhao, I. Alaefour, X. Li, Investigation of catalytic vs reactant transport effect of catalyst layers on proton exchange membrane fuel cell performance, Fuel 208 (2017) 321
328. Available from: https://doi.org/10.1016/j.fuel.2017.07.035. [68] A. Ozden, S. Shahgaldi, J. Zhao, X. Li, F. Hamdullahpur, Assessment of graphene as an alternative microporous layer material for proton exchange membrane fuel cells, Fuel 215 (2018) 726
734. Available from: https://doi.org/10.1016/j.fuel.2017.11.109. [69] J. Zhao, S. Shahgaldi, I. Alaefour, Q. Xu, X. Li, Gas permeability of catalyzed electrodes in polymer electrolyte membrane fuel cells, Appl. Energy 209 (2018) 203
210. Available from: https://doi.org/10.1016/j.apenergy.2017.10.087. [70] J. Zhao, S. Shahgaldi, X. Li, Z.(Simon) Liu, Experimental observations of microstructure changes in the catalyst layers of proton exchange membrane fuel cells under wet-dry cycles, J. Electrochem. Soc. 165 (2018) F3337
F3345. Available from: https://doi.org/ 10.1149/2.0391806jes. [71] X.-Z. Yuan, H. Li, S. Zhang, J. Martin, H. Wang, A review of polymer electrolyte membrane fuel cell durability test protocols, J. Power Sources 196 (2011) 9107
9116. Available from: https://doi.org/10.1016/j.jpowsour.2011.07.082. [72] J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. Available from: https://doi.org/10.1016/J.ENCONMAN. 2019.112022. [73] C. Wang, Q. Zhao, X. Zhou, J. Wang, Y. Tang, Degradation characteristics of membrane electrode assembly under drive cycle test protocol, Int. J. Green. Energy 16 (2019) 789
795. Available from: https://doi.org/10.1080/15435075.2019.1641712. [74] P. Novotny, M. Tomas, T. Nemec, L. Kullova, F. Marsik, On/off cycling test of lowtemperature PEM fuel cell at fully humidified conditions, Int. J. Green. Energy 16 (2019) 1189
1195. Available from: https://doi.org/10.1080/15435075.2019.1671394. [75] R. Masoudi, J. McPhee, Application of Karhunen
Loe´ve decomposition and piecewise linearization to a physics-based battery model, Electrochim. Acta 365 (2021) 137093. Available from: https://doi.org/10.1016/j.electacta.2020.137093.

This page intentionally left blank

Chapter 15

Fuel cell durability under automotive driving cycles— fundamentals and experiments Jian Zhao and Xianguo Li Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, Canada

15.1 Introduction Proton-exchange membrane (PEM) fuel cell is a power generation device that directly converts the energy of hydrogen fuels and oxygen gases into electricity, which can be used in a fuel cell vehicle because of its advantages in clean operation process, noiseless electrochemical reaction, rapid start-up and response to load changes, low gas emission, and high energy efficiency [1,2]. Many fuel cell vehicles were built by different auto manufacturers, such as Ford, General Motors, Honda, Hyundai, Nissan, Toyota, and Volkswagen [3]. However, the fuel cell vehicle still has not been widely accepted by customers due to its high cost, limited practical performance, and durability [4,5]. Among these three factors, the durability of fuel cells has drawn significant attention from both academia and industry. A typical single PEM fuel cell is composed of one membrane, dual catalyst layers (CLs), dual gas diffusion layers (GDLs), and dual bipolar plates (carved with flow channels for reactant gas supply) [6
10]. Fig. 15.1 shows the schematic diagram of a PEM fuel cell (part A) and the fuel cell stack integrated with auxiliary devices for automotive applications (part B). Details of the operation principle of a PEM fuel cell are included in Chapter 2. As shown in Fig. 15.1B, the automotive fuel cell stack requires four flow subsystems, including an H2 supply subsystem, an air supply subsystem, a coolant subsystem, and a humidification subsystem. Theoretically, the fuel cell lifetime is unlimited as long as fuels and oxidants are continuously supplied. However, all fuel cell components are subject to degradation under automotive driving cycles, which may lead to either irreversible longterm performance drop or the failure of different components. Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00004-6 © 2023 Elsevier Ltd. All rights reserved.

419

420

Fuel Cells for Transportation Electrons, e-

Electrons, eLOAD

eH2

DP

H+

AGDL

O2

eProtons, H+

O2 H+

ACL

PEM

AGDL: Anode gas diffusion layer ACL: Anode catalyst layer PEM: polymer electrolyte membrane

CCL

H2 O

CGDL

DP

CGDL: Cathode gas diffusion layer CCL: Cathode catalyst layer DP: Distribution plate

(A) U6 Traction Motor Contol

TM

U4 U5 Power Power Management Conditioning

Humidity Control

U3

Energy Storage (Battery)

Temperature Control

U1 Hydrogen Flow Control

S Hydrogen Tank

U2 Air Flow Control Motor Compressor Humidifier

S

Fuel Cell Stack

Water Separator Water Tank

(B) FIGURE 15.1 Schematic of a typical (A) single PEM fuel cell and (B) automotive fuel cell system [11,12]. PEM, Proton-exchange membrane. (A) Adapted from J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. https://doi.org/ 10.1016/J.ENCONMAN.2019.112022; (B) Adapted from J.T. Pukrushpan, A.G. Stefanopoulou, H. Peng, Background and introduction. Control. Fuel Cell Power Syst. (2004).

In a fuel cell vehicle, the fuel cell can be operated in two ways depending on its operation conditions, that is, power follow and soft run [13,14]. In the power-follow mode, the fuel cell stack is run under constant power conditions, which can minimize voltage and current fluctuations during automotive driving cycles. A battery will be adopted to store the electricity generated by the fuel cell stack and meet the dynamic energy demand. In the

Fuel cell durability Chapter | 15

421

soft-run mode, the fuel cell stack is used to directly provide dynamic power under different driving conditions, for example, acceleration and deceleration, load change, and cold start, which means dynamic changes in operational temperature, pressure, humidity, current, and voltage. Therefore the automotive driving conditions may significantly impact the fuel cell component degradation and lifetime. The proposed lifespan of PEM fuel cells for passenger cars is 5000 hours, and that for transit buses is five times longer according to the US Department of Energy (DOE) [15]. It is reported that the PEM fuel cell stack produced by Ballard Power Systems has been run continuously in a transit bus for more than 30,000 hours with no major repairs [16]. However, the definition of lifetime of fuel cells varies significantly in different studies. For instance, the US DOE defined the lifetime of fuel cells as the operation time when the rated power becomes less than 90% of the beginning of life (BOL) [15]. In Pei et al.’s studies [17], the lifetime of a fuel cell was defined as the timespan needed when the end-of-life (EOL) voltage is reduced by 10% under rated power, and their experimental results suggested that if the potential loss is larger than 10%, the power generation drops more rapidly. Marrony et al. [18] defined the lifetime of PEM fuel cells as the maximum service time before a component failure appears. Many experimental techniques have been developed to investigate the durability of PEM fuel cells to be used in automobiles. These methods can be classified into three categories—steady-state durability test, in situ accelerated stress test (AST), and ex situ AST. In the steady-state durability tests, the fuel cell is run under desired stationary driving conditions, which can directly determine its lifetime. However, the steady-state durability test is resources-demanding and time-consuming [11], and thus, in situ AST gained significant attention to shorten the durability test. In in situ ASTs, various accelerated stressors are imposed on actual fuel cell operations, including thermal fluctuation, humidification, current, voltage, startup
shutdown, and load changes. By understanding the relation between in situ AST and steadystate durability test, the lifetime of fuel cells can be estimated using less testing time and resources. To further reduce experimental efforts and directly recognize the causes of different failure modes, ex situ AST has been actively studied recently to test individual components without running an actual fuel cell stack. The ex situ AST can be used to investigate the effect of temperature, humidity, flow rate, current, voltage, water dynamics, freeze
thaw, clamping force, and vibration on component degradation. In this chapter, the fuel cell durability under automotive driving cycles is examined with a focus on fundamental mechanisms and various experimental techniques. First, the fundamental concepts and degradation mechanisms supported by experimental observations have been highlighted and discussed for the major fuel cell components, including membranes, CLs, GDLs, and bipolar plates. Then, the details of steady-state durability test and in situ AST have been discussed, including testing conditions, voltage drop rate,

422

Fuel Cells for Transportation

and test duration. The ex situ AST is also reported along with the discussion of root causes of various degradation modes in different components. Finally, a summary and concluding remarks are presented.

15.2 Fundamental degradation mechanisms under automotive driving cycles The lifetime of PEM fuel cells is theoretically unlimited as long as the fuels and oxidants are constantly supplied to flow channels. However, due to the existence of degradation in different components under automotive driving cycles, the long-term performance of PEM fuel cells will decay gradually. The rate of performance decay determines the fuel cell lifetime and is affected by many automotive driving conditions, typically, idling, dynamic load changes, and startup
shutdown [19]. Therefore the fundamental degradation mechanisms are examined in this section for different cell components. It should be mentioned that the durability studies reported in the literature are mainly focused on membrane, CL, GDL, and bipolar plate. The work on other cell components, such as sealing gaskets and endplates, and accessory equipment, such as pump, humidifier, compressor, and radiator, is rarely reported and therefore not discussed in this chapter.

15.2.1 Polymer electrolyte membrane Polymer electrolyte (proton-conducting membrane) is one of the most important components in PEM fuel cells, which lies between anode and cathode electrodes. The function of the membrane is to conduct protons, reject electrons, and separate fuels and oxidants. Thus, any structure and material defects may defeat the purpose of the solid membranes and shorten the lifetime of the whole fuel cell stack. Two types of membranes are widely used in PEM fuel cell applications—long- and short-side chain perfluoro sulfonic acid (PFSA) membranes. Examples of long-side chain membranes include Nafion, GoreSelect, Flemion, and Aciplex products [7,8,20
23], and shortside chain membranes include Aquivion, Hyflon, and Dow [24
29]. The membranes are subject to three types of degradations via mechanical, thermal, and electrochemical mechanisms [20], which can be accelerated under automotive driving cycles. The mechanical properties of membranes can be affected by pinholes, cracks, and interfacial delamination between CL and membrane, which are illustrated in Fig. 15.2 based on experimental work in [30,31]. The mechanical property deterioration is often initialized due to membrane electrode assembly (MEA) manufacturing deficiencies and propagated under automotive driving conditions [20]. The pinholes and cracks in membranes can increase fuel crossover, and when hydrogen is accumulated on the cathode side, direct combustions are possible, which will reduce the useful work produced by the fuel cell stack due to which the

Fuel cell durability Chapter | 15

423

FIGURE 15.2 Mechanical degradation modes of membranes: (A) pinhole and crack formation, and (B) delamination. (A) Adapted from C. Lim, L. Ghassemzadeh, F. Van Hove, M. Lauritzen, J. Kolodziej, G.G. Wang, et al., Membrane degradation during combined chemical and mechanical accelerated stress testing of polymer electrolyte fuel cells, J. Power Sources 257 (2014) 102
110. https://doi.org/10.1016/J.JPOWSOUR.2014.01.106; (B) Adapted from S. Kim, B.K. Ahn, M.M. Mench, Physical degradation of membrane electrode assemblies undergoing freeze/ thaw cycling: diffusion media effects, J. Power Sources 179 (2008) 140
146. https://doi.org/ 10.1016/j.jpowsour.2007.12.114.

direct combustion of H2 and O2 is possible and poor thermal and water management [32]. The interfacial delamination between CL and membrane can be caused by stress cycling resulting from thermal, humidification, and clamping force fluctuations under dynamic automotive operating conditions [31,33]. The CL
membrane interfacial delamination will increase the contact resistance to proton flows, which should be well controlled during MEA fabrication. Experimental studies indicated that the membranes can be mechanically damaged due to a substantial difference between the anode and cathode pressure, where nitrogen gas test can be used to detect gas crossover, and when the gas crossover is severe, the fuel cell must be stopped to avoid harmful operation [34]. Singh et al. [35] imposed an in situ AST on membranes by fluctuating the RH of reactants. In each cycle, the membrane is tested under 150% RH for 2 minutes and then under dry conditions for another 2 minutes. After a few thousand cycles, the membranes were observed using an X-ray computed tomography, and noticeable cracks were identified as a result of RH cycling. The thermal degradation of the membrane can cause material decomposition when the temperature is high enough. If the local temperature becomes higher than 280 C due to poor thermal management, the side acid groups of PFSA membranes can be decomposed from their PTFE-like molecular backbones, and without the side acid groups, the ability of proton and water transport in membranes will deteriorate. The rate of membrane decomposition can be measured via monitoring the rate of fluoride emission [36]. Many high-temperature PEM fuel cells are operated at more than 120 C,

424

Fuel Cells for Transportation

taking advantage of enhanced electrochemical kinetics, simple water management strategies, and high carbon monoxide tolerance [37]. However, a high temperature at low humidity conditions can lead to high resistance to proton flows in membranes, which will reduce the fuel cell performance significantly [38]. Commonly observed electrochemical degradations of the membrane can be membrane thinning and Pt band formation [20,39]. The membrane thinning can be caused by the material decomposition due to radical attack. The radical species, for example, peroxide (HO ) and hydro-peroxide (HOO ), can be produced with help of metal ions, for example, Cu21 and Fe31, which are often generated due to the corrosion of bipolar plates or end plates [20]. After a long-term operation, the impact of radical attack accumulates, and observable thickness reduction can be found by a scanning electron microscope, as shown in Fig. 15.3A [39,40]. Mukundan et al. [41] tested a Nafion XL membrane at open-circuit voltage (OCV) conditions with humidity cycling. In each cycle, the fuel cell is run under fully dry conditions for 30 seconds and then fully humidified conditions for 45 seconds. After about 600 hours, the rate of membrane thinning is measured to be about 2 nm per testing hour. The researchers also reported that the reduced thickness led to significant growth of hydrogen crossover. Other consequences of membrane thinning include electric short and nonuniform stress distribution, which are harmful to fuel cell performance. Under fuel cell operation, Pt band can be found in the membranes after a long-term operation, as shown in Fig. 15.3B. The formation of Pt band can resist the transport of protons, impact the stability and reliability of the membrane, and increase the ohmic overpotential. The Pt band is formed due to Pt dissolution and reduction. In the fuel cell environment, a small amount of Pt catalysts can be corroded by oxidants and dissolved in water, forming Pt ions, for example, Pt21 or Pt41 [42,43]. The Pt ions will be diffused into the membrane from cathode CLs, while a small amount of H2 gas will cross the membrane. When the Pt ions and H2 meet in the membrane, Pt will be reduced and Pt crystal will be formed [44].

15.2.2 Catalyst layer CLs, typically 5
30 μm thick, are assembled on both sides of membranes. The function of CLs is to provide places to catalyze electrochemical reactions, deliver reactants, expel produced water, and conduct electrons and ions. The electrochemical reactions in CLs can proceed naturally at a low rate, generating a small current density. To increase the current density for practical use in automotive applications, electrocatalysts are required to accelerate the reactions at the temperature of around 80 C [45,46], which is the optimal operating temperature for PEM fuel cells with relative low ionic resistance. The catalysts can be pure Pt nanoparticles [47,48], Pt
transition metal alloys [49,50], and nonprecious metals [51,52]. A good catalyst can

Fuel cell durability Chapter | 15

425

FIGURE 15.3 Scanning electron microscope images of (A) the membrane with reduced thickness due to material decomposition and (B) Pt band in membrane. (A) Adapted from X. Huang, R. Solasi, Y. Zou, M. Feshler, K. Reifsnider, D. Condit, et al., Mechanical endurance of polymer electrolyte membrane and PEM fuel cell durability, J. Polym. Sci. Part. B Polym Phys. 44 (2006) 2346
2357. https://doi.org/10.1002/polb.20863; (B) Adapted from S.V. Venkatesan, M. Dutta, E. Kjeang, Mesoscopic degradation effects of voltage cycled cathode catalyst layers in polymer electrolyte fuel cells, Electrochem. Commun. 72 (2016) 15
18. https://doi.org/10.1016/ J.ELECOM.2016.08.018.

promote a specific desirable reaction by reducing the Gibbs function of activation through enhancing the reactant gas adsorption and product water releasing. To maximize the usage of catalysts, precious metals are often

426

Fuel Cells for Transportation

deposited on less-expensive nonmetal supports, such as carbon particles [8,53], graphene [54], nanofiber [55], and composite supports [56]. Currently, the most commonly available catalyst is carbon-supported Pt (Pt/C), consisting of Pt nanoparticles and carbon particle supports, such as Vulcan, Ketjenblack, and Black Pearls, which yields good practical performance. The supporting materials, serving as the media for electron transfer, are stabilized in CLs by ionomers as the binding materials. The binding materials may wrap the catalyst and supporting materials; thus, the ionomers should be thin enough to allow the hydrogen and oxygen to reach the threephase boundaries, where electrochemical reactions occur [7,20,57]. Therefore the degradation of CLs can be caused by the malfunction of various materials, including catalyst nanoparticles, carbon particles, and ionomers. The commonly observed failure modes of catalyst nanoparticles can be dissolution [58,59], detachment [60], and sintering [5,61]. The dissolution of catalyst, for example, Pt, is caused by the oxidation of Pt which is then dissolved in water as Pt21 or Pt41 ions. Pt ions will lose their function as a catalyst, therefore reducing electrochemical surface area (ECSA) and the corresponding performance [58]. It is reported that a proper degree of dissolution of the catalyst may enhance the membrane stability and durability when a small amount of Pt crystal is formed in the membrane [62]. If a large amount of Pt is corroded, dissolved, and then deposited in the membrane, the protonic conductivity of the membrane will be significantly decreased, leading to increased ohmic overpotential [20]. The catalyst nanoparticles can also be physically detached from their carbon supports [60], as shown in Fig. 15.4. Some of the nanoparticles can be taken away from CLs by reactant flow or waterflooding, becoming catalytic inactive, while the others can merge into larger catalyst particles, causing Pt sintering. The Pt sintering is also known as catalyst agglomeration and particle growth, as shown in Fig. 15.5. Mayrhofer et al. [63] tested Pt/C catalysts in PEM fuel cells under potential cycling. In each cycle, the potential fluctuates from 0.4 to 1.4 V with a changing rate of 1 V per second. The catalyst is observed using a transmission electron microscope (TEM), and the mean diameter of the nanoparticles grew from 4.9 to 5.6 nm, as shown in Fig. 15.5. The particle growth will reduce the reactive surface area per unit mass of catalysts, which is conventionally quantified by ECSA [44,61]. The catalyst sintering and ECSA reduction involve two common pathways—Ostwald ripening and particle migration and coalescence [61], as shown in Fig. 15.4A and B. The Ostwald ripening refers to the particle growth due to the redeposit of molecules, atoms, or charged species released from the surface of small particles. Specifically, in the environment of fuel cell operation, electrochemical Ostwald ripening of Pt catalysts can be caused by the transport of Pt ions within ionomers, which is reduced by electrons conducted by carbon particles [64]. Another mechanism is called particle migration and coalescence,

Fuel cell durability Chapter | 15

427

FIGURE 15.4 Degradation mechanisms of Pt catalyst: (A) electrochemical Ostwald ripening, (B) particle migration and coalescence, (C) Pt dissolution, and (D) Pt detachment. Reprinted from P. Ren, P. Pei, Y. Li, Z. Wu, D. Chen, S. Huang, Degradation mechanisms of proton exchange membrane fuel cell under typical automotive operating conditions, Prog. Energy Combust. Sci. 80 (2020) 100859. https://doi.org/10.1016/j.pecs.2020.100859.

(A) 0 h

(B) 2 h

(C) 4 h

FIGURE 15.5 TEM images of a carbon-supported platinum (Pt/C) catalyst: (A) catalyst before electrochemical treatment, (B) after 2 h treatment, and (C) after 4 h treatment. The electrochemical treatment involved potential cycling with 1.0 V s21 between 0.4 and 1.4 V at room temperature. Adapted from K.J.J. Mayrhofer, J.C. Meier, S.J. Ashton, G.K.H. Wiberg, F. Kraus, M. Hanzlik, et al., Fuel cell catalyst degradation on the nanoscale, Electrochem. Commun. 10 (2008) 1144
1147. https://doi.org/10.1016/j.elecom.2008.05.032.

which is the particle growth due to the direct coalescence of small particles into large particles due to Brownian movements. Both mechanisms of catalyst sintering involve the decrease of small particles and the growth of large

428

Fuel Cells for Transportation

particles, which reduces the ECSA per unit mass of catalysts. The catalyst dissolution, detachment, and sintering happen throughout the PEM fuel cell lifetime, which can be accelerated under an automotive driving environment [19,65,66]. Despite running the fuel cells stationarily, the automotive fuel cells may have to rapidly and regularly change the output voltage and current to meet the load requirements for idling, acceleration, deceleration, startup, or shutdown [19]. Therefore the gas flow rates, water flow rates, reaction rates, and the number of radical species may change dynamically as the automotive fuel cell operates, and these factors can exacerbate the catalyst degradations. The degradation of carbon supports is mainly due to the corrosion of carbon materials. The corrosion of carbon materials can lead to the disconnection of various carbon particles and reduction in reaction surface area as Pt nanoparticles tend to agglomerate if the supporting particles become smaller due to corrosion. Carbon supports are usually in the diameter of a few microns, which can be unstable at elevated temperatures and easily corroded in an electrochemical environment via the following carbon corrosion reaction [67]: C 1 2H2 O 5 CO2 1 4H1 1 4e2 The rate of carbon corrosion reaction is negligible under normal operational conditions if fuels and oxidants are abundantly supplied. However, in automobiles, fuel cells may be started and stopped frequently to address the load changes under different road conditions. The frequent startup and shutdown of fuel cell engines can cause local fuel-rich and fuel starvation phenomena in CLs that lead to extremely high local potential, for example, 1.2
1.5 V [22], which can cause severe carbon corrosion. Specifically, when hydrogen starvation occurs at the anode as a result of H2 supply accessory failure, severe waterflooding at high current and high humidity environment, or ice blockage under a cold start, the phenomena of cell voltage reversal can be commonly found because of a temporally sharp potential rise at the anode [68]. Under high local potential conditions, the carbon corrosion reaction is enhanced, leading to a significant loss of carbon supports and hence catalytic activities, as shown in Fig. 15.6. Another consequence of carbon corrosion is disconnecting Pt/C particles from the electronic network and decreasing the electric conductivities in CLs [22], as shown in Fig. 15.7. The loss of carbon particles can also change the interior structure of CLs, which may weaken their mechanical strength and lower their mass transport capabilities. The degradation of binding materials, ionomer, can be caused by radical species attack or thermal decomposition [51]. Even under normal fuel cell operating conditions, a small number of radical species is always possible in CLs. For example, the metal Pt, water, and oxygen can be reacted to form

Fuel cell durability Chapter | 15

429

FIGURE 15.6 Pt agglomeration due to carbon corrosion in Pt/C after 400 h accelerated stress test. BOL, Beginning of life; EOL, end of life; LSAC, low surface area carbon. Adapted from N. Macauley, D.D. Papadias, J. Fairweather, D. Spernjak, D. Langlois, R. Ahluwalia, et al., Carbon corrosion in PEM fuel cells and the development of accelerated stress tests, J. Electrochem. Soc. 165 (2018) F3148
F3160. https://doi.org/10.1149/2.0061806jes.

FIGURE 15.7 Schematic of the degradation of carbon-supported Pt based on HSAC Vulcan XC72 and RG [69]. HSAC, High surface area carbon; RG, reinforced graphite. Reprinted from L. Castanheira, W.O. Silva, F.H.B. Lima, A. Crisci, L. Dubau, F. Maillard, Carbon corrosion in proton-exchange membrane fuel cells: effect of the carbon structure, the degradation protocol, and the gas atmosphere, ACS Catal. 5 (2015) 2184
2194. https://doi.org/10.1021/cs501973j with permission. Copyright (2015) American Chemical Society.

peroxide and hydro-peroxide, which can be further reacted with ionomer, decomposing and malfunctioning the binding materials. Similar to membranes, thermal decomposition can happen to ionomers in CLs due to local hot spots [51]. The ionomer corrosion may lead to ionomer agglomeration and deteriorate the ionomer network, inhibiting proton transport. Without ionomer connection, Pt/C particles tend to collide with each other and

430

Fuel Cells for Transportation

agglomerate, leading to catalyst sintering and surface area reduction. The loss of ionomer can also change the pore size and volume of CLs, make them mechanically weak, and block the reactant and water transport. Under cyclic conditions or due to manufacturing defects, the interfaces between Pt/ C and ionomers may be delaminated due to swelling and shrinking of materials. Although it is difficult to be experimentally observed, many numerical studies have been performed to investigate the mechanisms of the interfacial delamination under cyclic humidity, temperature, and clamping force conditions [33,70], which is typical in automotive driving cycles. It was found that humidity cycling is the dominant factor among others that caused the interfacial delamination between Pt/C and ionomer, and this is caused by the swell and shrinkage of ionomer under different humidification levels. Direct experimental evidence will be important to further understand this degradation mechanism.

15.2.3 Gas diffusion layer The GDL is a thin layer of carbon-based materials placed between CL and membrane. GDLs are designed to diffuse H2 and O2 gases from channel to CLs, expel product water from reaction sites to channel, transfer electrons, and protect CLs. The commonly used GDLs include the products fabricated by Sigracet, AvCarb, Toray, ELAT, Freudenberg, and Spectracarb. Currently, the GDL is typically composed of two thin layers. One is a macroporous layer composed of carbon fiber or cloth, and the other is a thinner microporous layer (MPL) containing carbon particles. The two thin sublayers are often treated by a hydrophobic agent, for example, PTFE, for the better capability of liquid water removal. The degradation of GDLs can be increased mass transfer resistance, reduced electric conductivity, and degraded water expel capability [71]. The mass transfer resistance can be caused by improper clamping forces at land
GDL interfaces and imperfect stack assembly process. Shen [72] investigated the effect of cyclic clamping force on the pore structure and transport properties of GDL experimentally and numerically. The results demonstrated that the clamping force can change the porosity, mass transport properties, and contact resistance, degrading the long-term operation of PEM fuel cells. The electric conductivity of GDLs can be reduced due to carbon corrosion [73], thermal expansion, and carbon fiber cracks [20]. The hydrophobicity of GDLs can be deteriorated due to the loss of the hydrophobic agents as a result of peroxide or hydro-peroxide attacks [32].

15.2.4 Bipolar plate Bipolar plates (a.k.a. gas distribution plate or flow field plate) for fuel cells are typically made of metals (e.g., aluminum, stainless steel, titanium, and

Fuel cell durability Chapter | 15

431

nickel), graphite, or graphite composites. They are used to compact the MEA, collect the electrons released from the reaction sites, separate reactants between adjacent single cells, and uniformly distribute fuels and oxidants. On one side or both sides of bipolar plates, distribution flow channels are carved for H2 and O2 supply. Further details of the bipolar plate’s functionality and flow field designs are included in Chapter 11. The degradation of bipolar plates can be corrosion and passivation layer formation [74
76]. Take stainless steel plates as an example, corrosion is a significant problem in the oxidative and reductive environment [20,75]. Wind et al.’s [75] experimental work demonstrated that the contaminant metal elements of Ni, Fe, and Cr exist in the MEA after running the fuel cell for more than 100 hours. These contaminant elements are released from the stainless steel plates, which can generate a large amount of radical species. These radical species will attack PTFE in the electrodes and ionomers in CLs and membranes, causing material decomposition. However, if the bipolar plates are made of graphite-based materials, the corrosion of bipolar plates and release of metal contaminants are insignificant compared to metal plates. The corrosion of graphite-based plates is difficult under normal fuel cell conditions, while the corrosion is only observable under frequent start
stop cycles where reactant starvation is likely to happen [65]. The passivation is another important degradation mechanism of bipolar plates. The high potential at the cathode bipolar plate can create a conducting passive layer chronically on the metal surface. The existence of a passivation layer can resist metal corrosion but will increase interfacial resistance between bipolar plates and GDLs. The passivation layer may grow throughout the fuel cell operation, and it takes thousands of hours before the contact resistance becomes noticeable [76].

15.3 Steady-state durability test If the PEM fuel cell in automotive applications is operated under desired stationary conditions to avoid fluctuations in operating conditions and component degradation, the PEM fuel cell is operated under constant current, voltage, or power. To investigate the durability of the PEM fuel cell in this driving scenario, a steady-state durability test is an appropriate method that can directly obtain the lifetime of the fuel cell and investigate the corresponding degradation modes. Therefore the testing procedures and conditions, as well as the degradation rates of various steady-state durability tests, are examined in this section.

15.3.1 Steady-state durability test protocols A steady-state durability test is to run the fuel cells under constant current, voltage, or power conditions. In practice, the most common steady-state

432

Fuel Cells for Transportation

durability test protocols are running the fuel cells under fixed current and recording the output voltage as a function of operational duration. An ideal steady-state durability test should run the fuel cells with no interruptions and maintenance, while the PEM fuel cells have to be stopped regularly. Specific maintenance service and refreshment will be imposed on fuel cells to recover reversibility [77,78]. In some reported studies, the fuel cells will be temporally stopped to conduct inspections or performance characterization through cyclic voltammetry, electrochemical impedance spectroscopy, and polarization curve measurements [77,79], although the majority of the testing time is spent on running fuel cells under stationary conditions, and the temporal interruptions is believed to have little impact on the lifetime of PEM fuel cells. Steady-state durability test is easy and simple but resource-intensive and time-consuming [11]. Under automotive driving conditions, the operational conditions of a fuel cell, including flow rate, temperature, pressure, humidity level, current, and voltage, can be very dynamic and significantly varied depending on local road conditions. Therefore the steady-state durability test may omit some important failure modes with a relatively low degradation rate. Therefore many accelerated durability experiment methods have been developed to accelerate the degradation of fuel cell components such that any potential failure modes can be identified before the fuel cell vehicle can become commercially available in the market. Knowing the degradation mechanisms, advanced control strategies can be developed to avoid harsh operation conditions and mitigate degradation rates.

15.3.2 Degradation rate by steady-state durability test The steady-state durability tests are often performed under a fixed current, where the voltage drop rates are recorded as a function of testing time. Table 15.1 summarizes the detailed steady-state durability test conditions and the voltage decay rate in various studies. It is found that at large current and moderate voltage conditions, the performance degradation rate is very small. For instance, St-Pierre et al. [80] tested a fuel cell stack with 8 cells using Nafion membranes fed by hydrogen and air for 6000 hours at the current density of 1.1 A cm22 (voltage of a single cell: B0.62 V), and the voltage drops at a slow rate of only 1 μV h21. Similar degradation rates were also reported in [81,82] when the current density varies from 0.86 to 0.50 A cm22. When the current density is decreased, the voltage is usually elevated, and the fuel cell is prone to be degraded faster, but the actual rate depends on the operational conditions. Ralph [83] tested a fuel cell stack with 4 cells with Nafion membranes at the temperature of 80 C for 5000 hours. Reformate fuel, which is composed of carbon dioxide and hydrogen, is used as anode reactant, while air is used as the cathode oxidant. The current

TABLE 15.1 Steady-state durability tests of proton-exchange membrane fuel cells. References

# of

Pt loading

Membrane

Membrane

cells

(an./ca.,

type

thickness

mg  cm22) [80]

8




T ( C)

Reactants

RH

Pressure

Stoichiometry

Voltage

Current

Test

Degradation

(an./ca.)

(an./ca.)

(an./ca., kPa)

ratio (an./ca.)

range (V)

density

time

rate

(A  cm22)

(h)

(μV  h21)

(μm) Nafion117

177.8a

75
85

H2/air

NAb b

308/308c

1.5/2.0

B0.62

1.076

6000

1

[81]

8

4/4

Dow




70

H2/O2

NA

480/480

1.5/2.0

0.85
0.75

0.861

10,000

1.3

[77]

1

0.45/0.60

GoreSelect

35

70

H2/O2

100%/ 100%

101/101c

1.2/2.0

0.65
0.5

0.80

26,300

4.2

[81]

8

4/4

Dow




70

H2/O2

NAb

480/480

1.5/2.0

0.85
0.75

0.538

1000

1.4

[80]

1




Nafion117

177.8a

80

H2/air

NAb

308/308c

1.5/2.0

0.67
0.52

0.538

2000

60

[82]

17










82

(H2 1 CO2)/ air

NA







B0.72

0.50

13,000

0.5e

[34]

3




GoreSelect




55

H2/air

0%/20%

101/101c

2.0/4.0

0.6
0.49

0.50

1000

113d,e

80

(H2 1 CO2)/ air

NA

b

308/308

1.5/2.0

B0.68

0.40

5000

4

a

[83]

4

0.25/0.55

Nafion117

177.8

[84]

1

0.26/1.46

Nafion112

50.8a

60

(H2 1 CO2)/ air

6%/4.3%

101/101c

1.2/4.0

B0.60

0.40

5100

B6d

[85]

30

0.1/0.4

Ionomer

25

80

H2/air

50%/30%

130/130

1.5/1.8

0.75
0.70

0.40

2000

18e

[86]

1




Nafion117

B178

80

H2/O2

100%/ 100%

308/308c

1.2/2.0

B0.7

0.40

1350

11

[87]

1

0.5/0.5

Nafion112

50.8a

60

(H2 1 CO2)/ air

100%/ 100%

101/101c




B0.6

0.40

4000

3

(Continued )

TABLE 15.1 (Continued) References

# of

Pt loading

Membrane

Membrane

cells

(an./ca.,

type

thickness

mg  cm22) [88]

50

0.3/0.4

T ( C)

Reactants

RH

Pressure

Stoichiometry

Voltage

Current

Test

Degradation

(an./ca.)

(an./ca.)

(an./ca., kPa)

ratio (an./ca.)

range (V)

density

time

rate

(A  cm22)

(h)

(μV  h21)

(μm) GoreSelect




55 a

H2/air

0%/5.7%b

130/101c c

1.0/5.0

B0.55

0.30

2500

20

[89]

1

0.5/0.5

Nafion112

50.8

75

H2/air

80%/80%

101/101

1.33/2.5

0.68
0.61

0.30

2700

2.5
50

[90]

20










75

(H2 1 CO2)/ air

57%/74%




1.33/2.5

0.72
0.70

0.25

5000

1.5

[91]

15

0.4/0.4

GoreSelect

B25

80

H2/air

100%/ 100%

150/150

2.0/2.0

0.77
0.72

0.20

200

25

[91]

15

0.4/0.4

GoreSelect

B25

80

H2/air

100%/ 100%

150/150

2.0/2.0

0.96
0.91

0.00

2000

20

[92]







FlemionSH50

50




H2/air

100%/ 100%







0.96
0.88

0.00

160

B200

[92]







FlemionSH50

50




H2/air

100%/0%







0.00

160

B2000

0.98
0.66 a

b

c

Note: # of cells means the total cell number in the fuel cell stack; an., denotes anode; ca., denotes cathode; RH, means relative humidity; T, denotes temperature, data taken from manufacturer, internal humidifier, assuming atmospheric pressure 5 101 kPa, ddata taken from plots, eaverage cell voltage,
no information available. Source: Adapted from J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. https://doi.org/ 10.1016/J.ENCONMAN.2019.112022.

Fuel cell durability Chapter | 15

435

density is fixed at 0.40 A cm22 (voltage of a single cell: B0.68 V), and the observed voltage decay rate is 4 μV h21. A similar voltage degradation rate is observed in Sishtla et al. [84] and Cheng et al.’s [87] studies, where reformate fuels, Nafion membrane, and a fixed current density of 0.40 A cm22 were applied. When the fuel cell current density is further reduced to 0.20
0.30 A cm22, the fuel cell degradation rates are increased to a level of tens of μV h21. For example, Scholta et al. [88], Ferreira et al. [91], and Yu et al. [89] ran the PEM fuel cell stacks with fully or partially humidified H2 and air for a few thousand hours, and the voltage degradation rate can be as high as 20
50 μV h21. When the current density is reduced to zero, the fuel cell is operated at OCV conditions, which is believed to be very harmful to the catalysts. The durability testing results were based on humidified H2 and air at a fixed current density of 0 A cm22, and the fuel cell degradation was observed to be extremely fast at the rate of 20
2000 μV h21. This indicates that when the automobiles are operated under idling conditions, the fuel cell should avoid high-voltage or OCV conditions even though minimum power is needed to be extracted from fuel cells. It is observed that many steady-state durability tests were conducted for a few thousand hours, as shown in Table 15.1, while Cleghorn et al. [77] reported that the fuel cell fueled by pure H2 and O2 gases in their studies had been run for 26,300 hours, with a mean decay rate of 4.2 μV h21. Verhage et al. [78] ran a cell stack for as long as 30,000 hours before the long-term performance degraded by 10%. They also reported that reversible voltage losses have been removed during the fuel cell testing through routine maintenance and recovery processes. It is found that the lifetime is rarely reported in durability tests, and this is likely because the definition of lifetime is still under debate. In many studies, the EOL of a PEM fuel cell is defined as the operating point when the voltage, current, or power is reduced by 10%. However, the performance indicator selection can affect the lifetime of PEM fuel cells significantly. Some other studies define the lifetime of PEM fuel cells as the maximum service time as long as no major component failure is found. To compare the durability of different fuel cells, the voltage drop rate in μV  h21 under specified operating current density is usually stated in the literature, which allows a direct comparison of different durability studies.

15.4 In situ accelerated stress test If the PEM fuel cell in automotive applications is the sole power source and dynamic power output is used to meet the transient energy demand under various driving conditions. The dynamic operation of a PEM fuel cell can be achieved by controlling reactant pressure, temperature, and humidity levels,

436

Fuel Cells for Transportation

while it can also be affected by environmental conditions if the air is used as the cathode oxidant. Therefore the durability behaviors of automotive fuel cells may vary at different locations with significantly different climate and weather conditions, and thus, the testing protocols reported in different regions, such as Canada, China, the European Union, Japan, South Korea, and the United States have been examined in this section [20,93].

15.4.1 In situ accelerated stress test protocols An in situ AST often involves cycling current, voltage, or power during fuel cell testing. Factors, including cyclic profile, duration of each cycle, upper and lower limits of each cycle, and the total number of cycles, will affect the durability and degradation modes significantly. Many in situ AST methods were established to accelerate the degradation of fuel cell components under dynamic operational conditions, which can be affected by the actual automotive driving conditions, for example, idling, acceleration, deceleration, startup, and shutdown. Uchimura and Kocha [42] investigated the effect of square and triangle cyclic voltage profiles on the degradation behavior of fuel cells. The upper and lower voltage limits are constrained between 0.95 and 0.60 V, and the duration of each cycle varies from 5 to 10 seconds, as shown in Fig. 15.8. Six square and symmetric triangle voltage profiles with different frequencies were applied on a 25 cm2 MEA with Pt loadings of 0.35 mg cm22 for both anode and cathode electrodes. It was found that the mass activity was reduced slightly faster when the cycling frequency was increased from 0.2 to 0.1 seconds21. Specifically, the hourly mass activity loss under square or triangle voltage cyclic profiles at the frequency of 0.2 seconds21 is almost twice higher than that of 0.1 second21. Under a given cycling frequency, the square-wave cycling caused almost 2.5 times faster performance decay

FIGURE 15.8 Accelerated stress test protocols proposed by Uchimura and Kocha [42] with potential cycling between 0.60 and 0.95 V of various cycle profiles and durations: (A) square 10 s; (B) square 5 s; (C) triangle 10 s; (D) triangle 5 s; (E) asymmetrical triangle, slow cathodic 10 s; (F) asymmetric triangle, slow anodic, 10 s. Adapted from M Uchimura, S.S. Kocha, The impact of cycle profile on PEMFC durability, ECS Trans. 11 (2007) 1215
1226. https://doi.org/ 10.1149/1.2781035.

Fuel cell durability Chapter | 15

437

0.95

Voltage [V]

Voltage [V]

compared to the triangle-wave profile. The authors also investigated two asymmetric voltage profiles by sweeping voltage from 0.95 to 0.6 V (named as slow cathodic scan in Fig. 15.8E) and from 0.6 to 0.95 V (named as slow anodic scan in Fig. 15.8F) in 10 seconds. It was found that the slow cathodic scan caused a faster deterioration in mass activity than the anodic scan, and the difference is attributed to the chemical/electrochemical Pt dissolution. Many in situ AST protocols were developed based on square-wave cycling in voltage. Kneer et al. [94] tested a 45-cm2 single PEM fuel cell (anode Pt loadings of 0.1 mg cm22 and cathode Pt loadings of 0.25 mg cm22) under square-wave voltage cycling with different voltage limits and different temperature. One test is conducted at 70 C with potential cycles between 0.4 and 0.95 V in 4 seconds (see Fig. 15.9A), and the other test is done at 90 C with potential cycles between 0.6 and 0.95 V in 4 seconds (see Fig. 15.9B). The experimental results suggested that the voltage decay rate was about 110 μV h21 in the first test, while the voltage decay rate can be as high as 525 μV h21 in the second test. The voltage was read

0.4 0

2

4

6

0.95

0.6

0

8

2

4

6

t [s]

t [s]

(A)

(B)

8

Voltage [V]

UPL: upper potential limit, 1.0-1.4 V

UPL

0.6 0

30

60

90 120 150 180 t [s] (C)

FIGURE 15.9 Accelerated stress test protocols based on potential cycling. (A) 0.4
0.95 V at 70 C by Kneer et al. [94], (B) 0.6
0.95 V at 90 C by Kneer et al. [94], (C) 0.6 to 1.0
1.4 V by Hitchcock et al. [95], and Venkatesan et al. [40]. Adapted from J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. https://doi.org/10.1016/J. ENCONMAN.2019.112022.

438

Fuel Cells for Transportation

Relative Humidity [%]

under the reference current density of 2.5 A cm22. The size and location of catalysts were investigated by TEM imaging, and it was observed that a large portion of Pt catalysts was redistributed within the MEA. Near the membrane
CL interface, Pt was diminished, while in the membrane, Pt bands were observed. The results suggested that about 8% of the Pt catalysts were removed from the electrodes after durability tests. Hitchcock et al. [95] also applied potential cycling on a 45-cm2 single PEM fuel cell (anode Pt loading of 0.1 mg cm22 and cathode Pt loading of 0.4 mg cm22). In each cycle, the fuel cell was run at a low voltage of 0.6 V for half a minute and then at a high potential of 1.0
1.4 V for 1 minute, as shown in Fig. 15.9C. The CL was characterized by a scanning transmission X-ray microscope, and the CL degradation was observed in terms of Pt sintering, relocation, Pt redeposition in the membrane, carbon oxidation, and cathode CL thinning. Venkatesan et al. [40] tested a similar PEM fuel cell using similar squarewave cyclic protocols. After running the fuel cell for 117.5 hours (equivalent to 4700 cycles), the degradation rate was observed to be as high as 2400 μV h21 (the voltage was evaluated at current density of 0.75 A cm22). The changeable conditions in cyclic testing are not limited to voltage, while humidity, power, current, and other variables can also be controlled. For instance, Panha et al. [96] developed a humidity cycling durability test under idling automotive driving conditions. In the authors’ experiment, the PEM fuel cell was first operated for 120 hours to test the leakage, crossover, and break-in performance. Subsequently, the fuel cell was subject to humidity cycling at a very small current density of 0.01 A cm22. In each cycle, the inlet reactants were fully dried for 10 minutes and then fully humidified for 40 minutes, as shown in Fig. 15.10. A 700-hour humidity cyclic testing on a fuel cell with 0.4 mg cm22 Pt loading in both electrodes suggested that the voltage dropped at a rate of 240 μV h21. To understand the humidity cycling

100

0

0 10 20 30 40 50 60 70 80 90 100

t [min] FIGURE 15.10 Accelerated stress test protocols based on hydrogen—air humidity cycling in Panha et al.’s [96] work. Adapted from K. Panha, M. Fowler, X.-Z.Z. Yuan, H. Wang, Accelerated durability testing via reactants relative humidity cycling on PEM fuel cells, Appl. Energy 93 (2012) 90
97. https://doi.org/10.1016/j.apenergy.2011.05.011.

Fuel cell durability Chapter | 15

439

on the degradation of CLs, Rong et al. [33,97] developed a model to simulate the thermal and humidity cycling during a startup
shutdown cycle, and their results suggested that the swelling and shrinking of binding materials can cause structural changes in CLs, including Pt/C
ionomer delamination and cracks, which is harmful to catalyst activity. Bae et al. [98] controlled the upper limit of output power during on
off cycling. During the test, constant power of 0.183 W cm22 lasted for 2
10 hours, followed by an off period of 2 hours, as shown in Fig. 15.11. The testing results based on a 25cm2 single PEM fuel cell (anode Pt loading of 0.45 mg cm22 and cathode Pt loading of 0.4 mg cm22) at 60 C suggested that shorter hold-on time yields faster degradation in performance (e.g., performance drop by 15% during 750 hours for the 2h-on/2h-off tests). In the case of 10h-on/2h-off cycles, the performance drops only 14% in a long testing period of 2600 hours. Many in situ AST procedures have been developed based on automotive driving cycles based on complicated startup
shutdown and acceleration
deceleration conditions. Among these, the New European Driving Cycle (NEDC) established by the United Nations Economic Commission for Europe has been widely used as a reference to develop in situ AST protocols

Power / Wcm-2 0.183

Power / Wcm-2

10hr 0.183

2hr

Time / h

(A) Mode 1

5hr

2hr

Time / h

(B) Mode 2

Power / Wcm-2 0.183

2hr

2hr

Time / h

(C) Mode 3 FIGURE 15.11 Accelerated stress test protocols proposed by Bae et al. [98] based on on
off cycles: (A) Mode 1 (constant power of 0.183 W  cm22 for 10 h and off for 2 h), (B) Mode 2 (constant power of 0.183 W  cm22 for 5 h and off for 2 h), and (C) Mode 3 (constant power of 0.183 W  cm22 for 2 h and off for 2 h). Adapted from S.J. Bae, S.J. Kim, J.I. Park, C.W. Park, J. H. Lee, I. Song, et al., Lifetime prediction of a polymer electrolyte membrane fuel cell via an accelerated startup-shutdown cycle test, Int. J. Hydrog. Energy 37 (2012) 9775
9781. https:// doi.org/10.1016/j.ijhydene.2012.03.104.

Fuel Cells for Transportation

(A) 176 A

180 Current [A] RH cathode [%] Temperature [oC]

160

Parameter

140

128 A

120 100

92 A

92 A 80 oC

80 59 A

64 A

60

50% RH

40

28 A

20

30% RH

22 A

0 0

100 200 300 400 500 600 700 800 900 10001100 1200

t [s]

(B) 100

Ratio of Rated Power [%]

440

90 80 70 60 50 40 30 20 10 0 0

100 200 300 400 500 600 700 800 900 10001100 1200

t [s]

(C) (Continued)

Fuel cell durability Chapter | 15

441

L

for PEM fuel cells to be installed in automotive applications. A single NEDC cycle is composed of four repetitive city driving cycles in the first 13 minutes and an extra high-velocity urban driving cycle in the next 7 minutes, as shown in Fig. 15.12A. Many in situ AST methods for fuel cell testing were established based on NEDC cycles. For instance, Nandjou et al. [85] established an NEDC-like AST protocol to test a 220-cm2 30-cell fuel cell stack at 80 C by controlling the operating current. In each AST cycle, a low current cycle was repeated four times in the first 13 minutes to simulate city driving conditions, while a high current cycle was used in the latter 7 minutes to meet the energy demands in highway cycles, as shown in Fig. 15.12B. In the first 13 minutes, the cathode relative humidity was controlled at 50%, while in the last 7 minutes, the cathode relative humidity was set at 30%. The 1500-hour durability test results suggested a voltage decay rate of 27 μV h21. Bloom et al. [99] developed a similar NEDC-like AST protocol by controlling the rated power, as shown in Fig. 15.12C. The cyclic power profile was related to the velocity of automobiles in NEDC cycles and controlled in square waves [99,100]. Lin et al. [101] applied a dynamic automotive driving cycle similar to NEDC to a 50-cm2 single PEM fuel cell, which is subject to cold start, idling, full power, and overload operation, as shown in Fig. 15.13. It was found that at the first 280 hours of ASTs, the voltage dropped at a rate of 276 μV h21 (voltage evaluated at 0.7 A  cm22), while in the following 70 hours, the voltage degradation rate significantly increased to 2300 μV h21. Dynamic stress test (DST) established by the US DOE is another commonly used AST cycle to test PEM fuel cells. This method is established based on the power requirements of an automobile to be run under a combined high-speed and hill-climbing scenarios in the United States [102], as shown in Fig. 15.14. Each cycle lasts for 360 seconds, and a few step changes bouncing between OCV to 0.6 V were imposed on the fuel cells. Garland et al. [103] also reported four DOE protocols aiming at the durability of catalyst, carbon support, membrane, and the whole MEA. The AST

FIGURE 15.12 Various driving cycles: (A) Reference European driving cycles (NEDC), (B) simplified NEDC/relative humidity by Nandjou et al. [85], and (C) adapted NEDC by Bloom et al. [99]. Parts (A) and (B) are adapted from F. Nandjou, J.-P. Poirot-Crouvezier, M. Chandesris, J.-F. Blachot, C. Bonnaud, Y. Bultel, Impact of heat and water management on proton exchange membrane fuel cells degradation in automotive application, J. Power Sources 326 (2016) 182
192. https://doi.org/10.1016/j.jpowsour.2016.07.004 and part (C) is adapted from I. Bloom, L.K. Walker, J.K. Basco, T. Malkow, A. Saturnio, G. De Marco, et al., A comparison of Fuel Cell Testing protocols—a case study: protocols used by the U.S. Department of Energy, European Union, International Electrotechnical Commission/Fuel Cell Testing and Standardization Network, and Fuel Cell Technical Team, J. Power Sources 243 (2013) 451
457. https://doi.org/10.1016/j.jpowsour.2013.06.026.

442

Fuel Cells for Transportation

FIGURE 15.13 Accelerated stress test protocols developed by Lin et al. for a single PEM fuel cell with an active area of 50 cm2. Adapted from R. Lin, B. Li, Y.P. Hou, J.M. Ma, Investigation of dynamic driving cycle effect on performance degradation and micro-structure change of PEM fuel cell, Int. J. Hydrog. Energy 34 (2009) 2369
2376. https://doi.org/10.1016/j. ijhydene.2008.10.054.

aiming to accelerate the degradation of the catalyst was developed using H2 as the fuel and N2 as the oxidant. In each cycle, the voltage was fixed at 0.7 V for half a minute and then 0.9 V for another 30 seconds. The test will be stopped when the catalyst activity is reduced by 60% or when the voltage is decreased by 0.03 V at 0.8 A cm22 [103,104]. The voltage is controlled under 0.9 V because a higher voltage can accelerate the corrosion of carbon support significantly, and it would be difficult to distinguish between the catalyst and carbon support degradation. The AST aiming to test carbon support is performed under high voltage fed by H2 and N2 to prevent chemical corrosion by air or oxygen. In this test, the voltage was fixed at 1.2 V (close to OCV) for a long period of 200 hours, and the test can be ceased early if the catalyst activity is reduced by over 60% or the voltage dropped 0.03 V at 1.5 A cm22 or rated power. The AST aiming to test membranes is a humidity cycling applied to PEM fuel cells fed by air in both anodic and cathodic channels. In each cycle, the relative humidity is controlled at 0% for 120 seconds and then over 100% for another 120 seconds. The test is targeted for 20,000 cycles but will stop if the gas crossover rate becomes as high as 10 sccm. The AST aiming to test the whole MEA is conducted at opencircuit conditions with a temperature of 90 C and relative humidity of 30% for both H2 and air. The test will be stopped if the voltage is reduced by 20% or if the fuel crossover rate is equivalent to 0.02 A  cm22.

Fuel cell durability Chapter | 15

443

FIGURE 15.14 Accelerated stress test protocols proposed by US DOE—DST for fuel cell testing. DOE, Department of Energy; DST, dynamic stress test. Adapted from I. Bloom, L. Walker, J. Basco, T. Malkow, G. De Marco, G. Tsotridis, et al., A comparison of fuel cell test protocols, ECS Trans. 30 (2011) 227
235. https://doi.org/10.1149/1.3562478.

FIGURE 15.15 Accelerated stress test protocols developed by Lu et al. based on Chinese typical bus driving cycle for hybrid PEM fuel cell buses with 80 cells in series and active area of 280 cm2. Adapted from L. Lu, M. Ouyang, H. Huang, P. Pei, F. Yang, A semi-empirical voltage degradation model for a low-pressure proton exchange membrane fuel cell stack under bus city driving cycles, J. Power Sources 164 (2007) 306
314. https://doi.org/10.1016/j. jpowsour.2006.10.061.

Lu et al. [105] established an AST method based on a typical hybrid fuel cell bus driving cycle in China, as shown in Fig. 15.15. In each cycle, a combination of idling, acceleration, deceleration, and overloading processes was performed. The testing results on a 280-cm2 80-cell PEM fuel cell for 640 hours suggested a voltage drop of 54
59 V (or mean voltage drop of 0.675
0.738 V) at the reference current density of 0.36 A  cm22 [105].

444

Fuel Cells for Transportation

15.4.2 Degradation rate by in situ accelerated stress test In automobiles, the PEM fuel cells may have to frequently start, stop, change flow rate, current, and voltage to meet the power demands. To simulate these scenarios and understand the degradation behaviors under automotive driving cycles, various in situ ASTs are conducted for different fuel cells. Table 15.2 summarizes the experimental conditions, test protocols, test duration, and voltage degradation rate of PEM fuel cells tested by different in situ AST protocols. Nandjou et al. [85] applied an NEDC-like AST on a 30-cell stack for about 1500 hours, and the corresponding voltage degradation rate was reported to be only 27 μV h21. Lu et al. [105] applied dynamic current cycling on an 80-cell fuel cell stack with Nafion112 membranes for 640 hours, and the voltage drop rate was found to be 72.5 μV h21 (voltage was measured at 0.36 A cm22). Wahdame et al.’s [34] potential and current cycling on a three-cell stack with GoreSelect membrane were run for 700 hours, and the potential is decreased at a rate of 100 μV h21 (voltage was measured at 0.5 A cm22). Panha et al. [96] ran a 25-cell PEM fuel cell stack with GoreSelect membranes for about 700 hours at idling conditions with an extremely low current density of 0.01 A cm22, and the voltage degradation rate was reported at 240 μV h21 under a humidity cycling protocol. Lin et al. [101] applied 280-hour current cycling on a single PEM fuel cell with Nafion112 membranes. The voltage decay rate was found to be 276 μV h21 (voltage was taken at 0.7 A cm22). When the test was run for another 90 hours under the same conditions, the degradation is significantly faster at the rate of 2300 μV h21. Kneer et al. [94] applied a potential cycling protocol between 0.6 and 0.95 V at the temperature of 90 C on a single PEM fuel cell, and the voltage decay rate was measured to be about 525 μV h21 (voltage was taken at 2.5 A cm22). Liu and Case [106] tested a single PEM fuel cell with Nafion112 membranes for 1000 hours following a current cycling procedure. The voltage decay rate was found as high as 540 μV h21 (voltage taken at 1.06 A cm22). Venkatesan et al. [40] applied a potential cycling protocol on a single PEM fuel cell for 117.5 hours, and the voltage drops much faster at the rate of 2400 μV h21 (voltage taken at 0.75 A cm22), while Selvaganesh et al. [107] reported a 4700 μV h21 degradation rate at 0.95 A cm22. It is seen that most AST studies tested the fuel cells for less than 1000 hours, while most steady-state durability tests are longer than 1000 hours (see Table 15.1). The maximum reported potential cycling AST duration in Table 15.2 is reported by Noto et al. [108] for about 7000 hours. The degradation rate in voltage is above 100 μV h21 in most AST studies, while most reported degradation rates in steady-state durability tests are among 1
20 μV h21. The degradation rate in AST measurements can be affected by many factors, including cyclic profile, operation conditions, and the materials applied. The total testing duration can also impact the reported

TABLE 15.2 Accelerated stress tests of proton-exchange membrane fuel cells. References

# of

Membrane

Membrane

Pt loading

T

Reactants

RH

Pressure

Stoichiometry

Cyclic

Cyclic

Test time

Degradation

Reference

cells

type

thickness

(an./ca.,

( C)

(an./ca.)

(an./

(an./ca.,

ratio or flow

variables

span

(h)

rate

current

(μm)

mg  cm22)

ca.)

kPa)

rate (an./ca.)

(μV  h21)

density

(min)

(A  cm22) [85]

30

Ionomer

25

0.1/0.4

80

H2/air

50%/ 50%30%

130/130

1.5/1.8

Current cycling Humidity cycling

20

B1500

27




[105]

80

Nafion112

50.8a




60c

H2/airc

60%
100%/ 60%
100%c

101/ 101b,c

1.5/2.5c

Dynamic driving cycles Current cycling

B59.6

640

72.5f

0.36

[34]

3

GoreSelect







55

H2/air

0%/ 100%

101/101b

2/4

Dynamic potential/ current cycling

B9.2

700

B100

0.50

[94]

1




15

0.10/0.25

70

H2/air

100%/ 100%

270/250

4.5/11.3 slpm

Potential cycling

0.067

460

B110d

2.5

[101]

1

Nafion112

50.8a

0.4/0.4

85

H2/air

100%/ 100%

101/101b

1.4/3.3

Dynamic load/ current cycling

20

280

104 188 270 276

0 0.2 0.5 0.7

[96]

1

GoreSelect

25

0.4/0.4

70

H2/air

100%/ 100%

101/101b

0.113/0.358 slpm

Humidity cycling

50

B700

240

0.01

[94]

1




15

0.10/0.25

90

H2/air

100%/ 100%

270/250

4.5/11.3 slpm

Potential cycling

0.067

400

B525d

2.5

(Continued )

TABLE 15.2 (Continued) References

# of

Membrane

Membrane

Pt loading

T

Reactants

RH

Pressure

Stoichiometry

Cyclic

Cyclic

Test time

Degradation

Reference

cells

type

thickness

(an./ca.,

( C)

(an./ca.)

(an./

(an./ca.,

ratio or flow

variables

span

(h)

rate

current

(μm)

mg  cm22)

ca.)

kPa)

rate (an./ca.)

(μV  h21)

density

(min)

(A  cm22) [106]

1

Nafion112

50.8a

0.5/0.5

80

H2/air

100%/ 100%

239/239b

200/500 sccm

Current cycling Constant flowrate

6

1000

540d

1.06

[40]

1










75

H2/air

100%/ 100%







Potential cycling

1.5

117.5

B2400d

0.75

[107]

1

Nafion1135

88.9

0.5/0.5

60

H2/O2

100%/ 100%

101/101b

2/[email protected] A  cm22

Potential cycling

0.67

78

B4700d B1026d

B0.95 B2.0

[108]




























Potential cycling

11.7

100
7000







Note: # of cells means the total cell number in the fuel cell stack; an., denotes anode; ca., denotes cathode; T, denotes temperature; RH, means relative humidity, adata taken from manufacturer, bassuming atmospheric pressure 5 101 kPa, cdata taken at rated conditions, ddata taken from plots, e average cell voltage, Ref. current density denotes the reference current density under which the degradation rate is calculated,
no information available. Source: Adapted from J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. https://doi.org/10.1016/J.ENCONMAN.2019.112022.

Fuel cell durability Chapter | 15

447

degradation rate. Lin et al.’s [101] AST results suggested a lower decay rate of 276 μV h21 (voltage was taken at 0.7 A cm22) in the first 280 hours, while in the last 90 hours under the same conditions, the degradation rate becomes 8.3 times higher in comparison with the first 280 hours. This means the early life of fuel cell operation may have a lower degradation rate, while as the fuel cell operation continues, the degradation becomes faster. However, more experimental work should be explored to understand the relationship between degradation rate and testing duration. The reported AST results also suggested that directly comparing the voltage decay rate in different studies is difficult because the voltage decay rate is impacted by different conditions, such as current, voltage, temperature, humidity, flow rate, and pressure [109]. Even in the same AST, the voltage measured at different current densities can also affect the reported degradation rate significantly.

15.5 Ex situ accelerated stress test Through in situ ASTs, the durability and degradation rate can be identified, and different failure modes can be observed using proper techniques. However, it is challenging to identify the causes of different failure modes or the degradation mechanisms of individual cell components, such as membranes, CLs, GDLs, and bipolar plates. Therefore many ex situ AST experimental methods have been developed to fundamentally understand the degradation mechanisms, and the stressors can be humidity, thermal, hygrothermal, wet
dry, freeze
thaw, clamping force, and vibration.

15.5.1 Humidity cycling The relative humidity of reactants at channel inlets under different operating conditions is controlled to make the membrane well hydrated for high ionic conductivity. However, in the actual fuel cell operation, the humidity can be varied due to local weather conditions, system design, and control strategies. The humidification of reactants can affect the voltage decay rate, and many ex situ AST methods have been developed to investigate its effect on individual component failure. Sethuraman et al. [110] developed a humidity AST to test the degradation of membranes, which is compacted by two GDLs. In each cycle, the anode and cathode were supplied with N2 and were fully humidified for 1 hour and then fully dried for another hour at the temperature of 100 C and pressure of 150 kPa. A pressure difference of 20.68 kPa was applied across the samples to measure the gas crossover rate. The experimental results after 100 hours of testing suggested that Nafion 112 membranes demonstrated superior mechanical stability with smaller gas crossover than bi-phenyl sulfone-H (BPSH) membrane. At the end of testing, the crossover rate of N2 was found to be over 680 cc min21 through BPSH membrane, while it is negligibly

448

Fuel Cells for Transportation

small through Nafion 112 membrane. The difference is attributed to the capabilities of different materials to resist the swelling and shrinking caused by humidification fluctuations. Miyatake et al. [111] applied an ex situ humidity cycling test on sulfonated polyimide (SPI-8) membranes. The test was conducted between fully dry and fully humidified conditions, and the cycle is repeated every 4 minutes at the temperature of 80 C for 667 hours. The authors reported no observable mechanical degradation, while a small amount of hydrogen permeation is detected.

15.5.2 Temperature cycling The temperature of the PEM fuel cell determines not only the instant performance but also the degradation of fuel cell components. If the PEM fuel cell temperature is too low, the catalyst activity remains low. If the temperature is too high, the membrane can be overdry, leading to a significant ohmic resistance increase. As discussed earlier, an extremely high operating temperature can also cause thermal, chemical, and electrochemical degradation of membranes. Therefore 60 C
90 C is often regarded as the optimal operating temperature for PEM fuel cells. However, during the fuel cell operation, the temperature can fluctuate under startup
shutdown and load-changing conditions. Many experimental techniques have been developed to investigate the impact of thermal cycles on the degradation of fuel cells. Kim et al. [31] applied a thermal cycling protocol on MEA with and without GDL supports. The temperature was cycled between 5 C and 70 C, and the testing results suggested that GDL can protect the CL and membrane from cracks, especially in the channel/land interfacial regions. Bi and Fuller [112] applied square-wave voltage cycling on PEM fuel cells under different temperatures from 40 C to 80 C. The experimental results suggested that when the temperature was increased, more Pt particles moved to the membranes, while the remaining catalysts become larger, which is due to the carbon corrosion at elevated temperatures.

15.5.3 Hygrothermal cycling Humidity and temperature are often coupled in actual fuel cell operation, and it is challenging to separate these two factors. Many experimental methods have been developed to study the effect of combined humidity and temperature cycles (or hygrothermal cycling) on the degradation of fuel cell components [31,113]. Silberstein and Boyce [113] applied a hygrothermal cycling procedure on a Nafion membrane/GDL bimaterial, as shown in Fig. 15.16. Under hydrated conditions, the bimaterial strip was significantly bent toward the GDL side (see Fig. 15.16B), while when the environment is fully dried, the bimaterial strip bounces back with a small curvature due to the plastic deformation

Fuel cell durability Chapter | 15

449

FIGURE 15.16 Side view images of Nafion211 membrane/GDL bimaterial swelling test (membrane thickness of 27 μm and GDL thickness of 180 μm): (A) initial configuration, (B) hydrated, and (C) dried. Membrane is on the right-hand side of the strip. GDL, Gas diffusion layer. Adapted from M.N. Silberstein, M.C. Boyce, Hygro-thermal mechanical behavior of Nafion during constrained swelling, J. Power Sources 196 (2011) 3452
3460. https://doi.org/ 10.1016/j.jpowsour.2010.11.116.

(see Fig. 15.16C). The authors also developed a finite element model, and the modeling results agree with the experiment well (see Fig. 15.17). The modeling results also suggested that the stress cycling in fuel cell components is affected by hygrothermal ramping and holding time. Banan et al. [114] developed a model to investigate the effect of hygrothermal cycling on crack propagation and interfacial delamination. The modeling results suggested amplitude of humidity cycling is the decisive factor that determines the rate of membrane
CL interfacial delamination propagation. The swelling of humified membranes will cause large in-plane stresses, thus accelerating the delamination process.

15.5.4 Liquid water wet
dry cycling Liquid water management is always a crucial problem in PEM fuel cells. A small amount of liquid water existing in the electrode may help hydrate the membrane to maintain a low ionic resistance, while too much liquid water in the electrode will block reactant transport pathways and occupy active reaction sites. Many experimental methods were developed to investigate the impact of liquid water wet
dry cycling on the degradation of fuel cell components. Lin et al. [115] established a GDL washing experiment to study the effect of liquid waterflooding on the structure of GDLs. During the experiment,

450

Fuel Cells for Transportation

FIGURE 15.17 Side view images of the finite element simulation of bimaterial swelling test with Nafion211 membrane and GDL: (A) initial configuration, (B) hydrated, and (C) dried. Membrane is indicated by dark gray, and GDL is indicated by light gray. GDL, Gas diffusion layer. Adapted from M.N. Silberstein, M.C. Boyce, Hygro-thermal mechanical behavior of Nafion during constrained swelling, J. Power Sources 196 (2011) 3452
3460. https://doi.org/ 10.1016/j.jpowsour.2010.11.116.

GDL samples were placed between two connected and sealed tubes, as shown in Fig. 15.18. The liquid water flow rate was controlled at 500 cm3 min21. It was observed that MPL contributed to distributing liquid water more uniformly within the MEA. The performance testing results suggested that with an MPL, the current density at a fixed voltage of 0.6 V is enhanced from 0.4 to 0.65 A cm22. If the washing experiment lasts long, the MPL may crack and penetrate GDL fibers by water, which may affect the ohmic resistance significantly. Zhao et al. [53] established two liquid water management experiments to mimic the liquid water formation and evaporation cycling and liquid water flow-through-and-dry cycling. The first testing method (see Fig. 15.19) followed a procedure of soaking CLs in deionized (DI) water under vacuum conditions for 1 hour and heating the CLs at 75 C for another hour until dried. The wet
dry cycling was repeated 15 times in 30 hours of durability tests. The second testing method (see Fig. 15.20) followed a procedure of forcing liquid water to penetrate the CLs on a porous hot metal plate using a vacuum pump from the bottom at 75 C for half an hour and heating the sample for another 30 minutes until dried. The second method was repeated 60

Fuel cell durability Chapter | 15

451

FIGURE 15.18 Schematic of GDL washing experiment with pure water. GDL, Gas diffusion layer. Adapted from J.H. Lin, W.H. Chen, S.H. Su, Y.J. Su, T.H. Ko, Washing experiment of the gas diffusion layer in a proton-exchange membrane fuel cell, Energy Fuels 22 (2008) 2533
2538. https://doi.org/10.1021/ef800116c. Copyright (2008) American Chemical Society.

FIGURE 15.19 Schematic of water intrusion
evaporation test for catalyst layers. Adapted from J. Zhao, S. Shahgaldi, X. Li, Z. Liu (Simon), Experimental observations of microstructure changes in the catalyst layers of proton exchange membrane fuel cells under wet-dry cycles, J. Electrochem. Soc. 165 (2018) F3337
F3345. https://doi.org/10.1149/2.0391806jes.

times with a total time of 30 hours. The experimental results suggested that the liquid intrusion
evaporation can cause catalyst sintering, pinhole, and cracks in CL. In comparison, the Pt/C particle did not obviously grow during the flow-through-and-dry cycling. This is because if the water is expelled efficiently from the electrode to channels (similar to the second experiment), liquid water is more likely to flow through large pores in CLs quickly, leaving the small pores unaffected. However, if the liquid water is accumulated in the CLs (similar to the first experiment), liquid water can penetrate small pores and accelerate the Pt/C degradation. The experimental results

452

Fuel Cells for Transportation

FIGURE 15.20 Schematic of water flow-through-dehydration test for catalyst layers. Adapted from J. Zhao, S. Shahgaldi, X. Li, Z. Liu (Simon), Experimental observations of microstructure changes in the catalyst layers of proton exchange membrane fuel cells under wet-dry cycles, J. Electrochem. Soc. 165 (2018) F3337
F3345. https://doi.org/10.1149/2.0391806jes.

suggested that managing water efficiently can not only enhance the performance of fuel cells but also inhibit Pt/C agglomeration and other irreversible damages.

15.5.5 Freeze
thaw cycling Operating PEM fuel cells at the temperature of subzero is a significant challenge, especially for automotive applications. Under subzero temperature, ice formed in electrodes and membranes can either block the pores for reactant transport or damage electrode structure. Many experimental methods were developed to investigate the impact of freeze
thaw cycles on the degradation of fuel cell components. McDonald et al. [116] applied 385 freeze and thaw cycles on the catalyst-coated membranes with no GDL supports between 180 C and 240 C. The experimental results suggested no noticeable physical damages in structure, but molecule-level damages were identified. Yan et al. [117] started their PEM fuel cell at 215 C, and holes in membranes and membrane
CL delamination were observed. Ozden et al. [54] applied 60 freeze
thaw cycles on fuel cell electrodes with electrodes soaked in liquid water, and the environmental temperature fluctuated between 240 C and 30 C. The experimental results suggested that large pieces of carbon particles and binding material mixtures collapsed from the CL surface.

15.5.6 Clamping force Clamping force when assembling the different components is one of the major factors that affect the performance of overall fuel cells. The clamping force can be nonuniformly distributed on GDLs, CLs, and membranes due to the torque applied by the assembly bolts. The optimal clamping force is

Fuel cell durability Chapter | 15

453

1.0
2.0 MPa [118] to minimize reactant leakage and reduce the contact resistance between adjacent components. A large clamping force can compact the electrode, increasing mass transfer resistance. The nonuniformly distributed clamping force can be caused by improper assembly process and internal stress cycling as a result of the cyclic hygrothermal, wet
dry, and freeze
thaw conditions, and this can lead to structural damages to MEA, such as delamination, holes, and cracks [119]. Most clamping-related studies emphasized instant performance, and its effect on component degradation has not been sufficiently studied. More efforts should be further dedicated to the impact of clamping force on the durability of fuel cells.

15.5.7 Vibration The automotive fuel cells may be operated in an environment of vibrations or shock attacks. Diloyan et al. [120] ran a fuel cell with and without mechanical vibrations for more than 300 hours. Under no vibration conditions, the Pt nanoparticles grew from 2
2.5 nm to about 6 nm; While with vibration conditions of 1 g and 20 Hz, the Pt nanoparticles grew to 5.47 nm. This means the vibration treatment increased the Pt particles by almost 10%. Rajalakshmi et al. [121] examined a 30-cell stack under vibrations and shock attacks. The experiment results demonstrated similar performance with or without vibration treatment, and this means the 30-cell stack is mechanically stable. However, low compression stress was released to lose the assembly bolts, and therefore the authors recommended using a padding or spring suspension to alleviate the vibrations and shock attacks. More experimental efforts are required to fully understand the impact of vibrations on fuel cell degradation.

15.6 Summary The durability of PEM fuel cells under automotive driving cycles is discussed in this chapter with a focus on fundamentals and experiments. The fundamental degradation mechanisms under automotive driving cycles in different components, such as polymer electrolyte membrane, catalyst layer, GDL, and bipolar plates, have been comprehensively highlighted. The experimental methods used to examine the durability and degradation mechanisms of PEM fuel cells have been examined in this chapter, including steady-state durability test, in situ AST, and ex situ AST. It was found that the general voltage decay rate can be as low as 1 μV h21 during steady-state durability tests, while that of in situ AST test can be much higher than 100 μV h21 depending on the actual cyclic conditions. It should be mentioned that directly comparing the lifetime and degradation rate of different fuel cells remains challenging as the test protocols, operating conditions, controllable variables, and reference state are not unified. Nevertheless, the detailed

454

Fuel Cells for Transportation

experimental conditions and procedures, as well as degradation rate, test duration, and failure modes, are comprehensively examined in this chapter. To identify the causes of various degradation mechanisms, many ex situ ASTs have been examined, and the stressors include temperature, humidity, hygrothermal, wet
dry, freeze
thaw, clamping force, and vibration conditions.

Acknowledgments This work received financial support from Canadian Urban Transit Research and Innovation Consortium (CUTRIC) via Project Number 160028, Natural Sciences and Engineering Research Council of Canada (NSERC) via a Discovery Grant and a CRD Grant (Project Number CRDPJ 522410-17), and Ballard Power Systems Inc. (Project Number SRA#077701).

References [1] Y. Zhao, Y. Liu, G. Liu, Q. Yang, L. Li, Z. Gao, Air and hydrogen supply systems and equipment for PEM fuel cells: a review, Int. J. Green. Energy 00 (2021) 1
18. Available from: https://doi.org/10.1080/15435075.2021.1946812. ¨ . Ekici, M. K¨oksal, Process modeling of an automotive pem fuel cell system, [2] A. Akroot, O Int. J. Green. Energy 16 (2019) 778
788. Available from: https://doi.org/10.1080/ 15435075.2019.1641105. [3] Y. Wang, K.S. Chen, J. Mishler, S.C. Cho, X.C. Adroher, A review of polymer electrolyte membrane fuel cells: technology, applications, and needs on fundamental research, Appl. Energy 88 (2011) 981
1007. Available from: https://doi.org/10.1016/j.apenergy.2010.09.030. [4] Y. Li, K. Moriyama, W. Gu, S. Arisetty, C.Y. Wang, A one-dimensional pt degradation model for polymer electrolyte fuel cells, J. Electrochem. Soc. 162 (2015) F834
F842. Available from: https://doi.org/10.1149/2.0101508jes. [5] Y. Li, C.Y. Wang, Modeling of transient platinum degradation in a low Pt-loading PEFC under current cycling, J. Electrochem. Soc. 164 (2017) F171
F179. Available from: https://doi.org/10.1149/2.0081704jes. [6] Y. Qin, Q. Du, M. Fan, Y. Chang, Y. Yin, Study on the operating pressure effect on the performance of a proton exchange membrane fuel cell power system, Energy Convers. Manag. 142 (2017) 357
365. Available from: https://doi.org/10.1016/j.enconman.2017.03.035. [7] S. Shahgaldi, I. Alaefour, J. Zhao, X. Li, Impact of ionomer in the catalyst layers on proton exchange membrane fuel cell performance under different reactant flows and pressures, Fuel 227 (2018) 35
41. Available from: https://doi.org/10.1016/j.fuel.2018.04.076. [8] S. Shahgaldi, J. Zhao, I. Alaefour, X. Li, Investigation of catalytic vs reactant transport effect of catalyst layers on proton exchange membrane fuel cell performance, Fuel 208 (2017) 321
328. Available from: https://doi.org/10.1016/j.fuel.2017.07.035. [9] A. Ozden, S. Shahgaldi, J. Zhao, X. Li, F. Hamdullahpur, Assessment of graphene as an alternative microporous layer material for proton exchange membrane fuel cells, Fuel 215 (2018) 726
734. Available from: https://doi.org/10.1016/j.fuel.2017.11.109. [10] L. Xing, W. Shi, H. Su, Q. Xu, P.K. Das, B. Mao, et al., Membrane electrode assemblies for PEM fuel cells: a review of functional graded design and optimization, Energy 177 (2019) 445
464. Available from: https://doi.org/10.1016/j.energy.2019.04.084.

Fuel cell durability Chapter | 15

455

[11] J. Zhao, X. Li, A review of polymer electrolyte membrane fuel cell durability for vehicular applications: degradation modes and experimental techniques, Energy Convers. Manag. 199 (2019) 112022. Available from: https://doi.org/10.1016/J.ENCONMAN.2019.112022. [12] J.T. Pukrushpan, A.G. Stefanopoulou, H. Peng, Background and introduction, Control. Fuel Cell Power Syst. (2004). [13] L. Xu, U. Reimer, J. Li, H. Huang, Z. Hu, H. Jiang, et al., Design of durability test protocol for vehicular fuel cell systems operated in power-follow mode based on statistical results of on-road data, J. Power Sources 377 (2018) 59
69. Available from: https://doi. org/10.1016/j.jpowsour.2017.11.075. [14] L. Xu, J. Li, U. Reimer, H. Huang, Z. Hu, H. Jiang, et al., Methodology of designing durability test protocol for vehicular fuel cell system operated in soft run mode based on statistic results of on-road data, Int. J. Hydrog. Energy 42 (2017) 29840
29851. Available from: https://doi.org/10.1016/j.ijhydene.2017.10.060. [15] U.S. Department of Energy. Fuel Cell Technologies Program Multi-Year Research, Development, and Demonstration Plan, 2017. [16] Ballard Power Systems Inc. Case study—Fuel cell zero-emission transit for the City of London. ,http://ballard.com/markets/transit-bus., n.d. (accessed 21.04.19). [17] P. Pei, Q. Chang, T. Tang, A quick evaluating method for automotive fuel cell lifetime, Int. J. Hydrog. Energy 33 (2008) 3829
3836. Available from: https://doi.org/10.1016/j. ijhydene.2008.04.048. [18] M. Marrony, R. Barrera, S. Quenet, S. Ginocchio, L. Montelatici, A. Aslanides, Durability study and lifetime prediction of baseline proton exchange membrane fuel cell under severe operating conditions, J. Power Sources 182 (2008) 469
475. Available from: https://doi.org/10.1016/j.jpowsour.2008.02.096. [19] P. Ren, P. Pei, Y. Li, Z. Wu, D. Chen, S. Huang, Degradation mechanisms of proton exchange membrane fuel cell under typical automotive operating conditions, Prog. Energy Combust. Sci. 80 (2020) 100859. Available from: https://doi.org/10.1016/j. pecs.2020.100859. [20] J. Wu, X.Z. Yuan, J.J. Martin, H. Wang, J. Zhang, J. Shen, et al., A review of PEM fuel cell durability: degradation mechanisms and mitigation strategies, J. Power Sources 184 (2008) 104
119. Available from: https://doi.org/10.1016/j.jpowsour.2008.06.006. [21] R. Borup, J. Meyers, B.S. Pivovar, Y.S. Kim, R. Mukundan, N. Garland, et al., Scientific aspects of polymer electrolyte fuel cell durability and degradation, Chem. Rev. 107 (2007) 3904
3951. Available from: https://doi.org/10.1021/cr050182l. [22] J. Zhang, PEM Fuel Cell Electrocatalysts and Catalyst Layers: Fundamentals and Applications, Springer London, London, 2008. [23] Fuel Cell Storer, NafionTM membranes ,https://www.fuelcellstore.com/fuel-cell-components/membranes/nafion., n.d. (accessed 7.08.19). [24] V. Arcella, C. Troglia, A. Ghielmi, Hyflon ion membranes for fuel cells, Ind. Eng. Chem. Res. 44 (2005) 7646
7651. Available from: https://doi.org/10.1021/ie058008a. [25] A. Ghielmi, P. Vaccarono, C. Troglia, V. Arcella, Proton exchange membranes based on the short-side-chain perfluorinated ionomer, J. Power Sources 145 (2005) 108
115. Available from: https://doi.org/10.1016/J.JPOWSOUR.2004.12.068. [26] A. Stassi, I. Gatto, E. Passalacqua, V. Antonucci, A.S. Arico, L. Merlo, et al., Performance comparison of long and short-side chain perfluorosulfonic membranes for high temperature polymer electrolyte membrane fuel cell operation, J. Power Sources 196 (2011) 8925
8930. Available from: https://doi.org/10.1016/j.jpowsour.2010.12.084.

456

Fuel Cells for Transportation

[27] S.J. Paddison, J.A. Elliott, Selective hydration of the ‘short-side-chain’ perfluorosulfonic acid membrane. An ONIOM study, Solid. State Ion. 178 (2007) 561
567. Available from: https://doi.org/10.1016/J.SSI.2007.01.007. [28] V. Arcella, A. Ghielmi, G. Tommasi, High performance perfluoropolymer films and membranes, Ann. N. Y. Acad. Sci. 984 (2003) 226
244. Available from: https://doi.org/ 10.1111/j.1749-6632.2003.tb06002.x. [29] K.D. Kreuer, M. Schuster, B. Obliers, O. Diat, U. Traub, A. Fuchs, et al., Short-side-chain proton conducting perfluorosulfonic acid ionomers: why they perform better in PEM fuel cells, J. Power Sources 178 (2008) 499
509. Available from: https://doi.org/10.1016/j. jpowsour.2007.11.011. [30] C. Lim, L. Ghassemzadeh, F. Van Hove, M. Lauritzen, J. Kolodziej, G.G. Wang, et al., Membrane degradation during combined chemical and mechanical accelerated stress testing of polymer electrolyte fuel cells, J. Power Sources 257 (2014) 102
110. Available from: https://doi.org/10.1016/J.JPOWSOUR.2014.01.106. [31] S. Kim, B.K. Ahn, M.M. Mench, Physical degradation of membrane electrode assemblies undergoing freeze/thaw cycling: diffusion media effects, J. Power Sources 179 (2008) 140
146. Available from: https://doi.org/10.1016/j.jpowsour.2007.12.114. [32] A.Z. Weber, Gas-crossover and membrane-pinhole effects in polymer-electrolyte fuel cells, J. Electrochem. Soc. 155 (2008) B521. Available from: https://doi.org/10.1149/1.2898130. [33] F. Rong, C. Huang, Z.S. Liu, D. Song, Q. Wang, Microstructure changes in the catalyst layers of PEM fuel cells induced by load cycling. Part I. Mechanical model, J. Power Sources 175 (2008) 699
711. Available from: https://doi.org/10.1016/j.jpowsour.2007.10.006. [34] B. Wahdame, D. Candusso, X. Franc¸ois, F. Harel, M.-C. Pe´ra, D. Hissel, et al., Comparison between two PEM fuel cell durability tests performed at constant current and under solicitations linked to transport mission profile, Int. J. Hydrog. Energy 32 (2007) 4523
4536. Available from: https://doi.org/10.1016/j.ijhydene.2007.03.013. [35] Y. Singh, R.T. White, M. Najm, T. Haddow, V. Pan, F.P. Orfino, et al., Tracking the evolution of mechanical degradation in fuel cell membranes using 4D in situ visualization, J. Power Sources 412 (2019) 224
237. Available from: https://doi.org/10.1016/J. JPOWSOUR.2018.11.049. [36] M. Bodner, B. Cermenek, M. Rami, V. Hacker, The effect of platinum electrocatalyst on membrane degradation in polymer electrolyte fuel cells, Membrane (Basel) 5 (2015) 888
902. Available from: https://doi.org/10.3390/membranes5040888. [37] D. Schonvogel, M. Rastedt, P. Wagner, M. Wark, A. Dyck, Impact of accelerated stress tests on high temperature PEMFC degradation, Fuel Cell 16 (2016) 480
489. Available from: https://doi.org/10.1002/fuce.201500160. [38] Z. Zhang, L. Jia, Parametric study of the porous cathode in the PEM fuel cell, Int. J. Energy Res. 33 (2009) 52
61. Available from: https://doi.org/10.1002/er.1471. [39] X. Huang, R. Solasi, Y. Zou, M. Feshler, K. Reifsnider, D. Condit, et al., Mechanical endurance of polymer electrolyte membrane and PEM fuel cell durability, J. Polym. Sci. Part. B Polym Phys. 44 (2006) 2346
2357. Available from: https://doi.org/10.1002/ polb.20863. [40] S.V. Venkatesan, M. Dutta, E. Kjeang, Mesoscopic degradation effects of voltage cycled cathode catalyst layers in polymer electrolyte fuel cells, Electrochem. Commun. 72 (2016) 15
18. Available from: https://doi.org/10.1016/J.ELECOM.2016.08.018. [41] R. Mukundan, A.M. Baker, A. Kusoglu, P. Beattie, S. Knights, A.Z. Weber, et al., Membrane accelerated stress test development for polymer electrolyte fuel cell durability

Fuel cell durability Chapter | 15

[42] [43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

457

validated using field and drive cycle testing, J. Electrochem. Soc. 165 (2018) F3085
F3093. Available from: https://doi.org/10.1149/2.0101806jes. M. Uchimura, S.S. Kocha, The impact of cycle profile on PEMFC durability, ECS Trans. 11 (2007) 1215
1226. Available from: https://doi.org/10.1149/1.2781035. vol. A.P. Young, J. Stumper, S. Knights, E. Gyenge, Ionomer degradation in polymer electrolyte membrane fuel cells, J. Electrochem. Soc. 157 (2010) B425
B436. Available from: https://doi.org/10.1149/1.3281899. C.G. Chung, L. Kim, Y.W. Sung, J. Lee, J.S. Chung, Degradation mechanism of electrocatalyst during long-term operation of PEMFC, Int. J. Hydrog. Energy 34 (2009) 8974
8981. Available from: https://doi.org/10.1016/j.ijhydene.2009.08.094. J. Zhao, S. Shahgaldi, A. Ozden, I.E. Alaefour, X. Li, F. Hamdullahpur, Geometric pore surface area and fractal dimension of catalyzed electrodes in polymer electrolyte membrane fuel cells, Int. J. Energy Res. 43 (2019) 3011
3019. Available from: https://doi.org/ 10.1002/er.4260. J. Zhao, S. Shahgaldi, I. Alaefour, S. Yang, X. Li, Pore structure and effective diffusion coefficient of catalyzed electrodes in polymer electrolyte membrane fuel cells, Int. J. Hydrog. Energy 43 (2018) 3776
3785. Available from: https://doi.org/10.1016/j. ijhydene.2018.01.019. J. Zhao, A. Ozden, S. Shahgaldi, I.E. Alaefour, X. Li, F. Hamdullahpur, Effect of Pt loading and catalyst type on the pore structure of porous electrodes in polymer electrolyte membrane (PEM) fuel cells, Energy 150 (2018) 69
76. Available from: https://doi.org/ 10.1016/j.energy.2018.02.134. J. Zhao, S. Shahgaldi, A. Ozden, I.E. Alaefour, X. Li, F. Hamdullahpur, Effect of catalyst deposition on electrode structure, mass transport and performance of polymer electrolyte membrane fuel cells, Appl. Energy 255 (2019) 113802. Available from: https://doi.org/ 10.1016/J.APENERGY.2019.113802. Y. Yin, J. Liu, Y. Chang, Y. Zhu, X. Xie, Y. Qin, et al., Design of Pt-C/Fe-N-S-C cathode dual catalyst layers for proton exchange membrane fuel cells under low humidity, Electrochim. Acta 296 (2019) 450
457. Available from: https://doi.org/10.1016/j. electacta.2018.11.048. J.K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J.R. Kitchin, T. Bligaard, et al., Origin of the overpotential for oxygen reduction at a fuel-cell cathode, J. Phys. Chem. B 108 (2004) 17886
17892. Available from: https://doi.org/10.1021/jp047349j. O.T. Holton, J.W. Stevenson, The role of platinum in proton exchange membrane fuel cells, Platin. Met. Rev. 57 (2013) 259
271. Available from: https://doi.org/10.1595/ 147106713X671222. M. Neergat, R. Rahul, Unsupported Cu-Pt core-shell nanoparticles: oxygen reduction reaction (ORR) catalyst with better activity and reduced precious metal content, J. Electrochem. Soc. 159 (2012) F234
F241. Available from: https://doi.org/10.1149/2.039207jes. J. Zhao, S. Shahgaldi, X. Li, Z.(Simon) Liu, Experimental observations of microstructure changes in the catalyst layers of proton exchange membrane fuel cells under wet-dry cycles, J. Electrochem. Soc. 165 (2018) F3337
F3345. Available from: https://doi.org/ 10.1149/2.0391806jes. A. Ozden, S. Shahgaldi, J. Zhao, X. Li, F. Hamdullahpur, Degradations in porous components of a proton exchange membrane fuel cell under freeze-thaw cycles: morphology and microstructure effects, Int. J. Hydrog. Energy 45 (2020) 3618
3631. Available from: https://doi.org/10.1016/j.ijhydene.2018.10.209.

458

Fuel Cells for Transportation

[55] E.N. Alvar, B. Zhou, S.H. Eichhorn, Carbon-embedded mesoporous Nb-doped TiO2 nanofibers as catalyst support for the oxygen reduction reaction in PEM fuel cells, J. Mater. Chem. A 4 (2016) 6540
6552. Available from: https://doi.org/10.1039/C5TA08801A. [56] E. Navaei Alvar, B. Zhou, S.H. Eichhorn, Composite-supported Pt catalyst and electrosprayed cathode catalyst layer for polymer electrolyte membrane fuel cell, Int. J. Energy Res. 41 (2017) 1626
1641. Available from: https://doi.org/10.1002/er.3746. [57] X. Li, Principles of Fuel Cells, Taylor & Francis Group, New York, 2006. [58] E. Pizzutilo, S. Geiger, J.-P. Grote, A. Mingers, K.J.J. Mayrhofer, M. Arenz, et al., On the need of improved accelerated degradation protocols (ADPs): examination of platinum dissolution and carbon corrosion in half-cell tests, J. Electrochem. Soc. 163 (2016) 1510
1514. Available from: https://doi.org/10.1149/2.0731614jes. [59] S. Ye, M. Hall, H. Cao, P. He, Degradation resistant cathodes in polymer electrolyte membrane fuel cells, ECS Trans 3 (2006) 657
666. Available from: https://doi.org/10.1149/1.2356186. vol. [60] Y. Shao, G. Yin, Y. Gao, Understanding and approaches for the durability issues of Ptbased catalysts for PEM fuel cell, J. Power Sources 171 (2007) 558
566. Available from: https://doi.org/10.1016/j.jpowsour.2007.07.004. [61] T.W. Hansen, A.T. Delariva, S.R. Challa, A.K. Datye, Sintering of catalytic nanoparticles : particle migration or Ostwald ripening ? Acc. Chem. Res. 46 (2013) 1720
1730. [62] N. Macauley, K.H. Wong, M. Watson, E. Kjeang, Favorable effect of in-situ generated platinum in the membrane on fuel cell membrane durability, J. Power Sources 299 (2015) 139
148. Available from: https://doi.org/10.1016/J.JPOWSOUR.2015.08.096. [63] K.J.J. Mayrhofer, J.C. Meier, S.J. Ashton, G.K.H. Wiberg, F. Kraus, M. Hanzlik, et al., Fuel cell catalyst degradation on the nanoscale, Electrochem. Commun. 10 (2008) 1144
1147. Available from: https://doi.org/10.1016/j.elecom.2008.05.032. [64] P. Parthasarathy, A.V. Virkar, Electrochemical Ostwald ripening of Pt and Ag catalysts supported on carbon, J. Power Sources 234 (2013) 82
90. Available from: https://doi.org/ 10.1016/J.JPOWSOUR.2013.01.115. [65] F.A. De Bruijn, V.A.T. Dam, G.J.M. Janssen, Review: Durability and degradation issues of PEM fuel cell components, Fuel Cell 8 (2008) 3
22. Available from: https://doi.org/ 10.1002/fuce.200700053. [66] L. Ma, Z.S. Liu, C. Huang, S.H. Chen, G.W. Meng, Microstructure changes induced by capillary condensation in catalyst layers of PEM fuel cells, Int. J. Hydrog. Energy 35 (2010) 12182
12190. [67] N. Macauley, D.D. Papadias, J. Fairweather, D. Spernjak, D. Langlois, R. Ahluwalia, et al., Carbon corrosion in PEM fuel cells and the development of accelerated stress tests, J. Electrochem. Soc. 165 (2018) F3148
F3160. Available from: https://doi.org/10.1149/ 2.0061806jes. [68] P. Mandal, B.K. Hong, J.-G. Oh, S. Litster, Understanding the voltage reversal behavior of automotive fuel cells, J. Power Sources 397 (2018) 397
404. Available from: https:// doi.org/10.1016/J.JPOWSOUR.2018.06.083. [69] L. Castanheira, W.O. Silva, F.H.B. Lima, A. Crisci, L. Dubau, F. Maillard, Carbon corrosion in proton-exchange membrane fuel cells: effect of the carbon structure, the degradation protocol, and the gas atmosphere, ACS Catal. 5 (2015) 2184
2194. Available from: https://doi.org/10.1021/cs501973j. [70] Y. Chang, J. Zhao, S. Shahgaldi, Y. Qin, Y. Yin, X. Li, Modelling of mechanical microstructure changes in the catalyst layer of a polymer electrolyte membrane fuel cell, Int. J. Hydrog. Energy 45 (2020) 29904
29916. Available from: https://doi.org/10.1016/j.ijhydene.2018.10.157.

Fuel cell durability Chapter | 15

459

[71] J. Zhou, S. Shukla, A. Putz, M. Secanell, Analysis of the role of the microporous layer in improving polymer electrolyte fuel cell performance, Electrochim. Acta 268 (2018) 366
382. Available from: https://doi.org/10.1016/J.ELECTACTA.2018.02.100. [72] Y. Shen, Mechanical Degradation of Membrane Electrode Assemblies in Proton Exchange Membrane Fuel Cells, University of Waterloo, 2017. [73] J.E. Owejan, P.T. Yu, R. Makharia, Mitigation of carbon corrosion in microporous layers in PEM fuel cells, ECS Trans. 11 (2007) 1049
1057. Available from: https://doi.org/ 10.1149/1.2781018. vol. [74] H. Tawfik, Y. Hung, D. Mahajan, Metal bipolar plates for PEM fuel cell—a review, J. Power Sources 163 (2007) 755
767. Available from: https://doi.org/10.1016/j. jpowsour.2006.09.088. [75] J. Wind, R. Sp¨ah, W. Kaiser, G. B¨ohm, Metallic bipolar plates for PEM fuel cells, J. Power Sources 105 (2002) 256
260. Available from: https://doi.org/10.1016/S0378-7753 (01)00950-8. [76] R.C. Makkus, A.H. Janssen, F.A. de Bruijn, R.K.A. Mallant, Use of stainless steel for cost competitive bipolar plates in the SPFC, J. Power Sources 86 (2000) 274
282. Available from: https://doi.org/10.1016/S0378-7753(99)00460-7. [77] S.J.C. Cleghorn, D.K. Mayfield, D.A. Moore, J.C. Moore, G. Rusch, T.W. Sherman, et al., A polymer electrolyte fuel cell life test: 3 years of continuous operation, J. Power Sources 158 (2006) 446
454. Available from: https://doi.org/10.1016/j.jpowsour.2005.09.062. [78] A.J.L. Verhage, J.F. Coolegem, M.J.J. Mulder, M.H. Yildirim, De, F.A. Bruijn, 30,000 h operation of a 70 kW stationary PEM fuel cell system using hydrogen from a chlorine factory, Int. J. Hydrog. Energy 38 (2013) 4714
4724. Available from: https://doi.org/ 10.1016/j.ijhydene.2013.01.152. [79] H. Xu, M. Wu, Y. Liu, V. Mittal, F. Kassim, B. Vieth, et al., Durability of NafionsTeflons-phosphotungstic acid composite membranes in PEM fuel cells at 100oC and 25% RH, ECS Trans. 3 (2006) 561
568. Available from: https://doi.org/10.1149/ 1.2356177. vol. [80] J. St-Pierre, D.P. Wilkinson, S. Knights, M. Bos, Relationships between water management, contamination and lifetime degradation in PEFC, J. N. Mater. Electrochem. Syst. 3 (2000) 99
106. [81] J. St-Pierre, N. Jia, Successful demonstration of Ballard PEMFCS for space shuttle applications, J. N. Mater. Electrochem. Syst. 5 (2002) 263
271. Available from: https://doi. org/10.1080/19397030902947041. [82] S.D. Knights, K.M. Colbow, J. St-Pierre, D.P. Wilkinson, Aging mechanisms and lifetime of PEFC and DMFC, J. Power Sources 127 (2004) 127
134. Available from: https://doi. org/10.1016/J.JPOWSOUR.2003.09.033. [83] T.R. Ralph, Proton exchange membrane fuel cells: progress in cost reduction of the key components, Platin. Met. Rev. 41 (1997) 102
113. Available from: https://doi.org/ 10.1016/0016-2361(96)82642-1. [84] C. Sishtla, G. Koncar, R. Platon, S. Gamburzev, A.J. Appleby, O.A. Velev, Performance and endurance of a PEMFC operated with synthetic reformate fuel feed, J. Power Sources 71 (1998) 249
255. Available from: https://doi.org/10.1016/S0378-7753(97)02719-5. [85] F. Nandjou, J.-P. Poirot-Crouvezier, M. Chandesris, J.-F. Blachot, C. Bonnaud, Y. Bultel, Impact of heat and water management on proton exchange membrane fuel cells degradation in automotive application, J. Power Sources 326 (2016) 182
192. Available from: https://doi.org/10.1016/j.jpowsour.2016.07.004.

460

Fuel Cells for Transportation

[86] M.W. Fowler, R.F. Mann, J.C. Amphlett, B.A. Peppley, P.R. Roberge, Incorporation of voltage degradation into a generalised steady state electrochemical model for a PEM fuel cell, J. Power Sources 106 (2002) 274
283. Available from: https://doi.org/10.1016/ S0378-7753(01)01029-1. [87] X. Cheng, L. Chen, C. Peng, Z. Chen, Y. Zhang, Q. Fan, Catalyst microstructure examination of PEMFC membrane electrode assemblies vs. time, J. Electrochem. Soc. 151 (2004) A48
A52. Available from: https://doi.org/10.1149/1.1625944. [88] J. Scholta, N. Berg, P. Wilde, L. J¨orissen, J. Garche, Development and performance of a 10 kW PEMFC stack, J. Power Sources 127 (2004) 206
212. Available from: https://doi. org/10.1016/J.JPOWSOUR.2003.09.040. [89] J. Yu, T. Matsuura, Y. Yoshikawa, M.N. Islam, M. Hori, In situ analysis of performance degradation of a PEMFC under nonsaturated humidification, Electrochem. Solid-State Lett. 8 (2005) A156
A158. Available from: https://doi.org/10.1149/1.1854781. [90] S. Yoshioka, A. Yoshimura, H. Fukumoto, O. Hiroi, H. Yoshiyasu, Development of a PEFC under low humidified conditions, J. Power Sources 144 (2005) 146
151. Available from: https://doi.org/10.1016/J.JPOWSOUR.2004.12.045. [91] P.J. Ferreira, G.J. la O’, Y. Shao-Horn, D. Morgan, R. Makharia, S. Kocha, et al., Instability of Pt/C electrocatalysts in proton exchange membrane fuel cells, J. Electrochem. Soc. 152 (2005) A2256
A2271. Available from: https://doi.org/10.1149/ 1.2050347. [92] E. Endoh, S. Terazono, H. Widjaja, Y. Takimoto, Degradation study of MEA for PEMFCs under low humidity conditions, Electrochem. Solid-state Lett. 7 (2004) a209
a211. Available from: https://doi.org/10.1149/1.1739314. [93] H. Chen, Z. Song, X. Zhao, T. Zhang, P. Pei, C. Liang, A review of durability test protocols of the proton exchange membrane fuel cells for vehicle, Appl. Energy 224 (2018) 289
299. Available from: https://doi.org/10.1016/J.APENERGY.2018.04.050. [94] A. Kneer, J. Jankovic, D. Susac, A. Putz, N. Wagner, M. Sabharwal, et al., Correlation of changes in electrochemical and structural parameters due to voltage cycling induced degradation in PEM fuel cells, J. Electrochem. Soc. 165 (2018) F3241
F3250. Available from: https://doi.org/10.1149/2.0271806jes. [95] A.P. Hitchcock, V. Berejnov, V. Lee, M. West, V. Colbow, M. Dutta, et al., Carbon corrosion of proton exchange membrane fuel cell catalyst layers studied by scanning transmission x-ray microscopy, J. Power Sources 266 (2014) 66
78. Available from: https://doi. org/10.1016/J.JPOWSOUR.2014.04.119. [96] K. Panha, M. Fowler, X.-Z.Z. Yuan, H. Wang, Accelerated durability testing via reactants relative humidity cycling on PEM fuel cells, Appl. Energy 93 (2012) 90
97. Available from: https://doi.org/10.1016/j.apenergy.2011.05.011. [97] F. Rong, C. Huang, Z.S. Liu, D. Song, Q. Wang, Microstructure changes in the catalyst layers of PEM fuel cells induced by load cycling. Part II. Simulation and understanding, J. Power Sources 175 (2008) 712
723. Available from: https://doi.org/10.1016/j. jpowsour.2007.10.007. [98] S.J. Bae, S.J. Kim, J.I. Park, C.W. Park, J.H. Lee, I. Song, et al., Lifetime prediction of a polymer electrolyte membrane fuel cell via an accelerated startup-shutdown cycle test, Int. J. Hydrog. Energy 37 (2012) 9775
9781. Available from: https://doi.org/10.1016/j. ijhydene.2012.03.104. [99] I. Bloom, L.K. Walker, J.K. Basco, T. Malkow, A. Saturnio, G. De Marco, et al., A comparison of fuel cell testing protocols—a case study: protocols used by the U.S. Department of Energy, European Union, International Electrotechnical Commission/Fuel Cell Testing and

Fuel cell durability Chapter | 15

[100]

[101]

[102]

[103]

[104]

[105]

[106]

[107]

[108]

[109] [110]

[111]

[112]

[113]

[114]

461

Standardization Network, and Fuel Cell Technical Team, J. Power Sources 243 (2013) 451
457. Available from: https://doi.org/10.1016/j.jpowsour.2013.06.026. I. Bloom, J.K. Basco, L.K. Walker, T. Malkow, G. De Marco, A. Saturnio, et al., Fuel cell testing protocols an international perspective, Publ. Office (2013). Available from: https://doi.org/10.2790/86676. R. Lin, B. Li, Y.P. Hou, J.M. Ma, Investigation of dynamic driving cycle effect on performance degradation and micro-structure change of PEM fuel cell, Int. J. Hydrog. Energy 34 (2009) 2369
2376. Available from: https://doi.org/10.1016/j.ijhydene.2008.10.054. I. Bloom, L. Walker, J. Basco, T. Malkow, G. De Marco, G. Tsotridis, et al., A comparison of fuel cell test protocols, ECS Trans. 30 (2011) 227
235. Available from: https:// doi.org/10.1149/1.3562478. vol. N. Garland, T. Benjamin, J. Kopasz, DOE Fuel cell program: durability technical targets and testing protocols, ECS Trans. 11 (2007) 923
931. Available from: https://doi.org/ 10.1149/1.2781004. vol. D. Bezmalinovic, B. Simic, F. Barbir, Characterization of PEM fuel cell degradation by polarization change curves, J. Power Sources 294 (2015) 82
87. Available from: https:// doi.org/10.1016/j.jpowsour.2015.06.047. L. Lu, M. Ouyang, H. Huang, P. Pei, F. Yang, A semi-empirical voltage degradation model for a low-pressure proton exchange membrane fuel cell stack under bus city driving cycles, J. Power Sources 164 (2007) 306
314. Available from: https://doi.org/ 10.1016/j.jpowsour.2006.10.061. D. Liu, S. Case, Durability study of proton exchange membrane fuel cells under dynamic testing conditions with cyclic current profile, J. Power Sources 162 (2006) 521
531. Available from: https://doi.org/10.1016/J.JPOWSOUR.2006.07.007. S.V. Selvaganesh, G. Selvarani, P. Sridhar, S. Pitchumani, A.K. Shukla, Durable electrocatalytic-activity of Pt
Au/C cathode in PEMFCs, Phys. Chem. Chem Phys 13 (2011) 12623
12634. Available from: https://doi.org/10.1039/c1cp20243j. H. Noto, M. Kondo, Y. Otake, M. Kato, Development of fuel cell hybrid vehicle by Toyota-durability-, SAE Tech. Pap. (2009) 1
6. Available from: https://doi.org/10.4271/ 2009-01-1002. M. Fowler, J.C. Amphlett, R.F. Mann, B.A. Peppley, P.R. Roberge, Issues associated with voltage degradation in a PEMFC, J. N. Mater. Electrochem. Syst. 5 (2002) 255
262. V.A. Sethuraman, J.W. Weidner, A.T. Haug, L.V. Protsailo, Durability of perfluorosulfonic acid and hydrocarbon membranes: effect of humidity and temperature, J. Electrochem. Soc. 155 (2008) B119
B124. Available from: https://doi.org/10.1149/ 1.2806798. K. Miyatake, H. Furuya, M. Tanaka, M. Watanabe, Durability of sulfonated polyimide membrane in humidity cycling for fuel cell applications, J. Power Sources 204 (2012) 74
78. Available from: https://doi.org/10.1016/J.JPOWSOUR.2011.12.039. W. Bi, T.F. Fuller, Temperature effects on PEM fuel cells Pt/C catalyst degradation, J. Electrochem. Soc. 155 (2008) B215
B221. Available from: https://doi.org/10.1149/ 1.2819680. M.N. Silberstein, M.C. Boyce, Hygro-thermal mechanical behavior of Nafion during constrained swelling, J. Power Sources 196 (2011) 3452
3460. Available from: https:// doi.org/10.1016/j.jpowsour.2010.11.116. R. Banan, J. Zu, A. Bazylak, Humidity and temperature cycling effects on cracks and delaminations in PEMFCs, Fuel Cell 15 (2015) 327
336. Available from: https://doi. org/10.1002/fuce.201400118.

462

Fuel Cells for Transportation

[115] J.H. Lin, W.H. Chen, S.H. Su, Y.J. Su, T.H. Ko, Washing experiment of the gas diffusion layer in a proton-exchange membrane fuel cell, Energy Fuels 22 (2008) 2533
2538. Available from: https://doi.org/10.1021/ef800116c. [116] R.C. McDonald, C.K. Mittelsteadt, E.L. Thompson, Effects of deep temperature cycling on Nafions 112 membranes and membrane electrode assemblies, Fuel Cell 4 (2004) 208
213. Available from: https://doi.org/10.1002/fuce.200400015. [117] Q. Yan, H. Toghiani, Y.-W. Lee, K. Liang, H. Causey, Effect of sub-freezing temperatures on a PEM fuel cell performance, startup and fuel cell components, J. Power Sources 160 (2006) 1242
1250. Available from: https://doi.org/10.1016/J. JPOWSOUR.2006.02.075. [118] P.K. Sow, S. Prass, P. Kalisvaart, W. Me´rida, Deconvolution of electrical contact and bulk resistance of gas diffusion layers for fuel cell applications, Int. J. Hydrog. Energy https://doi.org/10.1016/J. 40 (2015) 2850
2861. Available from: IJHYDENE.2014.12.110. [119] A.M. Dafalla, F. Jiang, Stresses and their impacts on proton exchange membrane fuel cells: a review, Int. J. Hydrog. Energy 43 (2018) 2327
2348. Available from: https:// doi.org/10.1016/j.ijhydene.2017.12.033. [120] G. Diloyan, M. Sobel, K. Das, P. Hutapea, Effect of mechanical vibration on platinum particle agglomeration and growth in polymer electrolyte membrane fuel cell catalyst layers, J. Power Sources 214 (2012) 59
67. Available from: https://doi.org/10.1016/j. jpowsour.2012.04.027. [121] N. Rajalakshmi, S. Pandian, K.S. Dhathathreyan, Vibration tests on a PEM fuel cell stack usable in transportation application, Int. J. Hydrog. Energy 34 (2009) 3833
3837. Available from: https://doi.org/10.1016/J.IJHYDENE.2009.03.002.

Chapter 16

Subzero startup of polymer electrolyte fuel cell—a battle between water and thermal management at low temperatures Jianbo Zhang1, Dechun Si2 and Kei Ono1 1

School of Vehicle and Mobility, Tsinghua University, Beijing, P.R. China, 2Beijing Huairou Laboratory, Beijing, P.R. China

16.1 Introduction Fuel cell (FC) has superior energy conversion efficiency at ambient temperature and even greater combined heat and power efficiency at high temperature, as it converts the chemical energy in the reactants directly into electricity via an electrochemical reaction. In addition, the energy source for the FC is diverse and independent of fossil fuels. Therefore, FC is one of the key technologies to enable sustainable growth. Of the four major types of FC, polymer electrolyte fuel cell (PEFC) based on perfluoro sulfonic acid (PFSA) membrane has low operation temperature (,100 C) and high power density and hence is particularly suitable as a power source for automotive applications, which has frequent startup/shutdown, dynamic load change, and limited space. The anode/cathode and overall reactions of the PEFC are listed in Eqs. (16.1
16.3). The outcome of the FC reactions is product water and heat. As there is no emission of NOx, COx, SOx, and particulate matter, fuel cell electric vehicles (FCEVs) constitute a major candidate for zero-emission vehicles (ZEVs). Anode: 0 H2 -2H1 1 2e2 EAnode 5 0V

Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00012-5 © 2023 Elsevier Ltd. All rights reserved.

ð16:1Þ

463

464

Fuel Cells for Transportation

Cathode: 0 O2 1 4H1 1 4e2 -2H2 O 1 heat ECathode 5 1:23 V

ð16:2Þ

1 0 H2 1 O2 -H2 O 1 heat ECell 5 1:23 V 2

ð16:3Þ

Overall:

The ambient operating temperature of the PEFC turns out to be a doubleedged sword. While allowing for dynamic load change and frequent startup/ shutdown, it makes the PEFC addicted to the use of platinum group precious metals as catalyst, particularly for the sluggish oxygen reduction reaction (ORR), on the one hand, and beleaguered by the complex, temperaturesensitive two-phase flow of water on the other hand. This chapter addresses the latter issue, that is, the water and thermal management at subzero temperatures. Water and thermal management are critical for the efficient, reliable, and durable operation of FCEVs. The task of water management is to maintain all the FC components in their suitable range of humidification across the FC stack (from cell to cell, from inlet to outlet) and across the operation conditions (from idling to peak power, from steady load to dynamic change, from summer to winter). Specifically, the proton-exchange membrane (PEM) needs to be well humidified to be conductive of protons, while the catalyst layer (CL) and gas diffusion layer (GDL) should have adequate dry pores left to allow for the transport of reaction gases to the catalyst. Product water needs to be removed from the reaction sites to prevent flooding. As the normal operating temperature for the PEFC in FCEVs is between 60 C and 90 C, two-phase flows of water across the thin, composite porous media with pore size in nanometers and micrometers and in flow channels with a cross-section in millimeters are the key processes to be regulated for proper water management. The task of thermal management is first to remove the waste heat from the PEFCs, so as to control the maximum temperature and to avoid the damage of the PEM due to the hot spot. Even though PEFCs have much higher efficiency than ICEs, the removal of waste heat is not a trivial task as the temperature difference between the PEFC and the ambient is much less than that of the ICEs. In fact, the nominal power of an FC stack is determined by the capability of the radiator to dissipate the heat. Second, the thermal management needs to adjust the temperature and its distribution to assist water management, as the saturation water vapor pressure becomes increasingly sensitive to temperatures from 60 C to 90 C. New challenges for water and thermal management emerge when the ambient temperature drops below zero. During the subzero startup, product water inside the CL may freeze before the FC temperature reaches 0 C. The frozen water may block the pores in the porous media, causing the failure of

Subzero startup of polymer electrolyte fuel cell Chapter | 16

465

a subzero startup. Moreover, the volume expansion of the frozen water can cause irreversible damage to FC components. For thermal management, while the heat dissipation issue is alleviated and the waste heat is adequate to maintain the nominal operation temperature, how to quickly start up the cells soaked in subzero before the product water freezes is rather demanding. In fact, the subzero startup problem constitutes one of the main obstacles to the diffusion of FCEVs in cold regions. Due to the importance as well as the complexity involved in the subzero startup, DOE formulated specific technical targets to guide the R&D in academia and industry in the early 2000s [1]. The cold startup time to 50% of rated power from 220 C ambient temperature should be less than 30 seconds, the startup and shutdown energy should be less than 5 MJ, and unassisted startup should be capable from ambient temperatures as low as 230 C. Extensive efforts have been taken to understand the damages from freeze/thaw cycles, the state of water in the membrane using ex situ characterization, and the startup failure mechanism using the water fill test (WFT). Based on these understandings, startup strategies and techniques are developed and culminated in the startup capability of Toyota’s passenger FCEV in 2015 [2], which is capable of starting up from 230 C, exceeding DOE’s ultimate targets. However, a number of issues still remain. To expand FC application in commercial vehicles, especially heavy-duty vehicles (HDVs), stacks using graphite bipolar plates (BPPs) may be needed due to durability requirements. Graphite-based BPPs have a much higher thermal mass than its metal counterpart, hence, the startup is harder. In addition, to extend the diffusion of FCEVs in cold areas which may experience subzero temperature for longer days, the degradation from repeated subzero startup, which is not specified in DOE’s targets, becomes important. Accordingly, a resurging of R&D activities on subzero startups in recent years can be revealed from the survey of papers and patents [3]. The current understanding of the subzero problem is achieved through the combined efforts of both experiment and simulation. A number of reviews [3
8] and chapters [9,10] are available to give extensive coverage of these experiments and simulation studies. This chapter aims to update and summarize these studies from the perspective of water and thermal management. Emphasis is put on experimental studies rather than modeling and simulation, as experimental observations are fundamental in identifying the critical phenomena, justifying or discrediting assumptions for developing models, parameterizing the models, and validating the simulation results. Due to the space limit, only representative results can be listed. Despite the best efforts of the authors to be objective and balanced, the update and summary will undoubtedly be incomplete and biased due to our preferences and prejudice. This chapter also includes our explanation and speculation of the

466

Fuel Cells for Transportation

observed results, and our opinion on the right direction for further study, hoping to facilitate comprehension and provoke discussion. The rest of the chapter is arranged as follows. Section 16.2 gives an overview of the subzero experiment, including four categories of experiments, major test fixtures, typical test procedures, and characterization techniques. Section 16.3 presents the study of the damages of FCs in subzero scenarios. These findings in the early years exposed the severity of the problem, caught the attention of the research community and the government funding agency, and set the foundation as well as directions for the subsequent in-depth exploration of the water and thermal issues at subzero. Sections 16.4 and 16.5 summarize systematically the knowledge of the states and behavior of water at subzero, and the temperature-dependent properties and thermal behavior at subzero, respectively. The knowledge and equations in these two sections form the basis for the modeling and optimization of water and thermal management at subzero. Section 16.6 compares the unassisted and assisted subzero startup strategies and techniques, applying the understanding in Sections 16.4 and 16.5 to address the issues in Section 16.3. Section 16.7 comments on the directions for further study.

16.2 Overview of subzero experiment This section presents an overview of the subzero experiment, with an emphasis on the remarkable advances in technique and milestones in understanding. The related tests are first classified into four categories, each with its specific issue and state-of-the-art understanding. Then, major test fixtures, typical test procedures, and featured characterization techniques will be introduced.

16.2.1 Four categories of experiment in subzero study The experimental study of the PEFC subzero startup can be grouped into four categories, according to the key issue addressed. The first category explores the damages and mitigation of the subzero startup, typically using freeze/thaw cycles. The second studies the states of water in and the properties of the PFSA membrane with ex situ characterization. The third examines the states and behavior of water in the membrane electrode assembly (MEA) as well as the failure mechanism, using a WFT typically under isothermal conditions. The fourth develops startup strategies and deals with the competition between water and thermal management, using insulated/ heated single cells or short stacks. These four categories also roughly follow the logical and actual research sequence. Sections 16.3, 16.4, and 16.6 address the first, second and third, and fourth categories, respectively. Table 16.1 lists the issues, related tests, and major understanding of these four categories.

TABLE 16.1 Issues, related tests, and major understanding for the four categories of subzero experiment. Category

Issues

Related tests

1

Damages and mitigation

G

G

Freeze/thaw of components and cells Subzero startup of cells and short stacks

Major understanding G

G

2

State of water in and properties of PFSA at subzero

G G G

States of water: DSC Mechanical property: DMA, TTM Measurement of water content, proton conductivity, water diffusivity, electroosmotic drag coefficient

G

G

G

Major modes of degradation: (1) freezing of residual/product water causes major damage to the interface and porous structure [11]; (2) low RH and low water content operation cause damages to the PFSA [12]. Mitigation via operation and design: (1) purging is effective to mitigate freezing-induced damage [13]; (2) defect-free catalyst layer coated on a thin/reinforced membrane, protected by stiff diffusion media and narrow channel flow field is more tolerant to freezing [14,15]; (3) incorporation of radical quencher Ce31 can prevent the decomposition of the PFSA membrane [16]. There exist three states of water in PFSA. The closely bonded water does not freeze, the loosely bonded freeze as low as 220 C, and the bulk-like water freeze as low as 25 C to 210 C due to melting point depression effect [17,18]. The mechanical properties of wet NafionTM membrane change dramatically with temperature, that is, from a rubber-like behavior at room temperature to a brittle behavior below 180K. The freezing of the nanoconfined water is complete only below 180K [19]. Saturation water content and transport properties decrease with temperature [20]. The drag coefficient is close to 1 at subzero [21]. (Continued )

TABLE 16.1 (Continued) Category

Issues

Related tests

3

States and behaviors of water in MEA and startup failure mechanism

G

Isothermal water fill test using cells

Major understanding G

G

G

4

Startup strategies and techniques

G

G

Unassisted start using cells and stacks Assisted start using cells and short stacks

G

G

G

Cumulative product water is greater for lower current density, lower initial water content, higher subzero temperature, hydrophilic MPL [22
24], and thin membrane at higher current density [25]. While the melting point depression for the typical pore sizes in the CL and GDL amounts to only several degrees below zero, supercooled water is observed in the fuel cell at temperatures as low as 220 C. Yet the supercooled water is quasi-stable and freezes stochastically [26,27]. The dispersion in results is high for higher subzero temperatures and smaller active area cells [28]. The failure of the subzero operation is caused by the blocking of the pores from the frozen water which chokes the supply of reactants, not by the covering of Pt by ice, as the ECSA remains unchanged [29]. For the unassisted start, low thermal mass, high waste heat generation rate per unit product water, and proper range of residual water in the membrane are the three pillars for success. Toyota’s MIRAI FCEV has met and exceeded DOE’s subzero startup targets [30]. Significant temperature distribution may develop. Cold endplates with large thermal mass cause adjacent cells to have lower temperatures. Anode end cells have lower voltage than cathode end cells [31]. The cold periphery of cells causes the current to concentrate in the central area [32]. For the assisted start, the alternating hydrogen pump can start the cell from subzero reliably, efficiently, and durably [33,34].

Subzero startup of polymer electrolyte fuel cell Chapter | 16

469

16.2.2 Test fixtures During the last two decades, researchers have employed and developed many levels of test fixtures, from the material and components to single cell and short stack, in order to address the issues in the above four categories of experimental study. Among these test fixtures, single cells under isothermal boundary conditions or freeze/thaw cycles are most frequently employed to study the behavior of water in MEAs, the failure, and the degradation mechanism. Multiple single cells were developed to take the statistics in the WFT. Single cells with enhanced insulation and external heating to simulate the central cells in a stack were developed to study the subzero startup strategies. Table 16.2 lists the levels of the test objects, corresponding categories, issue and purpose, and key references in the subzero study.

16.2.3 Test procedures and control of temperature Typical procedures for the WFT, freeze/thaw, and subzero startup are shown in Fig. 16.1. Purging and cooling/warming are special steps in the subzero experiment. The purpose of purging is to adjust the level of residual water inside the MEA so as to create water storage capacity and improve the repeatability of the test. While dry purge takes less time, equilibrium purge [25] controls the residual water content in the PFSA membrane more accurately and hence greatly increases the test repeatability. Tajiri et al. [51] conducted a systematic study on the stages of water removal and their controlling parameters in purging. Different from the FC experiment at normal temperature (60 C
80 C), the subzero experiments usually take a very long time. For example, a single freeze/thaw cycle usually takes about 10 hours [11]. The most time-consuming step is cooling down to subzero temperature. Four methods for realizing the subzero are listed in Table 16.3. The environmental SEM is capable of controlling both temperature and relative humidity (RH), hence phase changes among liquid water, vapor, and ice, yet it is only applicable to FC components. The Peltier element can lower the temperature of a small-sized single cell from room temperature to 25 C in several minutes, hence attractive for speeding up the freeze/thaw test, yet there is a limit for the achievable low temperature due to the weak cooling power of the Peltier element. In addition, the Peltier element is usually used in combination with a recirculating cooling system. The environment chamber is capable of reaching 230 C and is most frequently used for a single cell, yet it has a long cooling time, especially for short stacks. The recirculating coolant method can quickly quench the stack to subzero through the recirculating system, yet there is a limit for temperature lower than 220 C. Therefore, the most suitable method for temperature control depends on the test objects, the required cooling rate, and the cost.

TABLE 16.2 Test fixtures in subzero study. Levels of the test objects

Corresponding categories

Issue and purpose

Material and components

1, 2

G G

Single cell

Key references

The material properties at subzero Freeze/thaw damages and mitigation strategy

[11,12,14,15,17
20,35]

Isothermal boundary condition

1, 3

G

Simulating the end cells to examine failure mechanism and degradation in water fill test

[9,22,23,26
28,36
39]

Insulated, quasi-adiabatic, adiabatic thermal boundary condition

1, 4

G

Simulating the center cells in the stack to explore startup strategy

[40
42]

Multiple single cells

3

G

Statistics in water fill tests

[38,39]

Short stack

1, 4

G

Startup strategies, degradation

[32,42
48]

Stack and vehicle

4

G

Startup strategies

[30,49,50]

Subzero startup of polymer electrolyte fuel cell Chapter | 16

471

FIGURE 16.1 Typical test procedures for water fill test, freeze/thaw, and subzero startup.

16.2.4 Characterization techniques Various characterization techniques are employed and developed to explore the subzero startup process. Some of them have played a crucial role in uncovering or confirming new phenomena, gaining insight, and resolving controversy. For example, Saito et al. [20] discovered the three states of water in the PFSA membrane at subzero, laying the foundation for subzero studies. Ishikawa et al. [26] observed the supercooled water during the subzero startup with an optical microscope in 2008, thereby fundamentally changing the perception and modeling assumption of the subzero startup. The research group at Paul Scherrer Institute [28,39] used neutron radiography and multiple single cells to investigate the dispersion in the WFT, demonstrating the high randomness in the releasing of supercooling state. Wang et al. [29] combined the use of serial CV and dynamic EIS in the whole subzero startup process to reveal that the failure is caused by the blocking of pores in the CL, rather than the covering of the catalyst surface, by the frozen water. Table 16.4 lists the featured characterization techniques grouped into three types: physical, chemical, and electrochemical.

16.3 Damage and mitigation in subzero scenarios Unlike the ICE, the PEFC does not have harmful emissions like NOx, SOx, COx, or particulate matter. The product water of the PEFC is environmentally friendly; hence, the FCEV constitutes a major type of the ZEV.

TABLE 16.3 Methods for temperature control in subzero study. Devices

T

Cooling rate and time

Cost

Features and suitable objects

Key references

Environmental SEM

B 2 10 C

Quick (several minutes for 25 C)

Expensive

Control both T and RH, applicable to components

[43,52]

Peltier element

B 2 20 C

Quick (several minutes for 25 C)

Cheap

Limited to a single cell with small surface areas (,B40 cm2) Quick response if only mild subzero to effect freeze/ thaw cycling is needed

[39,53]

Environment chamber

B 2 30 C

Slow (B5 h for 220 C)

Expensive

Single cell, suitable for disassembling and observing with cyro-SEM

[22,23,36]

Recirculating coolant

B 2 20 C

Quick (B1 h for 220 C)

Expensive

Short stack

[38,54]

Subzero startup of polymer electrolyte fuel cell Chapter | 16

473

TABLE 16.4 Featured characterization techniques in subzero study. Type

Techniques

Main information derived

Key references

Physical

Differential scanning calorimetry

States of water in PFSA from the heat flux during thermal scanning

[19,55,56]

Dynamic mechanical analysis

Effects of water freezing on the mechanical properties of PFSA membrane

[57]

Optical microscope

Water/supercooled water/ice position and morphology (Bmm)

[26,36,37]

Neutron radiography

Water/supercooled water position (B100 μm) Distinguishing liquid water and ice

[38,39,58]

X-ray radiography and X-ray CT

Water/supercooled water/ice position (B10 μm)

[59,60]

Cryo-SEM

Ice position and morphology (Bμm)

[22,23]

Infrared thermograph

Temperature and its distribution

[22,26]

Microthermocouple

Local temperature difference between channel and rib, temperature jump near the freezing site

[33,59]

ICP, 19F-NMR

Pt ions Fluoride ions and sulfate ions in exhaust

[43,61]

CO2 sensor

Carbon corrosion during startup

[61]

Printed circuit board

Current distribution

[32]

Electrochemical impedance spectroscopy

Ohmic, activation, concentration resistance

[29,54,62]

Cyclic voltammetry

Electrochemical active area

[29,63]

Chemical

Electrochemical

474

Fuel Cells for Transportation

However, the environmentally benign product water can turn out to be destructive to the device itself at the subzero temperatures. The freezing of water not only stops the function of the FC but may also cause irreversible damage to the cell components due to the 9% expansion in volume. Therefore, much effort in the early stage of subzero study has been spent on the damage and mitigation of FCs in subzero scenarios. Depending on whether there is water produced in the process, the subzero scenarios can be divided into two categories. One is the freeze/thaw cycling in its narrow sense, where there is only residual water in the FC or its components; the other is a repetitive subzero startup where the current is drawn and water is produced at subzero. Due to the difficulty in thawing itself solely from the operation of FC (so-called unassisted start in Section 16.6), the majority of early studies fall into the first category, where the FC or component temperature is brought up above zero by external heat sources (so-called assisted start in Section 16.6). The freeze/thaw test examines the effects of residual water in the FC or its components (PEM, CL, and GDL) on the performance and structure integrity under temperature cycling between subzero and normal working temperatures. The water is introduced into the test objects by impregnating or operating the FC and generating the product water before cooling down. The amount of water can be adjusted by the difference in water level or vessel pressure during impregnation or by varying the operation time/current and subsequent purging. Freeze/thaw cycling takes a long time, expends a great amount of energy, and causes irreversible damage to the test pieces. Yet the results are difficult to compare as there are so many variables but no unified protocol. In particular, the way to control the residual water inside the test objects varies from test to test. In what follows, the major types of damage to the components will be first introduced. Then, the effects of design and operation parameters for mitigating these damages will be described. Finally, the cause of and the countermeasure to the damages will be discussed.

16.3.1 Damages to fuel cell components from freeze/thaw and subzero startup The observation of the damages and degradation in subzero cycling tests is somewhat controversial. While some studies find no measurable degradation, others report significant damages. Such discrepancy may be associated with the different component properties, MEA fabrication, cell design, cell assemblies, temperature control, test protocols, etc., which could impact the residual water and mechanical stress distributions inside the cell. For example, GDL having different loadings of PTFE may trigger the release of supercooling state at a different rate, decal transfer process using ionomer allowing for higher temperature (e.g., tributyl ammonium [TBA], 210 C) may result in a

Subzero startup of polymer electrolyte fuel cell Chapter | 16

475

more durable catalyst-coated membrane (CCM), CCM supported by GDLs in a cell is more resistant to delamination than that standing alone, the cooling rate is higher for the cell controlled by coolant than that by the environment chamber, and freezing of residual water is expected to be more gradual in freeze/thaw compared to that of the product water during subzero startup. Nevertheless, most of the subzero tests have demonstrated severe and irreversible damage to FCs if the residual or product water freezes. Table 16.5 summarizes the representative damages to the major components in FCs, for example, PEM, CL, microporous layer (MPL), GDL, and gasket. More detailed information is described in the following. In 2004, McDonald et al. [12] conducted an ex situ freeze/thaw study on NafionTM 112 and MEA assembled in the FC stack under ambient humidity conditions (low water contents of about 1.6%
3.4% wt). After 385 freeze/ thaw thermal cycling between 240 C and 80 C, they observed no catastrophic failures in the membranes and MEA. Lee et al. [65] observed no change to GDL properties after 50 freeze/ thaw cycles between 235 C and 20 C. This contradicts the results of Y. Lee et al. [66], who examined the degradation of 4 types of GDL under 10 repetitive freeze/thaw cycles between 220 C and room temperature. The porosity, bulk density, and contact angle were all found to decrease appreciably. The cracks on the MPL increased in width and length. Dislodge of PTFE from GDL and detachment of binder from carbon fibers were observed. The difference in the observation may have a number of origins, including GDL material, water content, cooling rate, and whether there is compression. In Lee’s work, the GDL was saturated by submerging in an 80 C deionized water bath, and the water content was estimated to be 12.3
14.0 mg cm22. The GDL was compressed by the flow field plate simulating the situation in a stack and placed into an environmental chamber with a temperature cycling rate of 0.3 C, starting from 20 C. In Y. Lee’s work, the water in GDLs was impregnated through the water level difference of the two containers connecting to the GDL (the exact amount of water content is not given), and the GDLs were placed in the environmental chamber whose temperature was kept below 220 C (hence higher thermal shock), and without compression (hence more severe expansion and contraction). Ko et al. [60] studied the effects of 40 freeze/thaw cycles on the water distribution in the porous media of the FC. The cell was operated at 65 C under H2 and air (both humidified at 50% RH) following a series of stepwise increases in current density till a maximum value of 1.5 A cm22. The cell was then directly cooled down in a freezer maintained at 220 C for at least 3 hours. After that, the cell was heated up to 65 C and repeated the cycle for 40 times. Water saturation in the operating cell was measured using synchrotron X-ray radiography. It was found the water saturation level in the CCM and GDL, as well as the saturation jump across the interface between the MPL and backing layer, decreased with the cycle number. While contact

TABLE 16.5 Representative damages to the components and structures in fuel cell. Components

Damages

MEA

G

PEM

G

CL

G G G

GDL

G G G G

Images

Pinhole Micro-cracking

Pinhole damage of membrane

Micro-cracking of membrane

Micro-cavities in CL

Growth of cracks on CL surface [44]

Growth of micro-cracks in MPL

Detachment of binder

Micro-cavities (void) Growth of cracks Delamination of CL from GDL

Growth of micro-cracks Detachment of binder from carbon fiber Delamination of MPL GDL opening to the channel becomes hydrophilic

Gasket

G G G G

Increase of hardness Decrease of weight Pinhole development Drop of the sealing performance

Pinholes on silicon rubber after freeze/thaw cycling [64] (u16-01-9780323994859 and u16-02-9780323994859) Adapted from Q. Yan, H. Toghiani, Y.W. Lee, K. Liang, H. Causey, Effect of sub-freezing temperatures on a PEM fuel cell performance, startup and fuel cell components, J. Power Sources 160(2) (2006) 1242
1250. (u16-03-9780323994859) Adapted from R. Alink, D. Gerteisen, M. Oszcipok, Degradation effects in polymer electrolyte membrane fuel cell stacks by sub-zero operation—an in situ and ex situ analysis, J. Power Sources 182(1) 175
187. (u16-04-9780323994859) Adapted from M. Oszcipok, M. Zedda, D. Riemann, D. Geckeler, Low temperature operation and influence parameters on the cold start ability of portable PEMFCs, J. Power Sources 154(2) (2006) 404
411. (u16-05-9780323994859 and u16-06-9780323994859) Adapted from Y. Lee, B. Kim, Y. Kim, X. Li, Degradation of gas diffusion layers through repetitive freezing. Appl. Energy 88(12) (2011) 5111
5119. (u16-07-9780323994859) Adapted from F. Wu, B. Chen, Y. Yan, Y. Chen, M. Pan, Degradation of silicone rubbers as sealing materials for proton exchange membrane fuel cells under temperature cycling, Polymers 10(5) (2018) 522.

478

Fuel Cells for Transportation

angles on both sides of the GDL showed no change, cracks on the MPL were observed on the millimeter-scale images. Such cracks were thought to facilitate water drainage from the CCM to GDL, thus reducing the water saturation in the CCM and leading to high ohmic overpotential. Oszcipok et al. [67] performed 10 isothermal, potentiostatic single-cell startups from 210 C. The flow field channel pattern on the backing layer and the MPL of the cathode GDL was visible from the water spray test in the dissembled cell. Contact angles in these areas showed decreased value, while that at the rib or on the anode side remained the same. CV, EIS, and reference performance tests at V 5 450 mV showed a decrease in the ECSA, an increase in high-frequency resistance (HFR), and a decrease in power with the number of startups, respectively. Hwang et al. [52] compared the durability of a dispersed Pt/C CCM and a nanostructure thin film (NSTF) CCM under wet/dry and freeze/thaw cycles, and isothermal startup. Multiple isothermal subzero startups resulted in a performance degradation for the dispersed CCM, while no such degradation was found in the NSTF. Stand-alone CCM tests in the ESEM (20 cycles along the three types of phase change near the triple point of water) confirmed a similar trend. Whereas the dispersed CL had an exponential increase in the number and size of cracks until it delaminated from the membrane, the NSTF CL showed minimal crack generation and no delamination possibly because the ice formed on top of the layer. Among the three types of phase changes, freeze/thaw between ice and liquid had the most serious impact. The CL coated on PTFE rather than on NafionTM delayed the onset of crack formation. Yan et al. [68] studied the effect of subzero temperatures on FC performance and cold start behavior using a 25-cm2 FC. Significant damage to the MEA was observed if the cell cathode temperature fell below 25 C. The FC could start from 25 C and experience no irreversible performance loss if the cell was initially dry. Freezing of the water generated in the FC operation damaged FC internal components. SEM images of failed MEAs at 215 C and 220 C demonstrated CL delamination from both the membrane and the GDL, some minor damage to the coating Teflon and binder structure in the backing layer, and roughened surface, micro-cracks and pinholes on the surface of the membrane. Alink et al. [43] conducted a comprehensive examination of the influence of 120 freeze/thaw cycles in a dry state and 62 freeze/thaw cycles in a wet state, and a series of 9 isothermal subzero startup experiments, on cell components and cell performance of a 6-cell stack. It was argued the step-like changes in the HFR near subzero are caused by the draining out of the water in the membrane into the electrode, rather than by the increase in contact resistance. For the stack dried before freezing, repeated subzero exposure led only to slight performance degradation, mainly at high current densities,

Subzero startup of polymer electrolyte fuel cell Chapter | 16

479

while the stack which was frozen in the wet state showed serious degradation in water management. SEM images demonstrated micro-cavities on the surface and increased pore size in the cathode CL. About 10 ex situ freeze/thaw cycles between 220 C and 0.5 C of the CCM in the ESEM found distinct damage on the surface, more severe than that of the in situ wet stack (62 freeze/thaw cycles and 9 subzero startups), which was attributed to the lack of compression (protection) from GDLs as is the case in a stack. In contrast, 10 wet/dry cycles between atmosphere pressure of 10 and 700 Pa at a temperature of 0.5 C, that is, liquid/vapor phase change without freezing below 0 C, found essentially no degradation. In addition to the degradation of the MEA, which is the core component in FCs responsible for the mass transport, electric conduction, and electrochemical reaction, the reliability and stability of peripheral components, for example, sealing, at subzero cycling also need to be examined. Wu et al. [69] investigated the degradation of methyl vinyl silicone rubbers submerged in solution simulating the actual PEFC environment (pH 5 3.35, 12 ppm H2SO4, 1.8 ppm HF) under 200 cycles between three low temperatures of 25 C, 210 C, and 220 C, and the same high temperature of 90 C. It was found that the loss in weight, the decrease in hardness, and the formation of surface holes were exacerbated at lower temperatures and a greater number of cycles. These physical damage was corroborated by the detection of increased calcium dissolution (from the filler material calcium carbonate in the rubber) and reduced adsorption peak in ATR-FTIR spectra. As a result, a higher compression ratio would be needed to prevent gas leakage from the degraded seals. Besides the physical damages from the freezing of residual or product water, chemical degradation from low humidity operation may also be a concern. It was reported that low humidity operation accelerated the decomposition of PFSA [70], possibly due to high O2 and H2 partial pressure (high crossover, hence the high rate of H2O2 generation) and the low water content in ionomer (weak C-S bond at the non-dissociated state of the sulfonic end group in the side chain) [71]. Since water vapor pressure is nearly zero at subzero, and the FC is usually purged before shutdown to lower the water content in the ionomer, the subzero scenario may accelerate chemical degradation starting from the sulfonic end group. Such concern was confirmed by Onoda [72], where greater degradation in the performance of the FC and the decomposition of the PFSA membrane was reported for the cell with longer purge time (higher HFR and lower water content in the PEM) after repeated cold starts from 230 C. Liphardt et al. [48] tested the degradation of three 20-cell stacks undergoing 100 freeze startups with different amounts of residual water. The stack with the highest initial water content was found to have the least loss in performance at a nominal temperature in low, intermediate, and high current densities.

480

Fuel Cells for Transportation

16.3.2 Mitigation effects of material and design Kim and Mench systematically examined the effects of microstructure [14] and the diffusion media [15] on the physical damages of PEFC materials subjected to thermal cycling with freezing (240 C/70 C) and without freezing (5 C/70 C). For the freeze/thaw cycle test, damages occurred almost exclusively under the channel, where the compression pressure is low and hence prone to the formation of frost heave, while that below the land is negligible, even though more water tends to reside under the land. The CL with initial cracks showed severe disruption in morphology due to the pooling of water in the cracks with a typical gap of 10
18 μm, where capillary pressure is small and the freezing point depression is negligible. For the thermal cycling without freezing (5 C/70 C), damages were negligible under the channel. However, the CL in cells without the protection of DM/MPL showed line cracks at the edge of the land, while the CL with DM/MPL protection did not show such line cracks. Stiffer diffusion media, enabling more uniform compression under the channels and lands, were found to be capable of mitigating surface cracks. In conclusion, the authors recommend a freeze-tolerable design to be a crackfree virgin CL on a thin/reinforced membrane, protected by a stiff diffusion media, and mounted in cells with a narrower channel. However, the authors also cautioned that the freeze-tolerable design alone cannot completely prevent the diffusion media/CL delamination. Liquid water in the CL should be removed before being stored under subzero to facilitate subsequent startup and permit a large number of cycling. Han et al. [73] investigated the effects of the GDL’s anisotropic bending stiffness on the degradation behavior of a 5-cell stack under 1000 temperature cycles between 210 C and 1 C. The I-V performances of the stack with a perpendicular alignment of the GDL machine direction to the major flow field direction of the BPP were higher than those of the parallel alignment both before and after the cycles. The higher performance and durability for the perpendicular alignment were attributed to the more uniform compression and intimate contact between component interfaces. In order to alleviate the chemical degradation of PFSA under low RH operation, Endoh et al. [16] doped Ce31 in the PFSA membrane and found that ion exchanging the 5 mol% of total SO3H group of the membrane by Ce31 ions was able to stop the detection of fluoride emission in the exhaust. The degradation rate of the peak power was much reduced during the 50 cycles of subzero startups, although the membrane resistance increased and the peak power decreased compared with the non-doped NafionTM membrane.

16.3.3 Mitigation effects of operation control As the volume expansion of the residual water inside the FC is the major cause of the degradation in the freeze/thaw test, it is natural that removing the

Subzero startup of polymer electrolyte fuel cell Chapter | 16

481

residual water, or more precisely the freezable residual water, from the FC through purge during the shutdown will be effective for mitigating the damage. Cho et al. [74] studied the freeze/thaw cycling between 210 C and  80 C. They reported that the unpurged initial condition resulted in a severe increase in activation and ohmic resistances, which was attributed to the deformation of the electrode structure and the increase in the contact resistance between the membrane and the electrode. Both phenomena were caused by the volume expansion from the freezing of the water that was produced during operation and remained in the cell afterward. Cho et al. [13] further found that supplying dry gases to remove the water or supplying antifreeze solution into the cell after each operation can effectively suppress the degradation. Hou et al. demonstrated that with appropriate gas purging, no performance degradation and MEA delamination in a PEFC occurred after 20 freeze/thaw cycles from 220 C to 60 C [75]. Song et al. [11] studied the effect of air purging and dry operation on the durability of a PEFC under 20 freeze/thaw cycles between 220 C and 260 C. Three operation conditions and purging logics, corresponding to different levels of residual water inside the cell during the subzero storage, were tested on three cells with commercial CCM with a size of 26 cm2. It was found that cathode air purging and operation with dry air feed before freezing in the environment chamber were effective to mitigate freeze damage. Omitting the air purging to the anode side of the three cells did not cause any damage to the anode CL, either in morphology or ECSA. This finding is important in the field implementation of purging. It not only saves time and energy in purging the anode with air but also avoids the formation of H2
air front in the anode, which could induce severe carbon corrosion. Unexpectedly, a Pt band was observed in all three cells, even though the freeze/thaw cycles did not contain much cycling in potential, which is the major stressor promoting Pt dissolution. It is worth noting that another study in subzero operation [76] also observed the Pt band in the PEM. The Pt band formation is generally associated with Pt dissolution at the cathode CL, Pt ion diffusion from the cathode CL into the membrane, and Pt ion reduction inside the membrane by the H2 crossover from the anode. The dissolution and diffusion of Pt ions are accelerated by potential cycling under wet conditions, which is not particularly a feature in the freeze/thaw test. Therefore, it is suspected that the Pt band observed in these two studies may result from the initial quick dissolution of Pt during the break-in and characterization at temperatures above zero. The catalysts used may also play a role, as carbon support without heat treatment and Pt particles with wide distribution in size are all prone to dissolution and degradation. Bradean et al. [77] investigated a new FC stack concept for achieving appropriate MEA water content for freeze startup. The idea is to maximize the water migration from the MEA to the flow channels due to temperature

482

Fuel Cells for Transportation

gradients during natural cooling after shutdown from a wet operation. Using a stack insulated everywhere except at one end and a heat reservoir at the other end of the stack, a significant amount of water migrates from the MEA to the flow channels after shutdown in each cell of the stack, due to the temperaturedriven water transport to be introduced in Section 16.4. Successful startup of a stack from
12 C was achieved without purge operation after shutdown. The startup capability can be further increased if a short purge is used to remove water from the flow channels just before freezing. Onoda et al. [72] studied the effect of two levels of residual water contents in the MEA on performance degradation from 50 repeated subzero startups from 220 C and 230 C to 70 C. The residual water was adjusted by purging the cell (25 cm2) with dry N2 after operation at 70 C till the HFR reaches 10 or 100 mΩ. The subzero start at a potentiostatic hold of 0.5 V was assisted by an electric heater at a rate of 1.7 C min21. It was found the drier MEA (100 mΩ HFR) after the purge had a higher performance degradation rate. Performance loss breakdown showed that the major difference came from the higher increase in the concentration overpotential. Pore size distribution showed a substantial increase in the range below 0.1 μm, more notable for the drier MEA. SEM images indicated a crushing of the pores in the cathode CL near the membrane side. Fitting of the EIS results with the transmission line model showed a 60% increase in cathode ionomer ionic resistance, while the resistance for the membrane and carbon remained the same. Fluoride and sulfonic ions were detected at both the cathode and anode sides. CO2 sensor detected a visible peak just after the rising of OCV (supply of air), even at 230 C under N2 atmosphere, possibly due to nonuniformity in the cell such as the humidification level across the plane. Distributed current density measurement with segmented cells demonstrated that the area with the highest degradation rate occurred near the outlet for 220 C, but near the inlet for 230 C. Such results were corroborated by the degree of ionomer degradation in CL as characterized via 19F-NMR. The authors speculated that the dry operation of the cell at subzero caused the ionomer degradation. Comments and speculation from the authors of this chapter are in order here. The extensive characterization in Onoda et al.’s work reveals the complexity of degradation behavior that could be encountered in repeated subzero startup tests. There are multiple degradation modes that can interact or compete with each other. While some of the observation remains unresolved, one message is clear: there exists a suitable range of water content of MEA for reliable, effective, and durable subzero startup. While too much residual water may undermine the startup performance and cause irreversible damage from physical volume change, too less residual water may expose the PFSA to the chemical degradation mode under low RH operation. In addition, too dry from purging may induce wet/dry stress for the MEA, where the ionomer in the CL could be the weakest component. One perplexing observation in their work is the crushing of the pore structure in the cathode CL near the

Subzero startup of polymer electrolyte fuel cell Chapter | 16

483

membrane side rather than the MPL side where product water tends to accumulate. It may be explained by the strengthening of the local CL porous structure by the frozen water. At subfreezing temperatures as low as 220 C or 230 C, the product water freezes almost immediately after it exudes from the ionomer into the pores. The whole freezing process is gradual so that the frozen water strengthens the pore structure nearby rather than destroys it. The increased pressure from this freezing causes the weaker part of the CL, the part near the membrane where the product water can move into the membrane, to collapse. For the startup from temperature . 2 10 C, where a massive amount of liquid water may flow into GDL and channel, the freezing of the supercooled water tends to be more abrupt; thus, the pores with more water may suffer more severe damage. The use of assisted-start strategies, such as heating the coolant using an external energy source and using the coolant to warm up the stack, effectively decouples the water and thermal management. The FC starts to work only after its temperature is above 0 C, thus eliminating the concerns about the freezing of product water. However, heating the coolant first and recirculating it to warm up the stack, though effective in avoiding freezing damages, are too consuming in both time and energy. In contrast, assisted start using an alternating hydrogen pump (AHP) [33,34] can start up a short stack made of graphite carbon BPP from 230 C efficiently in both time and energy, because ohmic heat from the membrane is utilized. As there is no product water involved, the AHP startup is expected to be both reliable and durable. Unassisted-start strategies, where the stack is warmed up by the heat released from the FC reaction, involves tight competition between water and thermal management to avoid the freezing of product water. Maximizing both the heat generation rate to overcome the heat loss and the amount of heat generation per unit of product water to fully utilize the water uptake potential of the membrane is paramount for startup performance and damage mitigation. Toyota restricts the air supply to maximize the waste heat due to concentration overpotential, thereby greatly enhancing the ratio between heat generation and product water. In the meantime, the water content of the membrane before shutdown is controlled in an appropriate range, neither too wet to allow for adequate water storage capacity nor too dry to allow for adequate current flow to ensure a high heat generation rate. Such a strategy is already implemented in the commercial FCEV MIRAI [78].

16.3.4 Discussion on the cause and mitigation of damages in subzero Subzero scenarios, including freeze/thaw and subzero startup, involve both phase change and thermomechanical cycles. The damage observed in most of the subzero cycles is the mixed outcome of these two stresses. Even though thermal cycling between subzero as low as 230 C to 210 C

484

Fuel Cells for Transportation

(without freezable residual water) and normal operating temperature as high as 60 C
90 C also cause considerable mechanical stress due to the difference in the thermal expansion coefficients of the components, the dominant cause of damages comes from the significant volume expansion of residual or product water accompanying the freezing. Compared with the high porosity of CL, MPL, and macroporous substrate (MPS) which falls in the range of 60%
80%, the volume expansion ratio of 9% from the freezing of water may seem insignificant. Such an impression may be valid if the phase change is a uniform and gradual process. In actuality, there is high nonuniformity in the distribution of water content, and the freezing can be highly transient. The composite, thin-layer porous media has distributed hydrophilicity and pore sizes, and may contain cracks/defects. The surface of both CL and MPL is rough with a roughness comparable to their thickness; therefore, the interface contains many gaps. Water tends to accumulate in the less hydrophobic site, in cracks, defects, and gaps. Water may exist in a supercooled state for a prolonged time. The release of the supercooling state is highly spontaneous, which is followed by rapid ice growth. The ice inside the pores or the cracks/gaps may draw more water toward it, forming an ice lens (frost heave) thereby causing structure damage, loss of PTFE coating, and delamination if the local compression force is inadequate. Low water content and low humidity, characteristics of the subzero condition, can cause chemical degradation to PFSA. Fluroride emission rate in the exhaust was found to be higher for cells with higher initial HFR (less water content). The addition of Ce31 into the membrane was found to be effective in mitigating such a mode of degradation. In stack, the additional complication comes from the inevitable nonuniformity and distribution. The high thermal mass of the endplates induced a large temperature gradient among the cells, and the heat loss through the periphery causes a temperature gradient across the cell. Restricted air supply leads to strong distribution in current, product water, and temperature from inlet to outlet, which may cause degradation. Due to the overwhelming damages from freeze/thaw, the stresses from nonuniformity and distribution are masked and have not received adequate attention. The degradation in subzero scenarios causes the increase in concentration overpotential, as well as activation and ohmic overpotential. The major damages from freezing are the destruction of porous media structures, such as crack growth in the CL and MPL, delamination of the CL from the MPL and PEM, loss of PTFE from the GDL (loss in the hydrophobicity, decrease in contact angle), holes on the gasket. As a result, the mass transport resistance increases, and the cell becomes prone to flooding. A significant loss in the electrochemical active area, hence an increase in activation overpotential, is also observed. Some studies even observed Pt band inside the membrane after the freeze/thaw test. Low humidity operation accelerates the decomposition of ionomer, increasing the ohmic resistance.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

485

Freeze-tolerable design, though helpful in increasing the tolerance to volume expansion, cannot avoid damages and usually incurs other restrictions, for example, the increase in cost, and the tradeoff in performance for nominal operation. Therefore, preventing freezing from operation control in the first place is always the better strategy. That is, the pores of the CL/MPL/ GDL should be free of residual water. The residual water in PEM should be less than the saturation water content to avoid freezing. Since the saturation water content of PEM decreases with temperature, the residual water content in PEM should be less at a lower temperature. For different membranes and the design of FCs, the maximum allowable residual water inside the membrane needs to be determined individually. Purging the cell so that only nonfreezable water is left in the membrane, combined with the startup technique such as air supply restriction, has been proven to be effective for a successful subzero startup. Early works have demonstrated the severe damages from the freezing of the water through aggressive freeze/thaw cycling. FCEVs with unassisted subzero startup capability has been commercialized. Future studies should shift the focus to identifying the major degradation modes from the successful subzero startup, and develop a target for the degradation rate of each subzero startup. For that purpose, repeated successful unassisted-start tests, rather than freeze/thaw cycles, become relevant. Degradation from water freezing needs to be kept to a minimum, so that other degradation modes, such as wet/dry cycles, low RH and low water content operation, carbon corrosion, and current/water/temperature distribution across and among the cells, become unmasked.

16.4 States and behavior of water at subzero The severe damage caused by the frozen water and the difficulty in starting up the FC from subzero without external heating while meeting DOE’s target [1] have made the subzero start a hot topic in both academia and automotive companies ever since the early 2000s. The nature of subzero start is the competition of water and thermal management at low temperatures. In this section, the states and phase change of water in free/confined space and membrane/ionomer, the water transport and icing dynamics, and the contribution and limitation of the WFT at subzero will be covered. In the next section, thermal management-related issues will be addressed.

16.4.1 States of water in fuel cell at subzero 16.4.1.1 Water in free space: states and thermodynamic properties The melting point of water at the pressure of 1 atm (relevant to the subzero storage and operation of PEFC) is 0 C. It means that the residual or product

486

Fuel Cells for Transportation

FIGURE 16.2 Density of ice, water, and supercooled water, extracted from [79].

water tends to freeze at ambient temperature below zero. The densities of water and ice at 0 C are 999.8 and 916.8 kg m23, respectively, as shown in Fig. 16.2. The 9% volume expansion accompanying the freezing, if occurred massively and suddenly, can cause severe damage to the component and structure of PEFC, leading to irreversible degradation in performance. However, when freezing occurs and how it proceeds are a kinetic phenomena depending on a multitude of parameters, for example, the degree of supercooling, the presence of nuclei, or interference. If the bulk is free of impurities and the surface is free of defects, water may remain in the liquid phase until the temperature drops to 248.3 C at standard pressure (1 bar), which is the crystal homogeneous nucleation temperature [80]. Water remaining at a liquid state below its melting point is termed supercooled water. Changing from the liquid state in supercooling to the thermodynamically stable solid state is called the release of the supercooling state. Though thermodynamically spontaneous, the release of the supercooling state is kinetically stochastic, hence highly random and unpredictable. The formula for the saturation vapor pressures of ice and supercooled water can be deduced from the Claperyron equation using the triple point properties and the molar heat capacity. The vapor pressure of both ice and liquid water at the triple point is 611.66 6 0.01 Pa at a temperature of 273.16K [81]. The latent heat of vaporization at the triple point is 51.059 kJ mol21 [82]. The molar heat capacity of ice can be calculated with Eq. (16.4) [82]. "
2 # T cp;ice 5 2 2:0572 1 0:14644T 1 0:06163T exp 2 ; T . 20K 125:1 ð16:4Þ The density and molar heat capacity of supercooled water over a range of temperatures are shown in Fig. 16.3.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

487

FIGURE 16.3 Molar heat capacity of supercooled water and ice as a function of temperature, extracted from [83,84].

FIGURE 16.4 Saturation vapor pressure of ice, supercooled water, and liquid water (above zero).

Using these values and the Clausius
Claperyron equation, the formula for the saturation vapor pressures of the ice and the supercooled water are deduced to be Eqs. (16.5) and (16.6), respectively, [82]. Eq. (16.7) [82] gives the saturation vapor pressure for liquid water above 0 C. The saturation vapor pressures of ice and water are shown in Fig. 16.4. The saturation vapor pressures at subzero are negligibly small, compared with that above zero. Supercooled water has a slightly greater vapor pressure than ice, since it is in a metastable state. That is, it has a greater Gibbs energy than ice.
 5723:265 1 3:53068 ln ðT Þ 2 0:00728332T ; T . 110K pice 5 exp 9:550426 2 T ð16:5Þ

488

Fuel Cells for Transportation

psup  pice

210368 1 131:438T 2 3:32373 3 exp 2 RT

105 T

2

41729:1 ln T RT

! ;

123K , T , 332K  pliq 5 611:21 exp

17:502ðT 2 273:15Þ 240:97 1 T 2 273:15



ð16:6Þ ð16:7Þ

The surface tension of supercooled water increases linearly with the decrease in temperature, which is 75.5 mN m21 at 0 C and 78.2 mN m21 at 220 C [83].

16.4.1.2 Water in porous media: melting point depression Water in confined spaces, like the porous media in FCs, has a depression of melting point. For an isolated spherical crystal of radius r in its own liquid, the Gibbs
Thomson equation for the melting point depression can be written as Eq. (16.8) [85]. ΔT 5 Tm0 2 Tm ðr Þ 5

2σsl Tm0 ρs Lr

ð16:8Þ

where Tm0 is the melting temperature (K) of the bulk, σsl is the solid
liquid interface energy (J cm22), L is the specific latent heat in the freezing and melting process (J g21), ρs is the density of ice (g cm23), and r is the radius of the pore (cm). Using the corresponding thermodynamic values for water, the dependence of melting point depression on the pore size in PEM/CL/MPL/MPS of FC is shown in Fig. 16.5. The water cluster inside PEM is about 2 nm [86],

FIGURE 16.5 Dependence of melting point depression on the pore size (σsl 5 3:17 3 1026 J cm22 [87], Tm0 5 273:15K, L 5 288:4 J g21 [88], ρs 5 0:9168 g cm23).

Subzero startup of polymer electrolyte fuel cell Chapter | 16

489

corresponding to a melting point depression of 65 C. The typical pore sizes in CL are B4 and B30 nm, corresponding to a melting point depression of B32 C and B4 C, respectively. The typical pore sizes in MPL and MPS are B100 and B1000 nm, corresponding to a melting point depression of B1 C and B0.1 C, respectively. It is clear that the melting point depression for the pores in MPL/MPS is negligibly small. Only in the pores in carbon support or ionomer/membrane can water exist stably in a liquid state at a temperature below 220 C.

16.4.1.3 Water in membrane/ionomer The PFSA-type of PEM consists of a hydrophobic main chain, which is the backbone to maintain its mechanical stability, and hydrophilic side chains ending with sulfonic acid to facilitate water adsorption and proton conduction. Equilibrium water content in PEM depends on the state and activity of the surrounding water. The structure and properties of the membrane evolve with the water content. Clusters of water form preferably around sulfonic end groups when the water content is relatively low (λB3) [89]. The membrane separates into hydrophobic and hydrophilic regions, which is called micro-phase separation, and the volume of the hydrophilic region increases as the water content increases. As more water is absorbed, the clusters grow in size and finally interconnect at some threshold value of water content (λB4:8) [8], which is called percolation when the membrane starts to conduct proton through a vehicular and hopping mechanism. Further increase in water content will lead to an increase in proton conductivity. The newly absorbed water behaves like bulk water in a confined space. The typical size of the water cluster in the membrane is about 1 B 3 nm for 2 , λ , 15 [86]. Three states of water in the perfluoro sulfonic acid membrane Differential scanning calorimetry (DSC) characterizations have shown the existence of three states of water inside the PFSA membrane [18,20,90]. Those strongly bonded to the sulfonic acid groups do not freeze at temperatures as low as 240 C [35]. The loosely bonded and bulk-like water has a melting point depression due to the colligative effects of the anionic side chains and nanosized curvature-induced Gibbs
Thompson effect. The existence of non-frozen water in PFSA at subzero temperature provides the capacity for water storage and enables proton conductivity at subzero, thereby laying the foundation for the possibility of a subzero startup. Saturation water content The water vapor sorption isotherm is a relationship between the concentration and the water activity in the PFSA ionomers. It is the most important phenomenon of ionomers which affects their phase-separated structure and transport properties. The water content λ, defined as the number of water molecules per mole of sulfonic acid groups, depends on a balance between the mechanical

490

Fuel Cells for Transportation

strength of the backbone of the membrane and the water affinity of the side chain. The dependence of equilibrium water content on the water activity is determined experimentally at normal temperature, as shown in Eq. (16.9) [91]. a is the water activity defined in Eq. (16.10) [92]. psat is the saturation vapor pressure of liquid water, pvap is the vapor partial pressure, and sl is the liquid water volume fraction in the pore regions of the CL.  0:043 1 17:81a 2 39:85a2 1 36:0a3 if 0 # a # 1 λequil 5 ð16:9Þ 14:0 1 1:4 ða 2 1Þ if 1 , a # 3 a5

pvap 1 2sl psat

ð16:10Þ

At subzero, two thermal effects influence the water sorption isotherm, as pointed out by Gallagher et al. in [21]. The first is due to the exothermic nature of the sorption, making the membrane absorb more water at constant activity as the system temperature decreases. The second is due to the lower vapor pressure of ice than that of the liquid water at the same temperature, leading to a maximum water content decreasing with temperature. For example, the maximum water content in equilibrium with water vapor is 14 at 0 C and decreases to around 8 at
20 C. These two effects can be recognized in Fig. 16.6. A correlation between saturation water content and temperature is given as Eq. (16.11) [92]. 8 if T , 223:15 K < 4:837 if 223:15 K # T , TN λsat 5 2 1:304 1 0:1479T 2 3:594 3 1025 T 2 : . λnf if T $ TN ð16:11Þ

FIGURE 16.6 Measured water uptake at multiple temperatures as a function of the water activity. po is the partial pressure of water, and po;l is the saturation pressure of liquid water at the system temperature. The line shows the measured isotherm at 30 C from Springer et al. [91] for comparison.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

491

where λsat and λnf are the saturation (maximum non-frozen water content) and non-frozen water content, and T and TN are the local temperature and the normal freezing temperature of water (273.15K), respectively. The nonfrozen membrane water content λnf in the ionomer is defined as Eq. (16.12). λnf 5

EW cnf ρmem

ð16:12Þ

where EW is the equivalent weight (the number of grams of dry polymer per mole of the anionic group), ρmem is the membrane density, and cnf is the non-frozen water concentration in the membrane. Vapor pressure over NafionTM of different water content at subzero can be found in [91].

16.4.2 Water transport dynamics in fuel cell at subzero While water management is critical for the efficient, reliable, and durable operation of the PEFC at normal temperatures, it becomes vital for the subzero startup. Therefore, it is essential to understand the water production and transport dynamics in the ionomer/PEM and porous media, for example, the CL, MPL, and MPS in FCs.

16.4.2.1 Overview of water production and transport During the operation of the FC, protons migrate from the anode to the cathode through the membrane while electrons move through the external load to do electrical work. Several water molecules are dragged with each proton from the anode to the cathode through the PEM (electro-osmotic drag). The states, transport, and phase changes of water in the ionomer/PEM and CL at subzero are shown in Fig. 16.7. In the cathode CL (cCL), water is produced from the electrochemical reaction of proton and O2 on the surface of the Pt catalyst that is covered by ionomer or liquid water. The water inside the PEM and ionomers in the CL will not freeze when the water content is below a certain level. Water in the ionomer of cCL will diffuse toward PEM and anode CL due to the water content gradient in the ionomer (back-diffusion). Water inside the ionomer may evaporate from the surface, yet its fraction is small due to the lower vapor pressure. The water content at the cCL ionomer surface will increase if the water production rate plus the dragged water is higher than the back-diffusion rate. If the water content at the cCL ionomer surface exceeds its maximum level, liquid water will exude into the micropores in the cCL. As water saturation grows in the cCL pores, the capillary pressure will build up. The water will break into the MPL if the capillary pressure exceeds the breakthrough pressure of the MPL. The liquid water in the pores of the CL/MPL/GDL at subzero may be stable if the pore size is sufficiently small so that the freezing point depression exceeds the local temperature. Otherwise, it is in a

492

Fuel Cells for Transportation

FIGURE 16.7 Schematic of states, transfer, and phase changes of water in ionomer/PEM and catalyst layer at subzero startup and freeze/thaw.

supercooling state. Whether such state releases or not depends on the availability of the nucleus which is sensitive to defects and disturbance. Supercooled water in the flow channels is observed via neutron radiography for as long as 1 hour [93]. Nucleation in one spot may lead to the freezing of supercooled water and the growth of ice in the surrounding area. In principle, three phases of water inside the pores of the CL, MPL, and MPS can convert into one another. However in practice, the dominant phase change is the freezing of the supercooled water.

16.4.2.2 Modes of liquid water transport in ionomer and membrane For the PFSA membrane, there are multiple modes of water transport, as shown in Fig. 16.8. Water may move under the gradient of water content (chemical osmosis), pressure (hydraulic permeation), temperature (thermoosmosis), and electric potential (electroosmosis), depending on the states of water on both sides of the membrane, operation conditions such as the pressure/temperature difference, and whether there is ionic current. Thermo-osmosis of liquid water across the membrane is distinct from other forms of water transport as it cannot be easily related to an electrochemical potential driving force; it is rather driven by the difference in the entropy of water under a temperature gradient. Kim and Mench [94] examined this phenomenon using the NafionTM, FlemionTM, and Gore-SelectTM

Subzero startup of polymer electrolyte fuel cell Chapter | 16

493

FIGURE 16.8 Four modes of water transport in membrane. Adapted from A. Kusoglu, A.Z. Weber, New insights into perfluorinated sulfonic-acid ionomers, Chem. Rev. 117 (3) (2017) 987
1104.

membranes and found the thermo-osmotic flow direction to be always from the cold to hot side, as anticipated for a small pore hydrophilic porous medium. The water flux was found to be proportional to the temperature gradient and increased with the average membrane temperature. The dependency of the thermo-osmotic diffusivity on average temperature showed predictable Arrhenius-type behavior. For operation at normal temperature, water transport due to thermo-osmosis may be small compared with other modes. Yet during shutdown or subzero startup, it may become nonnegligible or even dominant due to the increase in temperature difference and the decrease or absence of difference in water activity, pressure, or electric potential. The observed asymmetry that more anode end cells have low potential than the cathode end cells during subzero startup in a stack can be explained by the thermo-osmosis transport of water from anode to cathode in the anode end cells [95]. Fuller and Newman [17] treated the water
ionomer system as a concentrated binary electrolyte solution and defined three independent transport parameters, that is, ionic conductivity, diffusion coefficient of water, and electroosmosis drag coefficient, to model proton and water transport within the membrane of a PEFC. These parameters strongly depend on water content and temperature, as well as the types of PFSA used. The following lists the experimental formulae for the most widely used PFSA, the NafionTM membrane, focusing on the subzero temperature range. For modeling of ionomer in the CL or other types of PFSA chemistry, the same formulae are frequently used due to the lack of measurement results. Ionic conductivity The protons are conducted in the PFSA membrane via two mechanisms: the vehicle mechanism, in which protons are carried in the form of hydronium

494

Fuel Cells for Transportation

ions, and the hopping (Grotthuss) mechanism, in which protons are transported by a hydrogen bond-mediated long-range rearrangement of the bonding network among water molecules and hydronium ions, as in bulk liquid water [90]. In both mechanisms, the conductivity is influenced by water content and temperature. The correlation of ionic conductivity of Nafion with water content at normal temperature is given by Eq. (16.13) [91]. The proton conductivity decreases as the temperature decreases.

 1 1 2 σe 5 ð0:5139λ 2 0:326Þ exp 1268 ð16:13Þ 303:15 T where σe is the proton conductivity in the NafionTM membrane, λ is the membrane water content and is defined as the number of water molecules per ionic headgroups in ionomer, and T is the membrane temperature. For subzero temperature, the more accurate correlation of ionic conductivity of the NafionTM membrane with its water content is given by Eq. (16.14) [41]. The εm is the volume fraction of the ionomer in the CL.

 1 1 1:5 2 σe 5 εm ð0:432λ 2 0:662Þ exp 4029 ð16:14Þ 273:15 T

Water diffusion coefficient The water diffusion coefficient in the membrane depends on the temperature and the water content. The commonly used correlations in the subzero simulation of PEFC, as shown in Eq. (16.15), are determined experimentally in [96]. 8
 2346 > 27 > > ½  3:1 3 10 λ exp ð0:28λÞ 2 1 exp 2 ; ð0 , λ , 3Þ > < T
 Dmem 5 2346 > > > 4:17 3 1028 λ ½161exp ð 2 λÞ 1 1 exp 2 ; ð3 # λ , 17Þ > : T ð16:15Þ where Dmem is the diffusion coefficient of membrane water, λ is the membrane water content, and T is the membrane temperature. Electroosmotic drag coefficient The electroosmotic drag coefficient (nd ) is the ratio of the flux of water to that of protons in the absence of a concentration gradient. It is a measure of the number of water molecules carried with each proton, in its solvation shell, under only an electric potential gradient. Most of the measured values of nd for Nafion or other PFSA membranes above room temperature show a dependence on temperature and water content [97]. However, Gallagher’s

Subzero startup of polymer electrolyte fuel cell Chapter | 16

495

[21] measurement using vapor equilibrated methods gives a value of nd close to 1 for Nafion 112 and 117 for subzero temperatures of 210 C and 225 C, over a range of vapor pressure.

16.4.2.3 Modes of liquid water transport in porous media The driving force for liquid water transport in porous media is the capillary pressure gradient. The capillary pressure (pc ) is the pressure difference between liquid (pL ) and gas (pG ), as in Eq. (16.16) for the simple case of a circular tube. pc 5 pL 2 pG 5 2

2σ cosθ r

ð16:16Þ

where σ is the surface tension between the liquid and gas, θ is the water contact angle, and r is the pore radius. In porous media, the capillary pressure is related to the water saturation via the Leverett-J function, as in Eq. (16.17) [98]: 8
0:5 h



2

3 i > > ε > 3 1:42 1 2 slq 2 2:12 12slq 1 1:26 12slq ; θ , 90 σcosθ < K0 pc 5
0:5 h i > > > 3 1:42slq 2 2:12s2lq 1 1:26slq 3 ; θ . 90 : σcosθ Kε0 ð16:17Þ where σ is the surface tension, θ is the contact angle, slq is the water saturation (liquid volume fraction in the pore space), K0 (m2) is the intrinsic permeability, and ε is the porosity. It should be noted that the specific form of the Leverett-J function above in brackets (the constitutive relation) is taken from the experimental results on the sand/oil system with more uniform porosity and wettability in petroleum engineering. There is a scarcity of result for the types of thin/composite porous media in the FC [95,99], and the parameters and properties relevant to the subzero is even rarer.

16.4.2.4 Negligible water vapor transport As shown in Fig. 16.4, the saturation vapor pressure increases nearly exponentially with temperature. This means a significant amount of product water can be removed through the gas phase via reactant flows at normal operating temperatures above 70 C. As a secondary effect, when there is a temperature gradient across the PEM or porous media, there may be a significant flux of water from the high-temperature location to the low-temperature location via the evaporation
diffusion
condensation mechanism [100]. However, at subzero temperatures, the saturation vapor pressure for either supercooled water or ice is so small (Fig. 16.4) that the maximum water vapor carrying capacity of the cathode air flow accounts for only several percent of the product water [98], hence negligible. As a result, the majority of product water has to be

496

Fuel Cells for Transportation

stored temporarily in the ionomer and PEM and then exudes into pores of CL after the local water content exceeds the saturation level which depends on temperature and is much reduced at subzero. The supercooled water in the pores, in a metastable state, has to be removed in liquid form driven by a capillary pressure gradient before freezing.

16.4.2.5 Phase change and icing dynamics In normal operation of the PEFC, the liquid/gas two-phase flow of water is already a complex process to handle. In a subzero startup, processes involving three phases of water inside porous media and their interaction with ionomer need to be taken into account. Since the saturation vapor pressure at subzero is negligibly low, the phase changes between liquid and solid states of water, icing, in particular, will be the dominant processes in freeze/thaw and unassisted startup. Therefore, the following will focus on the icing dynamics at subzero. Discussion on and expressions for the phase change among the vapor/liquid/solid can be found in papers on modeling [101,102]. Freezing of supercooled water and melting of ice constitute a sink and a source term for the liquid water, respectively. Freezing upsets the balance of the capillary force of water in porous media and induces water movement. Massive and sudden freezing tends to damage the pores, cause crack growth, and accelerate delamination in the interface. Based on the classical theory of nucleation, Ishikawa [27] studied the freezing processing in an operating FC, and Durch studied the freezing and melting of water in the GDL [56]. Their major findings will be summarized as follows. Critical nucleation radius Ishikawa et al. [27] used heterogeneous nucleation theory to predict nucleation conditions in the CL. As shown in Fig. 16.9, liquid water resides on a solid surface, and a portion of this liquid water forms a cluster that differs from the liquid phase. In such a system, the free energy change ΔG to form a cluster of ice with a radius r is: ΔG 5 ΔGv Vc 1 ðσsc 2 σsl ÞAsc 1 σcl Acl

ð16:18Þ

where ΔGv is the free energy change of the phase transformation from liquid water to ice, Vc is the volume of the cluster, σcl , σsl , and σsc are the surface tensions between the cluster and the liquid water, the solid surface and the liquid water, and the solid surface and the cluster, respectively, and Asc and

FIGURE 16.9 Nucleation from water cluster formation on a solid surface.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

497

Acl are the interfacial areas between the solid and the cluster, the cluster and the liquid water, respectively. Note that ΔGv is related to the supercooling degree ΔT (K) and the specific latent heat of freezing L (J g21) via: ΔGv 5 2 ρc L

Tm0 2 T ΔT 5 2 ρc L 0 Tm0 Tm

ð16:19Þ

where ρc is the density of the cluster, Tm0 is the melting temperature of the bulk, and ΔT is the degree of supercooling defined as the difference between the melting temperature (Tm0 ) and the temperature (T) of the system. Geometrically, Vc , Asc , and Acl relate to the cluster radius and the contact angle as in the following expressions: Vc 5 πr 3 ð2 2 3cos θ 1 cos3 θÞ=3

ð16:20Þ

Asc 5 πr 2 ð1 2 cos2 θÞ

ð16:21Þ

Acl 5 2πr 2 ð1 2 cos θÞ

ð16:22Þ

where θ shown in Fig. 16.9 is defined according to Young’s Eq. (16.23). σsl 2 σsc cosθ 5 ð16:23Þ σcl Substituting Eqs. (16.19)
(6.23) into Eq. (16.18) yields: ΔG 5 ΔGv πr 3 ð2 2 3cos θ 1 cos3 θÞ=3 1 ðσsc 2 σsl Þπr 2 ð1 2 cos2 θÞ 1 2σcl πr 2 ð1 2 cos θÞ

ð16:24Þ

The cluster radius r at which ΔG attains maximum is defined as the critical cluster radius, r  , in Eq. (16.25). Differentiation and arrangement give rise to a formula for calculating the critical cluster radius r  in Eq. (16.26).

r  5 2σcl

@ΔG 5 0; r 5 r  @r

ð16:25Þ

Tm0 2 2 cosθ 2 cos3 θ ρc LΔT 2 2 3cosθ 1 cos3 θ

ð16:26Þ

Fig. 16.10 shows the change of the free Gibbs energy of the bulk [ΔGv , the first term of the right-hand side in Eq. (16.18)] and the interfaces [ΔGs , the last two terms of the right-hand side in Eq. (16.18)] with the cluster radius. The ΔG attains a maximum at the critical cluster radius r  . Once the cluster radius exceeds r  , the liquid cluster changes to a solid state spontaneously due to the decrease in the free Gibbs energy, causing the release of the supercooling state. A larger critical radius r  , through material choice, surface treatment, and structure modification, results in the delaying of the release of the supercooling state.

498

Fuel Cells for Transportation

FIGURE 16.10 Dependence of free Gibbs energy change on cluster radius and the critical nucleation radius r .

FIGURE 16.11 The dependence of critical nucleation radius on contact angle and degree of supercooling.

As shown in Fig. 16.11, a greater degree of supercooling gives a smaller critical nucleation radius, meaning that supercooled water is more prone to freeze at lower subzero temperatures. The critical nucleation radius depends on the contact angle non-monotonically. Conversely, the degree of supercooling can be solved from Eq. (16.26) as: ΔT 5 2σcl

Tm0 2 2 cosθ 2 cos3 θ ρc Lr  2 2 3cosθ 1 cos3 θ

ð16:27Þ

Table 16.6 compares the supercooling degree estimated from Eq. (16.27) and the melting point depression from Eq. (16.8) in the major components (PEM, CL, MPL, MPS) in a FC. The supercooling degree from heterogeneous

TABLE 16.6 Estimated supercooling degree and melting point depression for the PEM and porous media in fuel cell. PEM

Catalyst layer

Microporous layer

Macroporous substrate

Pore size

B2 nm

Within primary particles: B4 nm

Between secondary particles: B30 nm

B100 nm

B1 μm

Contact angle

B110 degrees

B145 degrees

B145 degrees

B140 degrees

B125 degrees

Supercooling degree [Eq. (16.26)]



52 C







28 C

3.7 C

1.1 C

0.1 C

Melting point depression [Eq. (16.8)]

65 C

33 C

4.3 C

1.3 C

0.13 C

500

Fuel Cells for Transportation

nucleation is smaller than the melting point depression from homogeneous nucleation, which means supercooled water is easier to freeze on the surface than in the bulk. At θ 5 180 , Eq. (16.26) reduces to the Gibbs
Thomson equation (Eq. (16.8)). Evidently, the degree of supercooling here is of the same nature as the melting point depression in Eq. (16.8). Nucleation rate In the nucleation theory in Ref [103], the nucleation rate is described by a parameter called critical cluster production rate, J, which is the production rate of the clusters reaching the critical cluster size in a certain water volume. J is a parameter reflecting the duration of the supercooling state and is defined as follows. J  Vw 5 v  n

ð16:28Þ

where v is the molecular frequency of the activated water, and n is the number of critical clusters in the volume of water Vw . The molecular frequency v here corresponds to the probability that a water molecule is included in an existing cluster, and is expressed using Planck’s number h, Boltzmann’s constant κ, and the water molecule activation energy Δg.
 kT Δg exp 2 v5 ð16:29Þ h kT The number of critical clusters r  also obeys the Arrhenius formula in terms of the total number of water molecules N and the critical cluster free energy ΔG .
 ΔG  n 5 N exp 2 ð16:30Þ kT where N can be given by the product of the number density of water nL and the volume of water Vw . N 5 n L Vw

ð16:31Þ

Substituting Eq. (16.29)
(6.31) into Eq. (16.28) yields the critical cluster production rate J:

 nL kT Δg 1 2 16πσ3cl T2  exp 2   2 2 e 2  f ðθÞ ð16:32Þ J5  exp 2 h kT kT 3 ρc L ΔT Here, the Δg is the activation energy for a water molecule immigrating in a self-diffusion process and can be estimated with two self-diffusion coeffi0 cients (D) at different temperatures of T and T .   k DðT Þ  ln 0 0 Δg 5 2  ð16:33Þ 1 1 D ðT Þ 2 0 T T

Subzero startup of polymer electrolyte fuel cell Chapter | 16

501

FIGURE 16.12 Dependence of nucleation rate on contact angle at three supercooling degree. Adapted from Y. Ishikawa, M. Shiozawa, M. Kondo, K. Ito, Theoretical analysis of supercooled states of water generated below the freezing point in a PEFC, Int. J. Heat. Mass. Transf. 74 (2014) 215
227.

Fig. 16.12 shows the dependence of the nucleation rate (J) on the contact angle for three supercooling degrees. At the same supercooling degree, the nucleation rate is higher for the smaller contact angle (more hydrophilic surface). The dependence is very sensitive, in that a slight change in the contact angle will cause a drastic change in the nucleation rate, particularly for a low degree of supercooling. For the same contact angle, the nucleation rate increases significantly with the increase in the supercooling degree. Integrating the product of the water volume Vw ðtÞ and the critical cluster nucleation rate JðT; θÞ with respect to time, a parameter I is defined in Eq. (16.34). When I reaches unity, the supercooling state is considered to be released. ð t0 I5 J ðT; θÞ  Vw ðtÞdt ð16:34Þ 0

Induction time of freezing and ice growth When the cluster size reaches the critical radius, ice will nucleate and grow, as shown in Fig. 16.13. Dursch et al. [55,56] studied the ice nucleation and the growth process in the CL and GDL components from experiments and modeling. They measured the heat flux from ice formation in the CL and GDL with DSC.

502

Fuel Cells for Transportation

FIGURE 16.13 Ice growth after nucleation at the critical radius r .

They utilized the theoretical framework of Johnson
Mehl
Avrami
Kolmogorov to define an ice crystallization rate ϕðtÞ as a convolution integral over nucleation and growth rates in Eq. (16.37), with the following assumption: (1) isothermal boundary condition; (2) semi-infinite space boundary condition; (3) uniform contact angle distribution. 0 1 ðt B 4π C ϕ 5 1 2 exp@ 2 gðθÞ J ½T ðt0 Þr 3 ðt 2 t0 Þdt0A ð16:35Þ 3 τi

where t is time, θ is contact angle, J is the nucleation rate given in Eq. (16.32), and gðθÞ is defined in Eq. (16.36). 1 gðθÞ 5 ð2 1 cosθÞð12cosθÞ2 4

ð16:36Þ

The induction time, predicted by Eq. (16.35), for the ice crystallization after the sample is subjected to a step decrease in temperature to subzero, showed good agreement with their DSC measurement [55]. The ice growth process is classically posed as a Stefan problem, which consists of two simple equations. The first is the energy conservation equation: @T 5 Dr2 T @t

ð16:37Þ

where D denotes a diffusion coefficient. The second is the ice/water interface evolution equation in the normal direction: @Γ @T n5D @t @n

ð16:38Þ

where n denotes the normal direction, and Γ is the position of the ice/water interface. First posed by Josef Stefan as a model of ocean ice forming in arctic regions, Stefan’s problem has since found applications in many fields. Due to the nonlinear behavior of the Stefan problem, it has attracted considerable interest, especially in mathematics. Many different flavors of the Stefan problem exist that impose various boundary conditions on the heat field and interface. However, only limited studies are conducted in the field of FCs. Dursch

Subzero startup of polymer electrolyte fuel cell Chapter | 16

503

et al. [19] studied the ice-melting process in the GDL formulated as a 1D moving-boundary Stefan problem. The model prediction of the ice-melting rate and ice-melting time agreed well with their DSC measurements. Freezing-induced liquid water movement Freezing not only constitutes a sink term in the liquid water transportation equation, but it may also induce the movement of nearby liquid as a secondary effect. The phase change from liquid water into solid alters the liquid water saturation inside the porous media, causing the change in the capillary pressure gradient, which will drive the transport of liquid water toward the ice. Ice lenses may form in the gaps of the interface in between or cracks in the CL and MPL, causing delamination or disintegration of the porous media like the frost heave phenomena in geology. The possibility of freezinginduced water movement may affect the degree to which the ice distribution observed from cryo-SEM can represent the original liquid water distribution before immobilization via freezing with liquid N2.

16.4.3 Water fill test: contribution and limitation Among the four major categories of the experimental study of subzero problems as introduced in Section 16.2, WFT is the one that occupied the most interest and has greatly enhanced our understanding of the subzero startup process. WFT typically uses a small single cell which is approximately at isothermal conditions, due to the high thermal mass of the thick endplates and the low current density involved. Most of the tests use a galvanostatic control at current densities between 10 and 100 mA cm22, some use potentiostatic control at potentials between 0.1 and 0.6 V, and some limited cases explore the effects of the combination of both modes of load control. The stages and features of an isothermal subzero startup are examined, with an aim to understand the location of ice formation and the startup failure mechanism. A metric called cumulative product water [104], which is the total amount of product water before the startup failure, is proposed as an indicator of the subzero startup capability. Cumulative product water is a reflection of the water storage capacity of the MEA. The effects of cell design and operation control on the effective water storage capacity have been the subject of extensive study.

16.4.3.1 Stages and features of water fill test Using galvanostatic control as an example (Fig. 16.14), three stages in the WFT can be delineated. In the first stage, the ionomer in the cCL and PEM is hydrated with product water, leading to a decrease in HFR and an increase in cell voltage. Some studies used a current ramp to transit to the constant current at the very beginning. In such a case, the cell voltage decreases

504

Fuel Cells for Transportation

FIGURE 16.14 Isothermal water fill tests. (A) Temperature effects (λ 5 6:2, 40 mA cm22), (B) membrane water content effects (220 C,100 mA cm22), (C) current density effects (220 C, λ 5 2:2). (A and C) Adapted from Y. Tabe, M. Saito, K. Fukui, T. Chikahisa, Cold start characteristics and freezing mechanism dependence on start-up temperature in a polymer electrolyte membrane fuel cell, J. Power Sources 208 (2012) 366
373 (B) Adapted from K. Tajiri, Y. Tabuchi, F. Kagami, S. Takahashi, K. Yoshizawa, C.Y. Wang, Effects of operating and design parameters on PEFC cold start, J. Power Sources 165 (1) (2007) 279
286.

during the ramp due to the increase in the current. In the second stage, after the water content in the ionomer of the cCL reaches saturation, the water inside the ionomer exudes into the pores of the CL and subsequently breaks into the MPL, MPS, and even flow channels in a supercooled state. The increase in water saturation in the porous media restricts the O2 transport and leads to a gradual decrease in cell voltage. The competition between the membrane hydration, which reduces the ohmic overpotential, and porous media blockage, which increases the concentration overpotential, may result in a maximum cell voltage in the second stage. In the third stage, supercooled water freezes, blocking the pores and leading to a quick drop in voltage, and the subzero startup fails. The voltage drop in the third stage tends to be abrupt at temperatures higher than 210 C and gentle at lower than 210 C. The reason is that at temperatures higher than 210 C, the product water tends to exist in a supercooled state for a longer time, flowing out from the CL into the MPL, GDL, and even flow field. The nucleation at a certain place will cause sudden ice formation and growth, leading to an abrupt failure. At temperatures lower

Subzero startup of polymer electrolyte fuel cell Chapter | 16

505

than 210 C, the probability of product water being in the supercooling state is drastically lower [22]; thus, it will freeze inside the CL and choke off the supply of O2 gradually. For temperatures above 25 C, the cell may continue to work for a prolonged time as the waste heat may warm up the core part of the cell above 0 C. Fig. 16.14 shows the effect of operating conditions, such as subzero temperature, current density, and initial water content, on the features and cumulative product water of WFT. Generally, higher temperature, lower initial water content, and smaller current density operate longer and generate more product water. Purging the cell during the shutdown process not only mitigates the damage from freezing [3,47] but also improves the subzero startup performance. Purging removes the residual water inside the pores of the CL, MPL, and MPS so that the possible nucleus triggering the release of the supercooling state of the product water will be reduced, leading to longer operation times. Purging also reduces the water content of the membrane so that more water can be stored during the subzero startup. The MEA design has a strong effect on the cumulative product water. While a thick membrane may seem to have a large capacity for storing product water, the effective water storage capacity is determined by the balance of the water uptake rate of the membrane and the water generation rate at the cCL. Since the FC with a thin membrane has a higher water uptake rate due to the higher water content gradient, it can have a larger effective water storage capacity at high current density [9]. Higher Pt loading is found to have higher effective water storage capacity [61]. The MPL design has a great effect on cumulative water production but sometimes gives inconsistent results. Some reports that a hydrophilic MPL helps to reduce the breakthrough pressure and allow product water to flow into the GDL thereby increasing the cumulative product water [24]. While others find that a hydrophobic MPL may help to facilitate the movement of water into the anode side, thereby also increasing the cumulative product water [54]. Electrospun MPL and CL are reported to have a larger amount of cumulative product water, attributed to the improved water removal from the CL to the MPL [54,62]. Large uncertainty in operating time and cumulative product water in the WFT were recognized and extensively examined by the research group in PSI [38]. It was also found that the cell with a larger active area had lower uncertainty [102], and the dispersion in operation time was higher for the subzero temperature close to zero [62]. Since liquid water was detected with in-situ neutron radiography in the flow channel at subzero, the uncertainty was attributed to the stochastic freezing of the supercooled water. The product water is pure, and the FC components are usually fabricated and assembled under clean conditions; therefore, the product water at subzero above 210 C can exist in a supercooled state for a prolonged duration. However, since supercooled water is thermodynamically metastable, it tends to freeze spontaneously.

506

Fuel Cells for Transportation

The release of the supercooling state of water can be triggered by defects, cracks, or hydrophilic sites in the FC. It was found the size of the critical nucleation cluster was much smaller (easier) on the hydrophilic surface than that on the hydrophobic surface [27]. Mechanical shock can also act as a trigger. Oberholzer et al. [38] found repetitively in the four trials that dropping the cell (1.5 kg) from 35 mm high to the base plate caused the collapse of the cell voltage. The energy input corresponding to the kinetic energy of the cell before it hits the base plate was estimated to be as low as 47 mJ. To avoid the influence of vibration from the environmental chamber on the freezing of water inside a FC, the cold start tests were performed after the chamber was powered off [105]. In contrast, a dedicated test using a vibration machine [53] found that vibration can delay the freezing of supercooled water and extend the startup time by facilitating the water movement into the small pores in the CL and inhibiting ice accumulation at the interface. The frequency near the cell’s natural frequency (B10 Hz), the amplitude of 2 mm, and orientation along the gravity were found to be most beneficial in increasing the cumulative product water. Such discrepancy in the effect of mechanical disturbance may also attest to the existence as well as the unpredictable nature of supercooled water. The discovery of the existence of supercooled water in 2007 [26] during the subzero startup marks a milestone. Before 2007, most of the modeling assumes no liquid water inside the FC. After the discovery, great interest has been directed to the understanding and modeling of the behavior of supercooled water in FCs via experiment [22,28,34], theory [27,55], and simulation [98], in a hope that supercooled water may greatly increase the cumulative product water and the water storage capacity, hence the startup performance. However, considering the intrinsic propensity of supercooled water to freeze and the severe damage of freezing to the FC, a startup strategy based on the supercooled water is inherently unreliable, which is unacceptable for field application. As a result, the design-available water storage capacity is limited to the non-frozen water in the PEM and ionomer in the CL, and the nanosized pores inside the carbon support. The temperature of the core part of the FC has to be raised above zero before this reliable water storage capacity is used up, which results in a tight competition between water and thermal management. Currently, Toyota’s commercialized FCEV MIRAI has already realized the subzero startup capability exceeding the DOE’s target, without relying on supercooled water (see Section 16.6 for further discussion). Therefore, it is the view of the authors of this chapter that the study of supercooled water may be academically interesting, but supercooled water cannot be counted on in any field application due to the demand on reliability and durability. A shift to more industrially relevant study is advocated to develop reliable and durable subzero startup capability.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

507

16.4.3.2 Failure mechanism of subzero startup There is convincing evidence from visualization [26,36], a sudden increase in temperature [33,59,106] and HFR [22] that ice formation accompanies the potential drop and triggers the failure of the startup. However, how exactly the ice caused the failure has been a subject of vague description and controversial explanation. Some believed that the ice chokes the pore for the passage of O2 [63], while others argued that the ice covers the Pt surface [37]. Early models usually contain an equation describing the decrease of the electrochemical active area due to the accumulation of ice in the CL [107]. To address this controversy, Wang et al. [29] broke the WFT into several stages (Fig. 16.15), using the dynamic EIS during each stage and the CV between the stages to characterize the cell. It was found that the ECSA essentially remains unchanged in all the stages even after the failure. In a WFT from 210 C, galvanostatic EIS (GEIS) was measured continuously (Fig. 16.16), and the EIS results were fitted to a transmission line model with an interface impedance consisting of a capacitor in parallel with a series connection of charge transfer resistance and finite-length Warburg impedance. There was a dramatic increase in the mass transport resistance at the final stage while the charge transfer resistance changed little. The serial CV, dynamic EIS, and GEIS characterization results, as well as the fact that melting point depression in ionomer is much greater than that in the CL pores, support the view that the choking of the O2 pathway by the ice is the culprit for the subzero startup failure, rather than the loss of ECSA from the covering of the Pt surface by ice. The water production, transport, and ice formation process are summarized in a schematic in [29]. A more detailed picture in the final step can be speculated as follows. Immediately before the release of the supercooling state, the water saturation in the pores of cCL is below 1.0. The freezing of

FIGURE 16.15 Evolution of CV during cold start from 220 C. (A) The applied current and the voltage response as a function of time; (B) CV curves at the end of each stage. The results are from the authors’ laboratory.

508

Fuel Cells for Transportation

FIGURE 16.16 GEIS measurements during cold start from 210 C: (A) the applying current and voltage response with time. (B) The EIS curve evolution. The results are from the authors’ laboratory.

the supercooled water immobilizes the water in the pores of cCL and/or GDL, reducing the local liquid water saturation (hence capillary pressure) and providing a base for ice growth. The reduced capillary pressure tends to draw supercooled water in the surrounding area to move to the ice and freeze. The ice growth (and the associated volume expansion) disconnects the liquid water pathway from the reaction site to MPL. Therefore, the newly formed product water cannot be transported away, but rather freezes in the pores around the reaction site, increasing the local ice saturation quickly close to 1.0, thereby choking all the supply of O2 to the catalyst. Several studies observed a significant increase in the HFR accompanying the collapse of the cell voltage [38,47,53,108] or the occurrence of freezing [33]. Such an increase may be attributed to the increase in the contact resistance inside the cCL or between the cCL/MPL due to the freezing of accumulated product water, or the increase in PEM resistance due to the draining out of the water from the PEM, depending on the amount of product water, clamping force, and MEA design and quality. Though the associated ohmic loss increment is insignificant (just about 10 mV, much less than the several hundred mV of the voltage collapse), understanding such phenomena helps to shed light on the failing process and failure mechanism. Li et al. [105] studied the effects of clamping stress on the constant current (0.04 A cm22) cold-start performance at two subzero temperatures. While all the startups at 23 C proceeded stably and similarly, that at 25 C showed sensitive dependence on the level of the clamping stress. Specifically, the lower the clamping stress, the shorter the operation time and the larger the increase in the HFR in the failure stage, probably due to the easier accumulation and subsequent freezing of water in the cCL/MPL interface at lower compression. Xie et al. [53] studied the effect of mechanical vibration on the cold-start performance (at 0.1 A cm22) and observed a significant HFR increase from startup at 213 C, but no HFR increase from

Subzero startup of polymer electrolyte fuel cell Chapter | 16

509

startup at 220 C. The difference may be attributed to the more than doubling of the amount of cumulative product water at 213 C than that at 220 C. For the same subzero temperature of 220 C, however, Pistono and Rice [47] found in WFTs (at 0.01 A cm22) that HFR increased significantly accompanying the cell voltage collapse. The discrepancy between Xie and Pistono’s observation may be caused by the delicate balance between the MEA design and operation control, here specifically the membrane thickness and startup current density, which are 50 and 20 μm, 0.1 and 0.01 A cm22, respectively. Higher current density means a higher heat generation rate and faster warming up of the membrane, leading to an increase in the backdiffusion coefficient of water and a larger storage capacity for the nonfrozen water. As a result, less product water will move into and freeze inside the cathode CL. The second cause for the increase in HFR accompanying the voltage collapse is the increase in PEM resistance. Alink et al. [43] compared the change of HFR with temperature during two freeze/thaw tests and one coldstart test from 240 C and attributed the step-like increase in HFR in the freeze/thaw cycle with relatively more residual water to the draining out of the water from the membrane, rather than the increase in the contact resistance. The test results from a simplified situation also support such a possibility [33] (Fig. 16.17). The HFR and temperature of a FC with an initial membrane water content of 5.0 and 11.0 were monitored during the cooling down from 20 C to 230 C in an environment chamber. While the change in HFR and temperature for the case of λinital 5 5.0 was gradual, that for the cell with inital water contents of 11.0 exhibited an abrupt rise near the temperature of 25.5 C. The rise in temperature is caused by the latent heat

FIGURE 16.17 HFR and temperature change during the cooling down process. The results are from the authors’ laboratory.

510

Fuel Cells for Transportation

released with the freezing of water. The water most probably froze outside the PEM, rather than inside it, due to the much greater melting point depression in PEM. Since the amount of water drained out was considered inadequate to freeze to an extent to cause an appreciable increase of the contact resistance, the concurrent step rise in HFR was most probably caused by the draining out of water from the PEM following the freezing of the water already outside the PEM. Most of the studies in the WFT use low current density and focus on the cathode pore blockage as the failure mode. Yao et al. [106] found through non-isothermal simulation of galvanostatic startup of the single channel FC that anode dehydration can be the mode of failure in the region with low initial water content and high current density in the operation chart of initial water content vs. startup current density (Fig. 16.18). Such regions will expand at lower subzero temperatures due to the reduced back-diffusion of product water. These results indicate that there is a lower limit for the initial membrane water content to allow for adequate startup current density. The model took into account the phase changes of the water among membrane and three states of water in the pores and treated the freezing process from supercooled water into ice as a random process using a probability based on classical nucleation theory. A random region (in light blue in the figure) between the anode dehydration region and cathode pore blockage region was predicted where the failure was stochastic. For guaranteed prevention of anode dehydration, this region also needs to be avoided. It has to be noted that the failure mode map predicted by Yao et al. is an attempt to understand systematically how the startup fails. It is valuable in that it accounts for the stochastic process of freezing and reveals the need to adjust the initial water

FIGURE 16.18 The map of failure modes for the cell starts from (A) 220 C and (B) 230 C. Adapted from L. Yao, J. Peng, J.B. Zhang, Y.J. Zhang, Numerical investigation of cold-start behavior of polymer electrolyte fuel cells in the presence of super-cooled water, Int. J. Hydrog. Energy 43 (32) (2018) 15505
15520.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

511

content in the membrane to avoid anode dehydration. Another quantitative or even qualitative conclusion depends on the specific cell design and startup strategy used in the simulation, not to mention the validity of the expressions and parameters in the model. The ultimate goal of the experiment and simulation study is to search for cell design and operation control to overcome such failure modes.

16.4.3.3 Contribution and limitation of water fill test Since WFTs use a single cell involving a small current density, it is essentially isothermal thereby decoupling the thermal issues and greatly simplifying the situation. Extensive WFTs have greatly enhanced our understanding of the states and behavior of water inside FCs at subzero. It helps to clarify the stages and features in the subzero startup and identify the major failure mechanism. It also helps to calibrate models and verify simulation results. The metrics like cumulative product water and water storage capacity help to compare the effects of the design and operation parameters on the subzero startup performance. However, for the same reason that the WFT is nearly isothermal, it is somehow disconnected from the technically relevant subzero startup, where high current density and fast warming up are the basic requirements [3]. Understanding the failure mechanism does not translate directly into a successful startup technique. The metric of cumulative product water may be misleading as the low current density tends to give a high value. Yet at low current density, the heat generation rate will be too low to warm up the cell, and the anode dehydration failure mode cannot be revealed. In addition, the surging academic curiosity on the behavior and modeling of the supercooled water may not help to solve the problem, as the reliability required in the field application precludes the exploitation of this thermodynamically metastable state. To achieve efficient, reliable, and durable subzero startup, the temperaturedependent properties, the thermal behavior, and the interplay of water and thermal management need to be explored. Startup tests using stack or single cell simulating adiabatic thermal boundary conditions [41], classified as the fourth of the four major categories of experimental subzero study in Section 16.2, are needed. These issues are addressed in Sections 16.5 and 16.6.

16.5 Temperature-dependent properties and thermal behavior at subzero Water and thermal management are like two wheels of a cart for the smooth operation of the FC, either in normal operation or in subzero startups. This section addresses the issues related to thermal management, particularly the temperature-dependent properties and thermal behavior at subzero.

512

Fuel Cells for Transportation

With its normal operating temperature at 60 C
90 C and the ambient temperature at 230 C
40 C, PEFC experiences a temperature range of about 100 C. The performance, reliability, and durability of PEFC depend sensitively on temperature, as the temperature has significant effects on the mechanical properties, transport and kinetic parameters, saturation water and non-frozen water content in the ionomer, etc. At subzero, all the transport and kinetic parameters become small, leading to slower processes and inadequate functionality of the FC. Furthermore, residual or product water tends to freeze at subzero, causing irreversible damage to the FC. While at normal temperature operation, every effort is made to reduce the overpotential, so as to increase the fuel efficiency and to alleviate the burden on the radiator to dissipate the heat, and the waste heat from the overpotential turns out to be the most useful ally for an unassisted startup. As the temperature is determined by the balance of heat source and heat loss, and the transient response is affected by the thermal mass, it is essential to understand the thermal properties and behavior of all the components and processes in the FC at subzero.

16.5.1 Temperature-dependent properties Thermodynamic, kinetic, transport, and thermomechanical parameters all have a temperature dependence. The equilibrium potential of ORR depends on temperature and reactants’ partial pressures according to the Nernst equation, as shown in Eq. (16.39): ! RT 1 0 ln ENernst 5 Ecell 2 0:009ðT 2 T0 Þ 2 ð16:39Þ

0:5 2F ðpH2 =p Þ  pO2 =p 0 is the equilibrium potential of the cell under the standard state at where Ecell T0 (V), T0 is the room temperature (25 C), R is the gas constant (J K21 mol21), p is the reference pressure (1 bar), and pH2 and pO2 are the H2 and O2 partial pressures (bar), respectively. It is clear that the equilibrium potential is higher at a lower temperature. For a general electrochemical reaction O 1 e2 5 R, the reaction current density depends on temperature and overpotential according to the Butler
Volmer equation, as shown in Eq. (16.40): 

 cO αF cR ð1 2 αÞF η 2 η ð16:40Þ exp 2 exp i 5 i0 RT RT cO  cR 

where i0 is the exchange current density, α is the charge transfer coefficient, η is the overpotential, cO and cR are the concentration of the species O and R at the electrode, and cO  and cR  are the concentration in the bulk. The charge transfer coefficient α (related to the Tafel slope via b 5 2:303RT αF ) and exchange current density i0 also have a temperature dependence. Refs.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

513

[109
111] discussed several possible mechanisms that contribute to the temperature dependence of the charge transfer coefficient. Tajiri et al. [25] found that the Tafel slope increased from the typical B66 mV decade21 under room temperature to B105 mV decade21 at 230 C. It means about 40 mV higher activation overpotential is needed to drive the current to increase by 10-fold at 230 C. Thompson et al. [112] found no fundamental change in the ORR mechanism at subzero temperatures. The exchange current density and the activation energy at zero overpotential were estimated to be 1:2 3 1028 A cm22Pt and 55 kJ mol21 under 353K and 1 atm, respectively. The exchange current density at other temperature T can be estimated according to Eq. (16.41) in an Arrhenius fashion.
  c 
1 1 2 i0 5 iref ð16:41Þ exp 2 1400 0 cref T 353:15 22 where iref 0 is the reference exchange current density (A cm ) at T 5 353; 21 c is the molar concentration of the reactant (mol L ), and cref is the reference molar concentration (mol L21). The major transport parameters in the FC and their temperature dependence are listed in Table 16.7. In membrane and ionomer, the water diffusivity and proton conductivity are highly dependent on temperature. In the CL, the gas diffusion coefficients are highly dependent on temperature. The thermal/mechanical/electrical properties of the typical materials in a FC are listed in Table 16.8 [113
117]. Different degrees of shrinking of the components as temperature drops below zero may impair the integrity of the sealing.

16.5.2 Thermal management issues highlighted from a lumped model A lumped thermal model of the FC is adequate to highlight the issues in thermal management critical to subzero startup. Assuming a uniform temperature in the FC, the law of energy conservation can be presented as: dT 5 Q_ gen 2 Q_ loss ð16:42Þ dt where mtherm is the thermal mass of the cell, Q_ gen is the heat generation rate, and Q_ loss is the heat loss rate. mtherm

16.5.2.1 Thermal mass The thermal mass in the lumped model is the sum of the thermal mass of all the components. It depends on the density and specific thermal capacity of the components as well as the cell design. Roughly speaking, BPP accounts

TABLE 16.7 Dependence of transport parameters on temperature. Parameters and description Dmem , diffusion coefficient of membrane water

σe , proton conductivity μl , dynamic viscosity of liquid water Dm binary molecule diffusion coefficient

Equation

Units


8

2436 > 27 > ; 0,λ#3 λ exp ð 0:28λ Þ 2 1 exp 2 3:1 3 10 > > T < 
Dmem 5

2436 > > > 4:17 3 1028 λ 161 exp ð 2 λÞ 1 1 exp 2 ; λ.3 > : T 1

2 T1 σe 5 ð0:432λ 2 0:662Þ exp 4029 273:15 lnμl 5 131:48 2 1:69T 1 8:10 3 1023 3 T 2 2 1:75 3 1025 3 T 3 1 1:42 3 1028 3 T 4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 2  μ1 1 μ1 Dm 5 435:7T 1 1 2 p VA3 1VB3

DK , Knudsen diffusion coefficient

DK 5

dk 3

qffiffiffiffiffiffiffi 8RT πM

A

m2 s21

S m21 Pa s m2 s21

B

m2 s21

TABLE 16.8 Properties of the typical materials in fuel cell. Material

Young’s modulus, kN mm22 (25 C)

Density, g cm23 (25 C)

Average specific heat, J g21 K21 (0 C
100 C)

Thermal conductivity, W m21 K21 (100 C)

Electrical resistivity, Ω m (1028) (25 C)

Average coefficient of thermal expansion, 1026/ C (0 C
100 C)

Nafion 112 (Dry)

0.24

1.98

1.05

0.259




400B470a

Silicon rubber

0.000005
1.90

0.700
3.80

0.32
1.26

0.06
6.50

50
6.0 3 1021

11.0
335

SUS316L

193

8.00

0.5

14.0
15.9

74

16.9

Pure titanium

106

4.51

0.52

17

51

8.4

Titanium alloy Ti-6Al-4V

113

4.42

0.52

7.5

171

8.6

BMC 940 Vinyl Ester Bipolar Plate

0.028

1.89

0.841

19.2

In plane 10000 Through plane 20000

30

a

The average coefficient of thermal expansion of NRE112 is estimated from Fig. 16.11 in Ref. [118].

516

Fuel Cells for Transportation

for nearly half of the total, and BPP made of graphite is higher than that of SS316L, which in turn is higher than that of titanium alloy. The density and specific thermal capacity of the major components in the FC at room temperature are listed in Table 16.9.

16.5.2.2 Heat source and heat generation rate Unassisted start depends on the heat generated from ORR while water is produced. The total heat generation rate includes three parts. Q_ gen 5 Q_ rev 1 Q_ irrev 1 Q_ pc

ð16:43Þ

The reversible heat generation rate, Q_ rev , can be written as
 iA @ENernst 5 2T Q_ rev 5 ð 2 TΔSÞ iA 2F @T

ð16:44Þ

where i is the current density, A is the area of the active region, ΔS is the entropy change, and ENernst is the equilibrium potential given in Eq. (16.39). The total irreversible heat generation rate due to the cell overpotential is
 ΔG 2 Vcell iA 5 ðENernst 2 Vcell ÞiA ð16:45Þ Q_ irrev 5 2 2F where ΔG is the free energy change of Eq. (16.3), and Vcell is the cell voltage. The heat release due to the phase change of water is calculated by Eq. (16.46). Q_ pc 5 n_ pc L

ð16:46Þ

where L is the latent heat of phase change, and n_ pc is the rate of phase change. The cell overpotential includes the activation overpotential (ηact ), ohmic overpotential (ηohmic ), and concentration overpotential (ηconc ), as shown in Eq. (16.47). These overpotentials can be estimated from the FC performance curves or cell design and operation parameters. Vcell 5 ENernst 2 ηact 2 ηohmic 2 ηconc

ð16:47Þ

where ENernst is given in Eq. (6.39), and ηact , ηohmic , and ηconc are calculated from Eqs. (16.48), (16.49), and (16.50), respectively.
 i ηact 5 b log ð16:48Þ i0 ηohmic 5 Rohm i 5 ðRBPP 1 RGDL 1 Rmem Þi 1
 i ηconc 5 2 b log 1 2 iL

RCL i 3

ð16:49Þ ð16:50Þ

TABLE 16.9 Density and specific thermal capacity of fuel cell components. Property

23



ρ kg m

Cp J kg21 K21

PEM

CLs

GDLs

BPPs

Coolant

PFSA

Carbon

Carbon

Graphite

SS316L

Ti

Glycol and water (1:1 mass ratio)

1980

1230

1230

2230

8000

4506

1050

833

462

462

710

500

544

3500

518

Fuel Cells for Transportation

Eqs. (16.48) and (16.50) are obtained from Eq. (16.40) when the first term on the right hand overwhelms the second term (at current densities much larger than the exchange current density), and assuming Nernst layer diffusion which defines a limiting current density iL . In Eq. (16.49), RBPP , RGDL , and Rmem are the area specific electrical resistance (Ω cm2) of the BPPs, GDLs, and membrane, respectively. RCL (Ω cm2) is the area specific proton conduction resistance of ionomer in the CL. The i0 is calculated from Eq. (16.51) [8,109]:  c  i0 5 iref ð16:51Þ 0 δCL cref 23 where iref 0 is the volumetric exchange current density (A cm ) of the CL at reference concentration, δCL is the thickness of the CL (cm), c is the mole concentration of O2 (mol dm23), and cref is the reference mole concentration of O2 (mol dm23). In the FC, the effective O2 concentration is affected by the air stoichiometric ratio (SR). The limiting current density iL is calculated from Eq. (16.52). f ðξca Þ is a function of air SR ξca in the cathode, and it can be expressed in Eq. (16.53), which is derived from Ref. [108]. Deff is the effective diffusion coefficient of oxygen in the GDL of a thickness lb . The limiting current density iL is affected by cell design, operation conditions, and volume fraction of liquid water and ice in the pores.

4FDeff c iL 5 f ξca lb


1 f ξca 5 2 ξ ca ln 1 2 ξca

ð16:52Þ ð16:53Þ

Both the exchange current density i0 and the limiting current density iL depend on the air SR ξ ca . Therefore, the air SR will have a significant impact on the magnitude of the activation overpotential (ηact ) and the concentration overpotential (ηconc ).

16.5.2.3 Heat loss Heat losses consist of conductive, convective, and radiative heat transfer. Radiative heat transfer is negligible due to the low temperature as well as the small temperature difference. The convective and conductive heat transfer include the enthalpy differences between the inlet and outlet gases and the heat dissipation through coolant and exterior surfaces, as shown in Eq. (16.54):

gas gas

gas

co co _ p out Tout _ p in Tingas 1 mC _ p out Tout 2 mC Q_ loss 5 mC ð16:54Þ

co _ p in Tinco 1 hext Aext ðTcell 2 Tambient Þ 2 mC

Subzero startup of polymer electrolyte fuel cell Chapter | 16

519

_ p is the thermal mass flow rate, the subscripts out and in mean the where mC outlet and inlet, respectively, the superscript co means coolant, hext is the conductive heat transfer coefficient between the FC and the ambient, Aext is the exterior area of the FC exposed to the ambient, Tcell is the FC temperature, and Tambient is the ambient temperature. The heat losses must be minimized to assist the rapid startup. Heat loss to the reactant gas can be reduced either by increasing the inlet gas temperature or decreasing the gas flow rate. Heat loss to the coolant and exterior surface can be reduced by deactivating the cooling, enhancing thermal insulation and adding external heating.

16.5.2.4 Thermal boundary conditions Depending on the location of cells in a stack, different thermal boundary conditions may apply. The cells in the center of the stack are nearly adiabatic, as the adjacent cells behave approximately the same. The cells adjacent to the endplates are nearly isothermal due to the large thermal mass of the endplate, if no external heating is applied. For other cells in between, their thermal boundary condition is asymmetric and more complex. At the periphery of large cells, the extent of heat loss depends on the thermal insulation. 16.5.3 Discussion on heat source, temperature difference, and thermal BC control 16.5.3.1 Heat source Of the three types of heat sources in Eq. (16.43), the irreversible heat, also referred to as waste heat at normal operation, is the most useful one for a subzero startup. The latent heat accompanying the freezing of product water should be avoided rather than utilized. The reversible heat generation rate is proportional to the current density with a coefficient determined by the reaction and temperature, rather than the cell design and operation control. In addition, the proportionality with current density means that there is no way to maximize the reversible heat-to-water ratio by adjusting the current density. In contrast, the irreversible heat depends nonlinearly on current density and air stoichiometry, which can be exploited to enlarge the heat generation rate as well as the heat-to-water ratio. Actually, Toyota developed a rapid warmup method for their FC stack by aggressively reducing the air stoichiometry to maximize the irreversible heat generation rate. When the stoichiometry is close to one, the chemical energy of the reactants is almost entirely converted to heat to warm up the FC. A simple estimation can give us insight into the magnitude and sensitivity of the three components of the irreversible heat, as illustrated in Table 16.10. It can be seen that the heat generation due to the concentration overpotential

TABLE 16.10 Estimation of the irreversible heat generation rate from three types of overpotential. Type of overpotential

Activation

Expressions for heat generation rate

  ηact i 5 bi log ii0

Estimated values (W cm22) at i 5 0.4 A cm22 Vcell 5 0.1 V and T 5 230 C

0.210

Ohmic

Concentration

ηohmic i 5 ðRBPP 1 RGDL 1 Rmem Þi 1 2

0.032

RCL i 2 3

  ηconc i 5 2 bi log 1 2 iiL

0.322

Notes: The Tafel slope b is taken to be B105 mV decade21 [25]. The i0 of the ORR is estimated to be 4 3 1026 A cm22 from Eq. (16.41). The ohmic resistance is mainly from the membrane and the resistance of the membrane with a water content at λ 5 6 is taken to be B0.2 Ω cm2. The ηconc is estimated from Eq. (16.47), with ENernst estimated from Eq. (16.39).

Subzero startup of polymer electrolyte fuel cell Chapter | 16

521

is the largest, accounting for 57%. Since air stoichiometry affects both the limiting current density and the exchange current density, it is expected to be the most sensitive control parameter in adjusting irreversible heat and heatto-water ratio.

16.5.3.2 Temperature difference In addition to the range of average temperature which varies by 100 C, it is interesting to know the magnitude of temperature difference that could develop during subzero startup across the MEA, the cell, and the stack. Several studies examined such difference and their effects on performance through estimation, simulation, and measurement. Using the lumped thermal model for CL, Wang et al. [119] estimated the temperature differences across the CL as: ΔTCL B

δCL iðE0 2 Vcell Þ 2κeff CL

ð16:55Þ

where ΔTCL is the temperature difference across the CL, δCL is the thickness 21 K21). of the CL (m), and κeff CL is the effective thermal conductivity (W m Less than 0.03 K difference across the CL was obtained assuming a cell voltage of 0.5 V at a current density of 0.1 A cm22, for a cathode CL with a thickness from 7.5 to 25 μm and effective thermal conductivity from 0.3 to 3 W m21 K21. Zhou et al. [107] simulated the temperature distribution under a small current density (0.1 A cm22) in a stack with different numbers of cells, as shown in Fig. 16.19A. The cells in the middle had the highest temperature;

FIGURE 16.19 Temperature distribution during the start process: (A) in the stacks with different amounts of cells (220 C, 0.1 A cm22), (B) in the MEA (220 C, 1 A cm22). (A) Adapted from Zhou Y, Luo Y, Yu S, Jiao K. Modeling of cold start processes and performance optimization for proton exchange membrane fuel cell stacks. J. Power Sources. 2014 Feb 1; 247: 738
748. (B) Adapted from M. Khandelwal, S. Lee, MM. Mench, One-dimensional thermal model of coldstart in a polymer electrolyte fuel cell stack. J. Power Sources 172 (2) (2007) 816
830.

522

Fuel Cells for Transportation

the difference from that at the two sides was larger for stacks with more cells. The peaks showed the temperature at the membranes, which were close to the site of the electrochemical reaction. The difference in temperature within the cell was less than 1 C, while that among the cells was about 3 C
4 C. At a higher current density of 1 A cm22, Khandelwal et al. [31] predicted a temperature difference across the MEA as 5 C and between the center cells and end cells as 20 C
25 C (Fig. 16.19B). In a startup test of a 5-cell stack, Lin et al. [32] measured a temperature difference of 5 C
10 C between the center cell and the two end cells, as shown in Fig. 16.20. They also measured a much higher current density in the central area than that in the periphery of the center cell via segmented PCB techniques, most probably due to the higher temperature in the central area. Even small temperature differences across the thickness direction in each cell in a stack can accumulate and cause notable differences in potential. In the cold start of a 30-cell stack from 210 C, Patterson et al. [95] observed a notable asymmetry in the cell potential near the endplates. More cells with the anode side facing the endplate (right side in Fig. 16.21) suffered from a declined potential than that with the cathode facing the endplate (left side). This result was attributed to the thermal-gradient-driven water transport (frost heave as introduced in Section 16.4). As the thermal gradient was opposite near the two endplates, water should move in different directions on the two sides of the stack, leading to more water at the cathode CL in the anode end cells (on the right) than that in the cathode end cells (on the left). It is worth noting that simulations not consider this effect can only predict a nearly symmetrical distribution of temperature and ice volume fraction.

FIGURE 16.20 Subzero startup from 215 C for a 5-cell stack under a potentiostatic hold of 0.2 V/cell: (A) temperature rise profile for the end cells (#1 and #5) and the center cell (#3), (B) current density distribution in cell #3 at the center segment (D4) and the periphery segments (A1, G1, A7, and G7). Adapted from R. Lin, Y. Zhu, M. Ni, Z. Jiang, D. Lou, L. Han, et al. Consistency analysis of polymer electrolyte membrane fuel cell stack during cold start, Appl. Energy. 241 (2019 May 1) 420
432.

523

Subzero startup of polymer electrolyte fuel cell Chapter | 16 1 Baseline

Replaced Cells 25-30 with MEA B

Cell Voltage(V)

0.8 0.6 0.4 0.2 0 0 -0.2

5 Cathode End

10

15 Cell Number

20

25

30 Anode

FIGURE 16.21 Subzero startup of a 30-cell stack from 210 C at 300 mA cm22. Adapted from T. Patterson, J. O’Neill, M. Perry, P. Hagans, C. York, R. Zaffou, DOE final technical report: PEM fuel cell freeze durability and cold start project. 2006, Company: UTC Fuel Cell, Contract Number: DE-FG36-06GO86042, 10.2172/921503.

MEA/cell/stack under repeated large transient temperature differences may suffer from severe thermomechanical stress with an adverse consequence in reliability and durability. Study in this respect is still lacking.

16.5.3.3 Control of thermal boundary conditions Either in simulation or experimental study of a subzero startup using a single cell, the control of thermal boundary conditions is important. Single cell with thick endplates at small currents (nearly isothermal) has been commonly employed to explore the behavior of water during startup and the associated failure mechanism, as in the isothermal WFTs. However, as pointed out by Rice [3], there exists a clear disconnect between published subzero isothermal WFT results and industrially relevant cold starts. For an isothermal WFT in a single cell at 220 C, the highest water fill capacity (7.8 mgH2O cm22 in 2100 seconds) was observed for the lowest applied galvanostatic load of 10 mA cm22 with the lowest initial water content (λinitial 5 2.2). In contrast, the real field startup, as specified in DOE’s target and realized in Toyota’s MIRAI, needs to attain 50% of rated power within 30 seconds. Therefore, a much higher current density, hence higher initial water content, is needed to ensure the subzero startup performance and thermal response of the stack. To resolve this disconnection, a non-isothermal single-cell fixture to emulate cells in a stack is needed. Early in 2006, Gupta et al. [42] pointed out that the subzero startup performance of interior cells in a stack is assisted by adjacent cell heating while the cells near the endplates suffer from the thermal sinks of the endplates. To minimize cost and increase the testing throughput, they developed a quasi-adiabatic single-cell fixture to allow for rapid testing of designs for

524

Fuel Cells for Transportation

enhanced subzero startup performance. Specifically, they separated the endplates from the cell with insulation and added resistive heating pads outside of the flow fields to emulate the adjacent cell heating in a stack. The evolution of cell voltage and cathode temperature of the quasi-adiabatic single cell was found to closely resemble that of the center cell in a 20-cell stack starting from 230 C under an applied load of 200 mA cm22. Using such a quasi-adiabatic cell fixture, Pistono et al. [40] demonstrated that subzero startup performance from 220 C was improved with increased λinitial (3.2 vs. 6.2), and the time required to attain usable power density was reduced from 43.2 to 23.1 seconds. In this work, the output of the heating pads was based on the heat generation rate of the cell but scaled by a multiplier of either 1 or 2 to compensate for the load-dependent fraction of the heat that would be lost, according to Q 5 Ai ð1:48 2 Vcell Þ, where A is the heat adjustment factor. The setting of the heat adjustment factor is somewhat arbitrary. Wang et al. [41] further improved the design to realize the adiabatic thermal boundary condition by incorporating two thermal adjustment boards (TABs) outside the cathodic and anodic flow field plates (BPP), as shown in Fig. 16.22. Two thermal couples were placed on both sides of the TAB. The power of the two heating plates (HPs) was controlled by PID so that T2 closely follows T1 . In this way, there will be a point of minimum temperature within each TAB, and that point is the location where the adiabatic condition is satisfied. Design consideration to minimize the thickness of TAB and to minimize additional thermal mass was discussed. With this adiabatic single cell, a subzero startup from 220 C was realized. Using feedback

FIGURE 16.22 Single-cell structure and its temperature distribution during subzero startup. This schematic is from the author’s laboratory.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

525

from the measured temperatures to control the heating rates, rather than the heat generation rate of the cell, has the advantage of separate control of the anode and cathode sides. It allows for the emulation of asymmetrical thermal boundary conditions for the anode side and cathode side, thereby enabling the exploration of the startup behavior for the cells at different locations in a stack. In addition, the distributed temperature in the cell plane emulating technically relevant large-sized cells can be implemented by using more HPs and thermocouples.

16.6 Subzero startup strategies and techniques The irreversible damages caused by the frozen water and the challenges in starting up the FC from subzero have motivated researchers from diverse backgrounds to delve into the fundamental issues in water and thermal management at low temperatures from different approaches. Continued efforts for decades have led not only to an enhanced understanding of the processes but also to a giant leap in startup capability. Indeed, the strategy has evolved amazingly from keeping warm to an assisted start, and unassisted start. Just in 2002, Kagami found that a successful unassisted start without external heating was only possible at a temperature above 25 C [120]. In 2010, Toyota already reported that their FCEV was capable of starting from 230 C and delivering 60% of nominal power within 30 seconds [49]. Amamou et al. gave a comprehensive review of the solutions and strategies for the subzero startup of automotive FCs [7]. They found that the keepingwarm strategy would consume more energy than the thaw-at-start strategy just 1 hour after the heating set in at the ambient temperature of 220 C [121], therefore such strategy was only applicable to regions of mildly cold climate. The following will focus on the thaw-at-start strategy, which can be further divided into the unassisted and assisted start. Unassisted start here means using the heat from the FC reaction to warm up the cell from subzero to above zero, while assisted start means using energy external to the FC systems instead.

16.6.1 Unassisted start: a battle between water and thermal management Unassisted start utilizes the heat from the electrochemical reaction to warm up the cell from subzero to above zero before the product water freezes. It is tight competition between the two outcomes of the ORR reaction: product water and liberated heat. According to DOE’s subzero startup targets, the FC system needs to deliver 60% of nominal power within 30 seconds, making subzero startup a highly dynamic process both in water and temperature. Specifically, the water content in MEA rises from relatively dry to nearly fully hydrated, and the temperature jumps from 230 C to above zero within just 30 seconds.

526

Fuel Cells for Transportation

In certain sense, the nature of the battle between the water and the heat is trading space for time. The space here is the water storage capacity, particularly the capacity to hold product water in a non-frozen state in the MEA, which is dominantly in membrane and ionomer. The time here is the window to heat up the cell to above zero before the product water overwhelms the storage space, flows into pores, and freezes stochastically. To achieve high power density with low Pt loading at normal operation, the design trend in the past decades has been moving toward thin membranes and less ionomer content in the CL, resulting in a narrow window for heating in a subzero startup. Therefore, sophisticated tactics to maximize the heat generation rate per unit of product water are needed to win the tight battle. Key factors that control subzero startup, especially those pertaining to water and thermal management, can be deduced from lumped thermal and water models. The needed average heat generation rate to raise the cell temperature from subzero by ΔT within the required time Δt is: P mcp ΔT Q_ gen 5 ð16:56Þ Δt where m is the mass of the FC component (kg), cp is the corresponding specific heat capacity (J K21 kg21), and Δt is the startup time (seconds). The constraint from reliability, that is, no supercooled water emerging into the pores of porous media, is that the total amount of product water ΔmH2O (g cm22), which needs to be less than the water uptake potential in the membrane and ionomer: ð Δt MH2O idt , ðδm 1 δCL εm ÞMH2O cf;dry Δλm ð16:57Þ ΔmH2O 5 2F 0 where MH2O is the molar mass of water (g mol21), δm is the PEM thickness (cm), cf;dry is the molar concentration of fixed charge in dry Nafion ionomers (mol cm23), Δλm 5 λsat 2 λinitial , λsat is the saturation water content of ionomer at temperature T (see Eq. (16.11)), λinitial is the water content after the purge, δCL is the CL thickness (cm), and εm is the mass fraction of ionomer in CL. Assuming a constant current density, Eq. (16.57) can be simplified to: iΔt MH2O , ðδm 1 δCL εm Þcf;dry MH2O Δλm ð16:58Þ 2F Dividing the needed heat generation rate in Eq. (16.56) with the limit on the amount of product water in Eq. (16.58) gives a criterion for a successful startup without the appearance of supercooled water: P Q_ gen mcp ΔT . ð16:59Þ i ðδm 1 δCL εm Þcf;dry Δλm 2F ΔmH2O 5

Subzero startup of polymer electrolyte fuel cell Chapter | 16

527

That is, the ratio between the heat and water generation rates [LHS in Eq. (16.59)] should be maximized in operation, the thermal mass should be minimized, and the effective water storage capacity should be maximized by design and operation [RHS in Eq. (16.59)]. Substituting the heat generation rate in Eqs. (16.41), (16.59) becomes: P
 TΔS mcp ΔT 2 ηohmic 2 ηconc 2 ηact . ð16:60Þ 2F 2 2F ðδm 1 δCL εm Þcf;dry Δλm Further substituting the overpotential Eqs. (16.48), (16.49), and (16.50), we get Eq. (16.61), which defines the condition for a reliable and successful subzero startup. P 

 TΔS i i mcp ΔT 2 Rohmic i 2 b log 1 2 2F 2 1 b log . 2F iL i0 ðδm 1 δCL εm Þcf;dry Δλm ð16:61Þ The variables in Eq. (16.61) can be divided into three types according to their attributes, as listed in Table 16.11. It is clear from Eq. (16.61) and Table 16.11 that successful subzero startup depends on three controlling factors: the thermal mass of the cell, the water storage capacity of MEA, and the irreversible heat generation rate. These factors need to be accounted for in both design and operation simultaneously. In the following, lessons learned from laboratory startup studies on the effects of design and operation will be first introduced. Subsequently, Toyota’s remarkable achievement in the field will be explained.

16.6.1.1 Lessons learned from laboratory startup study Most of the laboratory studies employed single cells to explore the effects of design and operation on startup performance. While the results obtained are helpful in identifying the direction or revealing the trend, there are sharp differences from the test using stack. Rice et al. [3] compared the startup results between single cells and short stacks and found that while the lowest successful start temperature for a single cell was 210 C, that for a 10-cell stack was 215 C, and that for a 20-cell stack was 230 C. Obviously, the higher effective thermal mass of the single cell and quasi-adiabatic thermal boundary condition in the stack were the major causes for such differences. The simulation result also indicated that a stack with more than 20 cells was required to overcome the thermal sink of the endplates [31]. Ideally, the evaluation of the startup strategy should be conducted on a stack or FC system. However, the high fuel cost and especially the severe damages from failed startups make the frequent use of stack impractical. Cell fixture capable of emulating the thermal environment of cells at different positions in a stack is a more practical choice, especially in the initial stages. Attempts in this direction can be found in [40
42].

TABLE 16.11 Group of variables with different attributes related to the subzero startup in Eq. (16.61). Attributes

Variables

Description

Ranges

Target

ΔT; Δt

Startup temperature and duration

230 C, 30 s (DOE targets)

Design

mcp

Thermal mass of the fuel cell

0:2
0:6 J K21 cm22

δm cf;dry

Membrane charge concentration

B4:5 3 1026 mol cm22

δCL εm cf;dry

Ionomer charge concentration in CL

B1:8 3 1027 mol cm22

RMEM 1 RBPP 1 RGDL

Components electrical resistance

B0.1
0.3 Ω cm2

i0

CL exchange current density

B4 3 1026 A cm22

i

Startup current density

0.1
1 A cm22

Δλm

Water content change defining the water uptake potential of membrane and ionomer

3
8

c

O2 concentration, affecting both i0 , jD

15
40 mol L21

SRair

Air stoichiometry, affecting both i0 , jD

1
2.5

Operation

Subzero startup of polymer electrolyte fuel cell Chapter | 16

529

Effects of membrane electrode assembly and cell design While MEA design determines the water storage capacity, the cell design, particularly the choice of BPP material, determines the thermal mass. A thick membrane may have a higher water uptake capacity than a thin one depending on the initial water content. Tajiri et al. [9] found that at a current density of 40 mA cm22 and temperature of 220 C, the 60-μm-thick membrane had a higher water storage capacity than that of the 30 μm membrane only when the initial water content λinitial was less than 4. On the other hand, a thin membrane can not only reduce the ohmic resistance and therefore allow for higher current density, but also improve the back-diffusion of product water to the anode. Miao et al. [122] found that adding silica to the cathode CL can increase the water storage capacity. Hiramitsu et al. [123] found that the CL having a higher ionomer content showed a small diffusion overpotential in the ice-blocked CL. Ko et al. [124] found from a 3D transient cold-start model that increasing the ionomer volume fraction in the cathode CL from 0.1 to 0.3, or decreasing the weight ratio of platinum to carbon support from 40% to 20%, significantly extended the cell shutdown time. For the BPP, Ahluwalia [125] found that the low-temperature limit of the successful startup of a stack with stainless steel is 240 C, much lower than that with graphite BPP which is 220 C. It has to be noted that most of the studies used the water storage capacity, typically calculated from the cumulative product water in the WFT, as a metric for the subzero startup performance. Unfortunately, the results from such studies are disconnected from the industrially relevant non-isothermal, dynamic subzero startup [3]. In addition, the effect of the MEA modification on the magnitude of cumulative product water is rather limited and frequently incurs adverse consequences on cost and performance at normal operation. Effects of operation Operation strategy determines the heat generation rate and water storage capacity and affects the effective water storage capacity and effective thermal mass. Two main elements in operation are the initial water content adjustment during shutdown and the way of load control during startup, each having its appropriate range. If the initial water content is too high, there will be inadequate water storage capacity. If it is too low, the initial current, hence the heat generation rate, will be too low and cannot meet the basic power needs of the vehicle. An insufficient startup current leads to a low heat generation rate, while an excessively large current may exceed the water uptake rate of the membrane so that product water flows into the pores of the cCL and freezes there. In addition, a too-high current may trigger the anode dehydration failure mode. Shutdown operation is critical for the success of the subsequent subzero startup. Purging the cell to remove the residual water in the pores of the CL,

530

Fuel Cells for Transportation

MPL, and GDL not only avoids freezing during the cold storage but also reduces the number of nuclei to trigger the release of the product water in the supercooling state during the subzero startup. Reducing the membrane water content below its saturation level at subzero avoids the freezing of bulk-like water inside the membrane and increases its water uptake capacity. The startup operation strategy includes the way of load control, the magnitude of the current/voltage/power, and stoichiometry. Of the four tested voltages (0.2, 0.3, 0.6, and 0.8 V) in a potentiostatic startup, 0.2, 0.3, and 0.6 V can raise the cell temperature from 25 C and 210 C to above zero, while none of them can raise the temperature when the ambient temperature is 215 C and 220 C. Lower voltage has a higher overpotential and higher heat generation rate, hence more effective heating [126]. Amamou et al. [46] found that in the subzero startup of a 36-cell stack from 220 C, the rate of temperature rise under potentiostatic control (0.38 V/cell) is faster than that under galvanostatic control (220 mA cm22), while an algorithm-controlled strategy performs the best. The difference comes from the difference in the heat generation rate of the three controls. Qing et al. [127] found that in a subzero startup from 212.5 C, the maximum power mode had a larger heat generation rate than the constant voltage mode at 0.3 V, which in turn was larger than the constant current mode at 0.15 A cm22. For a faster and more robust unassisted start, Ahluwalia and Wang [125] and Silva et al. [128] suggested operating the stack near short-circuit conditions, which maximizes hydrogen consumption and increases the production of irreversible heat. Although this method is effective for warming up the FC, short-circuiting the stack can be very dangerous and may cause severe damage to cell components. Additionally, no electricity is available for the vehicle. Restricting the supply of reactants is particularly effective in maximizing the irreversible heat generation rate per unit of product water. Two patents from Ballard [129,130] disclosed the effectiveness of reactant starvation on the subzero startup in 2001 and 2002. In 2005, Ballard demonstrated, in a startup of a 10-cell stack from 27 C, that doubling the current and halving the air stoichiometry from around 5 to around 2 nearly doubled the temperature rise rate [45]. In 2009, Toyota reported that their FCEV can achieve startup from 230 C within 30 seconds if the air stoichiometry is further reduced to around or even below 1 [49]. Operation strategies at the system level are also important for a practical subzero startup in FCEVs. Schießwohl et al. [131] found that the startup time for the system to reach 50% power from a temperature of 26 C was reduced from 10 minutes to 210 seconds by just deactivating the coolant. Deactivation of the coolant reduces the effective thermal mass of the stack. Several groups studied the effectiveness of exothermal catalytic combustion between H2 and O2 on Pt on the subzero startup, either by coating the tube or chamber with the catalyst ahead of the stack where the mixture of H2

Subzero startup of polymer electrolyte fuel cell Chapter | 16

531

and O2 is supplied and kept out of the explosive limits (4
74.2 vol% in O2) [132], or by utilizing the anode and cathode catalyst directly with H2 concentration as high as 40% (VH2/Vtotal) [133,134]. No explosion occurred in the latter case due to the radical quenching property of the microchannel reactor inside FC. While catalytic combustion can fully convert the chemical energy in the reactants into heat, the safety concern on the allowable mixing ratio between H2 and O2 in the tube or chamber and the possible freezing of the product water inside the FC limit their applicability in a vehicle.

16.6.1.2 Validation in the field: Toyota’s achievement Built on the lessons learned in the research community over the decades, Toyota developed subzero startup technology that cleared the DOE’s targets and implemented it in their commercial FCEV MIRAI in 2015. The technology addressed the three controlling factors, that is, the thermal mass of the cell, the water storage capacity of MEA, and the heat generation rate, from both design and operation. To reduce the thermal mass, metal (titanium) BPPs are used. To adjust the water content during the shutdown, a purge technique utilizing the cell impedance monitoring is developed. As shown in Fig. 16.23, the water content is maintained in a suitable range, which is low enough to secure the water storage capacity and high enough to allow for a large initial current. Onboard circuitry is developed to monitor the HFR, which is an indicator of the PEM water content. To maximize the concentration and activation overpotential during the startup, aggressive restriction of air supply is employed. Air SR close to or even below 1.0 is used to operate the cell in O2-depleted or O2-starved conditions to enable a sharp increase in concentration and activation overpotential at low current density, thereby maximizing heat

FIGURE 16.23 Water content adjustment for Toyota’s subzero startup.

532

Fuel Cells for Transportation

generation rate per unit product water. It is worth noting that air starvation is close to the limit to which the heating power of the reactants and the utilization of the water storage capacity can be maximized. Toyota’s engineers further demonstrated their in-depth understanding of the system by separating the startup operation into two stages, each with its own priorities. The first stage increases the stack temperature from subzero to above zero within 30 seconds while delivering minimal power (B5 kW) for the vehicle control. The cell operates near 0.1 V to maximize the heat generation rate. The second stage further warms up the stack while fulfilling the driving needs. A lower limit for the cell voltage is prescribed with a priority on meeting the load response. The level of the restriction of the air supply is adjusted by balancing the need to quickly warm up the stack and the need to control the H2 concentration in the cathode exhaust to be below 4% to ensure safety. The remarkable success of subzero startups in Toyota’s FCEV builds on the understanding of both academia and industry accumulated over decades, as well as on their innovation and engineering ingenuity. It achieved the targets of DOE for subzero startup well in advance and marked a milestone in the commercialization of FC vehicles. Nevertheless, there is still room to improve either in startup performance, or in safety, reliability, and durability. Stainless steel and graphite-based BPPs have a much higher specific heat capacity and will limit the lowest temperature of a successful startup. The air SR close to 1 means a strong distribution of current, product water, and temperature distribution from the inlet to the outlet, which could cause degradation due to thermal-induced stress or local freezing of product water. Cell-tocell variations may notably increase the H2 content in the exhaust gases and cause a safety concern. Further efforts to resolve these issues are needed.

16.6.2 Assisted start: decoupling thermal from water management Assisted-start strategy simplifies the problem by decoupling the thermal management from the water management; that is, it heats the stack using energy external to the FC system (other than the H2 in the H2 tank) to above zero and then starts the electrochemical reaction. Examples include heating the coolant [129], heating the inflow [130], embedding electric heaters close to or inside the cell [135], heat pump [130], all using direct current (DC). While such a strategy is more reliable and robust in that it effectively avoids the freezing of the product water, it is inefficient in terms of both time and energy, as it generally depends on the heating of coolant or the periphery parts and relies on the heat conduction to transmit heat into the cores of the stack. A new type of assisted-start strategy utilizes the alternating current (AC) from external power sources to heat up the cell with the ionic ohmic loss in the membrane. Since AC heats the cell right from the core, it has high efficiency in both time and energy while keeping the advantages of DC heating

Subzero startup of polymer electrolyte fuel cell Chapter | 16

533

FIGURE 16.24 Principles of alternate hydrogen pump method. Adapted from J. Wen, D. Si, S. Wang, H. Ding, C. Li, K. Ono, et al. Alternate hydrogen pump method enables start-up from 2 30 C for graphite-bipolar-plate proton exchange membrane fuel cells, J. Electrochem. Soc. 166 (14) (2019) F1112.

in high reliability and robustness. Zhang et al. reviewed the subzero startup of FCs and LIB using AC techniques [134]. Kurz et al. [133] applied AC at a frequency of 1 kHz to a cell in N2/N2 atmosphere and raised the stack temperature from 210 C to 40 C within 148 seconds at a current density of 1.2 A cm22. Such high frequency was needed as the electrode was under blocking conditions. Only high-frequency AC can bypass the electrical double layer in the CL and generate a meaningful magnitude of the current. The requirement of N2 atmosphere and high-frequency power sources limits its applicability in a vehicle. To address such issues, Zhang [33] proposed to introduce H2 on both sides of the cell and utilize an AHP to heat up the cell, as shown in Fig. 16.24. The frequency on the order of 1 Hz was adequate to heat up a cell [33] and stack [34] from 230 C. A 20-cell short stack with graphite-based BPPs was heated up from 230 C to 0 C within 80 seconds using a homemade alternating power of 3.6 kW, as shown in Fig. 16.25. The duration can be easily shortened with an AC power source having higher power. It is expected that the combination of assisted start with AHP and unassisted start with Toyota’s air restriction technique can meet the DOE’s subzero startup targets for the stack made of graphite-based BPPs.

16.7 Directions for further study Meta-analysis of the literature shows a resurgence of papers and patents on subzero startup of the PEFC, possibly driven by the need to expand FCEVs from the temperate into the cold regions, and from passenger cars to commercial vehicles, especially HDV. The higher durability requirement of

534

Fuel Cells for Transportation 30 T1: Near the end plate T2: Central single cell (gas outlet)

20

T3: Between T1 and T2 T4: Central single cell (gas intel)

T/ °C

10 0 -10 -20 -30

0

50

80 s 100 Time/ s

150

200

FIGURE 16.25 Subzero startup result of graphite bipolar plate fuel cell short stack based on alternating hydrogen pump method (λinitial 5 5; 5 slpm; 6 0.8 A cm22; 0.5 Hz). Adapted from D. Si, J. Ji, H. XU, S. Wang, C. Wang, J. Zhang, -30 C start-up of a PEFC stack based on alternate hydrogen pump method, J. Automot. Saf. Energy 12 (4) (2021) 604.

HDV allows the stacks made of graphite-based BPP to become a strong competitor against that of metal BPP. The much larger thermal mass of graphitebased BPP demands a further improvement in the subzero startup capability, which should be both reliable for each startup and durable for repeated startups. However, there are concerns about whether some research directions at the academy match the demands of the industry. As pointed out by Rice [3] in a review in 2021, there is a disconnect between the WFT and industrially relevant startups. While WFT has been valuable in the early days to understand the startup failure mechanism, it is not a suitable test to explore the startup techniques. Some papers still report damages caused by residual water on the FC, even though such effects have been recognized and effective measures have been known for years. Some attempts at seemingly novel technologies for subzero startups are impractical and do not have an advantage over the technologies already implemented in commercial vehicles. Another notable issue is how to treat the subject of supercooled water. While some researchers stress its paramount importance, we dispute its industrial relevance because of its inherently low reliability. The complex behavior of supercooled water in porous composite material is of profound academic interest, yet any startup method based on the utilization of supercooled water is impractical in field application. Supercooled water in the pores is something to be avoided, rather than to be exploited. Areas needing further study include the subzero startup capability of cell design using a new generation of materials, the reliable startup of stacks made of graphite-based BPP, and the degradation modes other than freeze/ thaw. Regarding materials, the effects of ultralow loading of Pt, the contribution of the mesopores in the porous accessible carbon to the water storage

Subzero startup of polymer electrolyte fuel cell Chapter | 16

535

capacity, and the effects of thin/reinforced membranes on the effective water storage capacity need to be examined. For the subzero startup of a stack with graphite-based BPP, assisted start with AHP is worthy of being implemented at the vehicle level. A combination of AHP assisted start and air restriction unassisted start is expected to meet the subzero startup target effectively (within 30 seconds from 230 C), economically (less than 5 MJ energy consumption), and reliably (no dependence on supercooled water). Energy and power requirements of the onboard battery and adaptation of onboard power circuitry need to be examined. The optimal transition from AHP to air restriction needs to be explored, for which a simulation study could prove to useful. Much work is needed to expose the degradation modes of a repetitive successful startup, where the severe damages from the massive freezing are eliminated. Possible degradation from low humidity and low water content operation needs to be explored. Issues from air restriction operation, for example, intercell differences (uneven thermal boundary condition, nonuniform gas allocation), intra-cell distribution (large variation in current density, product water, and temperature from inlet to outlet), and possible degradation from the H2O2 that tends to be generated at low voltage (B0.1 V) need to be studied. To separate different modes, a test fixture with adjustable compression is needed. The study of subzero startups is costly with respect to time, energy, and materials. The startup itself should be less than 1 minute, yet the purging and cooling down take hours. Startup failure tends to cause severe damage to the core components of the cell or the stack, limiting the repeated use of the same test object. To mitigate these issues, single-cell test fixtures capable of emulating distributed and asymmetric thermal boundary conditions of largesized cells at different locations in a stack, as well as being capable of warming up and cooling down quickly, need to be developed.

16.8 Summary The study on the subzero startup of PEFC has come a long way. What seemed to be an insurmountable task in 2000 was accomplished when the Toyota’s passenger FCEV, commercialized in 2015, achieved unassisted startup from 230 C, exceeding DOE’s targets. Such an achievement is built on the insight accumulated over decades from the four categories of experimental study: the freeze/thaw cycling to understand the damages from water freezing, the characterization of the states of water in the PFSA membrane to uncover the water storage capacity, the WFT to explore the failure mechanism, and the startup test to examine the interplay between water and thermal management. This chapter addresses the subzero startup of the FCEV from the perspective of water and thermal management at low temperatures. It first presents an overview of the subzero experiment, with an emphasis on the remarkable

536

Fuel Cells for Transportation

advances in experimental techniques and milestones in understanding. The four categories of subzero experiments, the major test fixtures, the typical test procedures, and featured characterization techniques are summarized in tables and flowcharts. The chapter then presents the study of the damages to the FCs in two subzero scenarios. One scenario is called freeze/thaw cycling in its narrow sense, where there is only residual water in the FC or its components. The other scenario is the repetitive subzero startup where current is drawn and water is produced at subzero. The first scenario played an important role in the early days to reveal the severity of the damage if excessive residual water is left inside the FC and to motivate subsequent efforts to resolve the issue. The second scenario is more relevant for developing practical startup techniques and hence needs to be focused on in further studies. Next, the chapter summarizes the current understanding of the states and behavior of water at subzero, including the states of water in the FC, the water transport dynamics, and the contribution and limitation of the WFTs. Following that, it summarizes the thermal properties and behavior at subzero, including the temperature-dependent properties, the thermal management issues as seen from a lumped model, and the discussion of heat source, temperature difference, and thermal boundary conditions. After that, the chapter compares the two types of thaw-at-start subzero startup strategies, that is, unassisted start, which is essentially a battle between water and thermal management, and assisted start, which effectively decouples thermal from water management. Finally, the chapter comments on the directions for further study. For the unassisted start, the possible long-term degradation from distributions, such as current, product water, and temperature, needs to be examined. For the assisted start using AHP, its implementation in real vehicles is recommended.

16.9 Exercise questions 1. Please compare the two concepts in Section 16.4.1: supercooled water and water with melting point depression in the small-sized pore. 2. In Section 16.4.3, the cumulative product water is an indicator of the water storage capacity of the MEA. Please calculate the maximum water storage capacity (mg cm22) of the membrane, the CL, and the whole MEA. Estimate the maximum amount of heat generated (J cm22) associated with the MEA water storage capacity. Assuming all the product heat is absorbed by MEA and the product water, estimate how many degrees the MEA (together with the product water) can be heated up. Discuss the scenarios when the cumulative product water is greater or smaller than the water storage capacity of the MEA. For calculation, the thickness of the CL and membrane are 5 and 15 μm, respectively. The dry density of the ionomer is 1.98 g cm23. The dry ionomer concentration is

Subzero startup of polymer electrolyte fuel cell Chapter | 16

537

1.8 3 1023 mol cm23. The mass fraction of dry ionomer in the CL is 0.3. The saturated ionomer water content is set to 14, and the initial ionomer water content is set to 6. The porosity of the CL is assumed to be 0.5. Refer to Fig. 16.3 and Table 16.8 for estimating the average heat capacity of the MEA and water. The cell startup voltage is set to 0.1 V. 3. In Section 16.6.1, the unassisted-start technique by Toyota is introduced as a remarkable subzero startup technique. Please analyze the possible safety and degradation issues using this technique.

References [1] Depart of Energy. 3.4 Fuel cells, The US. Fuel Cell Technologies Office Multi-Year Research, Development, and Demonstration Plan (2017). Available from: https://www. energy.gov/sites/default/files/2017/05/f34/fcto_myrdd_fuel_cells.pdf. [2] T. Yoshida, K. Kojima, Toyota MIRAI fuel cell vehicle and progress toward a future hydrogen society, Electrochem. Soc. Interface 24 (2) (2015) 45. [3] C.A. Rice, Subzero automotive fuel cells: water fill tests vs cold-starts, J. Electrochem. Soc 168 (4) (2021) 044513. [4] A. Pesaran, G. Kim, T. Markel, K. Wipke, Fuel Cells Vehicle Systems Analysis (Fuel Cell Freeze Investigation), National Renewable Energy Lab. (NREL), Golden, CO, 2005. [5] H. Meng, B. Ruan, Numerical studies of cold-start phenomena in PEM fuel cells: a review, Int. J. Energy Res 35 (1) (2011) 2
14. [6] Z. Wan, H. Chang, S. Shu, Y. Wang, H. Tang, A review on cold start of proton exchange membrane fuel cells, Energies. 7 (5) (2014) 3179
3203. [7] A.A. Amamou, S. Kelouwani, L. Boulon, K. Agbossou, A comprehensive review of solutions and strategies for cold start of automotive proton exchange membrane fuel cells, IEEE Access 4 (2016) 4989
5002. Aug 2. [8] Y. Luo, K. Jiao, Cold start of proton exchange membrane fuel cell, Prog. Energy Combust. Sci 64 (2018) 29
61. Jan 1. [9] K. Tajiri, C.Y. Wang, Cold start of polymer electrolyte fuel cells, Modeling and Diagnostics of Polymer Electrolyte Fuel Cells, Springer, New York, 2009, pp. 89
128. [10] A.K. Srouji, M.M. Mench, Freeze damage to polymer electrolyte fuel cells, Polymer Electrolyte Fuel Cell Degradation, Academic Press, Waltham, 2011, pp. 293
333. Aug 27. [11] K.Y. Song, H.T. Kim, Effect of air purging and dry operation on durability of PEMFC under freeze/thaw cycles, Int. J. Hydrog. Energy 36 (19) (2011) 12417
12426. [12] R.C. McDonald, C.K. Mittelsteadt, E.L. Thompson, Effects of deep temperature cycling on Nafions 112 membranes and membrane electrode assemblies, Fuel Cell 4 (3) (2004) 208
213. [13] E. Cho, J.J. Ko, H.Y. Ha, S.A. Hong, K.Y. Lee, T.W. Lim, et al., Effects of water removal on the performance degradation of PEMFCs repetitively brought to , 0 C, J. Electrochem. Soc 151 (5) (2004) A661. [14] S. Kim, M.M. Mench, Physical degradation of membrane electrode assemblies undergoing freeze/thaw cycling: micro-structure effects, J. Power Sources 174 (1) (2007) 206
220. [15] S. Kim, B.K. Ahn, M.M. Mench, Physical degradation of membrane electrode assemblies undergoing freeze/thaw cycling: diffusion media effects, J. Power Sources 179 (1) (2008) 140
146.

538

Fuel Cells for Transportation

[16] E. Endoh, N. Onoda, Y. Kaneko, Y. Hasegawa, S. Uchiike, Y. Takagi, et al., Membrane degradation mitigation of PEFC during cold-start application of the radical quencher Ce31-, ECS Electrochem. Lett 2 (10) (2013) F73. [17] T.F. Fuller, J. Newman, Water and thermal management in solid-polymer-electrolyte fuel cells, J. Electrochem. Soc 140 (5) (1993) 1218. [18] E.L. Thompson, T.W. Capehart, T.J. Fuller, J. Jorne, Investigation of low-temperature proton transport in Nafion using direct current conductivity and differential scanning calorimetry, J. Electrochem. Soc 153 (12) (2006) A2351. [19] T.J. Dursch, G.J. Trigub, J.F. Liu, C.J. Radke, A.Z. Weber, Non-isothermal melting of ice in the gas-diffusion layer of a proton-exchange-membrane fuel cell, Int. J. Heat. Mass. Transf 67 (2013) 896
901. Dec 1. [20] M. Saito, K. Hayamizu, T. Okada, Temperature dependence of ion and water transport in perfluorinated ionomer membranes for fuel cells, J. Phys. Chem. B 109 (8) (2005) 3112
3119. [21] K.G. Gallagher, B.S. Pivovar, T.F. Fuller, Electro-osmosis and water uptake in polymer electrolytes in equilibrium with water vapor at low temperatures, J. Electrochem. Soc 156 (3) (2008) B330. [22] Y. Tabe, M. Saito, K. Fukui, T. Chikahisa, Cold start characteristics and freezing mechanism dependence on start-up temperature in a polymer electrolyte membrane fuel cell, J. Power Sources 208 (2012) 366
373. Jun 15. [23] Y. Tabe, K. Yamada, R. Ichikawa, Y. Aoyama, K. Suzuki, T. Chikahisa, Ice formation processes in PEM fuel cell catalyst layers during cold startup analyzed by cryo-SEM, J. Electrochem. Soc 163 (10) (2016) F1139. [24] Y. Aoyama, K. Suzuki, Y. Tabe, T. Chikahisa, T. Tanuma, Effect of wettability of microporous layer on microscopic water transport phenomena in PEFC, ECS Trans 64 (3) (2014) 527. [25] K. Tajiri, Y. Tabuchi, F. Kagami, S. Takahashi, K. Yoshizawa, C.Y. Wang, Effects of operating and design parameters on PEFC cold start, J. Power Sources 165 (1) (2007) 279
286. [26] Y. Ishikawa, T. Morita, K. Nakata, K. Yoshida, M. Shiozawa, Behavior of water below the freezing point in PEFCs, J. Power Sources 163 (2) (2007) 708
712. [27] Y. Ishikawa, M. Shiozawa, M. Kondo, K. Ito, Theoretical analysis of supercooled states of water generated below the freezing point in a PEFC, Int. J. Heat. Mass. Transf 74 (2014) 215
227. Jul 1. [28] P. Oberholzer, P. Boillat, Local characterization of PEFCs by differential cells: systematic variations of current and asymmetric relative humidity, J. Electrochem. Soc 161 (1) (2013) F139. [29] S. Wang, Y. Sun, F. Huang, J. Zhang, Freezing site of super-cooled water and failure mechanism of cold start of PEFC, J. Electrochem. Soc 166 (13) (2019) F860. [30] T. Bono, M. Kizaki, H. Mizuno, Y. Nonobe, T. Takahashi, T. Matsumoto, et al., Development of new Toyota FCHV-adv fuel cell system, SAE Int. J. Engines 2 (1) (2009) 948
954. [31] M. Khandelwal, S. Lee, M.M. Mench, One-dimensional thermal model of cold-start in a polymer electrolyte fuel cell stack, J. Power Sources 172 (2) (2007) 816
830. [32] R. Lin, Y. Zhu, M. Ni, Z. Jiang, D. Lou, L. Han, et al., Consistency analysis of polymer electrolyte membrane fuel cell stack during cold start, Appl. Energy 241 (2019) 420
432. May 1.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

539

[33] J. Wen, D. Si, S. Wang, H. Ding, C. Li, K. Ono, et al., Alternate hydrogen pump method enables start-up from 2 30 C for graphite-bipolar-plate proton exchange membrane fuel cells, J. Electrochem. Soc 166 (14) (2019) F1112. [34] D. Si, J. Ji, H. XU, S. Wang, C. Wang, J. Zhang, 30 C start-up of a PEFC stack based on alternate hydrogen pump method, J. Automot. Saf. Energy 12 (4) (2021) 604. [35] M. Cappadonia, J.W. Erning, S.M. Niaki, U. Stimming, Conductance of Nafion 117 membranes as a function of temperature and water content, Solid. State Ion 77 (1995) 65
69. Apr 1. [36] S. Ge, C.Y. Wang, In situ imaging of liquid water and ice formation in an operating PEFC during cold start, Electrochem. Solid-State Lett 9 (11) (2006) A499. [37] S. Ge, C.Y. Wang, Characteristics of subzero startup and water/ice formation on the catalyst layer in a polymer electrolyte fuel cell, Electrochim. Acta 52 (14) (2007) 4825
4835. [38] P. Oberholzer, P. Boillat, R. Siegrist, R. Perego, A. K¨astner, E. Lehmann, et al., Coldstart of a PEFC visualized with high resolution dynamic in-plane neutron imaging, J. Electrochem. Soc 159 (2) (2011) B235. [39] J. Biesdorf, A. Forner-Cuenca, M. Siegwart, T.J. Schmidt, P. Boillat, Statistical analysis of isothermal cold starts of PEFCs: impact of gas diffusion layer properties, J. Electrochem. Soc 163 (10) (2016) F1258. [40] A.O. Pistono, C.A. Rice, Automotive subzero cold-start quasi-adiabatic proton exchange membrane fuel cell fixture: design and validation, Molecules. 25 (6) (2020) 1410. [41] S. Wang, Y. Zhang, H. Xu, J. Zhang, An adiabatic cell as a model for stack in PEFC cold start study, Int. J. Hydrog. Energy 46 (61) (2021) 31391
31399. [42] N. Gupta, C. York, J. Needham, P. Hagans, Platform for rapid prototyping of pem fuel cell designs with enhanced cold-start performance and durability, ECS Trans 1 (6) (2006) 383. [43] R. Alink, D. Gerteisen, M. Oszcipok, Degradation effects in polymer electrolyte membrane fuel cell stacks by sub-zero operation—an in situ and ex situ analysis, J. Power Sources 182 (1) (2008) 175
187. [44] M. Oszcipok, M. Zedda, D. Riemann, D. Geckeler, Low temperature operation and influence parameters on the cold start ability of portable PEMFCs, J. Power Sources 154 (2) (2006) 404
411. [45] J. St-Pierre, J. Roberts, K. Colbow, S. Campbell, A. Nelson, PEMFC operational and design strategies for sub zero environments, J. N. Mater. Electrochem. Syst 8 (3) (2005) 163
176. [46] A. Amamou, M. Kandidayeni, L. Boulon, S.J. Kelouwani, Real time adaptive efficient cold start strategy for proton exchange membrane fuel cells, Appl. Energy 216 (2018) 21
30. Apr 15. [47] A.O. Pistono, C.A. Rice, Subzero water distribution in proton exchange membrane fuel cells: effects of preconditioning method, Int. J. Hydrog. Energy 44 (39) (2019) 22098
22109. [48] L. Liphardt, K. Suematsu, G. Grundmeier, Kinetic studies of cathode degradation on PEM fuel cell short stack level undergoing freeze startups with different states of residual water and current draws, Int. J. Hydrog. Energy 46 (5) (2021) 4399
4406. [49] K. Sekizawa, N. Kitamura, K. Manabe, Y. Nonobe, M. Kizaki, K. Kojima, Recent advances in TOYOTA FCHV-adv fuel cell system, ECS Trans 33 (1) (2010) 1947.

540

Fuel Cells for Transportation

[50] G.M. Rı´os, J. Schirmer, C. Gentner, J.J. Kallo, Efficient thermal management strategies for cold starts of a proton exchange membrane fuel cell system, Appl. Energy 279 (2020) 115813. Dec 1. [51] K. Tajiri, C.Y. Wang, Y. Tabuchi, Water removal from a PEFC during gas purge, Electrochim. Acta 53 (22) (2008) 6337
6343. [52] G.S. Hwang, H. Kim, R. Lujan, R. Mukundan, D. Spernjak, R.L. Borup, et al., Phasechange-related degradation of catalyst layers in proton-exchange-membrane fuel cells, Electrochim. Acta 95 (2013) 29
37. Apr 15. [53] X. Xie, M. Zhu, S. Wu, C. Tongsh, X. Sun, B. Wang, et al., Investigation of mechanical vibration effect on proton exchange membrane fuel cell cold start, Int. J. Hydrog. Energy 45 (28) (2020) 14528
14538. [54] C. Li, D. Si, Y. Liu, J. Zhang, Y. Liu, Water management characteristics of electrospun micro-porous layer in PEMFC under normal temperature and cold start conditions, Int. J. Hydrog. Energy 46 (19) (2021) 11150
11159. [55] T.J. Dursch, M.A. Ciontea, C.J. Radke, A.Z. Weber, Isothermal ice crystallization kinetics in the gas-diffusion layer of a proton-exchange-membrane fuel cell, Langmuir. 28 (2) (2012) 1222
1234. [56] T.J. Dursch, G.J. Trigub, R. Lujan, J.F. Liu, R. Mukundan, C.J. Radke, et al., Icecrystallization kinetics in the catalyst layer of a proton-exchange-membrane fuel cell, J. Electrochem. Soc 161 (3) (2013) F199. [57] F. Teocoli, A. Paolone, O. Palumbo, M.A. Navarra, M. Casciola, A. Donnadio, Effects of water freezing on the mechanical properties of Nafion membranes, J. Polym. Sci. Part. B: Polym. Phys 50 (20) (2012) 1421
1425. [58] J. Biesdorf, P. Oberholzer, F. Bernauer, A. Kaestner, P. Vontobel, E.H. Lehmann, et al., Dual spectrum neutron radiography: identification of phase transitions between frozen and liquid water, Phys. Rev. Lett 112 (24) (2014) 248301. [59] Y. Otsuki, Y. Tamada, S. Inoue, K. Shigemasa, T. Araki, Measurement of solidification heat from supercooled water freezing during PEFC cold start and visualization of ice distribution, Int. J. Hydrog. Energy 45 (31) (2020) 15600
15610. [60] D. Ko, S. Doh, D.I. Yu, H.S. Park, M.H. Kim, The change of water distribution in porous media of the polymer electrolyte membrane fuel cell after freeze/thaw cycles, Fuel Cell 18 (4) (2018) 413
421. [61] Takagi Yasuo, Research on the characteristics of performance deterioration caused by freeze-starting of polymer electrolyte fuel cells and the application of diagnostic technology (Special feature: prediction, observation, and evaluation of PEFC characteristics), J. Fuel Cell Technol 9 (2) (2009) 82
88. [62] H. Liu, D. Si, H. Ding, S. Wang, J. Zhang, Y. Liu, Cold start capability and durability of electrospun catalyst layer for proton exchange membrane fuel cell, Int. J. Hydrog. Energy 46 (19) (2021) 11140
11149. [63] S. Ge, C.Y. Wang, Cyclic voltammetry study of ice formation in the PEFC catalyst layer during cold start, J. Electrochem. Soc 154 (12) (2007) B1399. [64] F. Wu, B. Chen, Y. Yan, Y. Chen, M. Pan, Degradation of silicone rubbers as sealing materials for proton exchange membrane fuel cells under temperature cycling, Polymers. 10 (5) (2018) 522. [65] C. Lee, W. Merida, Gas diffusion layer durability under steady-state and freezing conditions, J. Power Sources 164 (1) (2007) 141
153. [66] Y. Lee, B. Kim, Y. Kim, X. Li, Degradation of gas diffusion layers through repetitive freezing, Appl. Energy 88 (12) (2011) 5111
5119.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

541

[67] M. Oszcipok, D. Riemann, U. Kronenwett, M. Kreideweis, M. Zedda, Statistic analysis of operational influences on the cold start behaviour of PEM fuel cells, J. Power Sources 145 (2) (2005) 407
415. [68] Q. Yan, H. Toghiani, Y.W. Lee, K. Liang, H. Causey, Effect of sub-freezing temperatures on a PEM fuel cell performance, startup and fuel cell components, J. Power Sources 160 (2) (2006) 1242
1250. [69] F. Wu, B. Chen, M. Pan, Degradation of the sealing silicone rubbers in a proton exchange membrane fuel cell at cold start conditions, Int. J. Electrochem. Sci. 15 (2020) 3013
3028. Apr 1. [70] J. St-Pierre, D.P. Wilkinsor, S. Knights, M.L. Bos, Relationships between water management, contamination and lifetime degradation in PEFC, J. N. Mater. Electrochem. Syst 3 (2) (2000) 99
106. [71] E. Endoh, Development of highly durable PFSA membrane for PEMFC under high temperature and low humidity conditions, ECS Trans 25 (35) (2010) 293. [72] M. Onoda, S. Okuzono, Y. Matsuura, K. Nishimura, Y. Hirose, B. Zhou, et al., Effect of residual water contents in MEA on performance degradation caused by repeated startup of frozen cell in PEFCs, ECS Trans 25 (1) (2009) 295. [73] K. Han, B.K. Hong, S.H. Kim, B.K. Ahn, T.W. Lim, Influence of anisotropic bending stiffness of gas diffusion layers on the degradation behavior of polymer electrolyte membrane fuel cells under freezing conditions, Int. J. Hydrog. Energy 36 (19) (2011) 12452
12464. [74] E. Cho, J.J. Ko, H.Y. Ha, S.A. Hong, K.Y. Lee, T.W. Lim, et al., Characteristics of the PEMFC repetitively brought to temperatures below 0 C, J. Electrochem. Soc 150 (12) (2003) A1667. [75] J. Hou, B. Yi, H. Yu, L. Hao, W. Song, Y. Fu, et al., Investigation of resided water effects on PEM fuel cell after cold start, Int. J. Hydrog. Energy 32 (17) (2007) 4503
4509. [76] J.P. Sabawa, A.S. Bandarenka, Degradation mechanisms in polymer electrolyte membrane fuel cells caused by freeze-cycles: investigation using electrochemical impedance spectroscopy, Electrochim. Acta 311 (2019) 21
29. [77] R. Bradean, H. Haas, K. Eggen, C. Richards, T. Vrba, Stack models and designs for improving fuel cell startup from freezing temperatures, ECS Trans 3 (1) (2006) 1159. [78] Y. Naganuma, K. Manabe, H. Imanishi, Y. Nonobe, Development of system control for rapid warm-up operation of fuel cell, SAE Int. J. Altern. Powertrains 1 (1) (2012) 365
373. [79] Y. Zhang, K.H. Liu, M. Lagi, D. Liu, K.C. Littrell, C.Y. Mou, et al., Absence of the density minimum of supercooled water in hydrophobic confinement, J. Phys. Chem. B 113 (15) (2009) 5007
5010. [80] E.B. Moore, V. Molinero, Structural transformation in supercooled water controls the crystallization rate of ice, Nature. 479 (7374) (2011) 506
508. [81] L. Guildner, D.P. Johnson, F.E. Jones, Vapor pressure of water at its triple point. Journal of Research of the National Bureau of Standards, Sect. A Phys. Chem 80 (3) (1976) 505. [82] D.M. Murphy, T. Koop, Review of the vapour pressures of ice and supercooled water for atmospheric applications, Q. J. R. Meteorol. Soc.: A J. Atmos. Sci., Appl. Meteorol. Phys. Oceanogr. 131 (608) (2005) 1539
1565. [83] J. Hrubý, V. Vinš, R. Mareš, J. Hykl, J. Kalova´, Surface tension of supercooled water: no inflection point down to 2 25 C, J. Phys. Chem. Lett 5 (3) (2014) 425
428. [84] C.A. Angell, Insights into phases of liquid water from study of its unusual glass-forming properties, Science. 319 (5863) (2008) 582
587.

542

Fuel Cells for Transportation

[85] C.L. Jackson, G.B. McKenna, The melting behavior of organic materials confined in porous solids, J. Chem. Phys 93 (12) (1990) 9002
9011. [86] K. Malek, M. Eikerling, Q. Wang, Z. Liu, S. Otsuka, K. Akizuki, et al., Nanophase segregation and water dynamics in hydrated Nafion: molecular modeling and experimental validation, J. Chem. Phys 129 (20) (2008) 204702. [87] W.B. Hillig, Measurement of interfacial free energy for ice/water system, J. Cryst. Growth 183 (3) (1998) 463
468. [88] A. Bogdan, M. Kulmala, N. Avramenko, Reduction of enthalpy of fusion and anomalies during phase transitions in finely divided water, Phys. Rev. Lett 81 (5) (1998) 1042. [89] K. Jiao, X. Li, Three-dimensional multiphase modeling of cold start processes in polymer electrolyte membrane fuel cells, Electrochim. Acta 54 (27) (2009) 6876
6891. [90] A.Z. Weber, J. Newman, Transport in polymer-electrolyte membranes: I. Physical model, J. Electrochem. Soc 150 (7) (2003) A1008. [91] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell model, J. Electrochem. Soc 138 (8) (1991) 2334. [92] K. Jiao, X. Li, Water transport in polymer electrolyte membrane fuel cells, Prog. Energy Combust. Sci 37 (3) (2011) 221
291. [93] M. Siegwart, R. Woracek, J.I. Marquez Damian, A.S. Tremsin, V. Manzi-Orezzoli, M. Strobl, et al., Distinction between super-cooled water and ice with high duty cycle timeof-flight neutron imaging, Rev. Sci. Instrum 90 (10) (2019) 103705. [94] S. Kim, M.M. Mench, Investigation of temperature-driven water transport in polymer electrolyte fuel cell: thermo-osmosis in membranes, J. Membr. Sci 328 (1
2) (2009) 113
120. [95] T. Patterson, J. O’Neill, M. Perry, P. Hagans, C. York, R. Zaffou, DOE final technical report: PEM fuel cell freeze durability and cold start project, Company: UTC Fuel Cell (2006). Available from: https://doi.org/10.2172/921503. Contract Number: DE-FG3606GO86042. [96] S. Motupally, A.J. Becker, J.W. Weidner, Diffusion of water in Nafion 115 membranes, J. Electrochem. Soc 147 (9) (2000) 3171. [97] A. Kusoglu, A.Z. Weber, New insights into perfluorinated sulfonic-acid ionomers, Chem. Rev 117 (3) (2017) 987
1104. [98] L. Yao, J. Peng, J.B. Zhang, Y.J. Zhang, Numerical investigation of cold-start behavior of polymer electrolyte fuel cells in the presence of super-cooled water, Int. J. Hydrog. Energy 43 (32) (2018) 15505
15520. [99] K. Wu, X. Xie, B. Wang, Z. Yang, Q. Du, J. Xuan, et al., Two-dimensional simulation of cold start processes for proton exchange membrane fuel cell with different hydrogen flow arrangements, Int. J. Hydrog. Energy 45 (35) (2020) 17795
17812. [100] S. Kim, M.M. Mench, Investigation of temperature-driven water transport in polymer electrolyte fuel cell: phase-change-induced flow, J. Electrochem. Soc 156 (3) (2009) B353. [101] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, Wettability and capillary behavior of fibrous gas diffusion media for polymer electrolyte membrane fuel cells, J. Power Sources 194 (1) (2009) 433
444. [102] J. Biesdorf, P. Stahl, M. Siegwart, T.J. Schmidt, P. Boillat, When size matters: active area dependence of PEFC cold start capability, J. Electrochem. Soc 162 (10) (2015) F1231. [103] E.J. Langham, B.J. Mason, The heterogeneous and homogeneous nucleation of supercooled water, Proc. R. Soc. London. Ser. A. Math. Phys. Sci 247 (1251) (1958) 493
504. [104] K. Tajiri, Y. Tabuchi, C.Y. Wang, Isothermal cold start of polymer electrolyte fuel cells, J. Electrochem. Soc 154 (2) (2006) B147.

Subzero startup of polymer electrolyte fuel cell Chapter | 16

543

[105] L. Li, S. Wang, G. Wang, L. Yue, A.R. Takuto, Effect of clamping stress on Cold Start Performance of PEMFC, Energy Procedia 158 (2019) 1744
1749. [106] L. Yao, F. Ma, J. Peng, J. Zhang, Y. Zhang, J. Shi, Analysis of the failure modes in the polymer electrolyte fuel cell cold-start process—anode dehydration or cathode pore blockage, Energies. 13 (1) (2020) 256. [107] Y. Zhou, Y. Luo, S. Yu, K. Jiao, Modeling of cold start processes and performance optimization for proton exchange membrane fuel cell stacks, J. Power Sources 247 (2014) 738
748. Feb 1. [108] A.A. Kulikovsky, The effect of stoichiometric ratio λ on the performance of a polymer electrolyte fuel cell, Electrochim. Acta 49 (4) (2004) 617
625. [109] A. Parthasarathy, S. Srinivasan, A.J. Appleby, C.R. Martin, Temperature dependence of the electrode kinetics of oxygen reduction at the platinum/Nafions interface—a microelectrode investigation, J. Electrochem. Soc 139 (9) (1992) 2530. [110] S.J. Clouser, J.C. Huang, E. Yeager, Temperature dependence of the Tafel slope for oxygen reduction on platinum in concentrated phosphoric acid, J. Appl. Electrochem 23 (6) (1993) 597
605. [111] B.E. Conway, D.F. Tessier, D.P. Wilkinson, Temperature dependence of the Tafel slope and electrochemical barrier symmetry factor, β, in electrode kinetics, J. Electrochem. Soc 136 (9) (1989) 2486. [112] E.L. Thompson, J. Jorne, H.A. Gasteiger, Oxygen reduction reaction kinetics in subfreezing PEM fuel cells, J. Electrochem. Soc 154 (8) (2007) B783. [113] A. Kusoglu, A.M. Karlsson, M.H. Santare, S. Cleghorn, W.B. Johnson., Mechanical response of fuel cell membranes subjected to a hygro-thermal cycle, J. Power Sources 161 (2006) 987
999. [114] Japan Stainless Steel Association. Stainless Steel Data Book (2013). ISBN:4526045217. [115] S. Kobe, Characteristic of Titanium, (Online Product Info) (2022). Available from: https://www.kobelco.co.jp/english/titan/characteristic/index.html. [116] Product Information, DuPont Nafion PFSA Membranes N-112, NE-1135, N-115, N-117, NE-1110 Perfluorosulfonic Acid Polymer, NAE101 (2002). [117] A. Schulman, BMC T40(18)-LS die casting replacement grade molding compound : raw material for graphite separators, Material Data on MatWeb (2023). Available from: https://www.matweb.com/search/datasheet.aspx?matguid 5 31830d9d5bfc483892f8a0667 a3840a1&ckck 5 1. [118] Y. Wang, P.P. Mukherjee, J. Mishler, R. Mukundan, R.L. Borup, Cold start of polymer electrolyte fuel cells: three-stage startup characterization, Electrochim. Acta 55 (8) (2010) 2636
2644. [119] F. Kagami, T. Ogawa, Y. Hishinuma, T. Chikahisa, Simulating the performance of a PEFC at a temperature below freezing, Fuel Cell Semin. (2002) 239. [120] A. Amamou, L. Boulon, S. Kelouwani, K. Agbossou, P. Sicard, Thermal management strategies for cold start of automotive PEMFC, 2015 IEEE Veh. Power Propuls. Conf. (VPPC) (2015) 1
6. Oct. 19. [121] Z. Miao, H. Yu, W. Song, L. Hao, Z. Shao, Q. Shen, et al., Characteristics of proton exchange membrane fuel cells cold start with silica in cathode catalyst layers, Int. J. Hydrog. Energy 35 (11) (2010) 5552
5557. [122] Y. Hiramitsu, N. Mitsuzawa, K. Okada, M. Hori, Effects of ionomer content and oxygen permeation of the catalyst layer on proton exchange membrane fuel cell cold start-up, J. Power Sources 195 (4) (2010) 1038
1045.

544

Fuel Cells for Transportation

[123] J. Ko, H. Ju, Effects of cathode catalyst layer design parameters on cold start behavior of polymer electrolyte fuel cells (PEFCs), Int. J. Hydrog. Energy 38 (1) (2013) 682
691. [124] R.K. Ahluwalia, X. Wang, Rapid unassisted-start of polymer electrolyte fuel cell stacks from subfreezing temperatures, J. Power Sources 162 (1) (2006) 502
512. [125] R. Lin, Y. Weng, X. Lin, F. Xiong, Rapid cold start of proton exchange membrane fuel cells by the printed circuit board technology, Int. J. Hydrog. Energy 39 (32) (2014) 18369
18378. [126] Q. Du, B. Jia, Y. Luo, J. Chen, Y. Zhou, K. Jiao, Maximum power cold start mode of proton exchange membrane fuel cell, Int. J. Hydrog. Energy 39 (16) (2014) 8390
8400. [127] R.E. Silva, F. Harel, S. Jemei, R. Gouriveau, D. Hissel, L. Boulon, et al., Proton exchange membrane fuel cell operation and degradation in short-circuit, Fuel Cell 14 (6) (2014) 894
905. [128] J. Roberts, M. van der Geest, J. St-Pierre, D.P. Wilkinson, A. Lee, S. Moroz, inventors; Ballard Power Systems Inc, assignee. Method and apparatus for increasing the temperature of a fuel cell. United States Patent US 6,329,089, 2001 Dec 11. [129] S. Kim, H. Jeong, H. Lee, Cold-start performance investigation of fuel cell electric vehicles with heat pump-assisted thermal management systems, Energy 232 (2021) 121001. Oct 1. [130] L. Li, S. Wang, L. Yue, G. Wang, Cold-start method for proton-exchange membrane fuel cells based on locally heating the cathode, Appl. Energy 254 (2019) 113716. Nov 15. [131] W. Jiang, K. Song, B. Zheng, Y. Xu, R. Fang, Study on fast cold start-up method of proton exchange membrane fuel cell based on electric heating technology, Energies. 13 (17) (2020) 4456. [132] S. Sun, H. Yu, J. Hou, Z. Shao, B. Yi, P. Ming, et al., Catalytic hydrogen/oxygen reaction assisted the proton exchange membrane fuel cell (PEMFC) startup at subzero temperature, J. Power Sources 177 (1) (2008) 137
141. [133] T. Kurz, F. K¨ufner, D. Gerteisen, Heating of low and high temperature PEM fuel cells with alternating current, Fuel Cell 18 (3) (2018) 326
334. [134] Z. Zhang, J. Jian, S. Wang, C. Wang, J. Gong, J. Zhang, Subzero startup of electrochemical power sources using alternating current, Int. J. Green. Energy (2022) 1
22. Aug. [135] R.K. Ahluwalia, X. Wang, A.Z. Phoenix, Startup of PEFC stacks from sub-freezing temperatures. In DOE Workshop on Fuel Cell Operations at Sub-Freezing Temperatures, 2005, pp. 1
2.

Further reading M. Cappadonia, J.W. Erning, U. Stimming, Proton conduction of Nafions 117 membrane between 140 K and room temperature, J. Electroanal. Chem 376 (1
2) (1994) 189
193. K.M. Colbow, M. van der Geest, C.J. Longley, J. M¨uller, J. Roberts, J. St-Pierre, et al., inventors; Ballard Power Systems Inc, assignee. Method and apparatus for operating an electrochemical fuel cell with periodic reactant starvation. United States patent US 6,472,090, Oct 29 2002. H. Guo, S. Sun, H. Yu, L. Lu, H. Xu, Z. Shao, Proton exchange membrane fuel cell subzero start-up with hydrogen catalytic reaction assistance, J. Power Sources 429 (2019) 180
187. Jul 31. L. Hao, P. Cheng, Capillary pressures in carbon paper gas diffusion layers having hydrophilic and hydrophobic pores, Int. J. Heat. Mass. Transf 55 (1
3) (2012) 133
139. Y.S. Kim, L. Dong, M.A. Hickner, T.E. Glass, V. Webb, J.E. McGrath, State of water in disulfonated poly (arylene ether sulfone) copolymers and a perfluorosulfonic acid copolymer

Subzero startup of polymer electrolyte fuel cell Chapter | 16

545

(Nafion) and its effect on physical and electrochemical properties, Macromolecules. 36 (17) (2003) 6281
6285. M. Luo, J. Zhang, C. Zhang, C.S. Chin, H. Ran, M. Fan, et al., Cold start investigation of fuel cell vehicles with coolant preheating strategy, Appl. Therm. Eng 201 (2022) 117816. Jan 25. Y. Luo, B. Jia, K. Jiao, Q. Du, Y. Yin, H. Wang, et al., Catalytic hydrogen
oxygen reaction in anode and cathode for cold start of proton exchange membrane fuel cell, Int. J. Hydrog. Energy 40 (32) (2015) 10293
10307. R. Mukundan, J. Davey, R. Lujan, J. Spendelow, Y.S. Kim, D.S. Hussey, et al., Performance and durability of PEM fuel cells operated at sub-freezing temperatures, ECS Trans 16 (2) (2008) 1939. J. Sargianis, Out of plane expansion of PFSA membranes for fuel cell applications. 2010 May. (Doctoral dissertation, University of Delaware). E. Schießwohl, T. von Unwerth, F. Seyfried, D. Br¨uggemann, Experimental investigation of parameters influencing the freeze start ability of a fuel cell system, J. Power Sources 193 (1) (2009) 107
115.

This page intentionally left blank

Chapter 17

Solid oxide fuel cells for vehicles Haoyu Li , ThomasJae Garcia and Min Hwan Lee Department of Mechanical Engineering, University of California, Merced, CA, United States

17.1 Overview of fuel cell Due to rising carbon emissions, concern over global warming is constantly increasing. According to a recent report by the Intergovernmental Panel on Climate Change, the average temperature of our planet will reach 1.5 C higher than preindustrial levels by the year 2030 [1]. Concerted efforts by governments and industries are currently underway to cope with the unprecedented risk and achieve carbon net zero as soon as possible by shifting our fossil fuel-based energy ecosystem into a renewable energy-based one. Since transportation accounts for nearly a third of carbon emissions and its contribution grows rapidly, intensive efforts have been made to electrify power trains for vehicles mainly in the form of batteries or fuel cells. For the past few years, we have observed a significant change in the market landscape in the passenger car market. Triggered by the recent overwhelming success of Tesla, the demand for battery-powered vehicles surged. However, due to the limited scalability of batteries, it is challenging to apply them to large-scale transportation such as ships and aircraft. While batteries store energy by keeping metal ions within the cell, fuel cell systems store energy (fuel) in a fuel tank, which makes fuel cell-based systems intrinsically far more scalable. In addition to excellent scalability, hydrogen-based systems have a superior gravimetric energy density. Hydrogen has an energy density of B120 MJ kg21 approximately three times higher than gasoline or diesel and B100 times higher than that of Li-ion batteries. Even if we add the weight of a hydrogen tank for a fairer comparison, they (the H2 tank and H2 gas at 700 bar) are still more than five times lighter than a battery of the same energy content. For long-distance and large-scale vehicles, batteries are not 

Equal contribution.

Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00008-3 © 2023 Elsevier Ltd. All rights reserved.

547

548

Fuel Cells for Transportation

a viable option for the main energy supply due to their low energy density despite their high energy conversion efficiency. Fuel cells prove promising in this regard, as their systems provide energy densities significantly higher than batteries (per both volume and weight). However, solid oxide fuel cells (SOFCs) have not been widely explored for transportation compared to lowtemperature fuel cells due to their high-temperature operation and resultant drawbacks such as slow start-up and poorer cell durability. But the advantages of using SOFCs such as high efficiency and fuel flexibility are attractive in niche applications. In this chapter, we discuss the intrinsic advantages and representative applications of SOFCs for transportation, current technological issues, and recent efforts to address these challenges.

17.2 Solid oxide fuel cells for transportation 17.2.1 Solid oxide fuel cells: advantages and shortcomings A SOFC is a type of fuel cell operating at a high temperature usually ranging from 600 C to 1000 C [2]. In an H2-fed SOFC, the hydrogen oxidation reaction (Eq. 17.1) occurs at the anode while the oxygen reduction reaction (ORR; Eq. 17.2) occurs at the cathode. H2 1 O22 -H2 O 1 2e2

ð17:1Þ

1 O2 1 2e2 -O22 2

ð17:2Þ

A high operating temperature is required mainly to increase the oxygen ion conductivity of the usual electrolytes of SOFCs (zirconia or ceria doped with heteroatoms) to a level where the cell achieves a reasonable performance. This high-temperature operation brings several intrinsic advantages. First, the cell does not necessitate expensive noble metals as catalyst materials. Cheap and abundant metals and their oxides can be used instead. Second, a high operating temperature renders a generally high efficiency in fuel-to-electricity conversion due to fast kinetics in both electrochemical reactions and charge transport. When the waste heat is used as well, the overall efficiency for combined heat and power (CHP) is even higher. Stateof-the-art SOFCs achieve an efficiency of B60% for electricity and up to B85%
90% for CHP [3]. Third, for the same reason, the fuel for SOFCs is not limited to pure H2. They can run on various fuels such as natural gas, alcohol, LPG, diesel, and ammonia because the cell operational temperature provides the energy for internal reforming [4,5]. Carbon poisoning, the usual poisoning mechanism of the anode surface in low-temperature fuel cells such as the proton-exchange membrane fuel cell (PEMFC), is not a concern in SOFCs. While PEMFCs require gas purification due to CO poisoning, CO

Solid oxide fuel cells for vehicles Chapter | 17

549

becomes a fuel in a SOFC [6]. Lastly, while PEMFCs need additional modules for water management and active cooling, the overall balance of plant of SOFC systems is simpler and easier to fit in a limited space. However, SOFCs have been significantly less explored for vehicular applications compared to PEMFCs. Most fuel cells employed for current vehicular applications are PEMFCs [7]. Representative commercialized fuel cell vehicles—Hyundai Nexo and Toyota Mirai—are also PEMFC-based [8]. This is mainly thanks to their rapid start-up capability (due to their low operational temperature of B80 C), superior resilience to mechanical shock (due to the use of polymer and metal-based materials as the backbone), and high power density. On the other hand, SOFCs have several disadvantageous characteristics to overcome such as slow start-up, brittleness of cell components, and vulnerability to thermal cycle-induced cracking and delamination [9].

17.2.2 Leveraging advantages and overcoming shortcomings The fuel flexibility of SOFCs provides an important advantage for longrange, large-scale vehicles [6], such as ships, locomotives, and aircraft. This is because it is currently a challenge to load a large amount of hydrogen economically. Due to the low volumetric density of H2 gas, the hydrogen fuel for these large-scale vehicles needs to be liquefied in a vacuum-insulated tank by lowering the temperature down to 2253 C, which is currently not an economical approach. While there are recent efforts to store hydrogen in a chemically bound form as liquid organic hydrogen carrier (LOHC), the cost-competitiveness issue still persists along with technological immaturity [10]. The capability of taking a more energy-dense fuel such as ammonia, ethanol, and natural gas as the fuel will enlarge the usability of SOFCs for a wider range of vehicles, especially those for long-range and large-scale transportation. However, the intrinsic sluggishness in both start-up and load response makes it difficult to use SOFCs as the sole power source of any kind of vehicle. To address this, SOFCs need to be combined with batteries, gas turbines, or ICEs. The fuel flexibility is particularly beneficial for implementing a hybrid system with a combustion-based system (such as SOFC-gas turbine and SOFC-ICE systems) because they can use the same fuel used for the combustion-based energy generator. Lastly, to counteract the drawbacks caused by using ceramic-based components, various cell configurations have been tried. Metal-supported SOFCs (MS-SOFCs) provide significant mechanical integrity and resilience to mechanical shock thanks to the high thermal conductivity and ductility of metal support [11]. In addition, the so-called wet impregnation process (i.e., infiltration) will also enable us to achieve structural resilience to thermal gradient-induced mechanical fractures. We can use an electrode backbone

550

Fuel Cells for Transportation

that has a coefficient of thermal expansion (CTE) close to that of the electrolyte to minimize the mechanical stress from temperature changes while the infiltrated nanoparticles boost the electrode performance. The cell design in a tubular shape will also help suppress mechanical fracture caused by the thermal cycle. Each case is discussed below.

17.2.3 Cell configurations As shown in Fig. 17.1, two different configurations of SOFCs are possible depending on the ionic species the electrolyte conducts: O22 conducting electrolyte-based (namely, SOFC-O) and H1 conducting electrolyte-based cells (SOFC-H) [12]. In an SOFC-O, H2 reacts with the O22 species transported from the cathode side, producing steam at the anode. On the other hand, in an SOFC-H, H1 travels through the electrolyte to produce steam at the cathode by reacting with O2. SOFC-Hs are advantageous in achieving a higher power density at a given operational temperature mainly because of the higher ionic conductivity of proton-conducting electrolytes compared to their oxygen ion-conducting counterparts for SOFC-Os. Additionally, in SOFC-Hs, the fuel at the anode is not diluted during operation because the steam is formed at the cathode, enabling to maintain a high-power output regardless of load condition.

17.3 Fuel types As discussed above, SOFCs can run on various kinds of hydrocarbon fuels, alcohols, and ammonia as well as hydrogen. In SOFCs, the internal reforming of short-chain hydrocarbons (such as natural gas and ethanol) and ammonia is possible whereas long-chain hydrocarbons (such as gasoline and

FIGURE 17.1 Schematic diagrams of SOFC-H (left) and SOFC-O (right).

Solid oxide fuel cells for vehicles Chapter | 17

551

diesel) will need an external reforming, for which the waste heat from SOFC operation can be utilized [13]. In this section, we briefly discuss the advantages of each fuel type and methods of using them for SOFCs.

17.3.1 Hydrogen Hydrogen has been considered the main fuel of fuel cells including SOFCs. It has a high gravimetric energy density (120.2 MJ kg21), but an extremely low density (0.09 kg m23 at 1 bar), making it challenging to store and transport them in an economically viable way [10]; a table of density and energy density by fuel type is provided in Table 17.1. To address this, hydrogen is mostly compressed at a high pressure ranging from 200 to 700 bar for ground transportation. Most commercial fuel cell vehicles including Toyota Mirai (700 bar), Hyundai Nexo (700 bar), and Hyundai Xcient (350 bar) carry compressed hydrogen gas. It consumes . 4.1% of the stored energy content for compressing the gas and an additional 1.8%
3.6% for cooling the container (to avoid overheating of the vessel during compression) [14]. For larger vehicles such as ships and submarines, however, it is challenging to meet the range requirement by this approach, considering the usually limited space available for hydrogen tanks. A readily available approach is to liquefy hydrogen. The liquefaction will increase the volumetric density of hydrogen by 780 times (70.8 kg m23) at the same pressure of 1 bar, which is almost twice the density achieved by pressurizing hydrogen gas at 700 bar (38 kg m23). However, the liquefaction requires a cryogenic temperature of 2252.9 C, and it costs roughly a third of the energy content of H2 themselves being liquefied. Since the tank is not designed for high pressure, hydrogen should be allowed to leak through a relief valve [14]. The resulting loss of hydrogen from this “boil-off” is another major drawback of liquid hydrogen. A hybrid approach of compressing and cooling hydrogen is also explored. This cryo-compression method still necessitates vacuum-based thermal insulation but imposes a less stringent requirement in both pressure (compared to the pure compression gas-based storage) and temperature (compared to the sole liquefaction approach) [14]. These liquefaction approaches have hydrogen release kinetics, comparable to that of compressed hydrogen. However, other methods of storing hydrogen, such as the formation of metal hydrides, chemisorption of hydrogen on nanostructured carbon matrix, and hydrogenation of chemical compounds (namely, LOHCs), are slow and energy-consuming in releasing hydrogen [10,15,16].

17.3.2 Ammonia As discussed above, a liquefied form of fuel is the preferred option for bulk transport because gaseous fuels hardly achieve economically viable energy density [17]. An attractive alternative to circumvent the challenges with

TABLE 17.1 Storage pressure, density, and energy densities of various fuels. Fuel type

Storage temperature [ C]

Storage pressure [bar]

Density [kg m23]

Gravimetric energy density [MJ kg21]

Volumetric energy density [GJ m23]

Liquid ammonia (NH3)

20/ 2 33

10/1

610.3

18.7

11.4

Liquid hydrogen (H2)

2253

1

71.0

120.2

8.54

Gaseous hydrogen (H2)

20

700

50.0

120.2

6.02

Methanol (CH3OH)

20

1

787.0

22.7

17.9

Ethanol (C2H5OH)

20

1

785.0

29.7

23.3

Gaseous methane (CH4)

20

250

193.6

55.6

10.8

Liquid natural gas (LNG)

2162

1

468.1

53.6

25.1

Solid oxide fuel cells for vehicles Chapter | 17

553

hydrogen liquefaction is to use ammonia (NH3). Unlike hydrogen, ammonia can be easily liquefied either by increasing pressure to B10 bar at 20 C or decreasing temperature to 233 C at atmospheric pressure [17]. Liquefied ammonia has a volumetric energy density of 12.7 MJ L21, even higher than that of liquid hydrogen (8 MJ L21). Ammonia also contains 17.8 wt.% of hydrogen, which is greater than methanol (12.6 wt.%) by B40% [18]. In addition, ammonia can be readily synthesized through the well-established Haber
Bosch process using hydrogen and nitrogen as the feedstock [17]. While most hydrogen used in this process is currently obtained from steam methane reformation, it is possible to utilize electrolysis to cleanly produce hydrogen, making the formation of ammonia an entirely green process [17]. Ammonia is one of the most produced chemicals to date with an annual production of B175 million tons globally, equivalent to a market value of B$70 billion [17]. It is currently widely utilized in agricultural applications due to its presence in fertilizer. This is beneficial on a system-wide level, as the infrastructure for storing and transporting ammonia is already in place, while a hydrogen infrastructure would need to be built from scratch [19]. A large benefit to using ammonia as a fuel is that it can be directly applied to SOFCs without needing a separate reforming process because ammonia is cracked at a temperature similar to the usual operational temperature of SOFCs [20]. If one chooses to use ammonia directly as the fuel, the explicit energy-consuming step of cracking ammonia back to hydrogen can be avoided. The nickel already present in SOFC anodes acts as an efficient catalyst for ammonia cracking, that is, splitting the ammonia into nitrogen and hydrogen. Furthermore, all the byproducts of the electrode reactions are gaseous phase without causing anode surface poisoning [19]. However, strict safety measures, especially in terms of human health, need to be implemented before deploying public infrastructure for ammonia, since human exposure to high concentrations of ammonia can cause respiratory and ocular damage, possibly even resulting in death [21]. Both SOFC-O and SOFC-H configurations can be considered for ammonia-based SOFCs as well. In both configurations, NH3 first turns into N2 and H2 at the anode via a catalytic thermal decomposition. In the case of SOFC-H, the ammonia at the anode is not diluted during operation because the steam is formed at the cathode [20], enabling it to maintain a high-power output regardless of load conditions. Additionally, since steam (H2O) is formed at the cathode side, there is no risk of forming NOx, an environmentally vicious molecule. In the case of SOFC-O, on the other hand, a rational choice of anodic catalyst highly selective against NOx formation needs to be made [22]. The performance of ammonia-fueled SOFCs has proved excellent. Ma et al. earlier demonstrated a maximum power density of 355 mW cm22 at 700 C by using ammonia as fuel, which is comparable to that of a hydrogenfed SOFC (371 mW cm22 at 700 C) [2]. The degradation of the stack also

554

Fuel Cells for Transportation

needs to be considered, as the stack should have 1%
2% degradation over 1000 hours. Another study showed that the difference in power densities between ammonia and hydrogen was approximately 15% lower for ammonia at 900 C, which is believed to be due to the lower partial pressure of hydrogen at the anode [23]. However, this difference in power density is not large enough to reevaluate the use of ammonia as SOFC fuel. A real-life deployment of an ammonia-based SOFC system is also ongoing; a supply vessel for Equinor’s offshore operations named “Viking Energy” will be modified to run on an ammonia-fed SOFC by 2024 [24]. AlHamed et al. proposed to use a SOFC-based hybrid system using ammonia for railway applications [20]. The presented system is an ammonia-fed ITSOFC integrated with a gas turbine Brayton cycle and ammonia-organic Rankine cycle, which is combined to reuse waste heat. They suggested an optimum operating condition that can achieve a maximum energy efficiency of 79.88%.

17.3.3 Hydrocarbons Direct internal reforming SOFCs (DIR SOFCs) have been widely explored due to the abundance and cost-competitiveness of these fuels. The usual hydrocarbons being considered for this are natural gas and biogas. Natural gas is a naturally occurring mixture of hydrocarbon gases (primarily methane), and biogas is a mixture of methane, carbon dioxide, and small amounts of hydrogen sulfide, produced from plants, food waste, agricultural waste, and sewage. DIR SOFCs, therefore, can leverage the current industrial infrastructure and consume organic wastes for good use. One of the major challenges of direct-hydrocarbon SOFCs for wide adoption is the coking (carbon poisoning) on Ni-based anode from hydrocarbon decomposition (Eq. 17.3), which results in a drastic performance degradation [25]. Cn Hm -

m H2 1 nCðsÞ 2

ð17:3Þ

Sulfur poisoning is another major issue of natural gas [26]. Strongly adsorbed sulfur-containing species act as a passivation layer against electrochemical reactions. In a biogas-fed DIR SOFCs, dry reforming of methane (Eq. 17.4) is followed by electrochemical oxidation of CO and H2 (Eq. 17.5). In addition to coking, this approach causes the issue of building thermal stresses within the cell due to this series of an endothermic (dry reforming) and an exothermic reaction (CO and H2 oxidation) [27]. CH4 1 CO2 -2CO 1 2H2

ð17:4Þ

CO 1 H2 1 2O22 -CO2 1 H2 O 1 4e2

ð17:5Þ

Solid oxide fuel cells for vehicles Chapter | 17

555

17.3.4 Alcohols The physical state of alcohol fuels under atmospheric conditions is an undeniable advantage over other kinds of fuels, as liquid fuels can be conveniently transported and stored. Methanol and ethanol are the main candidates for alcohol fuel over higher liquid hydrocarbons, like butane or octane, due to containing less carbon atoms and having a high O/C ratio which lowers the rate of carbon coking. Primarily produced through the reformation of natural gas, methanol is considered a convenient carrier for syngas [28]. A Ni and ceramic composite, which is the state-of-the-art and widely used anode material, easily decomposes methanol into hydrogen-rich gas at the operating temperature of SOFCs using the composite phase as the catalyst (Eq. 17.6). Taking advantage of the easy decomposition for methanol at a Ni-contained anode, direct-fueled methanol SOFCs show little loss compared to the one fueled by a reformed gas mixture of CO and H2 [29]. Additional hydrogen can also be produced in the presence of steam through the water gas shift reaction (Eq. 17.7). CH3 OH2CO 1 2H2

ð17:6Þ

CO 1 H2 O2H2 1 CO2

ð17:7Þ

Ethanol is mainly produced from renewable biomass sources, such as agro-waste, industrial wastes, and agriculture residues [30]. The production of ethanol by renewable sources comprises most of the global biofuel production. Like methanol, a reforming process (Eq. 17.8) is essential before putting it in conventional fuel cell devices. C2 H5 OH 1 H2 O22CO 1 4H2

ð17:8Þ

Furthermore, the Ni-based anode in SOFCs with high operating temperatures has an advantage not only for the direct decomposition of ethanol fuel in the Ni catalyst but also tolerates incoming fuel of lower purity, which reduces the distillation requirement for bioethanol. Ni et al. [31] reviewed the effectiveness of a variety of catalysts for bioethanol reformation and indicated that Ni is considered the best candidate for steam reforming among all the transition metals. It is cheap and used industrially but suffers from high carbon coking [32], especially in the case of ethanol. Qu et al. [33] conducted an experiment on the coke formation rate between various carboncontaining fuel at 750 C and observed a higher coking rate when methane is used as the fuel. On the contrary, Eigenbrodt et al. [34] found that the carbon formation caused by methanol fuel is significantly higher than that by methane. The contradictory result indicates the formation of carbon coking by each mechanism might be facilitated by different thermodynamic conditions. The implementation of bioethanol as the fuel source for a vehicle power source has been intensively researched in recent years. Goncalves et al. [35]

556

Fuel Cells for Transportation

evaluated and compared the feasibility of ethanol and gasoline in Brazil, concluding that ethanol has more economic, environmental, and social advantages over gasoline. Ma et al. [36] designed a SOFC SPU system with bioethanol onboard reforming for vehicular applications. Dogdibegovic et al. [37] investigated the internal reforming of ethanol fuel on metal-supported cells for vehicular applications.

17.4 Applications 17.4.1 Auxiliary power units for heavy-duty trucks There has been an increasing demand for electrical power for safety and convenience features in automobiles. Special types of vehicles (e.g., commercial trucks, military vehicles, recreational vehicles) necessitate a high-power draw for nonpowertrain purposes such as cargo refrigeration, heating, and other electric appliances. The usual method of electrification has been through an alternator linked to the ICE, but with larger demand for onboard electricity, the method necessitates large electricity generators and complicated mechanical systems [38]. A separate electricity-generating module (i.e., auxiliary power units; APU) provides much better flexibility in addressing various power-consuming needs. The idling time of class 8 trucks in the United States is known to be 1500
2500 hours every year, which is equivalent to $5000
9000 per year [39]. An APU will also reduce a significant amount of sunk costs during idling. Fuel cells, along with batteries, have been widely considered for APUs. When a high and stable amount of power is required (e.g., for truck refrigeration unit; TRU), fuel cells are capable of constantly supplying the needed electricity independent of the propulsion system. While idling (when the engine is off), fuel cells can also provide heating and other conveniences for the driver during long-distance travel. Since fuel cells do not involve a moving part, the power supply does not cause noise or vibrations [40]. Delphi is a company that focuses on SOFC-based APU systems, involved not only in heavy-duty trucks but also in recreational vehicles, trucks, trailer refrigeration systems, military vehicles, etc., [41]. They started a collaborative project with BMW for a 5 kW SOFC APU in 1999 and demonstrated their first-generation model operating on a passenger car in 2001 [42]. Later in 2003, they reported their generation 2 SOFC APU system for a gasolinebased ICE car; the system weight and volume were reduced by a factor of 4 compared to their first proof of concept [38]. They used the metal-supported cell configuration with tape-casted anode and electrolyte and screen-printed cathode. Since conventional sintering of cell components necessitates high temperatures .1000 C, a process called vacuum plasma spraying was used to sinter the ceramic layers, not to cause complications from the metal substrate. In the report [38], they demonstrated a peak power density of

Solid oxide fuel cells for vehicles Chapter | 17

557

0.84 W cm22 at 750 C from a 7 3 7 cm2 cell using a fuel flow rate of 5 lpm (48.5% H2 and 3% H2O), but a lower peak power density of 350 mW cm22 from a 12 3 12 cm2 cell, which they ascribed to a nonoptimized flow distribution. They also presented a set of durability tests from cyclic and continuous operations. After 5 cycles between room temperature and 750 C (each spanning B100 hours), the power density at 0.7 V decreased from 0.45 W cm22 to 0.39 W cm22. The cell maintained B92% of current level after a continuous load at 0.7 V for 1000 hours. In 2009, the German Aerospace Center (DLR) group reported their progress in MS-SOFC systems for APU [43]. A power density of 400 mW cm22 was demonstrated on a reformate gas at 800 C from a large effective cell area of 100 cm2. There have been efforts to develop diesel APUs for trucks. SOFCs can use diesel and gasoline as fuel after reforming at 700 C
800 C, which is close to the temperature needed for SOFC operation, easing the system integration as an APU. The less strict requirements for reformer technology and the compatibility of fuel and temperature are significant advantages of SOFC-based APU systems over other fuel cell-based ones. Particularly, diesel-based APUs are attractive for ease of entry into the truck and leisure vehicle market. A related prototyping work was reported earlier by Webasco AG [44]. Their 1 kW APU stack was designed to have 60 cells, but their reported work is based on measurements from 30 cells. Their prototype APU system proved thermal self-sustainability, operation without external water supply, and tolerance to system transiency. They also demonstrated a stable operation for 4 hours at B700 C without observable degradation in stack performance even with dry reformate. In 2016, the first European SOFC APU for a heavy-duty truck was demonstrated and tested through the so-called DESTA project [39]. Diesel fuel is first fed into a separate reformer. The resultant H2 and CO molecules are introduced to a SOFC cell. The SOFC APU system was integrated into a Volvo class 8 truck to demonstrate a maximum electrical APU power output of 2.9 kW with an overall efficiency of B30%. The start-up time was less than 70 minutes at ambient temperature, and CO2 emissions were reduced by 73% compared to an idling engine. There have been efforts to model SOFC-based APU systems for vehicles [45
48]. Venkataraman et al. reported modeling of an SOFC heat-driven vapor absorption refrigeration system with a focus on its refrigerated truck application, suggesting that using both heat and power from an SOFC achieves a total efficiency up to 80% and removes a significant part of the load from the main diesel engine [49].

17.4.2 Aircraft Currently, the aircraft industry contributes approximately 2% to global CO2 emissions and is expected to continuously rise every year [50]. Along with

558

Fuel Cells for Transportation

the concern over global warming, recent advances in batteries and fuel cells have motivated efforts to electrify aircraft power train as well. The energy densities required for a short-range aircraft range from 750 to 2000 Wh kg21 while the state-of-the-art Li-ion batteries reach a specific energy of B250 W kg21 only [51]. The low energy density and disadvantageous scaling behavior are the primary reasons batteries cannot be used as the main power source in aircraft. The required energy density exponentially increases as the size and expected flying distance of the aircraft increase, leading to a greater need for high energy density devices such as SOFCs. Hydrocarbon turbines can afford a power density of B10 kW kg21, and SOFCs have been shown to approach this value, reaching B7 kW kg21 [52]. However, for aircraft, fuel cells have been more considered as APUs than as the main power train because their power densities are still not sufficient to serve the role [53]. An APU can provide power to assist takeoff/landing operations and provide power while the plane is on the ground. In the air, the APU can be also used as an emergency backup system but is off the majority of the flight [13]. The development of SOFC-based APU systems is motivated by the low efficiency of preexisting APU systems [13]. While records of actual implementation are lacking, there have been several recent reports on thermodynamic analysis of hybrid SOFC-GT (gas turbine) systems for unmanned air vehicles [54
56] and medium airplanes [57]. Waters et al. presented a modeling work of a hybrid SOFC-GT system, suggesting a synergistic effect between gas turbine and SOFC systems for combined propulsion and electric power on aircraft [58]. Incorporating both reformer and fuel cell directly into the flow path of the engine reduces the overall fuel consumption of a highaltitude long-endurance aircraft at a cruise by . 8% when the engine is mounted inside the fuselage. It also increases electrical generation capability by factors of . 2 compared to engine-driven mechanical generators. With regard to recent prototyping efforts, there are companies making strides to produce working APU systems. Siemens Westinghouse has created a 300 kW system [13], while Mitsubishi has created a 200 kW system [59]. Both systems were hybrid SOFC-GT APUs. Due to their capacity to use various types of fuels, SOFCs are advantageous in replacing conventional APUs compared to low-temperature fuel cells. However, their lower energy density compared to conventional APU systems is a drawback; the entire SOFC stack and complementary system weighs more than a conventional APU system.

17.4.3 Maritime applications The shipping industry is a significant contributor to global warming, responsible for 2%
3% of worldwide carbon emissions [6]. In the past decade, carbon emissions due to shipping have increased by nearly 20%. These emissions are due to the use of diesel as the primary fuel source in ships. The International

Solid oxide fuel cells for vehicles Chapter | 17

559

Maritime Organization sets a challenging target of reducing carbon emissions from shipping by at least 50% by 2050, forcing the maritime industry to actively seek energy sources cleaner than natural gas [60]. Baldi et al. recently investigated energy, cost savings, and emission reduction by using an SOFC system as the main power source of ships [6]. Their model suggests that SOFCs can reduce carbon emissions by up to 34%, and that when natural gas is used as the fuel, SOFCs are the most costoptimal solution for carbon emission. Sapra et al. simulated the potential of a SOFC-ICE integration in terms of system efficiency, carbon emissions, load sharing, space and weight considerations, and load response [61]. The study claims that integration with 67:33 power split between the SOFCs and ICE at 750 kW power output affords a promising improvement of efficiency by 5.2%, NOx reductions by B30%, and CO2 reductions by B12% with an increase in size and weight by 1.7 times. More recently, Kistner et al. studied the integration of batteries to supplement the slow load response of SOFCs [62]. A yacht requiring a maximum power of 487 kW with a flexible load shift can be implemented by an SOFC system of 251 kW with a small battery capacity of 129 kWh. With the help of a small supercapacitor, the requirement can be met with significantly smaller SOFC and battery systems; a peak power of 560 kW can be attained by a 195 kW SOFC system when a 49.4 kWh battery and 71 Wh supercapacitor are integrated. These hybrid systems are necessary to compensate for the sluggishness of the SOFC system during start-up or in a high-load scenario. Committing to a specific fuel type is yet another issue, as both storage and cost are large barriers to overcome with hydrogen. Inal et al. compared different fuel cell types for merchant ships smaller than 5 MW [63]. They stressed that hydrogen is not an ideal fuel for ships due to the lack of cheap and safe methods of loading them on board at a needed density. This makes high-temperature fuel cells such as the SOFC and molten carbonate fuel cell a significantly more favorable choice because they can take fuels of decent energy density that are easier to transport, such as hydrocarbons and ammonia. Since natural gas still generates CO2, it is not the eventual option in the long run. For long-term considerations, hydrogen and ammonia are preferred due to their carbon-free emissions. Ammonia has an increased benefit since infrastructure and expertise are already in place for both its transport and storage. In addition, due to the high operating temperatures of SOFC, the heat generated by the SOFC stack can be recovered and then used in the process of evaporating the liquid ammonia into gaseous form, allowing ammonia to be stored as a liquid [18].

17.4.4 Railways Land-based transportation of commodities and passengers is heavily reliant on railways. Nowadays, the dominant engine type for powering locomotive

560

Fuel Cells for Transportation

technology is diesel-based, and thus, a massive amount of greenhouse gases is produced; B1% of global CO2 emission is caused by locomotives [64]. Accompanied with the price hike in diesel fuel, innovations using alternative fuels and higher fuel-to-electric efficiency methods are required. Schroeder and Majumdar earlier performed a feasibility analysis of employing a SOFC-based system for railroad locomotives [65]. They concluded that SOFC-based systems using onboard gasified biodiesel are technically feasible but difficult to justify their economic viability. Recently, the National Fuel Cell Research Center proposed a hybrid SOFC and SOFC-GT [66] system and built a prototype on a railway system showing reduced operating costs compared to a diesel-electric and battery-based system [66]. According to their economic analyses, SOFC-GT systems lead to higher costs for delivering goods per ton-mile than the diesel-electric alternative, but incur lower operating costs compared to a catenary-electric alternative, and significantly lower operating costs compared to the battery electric alternative. More recently, Al-Hamed and Dincer proposed a series of SOFC-GT hybrid systems for rail transportation [67
70]. First, they performed a thermodynamics study for an ammonia-based SOFC-GT system made of two subsystems for a locomotive. In the main subsystem, both the SOFC and gas turbine systems provide electrical thrust to the locomotive. In the other subsystem, the waste heat from SOFC is used to produce heat energy for space cooling, heating, and hot water, along with additional electric power [68]. The two researchers further performed a thermodynamic analysis of SOFC-GT hybrid systems with cascaded Rankine cycles where natural gas is directly fed [69]. The system reaches energy and exergy efficiencies of 85.3% and 83.3%, respectively. Compared to the currently used diesel-based system, the use of methane is found to reduce CO2 emissions by 58.4% and fuel cost by a factor of 3.4. Fig. 17.2 shows a schematic diagram of a hybrid SOFC-GT system, in which the high-quality waste heat is utilized and unused reactants (H2, CO, or CH4) combust harvesting additional power. Recently, a start-up company, Fuel Cell Enabling Technologies (FCET) made a memorandum of understanding (MOU) with NextGenPropulsion, LLC (NGP) where FCET denoted a plan to provide SOFC systems for NGP’s locomotives [71].

17.4.5 Passenger vehicles Petroleum-based ICEs have played the dominant role as the main power source in passenger vehicles for the last B100 years. With recent efforts to be away from carbon-emitting energy sources, batteries and clean fuels are actively pursued. While battery electric vehicles are currently perceived as the most reasonable power source for clean passenger vehicles, there have also been efforts to develop hydrogen-based power trains. As a representative example, BMW Hydrogen 7 (2005
07) ran on a hydrogen ICE that was

Solid oxide fuel cells for vehicles Chapter | 17

561

FIGURE 17.2 A schematic diagram of a hybrid SOFC-GT system.

fueled by liquid hydrogen stored in a B170-L vacuum-insulated tank [72]. However, their low fuel-to-wheel efficiency and the need to address NOx emissions, which are intrinsic to ICEs, diverted the main attention to fuel cells. Among all the fuel cell types, PEMFCs are the most common type of fuel cells for vehicular applications thanks to their fast start-up capability and excellent resiliency to mechanical shock. However, SOFCs have also been developed as the power supply for passenger vehicles. Different from PEMFC-powered vehicles which require a quick start-up response, vehicles powered by SOFCs need a “warm-up” time before they can operate and deliver power. Therefore, a sufficiently sized battery is essential to act as the power source before the SOFC stacks warm up. However, in terms of longdistance travel, SOFC-powered vehicles have a greater advantage. The high volumetric density of the fuel, low purity requirement, and high efficiency all contribute to the advantages of SOFCs over other kinds of fuel cells. The effort by Nissan for the last decade is the most notable. In 2016, Nissan unveiled their first SOFC vehicle that is fueled by bioethanol. It is equipped with a 5 kW fuel cell stack along with a 24 kWh battery-based power train [73]. With 30 L of fuel, it achieves a driving range larger than 600 km. Nissan promotes the use of bioethanol in the context of the “Carbon-Neutral Cycle” where the carbon dioxide generated during the reforming stage in the car will be consumed to grow sugarcane, the main source of bioethanol. Furthermore, ethanol is easier and safer to store and transport than most of the other fuels including hydrogen, and it can be widely used without establishing new infrastructure. Like pure electrical vehicles, a battery is used as the main propulsion system and heater of the SOFC system to warm up. When it reaches its optimized operating

562

Fuel Cells for Transportation

TABLE 17.2 SOFC applications for transportation. Model

Application

Power

System

Delphi CALSTART AeroVironment [41]

Class 8 truck (APU)

5 kW

SOFCDiesel

DESTA project [39]

Heavy-duty truck (APU)

3 kW

SOFCDiesel

Nissan e-NV200 prototype [74]

Passenger car (main)

5 kW

SOFCBattery

Burlington Northern Santa Fe Railway [66]

Railway (main)

0.2
1 MW

SOFC-GT

Siemens Westinghouse [13], Mitsubishi [59]

Aircraft (APU)

200
300 kW

SOFC-GT

temperature, the SOFC system starts to constantly feed electricity back to the battery. A reformer is used to convert ethanol or ethanol-blended water to hydrogen and carbon monoxide at a high temperature, the same as that of SOFC system. A list of representative applications of SOFCs to transportation is listed in Table 17.2.

17.5 Challenges and related efforts 17.5.1 Mechanical integrity Vehicular applications require a fast and frequent start-up. These requirements impose a significant challenge in the cell design of SOFCs. Since fast heating can easily develop a significant thermal gradient in the cell, resulting in cracks and delamination [75], there have been efforts to alleviate this issue through different approaches.

17.5.2 Low-temperature operation An approach partially motivated by this was to reduce the cell operating temperature [76]. For the last few decades, intensive research efforts have been made to lower the operating temperature from a high ( . 800 C) to an intermediate (650 C
800 C) or low-temperature regime (,650 C) [77,78]. An excessively high operating temperature not only limits the use of costcompetitive parts but also makes it disadvantageous in terms of start-up time, heat management, and cell degradation [79]. In particular, a high operating temperature makes thermally induced stresses more significant during thermal cycles, resulting in the facilitated development of mechanical faults.

Solid oxide fuel cells for vehicles Chapter | 17

563

17.5.3 Metal-supported cells Another approach is to use metal as the support of SOFCs. MS-SOFCs afford various advantages for mechanical stability [11,75,80]. First, one can leverage the ductility of a metal. By using metal as the mechanical support of cells, as opposed to ceramic-based support, one can make the cell more resilient against mechanical shock. For stationary applications, anodesupported configurations have been widely employed for facile manufacturability and high performance. However, ceramic-based anodes, usually a composite of Ni and doped ceria (or doped zirconia), are intrinsically brittle and susceptible to mechanical stimuli, making them unsuitable for vehicular applications. Second, metal-based supports enable fast thermal conduction throughout the cell, minimizing the thermal gradient, which in turn reduces thermally induced mechanical stresses within the cell. Along with the low cost of raw materials and manufacturing process [11], these expected advantages have driven researchers to actively develop MS-SOFCs for vehicular applications. Ceres Power, a UK-based SOFC company, initiated the effort of realizing MS-SOFCs in the industry sector based on Steel’s low-temperature SOFC work [81]. They have demonstrated that a low-temperature MS-SOFC achieves a start-up time of ,10 minutes and has an insignificant change in the mean cell voltage even after 2500 thermal cycles between 175 C and 610 C [82]. In 2018, this company partnered with Nissan Motors and The Welding Institute to apply their modular MS-SOFC technology (named SteelCell) to automotive applications. Ceres Power is also known to have formed strategic partnerships with other automotive vendors including Cummins, Honda, and two unnamed ones [83,84]. Recently, Pirou et al. reported a monolithic SOFC stack design, a new metal-supported configuration to overcome the issues of low volumetric power density and specific density, poor resilience against thermal cycling, and mechanical vibrations [85]. With the expected cost-effectiveness and scalability in manufacturing processes, along with a high power density of 5.6 kW L21, the design demonstrated the viability of applying SOFCs to transport applications.

17.5.4 Tubular configuration A tubular configuration is more advantageous than planar counterparts in achieving mechanical stability under thermal cycles although it is less favorable due to its low volumetric power density and difficulty in manufacturing [86,87]. Westinghouse first commercialized tubular SOFCs by using electrochemical vapor deposition in depositing doped zirconia electrolytes [88]. Later, the finding that power density and thermal shock resistance are inversely proportional to the cell diameter led to the invention of so-called

564

Fuel Cells for Transportation

microtubular SOFCs, in which the tube-like cell is sized at a millimeter scale [89]. Thermal shock resistance of microtubular SOFCs is better than the planar counterpart by several orders of magnitude; a ramp rate of 8000K minutes21 is possible in microtubular SOFCs [89], which would make the cell more suitable for vehicular applications. More recently, tubular SOFCs have been reported to have a few centimeters to facilitate cell fabrication by reducing the number of cells in each stack of given power output [90].

17.5.5 Wet impregnation The cathode performance is dependent not only on the catalytic activity of the electrodes but also on geometric factors such as triple phase boundary (TPB) length. While a pure electron conductor would provide a TPB area limited to the interface between the electrode and electrolyte, a mixed ion and electron conductor (MIEC) would enlarge the area where continuous electrochemical reactions occur by a significant margin. However, MIEC cathodes have a CTE commonly much higher than electrolyte materials [91]. To circumvent CTE mismatch-derived issues, the so-called wet impregnation has been widely performed on a backbone with a CTE close to that of electrolyte materials [92]. A wet impregnation (i.e., infiltration) process forms a layer of dispersed nanoparticles of virtually any kind of solid oxides onto a presintered scaffold structure. Using an electrode scaffold such as LaNi12xFeO3 (LNF) will alleviate the risk of mechanical fracture due to the CTE mismatch with electrolyte [91]. However, LNF has poor ionic conductivity and low ORR activity compared to conventional MIECs such as La0.6Sr0.4Co0.2Fe0.8O32δ (LSCF) [93]. While stabilizing the mechanical integrity, other phases of high catalytic activity and ionic conductivity can be coated at the nanoscale to afford an overall increased electrode performance [94]. An infiltration process is performed by drop-casting a solution of metal precursor onto a porous backbone electrode followed by a firing process, which will result in the formation of highly dispersed metal oxide nanoparticles [95]. The firing temperature for the crystallization of the nanoparticle (B800 C) is much lower than the sintering temperature needed for a conventional electrode (B1100 C) [91]. The low firing temperature preserves a high-surface-area geometry for maximized electrode reactions.

17.5.6 Activity degradations As discussed above, repeated thermal cycling from on
off operations is detrimental to the mechanical integrity of multilayer-based cells. On the other hand, the prolonged continuous operation will cause other types of issues

Solid oxide fuel cells for vehicles Chapter | 17

565

such as loss of surface area for reaction, formation of insulating phases, and electrode surface passivation against reactions’ access. These issues are briefly summarized below.

17.5.7 Sintering and loss of active area A high-surface-area geometry is generally advantageous for electrode performance. A larger area will provide more sites available for reactant adsorption. In the case of MIEC-based electrodes, virtually the whole area can be utilized as the reaction site and is not limited to being the adsorption site. However, an electrode of a larger surface area has higher surface energy and thus is less stable because a larger number of atomic species are exposed to the surface, as opposed to being embedded in the lattice. Therefore, a geometry with a high surface area is more susceptible to morphological changes in a direction that lowers surface energy, thus decreasing the surface area [96,97]. This issue is exacerbated if tiny nanoparticles are decorated onto a backbone structure via infiltration. The usual size of nanoparticles is on the order of tens of nanometers, and therefore, even when the cell is operated at a relatively low temperature for an SOFC, infiltrated electrodes undergo severe changes in microstructure during operation. One approach to addressing this issue, atomic layer deposition (ALD)-based surface treatment, has recently attracted attention [98
100]. ALD is a new type of chemical vapor deposition. The ALD process for metal oxide deposition consists of the introduction and purging of metal precursors and an oxygen source for sequential reaction with the sample surface. Since the precursor molecules will react only with a surface (with oxygen-containing functional groups) without forming a chemical reaction with each other, this sequential introduction of precursors will result in reactions in a “self-limiting” fashion, enabling true atomic-scale thickness control even on a highly corrugated, high-aspect-ratio geometry [100]. Li et al. demonstrated that a ceria and yttria overcoat with a nom˚ over a ceria-infiltrated cell is highly effective in supinal thickness of 0.7
1.5 A pressing the thermally induced agglomeration of infiltrated nanoparticles and thus enhances the durability of electrode performance [99].

17.5.8 Dopant segregation The majority of SOFC cathodes are of a perovskite structure (chemical formula: ABO3) [101,102]. Most perovskite-based conventional cathodes have Sr doped in the A-site for electronic conductivity, and the cations in B-sites are transition metals with flexible valence states for catalytic activity [103,104]. Representative examples of air electrodes include (La, Sr)MnO3 (LSM) and (La, Sr)(Co, Fe)O3 (LSCF) where La and Sr are the host and dopant cations in the A-site, respectively. Upon sintering perovskite-based

566

Fuel Cells for Transportation

oxides in an oxidizing environment, A-site dopants (e.g., Sr) tend to segregate to the surface due to their larger size compared to the host cation and subsequently form insulating phases on the surface or rearrange the surface structure. This eventually leads to degraded surface oxygen exchange kinetics and electrochemical performance. This A-site dopant segregation, besides microstructural sintering and surface poisoning, has been regarded as the major factor of electrode performance decay [100,105,106]. Extensive research has been performed to suppress dopant segregation toward the surface. The main strategies can be divided into two categories: bulk engineering and surface engineering. Bulk engineering is usually achieved by controlling the composition during the material synthesis or deposition process [105,107,108] or performing external doping into A- and/ or B-site [109
113], which can stabilize the bulk structure. On the other hand, surface engineering of a perovskite-based electrode is performed by coating the electrode surface with a metal oxide [114
117].

17.5.9 Carbon coking For applications that use fuels containing carbon elements, carbon coking is one of the main concerns. Although the presence of a reformer may block most of the carbon-containing fuel, carbon poisoning is still inevitably occurring through the Boudouard reaction at low temperatures (Eq. 17.12) and the reduction of CO in the presence of H2 through the so-called reverse water gasification (Eq. 17.13). 2CO2CO2 1 CðsÞ ΔH 5 2 172 kJ mol21

ð17:12Þ

CO 1 H2 2H2 O 1 CðsÞ ΔH 5 2 131 kJ mol21

ð17:13Þ

The coking of carbon can be categorized into two different forms, graphite carbon and amorphous carbon. The formation of graphite carbon requires the reactions occur on the active site where the carbon is produced and precipitates on the surface, which results in the reduction of the active site and lowers the cell performance. Not only that, but the graphite is then dissolved into the metal bulk and precipitates as a whisker [118]. The volume expansion caused by the introduction of extra carbon in the bulk could cause serious structural stress [119] and even the failure of the system. At the same time, the amorphous carbon would be formed in the gas atmosphere even without the catalyst. The gaseous phase carbon would either deposit on the electrode or be exhausted as an output. Considering that the carbon tends to accumulate within the electrode pores, the concentration loss due to the decrease in effective flow rate will be the main concern in this case. Many efforts have been made to suppress coking at the anode. The approach is generally twofold. First, one can provide a gas environment thermodynamically less favorable for these poisoning reactions, for example,

Solid oxide fuel cells for vehicles Chapter | 17

567

providing a higher concentration of oxygen and water. However, a higher oxygen activity would reduce the overall cell efficiency and increase the risk of hydrocarbon combustion [120]. Alternatively, one can engineer the anodes by alloying, increasing basicity, or incorporating catalytically active species [120,121]. New oxide-based catalysts are explored as well. Chen et al. developed a robust SOFC that feeds nearly dry methane (3% H2O) for B550 hours at 500 C with no evidence of carbon coking. The high catalytic activity and coking resistance result from the synergistic effect of Ce0.9Ni0.05Ru0.05O2 catalyst [122].

17.6 Conclusion Forced by strong initiatives toward carbon neutrality by most major countries including EU countries, the United States, China, Japan, and Australia, the use of clean fuels for a whole range of transportation sectors recently became a pressing task. In the past few years, we have observed a drastic shift from conventional ICEs to electricity as the main power source for passenger vehicles. It is not difficult to foresee a continuous transition to clean energy systems at a faster rate. Between two different approaches to electrifying vehicle power train—batteries and fuel cells—there is a consensus that fuel cells are intrinsically more suitable for large-scale, long-range transportation while batteries are a better choice for small-scale transportation, the likes of passenger vehicles. While the PEMFC is currently the most widely adopted fuel cell type due to its fast start-up and mechanical stability, there are sectors where the SOFC can be a more reasonable choice—the areas where the high efficiency and fuel flexibility of SOFCs can be leveraged. The use of SOFCs for transportation is in its incipient stage, but many efforts on related research and commercialization are underway mainly for large-scale transportation such as heavy-duty trucks, railways, ships, and aircraft. For use as a main power source, SOFCs are often integrated with other types of power sources such as batteries and combustion-based systems to supplement the sluggish start-up and load response and low power density. SOFCs are most actively probed for ships and railways as the main power train while they are mainly considered as an APU for heavy-duty trucks and aircraft. For the last several decades, there has been a significant advancement in SOFC technology. However, before being widely commercialized in the transportation sector, SOFCs need further improvements mainly in system durability and cost reduction. To address the requirements of transportation, such as fast start-ups, frequent on
off operations, and resilience to mechanical and thermal shock, various approaches are being pursued including a rationale design of cell configurations, electrode surface treatments, and the development of new hybrid materials.

568

Fuel Cells for Transportation

References [1] IPCC. Climate Change 2022, Impacts, Adaptation and Vulnerability, 2022. [2] A.B. Stambouli, E. Traversa, Solid oxide fuel cells (SOFCs): a review of an environmentally clean and efficient source of energy, Renew. Sustain. Energy Rev. 6 (2002) 433
455. [3] T.M. Gur, T.M. G¨ur, Comprehensive review of methane conversion in solid oxide fuel cells: prospects for efficient electricity generation from natural gas, Prog. Energy Combust. Sci. 54 (2016) 1
64. [4] Q. Ma, R. Peng, Y. Lin, J. Gao, G. Meng, A high-performance ammonia-fueled solid oxide fuel cell, J. Power Sources 161 (2006) 95
98. [5] A. Atkinson, et al., Advanced anodes for high-temperature fuel cells, Nat. Mater. 3 (2004) 17
27. [6] F. Baldi, S. Moret, K. Tammi, F. Mare´chal, The role of solid oxide fuel cells in future ship energy systems, Energy 194 (2020) 116811. [7] A.G. Olabi, T. Wilberforce, M.A. Abdelkareem, Fuel cell application in the automotive industry and future perspective, Energy 214 (2021) 118955. [8] R.C. Samsun, L. Antoni, M. Rex, D. Stolten, Deployment status of fuel cells in road transport: 2021 update, 2021. [9] P. Boldrin, N.P. Brandon, Progress and outlook for solid oxide fuel cells for transportation applications, Nat. Catal. 2 (2019) 571
577. [10] P. Preuster, C. Papp, P. Wasserscheid, Liquid Organic Hydrogen Carriers (LOHCs): toward a hydrogen-free hydrogen economy, Acc. Chem. Res. 50 (2017) 74
85. [11] M.C. Tucker, Progress in metal-supported solid oxide fuel cells: a review, J. Power Sources 195 (2010) 4570
4582. [12] W. Wang, D. Medvedev, Z. Shao, Gas humidification impact on the properties and performance of perovskite-type functional materials in proton-conducting solid oxide cells, Adv. Funct. Mater. 28 (2018) 1802592. [13] M.D. Fernandes, et al., SOFC-APU systems for aircraft: a review, Int. J. Hydrog. Energy 43 (2018) 16311
16333. [14] E. Rivard, M. Trudeau, K. Zaghib, Hydrogen storage for mobility: a review, Materials 12 (2019) 1973. [15] B. Sakintuna, F. Lamaridarkrim, M. Hirscher, Metal hydride materials for solid hydrogen storage: a review, Int. J. Hydrog. Energy 32 (2007) 1121
1140. [16] J.L.C. Rowsell, O.M. Yaghi, Strategies for hydrogen storage in metal-organic frameworks, Angew. Chem. Int. Ed. 44 (2005) 4670
4679. [17] D.R. MacFarlane, et al., A roadmap to the ammonia economy, Joule 4 (2020) 1186
1205. [18] S. Giddey, S.P.S. Badwal, C. Munnings, M. Dolan, Ammonia as a renewable energy transportation media, ACS Sustain. Chem. Eng. 5 (2017) 10231
10239. [19] S.S. Rathore, S. Biswas, D. Fini, A.P. Kulkarni, S. Giddey, Direct ammonia solid-oxide fuel cells: a review of progress and prospects, Int. J. Hydrog. Energy 46 (2021) 35365
35384. [20] G. Jeerh, M. Zhang, S. Tao, Recent progress in ammonia fuel cells and their potential applications, J. Mater. Chem. A Mater 9 (2021) 727
752. [21] United States Environmental Protection Agency, Hazards of ammonia releases at ammonia refrigeration facilities, 2001.

Solid oxide fuel cells for vehicles Chapter | 17

569

[22] A. Wojcik, H. Middleton, I. Damopoulos, J. Van herle, Ammonia as a fuel in solid oxide fuel cells, J. Power Sources 118 (2003) 342
348. [23] A. Fuerte, R.X. Valenzuela, M.J. Escudero, L. Daza, Ammonia as efficient fuel for SOFC, J. Power Sources 192 (2009) 170
174. [24] T. Brown, Viking energy to be retrofit for ammonia fuel in 2024, Ammon. Energy Assoc. (2020). [25] K. Wei, et al., Progress in Ni-based anode materials for direct hydrocarbon solid oxide fuel cells, J. Mater. Sci. 53 (2018) 8747
8765. [26] W. Wang, C. Su, Y. Wu, R. Ran, Z. Shao, Progress in solid oxide fuel cells with nickelbased anodes operating on methane and related fuels, Chem. Rev. 113 (2013) 8104
8151. [27] M.A. Abdelkareem, et al., On the technical challenges affecting the performance of direct internal reforming biogas solid oxide fuel cells, Renew. Sustain. Energy Rev. 101 (2019) 361
375. [28] F. Dalena, et al., Methanol production and applications: an overview, Methanol: Sci. Eng. (2018) 3
28. Available from: https://doi.org/10.1016/B978-0-444-63903-5.00001-7. [29] K. Sasaki, K. Watanabe, Y. Teraoka, Direct-alcohol SOFCs: current-voltage characteristics and fuel gas compositions, J. Electrochem. Soc. 151 (2004) A965. [30] A. Taghizadeh-Alisaraei, A. Motevali, B. Ghobadian, Ethanol production from date wastes: adapted technologies, challenges, and global potential, Renew. Energy 143 (2019) 1094
1110. [31] M. Ni, D.Y.C. Leung, M.K.H. Leung, A review on reforming bio-ethanol for hydrogen production, Int. J. Hydrog. Energy 32 (2007) 3238
3247. [32] S. Bepari, D. Kuila, Steam reforming of methanol, ethanol and glycerol over nickel-based catalysts-a review, Int. J. Hydrog. Energy 45 (2020) 18090
18113. [33] J. Qu, et al., Ethylene glycol as a new sustainable fuel for solid oxide fuel cells with conventional nickel-based anodes, Appl. Energy 148 (2015) 1
9. [34] B.C. Eigenbrodt, M.B. Pomfret, D.A. Steinhurst, J.C. Owrutsky, R.A. Walker, Direct, in situ optical studies of Ni-YSZ anodes in solid oxide fuel cells operating with methanol and methane, J. Phys. Chem. C. 115 (2011) 2895
2903. [35] F. de Oliveira Gonc¸alves, E. Savioli Lopes, M. Savioli Lopes, R. Maciel Filho, Evaluation of the feasibility of ethanol and gasoline in solid oxide fuel cell vehicles in Brazil, Int. J. Hydrog. Energy 46 (2021) 36381
36397. [36] S. Ma, X. Hu, Y. Zhao, X. Wang, C. Dong, Design and evaluation of a metal-supported solid oxide fuel cell vehicle power system with bioethanol onboard reforming, ACS Omega 6 (2021) 29201
29214. [37] E. Dogdibegovic, Y. Fukuyama, M.C. Tucker, Ethanol internal reforming in solid oxide fuel cells: a path toward high performance metal-supported cells for vehicular applications, J. Power Sources 449 (2020) 227598. [38] P. Lamp, J. Tachtler, O. Finkenwirth, S. Mukerjee, S. Shaffer, Development of an auxiliary power unit with solid oxide fuel cells for automotive applications, Fuel Cell 3 (2003) 146
152. [39] J. Rechberger, A. Kaupert, J. Hagerskans, L. Blum, Demonstration of the first European SOFC APU on a heavy duty truck, Transp. Res. Procedia 14 (2016) 3676
3685. [40] T. Salameh, SOFC auxiliary power units (APUs) for vehicles SOFC auxiliary power units (APUs) for vehicles, 2015. [41] S.J. Mcphail, J. Kiviaho, B. Conti, The yellow pages of SOFC technology International Status of SOFC deployment, in: 2017 IEA Implementing Agreement Advanced Fuel Cells Annex 32-SOFC Energy and Sustainable Economic Development DOSSIER, 2017.

570

Fuel Cells for Transportation

[42] S. Mukerjee, et al., Solid oxide fuel cell auxiliary power unit—a new paradigm in electric supply for transportation, in: H. Yokokawa, S.C. Singhal (Eds.), Solid Oxide Fuel Cells VII (SOFC VII), 2001, pp. 173
179. [43] P. Szabo, et al., Progress in the metal supported solid oxide fuel cells and stacks for APU, ECS Trans. 25 (2009) 175
185. [44] J. Lawrence, M. Boltze, Auxiliary power unit based on a solid oxide fuel cell and fuelled with diesel, J. Power Sources 154 (2006) 479
488. [45] N. Lu, Q. Li, X. Sun, M.A. Khaleel, The modeling of a standalone solid-oxide fuel cell auxiliary power unit, J. Power Sources 161 (2006) 938
948. [46] Z. Wang, et al., Dynamic evaluation of low-temperature metal-supported solid oxide fuel cell oriented to auxiliary power units, J. Power Sources 176 (2008) 90
95. [47] S. G¨oll, R.C. Samsun, R. Peters, Analysis and optimization of solid oxide fuel cell-based auxiliary power units using a generic zero-dimensional fuel cell model, J. Power Sources 196 (2011) 9500
9509. [48] L. Barelli, et al., Stochastic power management approach for a hybrid solid oxide fuel cell/battery auxiliary power unit for heavy duty vehicle applications, Energy Convers. Manag. 221 (2020) 113197. [49] V. Venkataraman, A.W. Pacek, R. Steinberger-Wilckens, Coupling of a solid oxide fuel cell auxiliary power unit with a vapour absorption refrigeration system for refrigerated truck application, Fuel Cell 16 (2016) 273
293. [50] Z. Ji, M.M. Rokni, J. Qin, S. Zhang, P. Dong, Energy and configuration management strategy for battery/fuel cell/jet engine hybrid propulsion and power systems on aircraft, Energy Convers. Manag. 225 (2020) 113393. [51] A.W. Sch¨afer, et al., Technological, economic and environmental prospects of all-electric aircraft, Nat. Energy 4 (2019) 160
166. [52] J.M. Collins, D. McLarty, All-electric commercial aviation with solid oxide fuel cell-gas turbine-battery hybrids, Appl. Energy 265 (2020) 114787. [53] R. Peters, R.C. Samsun, Evaluation of multifunctional fuel cell systems in aviation using a multistep process analysis methodology, Appl. Energy 111 (2013) 46
63. [54] Z. Ji, et al., Determination of the safe operation zone for a turbine-less and solid oxide fuel cell hybrid electric jet engine on unmanned aerial vehicles, Energy 202 (2020) 117532. [55] Z. Ji, et al., Thermodynamic analysis of a solid oxide fuel cell jet hybrid engine for longendurance unmanned air vehicles, Energy Convers. Manag. 183 (2019) 50
64. [56] D.F. Waters, C.P. Cadou, Engine-integrated solid oxide fuel cells for efficient electrical power generation on aircraft, J. Power Sources 284 (2015) 588
605. [57] R.A. Evrin, I. Dincer, Development and evaluation of an integrated solid oxide fuel cell system for medium airplanes, Int. J. Energy Res. 44 (2020) 9674
9685. [58] D.F. Waters, L.M. Pratt, C.P. Cadou, Gas turbine/solid oxide fuel cell hybrids for aircraft propulsion and power, J. Propuls. Power 37 (2021) 349
361. [59] T. Gengo, Y. Kobayashi, N. Hisatome, T. Kabata, Y. Ando, Progressing steadily, development of high- efficiency SOFC combined cycle system, Mitsubishi Heavy Industries Technical Review 44 (2007) 1. [60] P. Serra, G. Fancello, Towards the IMO’s GHG goals: a critical overview of the perspectives and challenges of the main options for decarbonizing international shipping, Sustainability 12 (2020) 3220. [61] H. Sapra, et al., Integration of solid oxide fuel cell and internal combustion engine for maritime applications, Appl. Energy 281 (2021) 115854.

Solid oxide fuel cells for vehicles Chapter | 17

571

[62] L. Kistner, A. Bensmann, R. Hanke-Rauschenbach, Optimal design of power gradient limited solid oxide fuel cell systems with hybrid storage support for ship applications, Energy Convers. Manag. 243 (2021) 114396. [63] O.B. Inal, C. Deniz, Assessment of fuel cell types for ships: based on multi-criteria decision analysis, J. Clean. Prod. 265 (2020) 121734. [64] U.S. EPA, Fast Facts on Transportation Greenhouse Gas Emissions | Green Vehicle Guide | US EPA, 2019. [65] D.J. Schroeder, P. Majumdar, Feasibility analysis for solid oxide fuel cells as a power source for railroad road locomotives, Int. J. Hydrog. Energy 35 (2010) 11308
11314. [66] A. Ali, P. Ahrend, S. Samuelsen, J. Brouwer, Prototype design and evaluation of hybrid solid oxide fuel cell gas turbine systems for use in locomotives, 2019. [67] K.H.M. Al-Hamed, I. Dincer, Development and optimization of a novel solid oxide fuel cell-engine powering system for cleaner locomotives, Appl. Therm. Eng. 183 (2021) 116150. [68] K.H.M. Al-Hamed, I. Dincer, A new direct ammonia solid oxide fuel cell and gas turbine based integrated system for electric rail transportation, eTransportation 2 (2019) 100027. [69] K.H.M. Al-Hamed, I. Dincer, Natural gas as a transitional solution for railway powering systems: environmental and economic assessment of a fuel cell based powering system, J. Nat. Gas. Sci. Eng. 80 (2020) 103347. [70] K.H.M. Al-Hamed, I. Dincer, A novel integrated solid-oxide fuel cell powering system for clean rail applications, Energy Convers. Manag. 205 (2020) 112327. [71] Fuel Cell Works, Fuel Cell Enabling Technologies, Inc. Announces First Major Customer, Signs Memorandum of Understanding Regarding Purchase of Fuel Cells for Locomotives, 2021. [72] T. Wallner, et al., Fuel economy and emissions evaluation of BMW Hydrogen 7 MonoFuel demonstration vehicles, Int. J. Hydrog. Energy 33 (2008) 7607
7618. [73] N. Reeve, Nissan unveils world’s first solid-oxide fuel cell vehicle, Nissan Mot. Corp. (2016). [74] e-Bio Fuel-Cell | Innovation | Nissan Motor Corporation Official Global Website. [75] K.J. Kim, et al., Micro solid oxide fuel cell fabricated on porous stainless steel: a new strategy for enhanced thermal cycling ability, Sci. Rep. 6 (2016) 22443. [76] D.J.L. Brett, A. Atkinson, N.P. Brandon, S.J. Skinner, Intermediate temperature solid oxide fuel cells, Chem. Soc. Rev. 37 (2008) 1568. [77] A. Pesaran, A. Jaiswal, E.D. Wachsman, CHAPTER 1. Bilayer electrolytes for low temperature and intermediate temperature solid oxide, Fuel Cell—A Review. (2019) 1
41. Available from: https://doi.org/10.1039/9781788012959-00001. in. [78] C. Zhang, Solid oxide fuel cells: low temperature cathodes, Nat. Energy 1 (2016) 16200. [79] M.C. Tucker, Durability of symmetric-structured metal-supported solid oxide fuel cells, J. Power Sources 369 (2017) 6
12. [80] V.V. Krishnan, Recent developments in metal-supported solid oxide fuel cells, Wiley Interdiscip. Rev. Energy Env. 6 (2017) e246. [81] B. Steele, Appraisal of Ce1 2 yGdyO2 2 y/2 electrolytes for IT-SOFC operation at 500 C, Solid. State Ion. 129 (2000) 95
110. [82] R.T. Leah, et al., Development progress on the ceres power steel cell technology platform: further progress towards commercialization, ECS Trans. 78 (2017) 87
95. [83] M. Kane, Nissan teams with ceres for future fuel-cell tech for EVs. Inside Evs, 2018. [84] Businesswire, Ceres wins d7m UK funding to support electric vehicle application with Nissan. Business Wire, 2018.

572

Fuel Cells for Transportation

[85] S. Pirou, et al., Production of a monolithic fuel cell stack with high power density, Nat. Commun. 13 (2022) 1263. [86] H. Xing, C. Stuart, S. Spence, H. Chen, Fuel cell power systems for maritime applications: progress and perspectives, Sustainability 13 (2021) 1213. [87] K.S. Howe, G.J. Thompson, K. Kendall, Micro-tubular solid oxide fuel cells and stacks, J. Power Sources 196 (2011) 1677
1686. [88] U.B. Pal, S.C. Singhal, Electrochemical vapor deposition of yttria-stabilized zirconia films, J. Electrochem. Soc. 137 (1990) 2937
2941. [89] K. Kendall, Progress in microtubular solid oxide fuel cells, Int. J. Appl. Ceram. Technol. 7 (2010) 1
9. [90] O. Hodjati-Pugh, A. Dhir, R. Steinberger-Wilckens, The development of current collection in micro-tubular solid oxide fuel cells—a review, Appl. Sci. 11 (2021) 1077. [91] Z. Jiang, C. Xia, F. Chen, Nano-structured composite cathodes for intermediatetemperature solid oxide fuel cells via an infiltration/impregnation technique, Electrochim. Acta 55 (2010) 3595
3605. [92] J.M. Vohs, R.J. Gorte, High-performance SOFC cathodes prepared by infiltration, Adv. Mater. 21 (2009) 943
956. [93] H. Taguchi, et al., LNF SOFC cathodes with active layer using Pr6O11 or Pr-doped CeO2, J. Power Sources 241 (2013) 768
775. [94] H.-S. Kang, S. Grewal, H. Li, M.H. Lee, Effect of surface-specific treatment by infiltration into LaNi 6Fe4O3-δ cathodic backbone for solid oxide fuel cells, J. Electrochem. Soc. 166 (2019) F255
F263. [95] D. Ding, X. Li, S.Y. Lai, K. Gerdes, M. Liu, Enhancing SOFC cathode performance by surface modification through infiltration, Energy Env. Sci. 7 (2014) 552. [96] C. Wu, et al., Sintering kinetics and microstructure analysis of composite mixed ionic and electronic conducting electrodes, Int. J. Energy Res. 46 (2022) 8240
8255. [97] Y. Wang, et al., Morphology and performance evolution of anode microstructure in solid oxide fuel cell: a model-based quantitative analysis, Appl. Energy Combust. Sci. 5 (2021) 100016. [98] I. Chang, S. Ji, J. Park, M.H.M.H. Lee, S.W.S.W. Cha, Ultrathin YSZ coating on Pt cathode for high thermal stability and enhanced oxygen reduction reaction activity, Adv. Energy Mater. 5 (2015) 1402251. [99] H. Li, et al., How an angstrom-thick oxide overcoat enhances durability and activity of nanoparticle-decorated cathodes in solid oxide fuel cells, J. Mater. Chem. A Mater 8 (2020) 15927
15935. [100] A. Karimaghaloo, et al., Nanoscale surface and interface engineering of solid oxide fuel cells by atomic layer deposition, Int. J. Precis. Eng. Manuf.-Green Technol. 6 (2019) 611
628. [101] J. Sunarso, S.S. Hashim, N. Zhu, W. Zhou, Perovskite oxides applications in high temperature oxygen separation, solid oxide fuel cell and membrane reactor: a review, Prog. Energy Combust. Sci. 61 (2017) 57
77. [102] A. Jun, J. Kim, J. Shin, G. Kim, Perovskite as a cathode material: a review of its role in solid-oxide fuel cell technology, ChemElectroChem 3 (2016) 511
530. [103] B. Koo, et al., Sr segregation in perovskite oxides: why it happens and how it exists, Joule 2 (2018) 1476
1499. [104] C. Sun, R. Hui, J. Roller, Cathode materials for solid oxide fuel cells: a review, J. Solid. State Electrochem. 14 (2009) 1125
1144.

Solid oxide fuel cells for vehicles Chapter | 17

573

[105] W. Jung, H.L. Tuller, Investigation of surface Sr segregation in model thin film solid oxide fuel cell perovskite electrodes, Energy Env. Sci. 5 (2012) 5370
5378. [106] S.P. Jiang, X. Chen, Chromium deposition and poisoning of cathodes of solid oxide fuel cells—a review, Int. J. Hydrog. Energy 39 (2014) 505
531. [107] O. Celikbilek, et al., Enhanced catalytic activity of nanostructured, a-site deficient (La 0.7 Sr 0.3) 0.95 (Co 0.2 Fe 0.8)O 3 2 δ for SOFC cathodes, J. Mater. Chem. A Mater 7 (2019) 25102
25111. [108] S. Jiang, The electrochemical performance of LSM/Zirconia
Yttria interface as a function of a-site non-stoichiometry and cathodic current treatment, Solid. State Ion. 121 (1999) 1
10. [109] F. Wang, et al., Tailoring the chemical stability of cobalt-rich perovskite mixed conductor, Solid. State Ion. 288 (2016) 2
5. [110] F. Wang, T. Nakamura, K. Yashiro, J. Mizusaki, K. Amezawa, Effect of Nb doping on the chemical stability of BSCF-based solid solutions, Solid. State Ion. 262 (2014) 719
723. [111] S. Choi, S. Park, J. Shin, G. Kim, The effect of calcium doping on the improvement of performance and durability in a layered perovskite cathode for intermediate-temperature solid oxide fuel cells, J. Mater. Chem. A Mater 3 (2015) 6088
6095. [112] M. Li, et al., A niobium and tantalum co-doped perovskite cathode for solid oxide fuel cells operating below 500 C, Nat. Commun. 8 (2017) 13990. [113] Y. Chen, et al., An effective strategy to enhancing tolerance to contaminants poisoning of solid oxide fuel cell cathodes, Nano Energy 47 (2018) 474
480. [114] M. Li, W. Zhou, Z. Zhu, Highly CO 2 -tolerant cathode for intermediate-temperature solid oxide fuel cells: samarium-doped ceria-protected SrCo 0.85 Ta 0.15 O 3 2 δ Hybrid, ACS Appl. Mater. Interfaces 9 (2017) 2326
2333. [115] Y. Chen, et al., A highly efficient multi-phase catalyst dramatically enhances the rate of oxygen reduction, Joule 2 (2018) 938
949. [116] J.-J. Choi, W. Qin, M. Liu, M. Liu, Preparation and characterization of (La 0.8 Sr 0.2) 0.95 MnO 3 2 δ (LSM) thin films and LSM/LSCF interface for solid oxide fuel cells, J. Am. Ceram. Soc. 94 (2011) 3340
3345. [117] H. Li, et al., Effect of angstrom-level oxide overcoat on Sr segregation behavior of LSM electrodes, Int. J. Hydrog. Energy 47 (2022) 33058
33066. [118] Q. Xu, et al., A comprehensive review of solid oxide fuel cells operating on various promising alternative fuels, Energy Convers. Manag. 253 (2022) 115175. [119] S. McIntosh, R.J. Gorte, Direct hydrocarbon solid oxide fuel cells, Chem. Rev. 104 (2004) 4845
4865. [120] N. Shi, et al., Review of anodic reactions in hydrocarbon fueled solid oxide fuel cells and strategies to improve anode performance and stability, Mater. Renew. Sustain. Energy 9 (2020) 6. [121] H.A. Shabri, M.H.D. Othman, M.A. Mohamed, T.A. Kurniawan, S.M. Jamil, Recent progress in metal-ceramic anode of solid oxide fuel cell for direct hydrocarbon fuel utilization: a review, Fuel Process. Technol. 212 (2021) 106626. [122] Y. Chen, et al., A robust fuel cell operated on nearly dry methane at 500 C enabled by synergistic thermal catalysis and electrocatalysis, Nat. Energy 3 (2018) 1042
1050.

This page intentionally left blank

Chapter 18

Hydrogen refueling stations/ infrastructure Yiheng Pang1, Andrew Martinez2 and Yun Wang1 1

Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA, United States, 2California Air Resources Board, Sustainable Transportation and Communities Division, California Environmental Protection Agency, Sacramento, CA, United States

18.1 Introduction Polymer electrolyte fuel cells (PEFCs) are one major fuel cell type considered for transportation applications [1]. Proton-exchange membrane fuel cells (PEMFCs) primarily rely on high-purity hydrogen gas as fuel for durable and efficient operation. Thus, a reliable hydrogen station network is a crucial prerequisite for PEMFC’s commercialization in automobile applications [2,3]. Hydrogen is widely used in industrial applications, with vast quantities produced to support gasoline refining and fertilizer production operations [4,5]. In the transportation sector alone, the infrastructure includes the full supply chain for hydrogen fuel, from production, transport and delivery, and storage at hydrogen refueling stations (HRSs), to safe and reliable dispensing into customers’ vehicles. However, infrastructure for the production and delivery of high-purity and high-pressure renewable hydrogen required for transportation fuel applications is far less prevalent [6]. Infrastructure development is currently a major bottleneck for commercializing fuel cell electric vehicles (FCEVs) in the transportation sector. The light-duty sector has launched in multiple countries worldwide, given the long-range (300 1 miles) and fast-filling (typically within 5 minutes) capability of FCEVs while providing zero-emission operation. In the medium- and heavy-duty markets, infrastructure developments have recently begun and enjoyed a growing interest in recent years, though there are still few commercially available products. Heavy-duty trucks may achieve long range (around 1000 miles or more) thanks to the ability to carry large amounts of hydrogen fuel onboard ( . 50 kg) without affecting payload volume or weight limits. Because of their zero-emission operation, when paired with low- or zero-carbon Fuel Cells for Transportation. DOI: https://doi.org/10.1016/B978-0-323-99485-9.00009-5 © 2023 Elsevier Ltd. All rights reserved.

575

576

Fuel Cells for Transportation

hydrogen fuel sources, worldwide PEMFC’s deployment and infrastructure development may significantly contribute to achieving global decarbonization goals [7].

18.2 Hydrogen fuel 18.2.1 Hydrogen properties Hydrogen is a widely used chemical in the industry for crude oil upgrading and fertilizer production. It has the highest specific energy among fuels but low energy density under standard conditions due to its low density or hydrogen content (see Tables 18.1 and 18.2). Energy density is critical to vehicle operation as high energy density enables long range on a single tank of fuel and fuel storage space is limited on most vehicle platforms. There are multiple options to increase the energy density of stored hydrogen fuel. One is to utilize liquid hydrogen, which has an energy density of about one-quarter of gasoline. Another is to use compressed gaseous hydrogen. In today’s hydrogen fuel

TABLE 18.1 Hydrogen contents in storage method/materials [2,8]. Storage method/material

Number of H atoms cm23 ( 3 1022)

Wt.% hydrogen

H2 gas (1 bar)

0.005

100

H2 gas (70 MPa)

2.3

100

H2 gas (130 MPa)

2.8

100

H2 liquid (20K)

4.2

100

H2 solid (4K)

5.3

100

MgH2

6.5

7.6

Mg2NiH4

5.9

3.6

FeTiH1.95

6.0

1.89

LaNi5H6.7

5.5

1.37

ZrMn2H3.6

6.0

1.75

11.4

2.10

VH2 

H2O (liquid, 25 C)

6.67

11.1

CH4 (liquid, 2162 C)

6.35

24.9

AlH3

8.90

10.0

LiBH4

7.36

18.4

NaBH4

6.81

10.6

Hydrogen refueling stations/infrastructure Chapter | 18

577

TABLE 18.2 Specific energy and energy density of several popular fuels [9,10]. Fuel

Specific energy (LHV) (MJ kg21)

Energy density (LHV) (MJ m23)

H2 gas (1 atm)

119.9

10.05

a

4,500

H2 gas (130 MPa)

a

119.9

6,600

H2 liquid

119.9a

8,491

Gasoline

44.4

31,150

Diesel

43

31,435.8

Methanol

19.8

15,800.1

Ethanol

26.8

21,200

H2 gas (70 MPa)

119.9

a

Neglect compression or liquefaction energy.

TABLE 18.3 Physical properties of H2 fuel. Property

Value

Hydrogen vapor density/specific volume

0.0813 kg m23/11.9 m3 kg21

Liquid density (at normal boiling point)

70.8 kg m23

Molecular mass

2.02 g mol21

Critical pressure

188.5 psi or 12.8 atm

Critical temperature

2239.96 C

Isotopes

Hydrogen, deuterium, and tritium

Latent heat of evaporation

461 kJ kg21

market, the liquid is more often used for transportation and distribution of hydrogen fuel, while vehicles themselves store hydrogen onboard in a compressed gaseous state. The current standard in FCEVs is 70 MPa, which has 4500 MJ m23, about seven times smaller than gasoline (see Table 18.2). Fuel cells have an efficiency advantage over conventional internal combustion engines, so onboard hydrogen tanks do not need to carry as much energy as gasoline tanks. Still, hydrogen storage systems onboard FCEVs are typically much larger than gasoline systems because of hydrogen’s lower energy density. Some physical properties of H2 are summarized in Table 18.3.

578

Fuel Cells for Transportation

H2 can be stored in various forms depending on temperature and pressure. Fig. 18.1 shows the phase diagram of hydrogen. Hydrogen is a gas at a high temperature with a density of 0.089 kg m23 at 0 C and 1 bar and becomes solid as temperature decreases to 2262 C with a density of 70.6 kg m23. A small area between the triple point and the critical point exhibits liquid hydrogen with a density of around 70.8 kg m23 [11]. In addition, hydrogen deviates from the ideal gas under high pressure. A compressibility factor Z is conventionally used as a multiplier to adjust the ideal gas law, which is given in Fig. 18.2 through the fitting with experimental data: PV 5 nZRT

ð18:1Þ

where V is the volume, P is the gas pressure, T is the absolute temperature, n is the number of moles, and R is the gas constant (R 5 8.314 J (K  mol)21). In addition, the deviation of real gas behavior from ideal gas behavior can be described by the application of the van der Waals equation to hydrogen: Pð V Þ 5

nRT n2 2a 2 V 2 nb V

ð18:2Þ

where a is the dipole interaction or repulsion constant (a 5 0.02476 m6 Pa mol22), and b is the volume occupied by the hydrogen molecules of one mole (b 5 2.661 3 10
5 m3 mol21) [11].

FIGURE 18.1 Phase diagram of hydrogen.

Hydrogen refueling stations/infrastructure Chapter | 18

579

FIGURE 18.2 Compressibility factor Z of hydrogen gas [12].

18.2.2 Energy for compression or liquefaction Hydrogen fuel is usually transported and stored either as compressed gas in tube trailers typically under 200
500 bar or in a cryogenic liquid state, using trailers equipped with vacuum-insulated multilayer dewars [13]. The current hydrogen tank in FCEVs is for compressed hydrogen at 70 MPa or 700 bar. Hydrogen compression requires energy input, which can be calculated for a specific polytropic process: PV n 5 const 2N , n , N 8 P2 V2 2 P 1 V1 > > ; n 6¼ 1 > < 12n 1W 2 5 V2 > > > P1 V1 ln ; n 5 1 : V1

ð18:3Þ

ð18:4Þ

where 1W2 is the input mechanical energy. Note that Eq. (18.4) assumes hydrogen follows the ideal gas law. Table 18.4 gives the energy needed to compress 5 kg of hydrogen from 1 atm to 70 MPa and 130 MPa at 25 C calculated by Eq. (18.4) for three compression processes. To liquefy hydrogen, several basic cycles may be applied: simple Claude, precooled simple Claude, precooled dual-pressure Linde
Hampson, or helium-precooled Claude. A comparison of the basic hydrogen liquefaction cycles is shown in Table 18.5. In the liquefaction systems, the specific energy consumption (SEC) is defined as the energy consumption divided by

580

Fuel Cells for Transportation

TABLE 18.4 Energy needed for hydrogen compression from 1 atm to 70 MPa and 130 MPa assuming H2 follows the ideal gas law.

Compression

Isothermal compression (n 5 1)

Adiabatic compression @n 5 1.4

Polytropic compression @n 5 1.3

1 atm - 70 MPa

40.45 MJ

84.84 MJ

72.73 MJ

1 atm - 130 MPa

44.32 MJ

104.17 MJ

86.89 MJ

TABLE 18.5 Comparison of basic hydrogen liquefaction cycles. Basic hydrogen liquefaction cycle

Liquid yield [%]

SEC [MJ kg21]

COP [%]

FOM [%]

Ideal liquefaction (Carnot cycle) [15]




11.88

100

100

Simple Claude [14]

8

79.56

5.5

18.1

Precooled simple Claude [17]

16
20

100.8
141.12




9.2
13

Precooled dual-pressure Linde
Hampson [18]

41

43.7




27

Helium-precooled Claude [17]

100

120.96
201.6




6.5
11

the physical production, as follows [14]: SEC 5

W_ m_ LH2

ð18:5Þ

where W_ is the power for liquefaction, and m_ LH2 is the mass flow rate of liquid H2. In refrigeration, the coefficient of performance (COP) is defined to measure the ratio of the cooling to input power of the cycle and can be evaluated using the reverse Carnot cycle: COP 5

Q_ TL 5 _ TH 2TL W

ð18:6Þ

where TH and TL denote the temperatures at the initial and final states of hydrogen, respectively. In actual liquefaction, the figure of merit (FOM) is popularly used, defined as the ratio of the ideal work to the actual work [14]: FOM 5

W_ rev m_ LH2 ½ðhLH2 2 h0 Þ 2 T0 ðsLH2 2 s0 Þ 5 _ W act W_ act

ð18:7Þ

Hydrogen refueling stations/infrastructure Chapter | 18

581

where W_ rev is the ideal reversible liquefaction work, W_ act is the actual liquefaction work, T0 is the ambient temperature, and s0 and h0 refer to the entropy and enthalpy at ambient temperature, respectively. The minimum theoretical liquefaction SEC from ambient (300K, 1.01 bar) conditions is 11.88 MJ kg21 [15]. Actual liquefaction energy requirements are much higher, for example, 36
46.8 MJ kg21 H2 depending on the size of the liquefaction operation [16]. These only account for a considerable portion of hydrogen energy according to Table 18.2. For example, the theoretical energy is about 9.9% (11.88/119.9) while the practical accounts for 30%
39%.

18.3 Hydrogen refueling station Fig. 18.3 shows the hydrogen dispensed at light-duty hydrogen fueling stations in the United States through June 2020. As of February 2022, the United States had about 50 HRSs, almost all in California [19,20]. The goal for the State of California is to have at least 100 stations funded by 2024 and 200 by 2025. The public
private California Fuel Cell Partnership has published an aspirational goal of 1000 stations as early as 2030 to enable the deployment of one million FCEVs in California [21]. In Europe, there were 139 HRSs in 2019, and they will operate about 1500 stations by 2025 according to their roadmap [22]. In Japan, 137 HRSs were in operation and planned to open 320 and 900 hydrogen stations by the end of 2025 and 2030, respectively, [23,24]. According to the International Partnership for Hydrogen and Fuel Cells in the Economy, China had 43 HRSs with over 6000 FCEVs as of early 2021 [25]. However, the current hydrogen

FIGURE 18.3 Amount of hydrogen dispensed by quarters in the United States [26].

582

Fuel Cells for Transportation

infrastructure is far less than what will be needed to support the worldwide deployment of FCEVs.

18.3.1 Hydrogen storage and source in hydrogen refueling stations Most HRSs across the globe follow the business and station design model of gasoline fueling stations. In fact, HRSs are mostly located on existing gasoline fueling stations in California [13]. An example of an HRS layout is shown in Fig. 18.4. HRSs are generally classified into two groups according to the bulk storage technology on-site: (1). the hydrogen is delivered as gas at a low or medium pressure (usually no higher than 50 MPa) stored in banks of gas cylinders at low or medium pressure [27]; (2). hydrogen may be delivered to the station in liquid form and stored on-site in liquid form. Liquid phase delivery allows a single delivery truck to carry larger quantities of hydrogen. Therefore, compared with a typical compressed-gas delivery truck, a single liquid delivery trailer can either easily support a larger station or deliver hydrogen to more stations. However, liquefaction incurs significant energy, monetary, and potential emission costs and may not be as good of a fit in sparser networks in the early phases of local market development [28,29]. Some boil-off of the liquid hydrogen may also occur, which much be kept at cryogenic temperatures during the entire delivery and storage process [30,31]. In addition, some HRS operators are investigating a higher-pressure gaseous delivery technology that may prove to be a viable alternative to liquid delivery and storage [22]. This technology has not been deployed at scale, so the coming years will provide more insight as stations begin operating with it.

FIGURE 18.4 An example of hydrogen refueling station (HRS) design with gaseous bulk storage.

Hydrogen refueling stations/infrastructure Chapter | 18

583

Another way to obtain hydrogen for stations is from on-site hydrogen production. Small (compared to industrial-scale units) steam methane reforming or electrolyzer systems [32,33] sometimes incorporated into station designs can utilize on-site production with regular deliveries at the same time. Although few light-duty fueling stations have on-site production because of the increased capital costs and space restrictions, there are some successful examples in various countries of HRS designs that incorporate on-site production [34]. The concept may also help establish hydrogen fueling in more remote areas, for dedicated fueling of fleets when abundant renewable resources are available for electricity production, or for the first few stations’ initial entries into local markets [13]. Hydrogen is typically kept in the gaseous phase from production to dispensing when it is produced on-site. In addition to production and bulk storage, at least one additional bank of gaseous hydrogen storage is typically included in stations like the example of HRS in Fig. 18.4, kept at pressures between the delivery pressure (usually no higher than 50 MPa) and final fueling pressure (usually at 80 MPa or higher). Common refueling processes designed to meet standard protocols like SAE J2601 or ISO 19-880 typically involve pressure-cascade filling. Cascade filling uses sequentially higher-pressure hydrogen storage banks to fill a vehicle’s onboard tank through a controlled pressure equalization process [35]. The lower-pressure tanks provide a proportionally higher mass of hydrogen during the fill, preserving the high-pressure hydrogen storage to enable greater backto-back fueling capability. Compressors refill the higher-pressure banks with hydrogen from the lower-pressure banks when higher-pressure banks are depleted. Although more sophisticated processes are also applied in today’s stations, most of them follow this basic strategy [36
38]. In addition to gaseous bulk storage, several strategies have been employed for stations with liquid bulk storage. Some vaporize hydrogen and then compress the gaseous hydrogen to storage banks at multiple pressures, much like a delivered hydrogen station model [39]. Others use a liquid pump to increase the hydrogen pressure in the liquid phase and then vaporize it before storage in high-pressure banks. This strategy makes it easier to compress at less energy cost and save on-site associated costs and emissions by taking advantage of the available liquid phase’s thermodynamics [40].

18.3.2 Refueling at hydrogen refueling stations To ease new FCEV adopters’ transition, the hydrogen refueling process is expected to be similar to gasoline refueling, which means the hydrogen refueling process must be safe, quick (in less than 5 minutes), and reliable. However, because the Joule
Thomson inversion temperature of hydrogen is very low (about 280 C), hydrogen gas temperature increases as it is dispensed into the vehicle under room temperature. The vehicle’s hydrogen

584

Fuel Cells for Transportation

tanks may be adversely affected by the heating. To ease the potential for damage due to cyclical thermal stresses, the hydrogen has to be significantly cooled before refueling (e.g., to 240 C for fills meeting the requirements of SAE J2601) [34]. Therefore, a refrigeration unit is also typically needed in an HRS. The technology and design will be based on the overall station design, the chiller’s location in the flow of hydrogen, and even the bulk storage phase (cryogenic liquid storage offers opportunities for more tightly integrated and optimized thermal management). Design features in several other station components are employed to ease new FCEV adopters’ transition, such as the dispenser, dispensing hose, nozzle, and payment system. These components have been designed to provide familiar experiences (e.g., many standard nozzles on the market use holster-style designs similar to gasoline nozzles). These dispensers are supplied by the same companies that also supply gasoline dispensers. Nozzles for HRSs are not only responsible for forming a pressure-tight connection with the onboard vehicle fuel tank and transferring hydrogen-like gasoline refueling, but they also detect and control the pressure and temperature during refueling. There is a temperature and pressure ramp rate safety limit set by industryaccepted standards [41]. In addition to the business model described above (applied to most HRSs built around the century-old model of stationary, centralized fueling stations), some regions have incorporated other models. For example, Japan is notable for using a large proportion of mobile fueling [42]. In this model, a truck is outfitted with bulk storage and compact processing and dispensing equipment. The truck travels to different areas in a single day to provide fueling service to customers. Outside of California, greenfield development of HRSs has been more popular. Stations designed for colocated light-duty and medium- or heavy-duty fueling are also becoming more common. In 2020, the California Energy Commission announced awards for more than 100 new hydrogen fueling stations, including as many as 13 with colocated fueling for light-duty and medium-duty vehicles [22].

18.3.3 Hydrogen refueling stations capital cost The capital expenses and operational costs for HRSs can vary widely based on the daily fueling capacity of the station, the bulk storage and other technology options employed, and the location. Stations developed with the assistance of State grants in California are required to report data to the National Fuel Cell Technology Evaluation Center administered by the National Renewable Energy Laboratory [43]. Total station costs, including industry and government funds, have varied from approximately $1,000,000 to as much as $4,500,000 though the latter number is known to be an outlier from an early demonstration station. It has been shown that the cost per kilogram of installed dispensing capacity during California’s early HRS program varied with station

Hydrogen refueling stations/infrastructure Chapter | 18

capacity, as shown by a simplified model or correlation [44]:


Capacity 20:95 21 5 $28; 000 Cost $ kg with R2 5 0:81 100

585

ð18:8Þ

For 100 and 1000 kg day21 station capacities, the total station costs are about $28,000 and $3100 kg21, respectively, from this model. The California Energy Commission has reported a drop in cost per kilogram of installed dispensing capacity during its hydrogen station program from $10,273 kg21 in 2012 to $6409 kg21 in 2015 [45]. Their latest solicitation, designed to enable larger network planning and multiyear development programs, could further reduce costs by allowing larger-scale orders in the upstream supply chain for station equipment [46
48]. Stations that have been awarded funding more recently in California, which are more commonly designed for liquid hydrogen delivery and storage, may result in lower capital and installation costs. Costs in other regions vary based on local supply chains, local fees and permitting requirements, and local incentives like tax credits. Fig. 18.5 illustrates the total installed cost for light-duty HRS from an analysis completed by the California Air Resources Board (CARB), based on data from stations proposed between 2009 and 2015 [44]. From the data, an HRS that can serve a local fleet of nearly 1000 vehicles may cost approximately $3 million to install.

FIGURE 18.5 Light-duty hydrogen refueling station total installed cost based on data from California’s hydrogen station funding program [44].

586

Fuel Cells for Transportation

18.4 Hydrogen refueling station networks To deploy a fleet of hydrogen-fueled FCEVs, a supporting network of conveniently located HRSs must be developed [49]. Fig. 18.6 shows the most recent guidance on needed HRS network development in California. Strategies for HRS network development pursued by public, private, or partnership efforts have varied widely. Some strategies facilitate the growth of FCEVs in the privately owned passenger vehicle market, whereas others focus on state or local government fleets. Other varying aspects include the

FIGURE 18.6 Most recent hydrogen fueling network and development guidance in California [51].

Hydrogen refueling stations/infrastructure Chapter | 18

587

TABLE 18.6 Station network status and target in select regions with active development programs. Hydrogen station goals HRS reported open in Sept 2020

HRS goals 2020 s

2030 s

2040 and beyond

USA [19,54]

44

320
570a (2025)

1500
3300a (2035)

7800
21,000a (2050)

California [47]

42

200 (2025)

1000b (2030)




Germany [55]

85

100 (2020)

Additional stations based on demand

Japan [23]

137

320 (2025)

900 (2030)




South Korea [56]

34




500 (2030)

1200 (2040)

China [57,58]

52

300(2025)

1000 (2030)




a

Based on scenario evaluations completed for public
private H2USA. Not adopted in state policy; based on report from public
private California Fuel Cell Partnership

b

amounts of public versus private funding, the use of mobile versus stationary stations, ownership and operation strategies, and individual station siting methodologies [50]. Table 18.6 summarizes the targets for HRS network development in some of the leading regions worldwide. In California, approximately 50 HRSs are open through joint private and state funding that supports around 10,000 lightduty FCEVs. The target for California is to achieve at least 100 stations by 2024 and 200 by 2025, with state co-funding secured for more than 150 stations and an additional 23 stations fully privately funded [22,52,53]. In addition, the public
private California Fuel Cell Partnership is working to advance a goal of 1000 stations by 2030. Approximately, 50 fuel cell electric buses currently operate in the State of California, mostly at three transit agencies in the regions of San Francisco Bay, Palm Springs, and Orange County. Fig. 18.7 shows the current number and projected future number of HRSs in the United States and other regions around the world by year. As with several other consolidated resources for tracking status, exact numbers may vary in the figure from other sources referenced here and in other reports, especially as reporting and tracking standards vary among regions. In Europe, about 1500 stations will become available by 2025 according to their roadmaps [60]. Germany plays an important role in Europe in developing hydrogen station networks. In Germany, HRSs are built and managed by H2Mobility [55]. Both public and private funds are available to the

588

Fuel Cells for Transportation

FIGURE 18.7 Number of HRSs by year in California [59].

corporation, which develops and operates the stations. About 85 stations have been completed in Germany, and H2Mobility aims to build 100 stations in seven primary city regions by the end of 2020. Further station network development will be applied based on demand. The government of Germany recently announced an economy-wide investment and recovery initiative that will invest 9 billion Euros into hydrogen across all industry sectors. In Japan, 137 HRSs are currently in operation; approximately, one-third of these locations are not permanent installations but are sites where mobile hydrogen fuelers are available for designated portions of the day. In the past, private and public funds flow alongside each other into individual station projects, which is a similar funding model as California. In 2018, Japan adopted the H2Mobility model, which directs public and private organizations’ funds to JHyM directly. The targets for network development in Japan are 160 stations by the end of 2020, 320 by 2025, and 900 by 2030. They have focused on four major metropolitan areas and then branch out from these four areas forming regional connections and more local networks [61]. South Korea and China have also made significant progress in recent years in developing HRS networks and establishing plans for future development. In South Korea, there were 34 operational stations in the country in late 2020; more recent estimates yielded over 70 operating stations. The rapid development of the HRS network and strong government incentives in South Korea have also led to a recent surge in FCEV deployment. South Korea is now estimated to have the most FCEVs on the road of any global market, at over 10,000 operating FCEVs though California had previously led in FCEV deployment until 2020. The government invested $2.33 billion over 5 years

Hydrogen refueling stations/infrastructure Chapter | 18

589

FIGURE 18.8 Numbers of HRSs and FCEVs and electrolysis hydrogen production capacity in various countries [64].

(about 50% of station development costs and funds to enable hydrogen sale prices to be competitive) [62]. China has 52 HRSs with over 6000 FCEVs [59]. A major Chinese FCEV company, SAIC Motor, recently released its first hydrogen strategy plan and announced that it will develop 10,000 FCEVs per year by 2025 [63]. Fig. 18.8 shows the number of HRSs and FCEVs in the world.

18.5 Challenges in hydrogen refueling stations network development The development of HRS networks is a complicated endeavor involving several partners across the public and private sectors. Although there are some successful examples worldwide in developing the network, challenges in achieving greater market growth still remain. The first challenge for hydrogen station network development is the station equipment technology development to provide more efficient and costeffective station operation. Nowadays, HRS technology provides a similar experience to the gasoline fueling experience. However, the equipment for hydrogen refueling has been relatively expensive to procure [46,65,66]; individual stations have struggled to maintain high levels of equipment reliability [67]; and on-site energy intensity can be relatively high, especially for station designs with hydrogen gas compressors [43,68,69]. Other equipment that has proven costly includes fueling nozzles, hydrogen compressors and liquid pumps, dispensers, and high-pressure gaseous storage [70]. Fig. 18.9 provides the data from a recent study of hydrogen station economics and key features for an individually profitable station [71]. It shows the cost and price targets for profitable station development and operations from a survey of station developers and other hydrogen stakeholders. While the cost of station

590

Fuel Cells for Transportation

FIGURE 18.9 Cost and price targets for profitable station development and operations in a survey of hydrogen fueling industry stakeholders [71].

development has been high, there are signs that improving economies of scale are helping to reduce costs. Another challenge is the market, supply chain, and logistical operations improvements to deliver more cost-effective station capital and operating costs and more reliable and lower-cost fuel supply due to economies of scale. To improve the development of hydrogen station networks, growth in the number of equipment suppliers and multiple product offerings made available from more manufacturers is crucial. In addition to equipment, a currently limited hydrogen marketplace also presents a significant challenge. Some station networks have been supplied with hydrogen fuel from a few suppliers, especially in California. The Pacific Northwest National Lab provides a database of hydrogen-producing facilities in North America, their production capacities, and the markets served [4,5]. By analyzing the data, in 2017, the CARB found that only 26 tons per day of liquid and 29 tons per day of gaseous merchant hydrogen production capacity were located within the state. CARB estimated that these sources are only sufficient for lightduty FCEV demand through 2020 and require up to 26 additional tons per day capacity by 2023 [6]. To solve this problem, the California Energy Commission has funded projects to get an additional 3 tons per day, and

Hydrogen refueling stations/infrastructure Chapter | 18

591

industrial gas company Air Liquide has begun construction of a 30 tons per day hydrogen production facility [34,72]. However, more production facility development is required to enable longer-term goals for FCEV deployment. Moreover, fuel distribution is a huge challenge. The considerable distance between stations requires greater numbers of fuel delivery drivers than if stations were present at densities similar to today’s diesel, ethanol, and gasoline networks [73]. The third challenge is the commitment to supporting policies that enable demand growth, foster industry development, and minimize the financial risk to enter the market. Many technological solutions to solve a problem are often new ventures and require support from the government, including financial support and policy support. Due to the potential consequences of global climate change, governments worldwide are taking serious actions to address the sources of greenhouse gas emissions. FCEVs and hydrogen are among the newest technologies in transportation to decrease CO2 emissions. Compared with battery electric vehicles, FCEVs represent a dual challenge in that success requires simultaneously launching not only a new vehicle technology market but also doing the same for fueling infrastructure. Costs and risks are high, and the industry has often looked to governments for financial and policy support. In California, the Low Carbon Fuel Standard has attracted significant industry interest, especially since the addition of the Hydrogen Refueling Infrastructure provision, which reduces developer risk by offsetting station operation costs when FCEV volumes are low [74]. Even though the programs exist, governments have to make difficult funding priority decisions with limited available funds. Ensuring that existing programs deliver maximum benefit and secure additional future opportunities remains a challenge for FCEVs and hydrogen. Developing awareness and acceptance of HRSs and FCEVs is another challenge. Even though FCEVs and HRSs have become more and more popular and reliable, lack of acceptance with hydrogen (even fear of it) and fuel cell technology has often caused delays in the station development and building at a proposed host site [34,45]. To increase public awareness and acceptance of fuel cell and hydrogen technology, local permits and governing structures may give more support.

18.6 Summary In this chapter, the current state of HRSs is discussed, alongside relevant hydrogen fuel fundamentals, HRS technology, hydrogen station networks, and challenges in hydrogen station network development. Hydrogen fuel has a high specific energy, but low energy density under standard conditions. To increase its energy density, compressed or liquid hydrogen is a viable method, both of which are prevalent in today’s markets. Energy for hydrogen compression (70 MPa) and liquefaction is 35.23 and 11.88 MJ, respectively,

592

Fuel Cells for Transportation

which accounts for 29.4% and 9.9% of the hydrogen energy. Compressed hydrogen in the FCEV tank (70 MPa) deviates from ideal gas behavior with a compressibility factor Z of about 1.5 at 300K. Hydrogen station networks development attracts governments’ interest and gets more funding support for R&D. The United States currently has over 50 HRSs, nearly all in California, and the State targets 200 stations by 2025. Europe has over 140 HRSs, with 1500 stations planned by 2025. Japan has 137 HRSs operating, with a target of 900 by 2030. However, the HRS industry is still in the early stages of development, and HRS equipment continues to evolve. Improving efficiencies of components and reducing station operation costs are still a barrier. Improvement in compressor technology, higher-pressure storage, wider distribution, and liquid hydrogen storage are particularly active development areas as well. In addition to HRS equipment development, there is a significant ongoing effort to increase government support and public acceptance of the technology.

18.6.1 Review questions/worked examples 1. Calculate the energy input to compress hydrogen from 1 atm to 100 atm in an isothermal process at 25 C. 2. For 70 and 130 MPa hydrogen tanks at 20 gallons at 25 C, calculate the mass of hydrogen using both ideal gas and van der Waals equations. 3. Calculate the COP of the ideal refrigeration evaluated by reverse Carnot cycle between 25 C and 2253 C. 4. Estimate the cost of an HRS with an 800 kg day21 capacity.

Nomenclature h m_ n P Q_ R s T V W_

specific enthalpy (kJ kg21) mass flow rate (kg s21) number of mole pressure (bar) heat flow (kW) the universal gas constant (J (K  mol)21) specific entropy (kJ (kg  K)21) temperature volume (m3) power (kW)

Acronyms COP FCEV FOM HRS HX

coefficient of performance fuel cell electric vehicle figure of merit hydrogen refueling station heat exchanger

Hydrogen refueling stations/infrastructure Chapter | 18 PEFC PEMFC SEC

593

polymer electrolyte fuel cell proton-exchange membrane fuel cell specific energy consumption

References [1] Y. Wang, B. Seo, B. Wang, N. Zamel, K. Jiao, X.C. Adroher, Fundamentals, materials, and machine learning of polymer electrolyte membrane fuel cell technology, Energy AI (2020) 100014. [2] Y. Wang, D.F.R. Diaz, K.S. Chen, Z. Wang, X.C. Adroher, Materials, technological status, and fundamentals of PEM fuel cells—a review, Mater. Today 32 (2020) 178
203. [3] A. Martinez, The future is present in California: delivering on the promise of fuel cellpowered transportation, IEEE Electrification Mag. 6 (1) (2018) 37
47. [4] H2Tools, Captive, on-purpose, refinery hydrogen production capacities at individual U.S. refineries. Available from: https://h2tools.org/hyarc/hydrogen-data/captive-purpose-refinery-hydrogen-production-capacities-individual-us, 2019. [5] H2Tools, Merchant hydrogen plant capacities in North America. Available from: https:// h2tools.org/hyarc/hydrogen-data/merchant-hydrogen-plant-capacities-north-america, 2016. [6] California Air Resources Board, Hydrogen station network self-sufficiency analysis per assembly bill 8. Available from: https://ww2.arb.ca.gov/sites/default/files/2021-10/hydrogen_self_sufficiency_report. pdf, 2021. [7] Fuel Cells and Hydrogen 2 Joint Undertaking, Hydrogen roadmap Europe: a sustainable pathway for the European energy transition, 2019. [8] U. Eberle, G. Arnold, R. Von Helmolt, Hydrogen storage in metal
hydrogen systems and their derivatives, J. Power Sources 154 (2) (2006) 456
460. [9] College of the Desert, Hydrogen properties. Available from: https://www1.eere.energy. gov/hydrogenandfuelcells/tech_validation/pdfs/fcm01r0.pdf, December 2001. [10] G. Thomas, Overview of storage development DOE hydrogen program. Sandia National Laboratories, 9. Available from: http://large.stanford.edu/courses/2012/ph240/landreman2/ docs/storage.pdf, 2000. [11] J.E. Jensen, R.G. Stewart, W.A. Tuttle, H. Brechna, Brookhaven National Laboratory Selected Cryogenic Data Notebook: Sections, III (Vol. 1)., Brookhaven National Laboratory, 1980. [12] S. Makridis, Hydrogen storage and compression, ArXiv Prepr. arXiv 1702 (2017) 06015. [13] Y. Wang, H. Yuan, A. Martinez, P. Hong, H. Xu, F.R. Bockmiller, Polymer electrolyte membrane fuel cell and hydrogen station network for automobiles: status, technology, and perspectives, Adv. Appl. Energy (2021) 100011. [14] M. Aasadnia, M. Mehrpooya, Large-scale liquid hydrogen production methods and approaches: a review, Appl. Energy 212 (2018) 57
83. [15] M.T. Syed, S.A. Sherif, T.N. Veziroglu, J.W. Sheffield, An economic analysis of three hydrogen liquefaction systems, Int. J. Hydrog. Energy 23 (7) (1998) 565
576. [16] ENERGY.GOV, DOE H2A Delivery Analysis. Available from: https://www.hydrogen. energy.gov/h2a_delivery.html (accessed 4.6.22). [17] W. Peschka, Liquid Hydrogen: Fuel of the Future, Springer Science & Business Media, 2012. [18] T.K. Nandi, S. Sarangi, Performance and optimization of hydrogen liquefaction cycles, Int. J. Hydrog. Energy 18 (2) (1993) 131
139. [19] Hydrogen Fueling Stations, Alternative fuels data center. Available from: https://afdc. energy.gov/fuels/hydrogen_stations.html, 2020.

594

Fuel Cells for Transportation

[20] Hydrogen Fuel Cell Partnership, Retail Station map. Available from: https://cafcp.org/stationmap (accessed 4.6.22). [21] California Fuel Cell Partnership, The California Fuel Cell Revolution. Available from: https://cafcp.org/sites/default/files/CAFCR.pdf, 2018. [22] California Energy Commission, Notice of proposed awards grant funding opportunity 19
602 1st revised. California Energy Commission, Available from: https://www.energy.ca.gov/sites/ default/files/2020-12/NOPA_GFO-19
602_1st_Revised_12-08-2020_ADA.xlsx, 2020. [23] S&P Global Market Intelligence, Corporate credit risk trends in developing markets: a loss given default (LGD) perspective. Available from: https://www.spglobal.com/marketintelligence/en/news-insights/blog/corporate-credit-risk-trends-in-developing-markets-aloss-given-default-lgd-perspective, 2021. [24] J. Mishler, Y. Wang, P.P. Mukherjee, R. Mukundan, R.L. Borup, Subfreezing operation of polymer electrolyte fuel cells: ice formation and cell performance loss, Electrochim. Acta 65 (2012) 127
133. [25] International Partnership for Hydrogen and Fuel Cells in the Economy, IPHE Accomplishments. Available from: https://www.iphe.net/ (accessed 25.4.21). [26] NREL Transforming Energy, Hydrogen Dispensed by Quarter - Retail Stations. Available from: https://www.nrel.gov/hydrogen/assets/images/cdp-retail-infr-01.jpg, 2022. [27] Energy.gov. Hydrogen delivery technical team roadmap. Available from: https://www. energy.gov/sites/default/files/2017/08/f36/hdtt_roadmap_July2017.pdf, 2017. [28] K. Reddi, A. Elgowainy, N. Rustagi, E. Gupta, Impact of hydrogen refueling configurations and market parameters on the refueling cost of hydrogen, Int. J. Hydrog. Energy 42 (34) (2017) 21855
21865. [29] C. Yang, J. Ogden, Determining the lowest-cost hydrogen delivery mode, Int. J. Hydrog. Energy 32 (2) (2007) 268
286. [30] A. Bauer, T. Mayer, M. Semmel, M.A.G. Morales, J. Wind, Energetic evaluation of hydrogen refueling stations with liquid or gaseous stored hydrogen, Int. J. Hydrog. Energy 44 (13) (2019) 6795
6812. [31] T. Mayer, M. Semmel, M.A.G. Morales, K.M. Schmidt, A. Bauer, J. Wind, Technoeconomic evaluation of hydrogen refueling stations with liquid or gaseous stored hydrogen, Int. J. Hydrog. Energy 44 (47) (2019) 25809
25833. [32] Q. Chen, Y. Wang, F. Yang, H. Xu, Two-dimensional multi-physics modeling of porous transport layer in polymer electrolyte membrane electrolyzer for water splitting, Int. J. Hydrog. Energy 45 (58) (2020) 32984
32994. [33] F. Suleman, I. Dincer, M. Agelin-Chaab, Environmental impact assessment and comparison of some hydrogen production options, Int. J. Hydrog. Energy 40 (21) (2015) 6976
6987. [34] J. Baronas, G. Achtelik, et al., Joint agency staff report on assembly bill 8: 2018 annual assessment of time and cost needed to attain 100 hydrogen refueling stations in California. California Energy Commission and California Air Resources Board, 2018. Publication Number: CEC-600-2018-008. Available from: https://static1.squarespace.com/static/ 58e8f58d20099ea6eb9ab918/t/5e62a572f7b7df2653ed8ffa/1583523208173/CEC-600
2018008.pdf, 2018. [35] SAE International, Fueling protocols for light duty gaseous hydrogen surface vehicles J2601_202005. SAE International. ,https://www.sae.org/standards/content/j2601_202005/ . , 2020. [36] Y. Wang, X.C. Adroher, J. Chen, X.G. Yang, T. Miller, Three-dimensional modeling of hydrogen sorption in metal hydride hydrogen storage beds, J. Power Sources 194 (2) (2009) 997
1006.

Hydrogen refueling stations/infrastructure Chapter | 18

595

[37] N.A.A. Rusman, M. Dahari, A review on the current progress of metal hydrides material for solid-state hydrogen storage applications, Int. J. Hydrog. Energy 41 (28) (2016) 12108
12126. [38] M.V. Lototskyy, V.A. Yartys, B.G. Pollet, R.C. Bowman Jr, Metal hydride hydrogen compressors: a review, Int. J. Hydrog. Energy 39 (11) (2014) 5818
5851. [39] YouTube, Linde standard hydrogen filling station with IC90 compressor. Available from: https://www.youtube.com/watch?v 5 usaQrCDORFY, July 16, 2014. [40] YouTube, Linde, Linde standard hydrogen filling station with cryo pump. Available from: https://www.youtube.com/watch?v 5 Pjh639S2dek, February 21, 2017. [41] J. Zheng, L. Zhao, K. Ou, J. Guo, P. Xu, Y. Zhao, et al., Queuing-based approach for optimal dispenser allocation to hydrogen refueling stations, Int. J. Hydrog. energy 39 (15) (2014) 8055
8062. [42] T. Ikeda, Status of hydrogen fueling station technologies in Japan. The Association of Hydrogen Supply and Utilization Technologies. Available from: https://www.energy. gov/sites/prod/files/2018/10/f56/fcto-infrastructure-workshop-2018-16-ikeda.pdf, 2018 (accessed 21.09.20). [43] National Renewable Energy Laboratory, Next generation hydrogen station composite data products: retail stations. Available from: https://www.nrel.gov/hydrogen/infrastructurecdps-retail.html, 2020. [44] Hydrogen station network self-sufficiency analysis per assembly bill 8. https://ww2.arb.ca. gov/sites/default/files/2020-11/ab_8_self_sufficiency _report_draft _ac.pdf. [45] J. Baronas, G. Achtelik, Joint agency staff report on assembly bill 8: 2017 annual assessment of time and cost needed to attain 100 hydrogen refueling stations in California, Calif. Energy Comm (2017). [46] J. Baronas, G. Achtelik, Joint agency staff report on assembly bill 8: 2019 annual assessment of time and cost needed to attain 100 hydrogen refueling stations in California, Calif. Energy Comm (2019). [47] Air Resources Board, 2020 annual evaluation of fuel cell electric vehicle deployment and hydrogen fuel station network development. Available from: https://ww2.arb.ca.gov/sites/ default/files/2020-09/ab8_report_2020.pdf, 2020. [48] California Energy Commission, Grant funding opportunity 19-602 application manual. Available from: https://www.energy.ca.gov/solicitations/2019-12/gfo-19
602-hydrogenrefueling-infrastructure, 2019. [49] Wang et al., EES review, under review, 2022. [50] The International Council on Clean Transportation, Developing hydrogen fueling infrastructure for fuel cell vehicles: a status update. Available from: https://theicct.org/sites/default/files/ publications/Hydrogen-infrastructure-status-update_ICCT-briefing_04102017_vF.pdf, 2017. [51] Energy.gov, Grant funding opportunity clean transportation program hydrogen refueling infrastructure. Available from: https://www.energy.ca.gov/sites/default/files/2020-05/ 01_GFO-19
602_Application_Manual_Addendum_04_ADA.docx, May 2020. [52] Official California Legislative Information, Assembly Bill No. 8 (Perea, Statutes of 2013, Chapter 401). Available from: http://www.leginfo.ca.gov/pub/13-14/bill/asm/ab_00010050/ab_8_bill_20130928_chaptered.pdf, 2013. [53] California Governor’s Office of Planning and Research, Zero-Emission Vehicles,. Available from: https://www.opr.ca.gov/planning/transportation/zev.html, 2020. [54] M.B. Melaina, M. Bush, J. Muratori, et al., National hydrogen scenarios: how many stations, where, and when? Prepared by the National Renewable Energy Laboratory for the H2 USA Locations Roadmap Working Group. Available from: https://www.h2usa.org/ sites/default/files/H2USA_LRWG_NationalScenarios2017.pdf, 2017.

596

Fuel Cells for Transportation

[55] H2Mobility. Keep on rolling
you’ll go far with hydrogen. Available from: https://h2mobility.de/en/our-h2-stations/, 2020. [56] Government of the Republic of Korea, Government of the Republic of Korea unveils three key strategies to accelerate the future development of the car industry. 2019. Available https://english.motie.go.kr/common/download.do?fid 5 bbs&bbs_cd_n 5 2&bbs_ from: seq_n 5 742&file_seq_n 5 1. [57] Z. Xin, Hydrogen energy industry to speed up expansion in nation. China Daily Global. Available from: http://epaper.chinadaily.com.cn/china/2021-05/21, June 3rd 2020. [58] J. Li, Hydrogen energy country overview: China. US Department of Energy Hydrogen Infrastructure Workshop. Conference Presentation, 2018. [59] A. Martinez, California’s LDV hydrogen network: 2021 HFTO annual merit review. DOE Hydrogen Program 2021 Annual Merit Review, United States Department of Energy. Conference Presentation, 10 June 2021. [60] California Energy Commission, Notice of proposed awards grant funding opportunity 19602 1st Revised. Available from: https://www.energy.ca.gov/sites/default/files/2020-12/ NOPA_GFO-19
602_1st_Revised_12-08-2020_ADA.xlsx, 2020. [61] JHyM, 4 new HRS installation added in fiscal 2020. Available from: https://www.jhym. co.jp/en/wp-content/uploads/2020/11/EN-JHyM_20201125-HRS-installation-plan-added. pdf 2020. [62] Ministry of Trade, Industry and Energy, Accelerating the construction of a public-private hydrogen vehicle industrial ecosystem. Ministry of Trade, Industry and Energy. Available from: http://www.motie.go.kr/motie/ne/presse/press2/bbs/%20bbsView.do?bbs_seq_n 5 160578&bbs_cd_n 5 81¤tPage 5 1&search_key_n 5 &%20cate_n 5 &dept_v 5 &search_val_v 5 %20Translated, 2018. [63] Y. Haiyan. SAIC to make, sell 10,000 hydrogen cars per year in five years. Available from: https://www.yicaiglobal.com/news/saic-to-make-sell-10000-hydrogen-cars-per-yearin-five-years, 2020 (accessed 19.10.21). [64] IPHE, IPHE accomplishments. Available from: https://www.iphe.net/accomplishments2021, 2021. [65] Hydrogen Council, Path to hydrogen competitiveness a cost perspective. Available from: https:// hydrogencouncil.com/wp-content/uploads/2020/01/Path-to-Hydrogen-Competitiveness_FullStudy-1.pdf, 2020. [66] R. Schreffler, Costs check growth of fuel-cell infrastructure. Wards Auto. Available from: https://www.wardsauto.com/technology/costs-check-growth-fuel-cell-infrastructure, 2019. [67] J. Kurtz, S. Sprik, M. Peters, et al., Retail hydrogen station reliability status and advances, Reliab. Eng. Syst. Saf. (2020) 106823. Available from: https://www.osti.gov/servlets/purl/ 1603259. [68] G. Saur, S. Gilleon, S. Sprik, Next generation hydrogen station composite data products: retail stations data through quarter 3 of 2019. Department of Energy Hydrogen and Fuel Cells Program Annual Merit Review 2020. Conference Presentation, 2020. [69] J. Pratt, D. Terlip, C. Ainscough, et al. H2FIRST reference station design task project deliverable 2. Available from: https://www.energy.gov/sites/prod/files/2015/04/f22/ fcto_h2first_reference_station_design_report_april2015_0.pdf, 2015. [70] N. Rustagi, A. Elgowainy, J. Vickers. DOE hydrogen and fuel cells program record #18003: current Status of hydrogen delivery and dispensing costs and pathways to future cost reduction. Available from: https://www.hydrogen.energy.gov/pdfs/18003_current_ status_hydrogen_delivery_dispensing_costs.pdf, 2018.

Hydrogen refueling stations/infrastructure Chapter | 18

597

[71] California Air Resources Board, 2019 annual evaluation of fuel cell electric vehicle deployment and hydrogen fuel station network development. Available from: https://ww2. arb.ca.gov/sites/default/files/2019-07/AB8_report_2019_Final.pdf, 2019. [72] Air Liquide, Air liquide air liquide to build first world scale liquid hydrogen production plant dedicated to the supply of hydrogen energy markets. Available from: https://energies. airliquide.com/air-liquide-build-first-world-scale-liquid-hydrogen-production-plant-dedicatedsupply-hydrogen, November 28, 2018. [73] M. Penev, C. Hunter, Regional supply of hydrogen. Department of energy hydrogen and fuel cells program annual merit review. Conference Presentation, 2018. [74] California Air Resources Board, Unofficial electronic version of the low carbon fuel standard regulation. California Air Resources Board. Available from: https://ww2.arb.ca.gov/ sites/default/files/2020-07/2020_lcfs_fro_oal-approved_unofficial_06302020.pdf, 2020.

This page intentionally left blank

Index Note: Page numbers followed by “f” and “t” refer to figures and tables, respectively.

A AC. See Alternating current (AC) Academic fuel cell testing, 253
254 Acceleration pressure loss, 232 Acid washing, 154 Active site protonation, 189 Activity degradations, 564
565 Adaptive control system, 358
360 Adaptive low-pass filter method, 294 Additive manufacturing, 246f Adhesion force, 229
230 AEMFC. See Anion-exchange membrane fuel cell (AEMFC) Aerodynamic drag correlations, 229
230 Aerodynamic drag force, 284
285 Agent models, 106 Agglomerate density, 53
55 Agglomerate kinetics, 33
37 AHP. See Alternating hydrogen pump (AHP) AI. See Artificial intelligence (AI) Air cooling, 348
349, 349f Aircraft, 557
558 Air supply system, 13 Alcohols, 555
556 ALD. See Atomic layer deposition (ALD) Algebraic equations, 68 Alkaline fuel cells, 29 Allis-Chalmers farm tractor, 8 Alternating current (AC), 217
218, 269, 532
533 Alternating hydrogen pump (AHP), 483 Aluminum alloys, 334 Ammonia, 551
554 Ammonium chloride (AC), 181 Analytical models, 394 Ancillary equipment, 260 Anion-exchange membrane fuel cell (AEMFC), 11, 29 ANNs. See Artificial neural networks (ANNs) Anode starvation, 264

Apollo Space Program, 8 APU. See Auxiliary power units (APU) Archimedes’ principle, 216
217 Arc ion plating, 334
335 Arrhenius-type behavior, 492
493 Artemis driving cycles, 283 Artemis motorway driving cycle, 283, 283f Artificial intelligence (AI), 103
107, 114
122, 296 combination of, 122
123 fault diagnosis, 120
122 model predictive control, 117
118 parameter optimization, 115
116 prognostics and health management, 118
120 technology, 289 Artificial neural networks (ANNs), 95
96, 106f, 122
123, 358
361 Assisted-start strategy, 483, 532
533 Atomic layer deposition (ALD), 565 Automotive driving cycles ex situ accelerated stress test, 447
453 clamping force, 452
453 freeze
thaw cycling, 452 humidity cycling, 447
448 hygrothermal cycling, 448
449 liquid water wet
dry cycling, 449
452 temperature cycling, 448 vibration, 453 fundamental degradation mechanisms, 422
431 bipolar plate, 430
431 catalyst layer, 424
430 gas diffusion layer, 430 polymer electrolyte membrane, 422
424 in situ accelerated stress test, 435
447 degradation rate by, 444
447 protocols, 436
443 steady-state durability test, 431
435 degradation rate by, 432
435

599

600

Index

Automotive driving cycles (Continued) protocols, 431
432 of proton-exchange membrane fuel cells, 433t Automotive systems, 258
259 Auxiliary battery, 13 Auxiliary power units (APU), 556, 558 for heavy-duty trucks, 556
557 Average absolute error, 117
118 Aviation sector, 3, 17, 19t

B Backpropagation neural network (BPNN), 117
118 Backtracking search algorithm, 292 Balance of plant (BoP), 120
122 Ballard Power Systems, 9, 420
421 Battery electric vehicles (BEVs), 3, 199 Battery electric vehicle with green electricity (BEV-Green), 24f Beginning of life (BOL), 420
421 Benzyl disulfide, 132
133 BEVs. See Battery electric vehicles (BEVs) Bidirectional DC/DC converters, 279, 290 Biomimetic/fractal flow field, 240 Bi-phenyl sulfone-H (BPSH) membrane, 447
448 Bipolar plates (BPs), 115, 307
311, 340
341, 394, 430
431, 465 flow field design, 311
330 with guided flow path, 313
330 without guided flow path, 312
313 functions, 308
309 materials and manufacturing, 330
336 carbon composite, 333
334 chemical stability/compatibility, 330 corrosion-resistance, 330 dissolution-resistance, 330 electrical conductivity, 331 gas diffusivity/impermeability, 331 graphite, 332
333 lightweight, 331 low cost, 330
331 mechanical strength, 331 metallic bipolar plates, 334
336 recyclability, 331 surface finish and flatness, 331 thermal conductivity, 331 typical materials and classification, 331 requirements, 309
311 Black box model, 120
122 Boltzmann equation, 104
105

Boltzmann kinetic theory, 107 Boltzmann’s constant, 32, 79 Bond number, 240 BoP. See Balance of plant (BoP) BPNN. See Backpropagation neural network (BPNN) BPs. See Bipolar plates (BPs) Brinkman equation, 43 Brownian movements, 426
428 Bruggeman’s correlation, 374
375 Bruggeman’s effective medium theory, 209 Brunauer
Emmett
Teller (BET), 216
217 Butler
Volmer (BV) equation, 31
33, 68, 340
341, 399, 512
513 Butler
Volmer (BV) kinetics, 218
219

C California Air Resources Board (CARB), 585 California Energy Commission, 585 California LA-92, 283 Canadian Department of National Defense, 9 Capillary-driven flow, 226
228 Capillary fingering (CF) regime, 226
228 Capillary pressure, 66, 377 Capital cost, 584
585 CARB. See California Air Resources Board (CARB) Carbon, 133
134 Carbon aerogels (CAs), 247 Carbon black, 128 Carbon capture and storage (CCS), 4 Carbon
carbon composite, 373
374 Carbon coking, 566
567 Carbon composite, 333
334 Carbon dioxide (CO2), 1 Carbon fibers, 202
205 Carbon monoxide (CO), 271
272 Carbon nanofiber (CNF), 128, 380
381 Carbon nanotubes (CNTs), 128, 352, 380
381 Carbon-Neutral Cycle, 561
562 Carbon paper, 373
374 Carman
Kozeny constant, 212 Carman
Kozeny equation, 212 Carnot cycle, 3
4 Carnot efficiency, 3
4 CAs. See Carbon aerogels (CAs) Catalyst agglomeration, 426
428 Catalyst-coated membrane (CCM), 182, 474
475 Catalyst layers (CLs), 4
5, 29
30, 75, 109
110, 112
115, 201, 234
235,

Index 305, 340
341, 367, 394, 419, 424
430, 464 electrochemical reaction process of, 113 modeling, 82
83 optimal structural parameters of, 115
116 proton exchange membrane and ionomer of, 75
76 structure and composition of, 201f two-phase transport mechanism in, 114 Catalyst support degradation, 274
275 Cathode catalyst layer (cCL), 225, 380
386 Cathode inlet temperatures, 257
258 CCM. See Catalyst-coated membrane (CCM) CCS. See Carbon capture and storage (CCS) Cell configurations, 550 Cell design, 529 Central processing unit, 104
105 Ceria nanoparticles (CeO2), 191
192 Cetyltrimethylammonium bromide (CTAB), 131
132, 131f CFD. See Computational fluid dynamics (CFD) Channel geometry, 237
238 Characterization techniques, 471, 473t Charge transfer coefficient, 39
41, 512
513 Chemical etching, 148
149 Chemical passivation, 334
335 Chemical vapor deposition, 334
335 CHP. See Combined heat and power (CHP) Clamping force, 452
453 Claperyron equation, 486 Classic fuzzy control, 292 Clausius
Claperyron equation, 487
488 Climate change, 1 CLs. See Catalyst layers (CLs) CNF. See Carbon nanofiber (CNF) CNTs. See Carbon nanotubes (CNTs) Coalescence, 426
428 Cobalt phthalocyanine (CoPc), 179 Coefficient of performance (COP), 579
581 Coefficient of thermal expansion (CTE), 549
550 Cold start, 244, 361 Combined heat and power (CHP), 548
549 Commercial carbon supports, 128
136 functionalization methods of, 132
136 comparison of, 136t doping heteroatoms, 132
133 oxidation treatment, 132 loading Pt-based electrocatalysts on, 136
138 ex situ mixing method, 137
138

601

one-pot synthesis method, 137 Pt-based electrocatalysts, 138
153 Pt-based open nanostructures, 145
149 Pt-based polyhedrons, 141
145 Pt-based spherical nanoparticles, 140
141 postsynthesis treatment, 158 Pt-based electrocatalysts, 157
158 structure and surface properties of, 128
132 studies on functionalization of, 156 Compression molding, 333
334 Compression ratios (CR), 57
58 Computational fluid dynamics (CFD), 77
78, 104, 394 Conservation of charge, 38
39 Conservation of energy, 48
49 Conservation of momentum, 42
44 Contact angles, 478 Control strategy, 358
361 adaptive control, 359
360 artificial neural network, 360
361 fuzzy control, 360 model predictive control, 359 proportional
integral
derivative control, 359 robust control, 360 Convection heat transfer coefficient, 344 Conventional flow channels, 368
370 Coolant pipeline, 356 Coolant pump, 356 Coolant tank, 356 Cooling/thermal management system, 14 COP. See Coefficient of performance (COP) Counter-reference electrode, 271 Critical cluster production rate, 500 Critical nucleation radius, 496
500 CTAB. See Cetyltrimethylammonium bromide (CTAB) CTE. See Coefficient of thermal expansion (CTE) Current interrupt (CI), 272 CV. See Cyclic voltammetry (CV) Cyanamide (CM), 179
180 Cyclic voltammetry (CV), 146, 147f, 218
219, 271, 431
432

D Darcy’s law, 43, 209, 217
218, 385 Data-driven models, 106, 116
118, 394, 408 Data-driven surrogate modeling, 93
96 DC. See Direct current (DC)

602

Index

DC/DC converter, 13 DDAB. See Didecyldimethylammonium bromide (DDAB) DDAC. See Dimethyldioctadecylammonium chloride (DDAC) Dead-ended flow fields, 311
312 Dead-end mode, 258 Decyltrimethylammonium bromide (DTAB), 144 Deionization filter, 357
358 Department of Energy (DOE), 181
182, 420
421 Derivative coefficient, 359 Design variables, 83
85, 92
93 DESTA project, 557 Dew point, 225
226 Diagnostic techniques, 267
273 current interrupt, 272 electrochemical impedance spectroscopy, 269
270 electrochemically active surface area, 271
272 impurity analysis, 273 online gas and water analysis, 273 polarization curves, 268
269 voltammetry, 270
271 Dicyandiamide, 132
133 Didecyldimethylammonium bromide (DDAB), 149
151 Diesel engines, 17 Diethylene triamine, 132
133 Differential equations, 68 Differential scanning calorimetry (DSC), 489 Diffusion bridge, 217
218 Diffusivity, 209
210 Dimethyldioctadecylammonium chloride (DDAC), 149
151 Dip coating, 334
335 Direct current (DC), 29, 217
218, 269, 532 Direct ink writing (DIW), 247 Direct internal reforming SOFCs (DIR SOFCs), 554 Direct methanol fuel cells, 29 Direct numerical simulation (DNS), 214
215 DOE. See Department of Energy (DOE) Dopant segregation, 565
566 Driving cycles, 281 Droplet adhesion force, 218 Droplet formation, 228
229 Drying method, 379 DSC. See Differential scanning calorimetry (DSC)

DST. See Dynamic stress test (DST) DTAB. See Decyltrimethylammonium bromide (DTAB) Durability testing, 254
255, 274
275 constant load operation, 275 start/stop durability, 275 Dynamic load cycles, 274 Dynamic programming algorithm, 293
294 Dynamic stress test (DST), 441
442

E ECMS. See Equivalent consumption minimization strategy (ECMS) ECSA. See Effective catalytic surface area (ECSA); Electrochemically active surface area (ECSA) EDX. See Energy-dispersive X-ray spectroscopy (EDX) Effective catalytic surface area (ECSA), 113 Effective diffusivity, 209
210 EG. See Ethylene glycol (EG) EIS. See Electrochemical impedance spectroscopy (EIS) Electrical conductivity, 41
42, 212
213, 331 Electric non-conductivity, 358 Electric traction motor, 13 Electrification, 20 Electrochemical cleaning, 155 Electrochemical deposition, 140
141 Electrochemical energy converter, 3
4 Electrochemical impedance spectroscopy (EIS), 215
216, 269
270, 431
432 Electrochemical kinetics, 394 Electrochemical limiting-current method, 217
218 Electrochemically active surface area (ECSA), 128
130, 218
219, 271
272, 400
401, 426
428 Electrochemical parameters, 218
219 Electrochemical reaction processes, 123, 339
341, 362 Electrochemistry, 178
179 Electrode kinetics, 31
42 agglomerate kinetics, 33
37 Butler
Volmer kinetics, 31
33 charge transfer coefficient, 39
41 conservation of charge, 38
39 electrical conductivities, 41
42 exchange current density, 39
41 ionic conductivities, 41
42 Electrode optimization, 84
88 Electrodeposition, 334
335

Index Electrolyte membrane, 11, 201
202 Electromotive force, 345
346 Electron beam (EB), 88 Electron transport, 378
379 Electroosmotic drag (EOD), 60, 75, 234 coefficient, 218, 494
495 Electrospinning, 182 Empirical models, 394 EMS. See Energy management strategy (EMS) End-of-life (EOL), 420
421 Energy converters, 4 Energy-dispersive X-ray spectroscopy (EDX), 216
217 Energy management strategy (EMS), 281 Energy sector, 3, 24 Energy storage devices, 281 Energy transfer, 394 Enthalpy of reaction, 3
4 EOD. See Electroosmotic drag (EOD) Equilibration, 266 Equivalent consumption minimization strategy (ECMS), 294 Ethanol, 555 Ethylene diamine, 132
133 Ethylene glycol (EG), 140
141, 351
352 European Hydrogen Roadmap, 21 European Union Fuel Cells & Hydrogen Joint Undertaking HYDRAITE project, 255
256 Evaporative cooling, 354 Exchange current density, 39
41 Experimental visualization techniques, 103
104 Ex situ accelerated stress test (AST), 421, 447
453 clamping force, 452
453 freeze
thaw cycling, 452 humidity cycling, 447
448 hygrothermal cycling, 448
449 liquid water wet
dry cycling, 449
452 temperature cycling, 448 vibration, 453 Ex situ mixing method, 137
138, 138t

F Fabrication techniques, 9 Faraday’s constant, 79, 225 Faraday’s law, 37, 257
258 Fault diagnosis techniques, 120
122 FCEVs. See Fuel cell electric vehicles (FCEVs)

603

FCEV with carbon capture and storage (FCEV-CCS), 24f FCHEV. See Fuel cell hybrid electric vehicle (FCHEV) FCVs. See Fuel cell vehicles (FCVs) FDA. See Fisher discriminant analysis (FDA) Federal Test Procedure (FTP-75), 281
282 FEP. See Fluorinated ethylene propylene (FEP) Fick’s law, 209 Figure of merit (FOM), 579
581 Finite element (FE) model, 116 Finite volume method (FVM), 104 Fisher discriminant analysis (FDA), 120
122 Flooded interface approach, 82
83 Flooding problem, 75
76 Flow boiling cooling, 354 Flow channels, 367, 394 Flow distributor, 367 Flow field design, 311
330 Flow fields optimization, 89
91 Flow field with guided flow path, 313
330 flow field design with metallic bipolar plates, 326 interdigitated flow channels, 319 latest developments in practical flow field design, 326
330 parallel flow channels, 315 serpentine flow channels, 315
319 strategies for improvement and hybridization of flow channel designs, 319
326 Fluorinated ethylene propylene (FEP), 205
208, 379
380 FOM. See Figure of merit (FOM) Forane 365 HX, 354 Forchheimer inertial effect, 89
90 Forward pressure control, 258 Fourier’s law, 343
345 Fourier transform infrared spectroscopy, 273 Freeze/thaw cycling, 452, 536 Freeze-tolerable design, 485 Freezing-induced liquid water movement, 503 Friedel-Crafts reaction, 135f Fuel cell (FC), 3
6, 29 aviation, 6 backup power generation, 6 development, 6
10 future of, 19
24 global distribution, 5 marine transportation, 6 passenger vehicles, 6

604

Index

Fuel cell (FC) (Continued) portable power generation, 6 present status of, 15
19 public transportation, 6 rail transportation, 6 in situ testing of, 103
104 total cost of ownership for, 25
26 for transportation, 10
15 unmanned aerial and underwater vehicles, 6 warehouse logistics, 5 Fuel cell electric vehicles (FCEVs), 3, 177, 199, 463, 575
576 capital costs of, 21
23 emerging transportation sectors for, 17 projected stock of, 23f Fuel Cell Enabling Technologies (FCET), 560 Fuel cell hybrid electric vehicle (FCHEV), 279 architecture, 280f body modeling, 284
287 fuzzy logic control for, 301 hybrid powertrain in, 281 model parameters, 297t road testing profiles, 281
284 Fuel cell performance prediction model, 394 Fuel cell power demand from hybrid powertrain, 287
296 energy management strategy, 287
288 fuel cell operation in vehicle applications, 288
289 state of the art of fuel cell hybrid electric vehicles energy management strategies, 289
296 Fuel cell short stack testing measurement techniques, 262
275 characterization of test station, 262
264 diagnostics, 267
273 durability testing, 274
275 stack operation, 264
267 operation and testing, 254
259 testing requirements, 259
262 Fuel cell stack optimization, 13, 91
92 Fuel cell vehicles (FCVs), 279, 350 Fuel filler, 13 Fuel supply system, 13 Fuel types, 550
556 alcohols, 555
556 ammonia, 551
554 hydrocarbons, 554 hydrogen, 551 Functionalization methods, 132
136 comparison of, 136t

doping heteroatoms, 132
133 oxidation treatment, 132 Fuzzy control, 358
360 Fuzzy logic control, 289
290, 292 FVM. See Finite volume method (FVM)

G Gallagher’s measurement, 494
495 Galvanic replacement reaction, 146
147 Galvanostatic control, 503
504 Galvanostatic EIS (GEIS), 507 Gas chromatography, 273 Gas diffusion electrode (GDE), 182 Gas diffusion layers (GDLs), 4
5, 29
30, 77, 104, 107
110, 109f, 115, 182, 201, 203f, 205
209, 225, 235
237, 268, 305, 343, 367, 394, 419, 430, 451f, 464 capillary pressure in, 377 electron conduction of, 378
379 isotropic properties of, 374
375 liquid water dynamics in, 376f mass transfer of, 386 oxygen transport in, 375 permeability and microstructure of, 108 pore-scale simulation models, 108
109 pore structure and transport properties of, 430 structural parameters of, 374
375 V-shaped porosity distributions, 108
109 Gaseous voltaic battery, 7 Gas flow channels (GC), 225, 261
262 Gas
liquid two-phase flow mechanism, 104
105 Gas-phase mass flux, 228
229 Gauss divergence theorem, 38 GDE. See Gas diffusion electrode (GDE) GDLs. See Gas diffusion layers (GDLs) Gemini fuel cells, 9 Gemini space program, 8 Genesis, 9 Genetic algorithm (GA), 95
96, 115, 290
292, 295 Gibbs free energy, 3
4, 32, 487
488 Gibbs
Thompson effect, 489 Gibbs
Thomson equation, 488 Global optimization problems, 294
295 Global warming, 177 Governing equations, 68 Grade resistance force, 286 Gradient-based optimization algorithm, 89 Graphene, 128, 332
333

Index Graphics processing unit, 104
105 Graphite, 332
333 Gravitational pressure loss, 232 Greenhouse gas (GHG) emissions, 1
2, 3f, 279 Greenlight Innovation (Canada), 260

H HDBAC. See Hexadecyldimethylbenzyl ammonium chloride (HDBAC) HDVs. See Heavy-duty vehicles (HDVs) Health management, 118
120 Heat generation, 340
343 Heating plates (HPs), 524
525 Heat loss, 518
519 Heat pipe cooling, 355 Heat source, 519
521 Heat source and heat generation rate, 516
518 Heat spreader cooling, 350 Heat transport, 48
51, 343
348 conservation of energy, 48
49 specific heat capacity, 49
51 thermal conductivity, 49
51 Heavy-duty vehicles (HDVs), 465 Helium pycnometry, 202
205 Henry’s law, 37 HESS. See Hybrid energy storage system (HESS) Hexadecyldimethylbenzyl ammonium chloride (HDBAC), 149
151 HFCs. See Hydrofluorocarbons (HFCs) HFR. See High-frequency resistance (HFR) High-frequency resistance (HFR), 478 High-order nonlinear fuel cell models, 117
118 High-temperature proton-exchange membrane (HT-PEM), 95
96 Homogenous equilibrium model, 232 Homogenous models, 81
82 Honda Clarity Fuel Cell, 10 Honda FCX Clarity, 10 Hooke’s law, 59 HOR. See Hydrogen oxidation reaction (HOR) Horiba FuelCon (Germany), 260 HRSs. See Hydrogen refueling stations (HRSs) HT-PEM. See High-temperature protonexchange membrane (HT-PEM) Humidity cycling durability test, 438
439, 447
448

605

Hybrid energy storage system (HESS), 287
288 Hybrid fuel cell powertrain, 279
281 Hybridization of flow channel designs, 325 Hybrid modeling, 213 Hydraulic permeation, 75 Hydricity economy, 6
7 Hydrocarbons, 554 Hydrofluorocarbons (HFCs), 1 Hydrogen, 3
6, 259
260, 357, 551, 575
576 future of, 19
24 history of, 6
10 phase diagram of, 578f Hydrogen fuel cells, 17
19 energy for compression, 579
581 liquefaction, 579
581 properties, 576
578 in storage method/materials, 576t Hydrogen oxidation reaction (HOR), 33 Hydrogen-powered automobiles, 9 Hydrogen pumping, 266
267 Hydrogen refueling station, 581
585 capital cost, 584
585 network development, 589
591 networks, 586
589 refueling at hydrogen refueling stations, 583
584 storage and source, 582
583 Hydrogen refueling stations (HRSs), 575
576, 582f Hydrogen tank, 14 Hydrogen technologies, 3 Hydrophilicity, 107 Hydrophobic coatings, 245 Hydrophobicity, 107 Hygrothermal cycling, 448
449 Hyundai ix35 FCEV Fuel Cell, 10 Hyundai Nexo, 326
327

I ICEs. See Internal combustion engines (ICEs) IEA. See International Energy Agency (IEA) IEC. See Ion exchange capacity (IEC) Imbibition method, 202
205 IMO. See International Maritime Organization (IMO) Impinging jet, 259 Impurity analysis, 273 Induction time of freezing and ice growth, 501
503 Injection modeling, 333
334

606

Index

In-plane buckling, 60 In situ accelerated stress test, 435
447 degradation rate by, 444
447 protocols, 436
443 In situ accelerated stress test (AST), 421 Integral coefficient, 359 Interdigitated flow channels, 319, 325, 370 Interdigitated flow field, 238
239 Internal combustion engines (ICEs), 1 International Energy Agency (IEA), 1
2 International Maritime Organization (IMO), 17 Intrinsic diffusion coefficient, 45
46 Intrinsic permeability, 375 Ion exchange capacity (IEC), 88 Ion flow tube mass spectrometry, 273 Ionic conductivity, 41
42, 213, 493
494 Ionomer, 182
184, 201
202 Ionomer to carbon (I/C) ratio, 184

J Japanese 10 mode (J10), 281
282 Japanese JC08, 283 Johnson
Mehl
Avrami
Kolmogorov framework, 501
502 Joule heat, 340
341, 362

K Ketjen Black EC300J (EC-300), 128 Ketjen Black EC600JD (EC-600), 128 K-means clustering algorithm, 120
122 Knudsen diffusion coefficient, 45
47, 210 Knudsen effect, 210 Kozeny
Carman equation, 376
377 Kronecker delta, 59

L Laboratory testing, 260 Lattice Boltzmann method (LBM), 81
82, 103
107, 232
233, 385 combination of, 122
123 pore-scale research in catalyst layer, 112
114 in gas diffusion layer, 107
110 in microporous layer, 110
111 Lawrence Livermore National Laboratory, 247 LBM. See Lattice Boltzmann method (LBM) Leak testing, 265 Learning-based strategy, 295
296 Lennard
Jones potential parameters, 46, 46t Leverett function, 66

Leverett-J empirical correlation, 385 Leverett-J function, 495 Linear programming, 289
290 Linear sweep voltammetry (LSV), 153, 270
271 Liquefaction, 579
581 Liquid cooling, 350
352 Liquid organic hydrogen carrier (LOHC), 549 Liquid-phase mass flux, 228
229 Liquid water transport, 310, 376
378 channels and flow fields, 237
241 biomimetic/fractal flow field, 240 interdigitated flow field, 238
239 metal foams and meshes, 240 parallel flow field, 238 serpentine flow field, 238 liquid management concerns and strategies, 241
242 patterned and structured porous media, 246
247 PEMFC architecture, 233
237 catalyst layer, 234
235 gas diffusion layer, 235
237 membrane, 234 microporous layer, 235 pressure and flow control, 242
243 startup/shutdown, 243
244 surface coatings, 245 thermal regulation and humidification, 243 two-phase flow basics, 226
233 ultrathin electrodes, 245
246 water production, 225
226 Liquid water wet
dry cycling, 449
452 Lithium-ion batteries, 279 Local mass transport resistance, 210
211 LOHC. See Liquid organic hydrogen carrier (LOHC) Long-haul trucking, 10
11 Long short-term memory (LSTM), 118, 242
243 Loschmidt cell, 217
218 Low-temperature operation, 562 LSTM. See Long short-term memory (LSTM) LSV. See Linear sweep voltammetry (LSV)

M Machine learning (ML), 106, 114
115, 242
243 Macroporous substrate (MPS), 484 Macroscopic modeling, 213
214 MAE. See Mean absolute error (MAE) Magnetron sputtering, 334
335

Index Marine technology, 17, 19t Maritime applications, 558
559 Markov chain, 295 Markov decision process, 296 Mass spectrometers, 263
264 Mass transfer, 123 in cathode catalyst layer, 380
386 in bulk catalyst layers, 385
386 in local region near catalysts, 382
384 in cathode gas diffusion layer and microporous layer, 373
380 electron transport, 378
379 innovations in gas diffusion layer structure and material, 379
380 liquid water transport, 376
378 oxygen transport, 374
376 in cathode gas flow fields, 367
373 in conventional flow channels, 368
370 gas diffusion layer and operation conditions, 373 in novel flow fields, 370
373 oxygen distribution and water removal, 367
368 Mass transport resistance, 75
76 MATLAB/Simulink, 299, 399 Maxwell-Stefan diffusion, 44
46 MEA. See Membrane electrode assembly (MEA) Mean absolute error (MAE), 118
120 Measurement techniques, 262
275 characterization of test station, 262
264 diagnostics, 267
273 durability testing, 274
275 stack operation, 264
267 Mechanical integrity, 562 Melamine, 132
133 Membrane, 234 Membrane electrode assembly (MEA), 29
30, 85
88, 177
178, 182
187, 225, 305
306, 306f, 308
309, 311, 315, 340
341, 422
423, 466, 529 dispersion of ionomer in solvent, 184
185 engineering cathode to improve water management, 185
187 equivalent weight of ionomer, 184
185 ionomer loading, 183
184 primary particle size, 184
185 Memorandum of understanding (MOU), 560 Mercury intrusion porosimetry (MIP), 216
217 Mercury porosimetry, 202
205 Meta-analysis, 533
534

607

Metal foams and meshes, 240 Metallic bipolar plates, 334
336 Metal
organic framework (MOF), 179 Metal-supported cells, 563 Metal-supported SOFCs (MS-SOFCs), 549
550 Methane (CH4), 1 2-methylimidazole (2-MeIM), 181 Micro-arc alloying, 334
335 Micro-phase separation, 489 Micropore flooding, 188
189 Microporous layers (MPLs), 4
5, 29
30, 104, 110
111, 115, 186
187, 201, 210, 225, 235, 305, 373
380, 430, 474
475 Microtubular SOFCs, 563
564 MIP. See Mercury intrusion porosimetry (MIP) Mixed ion and electron conductor (MIEC), 564 ML. See Machine learning (ML) Model predictive control (MPC), 117
118, 289
290, 358
359 Molecular dynamics (MD) simulation, 382
384 Molten carbonate fuel cells, 29 Monte Carlo methods, 122
123 MOU. See Memorandum of understanding (MOU) MPC. See Model predictive control (MPC) MPLs. See Microporous layers (MPLs) MPS. See Macroporous substrate (MPS) Multicomponent mass transport, 42
47 conservation of momentum, 42
44 Knudsen diffusion, 46
47 Maxwell
Stefan diffusion, 44
46 Multiphase processes, 219 Multiphysics processes, 219 Multiscale transport processes, 219 Multivariable optimization, 93
96 Multiwall carbon nanotubes, 352

N Nafion ionomer, 75, 339
340 Nafion membrane, 60 Nanofibers (NFs), 149
152 Nanofluids cooling, 352 Nanoparticles, 352 Nanoplates (NPs), 152
153 Nanorods (NRs), 149
152 Nanosheets (NSs), 152
153

608

Index

Nanostructured thin film (NSTF), 201
202, 245
246, 478 Nanotechnology, 352 Nanowire networks (NNWs), 137 Nanowires (NWs), 149
152 National Fuel Cell Technology Evaluation Center, 584
585 Natural aging protocols, 274 Natural gas, 20 Navier
Stokes conservation equations, 104
105 NEDC. See New European Driving Cycle (NEDC) Nernst equation, 39, 399 Nernst
Planck equation, 79 Net-zero target, 2 Neural networks, 295, 394 Neutron imaging, 215
216, 230
232 Neutron scattering (NR), 382, 385 New European Driving Cycle (NEDC), 281
282, 439
441, 441f Newton’s law, 345 Newton’s method, 406
407 Newton’s second law, 285 New York City cycle, 283 NextGenPropulsion, LLC (NGP), 560 NFs. See Nanofibers (NFs) Nickel alloys, 334 Nitrous oxide (N2O), 1 Non-homogenous models, 81
82 Nonuniform temperature distribution, 339
340 Novel dynamic programming technology, 289
290 Novel flow fields, 370
373 NPs. See Nanoplates (NPs) NRs. See Nanorods (NRs) NSs. See Nanosheets (NSs) NSTF. See Nanostructured thin film (NSTF) Nucleation rate, 500
501 Numerical models, 96
97, 103
104 Numerical optimization, 83
96 data-driven surrogate modeling, 93
96 electrode optimization, 84
88 flow fields optimization, 89
91 fuel cell stack optimization, 91
92 multivariable optimization, 93
96 operating condition optimization, 92
93 Nusselt number, 345 NWs. See Nanowires (NWs) Nyquist plot, 269
270

O OCP. See Open-circuit potential (OCP) OCV. See Open-circuit voltage (OCV) ODE. See Ordinary differential equation (ODE) Ohmic losses, 30
31, 272 Ohmic resistance, 269
270 Ohm’s law, 38, 79 Onboard traction battery pack, 13 One-dimensional (1D) models, 77 One-dimensional (1D) voltage data, 120
122 1D computational fluid dynamics model, 395
401, 396t 1D multiscale model, 211f 1D Pt-based nanostructures, 149
152 One-pot synthesis method, 137 Open-circuit potential (OCP), 30
31 Open-circuit voltage (OCV), 266, 424 Open-ended flow fields, 311
312 Open-end mode, 258 Operating condition optimization, 92
93, 255 Operating fuel cell stack, 310 Operating principle, 254
255 Operation strategy, 529
531 Optical feedback cavity enhanced absorption spectroscopy, 273 Optimal control algorithms, 359 Optimal control theory, 289
290 Optimization algorithm, 83
84 Optimization-based EMS, 296 Optimization-based strategy, 292
295 Optimization objective function, 287
288 Ordinary differential equation (ODE), 406
407 ORR. See Oxygen reduction reaction (ORR) Ostwald ripening, 400
401, 400f, 426
428 Oxygen excess ratio (OER), 117
118 Oxygen reduction reaction (ORR), 30
31, 75, 127, 177
178, 367, 464, 548 Oxygen starvation, 117 Oxygen transport, 374
376 Ozone (O3), 1

P PALS. See Positron annihilation lifetime spectroscopy (PALS) PANI. See Polyaniline (PANI) Parallel flow channels, 315 Parallel flow field design, 238, 320 Parameter optimization, 115
116 Partial differential equation (PDE), 68, 406
407

Index Partially fluorinated sulfonic acid membranes (part-FSAs), 88 Particle growth, 426
428 Particle migration, 426
428 Particle swarm optimization algorithm, 294 Passenger vehicles, 560
562 Patterned and structured porous media, 246
247 PCMs. See Phase-changing materials (PCMs) PDE. See Partial differential equation (PDE) PDMS. See Polydimethylsiloxane (PDMS) PEFCs. See Polymer electrolyte fuel cells (PEFCs) Peltier element, 469 PEMFC. See Proton exchange membrane fuel cell (PEMFC) Pentaethylene hexamine, 132
133 Percolation, 489 Perfluoro sulfonic acid (PFSA), 339
340, 380
381, 422, 463 Permeability, 212, 217
218 Perry Energy Systems, 9 PFSA. See Perfluoro sulfonic acid (PFSA) PGM. See Platinum group metal (PGM) Phase-change cooling, 352
356 Phase-change-induced (PCI) flow, 225 Phase-changing materials (PCMs), 355
356 Phase connectivity, 113 Phenyl disulfide, 132
133 PHM. See Prognostics and health management (PHM) Phosphoric acid fuel cells, 29 Physical vapor deposition, 334
335 PID control, 289
290 Planck’s constant, 32 Plasma surface diffusion alloying, 334
335 Platinum (Pt), 201
202, 380
381 Platinum group metal (PGM), 177, 234
235 Platinum group metal-free catalyst development, 178
182 in membrane electrode assembly, 182
187 dispersion of ionomer in solvent, 184
185 engineering cathode to improve water management, 185
187 equivalent weight of ionomer, 184
185 ionomer loading, 183
184 primary particle size, 184
185 mitigation strategies, 190
192 stability and durability of, 187
190 active site protonation, 189

609

attack of peroxide and associated radicals, 189
190 carbon oxidation, 189
190 demetallation, 189
190 micropore flooding, 188
189 PMP. See Pontryagin minimum principle (PMP) Poisson’s ratio, 218 Polarization curves, 30
31, 114
115, 268
269, 412, 431
432 Polyaniline (PANI), 179
180 Polydimethylsiloxane (PDMS), 66, 205
208, 379
380 Polymer electrolyte fuel cells (PEFCs), 11
12, 12f, 29, 177, 199
201, 225, 393, 422
424, 463, 575
576 ambient operation temperature of, 464 experimental techniques, 215
219 gas diffusion layer for, 201f model accuracy and computing speed, 408
411 modeling techniques, 213
215 1D computational fluid dynamics model, 395
401, 396t porous transport layers in, 201
208 pore size, 202
205 porosity, 202
205 wettability, 205
208 pseudo-2D computational fluid dynamics model, 401
408, 402t transport properties, 208
213 diffusivity, 209
210 electrical conductivity, 212
213 ionic conductivity, 213 local mass transport resistance, 210
211 permeability, 212 thermal conductivity, 212 Polymeric materials, 218 Polymeric matrix, 60 Polynomial equation, 64 Polytetrafluoroethylene (PTFE), 107
108, 186
187, 202, 227f, 260
261 Polyvinylidene fluoride (PVDF), 187 Pontryagin minimum principle (PMP), 294 Pore network (PN) method, 104 Pore network modeling (PNM), 122
123, 214
215 Pore-scale modeling techniques, 213, 215 Pore size distribution (PSD), 57
58, 202
205, 216
217 Pore/water interface, 210
211 Porosity of catalyst layer, 52
53

610

Index

Porous carbon black, 128 Porous electrode, 307 Porous transport layers, 200
208 pore size, 202
205 porosity, 202
205 wettability, 205
208 Positive temperature coefficient (PTC), 356 Positron annihilation lifetime spectroscopy (PALS), 216
217 Postpandemic economic recovery, 305 Postsynthesis treatment, 158 Potentiometry technique, 272 Power demand for fuel cell system hybrid electric vehicle body modeling, 284
287 hybrid electric vehicle road testing profiles, 281
284 hybrid electric vehicles energy management strategy, 296
300 hybrid fuel cell powertrain, 279
281 power demand from hybrid powertrain, 287
296 energy management strategy, 287
288 fuel cell operation in vehicle applications, 288
289 state of the art of fuel cell hybrid electric vehicles energy management strategies, 289
296 Power distribution coefficient, 298 Power electronics controller, 14 Prandtl number, 345 Predictive prevention strategy, 118 Pressure and flow control, 242
243 Prognostics, 118
120 Prognostics and health management (PHM), 118
120 Proportional coefficient, 359 Proportional-integral-derivative (PID) control, 117
118, 358
359, 393 Proton conduction equations, 394, 399 Proton exchange membrane fuel cell (PEMFC), 8, 11
12, 12f, 29
30, 73
74, 92, 115, 225, 253, 279, 305, 339
348, 367, 419, 464, 548
549, 575
576 accelerated stress tests of, 445t architecture, 233
237 catalyst layer, 234
235 gas diffusion layer, 235
237 membrane, 234 microporous layer, 235 chronological development of, 80t, 96

cold start, 361 conservation of water in different domains of, 67t control strategy, 358
361 adaptive control, 359
360 artificial neural network, 360
361 fuzzy control, 360 model predictive control, 359 proportional
integral
derivative control, 359 robust control, 360 electrode properties, 52
59 agglomerate density, 53
55 deformation of porous electrode, 57
59 porosity of catalyst layer, 52
53 specific area, 56
57 thicknesses of ionomer and liquid water films, 55
56 energy conservation equation in, 341
342 fuel cell modeling approach, 74
83 catalyst layer modeling, 82
83 modeling approaches, 76
79 modeling of water transport through membrane, 79
81 modeling of water transport through porous electrodes, 81
82 water formation and transport in fuel cells, 75
76 heat generation, 340
343 heat transport, 343
348 numerical optimization of, 83
96 data-driven surrogate modeling, 93
96 electrode optimization, 84
88 flow fields optimization, 89
91 fuel cell stack optimization, 91
92 multivariable optimization, 93
96 operating condition optimization, 92
93 operating temperature of, 354 operation principle of, 29
31 reaction kinetics and transport processes, 31
51 electrode kinetics, 31
42 heat transport, 48
51 multicomponent mass transport, 42
47 thermal management, 348
361 air cooling, 348
349, 349f heat spreader cooling, 350 liquid cooling, 350
352 nanofluids cooling, 352 phase-change cooling, 352
356 thermal management subsystem, 356
358, 357f

Index 3D neutron tomography of, 231f total waste heat generated in, 343 water management, 59
68 in Nafion ionomer with different membrane water content, 63
65 two-phase flow of gas
water mixture, 65
68 water phase-transfer, 59
62 water transport through membrane, 59
62 Proton transport, 234 Pseudo-2D computational fluid dynamics model, 401
408, 402t Pt-based electrocatalysts, 138
153 postsynthesis treatments of, 153
155 production of, 157
158 Pt-based open nanostructures, 145
149 1D Pt-based nanostructures, 149
152 2D Pt-based nanostructures, 152
153 chemical etching, 148
149 galvanic replacement reaction, 146
147 Pt-based polyhedrons, 141
145 metal precursors, 144 reaction temperature, 145 reductants, 142
144 secondary species, 145 solvents, 142 structure-directing agents, 144 support, 145 Pt-based spherical nanoparticles, 140
141 PTC. See Positive temperature coefficient (PTC) PTFE. See Polytetrafluoroethylene (PTFE) Public transportation, 10
11 PVDF. See Polyvinylidene fluoride (PVDF)

Q Q-learning algorithm, 296 Quantitative analysis, 269
270 Quasi-static transport, 218

R Radial basis function neural network (RBFNN), 116 Rail transport, 17, 19t Railways, 559
560 Random fiber model, 209 RBFNN. See Radial basis function neural network (RBFNN) RDE. See Rotating disk electrode (RDE) Reaction kinetics, 254
255 Recirculation, 259

611

Recurrent neural network (RNN), 118 Reduced graphene oxide (r-GO), 134f Reinforcement learning, 296 Relative effective diffusivity, 375 Relative humidity (RH), 469 Remaining useful life (RUL), 118 Reverse Carnot cycle, 579
581 Reverse water gasification, 566 Reversible degradation, 274
275 Reynolds number, 345 RH. See Relative humidity (RH) RMSE. See Root mean square error (RMSE) RNN. See Recurrent neural network (RNN) Robust control, 358
360 Robustness, 360 Robust phase-change cooling method, 355 Rolling resistance force, 284
285 Root mean square error (RMSE), 118
120 Rotating disk electrode (RDE), 184
185 RUL. See Remaining useful life (RUL) Rule-based EMS, 296 Rule-based methods, 294
295 Rule-based strategy, 290
292

S Scanning electron microscope (SEM), 202, 203f, 204f, 206f, 425f Scanning transmission X-ray microscope, 437
438 SEC. See Specific energy consumption (SEC) Second law of thermodynamics, 340
341 Seed-mediated growth, 149
151 Self-limiting fashion, 565 SEM. See Scanning electron microscope (SEM) Sequential-based programming approach, 294 Sequential dynamic programming, 292 Serpentine flow channels, 315
319, 370 Serpentine flow field, 238, 323 Shear modulus, 218 Shipping industry, 558
559 Short-stack testing, 253
256, 254f, 258 Single cell testing, 256
257, 259
260 Single fuel cell, 4
5, 29 Single-phase cooling method, 352
354 Single-phase flow model, 74
75 Single-phase mass transfer, 109
110 Single-task learning techniques, 120
122 Single-wall carbon nanotubes, 352 Sintering and loss of active area, 565 SL. See Stereolithography (SL) SOC. See State of charge (SOC)

612

Index

SOFCs. See Solid oxide fuel cells (SOFCs) SoH. See State-of-health (SoH) Solid carbon black, 128 Solid oxide fuel cells (SOFCs), 29, 122
123 applications, 556
562 aircraft, 557
558 auxiliary power units for heavy-duty trucks, 556
557 maritime applications, 558
559 passenger vehicles, 560
562 railways, 559
560 challenges and related efforts, 562
567 activity degradations, 564
565 carbon coking, 566
567 dopant segregation, 565
566 low-temperature operation, 562 mechanical integrity, 562 metal-supported cells, 563 sintering and loss of active area, 565 tubular configuration, 563
564 wet impregnation, 564 fuel types, 550
556 alcohols, 555
556 ammonia, 551
554 hydrocarbons, 554 hydrogen, 551 for transportation, 548
550 advantages and shortcomings, 548
549 cell configurations, 550 leveraging advantages and overcoming shortcomings, 549
550 Solid Polymer Electrolyte, 8 Solid polymer fuel cell (SPFC), 8 Solid polymer membrane, 279 Specific energy consumption (SEC), 579
581 Specific heat capacity, 49
51 SPFC. See Solid polymer fuel cell (SPFC) Stack operation, 264
267 break-in, 266
267 conditioning, 266
267 leak testing, 265 operation, 266
267 shutdown, 267 start-up, 265
266 STACKTEST project, 265 Start-up and shut-down (SU/SD) processes, 91
92 Start-up protocol, 265
266 Startup/shutdown, 243
244 State machine control, 290 State of charge (SOC), 289 State-of-health (SoH), 118

State of the art of fuel cell hybrid electric vehicles energy management strategies, 289
296 learning-based strategy, 295
296 optimization-based strategy, 292
295 rule-based strategy, 290
292 State-of-the-art technology, 336 Steady-state durability test, 431
435 degradation rate by, 432
435 protocols, 431
432 of proton-exchange membrane fuel cells, 433t Stefan
Boltzmann constant, 345 Stefan
Boltzmann law, 345 Stefan’s problem, 502
503 Stereolithography (SL), 247 Stochastic algorithm, 107
108 Stoichiometric-based gas control, 242 Stoichiometry, 257
258 Suboptimal design, 84 Subzero experiment, 466
471 characterization techniques, 471 control of thermal boundary conditions, 523
525 damage and mitigation in subzero scenarios, 471
485 damages to fuel cell components from freeze/thaw and subzero startup, 474
479 mitigation effects of material and design, 480 mitigation effects of operation control, 480
483 mitigation of damages in subzero, 483
485 decoupling thermal from water management, 532
533 four categories of experiment in subzero study, 466
468 heat source, 519
521 startup strategies and techniques, 525
533 lessons learned from laboratory startup study, 527
531 validation in the field, 531
532 states of water in fuel cell at subzero, 485
491 melting point depression, 488
489 modes of liquid water transport in ionomer and membrane, 492
495 modes of liquid water transport in porous media, 495

Index negligible water vapor transport, 495
496 phase change and icing dynamics, 496
503 states and thermodynamic properties, 485
488 water fill test, 503
511 water in membrane/ionomer, 489
491 water production and transport, 491
492 water transport dynamics in fuel cell at subzero, 491
503 temperature-dependent properties, 512
513 temperature difference, 521
523 test fixtures, 469, 470t test procedures and control of temperature, 469
470 thermal management issues highlighted from a lumped model, 513
519 heat loss, 518
519 heat source and heat generation rate, 516
518 thermal boundary conditions, 519 thermal mass, 513
516 Sulfonated polyimide (SPI-8) membranes, 447
448 Sulfur-cured polymers, 260
261 Supercapacitors, 279 Support vector machine (SVM), 95
96, 115 Surface coating techniques, 245, 334
335 Surface morphology, 344 Surface tension force, 66, 229
230 Surface wettability, 323 Surface X-ray scattering (SXS), 382 SVM. See Support vector machine (SVM) SXS. See Surface X-ray scattering (SXS) Synchrotron X-ray imaging, 57
58

T TABs. See Thermal adjustment boards (TABs) Tantalum
titanium oxide (Ta
TiOx), 191
192 TCO. See Total cost of ownership (TCO) TEAB. See Tetraethylammonium bromide (TEAB) TEM. See Transmission electron microscope (TEM) Temperature cycling, 448 Temperature difference, 521
523 Temporal Monte Carlo, 296 Test fixtures, 469, 470t Test stations, 260, 262
264

613

hydrogen utilization, 263
264 recirculation loop volume, 263 recirculation rate, 263 Tetraethylammonium bromide (TEAB), 144 Tetraethylene pentamine, 132
133 TGA. See Thermogravimetric analysis (TGA) Thermal adjustment boards (TABs), 524
525 Thermal annealing, 154 Thermal boundary conditions, 519 Thermal boundary layer, 344 Thermal conductivity, 49
51, 212, 326, 352 Thermal management subsystem, 356
358, 357f Thermal mass, 513
516 Thermal regulation and humidification, 243 Thermal reservoirs, 3
4 Thermogravimetric analysis (TGA), 131f Thermo-osmosis, 492
493 Thermostat control, 290
292, 356, 358 Thin film fully flooded approach, 82
83 3D 1 1D PEMFC model, 78 3D CFD fuel cell model, 115 3D CFD models, 394 3D multicomponent multiphase LB model, 108 3D printing, 247 Three-phase reaction interface, 103 Titanium alloys, 334 Titanium carbide (TiC), 334
335 Titanium nitride (TiN), 334
335 Top-level control strategy, 287 Toray gas diffusion layers, 204t Total cost of ownership (TCO), 25
26 Toyota Mirai (Mirai 1), 10, 326
327 Toyota Mirai 2, 329
330 Toyota Motor Corporation, 200 Toyota’s achievement, 531
532 TPBs. See Triple-phase boundaries (TPBs) Transfer coefficients, 40 Transfer molding, 333
334 Transmission electron microscope (TEM), 426
428 Transportation sector, 2, 575
576 Transport properties, 208
213 diffusivity, 209
210 electrical conductivity, 212
213 ionic conductivity, 213 local mass transport resistance, 210
211 permeability, 212 thermal conductivity, 212 Triethylene tetramine, 132
133 Triple-phase boundaries (TPBs), 33
34, 564

614

Index

TRU. See Truck refrigeration unit (TRU) Truck refrigeration unit (TRU), 556 Tubular configuration, 563
564 2D Pt-based nanostructures, 152
153 Two-phase flow, 226
233, 412 Two-phase modeling, 232
233

U Ultrathin electrodes, 245
246 Unassisted start strategy, 525
532 Unidirectional DC/DC converters, 279, 290 United Nations Economic Commission for Europe, 439
441 Unmanned aerial vehicles, 10
11 Urban Dynamometer Driving Schedule, 283 U.S. Department of Energy Hydrogen Program, 9
10 U.S. Federal Highway HFET, 283 UV
ozone irradiation, 154
155

V Vacuum plasma spraying, 556
557 Van der Waals interactions, 132 Variable heating and load control (VHLC), 91 VHLC. See Variable heating and load control (VHLC) Vibration, 453 VOF. See Volume of fluid (VOF) Voltage logging, 260
261 Voltage-sensing electrodes, 217
218 Voltammetry, 270
271 Volume displacement method, 218 Volume of fluid (VOF), 78, 229
230 Volumetric measurement techniques, 264 Vulcan XC-72 carbon black, 41
42, 56
57 Vulcan XC-72R (VXC-72R), 128

W Warehouse logistics, 10
11 Warming implicit scheme, 406
407 Water diffusion coefficient, 494 Water fill test (WFT), 465, 503
511 contribution and limitation of, 511 failure mechanism of subzero startup, 507
511

stages and features of, 503
506 Water in membrane/ionomer, 489
491 saturation water content, 489
491 three states of water in perfluorosulfonic acid membrane, 489 Water/ionomer interface, 210
211 Water production, 225
226 Water vapor (H2O), 1 Wavy flow channel, 321f Weber number, 240 Wet impregnation process, 549
550, 564 Wettability, 205
208 WFT. See Water fill test (WFT) Wheels traction force, 284
285 WLTC. See Worldwide Harmonized Light Vehicle Test Cycle (WLTC) Worldwide Harmonized Light Vehicle Test Cycle (WLTC), 281
282

X XPS. See X-ray photoelectron spectroscopy (XPS) X-ray computed tomography, 57
58, 186, 215
216, 422
423 X-ray imaging techniques, 236
237 X-ray photoelectron spectroscopy (XPS), 132, 216
217 X-ray radiography, 237 X-ray scattering, 385 X-ray tomography, 104, 230
232, 376f

Y Young
Laplace equation, 226
228 Young’s modulus, 59, 218

Z Zeolitic imidazolate framework (ZIF), 179 Zero-carbon electricity sources, 3 Zero-emission operation, 575
576 Zero-emission vehicles (ZEVs), 463 Zero GHG emission, 279 ZEVs. See Zero-emission vehicles (ZEVs) ZIF. See Zeolitic imidazolate framework (ZIF) Zigzag flow channel, 321f