Solid-state Hydrogen Storage Materials and Chemistry 1845692705, 9781845692704, 1420077880, 9781420077889

Hydrogen fuel cells are emerging as a major alternative energy source in transportation and other applications. Central

292 48 9MB

English Pages 583 Year 2008

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Cover Page ......Page 1
Title Page ......Page 2
Solid-state hydrogen storage: Materials and chemistry......Page 4
Contents......Page 6
Contributor contact details......Page 13
Preface......Page 17
Part I: Introduction......Page 19
1.1 Introduction......Page 20
1.2 High-pressure gas storage......Page 23
1.3 Liquid hydrogen......Page 24
1.4 Physically bound hydrogen......Page 25
1.5 Chemically bound hydrogen......Page 27
1.6 Hydrolytic evolution of hydrogen......Page 31
1.8 References......Page 32
2.2 Hydrogen technologies......Page 35
2.3 Hydrogen scenarios: from production to applications......Page 36
2.4.1 Hydrogen production......Page 40
2.4.2 Hydrogen storage......Page 41
2.4.3 Hydrogen transmission and distribution (T&D)......Page 42
Whole system incremental approach......Page 44
Step-change approach......Page 45
2.5.2 Portable fuel cells......Page 46
2.5.3 Stationary power......Page 48
2.5.4 Fuel cell vehicles (FCVs)......Page 49
2.5.5 Conclusions on applications......Page 51
2.6.1 Attitudes of the public......Page 52
2.7 Policy implications......Page 58
2.8 Conclusions......Page 61
2.9 References......Page 65
3.2 Materials challenges in hydrogen containment......Page 68
3.3 Hydrogen permeation......Page 70
3.4 Hydrogen embrittlement......Page 71
3.4.2 Effect of welding......Page 73
3.4.4 Effect of gas mixtures......Page 76
3.4.5 Effect of temperature......Page 78
3.4.6 Effect of cyclic loading......Page 79
3.5.1 Pressure vessels......Page 81
3.5.2 Pipelines......Page 82
3.6.1 Low-alloy and carbon steels......Page 83
3.6.2 Austenitic stainless steels......Page 86
3.6.4 Nickel alloys......Page 88
3.6.5 Aluminum alloys......Page 89
3.6.6 Other non-ferrous alloys......Page 90
3.6.8 Summary of materials selection......Page 91
3.7 Future trends......Page 92
3.8 Other sources......Page 93
3.10 References......Page 94
4.1 Introduction......Page 99
4.2 The behavior of solid-state hydrogen storage materials in systems......Page 101
4.3 Thermodynamic properties of hydrogen storage materials......Page 102
4.4 Thermal properties of hydrogen storage materials......Page 103
4.4.1 Thermal conductivity of a hydrogen storage packed particle bed......Page 104
Gas flow......Page 109
High thermal conductivity structures......Page 110
4.5.1 Absorption heat exchange......Page 111
4.5.2 Desorption heat exchange......Page 112
4.6.1 Volume expansion and decrepitation......Page 113
4.6.2 Contamination of metal hydrides – oxygen and water reactivity......Page 114
The thermochemistry characteristics of sodium alanates......Page 115
System implications of the oxidation reactions......Page 117
4.8 Future trends......Page 118
4.10 References......Page 119
Part II: Analysing hydrogen interactions......Page 121
5.1 Introduction......Page 122
5.2 Principles of diffraction......Page 123
5.3 X-ray and neutron diffraction......Page 127
5.3.1 Sources......Page 130
5.3.2 Powder diffractometers......Page 132
5.4 The use of powder diffraction data......Page 134
5.4.2 Powder indexing and extraction of Bragg intensities......Page 135
5.4.3 Structure solution from powder diffraction data......Page 136
5.4.4 Structure refinements – the Rietveld method......Page 137
5.5.1 Structure of Mg(BD4)2......Page 141
5.5.2 In situ diffraction studies of AlH3......Page 143
5.6 Future trends......Page 146
5.7 Sources of further information and advice......Page 147
5.8 References......Page 148
6.1 Introduction......Page 150
6.2.1 Advantages of the neutron scattering technique for investigating solid-state hydrogen storage systems......Page 151
6.3 Studies of light metal hydrides......Page 152
6.5 The basic theory of neutron scattering......Page 153
Incoherent inelastic scattering from a proton in a harmonic potential well: the Einstein oscillator model......Page 157
Comparison with ab initio calculations......Page 159
Incoherent scattering from systems with interacting protons......Page 160
6.6.2 Inelastic neutron scattering from molecular hydrogen......Page 163
6.6.3 The theory of quasi-elastic neutron scattering......Page 165
The Chudley–Elliott model......Page 167
Localised diffusion: the elastic structure factor......Page 168
6.7.1 Inelastic neutron scattering (INS) measurements on Laves phase hydrides......Page 169
6.7.2 Inelastic neutron scattering (INS) on alanates......Page 171
6.7.3 Inelastic neutron scattering (INS) measurements on hydrides of magnesium and its compounds......Page 172
6.7.4 Inelastic neutron scattering (INS) measurements on borohydrides......Page 173
6.8.1 Hydrogen trapped in carbon nanotubes......Page 174
6.8.2 Hydrogen trapped in zeolites......Page 178
6.8.3 Hydrogen trapped in ice clathrates......Page 180
6.8.4 Molecular hydrogen storage in metal oxide frameworks......Page 181
6.9.1 Quasi-elastic scattering studies of hydrogen in Laves phases......Page 182
6.9.2 Quasi-elastic scattering from hydrogen in alanates......Page 183
6.10 Conclusions......Page 184
6.11 References......Page 185
7.1 Introduction......Page 189
7.2 Compressibilities of hydrogen and deuterium......Page 190
7.3 Measurement regimes......Page 192
7.4 Measurement techniques......Page 194
7.4.1 Sieverts technique......Page 195
Large-aliquot effect......Page 198
Isobaric operation of a Sieverts hydrogenator......Page 200
7.4.2 Gravimetric technique......Page 202
7.5.1 Sieverts apparatus......Page 204
Equivalent volume model......Page 205
Divided volume model......Page 206
7.5.2 Gravimetric apparatus......Page 207
Two-sided balance with symmetric temperature distribution......Page 208
Single-sided balance......Page 209
7.6 The sample volume problem......Page 210
7.6.1 Is the calibrating gas inert?......Page 211
7.6.2 Sieverts technique......Page 213
7.6.3 Gravimetric technique......Page 214
7.7.1 Variable-volume technique......Page 215
7.8 Summary and conclusions......Page 217
7.10 References......Page 218
8.1 Introduction......Page 220
8.2 Hydrogen interactions with carbons: physisorption and chemisorption......Page 222
8.2.1 Physisorption......Page 223
8.2.2 Chemisorption......Page 225
8.3.1 Theoretical predictions of an ideal hydrogen storage medium......Page 226
8.3.2 The metal—H2 binding mechanism......Page 227
8.3.3 Further theoretical investigations......Page 229
8.3.4 Experimental attempts at realizing the theoretical predictions......Page 230
8.4 Conclusions and future trends......Page 231
8.6 References......Page 232
Part III: Physically bound hydrogen storage......Page 236
9.1 Introduction......Page 237
9.2 Hydrogen encapsulation at high temperatures......Page 238
9.3 Low-temperature physisorption......Page 241
9.4 Storage at room temperature: encapsulation, physisorption, chemisorption and spillover......Page 245
9.5 Spectroscopic studies......Page 248
9.6 Theoretical studies and modelling......Page 257
9.7 Other potential applications of zeolites in a hydrogen energy system......Page 262
9.8 Prospects for the use of zeolites in a hydrogen energy system......Page 263
9.9 Acknowledgements......Page 265
9.10 References......Page 266
10.1 Introduction......Page 275
10.2 Storage of hydrogen in solids......Page 276
10.3 Carbon nanostructures and hydrogen storage......Page 277
10.4 Supercritical adsorption in nanoporous materials......Page 281
10.5 Theory......Page 283
10.5.1 Virial expansion......Page 284
10.5.2 Langmuir model......Page 286
10.5.4 Dubinin pore-filling approach......Page 287
10.6.1 Activated carbons......Page 288
10.6.2 Single-wall nanotubes......Page 291
10.6.3 Other carbon nanostructures......Page 295
10.7 Beyond carbon nanostructures......Page 297
10.10 References......Page 298
11.1 Introduction......Page 302
11.1.1 Design principles......Page 303
Metal nodes......Page 304
11.2 Hydrogen storage in particular metal–organic framework (MOF) materials......Page 305
11.2.1 Prussian Blue analogues......Page 306
11.2.3 Isoreticular Zn(II) carboxylates: the iso-recticular metal–organic framework (IRMOF) series......Page 307
11.2.4 Cu(II) carboxylates......Page 312
11.2.6 Non-transition metal carboxylates......Page 314
11.2.7 Metal–organic frameworks (MOFs) based upon pyridine-carboxylate linkers......Page 315
11.2.8 Mn and Cu tetrazolates and related ligands......Page 316
11.3 Interactions of H2 with metal–organic frameworks: experiments and modelling......Page 317
11.3.1 Interactions of hydrogen with exposed metal sites......Page 318
11.3.2 Interactions of hydrogen with metal–organic frameworks (MOFs) without exposed metal sites......Page 320
11.3.3 Modelling......Page 321
11.5 References......Page 322
Part IV: Chemically bound hydrogen storage......Page 327
12.1 Introduction......Page 328
Pressure–composition isotherms......Page 330
12.2.2 Crystal structure aspects of metal hydrides......Page 332
12.2.3 Binary alloy/Intermetallic hydrides......Page 334
12.2.4 Effect of solid-state impurities – microalloying......Page 335
12.3 Long-term stability of metal hydrides......Page 338
12.4.1 Examples of thermal and pressure cycling of classical hydrides......Page 342
12.4.2 Decrepitation of hydrides – disproportionation aspects......Page 350
12.4.3 Intrinsic thermal cycling elemental vanadium alloyed with 0.5 at.%C hydrides......Page 351
12.5.1 Extrinsic gaseous impurity effects on LaNi5 and Fe-Ti......Page 354
12.5.2 Cycling test and response to oxygen as a minor impurity degradation behavior of AB5 hydrides......Page 356
12.6.1 Pressure extrinsic cycling studies on imide/amide system......Page 358
12.7 Conclusions......Page 359
12.9 References......Page 361
13.2 Background to magnesium and magnesium hydride......Page 370
13.2.1 Properties and structure......Page 372
13.3 Thermodynamics and hydride mechanisms......Page 375
13.4 Ball milling to improve hydrogen sorption behaviour......Page 376
13.5 Metal and alloy additives......Page 379
13.6 Metal oxide catalysts......Page 381
13.7 Kinetic models of hydrogen absorption......Page 385
13.8 Conclusions and future work......Page 387
13.9 References......Page 389
14.1 Introduction......Page 394
14.2.1 NaAlH4......Page 395
14.2.3 KAlH4......Page 396
14.2.5 Na3AlH6 and Li3AlH6......Page 401
Dehydrogenation reactions......Page 403
Rehydrogenation reactions......Page 405
14.3.2 Doped NaAlH4......Page 406
Mechanistic studies......Page 410
14.4 Density-functional calculations of alkali and alkaline-earth alanates......Page 417
14.4.1 Crystal structures of alanates......Page 418
14.4.3 Stability and thermodynamics of decomposition......Page 419
14.4.5 Mechanism of the catalytic effects......Page 422
14.4.6 Destabilization......Page 423
14.5.1 Thermodynamic considerations......Page 424
14.5.2 Kinetic considerations......Page 426
14.5.4 Safety......Page 427
14.7 References......Page 428
15.2.1 Alkali borohydrides......Page 433
15.2.2 Alkaline-earth borohydrides, trivalent and tetravalent borohydrides......Page 436
Alkali borohydrides, MBH4 (M = Li, Na, K, Rb and Cs)......Page 438
Alkaline-earth borohydrides, M(BH4)2 (M = Be, Mg and Ca)......Page 439
Trivalent and tetravalent borohydrides, M(BH4)3 (M = Al) and M(BH4)4 (M = Zr and Hf)......Page 443
Alkali borohyrides, MBH4 (M = Li, Na, K, Rb and Cs)......Page 444
Alkaline-earth borohydrides, trivalent and tetravalent borohydrides, M(BH4)n (n…......Page 446
Alkali borohydrides......Page 448
Alkaline-earth borohydrides......Page 451
Trivalent and tetravalent borohydrides......Page 452
15.4.2 Controlling thermodynamics......Page 453
15.4.3 Promoting kinetics......Page 456
15.7 References......Page 458
16.2 The lithium–nitrogen–hydrogen system......Page 463
16.2.1 Lithium nitride and hydrogen: a historical perspective......Page 464
16.2.2 Lithium imide and lithium amide......Page 466
16.2.3 Hydrogen storage in the Li–N–H system......Page 469
16.2.4 The effect of additives and particle size on hydrogen sorption......Page 473
16.3 The imides and amides of the group 2 elements......Page 477
16.4.1 The Li–Mg–N–H system......Page 479
16.4.2 The Li–Ca–N–H, Mg–Ca–N–H and Na–Mg–N–H systems......Page 484
16.5 Future trends and conclusions......Page 485
16.6 Acknowledgements......Page 486
16.7 References......Page 487
17.1 Introduction......Page 491
17.2 Thermodynamic destabilisation......Page 493
17.3 Complex hydride–metal hydride systems......Page 495
17.4 Complex hydride–non-hydride systems......Page 504
17.5 Complex hydride–complex hydride systems......Page 506
17.6 Other destabilisation multicomponent systems......Page 508
17.7 Future trends......Page 509
17.8 References......Page 510
18.1 Introduction......Page 513
18.2 Organic hydrides: chemistry and reactions for hydrogen storage and supply......Page 514
18.3 Spray-pulsed reactors for efficient hydrogen supply by organic hydrides......Page 519
18.4 Hydrogen storage and supply by organic hydrides......Page 525
18.5 Hydrogen delivery using organic hydrides for fuel-cell cars and domestic power systems......Page 531
18.6 High-density electric power delivery using organic hydride carriers......Page 534
18.7 Rechargeable direct fuel cells using organic hydrides......Page 536
18.8 Hydrogen delivery networks using organic hydrides......Page 541
18.9 References......Page 544
19.1 Introduction......Page 546
19.2 Indirect hydrogen storage in ammonia......Page 547
19.3 Compact storage in solid metal ammine materials......Page 550
19.4.1 Coordination complexes: a well-known class of materials......Page 554
19.4.3 Thermal release of ammonia......Page 558
19.4.4 Trends in metal ammine stabilities......Page 559
19.5 Nano- to macro-scale design of metal ammines......Page 561
19.5.1 Atomic structures and ab-/desorption pathways......Page 562
19.5.2 Diffusion and porosity......Page 563
19.5.3 Stabilities and binding energies......Page 564
19.6.1 From research project to real systems and applications......Page 566
Integrated fuel cell system......Page 567
Solid oxide fuel cell (SOFC)......Page 569
19.6.3 Ammonia decomposition......Page 570
19.6.4 Fuel supply: bulk production of storage materials......Page 572
19.6.5 Sources of further information and advice......Page 573
19.7 References......Page 574
20.1 Challenges in hydrogen applications......Page 578
20.2.1 Porous materials......Page 579
20.2.3 Ammines and liquid hydrides......Page 580
20.2.4 Metal hydrides......Page 581
20.4 References......Page 582
Recommend Papers

Solid-state Hydrogen Storage Materials and Chemistry
 1845692705, 9781845692704, 1420077880, 9781420077889

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

i

Solid-state hydrogen storage

© 2008, Woodhead Publishing Limited

ii

Related titles: Materials for energy conversion devices (ISBN 978-1-85573-932-1) As the finite capacity and pollution problems of fossil fuels grow more pressing, new sources of more sustainable energy are being developed. Materials for energy conversion devices summarises the key research on new materials which can be used to generate clean and renewable energy or to help manage problems from existing energy sources. The book discusses the range of materials that can be used to harness and convert solar energy in particular, including the properties of oxide materials and their use in producing hydrogen fuel. It covers thermoelectric materials and devices for power generation, ionic conductors and new types of fuel cell. Materials for fuel cells (ISBN 978-1-84569-330-5) This authoritative reference work provides a comprehensive review of the materials used in hydrogen fuel cells, which are predicted to emerge as an important alternative energy option in transportation and domestic use over the next few years. The design and selection of the materials are critical to the correct and long-term functioning of fuel cells and must be tailored to the type of fuel cell. The book looks in detail at each type of fuel cell and the specific material requirements and challenges. Chapters cover material basics, modelling, performance and recyclability. Microjoining and nanojoining (ISBN 978-1-84569-179-0) Many recent advances in technology have been associated with nanotechnology and the miniaturisation of components, devices and systems of electronic, precision and medical products. Among the key technical prerequisites, effective microjoining plays an essential role. Microjoining, having been closely associated with the evolution of microelectronic packaging, covers a much broader area, and is essential for manufacturing many electronic, precision and medical products. The book reviews the basics of microjoining and microjoining processes. Chapters discuss microjoining of materials and the applications of microjoining. Details of these and other Woodhead Publishing materials books, as well as materials books from Maney Publishing, can be obtained by: • visiting our web site at www.woodheadpublishing.com • contacting Customer Services (e-mail: [email protected]; fax: +44 (0) 1223 893694; tel.: +44 (0) 1223 891358 ext. 130; address: Woodhead Publishing Ltd, Abington Hall, Granta Park, Great Abington, Cambridge CB21 6AH, England) If you would like to receive information on forthcoming titles, please send your address details to: Francis Dodds (address, tel. and fax as above; e-mail: [email protected]). Please confirm which subject areas you are interested in. Maney currently publishes 16 peer-reviewed materials science and engineering journals. For further information visit www.maney.co.uk/journals

© 2008, Woodhead Publishing Limited

iii

Solid-state hydrogen storage Materials and chemistry Edited by Gavin Walker

WPTF2005

Woodhead Publishing and Maney Publishing on behalf of The Institute of Materials, Minerals & Mining

CRC Press Boca Raton Boston New York Washington, DC

WOODHEAD

PUBLISHING LIMITED

Cambridge England

© 2008, Woodhead Publishing Limited

iv Woodhead Publishing Limited and Maney Publishing Limited on behalf of The Institute of Materials, Minerals & Mining Woodhead Publishing Limited, Abington Hall, Granta Park, Great Abington, Cambridge CB21 6AH, England www.woodheadpublishing.com Published in North America by CRC Press LLC, 6000 Broken Sound Parkway, NW, Suite 300, Boca Raton, FL 33487, USA First published 2008, Woodhead Publishing Limited and CRC Press LLC © 2008, Woodhead Publishing Limited The authors have asserted their moral rights. This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. Reasonable efforts have been made to publish reliable data and information, but the authors and the publishers cannot assume responsibility for the validity of all materials. Neither the authors nor the publishers, nor anyone else associated with this publication, shall be liable for any loss, damage or liability directly or indirectly caused or alleged to be caused by this book. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming and recording, or by any information storage or retrieval system, without permission in writing from Woodhead Publishing Limited. The consent of Woodhead Publishing Limited does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific permission must be obtained in writing from Woodhead Publishing Limited for such copying. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation, without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. Library of Congress Cataloging in Publication Data A catalog record for this book is available from the Library of Congress. Woodhead Publishing ISBN 978-1-84569-270-4 (book) Woodhead Publishing ISBN 978-1-84569-494-4 (e-book) CRC Press ISBN 978-1-4200-7788-9 CRC Press order number WP7788 The publishers’ policy is to use permanent paper from mills that operate a sustainable forestry policy, and which has been manufactured from pulp which is processed using acid-free and elementary chlorine-free practices. Furthermore, the publishers ensure that the text paper and cover board used have met acceptable environmental accreditation standards. Typeset by Replika Press Pvt Ltd, India Printed by T J International Limited, Padstow, Cornwall, England

© 2008, Woodhead Publishing Limited

v

Contents

Contributor contact details

xiii

Preface

xvii

Part I Introduction 1

Hydrogen storage technologies

3

G. WALKER, University of Nottingham, UK

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

Introduction High-pressure gas storage Liquid hydrogen Physically bound hydrogen Chemically bound hydrogen Hydrolytic evolution of hydrogen Summary References

3 6 7 8 10 14 15 15

2

Hydrogen futures: emerging technologies for hydrogen storage and transport

18

P. EKINS, King’s College London and P. BELLABY, University of Salford, UK

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9

Introduction Hydrogen technologies Hydrogen scenarios: from production to applications Hydrogen economics Hydrogen end-use applications Public acceptability of hydrogen Policy implications Conclusions References

© 2008, Woodhead Publishing Limited

18 18 19 23 29 35 41 44 48

vi

Contents

3

Hydrogen containment materials

51

B. P. SOMERDAY and C. SAN MARCHI, Sandia National Laboratories, USA

3.1 3.2 3.3 3.4 3.5 3.6

51 51 53 54 64

3.7 3.8 3.9 3.10

Introduction Materials challenges in hydrogen containment Hydrogen permeation Hydrogen embrittlement Service experience with hydrogen containment Materials used in the design of hydrogen containment structures Future trends Other sources Acknowledgements References

4

Solid-state hydrogen storage system design

82

66 75 76 77 77

D. E. DEDRICK, Sandia National Laboratories, USA

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10

Introduction The behavior of solid-state hydrogen storage materials in systems Thermodynamic properties of hydrogen storage materials Thermal properties of hydrogen storage materials System heat exchange design Safe systems design Enabling safe systems based on hydrogen sorption materials Future trends Sources of further information and advice References

82 84 85 86 94 96 101 101 102 102

Part II Analysing hydrogen interactions 5

Structural characterisation of hydride materials

107

B. C. HAUBACK, Institute for Energy Technology, Norway

5.1 5.2 5.3 5.4 5.5 5.6

Introduction Principles of diffraction X-ray and neutron diffraction The use of powder diffraction data Examples of structures and results from powder diffraction studies Future trends

© 2008, Woodhead Publishing Limited

107 108 112 119 126 131

Contents

vii

5.7 5.8

Sources of further information and advice References

132 133

6

Neutron scattering studies for analysing solid-state hydrogen storage

135

D. K. ROSS, University of Salford, UK

6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 6.11 7

Introduction The neutron scattering method Studies of light metal hydrides Studies of molecular hydrogen trapping in porous materials The basic theory of neutron scattering Theory of inelastic neutron scattering Inelastic scattering measurements on solid-state hydrides Inelastic neutron scattering from molecular hydrogen trapped on surfaces Quasi-elastic scattering measurements on hydrogen diffusing in hydrides Conclusions References Reliably measuring hydrogen uptake in storage materials

135 136 137 138 138 142 154 159 167 169 170

174

E. MACA. GRAY, Griffith University, Australia

7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10

Introduction Compressibilities of hydrogen and deuterium Measurement regimes Measurement techniques System characterisation The sample volume problem The variable-volume hydrogenator Summary and conclusions Acknowledgements References

174 175 177 179 189 195 200 202 203 203

8

Modelling of carbon-based materials for hydrogen storage

205

J. ÍÑIGUEZ, Institut de Ciencia de Materials de Barcelona (ICMAB-CSIC), Spain

8.1

Introduction

© 2008, Woodhead Publishing Limited

205

viii

Contents

8.2

Hydrogen interactions with carbons: physisorption and chemisorption Predictions for hydrogen storage in carbon nanostructures coated with light transition metals Conclusions and future trends Sources of further information and advice References

8.3 8.4 8.5 8.6

207 211 216 217 217

Part III Physically bound hydrogen storage 9

Storage of hydrogen in zeolites

223

P. A. ANDERSON, University of Birmingham, UK

9.1 9.2 9.3 9.4

9.9 9.10

Introduction Hydrogen encapsulation at high temperatures Low-temperature physisorption Storage at room temperature: encapsulation, physisorption, chemisorption and spillover Spectroscopic studies Theoretical studies and modelling Other potential applications of zeolites in a hydrogen energy system Prospects for the use of zeolites in a hydrogen energy system Acknowledgements References

249 251 252

10

Carbon nanostructures for hydrogen storage

261

9.5 9.6 9.7 9.8

223 224 227 231 234 243 248

P. BÉNARD and R. CHAHINE, Institut de recherche sur l’hydrogène, Canada

10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8 10.9 10.10

Introduction Storage of hydrogen in solids Carbon nanostructures and hydrogen storage Supercritical adsorption in nanoporous materials Theory Adsorption of hydrogen on activated carbons and carbon nanostructures Beyond carbon nanostructures Conclusions Acknowledgments References

© 2008, Woodhead Publishing Limited

261 262 263 267 269 274 283 284 284 284

Contents

11

Metal–organic framework materials for hydrogen storage

ix

288

X. LIN, J. JIA, N. R. CHAMPNESS, P. HUBBERSTEY and M. SCHRÖDER, University of Nottingham, UK

11.1 11.2 11.3 11.4 11.5

Introduction Hydrogen storage in particular metal–organic framework (MOF) materials Interactions of H2 with metal–organic frameworks: experiments and modelling Conclusions and future trends References

288 291 303 308 308

Part IV Chemically bound hydrogen storage 12

Intermetallics for hydrogen storage

315

D. CHANDRA, University of Nevada, USA

12.1 12.2 12.3 12.4 12.5 12.6 12.7 12.8 12.9

Introduction Metal hydrides Long-term stability of metal hydrides Intrinsic testing of intermetallic hydrides Extrinsic testing of intermetallic hydrides Extrinsic cycling of complex hydrides Conclusions Acknowledgements References

315 317 325 329 341 345 346 348 348

13

Magnesium hydride for hydrogen storage

357

D. GRANT, University of Nottingham, UK

13.1 13.2 13.3 13.4 13.5 13.6 13.7 13.8 13.9

Introduction Background to magnesium and magnesium hydride Thermodynamics and hydride mechanisms Ball milling to improve hydrogen sorption behaviour Metal and alloy additives Metal oxide catalysts Kinetic models of hydrogen absorption Conclusions and future work References

© 2008, Woodhead Publishing Limited

357 357 362 363 366 368 372 374 376

x

Contents

14

Alanates as hydrogen storage materials

381

C. JENSEN, University of Hawaii at Manoa, Hawaii, Y. WANG and M. Y. CHOU, Georgia Institute of Technology, USA

14.1 14.2 14.3

381 382

14.5 14.6 14.7

Introduction Atomic structure of alanates Dehydrogenation and rehydrogenation reactions in alanates Density-functional calculations of alkali and alkaline-earth alanates Future trends Conclusions References

15

Borohydrides as hydrogen storage materials

420

14.4

390 404 411 415 415

Y. NAKAMORI and S. ORIMO, Tohoku University, Japan

15.1 15.2 15.3 15.4 15.5 15.6 15.7

Introduction Synthesis of borohydrides Structure of borohydrides Dehydrogenation and rehydrogenation reactions Future trends Acknowledgements References

420 420 425 435 445 445 445

16

Imides and amides as hydrogen storage materials

450

D. H. GREGORY, University of Glasgow, UK

16.1 16.2 16.3 16.4 16.5 16.6 16.7

Introduction The lithium–nitrogen–hydrogen system The imides and amides of the group 2 elements Mixed metal imides and amides Future trends and conclusions Acknowledgements References

450 450 464 466 472 473 474

17

Multicomponent hydrogen storage systems

478

G. WALKER, University of Nottingham, UK

17.1 17.2 17.3 17.4 17.5 17.6

Introduction Thermodynamic destabilisation Complex hydride–metal hydride systems Complex hydride–non-hydride systems Complex hydride–complex hydride systems Other destabilisation multicomponent systems

© 2008, Woodhead Publishing Limited

478 480 482 491 493 495

Contents

xi

17.7 17.8

Future trends References

496 497

18

Organic liquid carriers for hydrogen storage

500

M. ICHIKAWA, Hokkaido University, Japan

18.1 18.2

18.7 18.8 18.9

Introduction Organic hydrides: chemistry and reactions for hydrogen storage and supply Spray-pulsed reactors for efficient hydrogen supply by organic hydrides Hydrogen storage and supply by organic hydrides Hydrogen delivery using organic hydrides for fuel-cell cars and domestic power systems High-density electric power delivery using organic hydride carriers Rechargeable direct fuel cells using organic hydrides Hydrogen delivery networks using organic hydrides References

521 523 528 531

19

Indirect hydrogen storage in metal ammines

533

18.3 18.4 18.5 18.6

500 501 506 512 518

T. VEGGE, R. Z. SØRENSEN, A. KLERKE, J. S. HUMMELSHØJ, T. JOHANNESSEN, J. K. NØRSKOV and C. H. CHRISTENSEN, Technical University of Denmark, Denmark

19.1 19.2 19.3 19.4 19.5 19.6 19.7

Introduction Indirect hydrogen storage in ammonia Compact storage in solid metal ammine materials Selecting metal ammine storage materials Nano- to macro-scale design of metal ammines Commercial applications and future trends References

533 534 537 541 548 553 561

20

Conclusion: technological challenges in hydrogen storage

565

G. WALKER, University of Nottingham, UK

20.1 20.2 20.3 20.4

Challenges in hydrogen applications Challenges in materials for storage Conclusions References

© 2008, Woodhead Publishing Limited

565 566 569 569

xiii

Contributor contact details

(* = main contact)

Editor and Chapter 1, 17 and 20

Chapter 3

Dr Gavin Walker School of Mechanical, Materials and Manufacturing Engineering University of Nottingham University Park Nottingham NG7 2RD UK

Dr B. P. Somerday* and Dr C. San Marchi Sandia National Laboratories PO Box 969 MS 9402 Livermore CA 94551-0969 USA

Email: [email protected]

Email: [email protected]

Chapter 2

Chapter 4

Professor Paul Ekins* Policy Studies Institute 50 Hanson Street London W1W 6UP UK

Daniel E. Dedrick Senior Member of Technical Staff Sandia National Laboratories PO Box 969 MS 9409 Livermore CA 94551-0969 USA

Email: [email protected] [email protected]

Professor Paul Bellaby University of Salford Manchester M5 4WT UK Email: [email protected]

© 2008, Woodhead Publishing Limited

Email: [email protected]

xiv

Contributor contact details

Chapter 5

Chapter 8

Professor Bjørn C. Hauback Physics Department Institute for Energy Technology PO Box 40 NO-2027 Kjeller Norway

Dr Jorge Íñiguez Institut de Ciencia de Materials de Barcelona (ICMAB-CSIC) Campus UAB 08193 Bellaterra (Barcelona) Spain

Email: [email protected]

Email: [email protected]

Chapter 6

Chapter 9

Professor D. K. Ross Functional Materials Centre Director Institute for Materials Research 108a Maxwell Building University of Salford Salford Manchester M5 4WT UK

Dr Paul A. Anderson School of Chemistry University of Birmingham Edgbaston Birmingham B15 2TT UK

Email: [email protected]

Chapter 7 Professor Evan Gray Griffith University Nanoscale Science and Technology Centre Nathan 4111 Australia Email: [email protected]

© 2008, Woodhead Publishing Limited

Email: [email protected]

Chapter 10 Professor Pierre Bénard* and Professor Richard Chahine Institut de recherche sur l’hydrogène Université du Québec à TroisRivières Québec Canada G9A 5H7 Email: [email protected] [email protected]

Contributor contact details

Chapter 11

Chapter 14

Xiang Lin*, Junhua Jia, Neil R. Champness, Peter Hubberstey and Martin Schröder School of Chemistry University of Nottingham Nottingham NG7 2RD UK

Professor Craig Jensen* University of Hawaii at Manoa Department of Chemistry 2545 The Mall Honolulu Hawaii HI 96822-2275

Email: [email protected] [email protected] [email protected]

Email: [email protected]

Chapter 12 Professor Dhanesh Chandra Engineering Department Metallurgical and Materials Sciences Division MS 388 University of Nevada Reno NV 89557 USA

xv

Dr Y Wang Georgia Institute of Technology GA 30332-0295 USA Email: [email protected] Professor Mei-Yin Chou Georgia Institute of Technology School of Physics 837 State Street Atlanta GA 30332-0430 USA

Email: [email protected]

Chapter 13 Professor D. Grant Professor of Materials Science School of Mechanical, Materials and Manufacturing Engineering University of Nottingham University Park Nottingham NG7 2RD UK

Email: [email protected]

Chapter 15 Professor Dr Y. Nakamori* and S. Orimo Institute for Materials Research Tohoku University 2-1-1 Katahira Sendai 980-8577 Japan

Email: [email protected] Email: [email protected] [email protected]

© 2008, Woodhead Publishing Limited

xvi

Contributor contact details

Chapter 16

Chapter 19

Professor Duncan H. Gregory West CHEM Department of Chemistry University of Glasgow University Avenue Glasgow G12 8QQ UK

Dr Tejs Vegge*, R. Z. Sørensen, A. Klerke, J. S. Hummelshøj, T. Johannessen, J. K. Nørskov and C. H. Christensen Materials Research Department Risø National Laboratory for Sustainable Energy Technical University of Denmark DK-4000 Roskilde Denmark

Email: [email protected]

Chapter 18 Professor Masaru Ichikawa Hokkaido University Kita 8 Nishi 5 Kita-ku Sapporo 060-0808 Japan Email: [email protected]

© 2008, Woodhead Publishing Limited

Email: [email protected]

xvii

Preface

Energy is a requirement for any civilisation, whether from wood, fossil fuels, nuclear or renewable sources (such as solar, wind and tidal). The more developed a nation, the higher the energy need per capita as energy consumption moves away from being primarily for heating and cooking and pervades all aspects of life (domestic, work and leisure). For example, the greater wealth of a nation leads to higher rates of consumption and increased demand for transportation. Cheap and abundant reserves of oil and coal have fuelled an extraordinarily rapid rate of technological development in the Western world over the past century. Developing countries such as China and India are set to emulate this transition which, owing to the large populations of both countries, will lead to unprecedented global energy demand. The Intergovernmental Panel on Climate Change (IPCC) has shown that the increasing concentrations of carbon dioxide are caused by human activities, and that increased carbon dioxide levels lead to global warming (with the associated problems of rising sea levels and more frequent and extreme adverse weather events). These are global problems which will affect all countries. A potential solution is to develop a low-carbon future, where fossil fuel use is reduced, replacing it with zero-carbon energy sources such as from renewables. Hydrogen and electricity can both be used as convenient energy carriers for renewable energy sources. In a low-carbon future, a more robust energy network will probably incorporate both. One sector where hydrogen is likely to have a major impact is transport because hydrogen fuel cells have a higher energy density than current battery technologies, which severely limits the range for current electric vehicles. The importance of hydrogen for our low-carbon future is highlighted by the International Energy Agency’s (IEA) Hydrogen Implementing Agreement. IEA activities bring together scientists from around the globe to collaborate on solutions to the current technical barriers hindering our transition to a hydrogen economy. One focus is the compact and lightweight storage of hydrogen. Over the past decade, there have been many significant advances in the storage of hydrogen in porous materials, complex hydrides and liquid hydrides,

© 2008, Woodhead Publishing Limited

xviii

Preface

and in catalysts to accelerate the cycling kinetics for these materials. It is therefore very timely to bring together many of the leading experts in this field and have them report the exciting developments for these systems. There are so many different types of materials being investigated for hydrogen storage applications that it is beyond the scope of this book to include chapters on them all; hence the focus is on the new materials which have been intensively investigated over the past decade. The book also examines some of the techniques to characterise these materials to determine physical and structural changes, investigate the interactions of hydrogen with substrates and the accurate measurement of hydrogen storage capacities. In addition to the science and engineering related specifically to the storage medium, there are also chapters on the effect of hydrogen on structural materials (e.g. the walls of a pressure vessel where hydrogen embrittlement can be a significant problem) and the socio-economic factors that may influence our transition to a low-carbon future. This collection will inevitably be of interest to experienced scientists and engineers in the field as well as postgraduate and undergraduate students keen to explore either energy and/or hydrogen technologies. Global solutions are urgently needed if we are to avoid a global catastrophe. The range of nationalities of the contributing authors indicates the effort from around the world being devoted to hydrogen storage and it is the hope of the editor that this book will be a valuable resource to, and help inspire, young researchers in this exciting and challenging field. Dr Gavin Walker University of Nottingham

© 2008, Woodhead Publishing Limited

Part I Introduction

1 © 2008, Woodhead Publishing Limited

1 Hydrogen storage technologies G. W A L K E R, University of Nottingham, UK

1.1

Introduction

There are significant concerns about the rising level of CO2 emissions and the impact this is having on our environment (IPCC, 2007). The main source of the increase in CO2 is from our increasing energy demands. Global energy needs are expanding rapidly: in 1973 the global demand for energy was 6128 Mtoe (million tonnes of oil equivalent), which almost doubled over the following three decades to 11 435 Mtoe in 2005 (IEA, 2007a). Of this, the percentage contribution from countries in the Organization for Economic Cooperation and Development (OECD) fell from 61.3% to 48.5% while that for Asia and the Middle East more than doubled from 13.8% to 30.8% (IEA, 2007a). If energy policies do not change, energy scenarios predict demand will reach 17 100 Mtoe by 2030, with the developing countries contributing to 74% of that increase (IEA, 2007b). Fossil fuels currently account for 81% of the world’s energy needs and the scenario for 2030 shows little change in this percentage because of the reliance of the developing economies on coal. If we consider the energy consumption per capita we see a different picture emerging. Asia and the Middle East energy consumption is 1.0 toe/capita and that for OECD countries is 4.7 toe/capita. These results help illustrate that the ‘problem’ is not simply the increasing demand from the developing countries, but that the demand per capita for developed nations is almost five times that of the developing nations. This is a global problem, with developed countries needing to reduce their energy consumption per capita while enabling developing economies to expand (and the inevitable increase in energy consumption that that will result in). In addition to limiting energy consumption, low-carbon energy technologies are needed to further reduce CO2 emissions and alternatives to burning fossil fuels are required. Hydrogen has a high calorimetric value, with a lower heating value (LHV, i.e. the heat released upon combustion without recovery of the latent heat of vaporisation of any water produced) of 120 MJ kg–1, compared with petrol, which is approximately a third of this at 43 MJ kg–1. 3 © 2008, Woodhead Publishing Limited

4

Solid-state hydrogen storage

Hydrogen can be used for power generation either by combustion in for example an internal combustion engine (ICE), producing mechanical power, or electrochemically by using a fuel cell, producing electrical power. In both examples the hydrogen reacts with oxygen to form water. For fuel cells this is the only emission, but with ICE, NOx can be formed if the combustion is too hot, hence ICE engines are often run lean to avoid this. Although hydrogen is a very abundant element, it exists on Earth primarily in combination with oxygen, as water. It can, though, be generated in a number of ways, such as electrolysis, reforming, fermentation and direct splitting of water either thermochemically or photocatalytically. If a renewable source of energy is used, the hydrogen will be CO2 neutral. If fossil fuels are used then carbon capture and sequestration will be needed for there to be a significant CO2 reduction. Hydrogen is thus a versatile energy carrier and is seen as an important part of the solution to lowering CO2 emissions. Hydrogen fuel cells have a range of potential applications as indicated in Fig. 1.1, operating from a few watts up to gigawatts. Hydrogen microfuel cells have great potential to power portable electronics such as laptops and mobile phones. The superior energy density of a hydrogen polymer electrolyte membrane (PEM) microfuel cell compared with a lithium ion rechargeable battery opens the way to longer battery lifetimes. Small fuel cells or ICE generation sets (0.1–1 kW size) offer back-up power when disconnected from the national grid (either from outages or when working in the field). For mobile applications, fuel cells could provide auxiliary power for electrical units such as air conditioning and refrigerators (e.g. for food delivery vehicles). Both PEM fuel cells and ICEs are competing technologies for the vehicle Application Phone Laptop

Portable charger

Auxiliary power unit

Small stationary

Automobile powertrain

Medium stationary

Power W

100 W

kW

100 kW

10 g

100 g

kg

Mass of hydrogen 1g Design activities

Scale-up

Miniaturisation

1.1 A scheme indicating the power requirement and corresponding mass of hydrogen for various hydrogen applications.

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

5

powertrain. There is also interest in using hydrogen fuel cells for distributed generation for domestic power and using solid oxide fuel cells or combined cycle gas turbines for small power stations in the MW to GW range. Vehicle powertrain is seen as one of the biggest potential markets for hydrogen, replacing the use of petrol and diesel, but automotive hydrogen storage is also one of the most challenging applications. Transport accounts for 30.3% of global energy use (IEA, 2007a) and is the source of a significant amount of CO2 emissions. Hydrogen, battery or most likely hybrid hydrogen/ battery vehicles are the leading contenders for zero emission vehicles, an attraction for hydrogen vehicles is the combined high energy density and durability of hydrogen fuel cells in comparison with current battery technologies. The challenge with hydrogen is that it is a low-density gas and it is difficult to efficiently store enough hydrogen on-board a vehicle to give the vehicle an adequate range, e.g. c. 5 kg of hydrogen for a range of 500 km (Schlapbach and Züttel, 2001). In the United States, the Department of Energy has set tough targets for the development of hydrogen vehicles (DOE, 2006). The targets for hydrogen storage systems (i.e. this includes the container and any balance-of-plant) are given in Table 1.1. These targets show the importance of both gravimetric and volumetric capacity of the store, as one does not want the store to be too heavy (limiting the range of the vehicle) or too voluminous (limiting the cabin and/or the luggage space). These are the most commonly reported targets. However, in contrast the IEA Hydrogen Implementing Agreement has a 2009 hydrogen storage target of 5 wt%, which is a material target not a system target. Both the DOE and IEA targets assume the powertrain is a PEM fuel cell and therefore the temperature of operation of the store bed is limited to 80 °C (the waste heat generated from a PEM fuel cell). Fuel cells are preferred over ICEs because the former is more efficient (50–60%) (Schlapbach and Züttel, 2001) compared with ICE, for which early engine efficiencies were only c. 20%. Recent work, however, has shown efficiencies comparable with that of a diesel engine are possible, >40% (Verhelst and Sierens, 2007). It is also important that the store can be repeatedly cycled (at

Table 1.1 The US DOE hydrogen storage targets for automobile applications (DOE, 2006) Year

2010

2015

Gravimetric capacity

2 kW h kg system (6 wt% H2)

3 kW h kg–1 system (9 wt% H2)

Volumetric capacity

1.5 kW h l–1 (0.045 kg H2 l–1)

2.7 kW h l–1 (0.081 kg H2 l–1)

Cost

US$4 per kW h

US$2 per kW h

© 2008, Woodhead Publishing Limited

–1

6

Solid-state hydrogen storage

least 500 times) and it is also perceived that the charging time should be no longer than 3–5 minutes. The charging time assumes that consumers’ behaviour will remain the same, but ignores the potential of overnight charging scenarios, e.g. fleet vehicles at a depot or plugging your vehicle in at home to a domestic hydrogen supply (which would be realistic if hydrogen were also used for domestic distributed generation). The conventional hydrogen storage technologies are compressed gas and liquid storage. The following sections will briefly describe each and will then discuss alternative hydrogen storage material technologies. These materials can be split into three categories: physically bound hydrogen, where the hydrogen gas is physisorbed to a high surface area substrate; chemically bound hydrogen, where the hydrogen has formed a chemical compound with the substrate (e.g. metal hydrides and complex hydrides) and the hydrogen is released through a thermal decomposition; and lastly hydrolytic evolution of hydrogen (sometimes referred to as chemical hydrides), a variation on chemically bound hydrogen where the hydrogen is released through a chemical reaction, typically a hydrolysis reaction, e.g. the hydrolysis of metal powders and hydrides.

1.2

High-pressure gas storage

The most common hydrogen storage solution currently uses pressurised cylinders to store the hydrogen. Composite cylinders are used as these are lighter than metal only, and typically have a working pressure of 350 bar although 700 bar tanks now also exist. The construction of the cylinder consists of a liner (which can be made from aluminium, steel or polymer) around which carbon fibres are wound (other fibres can be used like glass, but these result in heavier tanks) and sealed in a polymer resin. The gravimetric storage capacity for a 700 bar tank is 4.5 wt% which is close to the DOE 2010 target; however, the volumetric capacity is only 0.025 kg H2 l–1 (Takeichi et al., 2003) which is equivalent to 0.83 kW h l–1 (cf. 0.58 kW h l–1 reported by DaimlerChrysler for a 350 bar tank; DOE, 2007a). Going to higher pressures will help improve the volumetric capacity, but the energy required for 700 bar compression is 15% of LHV of the hydrogen stored in the vessel and there are practical and safety concerns with refuelling vehicles at 700 bar (Felderhoff et al., 2007), which is presumably why most demonstration vehicles still use 350 bar tanks. Going to even higher pressures is a possibility, but given the current concerns with regards to 700 bar delivery and storage, it is probably unlikely that higher pressures will be used for automotive applications. Conformability of the pressure vessel is also an issue. A cylindrical storage tank is not the most efficient shape to fit into the limited space available in a vehicle. With demonstration buses this is not so much of an issue as the bus can have a rack of pressurised cylinders situated on the roof. However, in a

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

7

car a significant amount of luggage space has to be sacrificed. Alternative shapes are being investigated so that the gas tank utilises available space around the car (much like the design of fuel tanks for petrol/diesel vehicles) but these will be more difficult to manufacture because of the added complexity to the filament winding process and will lead to lower gravimetric hydrogen capacities. Compressed gas storage is currently the best compromise for on-board hydrogen storage, but it will not be able to meet the DOE targets. Incremental improvements can still be made through a more integrated design of the vehicle and the storage vessel, but to store 5 kg of H2 to give an adequate range will require allocating at least 220 l of space to hydrogen storage. For applications where space is less of an issue, for example stationary storage at a fuelling station and distributed generation, compressed gas storage is a relatively cheap solution (especially when lower pressures will suffice, but, for portable applications (much like mobile applications), compressed gas storage is too bulky.

1.3

Liquid hydrogen

Cryogenic storage of hydrogen is another mature technology, but very low temperatures are needed to liquefy hydrogen. The critical temperature for hydrogen is 33 K, above which it is a non-condensable gas. At 1 atmosphere hydrogen has a boiling point of 20 K and the density of the liquid under these conditions is 0.0708 kg l–1, which gives a respectable volumetric capacity of 2.35 kW h l–1. Keeping the contents of the store at 20 K is an imperative to minimise boil-off. The latter is primarily a result of thermal conduction (through cables and fixtures) and thermal radiation from the environment. Although boil-off is kept to a minimum through the design of the cryostore, it cannot be prevented. This is highly undesirable in automotive applications as it is wasteful of the hydrogen and could have serious safety issues for a vehicle left in an enclosed space for a few days (e.g. a garage). Thermal management and minimising the flow of heat from the surroundings to the liquid hydrogen are crucial; therefore the design of tanks aims to minimise the surface area of the liquid and thus minimise heat transfer to the liquid. This impacts on the conformability of the store, since increasing the surface area of the liquid will increase the boil-off rate. Hence stores are usually cylinders, with similar packing inefficiencies as the compressed gas cylinders. Researchers at the Lawrence Livermore National Laboratory have recently reported a pressurised cryostore with a 10.7 kg hydrogen charge (Aceves et al., 2007). They reported that a modified Toyota Prius installed with the tank had a range in excess of 800 km (500 miles) under normal driving conditions and the system capacities were 6.1 wt% and 0.033 kg H2 l–1. The cost of liquefaction is relatively high, requiring c. 30% the LHV of the stored hydrogen.

© 2008, Woodhead Publishing Limited

8

Solid-state hydrogen storage

There are further inefficiencies during refuelling as the delivery tube needs to be cooled and the internal walls of the store will need to be cooled, all leading to further evaporation of hydrogen. BMW has developed dual-fuel ICE vehicles that use liquid hydrogen stores, as this provides the possibility of directly injecting liquid hydrogen into the engine (BMW is one of the few companies developing H2 ICEs as opposed to a fuel cell powertrain). Liquid hydrogen has been used for some fuelling stations (for example, a BP refuelling station at Hornchurch, UK) as the storage is more compact than compressed gas and there already exists an infrastructure for liquid hydrogen delivery by truck. Liquid hydrogen at the fuelling station can be used to deliver high-pressure hydrogen if required. A benefit of this is that there is no extra need for a mechanical compressor as the pressurised gas can be obtained thermally through the evaporation of the hydrogen. However, the very low temperatures and boil-off issues mean it is not a suitable technology for portable applications.

1.4

Physically bound hydrogen

Hydrogen, like any other gas, will physisorb onto a surface; the diatomic molecule does not dissociate and is held to the surface through weak van der Waals interactions. The strength of these interactions for hydrogen is very weak with an enthalpy of adsorption, ∆Ha, of between 4 and 10 kJ mol–1. These weak interactions mean low temperatures are needed to obtain significant amounts of adsorbed gas, so that the hydrogen molecules do not have too much thermal energy which easily overcomes these van der Waals interactions. Hydrogen physisorption is normally measured at 77–80 K because of the convenience of using liquid nitrogen, which has a boiling point of 77 K at 1 atmosphere. Because physisorption is a surface phenomenon, research interest has focused on high surface area materials. A means of maximising the available surface is to increase the porosity. Porous materials which have received considerable attention are high surface area carbons (Yang et al., 2007), carbon nanotubes (CNTs) (Poirier et al., 2004), zeolites (Langmi et al., 2005), metal–organic frameworks (MOFs) (Lin et al., 2006) and, more recently, polymers of intrinsic microporosity (PIMs) (Budd et al., 2007). Table 1.2 compares some typical characteristics for these different types of materials, illustrating the correlation between hydrogen uptake and surface area. Züttel et al. (2004) have proposed that the correlation is 1.5 wt% for every 1000m2 g–1 at a pressure of 1 bar hydrogen. However, more recent examples in the literature show uptakes in excess of this, such as the porous carbon and MOF examples in Table 1.2. It is recognised that surface area alone does not dictate the hydrogen capacity but the pore geometry and volume has an effect (Nijkamp et al., 2001; Lin et al., 2006) as do the framework

© 2008, Woodhead Publishing Limited

Table 1.2 Characteristics and properties of typical examples of porous materials for hydrogen storage Material

Typical framework elements

Surface area

Porosity

(m2 g–1)

(cm3 g–1)

Porous carbons C

3150

1.95

6.9



Yang et al. (2007)

CNTs

C

1160



3.8a



Poirier et al. (2004)

Zeolites

O, Al, Si

670



2.2

0.031

Langmi et al. (2005)

MOFs

H, O, C, TM

2200

0.89

6.1

0.039

Lin et al. (2006)



Budd et al. (2007)

a

H, O, C

Measured at 1 bar hydrogen. Measured at 15 bar hydrogen.

b

1050

0.40

2.7

b

Reference

Hydrogen storage technologies

PIMs

Excess hydrogen capacity at 77 K and 20 bar (wt%) (kg H2 l–1)

9

© 2008, Woodhead Publishing Limited

10

Solid-state hydrogen storage

elements where heteroatoms can provide stronger adsorption sites (Liu et al., 2007). When considering hydrogen stores using these materials it can be seen from Table 1.2 that there are materials that have gravimetric and volumetric capacities which meet or are close to the DOE targets. In general porous materials show excellent cyclability as the material itself does not undergo any significant change through the adsorption/desorption process. A cryotank is needed as with liquid hydrogen and there are similar issues with boil-off and tank conformability. However, there is a significant energy saving in having to keep the store at only 77 K as opposed to 20 K (about a quarter of the energy), but one must not forget that the hydrogen physisorption process is exothermic, creating a thermal management issue during refuelling, which will require extra liquid nitrogen to keep the store at 77 K (Felderhoff et al., 2007). These combined reasons make porous hydrogen storage materials less attractive for automobile on-board storage, but there is interest in their use as a cheaper alternative to liquid hydrogen storage for lorry tankers and small to medium sized stationary stores. The problems associated with having to operate at cryogenic temperatures would be overcome if the hydrogen coordinated more strongly with the substrate, with the ideal enthalpy of interaction being around 30 kJ mol–1 for room temperature storage. This is a much stronger interaction than physisorption, requiring either undissociated hydrogen to coordinate with an active site on the substrate (e.g. a metal centre), or for hydrogen to be dissociated by a catalyst and for the hydrogen atoms to diffuse onto the support, known as the spillover effect. For hydrogen storage materials there is much debate about the physical basis of these phenomena and the experimental validation that these interactions are indeed occurring on hydrogen storage materials. The interested reader is directed toward a recent review by Kubas (2007) on hydrogen–metal complexes and papers by Li and Yang (2007), Chen et al. (2007) and Jain et al. (2007) with regards to hydrogen spillover.

1.5

Chemically bound hydrogen

The classic chemically bound hydrogen storage material is that of metal hydrides. Many metals and alloys will reversibly react with hydrogen to form a hydride. A generic reaction is given in Fig. 1.2. The regeneration of the metal can be accomplished either by increasing the temperature or by reducing the pressure. To understand this behaviour it is helpful to consider the pressure–composition isotherm (PCI) for a metal hydride. An example is given in Fig. 1.3, which shows that as the pressure increases the hydrogen uptake increases. The PCI plot also shows that there is a plateau above which pressure the metal will hydride and in a closed

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

11

system will continue to hydride until the pressure of the system decreases down to that of the plateau pressure. The stored hydrogen can be released by reducing the pressure of the system to a level below that of the plateau pressure. The plateau pressure is also temperature dependent, and increases with temperature. Thus a hydride stable under a certain temperature and pressure will decompose when the temperature is increased to a level where the plateau pressure is now higher than the system pressure. The temperature needed for a 1 bar plateau pressure, T(1 bar), is a useful characteristic of metal hydrides as this gives an indication of the minimum working temperature for a store based on that material. Unfortunately, the metals and alloys with a T(1 bar) value less then 80 °C are based on heavy transition metals (e.g. LaNi5 and FeTi) and therefore have low gravimetric hydrogen storage capacities Endothermic

2MHx

2M + x H2

Exothermic

H2 pressure

1.2 A generic reaction equation for the reversible hydrogenation of a metal, illustrating the exothermic and endothermic processes.

A

H2 content (wt.%)

1.3 An illustration of an idealised PCI graph, showing the effect of pressure on the amount of hydrogen stored in a metal hydride at a constant temperature. Point A denotes the plateau pressure, which will be higher at higher temperatures.

© 2008, Woodhead Publishing Limited

12

Solid-state hydrogen storage 0.16

Mg(BH4)2

Volumetric capacity (kg H2 l–1)

0.14 0.12

LaNi5H6

0.10

TiFeH1.7 Mg2NiH4

MgH2

MgCl2.6NH3

NaAIH4

LiNH2

Ca(BH4)2 NaBH4

LiAIH4

Petrol

LiH

0.08

MOF

0.04 Alanate tank 0.02 0

ρ H2(I)@20 K

C6H12

0.06

0

ρ H2(g)@700 bar

H2(I) tank 700 bar tank

5

LiBH4

10 Gravimetric capacity (wt%)

15

20

1.4 Hydrogen storage capacities for a range of storage media, including system capacities for stores based on liquid hydrogen, compressed hydrogen and sodium alanate. Values given for C6H12 are the reversible capacities when forming C6H6. For reference the density of liquid hydrogen and hydrogen gas at 700 bar are given on the right of the graph. (NB actual gravimetric capacity for both is 100 wt%.)

as shown in Fig. 1.4. It can be seen though that these materials have volumetric capacities far superior to that of liquid hydrogen. Because of the low gravimetric capacity for these interstitial metal hydrides, there has been a lot of interest in higher capacity materials such as magnesium hydride (with a storage capacity of 7.6 wt%) even though this has a T(1 bar) of 280 °C. Interest grew in the use of complex hydrides when Bogdanovich and Schwickardi (1997) showed that titanium-based catalysts enable sodium alanate to be reversibly dehydrogenated (having a theoretical capacity of 7.5 wt%). In addition to sodium alanate there has been much interest in lithium amide and tetrahydridoborates of lithium, magnesium and a number of transition elements. A comparison of the data in Fig. 1.4 with the DOE targets for 2010 and 2015 shows that the number of potential materials able to meet these targets is very limited. However, of the materials that do meet these targets, the dehydrogenation enthalpy is high, resulting in a high T(1 bar) because thermodynamics dictates that T(1 bar) = ∆H/∆S where ∆H and ∆S are the change in enthalpy and entropy for the dehydrogenation. Destabilisation is a strategy by which the dehydrogenation enthalpy can be lowered. Examples include multicomponent systems where a destabilising agent is added to a hydride or complex hydride. An example is LiBH4 plus MgH2 (Vajo et al., 2005) which, when combined, undergo the following reversible reaction:

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

13

2LiBH4 + MgH2 s 2LiH + MgB2 + 4H2 which has an associated T(1 bar) value of 165 °C in comparison with 280 °C and 410 °C for the binary and complex hydrides respectively. Hydrides of non-metals can also be used such as those of carbon (organic hydrocarbons) and nitrogen (ammonia). Examples of the former most commonly utilise the dehydrogenation of a saturated hydrocarbon forming an aromatic, e.g. cyclohexane → benzene. These reversible reactions are well-known heterogeneous catalytic reactions in the petroleum industry utilising PGM catalysts. The example of cyclohexane gives a theoretical reversible hydrogen capacity of 7.2 wt%, but requires a reactor temperature of c. 300 °C for the dehydrogenation reaction (Kariya et al., 2003). Lower temperature organic hydrides have been investigated by Air Products, which tend to be based on nitrogen doped rings such as N-ethylcarbazole (Cooper et al., 2006). This example dehydrogenates at 130 °C and has a theoretical capacity of 5.7 wt% and 0.054 kg H2 l–1. There is also interest in ammonia as this can be catalytically decomposed at temperatures over 400 °C using nickel or ruthenium catalysts (Li et al., 2005). Although liquid ammonia can be readily formed, ammonia is very corrosive and so a preferred method of storage is as an inorganic adduct such as MgCl2·6NH3 so that the corrosive ammonia is bound within an inorganic salt, making it much easier to handle. Magnesium chloride theoretically stores 9.2 wt% of hydrogen equivalent, but temperatures up to 350 °C are needed to release all the ammonia, although c. 6 wt% hydrogen equivalent is released by 220 °C. Thermal management for a hydride-based store is particularly challenging for automotive applications where fast, 3–5 minute, loading times are desired. For a store with a ∆H of dehydrogenation of 40 kJ mol–1 H2, this equates to c. 0.6 MW of heat dissipation (Felderhoff et al., 2007). The cooling system needed for this would be too large to be accommodated on-board and would have to be provided at the fuelling station. For this reason there has been a change in direction for the US hydrogen research programme, and instead of on-board hydrogenation of the store they are envisaging off-board recycling. However, this has logistical issues of requiring common exchangeable fuel tanks between manufacturers and an infrastructure for collection, recycling and distribution of the stores. Where slower recharging times could be accommodated, the thermal management would not be as much of an issue. Other potential applications for chemically bound hydrogen storage materials include portable electronics and in-field back-up power where these materials offer potentially far superior gravimetric and volumetric store capacities compared with any of the other technologies. Metal hydrides and complex hydrides are less likely to be favoured for stationary applications if there is no weight or space constraint as these solutions will be more expensive than

© 2008, Woodhead Publishing Limited

14

Solid-state hydrogen storage

pressurised cylinders, although the fact these stores can operate below 10 bar could be an attraction, avoiding any perceived concerns of safety with highpressure storage. Organic hydrides and ammonia both require a catalytic dehydrogenation reformer unit to extract the hydrogen. This adds further complexity to the system, in the form of an on-board reformer, which makes it less attractive for automotive applications. Both storage materials are potentially more attractive for large-scale stationary stores and the catalytic dehydrogenation process is less demanding in terms of matching greatly varying loads as would be needed for automotive systems.

1.6

Hydrolytic evolution of hydrogen

It is well known that adding a metal that has a more negative reduction potential than hydrogen to water will react with the water forming either the oxide or the hydroxide of the metal and liberating hydrogen via one of the following reactions: M + xH2O → MOx + xH2 M + 2xH2O → M(OH)2x + xH2 Group 1 metals follow the latter reaction and form soluble hydroxides, which means that fresh metal is continuously exposed during the reaction, leading to complete reaction. However, most other metals do not react as efficiently because the oxides and hydroxides for these elements have very low solubilities and particles of these metals often quickly build up passivating layers, which hinder further reaction. A claimed breakthrough to overcome this problem is the use of metals such as gallium alloyed with aluminium, which helps to minimise the effect of the passivating layer. It is claimed that the aluminium alloy completely reacts with water, yielding hydrogen and alumina (Cuomo and Woodall, 1982). The hydrolysis of this and other aluminium alloys have been investigated by Kravchenko et al. (2005). There are some drawbacks with the logistics for such systems, controlling the hydrogen evolution requires careful control of the quantities of the reactants being added together. The end-products are thermodynamically very stable and the reaction cannot be easily reversed to re-form the metals. This approach will require centralised regeneration facilities and is likely to be realistic only where there are cheap and abundant sources of electricity. Also reloading the vehicle will require transferring highly pyrophoric metal powders into the fuel tank of the vehicle which is no trivial task. An alternative to using metals is to use sodium tetrahydridoborate, NaBH4. A significant advantage over the use of metal powders is that sodium tetrahydridoborate can form stable aqueous solutions. The hydrolysis reaction

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

15

is controlled catalytically and thereby gives an easier method for controlling the reaction. The gravimetric hydrogen storage capacity, though, is not that high since enough water needs to be used to dissolve the end-products. This drops the overall system capacity down to 3.3 wt% and 0.026 kg H2 l–1 (DOE, 2007b) (cf. the theoretical capacity for solid NaBH 4: 10.7 wt% and 0.115 kg H2 l–1). The Millennium Cell project in the Unites States is based on this technology, but there are still the same concerns with the costs associated with recycling the end-products as with the metal powder approach and in light of the poor capacities the DOE hydrogen program announced a no-go recommendation for this project (DOE, 2007b).

1.7

Summary

This chapter has given an introduction to the potential applications for hydrogen storage and an overview of the different technologies available for storing hydrogen. The various applications have different requirements in terms of the mass of hydrogen needed, volumetric capacity, gravimetric capacity and, of course, cost. Compressed gas and liquid hydrogen storage technologies are the current state-of-the-art, but more compact means of storing hydrogen are needed for portable and mobile applications, solid-state hydrogen storage materials would appear to be the most promising solution. The following chapters in Part I help set the scene with chapters discussing the socioeconomic issues concerning the transition to a hydrogen economy. This is followed by chapters considering the suitability of materials used for the pressurised containment of hydrogen and design issues for solid-statebased hydrogen stores. Our understanding of hydrogen storage materials is underpinned by the accurate measurement of hydrogen capacities and the detailed characterisation of the chemical and structural changes during cycling, which is the subject of Part II. This is followed by sections discussing the recent advances in porous materials, complex hydrides, metal hydrides and liquid hydrides for hydrogen storage and there is a chapter describing the exciting developments in the solid-state storage of ammonia as an alternative hydrogen carrier. Finally, the book is concluded with a chapter highlighting some of the challenges ahead and current direction in hydrogen storage materials.

1.8

References

Aceves S M, Berry G, Espinosa-Loza F, Ross T and Weisberg A (2007), Storage of hydrogen in cryo-compressed vessels, DOE Hydrogen Program 2007 Annual Report, Section VI.A.6 (www.hydrogen.energy.gov/annual_progress07.html). Bogdanovic B and Schwickardi M (1997), Ti-doped alkali metal aluminium hydrides as potential novel reversible hydrogen storage materials, J. Alloys Compounds, 253, 1– 9.

© 2008, Woodhead Publishing Limited

16

Solid-state hydrogen storage

Budd P, Butler A, Selbie J, Mahmood K, McKeown N B, Ghanem B, Msayib K, Book D and Walton A (2007), The potential of organic polymer-based hydrogen storage materials, Phys. Chem. Chem. Phys., 9, 1802–1808. Chen L, Cooper A C, Pez G P and Cheng H (2007), Mechanistic study on hydrogen spillover onto graphitic carbon materials, J. Phys. Chem. C, 111, 18995–19000. Cooper A C, Campbell K M and Pez G P (2006), An integrated hydrogen storage and delivery approach using organic liquid-phase carriers, Proc. World Hydrogen Economy Conference, June 2006, Lyon, France. Cuomo J J and Woodall J M (1982), Solid state renewable energy supply, US Patent 4358291. DOE (2006), Hydrogen Posture Plan, An Integrated Research, Development and Demonstration Plan (www.hydrogen.energy.gov/pdfs/hydrogen_posture_ plan_dec06.pdf). DOE (2007a), DOE Hydrogen Program 2007 Annual Report (www.hydrogen.energy.gov/ annual_progress07.html). DOE (2007b), Go/No-go Recommendation for Sodium Borohydride for On-board Vehicular Hydrogen Storage, article NREL/MP-150–42220 (www.nrel.gov/docs/gen/fy08/ 42220.pdf). Felderhoff M, Weidenthaler C, von Helmolt R and Eberle U (2007), Hydrogen storage: the remaining scientific and technical challenges, Phys. Chem. Chem. Phys., 9, 2643– 2653. IEA (2007a), Key World Energy Statistics 2007 (www.iea.org/Textbase/publications/ free_new_Desc.asp?PUBS_ID=1199). IEA (2007b), World Energy Outlook 2007 – China and India Insights, IEA publications, Paris, France. IPCC (2007), Intergovernmental Panel on Climate Change Fourth Assessment Report – Climate Change 2007: Synthesis Report (www.ipcc.ch/ipccreports/ar4-syr.htm). Jain P, Fonseca D A, Schaible E and Lueking A D (2007), Hydrogen uptake of platinumdoped graphite nanofibers and stochastic analysis of hydrogen spillover, J. Phys. Chem. C, 111, 1788–1790. Kariya N, Fukuoka A, Utagawa T, Sakuramoto M, Goto Y and Ichikawa M (2003), Efficient hydrogen production using cyclohexane and decalin by pulse-spray mode reactor with Pt catalysts, Appl. Catal. A – General, 247, 247–259. Kravchenko O V, Semenenko K N, Bulychev B M and Kalmykov K B (2005), J. Alloys Compounds, 397, 58–62. Kubas G J (2007), Dihydrogen complexes as prototypes for the coordination chemistry of saturated molecules, Proc. Nat. Acad. Sci. USA, 104, 6901–6907. Langmi H W, Book D, Walton A, Johnson S R, Al-Mamouri M M, Speight J D, Edwards P P, Harris I R and Anderson P A (2005), Hydrogen storage in ion-exchanged zeolites, J. Alloys Compounds, 404, 637–642. Li X-K, Ji W-J, Zhao J, Wang S-J and Au C-T (2005), Ammonia decomposition over Ru and Ni catalysts supported on fumed SiO2, MCM-41, and SBA-15, J. Catal., 236, 181–189. Li Y and Yang R T (2007), Gas adsorption and storage in metal-organic framework MOF177, Langmuir, 23, 12937–12944. Lin X, Jia J, Zhao X, Thomas K M, Blake A J, Walker G S, Champness N R, Hubberstey P and Schröder M (2006), High H2 adsorption by co-ordination framework materials, Angewandte Chem. International Edition, 45, 7358–7364.

© 2008, Woodhead Publishing Limited

Hydrogen storage technologies

17

Liu Y, Brown C M, Neumann D A, Peterson V K and Kepert C J (2007), Inelastic neutron scattering of H2 adsorbed in HKUST-1, J. Alloys Compounds, 446, 385–388. Nijkamp M G, Raaymakers J E M J, van Dillen A J and de Jong K P (2001), Hydrogen storage using physisorption – materials demands, Appl. Phys. A, 72, 619–623. Poirier E, Chahine R, Bénard P, Cossement D, Lafi L, Mélançon E, Bose T K and Désilets S (2004), Storage of hydrogen on single-walled carbon nanotubes and other carbon structures, Appl. Phys. A, 78, 961–967. Schlapbach L and Züttel A (2001), Hydrogen-storage materials for mobile applications, Nature, 414, 353–358. Takeichi N, Senoh H and Yokota T et al. (2003), Hybrid hydrogen storage vessel, a novel high pressure hydrogen storage vessel combined with hydrogen storage material, Int. J. Hydrogen Energy, 28, 1121–1129. Vajo J J, Skeith S L and Mertens F (2005), Reversible storage of hydrogen in destabilized LiBH4, J. Phys. Chem. B, Lett., 109, 3719–3722. Verhelst S and Sierens R (2007), A quasi-dimensional model for the power cycle of a hydrogen-fuelled ICE, Int. J. Hydrogen Energy, 32, 3545–3554. Yang Z, Xia Y and Mokaya R (2007), Enhanced hydrogen storage capacity of high surface area zeolite-like carbon materials, J. Am. Chem. Soc., 129, 1673–1679. Züttel A, Sudan P, Mauron P and Wenger P (2004), Model for the hydrogen adsorption on carbon nanostructures, Appl. Phys. A, 78, 941–946.

© 2008, Woodhead Publishing Limited

2 Hydrogen futures: emerging technologies for hydrogen storage and transport P. E K I N S, King’s College London, and P. B E L L A B Y, University of Salford, UK

2.1

Introduction

Common to all discussions and projections of ‘a hydrogen economy’ is the widespread use of hydrogen as an energy carrier. Beyond that the projections differ greatly. The use of hydrogen as an energy carrier requires the development and use of a range of different technologies, some similar to and some quite different from those in common use in current energy systems. Section 2.2 very briefly introduces and describes these technologies. Section 2.3 then discusses some of the hydrogen futures that have been proposed, which envisage the use of different sets of hydrogen technologies. Most require that the technologies are considerably developed before the uses become competitive with alternative means of delivering the same energy service. For these technologies to become widely deployed in these or any other ‘hydrogen future’ their envisaged performance and end-use applications will need to become both technologically feasible and publicly acceptable, in both economic and social terms. The technical performance of different storage options, where storage is one of the areas in need of most improvement, is considered in detail elsewhere in this book. Sections 2.4 and 2.5 explore the economics of the different technologies for hydrogen production, distribution, storage and the end-use applications. Section 2.6 presents the results of new research on the public acceptability of hydrogen, and Section 2.7 discusses the policies that are likely to be necessary if hydrogen is to become widely used.

2.2

Hydrogen technologies

Padró and Putsche (1999) list and survey the economics of all the key technologies which might be included in a hydrogen economy. However, as the survey by McDowall and Eames (2006) made clear, there is not a single hydrogen economy, but a range of views as to precisely what technologies, in different combinations, a hydrogen economy might include. 18 © 2008, Woodhead Publishing Limited

Hydrogen futures

19

Hydrogen (H2) is often cited as the most common element in nature, but such citations do not always say that it is also reactive and therefore in nature does not exist in elemental form, but needs to be produced from compounds that contain it. There are a range of means of hydrogen production. Currently the most common method is the ‘steam methane reforming’ (SMR) of natural gas. All production technologies have in common that they require energy to produce hydrogen. A key question for any production technology is whether the energy used in producing hydrogen might not be better used to satisfy the demand for energy services itself, and what the energy cost of producing hydrogen actually is for different technologies. Once produced, hydrogen may need to be either stored or distributed or both. It may be stored or distributed as a gas or a liquid, or in the molecular structure of a variety of solid media. Means of distribution include pipelines (where it is a gas) and truck, rail and ship transport for hydrogen in all its forms. Finally, hydrogen may be put to a range of final uses to satisfy the demand for energy services. Some of these involve hydrogen fuel cells (but not all fuel cells use hydrogen as their fuel), devices which convert hydrogen to electric power with high efficiency; some of them involve hydrogen being burned in internal combustion engines or to power turbines for electricity generation. A major category of end-use is in vehicles, and one of the most active fields of research and development relates to fuel cell vehicles (FCVs). While in the past the development of vehicles capable of ‘on-board’ reforming of hydrogen from conventional fuels was seen as a possible way of avoiding the problems of developing hydrogen infrastructure for vehicles, this approach has in the past few years been largely abandoned by manufacturers due to significant technical challenges (Hughes, 2006). Thus, hydrogen will need to be stored on-board vehicles (either as a compressed gas, a liquid or in a solid-state form) raising a number of specific technological problems relating to the performance of the vehicle concerned.

2.3

Hydrogen scenarios: from production to applications

Table 2.1 sets out the ‘hydrogen futures’ developed by McDowall and Eames on the basis of their review of the literature (McDowall and Eames, 2006) and an expert workshop (for further details of which see McDowall and Eames, 2004). Eames and McDowall subsequently compressed their six futures into four ‘transition scenarios’, which describe both an end-point vision and a description of how such an end-point might feasibly be reached from the present day, drawing on insights from the transition theory literature (see Eames and McDowall, 2006). Most elements of the original six ‘visions’ or futures have been integrated

© 2008, Woodhead Publishing Limited

20

Solid-state hydrogen storage

Table 2.1 Potential UK hydrogen futures (Source: Eames and McDowall 2005, p. 1) UK hydrogen futures

Transport only Central pipeline

Brief description Hydrogen has become the dominant transport fuel, and is produced centrally from a mixture of clean coal and fossil fuels (with C-sequestration), nuclear power and large-scale renewables. Hydrogen is distributed as a gas by dedicated pipeline.

Forecourt reforming

Hydrogen produced locally from natural gas is the dominant road transport fuel. The existing natural gas network provides the delivery infrastructure, and hydrogen is generated on-site by steam methane reforming at the refuelling station.

Liquid hydrogen

Liquid hydrogen produced by nuclear power and large-scale renewable installations has become the dominant transport fuel. There is an international market in liquid hydrogen. This is largely a scenario of substitution, with current energy and transport paradigms remaining unchanged.

Synthetic liquid fuels

Renewably produced hydrogen again provides the dominant transport fuel. In this case, however, it is ‘packaged’ in the form of a synthetic liquid hydrocarbon, such as methanol, to overcome the difficulties of hydrogen storage and distribution. The carbon for fuel synthesis comes from biomass and from the flue gases of carbon-intensive industries.

Transport and other energy services Ubiquitous hydrogen Renewably produced hydrogen is a major energy carrier for heat and power as well as the dominant transport fuel. A hydrogen pipeline grid serves most buildings. Many homes and businesses use fuel cell CHPa systems running on hydrogen, and it is common to refuel your vehicle at home. Hydrogen is produced from a mix of larger centralised and smaller-scale distributed renewables and biomass. Electricity store

a

Combined heat and power.

© 2008, Woodhead Publishing Limited

Hydrogen, produced through on-site electrolysis, is the dominant road transport fuel, and also plays a vital role overcoming the intermittency problems of a renewables-based electricity system. Hydrogen production is flexible, and can respond to variable electricity supply conditions, easing load-balancing. Since hydrogen is produced on site it requires no distribution infrastructure. Locally stored hydrogen provides back-up power for domestic and commercial CHP units at times of peak electricity demand/limited supply.

Hydrogen futures

21

into the four transition scenarios. Synthetic liquid fuels and Electricity store are much as described in Table 2.1. So is Ubiquitous hydrogen, but the Forecourt Reforming future is folded within it as part of the transition to the end-point. The Central Pipeline and Liquid Hydrogen futures are amalgamated into a Central hydrogen for transport scenario. Table 2.2 lists the technologies, many of which are also discussed in Padró and Putsche (1999), that would be required for the four ‘transition scenarios’. From these tables some broad conclusions can be drawn: •

• •



For hydrogen to be ‘low-carbon’ it must be produced from renewable energy sources (electrolysis, biomass gasification/fermentation), nuclear power (electrolysis, thermal, thermochemical) or fossil fuels (SMR, gasification, electrolysis) with carbon capture and storage (CCS). Where local (‘on-site’) hydrogen production is envisaged, CCS is infeasible. Local SMR is therefore, at best, a transitional technology if hydrogen is intended to contribute to a ‘low-carbon’ economy. Most envisaged transport applications depend on fuel cell technology, although the fuel in this case could be methanol rather than hydrogen. All transport applications require considerable technical advances with on-board storage before they are likely to become competitive with fossil fuel vehicles (which may be hybrid fossil fuel/electric), at least in mainstream applications. The different futures have very different implications for infrastructure, some requiring highly developed hydrogen distribution networks (pipelines, refueling stations), while others can use existing (gas, electricity, road) networks.

This listing of technologies allows the identification of the kinds of technological development that are likely to be required for the scenarios to become realistic. •





Transmission and distribution (T&D) infrastructure. Pipelines and metering are necessary for both Central hydrogen for transport and Ubiquitous hydrogen. Liquid hydrogen technologies could also be applied to both. Neither of the other scenarios require hydrogen T&D infrastructure. Storage. Only Synthetic liquid fuels does not require on-board hydrogen storage (this was the reason this scenario was developed). The other three scenarios also require stationary storage, Electricity store on a small scale, the other two on a large scale. Applications. For transport applications, as already noted, only Synthetic liquid fuels does not require onboard storage. All the scenarios require fuel cells, DMFCs for Synthetic liquid fuels, and (probably) PEMFCs for the other three scenarios. For stationary power, Electricity store and Ubiquitous hydrogen require fuel cells for combined heat and power (CHP), which could be PEMFCs for residential applications with low

© 2008, Woodhead Publishing Limited

Transport only Central hydrogen for transport

Synthetic liquid fuels

Hydrogen production

Hydrogen transport

Hydrogen storage

Large-scale electrolysis (including nuclear power, renewables, fossil fuels with CCS) Thermal or thermochemical production in hightemperature reactors (HTRs) Gasification (coal, waste, biomass, with CCS) SMR (with CCS) Pyrolysis Synthetic liquid fuel synthesis, with CCS Renewables

Pipelines and metering Liquefaction

Stationary bulk storage Chemical or solidstate on-board storage Handling cryogenic technologies Liquid H2 storage

Transport, other energy services Ubiquitous Large and small scale (central hydrogen and local) production from a variety of sources, including small-scale SMR, gas separation/gasification (coal, waste, biomass, with CCS) Renewables Pyrolysis Electricity store Renewables for electricity Small-scale electrolysis

a

Transportation applications On-board storage (including liquid H2) Hydrogen fuel cell (likely to be proton exchange membrane) vehicles (HFCVs)

Direct methanol fuel cells (DMFCs)

Pipelines and metering Liquefaction

Stationary bulk storage

Hydrogen CHP (probably PEMFCs or SOFCs)a ‘Smart’ networks (electricity grid and metering)

On-board storage HFCVs

Small-scale stationary storage and handling

Hydrogen CHP (probably PEMFCs or SOFC) ‘Smart’ networks (electricity grid and metering)

On-board storage HFCVs

PEMFC: proton exchange membrane fuel cell; SOFC: solid oxide fuel cell

© 2008, Woodhead Publishing Limited

Stationary power

Solid-state hydrogen storage

UK hydrogen future

22

Table 2.2 Technologies required for different hydrogen futures

Hydrogen futures

23

heat demand, or high-temperature SOFCs for industrial applications. These two scenarios also require ‘smart’ networks for electricity and metering. This technological characterisation enables the scenarios to be placed in a rough hierarchy in terms of the technologies that are required to make them operational. •







Synthetic liquid fuels is stand-alone and only requires the large-scale centralised production of hydrogen from any of the available technologies, plus the development and application of methanol technologies, including DMFCs. Large-scale centralised production technologies, plus the development of bulk storage technology and pipeline infrastructure (perhaps with local SMR as a transitional technology), and on-board storage and PEMFCs, would permit the Central hydrogen for transport scenario. Abundant local renewables plus small-scale electrolysis, plus on-board storage and PEMFCs, plus small-scale storage and SOFCs, would make the Electricity store scenario feasible. The Ubiquitous hydrogen scenario requires elements from all the other scenarios, and may perhaps best be described as Central hydrogen for transport plus the application technologies (small-scale storage and SOFCs) of Electricity store.

Such a description enables an exploration of the economic implications of these scenarios, based on an investigation of the economics of their necessary component technologies, prior to a discussion of wider social aspects that will need to be addressed for hydrogen technologies to achieve public acceptability.

2.4

Hydrogen economics

2.4.1

Hydrogen production

Only for Electricity store is small-scale electrolysis a necessary production technology, to allow local renewables to produce hydrogen with their electricity, rather than feed it into the grid. All the other scenarios depend on large-scale centralised production of hydrogen, from electrolysis, high-temperature reactors, SMR or gasification (with CCS where necessary). Synthetic liquid fuels then requires the hydrogen to be converted to methanol, but this technology is already well developed. A transitional technology for Ubiquitous hydrogen is small-scale SMR (for example, on filling station forecourts), while pipeline infrastructure is being developed. Hawkins and Joffe (2005) have reviewed the literature on the costs of four major hydrogen production technologies, SMR (large and small scale),

© 2008, Woodhead Publishing Limited

24

Solid-state hydrogen storage

gasification, pyrolysis and electrolysis (large and small scale). Throughout their review they emphasise the uncertain and contingent nature of their results – the studies reviewed make a wide range of different assumptions about current technologies and how they might develop – but the range of costs they thereby derive is nonetheless instructive and give insights into the relative economics of four transition scenarios discussed in the previous section. Table 2.3 reproduces their range of costs with relevant comments.

2.4.2

Hydrogen storage

Hydrogen can be stored as a compressed gas, as a liquid, in a chemical compound (e.g. chemical hydrides or metal hydrides), or physically held within nanoporous structures. A major element of the cost of most of these storage modes (and a major consideration in terms of their energy efficiency) is the energy required to get the hydrogen in and out of storage. Table 2.4 shows the cost of a number of means of storage, including liquefaction, gas compression above ground and underground, and chemical and metal hydrides. Table 2.3 Costs for various hydrogen production technologies Technology

Transition scenarioa

Cost range, US$(2000)/ kg H2b

Comments

SMR, large scale (>100 MW)

SLF, CHT, UH

5.25–7.26

Cost highly dependent on natural gas price

Coal gasification (min. 376 MW)

SLF, CHT, UH

5.4–6.8

Coal price more stable and predictable than natural gas

Biomass gasification (>10 MW)

SLF, CHT, UH

7.54–32.61 (av. 14.31)

Size ranges from 25 to 303 MW and affects cost

Biomass pyrolysis (>10 MW)

SLF, CHT, UH

6.19–14.98

Size ranges from 36 to 150 MW; cost reduced by sale of co-products

Electrolysis, large scale (>1 MW)

SLF, CHT, UH

11–75 (20–60 is preferred range)

Size ranges from 2 to 376 MW, but little effect on cost; cost very dependent on assumed price of electricity

Electrolysis, small scale (45 kg/h)

CHT, UH

1–1.5

Cost highly dependent on scale, efficiency, cost of electricity

Compressed gas (2.5 cm) used for local distribution of hydrogen gas in large-scale industrial operations also make extensive use of low-alloy steels, such as the chromium–molybdenum steels, particularly for service at elevated temperature. Pipelines and piping fabricated from low-alloy and carbon steels have functioned safely and reliably in hydrogen gas service. Steel pipelines can require longitudinal (seam) welds to manufacture sections of pipeline as well as girth welds to construct the pipeline system. Consequently, the properties of welds are carefully controlled to preclude hydrogen embrittlement. Transmission pipeline systems have been fabricated from relatively low-strength carbon steels, such as ASTM A53 Grade B, ASTM A106 Grade B, ASTM A333 Grade 6, as well as the API 5L steels (e.g. Grades A, B, X42, and X52) [13, 31, 49]. The maximum tensile strength (Su)

© 2008, Woodhead Publishing Limited

66

Solid-state hydrogen storage

recommended for hydrogen gas pipeline steel is 800 MPa [31]. The maximum pressure in hydrogen gas transmission pipelines is 21 MPa [13], and wall stresses in these pipelines are designed to be less than 30% of the specified minimum yield strength of the steel [31]. The diameter of steel piping and pipelines is commonly in the range 5–30 cm. Hydrogen gas transmission pipelines are operated at near constant pressure, therefore cyclic loading and associated fatigue crack propagation have not been a concern [31].

3.5.3

Small-diameter piping

Austenitic stainless steels are routinely used in tubing for localized hydrogen gas distribution, such as pressure manifolds. The most common stainless steels for tubing are type 304 and 316 alloys [13]. Tubing is most commonly seamless and used in the annealed (low-strength) condition, although there is one report of welded type 304 stainless steel tubing for hydrogen gas service [13]. Because tubing for hydrogen gas service is relatively small diameter (e.g. < 2 cm) and austenitic stainless steels are considered less susceptible to hydrogen embrittlement than other families of steels, relatively high operating pressures are reported: up to 42 MPa for type 304 [13] and greater than 150 MPa for type 316 [27]. Low-pressure piping systems use a variety of materials in addition to the austenitic stainless steels. Copper and various brass alloys have been used for hydrogen gas distribution at ambient temperature and operating pressures up to 18 MPa [13].

3.6

Materials used in the design of hydrogen containment structures

In the following section, we describe in more detail the basic hydrogen compatibility trends associated with the major classes of structural materials. There are many sources of comparative information on materials compatibility with hydrogen gas available in the literature [2–14], as well as general guidance for high-pressure hydrogen systems from the industrial gas companies [31, 45] and NASA [50]. However, for most materials the knowledge base is insufficient for designing low-cost, structurally efficient high-pressure components for hydrogen gas systems. Existing laboratory data should be reviewed critically as not all test techniques are appropriate for establishing long-term effects of hydrogen embrittlement [51].

3.6.1

Low-alloy and carbon steels

Low-alloy and carbon steels represent the bulk of structural metals, and we refer to them collectively here as ferritic steels to distinguish them from the

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

67

austenitic grades. They have relatively low alloying content, particularly the carbon steels, which accounts for their low cost in comparison with highly alloyed steels such as the stainless steels. The microstructure of these alloys, however, is quite varied and not limited to predominantly ferrite. The ferritic steels are easily formed and welded, and there are numerous ways to control the microstructures and properties by heat treatment and mechanical deformation. It is the combination of manufacturing flexibility, availability, and low cost that makes ferritic steels the most widely used steels for pressure systems. One limitation of ferritic steels is that the fracture toughness can decrease precipitously at low temperature, a phenomenon known as the ‘ductileto-brittle transition’ in steels. Carbon steels are generally the least expensive and are used for large volume applications such as pipelines. Carbon steels generally contain manganese in concentrations up to 2 wt% and carbon with concentrations that vary from 0.02 to 0.5 wt%. Silicon is present as a consequence of the steel-making process in contents of around 0.5 wt%. Sulfur and phosphorus are common undesirable elements that exist in all steels in small concentrations and can have a deleterious effect on properties if not controlled. In the last few decades, microalloying additions have been commonly employed in carbon steels to improve mechanical properties and manufacturing processes. The total content of microalloying elements is generally less than 0.5 wt% and commonly consists of niobium, vanadium and titanium. Carbon steels for hydrogen service are processed to produce uniform, fine-grained microstructures [31]. A normalizing heat treatment, consisting of heating steel in the austenite phase field followed by air cooling [52], can yield the desired microstructure in conventional steels. More sophisticated thermomechanical processes are used for microalloyed steels, consisting of hot rolling in the austenite-ferrite phase field to produce the desired finegrained microstructure in the finished product [52]. The low-alloy steels are distinguished by additions of chromium, molybdenum, and, in some cases, nickel. These elements are added in concentrations of a few percent, but the total alloy content is generally less than 5 wt%. Modern steel pressure vessels are commonly fabricated from low-alloy steels with a ‘quenched and tempered’ microstructure. The heat treatment sequence to produce this microstructure consists of heating in the austenite phase field, rapidly cooling (quenching) to form martensite, then tempering at an intermediate temperature [52]; this process can be used for a given steel to produce a wide range in strength properties. For hydrogen gas vessels, the heat treatment parameters are selected to produce a uniform tempered martensite microstructure with a tensile strength (Su) less than 950 MPa (section 3.5.1) [45]. The low-alloy steels have improved strengthening characteristics compared with the carbon steels at an equivalent carbon content, and their microstructure is more stable at elevated temperature.

© 2008, Woodhead Publishing Limited

68

Solid-state hydrogen storage

The hydrogen embrittlement susceptibility of ferritic steels is a particularly complex function of the cross-section of microstructural, mechanical, and environmental variables because microstructure and strength can vary over such wide ranges. Microstructures in low-alloy and carbon steels that are resistant to hydrogen embrittlement have not been fully elucidated, although as mentioned previously martensite is generally viewed as highly susceptible to hydrogen embrittlement, particularly if non-tempered [7, 8]. As described in Section 3.4.1 (Fig. 3.2), the generally accepted trend is that hydrogen embrittlement susceptibility increases with the strength of the material. Consequently, quenched and tempered, low-alloy steels for pressure vessels are restricted by codes and standards to relatively low-strength conditions to ensure resistance to hydrogen embrittlement compared with the same alloy with higher strength. In fact, high-strength steels can be severely embrittled when exposed to hydrogen partial pressures of less than 1 atm. Strength trends, however, must be qualified as alloy composition and microstructure can dominate hydrogen embrittlement in some cases [8]. For example, highcarbon steel with a low-strength microstructure can be more susceptible to hydrogen embrittlement than medium-strength, low-alloy steel. Thus, careful consideration of the microstructure in relation to the hydrogen environment is extremely important for robust design. For gas storage applications, fatigue crack propagation aided by hydrogen embrittlement must be quantitatively addressed as the existing fatigue data in hydrogen gas show that carbon steels can be significantly affected even at low gas pressure [29, 43, 44, 53, 54]. For low-alloy steels, data on fatigue crack propagation in hydrogen gas are largely lacking [34, 55–57] and most engineering experience is for lower numbers of pressure cycles (e.g. industrial gas cylinders) than is expected for a refueling application. Therefore, for both low-alloy and carbon steels, fatigue crack propagation is anticipated to be an important, if not limiting, consideration for service in hydrogen gas. Welding is an area of concern for low-alloy and carbon steels in any pressure system. Many pressure components are designed using seamless construction to minimize the potential of failure at welds, in particular the HAZ, which can have significantly lower resistance to fracture than the base metal. In hydrogen environments, the HAZ can be more susceptible to hydrogen embrittlement than the base metal. Relatively few studies have addressed hydrogen embrittlement of welds in ferritic steels [29, 30] and more comprehensive study is necessary. One of the important material characteristics governing weld properties is the carbon equivalent (CE). The CE is a weighted average of elements, where concentrations of carbon and manganese are significant factors [31]. Higher values of CE increase the propensity for martensite formation during welding. Since non-tempered martensite is extremely vulnerable to hydrogen embrittlement [7, 8], welds can become an area of significant concern if post-weld heat treatment cannot be applied.

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

69

Although low values of CE are specified to prevent martensite formation in welds [31], these regions are often still harder than the surrounding pipeline base metal. As documented in Section 3.4.1, higher hardness (i.e. higher yield strength) makes steel welds more susceptible to hydrogen embrittlement. Elevated temperature is a particularly limiting service condition for inexpensive steels in contact with hydrogen gas as an irreversible hydrogen degradation mechanism called hydrogen attack can operate (Section 3.4). Atomic hydrogen in steel will react with carbon under specific conditions forming methane, which is insoluble in metals forming porosity and blistering [2]. The conditions under which these reactions take place depend sensitively on the activity of carbon and therefore depend on the composition of the steel and the temperature. Exposure time also plays an important role, although hydrogen diffusion is quite rapid in ferritic steels; thus degradation can activate during relatively short exposure time. Carbon steels with low transition metal content (e.g. chromium, molybdenum) are the most susceptible to this degradation mechanism and generally not recommended for use at temperatures greater than about 473 K. Alloying elements that display a strong affinity for carbon such as chromium and molybdenum (the alloying agents in low-alloy steels) increase resistance to hydrogen attack effectively increasing the temperature at which these alloys can be used in hydrogen environments. The microalloying elements, such as vanadium and niobium, are added for their strong affinity for carbon and thus have the greatest benefit per weight. It is not clear to what extent microalloying elements specifically reduce hydrogen attack, but the low content of these alloying elements may not have a substantial effect. Document RP941 from the American Petroleum Institute (API) provides empirically established bounds for various alloy classes based on many decades of engineering experience [58]; these bounds are commonly referred to as the Nelson curves.

3.6.2

Austenitic stainless steels

Austenitic stainless steels are distinguished from other highly alloyed steels and ferritic steels by their crystal structure: the austenitic alloys have an FCC crystal structure compared with the BCC structure of the ferritic alloys. A high alloying content consisting of greater than 25 wt% of nickel, chromium, and molybdenum is another characteristic of the austenitic stainless steels. This class of alloys can be divided into several sub-classes: stable austenitic stainless steels; metastable austenitic stainless steels; precipitation-hardened stainless steels; and duplex stainless steels. The alloy composition specifications of austenitic stainless steels are quite wide; moreover, common grades such as type 304 stainless steels actually consist of a family of alloys: 304L (lowcarbon grade), 304N (nitrogen-strengthened), 304LN, and other variants. Strain hardening is an effective method of strengthening austenitic stainless

© 2008, Woodhead Publishing Limited

70

Solid-state hydrogen storage

steels. Heat treating cannot be used to strengthen the majority of austenitic stainless steels, with the exception of the precipitation-hardening alloys and, to some extent, the duplex alloys. In the metastable alloys, strain hardening can induce the formation of martensite in the microstructure, referred to as strain-induced martensite. This is an important characteristic since tubing is one common application of stainless steels and it is typically bent during fabrication of hardware (e.g. gas manifolding) and as mentioned previously martensite adversely affects hydrogen embrittlement. The distinction between the stable and metastable alloys is somewhat arbitrary, as the amount of straininduced martensite is a strong function of temperature and alloy composition. Common stable austenitic stainless steels include type 310 and the highMn stainless steels (which are often called nitrogen-strengthened stainless steels). Common metastable austenitic alloys include type 304 stainless steels and the type 316 stainless steels. Type 316 alloys are sometimes considered stable since they are significantly more stable than the 304 alloys, although strain-induced martensite can form in type 316 alloys deformed extensively at room temperature and will be present if highly deformed at sub-ambient temperature. In the context of hydrogen embrittlement, A-286 is the most common precipitation-strengthened austenitic stainless steel. This alloy has additions of titanium and aluminum that form strengthening precipitates during controlled heat treatment. The duplex stainless steels, which consist of a well-balanced microstructure of austenite and ferrite, are included here because they have fracture resistance in hydrogen environments that appears to be characteristic of the austenitic alloys [59]. Alloy composition appears to be the primary distinguishing feature for hydrogen embrittlement resistance among the austenitic stainless steels. In particular, hydrogen embrittlement resistance increases with nickel content and several studies have demonstrated critical nickel content in the range of 10–12 wt% nickel [11, 12, 60], below which the embrittlement susceptibility increases dramatically, as demonstrated by the tensile ductility data in Fig. 3.9. This critical composition of nickel depends on other compositional variables and temperature [60], as well as on the metric for assessing hydrogen embrittlement. Austenitic stainless steels have a number of important advantages over the ferritic alloys [11, 12]: (i) generally the most resistant to hydrogen embrittlement of all the classes of steel, (ii) good baseline properties over a wide range of temperature from cryogenic to elevated temperature, and (iii) very low hydrogen permeability. The primary disadvantage of the austenitic stainless steels is cost; these alloys have a large alloying content (primarily chromium, nickel and molybdenum) resulting in cost premiums of 5 to 10 times that of carbon steels. In addition, austenitic stainless steels are relatively low strength compared to low-alloy ferritic steels, although some alloys can be strengthened by thermomechanical processing [11, 12].

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

71

Retained tensile ductility (%)

100

80

60

40

20

Commercial alloys High-purity alloys

0 0

10

20

30 Nickel (wt%)

40

50

60

3.9 The effect of 69 MPa hydrogen gas on the tensile ductility of Fe– Cr–Ni alloys at room temperature (reproduced from Caskey [11]).

Austenitic stainless steels are generally the first choice for hydrogen gas systems. Type 316 stainless steels, in particular, are emerging as the preferred material of construction for tubing and small-bore devices, such as valve bodies. A-286 has been used extensively in high-pressure hydrogen gas systems because of its comparatively high strength; however, fracture measurements indicate that at constant strength level it is not superior to other austenitic stainless steels in high-pressure hydrogen [27, 59] and can be much worse if the microstructure is not carefully controlled [61].

3.6.3

Highly alloyed steels

Generally, highly alloyed steels (with the exception of austenitic stainless steels) are not compatible with hydrogen and caution should be exercised in considering these materials in containment structures [8]. Ferritic and martensitic stainless steels, for example, are highly susceptible to hydrogen embrittlement in high-pressure hydrogen gas [4, 55]. Figure 3.10 shows that common highly alloyed steels demonstrate little resistance to hydrogen embrittlement. NASA guidance precludes the use of high-nickel steels [50], such as HP9-4-20.

3.6.4

Nickel alloys

Nickel-based alloys are not expected to be materials of construction for hydrogen gas containment for two important reasons: (i) very high cost and

© 2008, Woodhead Publishing Limited

72

Solid-state hydrogen storage

Strength ratio from notched tensile tests in 69 MPa gas (H2 : He)

1

0.8

0.6

0.4

Fe-base Highly-alloyed Fe Austenitic Fe Ni-base Ti-base Cu-base Al-base

0.2

0

0

500 1000 1500 Yield strength (MPa) from smooth tensile tests

2000

3.10 Hydrogen compatibility of several engineering alloys in 69 MPa hydrogen gas at room temperature plotted as a function of yield strength of the material. The metric for resistance to hydrogen embrittlement is the ratio of the notched tensile strength in hydrogen gas to that in helium, such that a value near 1 shows essentially no effect of hydrogen. The arrows indicate pure metals.

(ii) susceptibility to hydrogen embrittlement. The single-phase nickel alloys have shown similar ductility loss as some austenitic stainless steels after hydrogen gas exposure [62, 63]; however, these alloys have relatively low strength and thus offer no structural advantage over austenitic stainless steels. The nickel-base superalloys (precipitation-hardened, thus having a multiphase microstructure) feature substantially higher strength, particularly at elevated temperature, than the austenitic stainless steels, but their resistance to hydrogen embrittlement is comparatively low [64, 65]. While nickel-based alloys are unlikely to be used for hydrogen gas containment, they are used for specialized components in devices for hydrogen service, such as valve springs that require high-strength, corrosion-resistant alternatives to iron-based alloys as well as for elevated temperature applications.

3.6.5

Aluminum alloys

Aluminum alloys are desirable in many applications because of their high strength to weight ratio and corrosion resistance. This is particularly true in

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

73

transport applications and in pressure systems where the strength of materials is often purposefully kept low. Aluminum, however, has its own set of disadvantages: (i) high cost compared with steels; (ii) relatively low fracture toughness in high strength conditions; and (iii) limited performance at elevated temperature. Aluminum is a low melting point metal, therefore microstructural stability (and consequently creep) is an important issue even for near ambient temperature. Many common structural aluminum alloys are precipitation strengthened at heat treatment temperatures on the order of 423 K, implying that the service temperature must be considerably lower. There are a few aluminum alloys designed for ‘high-temperature’ applications; however, creep makes them unacceptable as materials of construction for containment of pressurized gas at elevated temperature. On the other hand, aluminum alloys are commonly employed at cryogenic temperatures. The available laboratory data show that aluminum alloys have excellent hydrogen embrittlement resistance in dry hydrogen gas [2, 4, 6, 38, 66]. However, aluminum alloys become susceptible to hydrogen embrittlement in gases that contain water vapor [38]. The water vapor creates a thermodynamic condition on the surface of aluminum equivalent to extremely high-fugacity hydrogen gas, which increases the amount of atomic hydrogen in the aluminum lattice thus facilitating hydrogen embrittlement. Therefore, the moisture content is an important consideration for gas storage systems that utilize aluminum components.

3.6.6

Other non-ferrous alloys

Titanium is commonly employed as a structural material in the aerospace industry because of its high strength to weight ratio. It is less common in the automotive industry as a structural material because of its high cost. The compatibility of titanium with hydrogen gas strongly depends on the microstructure of the alloy. Hydrogen diffusion in α-titanium is very slow; however, brittle hydrides form in α-titanium exposed to hydrogen, severely embrittling the material [2]. Hydrogen diffusion in β-titanium is rapid and this phase is resistant to hydride formation. Therefore, β-titanium is often considered to be resistant to hydrogen embrittlement, although it can be embrittled under certain conditions [2]. Most engineering titanium alloys consist of both the α- and β-phases, therefore, titanium alloys are, in general, considered highly susceptible to hydrogen embrittlement. Copper and gold are not considered structural materials since they tend to have very low strength. However, they have an important technological niche in hydrogen gas systems related to their resistance to hydrogen permeation [22, 67]. Permeation of hydrogen through gold is especially low, being many orders of magnitude less than structural metals (Fig. 3.1). This feature combined

© 2008, Woodhead Publishing Limited

74

Solid-state hydrogen storage

with their common use in sealing applications makes them ideal for gaskets and seals in hydrogen gas systems. Copper alloys can be significantly embrittled by hydrogen if oxygen is present in the alloy [68, 69]. Oxygen (or oxides) in copper will react with hydrogen forming pores and blistering, since the reaction product (water) is insoluble in copper. This problem is easily avoided by using oxygen-free grades of copper in systems exposed to hydrogen gas and preventing thermal exposure (since hydrogen embrittlement of oxygen-containing copper is particularly insidious at elevated temperature).

3.6.7

Composites

Composite pressure vessels have been used extensively for storing compressed natural gas on vehicles and similar storage concepts are being developed for hydrogen gas storage. In the context of hydrogen gas pressure systems, composites are fiber-reinforced polymer-matrix materials, commonly referred to as fiber-reinforced polymers (FRP). Carbon fibers are the structural fiber of choice, although they are currently prohibitively expensive for high-volume applications; thus less expensive fibers, such as glass, are often used to construct composite pressure vessels. The effects of hydrogen gas on the structural composite are not considered to be a major obstacle to the use of composites because a liner is typically used as a permeation barrier. The liner material is selected to have low hydrogen permeability and can be either metal (typically 6061 aluminum) or a polymer (commonly high-density polyethylene, HDPE). Therefore, interactions between hydrogen and the composite material are mitigated. Considering solid storage of hydrogen, composite containment structures will be limited by the thermal management associated with engineering of the system. In particular, FRP are more sensitive to temperature than most metals and the thermal conductivity can be low or highly directional, making heat management on short timescales a challenging engineering task.

3.6.8

Summary of materials selection

Qualitative measures of resistance to hydrogen embrittlement should not be overemphasized since they do not provide useful information for component design. Efforts to categorize structural metals with respect to hydrogen embrittlement, such as the large volume of data generated by NASA [4, 55, 56], nevertheless provides a means of assessing the relative compatibility that can be expected of broad classes of alloys. The relative notched tensile strengths (the ratio of the strength of notched tensile specimens in highpressure hydrogen gas relative to helium gas, such that a value of one is

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

75

desirable for hydrogen compatibility) of a number of structural metals are shown in Fig. 3.10 as a function of yield strength. This plot clearly demonstrates that, in a general sense, high yield strength materials tend to be more susceptible to hydrogen embrittlement than materials with yield strength less than about 700 MPa. This trend underscores the importance of austenitic stainless steels and the potential of aluminum alloys for a good combination of strength and ductility in hydrogen environments.

3.7

Future trends

Any hydrogen containment system will nominally consist of a storage vessel and gas manifold. For storage vessels, the candidate structural materials are likely ferritic steel, aluminum, and lined composites. Aluminum resists hydrogen embrittlement in dry hydrogen gas, and no hydrogen-induced degradation has been reported for composite materials in lined vessels. While steels are susceptible to hydrogen embrittlement, the effect of embrittlement on vessel performance can be controlled by considering variables such as material strength, welding, and load cycling. Although safe and reliable vessels can be designed from all three materials, materials selection will balance manufacturing viability and cost as well as constraints imposed by operating parameters. Operating parameters such as temperature, thermal management, and gas pressure must be considered in materials selection for hydrogen storage vessels. Operating temperatures above 473 K will likely preclude both aluminum and composite materials, since such elevated temperatures can degrade mechanical properties of the materials. In addition, the need for effective thermal management could eliminate composite materials, which may not have adequate thermal conductivity. Vessels that contain hydrogen gas at very high pressure (e.g. 100 MPa) may not be safely constructed from aluminum alloys, since these materials have relatively low strength and fracture toughness. Ferritic steels also present complications for high-pressure vessels, since the vessels would have relatively thick walls (e.g. greater than 2.5 cm) to ensure the wall stresses meet the design criteria. Such a vessel would be heavy and impractical as a fuel tank or transport cylinder. A highpressure steel vessel could also present challenges in manufacturing, as heat treating a thick-walled vessel while maintaining uniform properties could be difficult. Assuming thick-walled steel vessels could be manufactured, these vessels would most likely be considered only for stationary storage. Lined composites may therefore be attractive for high-pressure vessels, particularly as fuel tanks or transport cylinders; however the compromise may be higher component cost. Similar to storage vessels, the process of materials selection for manifold components will balance performance, manufacturing, and cost. The current

© 2008, Woodhead Publishing Limited

76

Solid-state hydrogen storage

trend is that gas manifold components, i.e. tubing and valves, will be fabricated from austenitic stainless steels. Manifold components can be readily fabricated from austenitic stainless steels, and these steels are relatively resistant to hydrogen embrittlement. Although stainless steels are costly owing to the alloying elements nickel, chromium, and molybdenum, manifold components do not require large quantities of material. Because material costs could be reduced by using ferritic steels and aluminum in manifold components, these materials are attractive to manufacturers. However, there is no documented experience of ferritic steels and aluminum alloys in components such as tubing and valves for hydrogen gas systems, so the performance of these materials will have to be demonstrated for such applications.

3.8

Other sources

One of the most prominent concerns for structural metals in hydrogen gas containment systems is hydrogen embrittlement. Research on hydrogen embrittlement has been extensive, and a broad sampling of the work conducted on hydrogen embrittlement can be found in the proceedings from a series of international topical conferences on the subject [70–75]. Many of the mechanical property measurements conducted on materials exposed to hydrogen gas are not useful for the design of hydrogen containment structures, since the material, environmental, or mechanical variables in the testing do not represent typical service conditions. However, an effort has been initiated to identify and document data from the technical literature that could be useful in design. The data have been compiled in an Internet-based resource entitled ‘Technical Reference on Hydrogen Compatibility of Materials’ (www.ca.sandia.gov/matlsTechRef). Recognizing the need for guidance in the design of hydrogen containment structures, a number of industrial and standards development organizations are developing technical documents. The European Industrial Gases Association (EIGA) has published two documents on hydrogen gas cylinders and hydrogen gas pipelines [31, 45]. The latter document was published in collaboration with the Compressed Gas Association (CGA). In addition, the American Society of Mechanical Engineers (ASME) has generated a comprehensive review of the documented experience with hydrogen systems [13] and formed project teams on hydrogen tanks as well as on hydrogen piping and pipelines. These teams are producing new standards for the construction of vessels for hydrogen gas storage and of piping and pipelines for hydrogen gas transport. The project team on hydrogen tanks is also developing a document specifically for hydride containment vessels.

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

3.9

77

Acknowledgements

This work was supported by the US Department of Energy under Contract DE-ACO4-94AL85000.

3.10

References

1. HG Nelson, ‘Testing for hydrogen environment embrittlement: primary and secondary influences’, in Hydrogen Embrittlement Testing, ASTM STP 543, American Society for Testing and Materials, Philadelphia, PA, 1974, pp. 152–169. 2. HG Nelson, ‘Hydrogen embrittlement’, in Embrittlement of Engineering Alloys, vol. 25, CL Briant and SK Banerji, Eds., Academic Press, New York, 1983, pp. 275–359. 3. MR Louthan, GR Caskey, JA Donovan, and DE Rawl, ‘Hydrogen embrittlement of metals’, Materials Science and Engineering, vol. 10, 1972, pp. 357–68. 4. RP Jewitt, RJ Walter, WT Chandler, and RP Frohmberg, Hydrogen Environment Embrittlement of Metals, NASA CR-2163, Rocketdyne for the National Aeronautics and Space Administration, Canoga Park, CA, March 1973. 5. AW Loginow and EH Phelps, ‘Steels for seamless hydrogen pressure vessels’, Corrosion, vol. 31, 1975, pp. 404–12. 6. MR Louthan and G Caskey, ‘Hydrogen transport and embrittlement in structural metals’, International Journal of Hydrogen Energy, vol. 1, 1976, pp. 291–305. 7. AW Thompson, ‘Materials for hydrogen service’, in Hydrogen: Its Technology and Implications: Volume II Transmission and Storage, KE Cox and KD Williamson, Eds., CRC Press, Cleveland, OH, 1977, pp. 85–124. 8. AW Thompson and IM Bernstein, ‘Selection of structural materials for hydrogen pipelines and storage vessels’, International Journal of Hydrogen Energy, vol. 2, 1977, pp. 163–73. 9. AW Thompson and IM Bernstein, ‘The role of metallurgical variables in hydrogenassisted environmental fracture’, in Advances in Corrosion Science and Technology, vol. 7, MG Fontana and RW Staehle, Eds., Plenum Publishing Corporation, New York, 1980, pp. 53–175. 10. GR Caskey, ‘Hydrogen damage in stainless steel’, in Environmental Degradation of Engineering Materials in Hydrogen, MR Louthan, RP McNitt, and RD Sisson, Eds., Laboratory for the Study of Environmental Degradation of Engineering Materials, Virginia Polytechnic Institute, Blacksburg, VA, 1981, pp. 283–302. 11. GR Caskey, Hydrogen Compatibility Handbook for Stainless Steels (DP-1643), EI du Pont Nemours, Savannah River Laboratory, Aiken, SC, June 1983. 12. GR Caskey, ‘Hydrogen effects in stainless steels’, in Hydrogen Degradation of Ferrous Alloys, RA Oriani, JP Hirth, and M Smialowski, Eds., Noyes Publications, Park Ridge, NJ, 1985, pp. 822–62. 13. Hydrogen Standardization Interim Report for Tanks, Piping, and Pipelines, American Society of Mechanical Engineers, New York, May 2005. 14. Technical Reference on Hydrogen Compatibility of Materials, available at http:// www.ca.sandia.gov/matlsTechRef/, Sandia National Laboratories, Livermore, CA. 15. J Crank, The Mathematics of Diffusion. Oxford University Press, Oxford, 1975. 16. JP Hirth, ‘Effects of hydrogen on the properties of iron and steel’, Metallurgical Transactions, vol. 11A, 1980, pp. 861–890. 17. C San Marchi, BP Somerday, and SL Robinson, ‘Permeability, solubility and diffusivity

© 2008, Woodhead Publishing Limited

78

18. 19.

20. 21.

22.

23.

24. 25.

26.

27.

28.

29.

30.

31. 32.

Solid-state hydrogen storage of hydrogen isotopes in stainless steels at high gas pressure’, International Journal of Hydrogen Energy, vol. 32, 2007, pp. 100–16. MR Louthan and RG Derrick, ‘Hydrogen transport in austenitic stainless steel’, Corrosion Science, vol. 15, 1975, pp. 565–77. HG Nelson and JE Stein, Gas-phase Hydrogen Permeation Through Alpha Iron, 4130 Steel, and 304 Stainless Steel from less than 100°C to near 600°C (NASA TN D-7265), Ames Research Center, National Aeronautics and Space Administration, Moffett Field, CA, April 1973. MR Louthan, JA Donovan, and GR Caskey, ‘Hydrogen diffusion and trapping in nickel’, Acta metallurgica, vol. 23, 1975, pp. 745–9. W Song, J Du, Y Xu, and B Long, ‘A study of hydrogen permeation in aluminum alloy treated by various oxidation processes’, Journal of Nuclear Materials, vol. 246, 1997, pp. 139–43. DR Begeal, ‘Hydrogen and deuterium permeation in copper alloys, copper-gold brazing alloys, gold, and the in situ growth of stable oxide permeation barriers’, Journal of Vacuum Science and Technology, vol. 15, 1978, pp. 1146–54. RW Schefer, WG Houf, C San Marchi, WP Chernicoff, and L Englom, ‘Characterization of leaks from compressed hydrogen dispensing systems and related components’, International Journal of Hydrogen Energy, vol. 31, 2006, pp. 1247–60. HK Birnbaum, ‘Hydrogen effects on deformation and fracture: science and sociology’, MRS Bulletin, July 2003, pp. 479–85. HK Birnbaum, IM Robertson, P Sofronis, and D Teter, ‘Mechanisms of hydrogen related fracture – a review’, in Corrosion–Deformation Interactions, T Magnin, Ed., Woodhead Publishing Limited, Cambridge, 1997. RP Gangloff, ‘Hydrogen assisted fracture of high strength alloys’, in Comprehensive Structural Integrity, vol. 6, I Milne, RO Ritchie, and B Karihaloo, Eds., Elsevier Science, New York, 2003. MW Perra, ‘Sustained-load cracking of austenitic steels in gaseous hydrogen’, in Environmental Degradation of Engineering Materials in Hydrogen, MR Louthan, RP McNitt, and RD Sisson, Eds., Laboratory for the Study of Environmental Degradation of Engineering Materials, Virginia Polytechnic Institute, Blacksburg, VA, 1981, pp. 321–33. BP Somerday, SX McFadden, DK Balch, JD Puskar, and CH Cadden, ‘Hydrogenassisted fracture of inertia welds in 21Cr–6Ni–9Mn stainless steel’, in Environmentinduced Cracking of Materials, S Shipilov, RH Jones, J-M Olive, and R Rebak, Eds., Elsevier, Amsterdam, vol. 1, 2008, pp. 305–314. HJ Cialone and JH Holbrook, ‘Sensitivity of steels to degradation in gaseous hydrogen’, in Hydrogen Embrittlement: Prevention and Control, ASTM STP 962, EL Raymond, Ed., American Society of Testing and Materials, Philadelphia, PA, 1988, pp. 134– 52. WR Hoover, SL Robinson, RE Stoltz, and JR Spingarn, ‘Hydrogen Compatibility of Structural Materials for Energy Storage and Transmission. Final Report’, SAND81– 8006, Sandia National Laboratories, Livermore, CA, 1981. Hydrogen Transport Pipelines, IGC 121/04/E, European Industrial Gases Association (EIGA), Brussels, 2004. SL Robinson and RE Stoltz, ‘Toughness losses and fracture behavior of low strength carbon–manganese steels in hydrogen’, in Hydrogen Effects in Metals, IM Bernstein and AW Thompson, Eds., The Metallurgical Society of AIME, New York, 1981, pp. 987–95.

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

79

33. WT Chandler and RJ Walter, ‘Testing to determine the effect of high-pressure hydrogen environments on the mechanical properties of metals’, in Hydrogen Embrittlement Testing, ASTM STP 543, American Society for Testing and Materials, Philadelphia, PA, 1974, pp. 170–97. 34. S Fukuyama and K Yokogawa, ‘Prevention of hydrogen environmental assisted crack growth of 2.25Cr-1Mo steel by gaseous inhibitors’, in Pressure Vessel Technology, vol. 2, Verband der Technischen Uberwachungs-Vereine, Essen, Germany, 1992, pp. 914–23. 35. HG Nelson, ‘Hydrogen-induced slow crack growth on a plain carbon pipeline steel under conditions of cyclic loading’, in Effect of Hydrogen on Behavior of Materials, AW Thompson and IM Bernstein, Eds., The Metallurgical Society of AIME, New York, 1976, pp. 602–11. 36. WG Clark, ‘Effect of temperature and pressure on hydrogen cracking in high strength type 4340 steel’, Journal of Materials for Energy Systems, vol. 1, 1979, pp. 33–40. 37. WG Clark and JD Landes, ‘An evaluation of rising load KIscc testing’, in Stress Corrosion – New Approaches, ASTM STP 610, American Society for Testing and Materials, Philadelphia, PA, 1976, pp. 108–27. 38. MO Speidel, ‘Hydrogen embrittlement and stress corrosion cracking of aluminum alloys’, in Hydrogen Embrittlement and Stress Corrosion Cracking, R Gibala and RF Hehemann, Eds., American Society for Metals, Metals Park, OH, 1984, pp. 271–96. 39. HG Nelson and DP Williams, Quantitative Observations of Hydrogen-Induced, Slow Crack Growth in Low Alloy Steel, NASA TMX–62253, Ames Research Center, National Aeronautics and Space Administration, Moffett Field, CA, 1973. 40. HG Nelson and DP Williams, ‘Quantitative observations of hydrogen-induced, slow crack growth in low alloy steel’, in Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Base Alloys, RW Staehle, J Hochmann, RD McCright, and JE Slater, Eds., NACE, Houston, TX, 1977, pp. 390–404. 41. JH Holbrook and AJ West, ‘The effect of temperature and strain rate on the tensile properties of hydrogen charged 304L, 21-6-9, and JBK 75’, in Hydrogen Effects in Metals, IM Bernstein and AW Thompson, Eds., The Metallurgical Society of AIME, New York, 1981, pp. 655–63. 42. LM Ma, GJ Liang and YY Li, ‘Effect of hydrogen charging on ambient and cryogenic mechanical properties of a precipitate-strengthened austenitic steel’, in Advances in Cryogenic Engineering, vol. 38A, FR Fickett and RP Reed, Eds., Plenum Press, New York, 1992, pp. 77–84. 43. HF Wachob and HG Nelson, ‘Influence of microstructure on the fatigue crack growth of A516 in hydrogen’, in Hydrogen Effects in Metals, IM Bernstein and AW Thompson, Eds., The Metallurgical Society of AIME, New York, 1981, pp. 703–11. 44. RJ Walter and WT Chandler, ‘Cyclic-load crack growth in ASME SA-105 grade II steel in high-pressure hydrogen at ambient temperature’, in Effect of Hydrogen on Behavior of Materials, AW Thompson and IM Bernstein, Eds., The Metallurgical Society of AIME, New York, 1976, pp. 273–86. 45. Hydrogen Cylinders and Transport Vessels, IGC 100/03/E, European Industrial Gases Association (EIGA), Brussels, 2003. 46. RS Irani, ‘Hydrogen storage: high-pressure gas containment’, MRS Bulletin, vol. 27, 2002, pp. 680–2. 47. C San Marchi, DK Balch, K Nibur, and BP Somerday, ‘Effect of high-pressure hydrogen gas on fracture of austenitic steels’, Journal of Pressure Vessel Technology,. accepted.

© 2008, Woodhead Publishing Limited

80

Solid-state hydrogen storage

48. JA Harris and MCV Wanderham, Properties of Materials in High Pressure Hydrogen at Cryogenic, Room, and Elevated Temperatures, NASA CR-124394, Pratt and Whitney Aircraft (report no. FR-5768) for the National Aeronautics and Space Administration (Marshall Space Flight Center), West Palm Beach, FL, July 1973. 49. Hydrogen Pipeline Working Group Workshop, US Department of Energy, Augusta, GA, 2005. 50. PM Ordin, Safety Standard for Hydrogen and Hydrogen Systems: Guidelines for Hydrogen System Design, Materials Selection, Operations, Storage, and Transportation, Office of Safety and Mission Assurance, National Aeronautics and Space Administration, Washington DC, 1997. 51. C San Marchi and BP Somerday, ‘Effects of high-pressure gaseous hydrogen on structural metals (2007-01-0433)’, in SAE 2007 Transactions, SAE, Warrendale PA, 2008. 52. G Krauss, Steels: Heat Treatment and Processing Principles. ASM International, Materials Park, OH, 1990. 53. HJ Cialone and JH Holbrook, ‘Effects of gaseous hydrogen on fatigue behavior of low strength carbon–manganese steels in hydrogen’, Metallurgical Transactions, vol. 16A, 1985, pp. 115–22. 54. HG Nelson, ‘On the mechanism of hydrogen-enhanced crack growth in ferritic steels’, in Proceedings of the Second International Conference on Mechanical Behavior of Materials, ASM, Metals Park, OH, 1976, pp. 690–4. 55. RJ Walter and WT Chandler, Effects of High-pressure Hydrogen on Metals at Ambient Temperature: Final Report, NASA CR-102425, Rocketdyne (report no. R-7780-1) for the National Aeronautics and Space Administration, Canoga Park, CA, February 1969. 56. RJ Walter and WT Chandler, Influence of Gaseous Hydrogen on Metals: Final Report, NASA CR-124410, Rocketdyne for the National Aeronautics and Space Administration, Canoga Park, CA, October 1973. 57. WG Clark, ‘The effect of hydrogen gas on the fatigue crack growth rate behavior of HY-80 and HY-130 steels’, in Hydrogen in Metals, IM Bernstein and AW Thompson, Eds., ASM, Metals Park, OH, 1974, pp. 149–64. 58. Steels for Hydrogen Service at Elevated Temperatures and Pressures in Petroleum and Petrochemical Plants, API RP941, American Petroleum Institute, Washington DC, 2004. 59. C San Marchi, BP Somerday, J Zelinski, X Tang, and GH Schiroky, ‘Mechanical properties of super duplex stainless steel 2507 after gas phase thermal precharging with hydrogen’, Metallurgical and Materials Transactions, accepted. 60. L Zhang, M Wen, M Imade, S Fukuyama, and K Yokogawa, ‘Effect of nickel equivalent on hydrogen environment embrittlement of austenitic stainless steels at low temperatures’, in Fracture of Nano and Engineering Materials and Structures, Alexandroupolis, Greece, 2006. 61. MW Perra and RE Stoltz, ‘Sustained-load cracking of a precipitation-strengthened austenitic steel in high-pressure hydrogen’, in Hydrogen Effects in Metals, IM Bernstein and AW Thompson, Eds., The Metallurgical Society of AIME, New York, 1981, pp. 645–53. 62. DM Symons, ‘Hydrogen embrittlement of Ni–Cr–Fe alloys’, Metallurgical and Materials Transactions, vol. 28A, 1997, pp. 655–63. 63. DM Symons, ‘The effect of carbide precipitation on the hydrogen-enhanced fracture behavior of alloy 690’, Metallurgical and Materials Transactions, vol. 29A, 1998, pp. 1265–77.

© 2008, Woodhead Publishing Limited

Hydrogen containment materials

81

64. DM Symons, ‘A comparison of internal hydrogen embrittlement and hydrogen environment embrittlement of X-750’, Engineering Fracture Mechanics, vol. 68, 2001, pp. 751–71. 65. DM Symons and AW Thompson, ‘The effect of hydrogen on the fracture of alloy X750’, Metallurgical and Materials Transactions, vol. 27A, 1996, pp. 101–10. 66. RM Vennett and GS Ansell, ‘A study of gaseous hydrogen damage in certain FCC metals’, Transactions of the ASM, vol. 62, 1969, pp. 1007–13. 67. WG Perkins, ‘Permeation and outgassing of vacuum materials’, Journal of Vacuum Science and Technology, vol. 10, 1973, pp. 543–56. 68. E Mattsson and F Schueckher, ‘An investigation of hydrogen embrittlement in copper’, Journal of the Institute of Metals, vol. 87, 1958–59, pp. 241–7. 69. GR Caskey, AH Dexter, ML Holzworth, MR Louthan, and RG Derrick, ‘The effect of oxygen on hydrogen transport in copper’, Corrosion, vol. 32, 1976, pp. 370–4. 70. IM Bernstein and AW Thompson, ‘Hydrogen in metals’, in Proceedings of the International Conference on the Effects of Hydrogen on Materials Properties and Selection and Structural Design (Champion, PA, 1973) American Society for Metals, Metals Park, OH, 1974. 71. AW Thompson and IM Bernstein, ‘Effect of hydrogen on behavior of materials’, in Proceedings of an International Conference on Effect of Hydrogen on Behavior of Materials (Moran, WY, 1975) The Metallurgical Society of AIME, New York, 1976. 72. IM Bernstein and AW Thompson, ‘Hydrogen effects in metals’, in Proceedings of the Third International Conference on Effect of Hydrogen on Behavior of Materials (Moran, WY, 1980) The Metallurgical Society of AIME, New York, 1981. 73. NR Moody and AW Thompson, ‘Hydrogen effects on material behavior’, in Proceedings of the Fourth International Conference on the Effect of Hydrogen on the Behavior of Materials (Moran, WY, 1989) TMS, Warrendale, PA, 1990. 74. AW Thompson and NR Moody, ‘Hydrogen effects in materials’, in Proceedings of the Fifth International Conference on the Effect of Hydrogen on the Behavior of Materials (Moran WY, 1994) TMS, Warrendale, PA, 1996. 75. NR Moody, AW Thompson, RE Ricker, GW Was, and RH Jones, ‘Hydrogen effects on materials and corrosion deformation interactions’, in Proceedings of the International Conference for Hydrogen Effects on Material Behavior and Corrosion Deformation Interactions (Moran, WY, 2002) TMS, Warrendale, PA, 2003.

© 2008, Woodhead Publishing Limited

4 Solid-state hydrogen storage system design D. E. D E D R I C K, Sandia National Laboratories, USA

4.1

Introduction

Efficient hydrogen storage is a significant challenge inhibiting the use of hydrogen as a primary energy carrier. In general, three main categories of energy storage applications require the use of energy-dense hydrogen storage materials; transportation, portable auxiliary power, and personal electronics. Each of these applications requires relatively high energy and power density capability. For transportation applications, the US Department of Energy (US DoE) has specified challenging volumetric and gravimetric energy density targets for technology acceptance (3 kW h/kg and 2.7 kW h/l) [1]. Portable power systems, including man-portable and auxiliary power systems, need to compete with the power densities of other technologies such as gas turbine and lithium-based battery technologies to be viable (0.100–1.5 kW h/kg and kW h/l) [2]. Personal electronics power and energy density requirements are continually increasing as portable computation demands increase. Lithium ion batteries are the current technology baseline and are characterized by energy densities of 0.2 kW h/kg and 0.3 kW h/l. This chapter will focus on automotive applications as these will likely have the most extensive impact on consumer-scale energy storage technologies. Automotive systems are arguably the most challenging energy storage application since the low cost and high performance of current fossil fuel technologies cause consumers to expect similarly plentiful energy and power from new transportation technologies. For automotive systems, current hydrogen storage technologies (liquid and compressed gas) fall short in terms of energy density and cost and do not provide a clear path for efficiency improvement and the resulting increase in energy density (Fig. 4.1). Hydrogen storage technologies that are efficient, low cost, and robust must be developed to enable the use of hydrogen fuel in transportation applications. Solid-state hydrogen storage solutions are theoretically able to store more hydrogen per unit volume than liquid or solid storage systems [3]. Given this potential for high energy density reversible 82 © 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

83

System volumetric capacity (g/l)

100

80

2015 US DoE targets

60 2010 US DoE targets Chemical hydride

40 Complex hydride

20

Liquid hydrogen

700 bar compressed

300 bar compressed

0 0

2

4 6 8 System gravimetric capacity (wt%)

10

4.1 The current status of automotive hydrogen storage technologies indicates that significant performance improvements must be realized prior to attaining 2010 and 2015 goals (adapted from [1]).

hydrogen storage, significant effort has been applied to develop storage solutions based on this technology. Some automotive-scale solid-state hydrogen storage systems have been developed and demonstrated. Sandia National Laboratories and General Motors Corporation developed automotive-scale hydrogen storage systems based on metal hydrides [4]. United Technologies Research Corporation was funded by the US DoE to demonstrate a large scale (~0.5 kg H2) hydrogen storage tank based on sodium alanates [5]. A few larger-scale demonstrations exist including a fuel-cell powered Class 212 German submarine [6, 7] and a Snowcat [8]. The submarine, built by Howaldtswerke–Deutsche Werft (HDW), utilizes Fe–Ti metal hydride tanks capable of storing a tonne of hydrogen. The storage of hydrogen in solid-solution materials is characterized by endothermic and exothermic phase change reactions. The thermodynamic nature of these reactions requires the full characterization of influential material properties to enable the optimization of heat and mass transfer within the system. Additionally, the thermodynamic nature of the materials will define the containment technologies required to withstand the operational pressures and temperatures. In addition to optimizing the hydrogen uptake and delivery performance of the system, safe operation must be ensured through hazard-minimizing design. Since some solid-solution hydrogen storage materials tend to be highly reactive in nature, extra care must be taken to ensure that the system user is not exposed to an unnecessary amount of risk. The system designer must fully understand the risks involved in order to design appropriately safe systems and controls.

© 2008, Woodhead Publishing Limited

84

Solid-state hydrogen storage

4.2

The behavior of solid-state hydrogen storage materials in systems

Metrics for system performance are typically described in terms of gravimetric and volumetric energy density. These two quantities define the mass and volume required within the application to provide sufficient hydrogen storage and delivery. These two performance metrics are directly influenced by hydrogen storage material properties including thermodynamic, physical, and chemical/kinetic. Additional influences on energy densities include, but not limited to, properties of the containment material, available energy, operational environment, and required hydrogen delivery rates. Figure 4.2 describes the hierarchical nature of the material properties’ influence on system energy densities. The thermodynamic characteristic (∆H – enthalpy of reaction) of the phase change material has a two-fold influence on the system. First, this property defines the amount of required heat transfer during hydrogen uptake and release. Secondly, the containment structure must be constructed based on the thermodynamic nature of the material since the enthalpy of reaction is defined by the equilibrium pressures and temperatures. The physical properties of the storage material, such as thermal conductivity, are highly influential on the design of a solid-state hydrogen storage system as these define the temporally varying rate at which heat can be removed or added to the system to support the phase change reaction. The chemical/ kinetic properties of the hydrogen storage material also influence the system energy densities by defining the operational conditions, such as temperature, and quantities of reacting material required to supply the application-specific Thermodynamic properties

Physical properties

Chemical/kinetic properties

∆H

Pressure

Temperature

Containment technology, e.g. metal, composite

Solid density

Geometry, defined by heat transfer

Thermal properties

Inefficiencies and losses

Uptake and release rates

Gravimetric and volumetric system energy density

4.2 The hierarchical influence of various hydrogen storage material properties on the system characteristics and efficiencies.

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

85

hydrogen flow demands. Other metrics, such as cost, are also important and must be considered based on the target application. Although less quantifiable, the safety performance of the system can be equally or exceedingly important to the system design. For consumer applications, the hazards associated with the use and operation of systems must be minimized to meet levels of acceptable risk. The optimum system design for any hydrogen storage application will vary based on acceptable cost, operational environment, and the safety and performance demands.

4.3

Thermodynamic properties of hydrogen storage materials

Reversible hydrogen storage materials are characterized by endothermic decomposition and exothermic recombination. The enthalpy of the reversible reaction defines the quantity of heat that will be moved in and out of the system during one complete fueling cycle. The pressure and temperature operating regime of an on-board automotive hydrogen storage system limits the possible variation of the enthalpy of reaction regardless of the material chosen. In Fig. 4.3, equilibrium pressures are plotted as a function of temperature for various hydrogen storage materials. The boxed area represents the approximate regime appropriate for automotive applications based on

Pressure (atm)

100

10

1

0.1 1.5

2.0

2.5 3.0 1000/T (1/K) Thermodynamic data include: V LaNi4.6Mn0.4 TiFe MmNi4.5Mn0.5 LaNi5 TiMn1.5 LaNi4.7Al0.3 ZrFe1.5Cr0.5 MmNi4.5Al0.5 MmNi3.5Co0.7Al0.8 CaNi5 MmNi4.2Co0.2Mn0.3Al0.3 Ca0.7Mm0.3Ni5 LaNi4.8Sn0.2 TiFe0.8Ni0.2 TiV0.62Mn1.5 TiFe0.9Mn0.1 Zr0.8Ti0.2MnFe

3.5

4.0 ZrMn2 ZrCr2 TiCo Mg2Ni ZrNi NaAIH4 Na3AIH6

4.3 Van’t Hoff data of a variety of hydrogen storage materials. The shaded area describes approximate area appropriate for automotive applications.

© 2008, Woodhead Publishing Limited

86

Solid-state hydrogen storage

system temperature and delivery pressure. For on-board hydrogen storage, the boundaries lie approximately between 20 and 120 °C and 1 and 10 atm. The plateau pressure expression for any of these materials is described as follows:

ln( p H 2 ) = – 1 ( ∆H – T∆S ) = – ∆H + ∆S RT RT R

[4.1]

where p H 2 is the plateau pressure, R is the gas constant, T is the temperature, ∆H is the enthalpy of reaction, and ∆S is the change in entropy. The change in entropy is approximately equivalent for any hydride solid as the hydrogen gas has lost nearly all its translational degrees of freedom when bonding to the metal species. Since the intercept of the plateau pressure expression is ∆S/R, hydride solids that operate within a limited pressure and temperature regime have similar enthalpies of reaction (∆H). If higher temperature fuel cells or off-board refueling storage systems are utilized, this region could expand, but for our applications of interest, these approximate boundaries on thermodynamics are appropriate. Upon observation of the associated enthalpies of reaction within this regime, approximately 40 kJ per mol of hydrogen sorbed must be accommodated during each fueling and delivery cycle. The most the enthalpy of reaction can vary in this automotive-specific regime is approximately ±10 kJ/mol. Considering a 5 kg hydrogen system, approximately 100 MJ of heat must be moved into and out of the bed. Due to the refueling goals of 2 kg H2/min [1], absorption is the most demanding energy transfer process, requiring a cooling rate of nearly 0.7 MW on average. If you consider that the natural absorption of hydrogen into these materials is typically nonlinear, then you can assume that up to three times that heat rate must be removed from the bed. This amount of heat in a compact automotive system requires careful engineering and a quantitative understanding of all associated heat transfer properties. From an infrastructure point of view, each hydrogen refueling station that is capable of refueling hydride-containing storage systems must include the necessary industrial cooling hardware to remove this quantity of heat. For example, a station capable of refueling 100 automobiles per day must dissipate and/or utilize up to 10 000 MJ of low-quality heat per day at an average rate of ~300 kW over a 10 hour operational period. Significant engineering design is required to enable an efficient, automotive hydrogen refueling station that is capable of refueling hydride-containing storage systems in a cost-effective manner.

4.4

Thermal properties of hydrogen storage materials

So far in this chapter we have described the sorption of hydrogen in reversible hydrogen storage materials as being a significant challenge in heat transfer.

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

87

In this section, we will explore the modes of heat transfer that the system designer must learn to accommodate to design an efficient hydrogen storage system. The thermal properties of hydrogen storage materials directly influence the efficiency, performance, and cost of hydrogen storage systems based on complex metal hydride materials. Thermal design is the most challenging aspect of hydride-based automotive hydrogen storage system design due to the thermodynamic characteristics of typical hydrogen storage materials operating in the pressure and temperature regime of conventional polymer electrolyte membrane (PEM) fuel cells. Hydrogen uptake is the most thermally challenging operational state as the reaction is exothermic and on-board automotive refueling scenarios require rapid refueling rates. As discussed in the introductory section, during rapid hydrogen uptake of an automotive scale system, nearly 0.7 MW of cooling may be required to maintain a constant system temperature. These intensive heat flux conditions demand a detailed understanding and optimization of the thermal transfer properties present within a hydrogen storage system. High thermal conductivity of the metal hydride and low thermal resistance at the wall mitigate the effects of thermally limited reactions, enable the use of larger containment vessel diameters, and reduce the number of parts required for efficient operation. Since thermal resistance at the vessel wall can generally be accommodated during exothermic hydrogen absorption with the temperature and flow characteristics of the coolant fluid, the effective thermal conductivity of the system is of specific interest as this can limit the rate of the absorption reaction most significantly.

4.4.1

Thermal conductivity of a hydrogen storage packed particle bed

Solid hydrogen storage materials generally take the form of packed particles or porous structures immersed in hydrogen gas. The effective thermal conductivity of a packed particle bed can be described in terms of three distinct gas pressure regimes: low, intermediate, and high. The low-pressure regime is characterized by molecular or rarified gas transport in the interparticle space and typically exists at pressures much less than 1 atm owing to the small characteristic distance between each solid particle. The thermal properties of the bed are relatively invariant with gas pressure in the lowpressure regime as the gas has a nearly insignificant role in heat transfer. The intermediate-pressure regime is characterized by transition between molecular and continuum transport within the void spaces. In this pressure regime, thermal properties are significantly affected by the gas pressure as the gas plays an important role in the bed heat transfer. The high-pressure regime is characterized by continuum gas transport within the void space and thus

© 2008, Woodhead Publishing Limited

88

Solid-state hydrogen storage

pressure has no further influence on the thermal properties of the bed. The inflection point between transition and continuum transport is called the critical pressure. The boundaries of each of these regimes depend on the characteristic void space or pore size of the bed. The intermediate- and highpressure regimes above 1 atm are of specific interest for automotive hydridebased hydrogen storage systems. There are a total of six modes of heat transfer within the bulk metal hydride packed particle structure as described in Oi et al. [9]: 1. 2. 3. 4. 5. 6.

Heat Heat Heat Heat Heat Heat

conduction at the particle contacts. conduction through a thin hydrogen ‘film’. radiation between the particles. conduction through the particles. conduction through hydrogen in larger void spaces. radiation between vacant spaces.

Models have been constructed describing each of these heat transfer mechanisms. Yagi and Kunii [10] developed generalized resistance models for packed beds which others adapted for application to metal hydride beds [9, 11–13]. For lower and moderate temperature applications of these models, radiation heat transfer can be neglected [9, 11, 12, 14, 15]. In general, the resistance model of effective thermal conductivity of a packed metal hydride bed can be described as:

1 – φv  K eff = K H 2 φv +  K   H 2 γ+ 2     3  K p   

[4.2]

where K H 2 is the thermal conductivity of the hydrogen gas, Kp is the thermal conductivity of the metal hydride solid, φ v is the void fraction of the bed. The term γ is a constant defined as the ratio of the effective length of the solid particle relating to conduction to the average hydride particle diameter. This ratio is difficult to quantify and is a function of particle contact angle, shape, and roughness. For metal hydrides, typical values fall between 0.01 and 0.1 Figure 5 in [16]. Using this resistance model, Oi et al. [9] predicted the thermal conductivity of a typical metal hydride bed as a function of metal hydride particle thermal conductivity for various porosities. In general, resistance models predict that metal hydride beds exhibit thermal conductivities between 0.5 and 2 W/m K for typical porosities and particle thermal conductivities. Resistance models are useful for making quick estimations of packed bed effective thermal conductivities, yet they tend to be highly subjective and, in practice, lack accuracy. Other types of models have been developed including numeric methods [17]. The most useful and accurate model was developed

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

89

by Zehner, Bauer, and Schlünder and reported in Tsotsas and Martin [18] and adapted by Rodriguez Sanchez et al. [19] which consists of a unit cell of two particle halves of equivalent shape encased in a cylinder of fluid (hydrogen). The model presented here neglects radiation and assumes spherical particles. The effective thermal conductivity of the bed is calculated by:

K eff = K H 2

1 (1– 1 – Ψ ) ⋅ Ψ ⋅    Ψ–1+ 1 KH2      + 1 – Ψ ⋅ [ϕ ⋅ K p + K pgp ⋅ (1 – ϕ )]

[4.3]

In this expression, K H 2 is the thermal conductivity of hydrogen, Ψ is the void fraction, φ is a flattening coefficient which defines contact quality, Kp is the particle thermal conductivity, and Kpgp is an expression for the thermal conductivity at the particle–gas–particle interface and includes particle diameter Dp. The effective thermal conductivity is highly influenced by Kpgp as it describes the contribution of the fluid to the particle thermal contact quality. For a complete description of the model refer to Rodriguez Sanchez et al. [19]. Using the Rodriguez Sanchez model, effective thermal conductivities (Kth) can be calculated as a function of hydrogen pressure, particle thermal conductivity, particle diameter, and void fraction (Fig. 4.4). Ranges selected for these variables were chosen based on typical metal hydride packed bed characteristics. Other model inputs include the deformation factor and contact flattening coefficient of the particle – indicating the area of contact between the particles. These factors are difficult to calculate or predict and are usually estimated from experimental results. Overall, the effective conductivity of the bed is less sensitive to the deformation factor and contact flattening coefficient compared with the other variables. For these calculations, typical values of these coefficients were used as in Rodriguez Sanchez et al. [19]. Effective thermal conductivity as a function of hydrogen pressure and void fraction is described in Fig. 4.4(a). In general, higher pressures and lower void fractions lead to higher conductivities up to the critical pressure. The critical pressure for these calculated geometries is near 1 × 108 Pa as observed by a reduction in rate of effective thermal conductivity increase with pressure. The effective thermal conductivity as a function of particle thermal conductivity is shown in Fig. 4.4(b). In general, particle thermal conductivities above 50 W/mK do not significantly influence bed effective thermal conductivity due to dominance in this regime by porosity and the particle to particle thermal contact. The effective thermal conductivity as a function of particle diameter is presented in Fig. 4.4(c). Larger particle diameters lead to higher effective conductivities due to the reduction of particle to

© 2008, Woodhead Publishing Limited

90

Solid-state hydrogen storage 4

Dp = 15 µm Kp = 100 W/m K 4

2

0 1×105 1×106 1×107 1×108 Hydrogen pressure (Pa) (a)

Effective K th (W/m K)

Effective K th (W/m K)

6

2

Dp = 15 µm pH2 = 10.3 MPa 0

100 200 Particle Kth (W/m K) (b)

Effective K th (W/m K)

6

pH2 = 10.3 MPa Kp = 100 W/m K 4

Void fraction = 0.3 Void fraction = 0.4

2

Void fraction = 0.5

0 1×10–6 1×10–5 1×10–4 Particle diameter (m) (c)

4.4 Calculated effective thermal conductivity of a packed particle bed as a function of void fraction, hydrogen pressure, and particle characteristics.

particle contacts and an increase in continuum-regime gas transport relative to transition-regime gas transport. The calculations presented here are consistent with many models and measurements described in the literature [11–14, 17, 20, 21]. Models and measurements indicate that the effective thermal conductivity of particles loaded in a packed bed is generally limited to values below ~5 W/m K, even with significant increases in the particle thermal conductivity (Fig. 4.4(b)). More clever methods must be employed to enhance thermal conductivity to levels above 5 W/m K. Additionally, the models discussed above have been developed for distinct particles typical of classic/interstitial hydride materials. These classic/interstitial beds are generally characterized as unsintered powders while complex hydrides, such as sodium alanates, can become porous sintered solids as seen in Fig. 4.5. Application of packed particle models have not been directly applied to sintered solid materials.

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

91

10 µm 2000× (a)

(b)

4.5 (a) The sintered solid resulting from hydrogen cycling of sodium alanates. (b) A scanning electron microscope (SEM) image of the sintered solid.

Measurements at Sandia National Laboratories, California, of sodium alanate sintered solids have demonstrated that the effective conductivities are similar to packed particle beds [22]. The thermal conductivities of stoichiometric sodium alanates compacted at 40% of the single crystal density were found to vary between 0.5 and 1.0 W/m K depending on cycle, hydrogen content, and gas pressure. Effective thermal conductivities as a function of gas pressure for fully cycled stoichiometric sodium alanates are shown in Fig. 4.6. As with packed particle beds, low thermal conductivities of complex hydrides such as sodium alanates present an engineering challenge when integrated within a system. Although the physical appearance of a sintered sodium alanate particle is dissimilar to a bed of close packed spheres, the thermal transport behaviors of both cases are similar. Both cases contain a characteristic thermal path length that influences the thermal transport within the bed when compared to the mean free path of the gas. Additionally, the sintered sodium alanate bed can be modeled as spheres (or islands) of material with improved thermal contact between each sphere or ‘island’. Given the similarities in fundamental mechanisms of heat transport of each case, we can assume that many of the same thermal conductivity enhancement strategies will work for both a packed particle and a sintered solid bed. Using our understanding of heat transfer within metal hydride beds, the following section will explore heat transfer enhancement mechanisms that can be exploited to improve the overall performance of metal hydride systems.

© 2008, Woodhead Publishing Limited

92

Solid-state hydrogen storage NaH + Al (H2)

1.00

NaH + AI

Kth (W/m K)

0.80

Na3AIH6

0.60

NaAIH4

0.40 Characteristic equations NaH + Al: Kth = 0.068 ln(PH2) + 0.71 Na3AIH6: Kth = 0.061 ln(PH2) + 0.50 NaAIH4: Kth = 0.037 ln(PH2) + 0.51

0.20

0.00 1

10 Pressure (atm)

100

4.6 Effective thermal conductivity (Keff) as a function of gas pressure for fully cycled stoichiometric sodium alanates [22].

4.4.2

Thermal properties enhancement methods for solid storage materials

Several authors have described the importance of thermal conductivity enhancement to enable improved performance of metal hydride-based hydrogen storage systems [23–25]. Lower thermal resistances enable the use of larger thermal length scales and an increased rate of hydrogen uptake. Gas flow Flow through the porous bed enhances the radial effective or apparent thermal conductivity of packed beds [10, 26]. Winterberg and Tsotsas [26] developed models and heat transfer coefficients for packed spherical particle reactors that are invariant with the bed-to-particle diameter ratio. The radial effective thermal conductivity is defined as the summation of the thermal transport of the packed bed and the thermal dispersion caused by fluid flow, or: Kbed + flow = Kbed + Kflow

= K bed + X1 ⋅ Pe 0 ⋅

uc ⋅ f ( r ) ⋅ K gas uave

[4.4]

The coefficient X1 is a correlation function that describes the rate of increase of the effective thermal conductivity with flow velocity, Pe0 is the Péclet number, which describes the contribution of forced convection relative to hydrogen heat conduction, uc is the velocity at the centerline of the bed, uave is the average velocity, f(r) describes the radial variation in dispersion, and

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

93

Kgas is the thermal conductivity of the fluid (hydrogen). In many cases, uc and uave are assumed to be near equivalent due to the large vessel-to-particle diameter ratios. The radial dispersion variation f(r) is assumed to be unity for similar reasons. For a complete description of the model and correlations, refer to Winterberg and Tsotsas [26]. Although the model is promoted to be invariable to the vessel diameter to particle diameter ratio, the correlations were built using experimental data with ratios from 5.5 to 65. High conductivity coatings Researchers have experimented with enhancing heat transport by coating metal hydride pellets with high-conductivity, high-ductility metals. Copper has been frequently used for this purpose [27–30]. Kim et al. [28] reported conductivities as high as 9 W/m K for higher packing densities of 25% by weight copper-coated LaNi5 powders (thermal conductivity of uncoated powders typically less than 1 W/m K). This corresponds well with results published by Kurosaki et al. [27], which describes effective thermal conductivity of LaNi5 as a function of copper mass percent. This approach may not be the most favorable method of thermal properties enhancement as significant copper mass is required to attain moderate increases in thermal conductivity. Additionally, gas and solid mass transport may be adversely affected when this technology is applied to complex hydrides. High thermal conductivity composite alloying Various high conductivity materials have been alloyed with metal hydrides to form enhanced heat transport composite materials. Eaton et al. [31] experimented with various alloyed metal additives including copper, aluminum, lead, and lead–tin. The samples were alloyed at elevated temperature (200– 600 °C) and cycled. In many samples, cycling resulted in the separation and fracture of the alloy and thus a reduction in composite thermal conductivity. Sintered aluminum structures of 20% solid fraction have been integrated with LaNi5 hydride materials with success, resulting in effective thermal conductivities of 10–33 W/m K [32–34]. Temperatures required for this process and added mass and volume may exclude application to some complex hydrides. High thermal conductivity structures High thermal conductivity structures have been used to enhance thermal conductivity including copper wire matrices [35], periodic plates [25], nickel foams [30], and aluminum foams [36]. Nagel et al. [35] designed and integrated a 90% porous corrugated copper wire matrix with a MmMi4.46Al0.54 hydride

© 2008, Woodhead Publishing Limited

94

Solid-state hydrogen storage

bed and improved the overall conductivity by a modest 15%. Aluminum and nickel foams have been used with some success [30]. Practically, metal foams tend to be costly and are a significant challenge to integrate with metal hydrides without the creation of high void fractions. High thermal conductivity additives Expanded natural graphite fibers have been described as appropriate additives due to their characteristic high thermal conductivity, porosity, dispersibility, and low cost [19,38,39]. Expanded natural graphite (ENG) fibers are produced from natural graphite that is treated in sulfuric acid and heated to high temperatures, thereby expanding to very fine flakes. The thermal properties of ENG fibers are highly anisotropic; in-plane thermal conductivities in excess of 500 W/m K with through-plane conductivities of less than 5 W/m K. Effective thermal conductivities as high as 20 W/m K are theoretically attainable for classic metal hydrides combined with 10% volumetric fraction of ENG fibers assuming 100% fiber alignment with the direction of heat transfer. Experimentally, composite thermal conductivities as high as 10 W/ m K were measured with mass fractions as low as 5% although these results were obtained with very small sample sizes and presented with significant measurement scatter [19]. The measurement of larger composite samples may result in somewhat lower thermal conductivities (as much as half) due to randomization of the ENG fiber orientation. Many options exist for the enhancement of the thermal properties of reversible hydrogen storage materials. The optimum method must be chosen based on the optimization of the total system energy densities, hydrogen uptake and delivery performance, and cost targets.

4.5

System heat exchange design

The combined influence of the thermodynamic, chemical/kinetic, physical properties, and application environment define the heat exchange requirements for the entire system. With this information the system designer is able to evaluate various methods for heat transfer within the system to obtain an overall optimized design based on the end-use requirements.

4.5.1

Absorption heat exchange

The most significant challenge arises from removing the heat of absorption as discussed in earlier sections. Many standard heat exchanger designs are viable; however, the optimal solution depends on the application. Heat exchange systems include but are not limited to the designs listed below. In these examples, ‘vessel’ refers to the hydride-containing pressure vessel.

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design









95

Internally finned vessels: Fins enhance heat transfer out of the containment vessel and thus allow for more monolithic tank designs. Monolithic tank designs are generally less complex systems, which is attractive for automotive applications. Attaining adequate heat transfer without adding excessive weight and volume can be challenging with finned vessels. Vessels in cross or axial flow: Vessels distributed within a fluid flow field can improve system heat transfer. The distributed nature of the vessels can result in increased system complexity and a reduction in volumetric and gravimetric efficiency. Internal flow cooled vessel: Designs that incorporate internal flow provide similar benefits of finned vessels by allowing for monolithic tank designs. Some clever engineering may be required to allow for penetration of the containment vessel with the cooling plumbing. Fluidized beds: Fluidization of the bed can result in enhanced heat and mass transfer performance. Unfortunately, many fluidized bed designs tend to be voluminous, which reduces overall storage efficiency.

This is not a comprehensive list of examples, and the designer may determine that a hybrid, or combinatorial, solution is appropriate, depending on the application.

4.5.2

Desorption heat exchange

Although less thermally rigorous, delivery of hydrogen to the fuel cell requires the transfer of heat into the bed to support the endothermic reaction. If the quality of available waste heat in the system from a fuel cell stack or other energy conversion device is higher than the quality of heat required to support the endothermic reaction, this heat can be transferred to the hydride at low cost. On the other hand, if the required quality of heat is greater than that of the available waste stream, then auxiliary heating will be needed to support the endothermic reaction. Depending on the heat of reaction, this can contribute a significant loss to the overall system efficiency. For example a ~20% loss to efficiency can be expected for a 40 kJ/mol material with an 80% efficient auxiliary heater. If resistive heating is used in conjunction with a fuel cell, losses can begin to overwhelm the benefit of using phase-change materials in place of lower energy density solutions such as compressed gas. It should be noted that the quality of the waste heat must be significantly higher than that of the required heat to enable heat transfer. Heat flux is defined as HA∆T, where H is the heat transfer coefficient in W/m2 K, A is the active surface area in m2, ∆T is the temperature difference in K between the available and required temperatures. Thus, if ∆T is very small, the heat flux is similarly minimized and will not allow for sufficient hydrogen fuel delivery to the conversion device.

© 2008, Woodhead Publishing Limited

96

Solid-state hydrogen storage

Identification of the most efficient design for a combination of material properties, fuel demand requirements, and operating environment parameters, requires the consideration of all operational modes including hydrogen refueling and hydrogen delivery. No single storage system design can be recommended without a detailed understanding of the material properties and application requirements.

4.6

Safe systems design

Although hydrogen uptake and delivery properties of storage systems are critical to the success of solid-state hydrogen storage systems, operator and consumer safety is of highest importance of all engineering-related properties. Safety properties are defined as the potential health and safety hazards realized during both abnormal scenarios (such as accidents) and also normal operation over the entire life cycle of the system. These properties have direct implications on the health and well-being of the vehicle worker and operator. Considering the automotive platform, approximately 12 million new motor vehicles are produced every year within the United States alone. Owing to the technology usage density and diverse user demographic, robust and safe systems must be designed and demonstrated. By definition, high-efficiency energy storage systems are capable of releasing significant amounts of the stored energy and present a potential hazard. Although it is improbable that this hazard can be eliminated entirely, engineering controls and appropriate protocols can minimize the potential hazard to a tolerable level. In the case of metal hydrides, the reaction processes with contaminants, especially air and water vapor, must be understood to enable the implementation of engineering controls and the definition of appropriate protocols. Another safety-related property, volumetric expansion of the solid materials during normal hydrogen sorption, is also an important property that must be considered during the design of solid-state hydrogen storage systems.

4.6.1

Volume expansion and decrepitation

Volumetric expansion-induced pressure due to expansion of the storage material during hydrogen sorption is a classic metal hydride safety system issue. This pressure originates from expansion of the crystal structure during hydrogen sorption and can be exacerbated by a physical process called decrepitation. Decrepitation is the systematic physical breakdown of larger particles into smaller fines due to high strain induced by hydrogen cycling [39]. Following decrepitation, the fines tend to settle under the influence of gravity and/or vibration, thus causing a density gradient to form. Upon subsequent cycling, these high-density sections of the bed expand. Owing to a lack of void space in these regions, the expansion is directed outward, exerting large forces on

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

97

the wall of the containment vessel. In extreme cases, structural failures of the containment vessel can occur, causing a release of hydrogen gas and exposure of the reactive storage materials to the surrounding environment. Every hydrogen sorption material will behave differently with respect to expansion forces. Some materials will not establish large density gradients due to sintering, while others will experience significant decrepitation events with just a few cycles. This property disparity requires the systems designer to characterize each material as a function of sorption cycle by monitoring the strain in the containment vessel wall. Nasako and colleagues have described methods for measuring stress and strain in a hydride-containing vessel wall [40]. The strain measurement of a hydride-containing vessel wall can be difficult for two reasons: (1) the temperature excursions intrinsic to the uptake and release of hydrogen can cause significant strain measurement errors, and (2) the strain induced by the gas pressure can complicate the measurement of the mechanical expansion-induced strain. In practice, these difficulties can be overcome with clever instrumentation, including temperature compensation and gas pressure-induced signal calibration.

4.6.2

Contamination of metal hydrides – oxygen and water reactivity

Another important set of safety properties can be described as the health, safety, and performance effects of the products, pathways, and rates of reactions occurring between hydrogen storage materials and their potential contaminants. The automotive platform is a dynamic hydrogen storage application subject to a highly variable operating environment, uncertain hydrogen refueling quality control, and accidents, such as collisions and externally fueled fires. The nature of this duty cycle forces the system designer to consider that the storage materials or containment structures may be somehow compromised during the lifetime of the hydrogen storage system. In an accident or contaminated refueling scenario, the hydride material may be exposed to air, water, and other contamination. Owing to the requirements for high-energy density hydrogen storage solutions, hydrides (complexes) comprising light metal elements are of specific interest – including, but not limited to, lithium (Li), sodium (Na), magnesium (Mg), calcium (Ca), boron (B), and aluminum (Al). Hydrides and complexes synthesized from these elements are typically found in the form of finely divided powders that can be pyrophoric and water reactive. Additionally, the oxidation reaction products may also present hazards. In the case of alkali metal-based materials, the oxidation products may include materials that present health hazards or form hydrates in a humid environment that can decompose rapidly as a result of friction or heat. The associated chemical pathways and reaction rates must be quantified to enable the design and implementation of safe hydride-based automotive

© 2008, Woodhead Publishing Limited

98

Solid-state hydrogen storage

hydrogen storage systems. In addition, chemical hazard mitigation methods must be developed to enable the safe handling of hydride materials during the entire life cycle of the storage system. In this section, we will explore the safety properties of a specific set of complex hydrides, sodium alanates. These materials represent the most advanced solid-state hydrogen system developed to date and allow us to understand the safety properties associated with this class of materials.

4.6.3

Sodium alanates: prototypical complex metal hydrides

Sodium aluminum tetrahydride (NaAlH4) is a complex hydride historically used as a reducing agent in chemical synthesis, which was found to reversibly release and accept hydrogen in a two-step phase change process in 1997. This discovery of destabilization by titanium doping made sodium alanates (NaAlH4, Na3AlH6) promising materials for hydrogen storage systems [41,42]. The thermochemistry characteristics of sodium alanates This complex hydride absorbs and releases hydrogen in a two-step decomposition and recombination reaction shown in the following reaction: NaAlH4 s 1/3Na3AlH6 + 2/3Al + H2 s NaH + Al + 3/2H2 Sodium alanate-based systems are currently under development for use as high energy density hydrogen storage solutions [4,5]. Reversible hydrogen storage materials are characterized by endothermic decomposition and exothermic recombination and the thermodynamics of these reactions are well characterized [43,44]. It should be noted that sodium alanates will not meet the current US DOE goals for hydrogen storage performance; however, similar higher energy density complexes are actively being investigated that may have structural and chemical reactivity characteristics similar to that of sodium alanates. Sodium alanates are similar to traditional metal hydrides in the fact that they are typically found as finely divided metal powders and undergo significant morphological changes during hydrogen sorption. Figure 4.7 illustrates this morphology change using SEM imaging of the highly crystalline sodium aluminum tetrahydride (hydrogen charged phase, A) compared to the highly porous form of the fully decomposed material (hydrogen released phase, B). The sodium alanates are high surface area materials that are pyrophoric and react readily with water. Intuitively, it would be expected that the system exothermically reacts to form NaOH and Al2O3 upon exposure to an oxidizing species. Despite the relatively benign oxidation products, some researchers have observed samples becoming friction sensitive after uncontrolled oxygen

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

A

99

B 5 µm 5000×

5 µm 5000×

4.7 SEM images of the highly crystalline sodium aluminum tetrahydride (A) compared with the highly porous form of the fully decomposed material (B).

exposure. Researchers as early as Ashby in 1969 [45] reported sensitivities of oxidized alanates to friction. This behavior could indicate that other reactions could occur, perhaps resulting in oxides such as sodium superoxide as determined by Desreumaux [46]. The oxidation reactions of these materials are not fully understood – which causes difficulties when designing safe and effective controls for systems. Since large quantities of reactive sorption materials must be integrated into the automotive platform, the hazards associated with contamination and environmental exposure must be fully quantified and understood so appropriate codes and standards can be specified and engineering controls can be designed and integrated. A large number of oxidation reactions are possible between sodium alanates and the oxidizing gases present in the atmosphere (oxygen and water vapor). Many reactions are thermodynamically favorable as shown by free energy and enthalpy changes of a few selected reactions in Table 4.1. The thermodynamic property data were obtained from Barin et al. [47]. Reactions that yield sodium oxides are of particular interest as some of these compounds can exhibit structural instability. The oxidation of sodium alanates by oxygen to yield various Na–O compounds and either hydrogen or water vapor are reactions that are particularly interesting. Oxidation to sodium oxides such as peroxide, Na2O2, and superoxide, NaO2, are both favorable according to the free energy changes although less so than to sodium oxide, Na2O. The water vapor product is capable of further oxidizing either the alanate or some of the oxidation products of the initial reactions.

© 2008, Woodhead Publishing Limited

100

Solid-state hydrogen storage

Table 4.1 Thermochemistry and volumetric changes of a selection of oxidation reactions of sodium alanate by oxygen or water vapor. Free energy and enthalpy changes are given per mol of NaAlH4. Volumetric changes refer to the solid compounds only and are percentages based on NaAlH4 ∆G (kcal/mol)

∆H (kcal/mol)

–22.59

–25.70

–10.4

Oxidation by O2 to Na–O compounds – H2O product O2 + 2NaAlH4 → Na2O + 2AlH3 + H2O –49.98

–54.68

–10.4

Proposed processes (abbreviated list)

Oxidation by O2 to Na–O compounds – H2 product 1 /2 O2 + 2NaAlH4 → Na2O + 2AlH3 + H2

∆V (%)

Oxidation to Al2O3 as one product 4O2 + 2NaAlH4 → Na2O + Al2O3 + 4H2O 4H2O + 2NaAlH4 → Na2O + Al2O3 + 8H2

–332.14 –112.99

–338.85 –107.05

–19.0 –19.0

Oxidation to NaOH 4O2 + 2NaAlH4 → 2NaOH + Al2O3 + 3H2O H2O + NaAlH4 → NaOH + AlH3 + H2

–350.71 –13.76

–362.23 –20.10

–12.7 –8.3

Calculations of volumetric changes of the solid compounds associated with these reactions were also performed to determine potential effects on the bed, such as consolidation or unpacking, or expansion that might affect the container. These results were obtained from published values of the bulk density of the various solid oxidation products and data for the phases of sodium alanates [47]. The volumetric calculations are summarized in Table 4.1. The volumetric changes are percentages based on the presence of only NaAlH4 initially. In general, significant volumetric changes are not realized during bed oxidation, with the prevailing trend toward volumetric reduction. System implications of the oxidation reactions Since the reactions described in Table 4.1 are exothermic, the rate of the oxidation processes must be controlled to avoid rapid heating and, possibly, ignition of the metal hydride. In the case of the thermodynamically aggressive postulated process: 4O2 + 2NaAlH4 → 2NaOH + Al2O3 + 3H2O the heat of oxidation reaction is –1516 kJ/mol NaAlH4. For a sodium alanatebased system that reversibly stores 5 kg of hydrogen, the total quantity of heat removal required is approximately 5 GJ – the amount of energy released when burning 38 kg of hydrogen fuel. If the majority of this heat is not removed, it is plausible that ignition can occur. In a postulated event in which the oxidation process is active for 4 hours, the average heat rate is 333 kW. Since this is a significant amount of energy and power to manage during

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design

101

an automotive accident scenario, controls must be in place to limit the impact of such an oxidation event to avoid ignition. In addition to the ignition and fire hazard posed by the oxidation of the metal hydrides, other safety factors must also be considered. The oxidation reaction products may be unstable, toxic, or otherwise hazardous. For example, the reactions described in Table 4.1 that result in sodium superoxide could result in the production of friction-sensitive materials and may add risk to the disassembly of handling contaminated hydrogen storage materials.

4.7

Enabling safe systems based on hydrogen sorption materials

Although sodium alanates have been the focus of this discussion, other chemically similar materials will likely have very similar safety hazards. Given thermochemistry considerations of complex metal hydrides, some effort is required to address the contamination issues and demonstrate that hazards associated with the safe production and operation of the automotivescale complex hydride-based hydrogen storage systems can be effectively minimized. To enable the realization of safe systems based on reactive solids, the developers must identify and characterize chemical processes and hazards associated with hydride exposure to air, water vapor, and other relevant contaminants. These hazards could include human toxicity and environmental pollutants. From this information, system designers can predict system contamination scenarios and identify hazard mitigation strategies to allow for handling and disposal of contaminated materials. Eventually, full-scale testing using standardized methods will be required to obtain regulatory approval for consumer use of new hydrogen storage technologies.

4.8

Future trends

As novel hydrogen storage materials are developed and become better understood, engineers will be responsible with developing safe, efficient, and application-relevant systems. As energy densities of the materials increase, system design becomes more challenging as more heat must be transferred per unit volume. Fortunately, as we described earlier in this chapter, we already have a general understanding of the physical and thermodynamic characteristics of many potential solid-state hydrogen storage materials, which allows system designers to develop engineering techniques and methods to enable the utilization of this future technology. Although no single system design is appropriate for all hydrogen storage materials and applications, intelligent engineering and optimization efforts will enable safe, high energy density storage solutions based on solid sorption materials.

© 2008, Woodhead Publishing Limited

102

4.9

Solid-state hydrogen storage

Sources of further information and advice

Currently the US DOE is funding efforts to develop hydrogen storage systems based on complex metal hydrides. Annual review proceedings that discuss US DOE-funded hydrogen storage engineering efforts can be accessed at http://www.eere.energy.gov/. Although references on hydrogen storage systems are very limited, hydrogen storage system design is discussed in brief in Fuel Cell Systems Explained by James Larminie and Andrew Dicks (Wiley).

4.10

References

1. The Hydrogen, Fuel Cells & Infrastructure Technologies Program 2006 Multi-year Research Demonstration and Development Plan (MYRDDP) – available at http:// www1.eere.energy.gov/hydrogenandfuelcells/mypp/ 2. Meeting the Energy Needs of Future Warriors, Committee of Soldier Power/Energy Systems, National Research Council, (2004). 3. G. Sandrock, Journal of Alloys and Compounds, 293–295 (1999) 877–888. 4. B. Mason, GM seeks help from Sandia Labs. The Valley Times, 93(281) (2005). 5. D. Mosher et al. Proceedings of the 2006 US DOE Hydrogen Program Review, May 16–19, 2006, DE-FC36-02AL67610. 6. G. Sattler, Journal of Power Sources, 71 (1998) 144. 7. G. Sattler, Journal of Power Sources, 86 (2000) 61–67. 8. A. Reller, Soilde Tatsachen Schaffen, ENET news, March 2004, Energy Switzerland. 9. T. Oi et al., Journal of Power Sources, 125 (2004) 52–61. 10. S. Yagi, S. Kunii, A.I.Ch.E Journal, 3(3) (1957) 373–381. 11. D. Sun, International Journal of Hydrogen Energy, 15(5) (1990) 331–336. 12. Y. Ishido, M. Kawamura, S. Ono, International Journal of Hydrogen Energy, 7(2) (1982) 173–182. 13. E. Suissa, I. Jacob, Z. Hadari, Journal of the Less-Common Metals, 104 (1984) 287– 295. 14. A. Isselhorst, Journal of Alloys and Compounds, 231 (1995) 871–879. 15. J. Kapischke, J. Hapke, Experimental Thermal and Fluid Science, 17 (1998) 347– 355. 16. D. Kunii, S. Smith, A.I.Ch.E Journal, 6(1) (1960) 71–79. 17. Y. Asakuma et al., International Journal of Hydrogen Energy, 29 (2004) 209–216. 18. E. Tsotsas, H. Martin, Chemical Engineering and Processing, 22 (1987) 19–37. 19. A. Rodriguez Sanchez et al., International Journal of Hydrogen Energy, 28 (2003) 515–527. 20. S. Suda, N. Kobayashi, International Journal of Hydrogen Energy, 6(5) (1981) 521– 528. 21. A. Kempf, W. Martin, International Journal of Hydrogen Energy, 11(2) (1986) 107– 116. 22. D.E. Dedrick et al., Journal of Alloys and Compounds, 389 (2005) 299–305. 23. M. Gopal, S. Murthy, Chemical Engineering and Processing, 32 (1993) 217–223. 24. M. Nagel, Y. Komazaki, S. Suda, Journal of the Less-Common Metals, 120 (1986) 35–43. 25. D. Sun, International Journal of Hydrogen Energy, 17(12) (1992) 945–949. 26. M. Winterberg, E. Tsotsas, Chemical Engineering Science, 55 (2000) 967–979.

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage system design 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47.

103

K. Kurosaki et al., Sensors and Actuators A, 113 (2004) 118–123. K.J. Kim et al., Powder Technology, 99 (1998) 40–45. H. Ishikawa et al., Journal of the Less-Common Metals, 120 (1986) 123–133. Y. Chen et al., International Journal of Hydrogen Energy, 28 (2003) 329–333. E. Eaton et al., International Journal of Hydrogen Energy, 6(6) (1981) 609–623. M. Ron et al., Journal of the Less-Common Metals, 74 (1980) 445–448. E. Bershadski et al., Journal of the Less-Common Metals, 153 (1989) 65–78. Y. Josephy et al., Journal of the Less-Common Metals, 104 (1984) 297–305. M. Nagel et al., Journal of the Less-Common Metals, 120 (1986) 35–43. W. Supper et al., Journal of the Less-Common Metals, 104 (1984) 219–286. K.J. Kim et al., International Journal of Hydrogen Energy, 26 (2001) 609–613. H.-P. Klein, M. Groll, International Journal of Hydrogen Energy, 29 (2004) 1503– 1511. B. Fultz, Journal of Alloys and Compounds, 335 (2002) 165–175. K. Nasako, Journal of Alloys and Compounds, 264 (1998) 271–276. B. Bogdanovic et al., Journal of Alloys and Compounds, 253–254 (1997) 1. B. Bogdanovic et al., Journal of Alloys and Compounds, 302 (2000) 36. G. Thomas et al., Proceedings of the 1999 US DOE Hydrogen Program Review, NREL/CP-570-26938. C. Jensen et al., Proceedings of the 1999 US DOE Hydrogen Program Review, NREL/CP-610-32405. E. Ashby, Chemical and Engineering News, 47(1) (1969). J. Desreumaux, European Journal of Inorganic Chemistry, (2000) 2031–2045. I. Barin, O. Knacke, O. Kubaschewski, Thermochemical Properties of Inorganic Substances, Springer, Berlin (1977).

© 2008, Woodhead Publishing Limited

Part II Analysing hydrogen interactions

105 © 2008, Woodhead Publishing Limited

5 Structural characterisation of hydride materials B. C. H A U B A C K, Institute for Energy Technology, Norway

5.1

Introduction

In order to understand the properties of materials and also to be able to determine new compounds, detailed knowledge about the position of the atoms is of major importance. When Friedrich et al. in 1912 [1] found the direct relationship between the observed diffraction pattern when X-rays irradiated a crystalline solid compound and the corresponding arrangement of atoms in the structure, the interdisciplinary field of crystallography started to develop. Today crystallography is very important in many areas of science, for example related to health, energy and information technology. Sources for diffraction studies are X-rays, neutrons and electrons, and each of the methods has specific properties and challenges. For studying hydrogen storage materials both neutron and X-ray diffraction are important, but in particular neutron scattering techniques play a crucial role because of the significant scattering from hydrogen atoms. The field of crystallography and diffraction studies of crystalline materials has developed significantly during the past 10–20 years. This is both due to improved sources for X-rays and neutrons and because of significant improvements in the tools used to analyse diffraction data. Since hydrides normally exist as powdered samples, powder diffraction methods, with both X-rays and neutrons, are the main methods of finding their atomic arrangements. However, powder diffraction is also important to determine information about phase compositions, crystallite sizes, strain and defects. In this chapter diffraction and in particular X-ray and neutron diffraction will be described in general, with an emphasis on powder diffraction techniques. The specific properties of X-ray and neutron diffraction and a description of sources and instruments for powder diffraction studies will be presented. Furthermore the use of powder diffraction data, from the simple use for phase identification to structure solution and refinements with the Rietveld methods, will be described. Two examples showing the potential for powder 107 © 2008, Woodhead Publishing Limited

108

Solid-state hydrogen storage

diffraction will be followed by a description of future trends and sources of further information.

5.2

Principles of diffraction

Diffraction is the phenomenon associated with the interference processes that occur when electromagnetic radiation is scattered by the atoms in solids, liquids and gases. For electromagnetic radiation with a wavelength comparable to interatomic distances in the substance, the resulting scattered waves will be a result of the atomic arrangements. The measured scattering data will therefore include information about the positions and distances between the atoms in the compound. X-rays and neutrons (from a research reactor or a spallation source) have wavelengths similar to typical interatomic distances, approximately 1Å (0.1nm). Thus, X-ray and neutron diffraction are important techniques for the determination of atomic arrangements in materials. In general the relationship between the recordable diffraction pattern and the atomic arrangement is complicated, and mathematically this is described by Fourier transformations. A crystalline material consists of a long-range order arrangement of atoms in three-dimensional space. It is the most likely arrangement of atoms since it corresponds to the lowest energy states. The repeating unit is called the unit cell. The unit cell is defined as the smallest unit that includes the full symmetry of the crystal structure. The overall structure of a crystalline material can therefore be described by the content of atoms and their individual arrangements in the individual unit cells and the size and shape of the unit cells. In practice the unit cell contains more than one molecule or group of atoms that are converted to each other by symmetry operations. The smallest part in the unit cell that will generate the whole cell is called the asymmetric unit. There are seven possible unit cell shapes, giving the seven crystal systems: cubic, tetragonal, orthorhombic, monoclinic, triclinic, hexagonal and trigonal. The most general unit cell is the triclinic system defined by the a, b and c axes (usually given in Å or nm) and the angles α, β and γ (in degrees). The different ways (from group theory) of combining the types of lattices and symmetry elements give the 230 space groups. It should be added that many materials, including hydrogen storage materials, are amorphous, and thus without long-range order. In these cases so-called total scattering experiments can give information about the local ordering, but such experiments will not be covered here. Scattering from crystalline materials irradiated by X-rays or neutrons will be strong in some directions (constructive interference) and much less (almost perfect destructive interference) in other directions. The directions for the constructive interference are determined from the size and shape of the unit cells. The intensities of the constructive interference peaks are given by the

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

109

atomic arrangement within the unit cells. A simple and elegant way of describing the conditions for constructive interference was presented by W. L. Bragg in 1913, resulting in the famous Bragg’s law [2]. His approach was to regard crystals as being built up of crystallographic planes and then to explain the diffraction as the consequence of subsequent reflections from succeeding planes belonging to the same family (physically, this involves scattering from the atoms lying on these planes). The lattice planes are labelled by three numbers, hkl, denoted as the Miller indices. The hkl numbers are derived from how many times the planes cut the a, b and c axes, respectively, and then given by the reciprocal of these fractions. Figure 5.1 illustrates Bragg’s law. Two X-ray beams are reflected from adjacent planes. To obtain constructive interference the difference in the travelling distance from the two planes has to be equal to an integer number of wavelengths. Mathematically Bragg’s law then becomes (from simple geometry): 2dhkl sin θhkl = nλ

[5.1]

where dhkl is the spacing between lattice planes with indices hkl, θhkl is half of the scattering angle 2θ between the incoming and scattered beams, n is an integer and λ is the wavelength of the incoming beam. For crystalline materials scattering should then be observed only in the directions (in three dimensions) given by Bragg’s law. In Eq. (5.1) the integer n can be omitted since dhkl = nd nh,nk,nl . Even though this simple derivation of Bragg’s law is an oversimplification of a complex scattering process, it has been shown that rigorous mathematical treatments of the complicated interactions between X-rays/neutrons and atoms give the same answer. Another way to visualise the phenomenon of diffraction is to introduce the concept of the so-called reciprocal lattice. Mathematically, diffraction involves Fourier transformations between the physical arrangement of atoms and the resulting intensities observed in directions determined by Bragg’s

θhkl

dhkl

2θhkl

5.1 Illustration of Bragg’s law. The difference in path travelled by the two beams is equal to 2dhkl sin θhkl .

© 2008, Woodhead Publishing Limited

110

Solid-state hydrogen storage

law. Distances in real space, e.g. unit cell dimensions a, b, c, and distances between lattice planes dhkl appear as so-called reciprocal lattice vectors in * = h a* + k b* + l c*, respectively. For the reciprocal space, a*, b*, c* and d hkl * | = | a* | = 1 a and d hkl = a / ( h 2 + k 2 + l 2 ) . example for a cubic crystal | d100 The corresponding scattering from hkl planes will be in well-defined directions determined from Bragg’s law in the three-dimentional space and the positions of the diffraction peaks are given solely from the axis and the angles of the unit cell. For a single crystal which by definition consists of a single crystallite with the same orientation and periodicity of the unit cells throughout the whole crystal, the scattering corresponding to lattice planes hkl will be in unique directions in the three-dimensional space. However, many compounds, for instance most metal hydride materials, exist only in the form of polycrystalline or powder compounds consisting of many differently oriented crystals. An ideal powder is defined as a compound with an infinite number of crystallites with random orientation. In this case the orientations of the reciprocal lattice vectors are random, but their angle with the incoming beam is constant. The left side of Fig. 5.2 illustrates an example of a diffraction pattern on a flat plate detector. Each ring, often called a Debye ring, corresponds to scattering from sets of hkl planes with the same 2θ angle. The usual representation of a powder diffraction pattern or a diagram is a plot showing measured intensity as a function of the scattering angle 2θ; an example is shown in the right side of Fig. 5.2. This can be obtained by integration along the rings in the left hand side. However, for most powder diffraction experiments intensities are measured as a function of scattering angle with a point detector or a linear detector. The powder diagram can be considered as a one-dimensional representation of the three-dimensional diffraction pattern obtained from a single crystal. Significant challenges are related to overlap of reflections with equal or nearly equal Bragg angles. For single-crystal diffraction the scattering will be in different directions and separate intensities can be collected; for powder diffraction such reflections with equal or similar Bragg angles will overlap. The intensities of the Bragg diffracted beams or reflections, Ihkl, are determined by the atomic arrangement within the unit cells. This gives the relative positions of the atoms and thereby also information about bonding, distances and angles between the individual atoms in the crystalline compound. Mathematically this can be described as: Ihkl ∝ | Fhkl |2

[5.2]

where the structure factor Fhkl is defined by: N

Fhkl = Σ f j exp[2 π i ( hx j + ky j + lz j )] exp (– B j sin 2 θ / λ 2 ) j =1

© 2008, Woodhead Publishing Limited

[5.3]

0

500

1000

1500

2000

40 000 2000 35 000

1500

Rows

1500

1000

1000

Intensity (arb. units)

30 000 25 000 20 000 15 000 10 000 5000 500

500

0 0

0

0

500

1000 1500 Columns

2000

5

10

15

20 2θ (°)

25

0

5.2 Left: Illustrated diffraction pattern measured with a two-dimensional MAR345 image plate. Right: The onedimensional representation of the same diagram.

30

35

Structural characterisation of hydride materials

2000

111

© 2008, Woodhead Publishing Limited

112

Solid-state hydrogen storage

where fj is the atomic scattering factor that is strongly dependent on both the type of radiation and the elements. For neutrons it can also strongly vary with different isotopes. N is the number of atoms in the unit cell, xj, yj, zj are the fraction coordinates of atom j in the unit cell, Bj is the isotropic displacement parameter of atom j defined as B j = 8π 2 ( u 2 ) j , ( u 2 ) j is the root mean square deviation of atom j from its equilibrium position in Å2. In general the displacement parameters are anisotropic, but in most cases isotropic parameters are used for powder diffraction data. From known atomic coordinates the scattered intensities for every hkl can be calculated from Eqs (5.2) and (5.3), and thus the calculated intensity diagram can be simulated. However, the opposite way, to determine the atomic coordinates from measured intensities is not straightforward since the information about the phases and thus the atomic coordinates is lost in the diffraction experiment. This is the so-called phase problem in crystallography. However, several methods have been developed to overcome this problem; see below.

5.3

X-ray and neutron diffraction

X-rays have wavelengths in the range 0.1–100 Å. The wavelengths used for crystallographic studies are typically in the range 0.5–2.5 Å. Because of the availability and properties X-ray diffraction is the most common technique for phase identification and structural characterization of crystalline materials. X-rays are scattered by the electrons of the atoms. Important consequences of this are that: • • •

The atomic scattering factor, fj in Eq. (5.3), increases linearly with the atomic number Z; The X-ray scattering factors have a significant dependence on the scattering angle, the maximum value (equal to the number of electrons) is at zero scattering angle and decreases with increasing scattering angles (Fig. 5.3); X-rays interact strongly with matter.

The neutron is an uncharged particle found in all atomic nuclei except the H nucleus. From the de Broglie relation a particle with momentum p behaves as a wave with wavelength λ = h/p where h is the Planck’s constant. A similar relationship is valid for electrons. Neutrons with wavelength in the typical range 1–2.5 Å are available from nuclear reactors (see below). The scattering process with neutrons is different from X-rays. Neutrons are scattered by the atomic nuclei in a complex process. The atomic scattering factor is usually named the scattering length for neutrons. Important properties related to scattering by neutrons are as follows:

1



There is no correlation between the atomic number and the scattering length, as shown in Fig. 5.4. Even scattering from different isotopes of the same element can differ significantly.

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

113

45 40

Zr

fo atomic scattering factor

35 30 25 20 15

Ca

Al

10 5

B H

0 0.0

0.2

0.4

0.6 0.8 (sin θ )/λ (Å–1)

1.0

1.2

1.4

5.3 X-ray atomic scattering factors as a function of (sin θ) λ for some selected elements H, B, Al, Ca and Zr.

2.0

Scattering length (10–14 m)

1.5

1.0

0.5

0.0

–0.5 –1.0 0

10

20

30 40 50 Atomic number, Z

60

70

5.4 Neutron scattering lengths as a function of atomic number Z.

© 2008, Woodhead Publishing Limited

80

114

• •



Solid-state hydrogen storage

The neutron scattering lengths do not decrease or change with changing scattering angles. Neutrons interact usually weakly with matter (a few elements, including boron, have a strong absorption). An important effect of this is that complex sample environments such as furnaces, cryostats and pressure cells can easily be used since neutron beams penetrate the material used in such devices. Neutrons have a magnetic moment, and neutron scattering is the main method to determine magnetic ordering and magnetic structures.

A comparison of X-ray and neutron diffraction is presented in Table 5.1. The difference in the neutron scatting lengths between neighbouring elements and the overall nearly constant value going from light to heavy elements, gives two advantages with neutron diffraction: •



Neighbouring elements in the periodic table that are very difficult to distinguish by X-ray diffraction may have significantly different neutron scattering lengths, and thus they can be individually distinguished by neutrons. Unlike X-ray diffraction, light and heavy elements often have comparable neutron scattering lengths. This is particularly important for hydrogencontaining materials since scattering from hydrogen with neutrons is comparable to scattering from most other elements in the periodic table.

Figure 5.5 illustrates the structure of the complex hydride Li3AlD6 from Xray and neutron ‘views’. The X-ray view mostly shows Al and Li with reduced accuracy. Neutron diffraction gives the complete structure including the hydrogen atoms. Table 5.1 Comparison of some of the properties of X-ray and neutron diffraction

Availability

X-rays

Neutrons

Easy (laboratory sources), more limited for synchrotrons

Limited

Interaction with matter

Strong

Weak

Scattering by

Electron density

Nuclei

Scattering function of elements in periodic table

∝Z

Complex behaviour

Scattering function of 2θ

Decreasing

Constant

Absorption

Increase with Z, depend on wavelength

Complex behaviour, even between isotopes

Accessories (sample environment)

Difficult (because of strong absorption)

Easy

Magnetic scattering

Very weak

Strong

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

Li3AID6 seen by X-rays

115

Li3AID6 seen by neutrons

5.5 The structure of Li3AlD6 as seen by X-ray diffraction and neutron diffraction. The size of the spheres illustrates the contribution from the different elements to the scattering. With X-rays, scattering from lithium and deuterium is nearly invisible, and X-ray diffraction gives only the position of the aluminium atoms. With neutrons, the strongest scattering is from deuterium, with both aluminium and in particular lithium weaker. This illustrates clearly the importance of combined X-ray and neutron diffraction experiments.

The scattering consists both of a coherent and an incoherent part. The 1H isotope protium (99.985% of natural hydrogen) gives very strong incoherent scattering, and this gives mainly a big background in the neutron diffraction data. On the other hand the isotope deuterium gives mainly coherent scattering which will contribute to the Bragg scattering. Deuterium is therefore used instead of hydrogen for structural studies of hydrides.

5.3.1

Sources

Laboratory source X-ray diffraction equipment, both for single crystal and powder diffraction studies, is widely available. In such set-ups X-rays are produced when high-energy electrons accelerated through a high voltage (typically several kV) collide with a metal target that is often Cu. The energy of the incident electrons is sufficient to ionise some of the Cu 1s electrons (K shell). An electron in an outer orbital will immediately drop into the vacant 1s level, and the released energy appears as X-rays. The radiation consists of characteristic X-rays determined by the anode material of the X-ray tube and a more continuous spectrum, called white radiation or Bremsstrahlung. The characteristic X-rays are much more intense than the white radiation. Commonly used target materials are Cu, Mo and Cr with X-ray wavelengths corresponding to the Kα1 transitions equal to 1.5404, 0.7093 and 2.2896 Å, respectively. During the past decade it has been a significant instrumental

© 2008, Woodhead Publishing Limited

116

Solid-state hydrogen storage

development both to achieve improved resolution and higher intensities. This includes higher intensities from the X-ray tube by using rotating anodes, improved collimation, focusing of the X-rays and use of linear and positionsensitive detectors with good spatial resolution. In addition, the instrumentation is also more flexible, allowing both flat-plate and capillary geometries, use of cryostats, furnaces, pressure cells, etc. and thus also allowing different types of in situ experiments. During the past decades synchrotron X-ray sources have been available in many countries. Today such facilities are the most powerful X-ray radiation sources. In a synchrotron the electrons are accelerated to relativistic speeds, achieving a final energy in the GeV range in a so-called booster. The electrons are then injected into an ultra-high vacuum storage ring that can be several hundreds of metres in diameter. The electrons are forced to travel in a closed loop by strong magnetic fields. Electromagnetic radiation is produced owing to the acceleration of the charged particles resulting from the bending magnets. Depending on the energy of the particles, soft to hard X-rays can be produced (wavelength as short as 0.1 Å). In the first-generation synchrotrons, radiation from bending magnets were used for diffraction experiments. In the current third-generation synchrotrons, insertion devices such as wigglers and undulators are used in order to significantly enhance the intensity of the X-ray beam. At present there are more than 50 synchrotrons worldwide. The three most powerful synchrotron X-ray sources are the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, the Super Photon Ring (Spring-8) in Hyogo, Japan, and the Advance Photon Source (APS) in Argonne, USA, with electron energies 6.0, 8.0 and 7.0 GeV, respectively. Both the intensity of the X-ray beam from the synchrotron and its divergence are much better than from a conventional X-ray tube. The brilliance of the X-rays, defined as the number of photons per mrad, per unit of bandwidth and per unit of source area, from the third-generation of synchrotrons, is up to 12 orders of magnitude higher than from a conventional X-ray tube. This extremely high brilliance gives significantly improved possibilities for both high resolution and time-resolved in situ experiments with synchrotrons. Another important advantage of the synchrotrons is the broad continuous distribution of the beam intensity as a function of the wavelength. This allows an easy selection of optimal wavelength, e.g. to reduce absorption either in sample and sample cells or for experiments at varying wavelengths. As shown above, neutrons are in particular important for studies of hydride materials. Neutrons can be produced either in nuclear reactors or by pulsed (spallation) neutron sources. In the research nuclear reactors neutrons are produced by fission processes based on U-235 (which is 0.7% in natural uranium, but usually enriched as fuel for reactors). Since the released neutrons from these processes are very energetic, the required chain reaction for continuous production of neutrons requires moderation (to reduce the energy)

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

117

of the neutrons. Usually heavy or light water is used as moderator. Neutrons are supplied to the experimental set-ups via beam channels or also via neutron guides where the neutron beam can be moved tens of metres without losing much intensity. This allows neutron instrumentation in separate buildings, in so-called guide-halls with significantly reduced background from the reactor source. In addition, this also makes space for more instruments. One of the advantages of the neutrons is the continuous energy spectrum determined solely by the temperature of the moderator (a Maxwell–Boltzmann distribution). Thermal neutrons are produced in research reactors with water (heavy or light) as the moderator. The available wavelengths are typically in the range 1–2.5 Å with a maximum in intensity at about 1.3 Å. Longer wavelengths can be obtained by using a cold moderator (usually a sphere filled with liquid hydrogen at 20 K) in some of the beam tubes. Shorter wavelength can be obtained by for example using a hot moderator using a graphite block. At present there are worldwide about 30 research reactors. Examples of the most intense reactors are ILL (Institut Laue-Langevin) in Grenoble, France, FRM II in Munich, Germany, BENSC (Hahn-Meitner Institut) in Berlin, Germany, NIST in Gaithersburg, USA, HFRI at Oak Ridge National Laboratory, USA, and ANSTO in Australia. Another method to produce neutrons is by accelerator-based neutron sources, usually named spallation sources. When a high-energy proton bombards a heavy atomic nucleus such as lead, tungsten or mercury, some neutrons are knocked out in a nuclear spallation reaction. For every proton striking the nucleus, 20–30 neutrons are expelled. The neutrons are moderated to obtain the useful wavelength distribution. The spallation method is the most efficient way to produce neutrons, and the neutron flux is higher than for neutrons produced in reactors. The neutrons can be used in different diffraction studies, and mostly in so-called time-of-flight (TOF) experiments. The most powerful spallation source at present (2008) is the Spallation Neutron Source (SNS) at Oak Ridge in the United States. Other important sources are ISIS in Didcot, UK, LANCE at Los Alamos National Laboratory, USA, and IPNS at Argonne National Laboratory, USA. At present several spallation sources are planned and under construction and the most important is the Japan Proton Accelerator Research Complex (J-PARC) in Tokai, Japan, which will be in operation by the end of 2008. Finally, the continuous spallation neutron source SINQ at the Paul Scherrer Institut in Switzerland should be mentioned. It is the first of its kind in the world.

5.3.2

Powder diffractometers

Metal hydrides are usually in the form of powders, and here we therefore will describe only instrumentation for studies of powder samples, so-called powder diffractometers. In principle, this experimental set-up is very simple,

© 2008, Woodhead Publishing Limited

118

Solid-state hydrogen storage

with an incoming beam (X-rays or neutrons), a monochromator to select a beam with a specific wavelength, the sample position and a detector system to measure the scattered photons or neutrons. In practice, set-ups will also include collimators, slit-systems, sometimes focusing devices (monochromators and mirrors) and different types of shielding materials, and therefore powder diffractometers can be rather complicated. Two important parameters for powder diffraction experiments are intensity and resolution of the measured data. It is not possible to obtain high resolution and high intensity at the same time. High resolution requires a very well-collimated beam, resulting in a rather low flux on the sample. To increase the intensity, the resolution (collimation) has to be reduced. However, a very effective detector system can allow both a reasonable high intensity and good resolution. The function of the monochromator is to select a wavelength (or in practice a distribution of wavelengths) for the experiment. For neutron powder diffractometers the monochromator is usually focusing to increase the flux on the sample and with some mosaicity also to increase the scattered intensity. Typical monochromator materials are germanium, silicon and graphite. For synchrotron X-ray diffraction, typically a so-called channel-cut perfect silicon monochromator is used. For modern laboratory source X-ray powder diffractometers the characteristic X-ray lines are used directly; typically the Kβ line is ‘removed’ by a filter and Kα2 is ‘removed’ electronically. Traditionally detector systems for X-rays are scintillation point detectors. A complete diagram can be measured by stepping the detector from the start to the final 2θ angle. This is rather time-consuming. By a selection of slits and collimators, the resolution by using a point detector can be very high, and therefore scintillation point detectors are still commonly used. The efficiency can be improved by using banks of detectors. The traditional detectors for neutron diffractometers are vertical 3He detectors. Since the flux of neutrons is significantly lower than for X-rays, a large number of detectors can be used: for example, at the D2B high-resolution diffractometer at ILL there are in total 64 detectors spaced by 2.5° in 2θ. With positionsensitive detectors (PSD) both the position of the X-ray photon or the neutron are measured in addition to the number of hits at that position. This significantly increases the efficiency, but at the same time the available resolution is lower since no collimation can be used in front of the detector. For X-ray setups, both so-called image plates and CCD (charge coupled device) cameras are used as two-dimensional detectors in addition to micro-strip linear detectors. For neutron instrumentation typically linear detectors (e.g. micro-strip, 3Hebased PSDs or scintillation counters) are available for powder diffraction. With such detectors, measurements of complete diagrams can be done in seconds. Thus it is possible to follow reactions, e.g. hydrogenation and dehydrogenation processes, as a function of time. However, the angular resolution will be limited.

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

119

Powder diffraction experiments can be performed with basically three main geometries: flat sample reflection geometry, flat sample transmission geometry and cylindrical transmission sample. For powder neutron diffraction experiments the sample is usually placed in cylindrical sample holders usually made of vanadium since vanadium has a nearly neglectable neutron scattering power. For X-ray powder diffraction both reflection and transmission geometries are used. A major challenge with flat-plate geometry (often as the Bragg–Brentano focusing geometry) is the problem of preferred orientation. This is strongly reduced by using a rotating capillary/cylindrical sample holder. However, for heavy elements with inherent strong absorption, flatplate geometry should usually be chosen. With synchrotron radiation also capillary geometry can be used for heavy elements by tuning the wavelength to minimise the absorption or in general by using hard X-rays (i.e. λ < 0.5 Å). To benefit from the pulsed nature of neutron spallation sources, powder diffractometers on such sources are different from the standard construction on a research reactor. Instead of measuring intensities as a function of scattering angle at constant wavelength, the intensities are measured as a function of the wavelength at a constant scattering angle with TOF instruments. The wavelengths of the neutrons are inversely proportional to their speed and thus given by the time the neutrons spend from the source to the detector via the sample. This will then correspond to a certain d-value in Bragg’s law. Instruments with a long flight path, like the instrument HRPD at ISIS with a flight path of nearly 100 m, will give data with a resolution that is comparable to high-resolution synchrotron data.

5.4

The use of powder diffraction data

Equation (5.2) shows that the intensity in the diffraction diagram is proportional to |Fhkl|2. However, there are also several other factors contributing to the intensity, including temperature- and angle-dependent parameters. More generally, the integrated intensity in powder diffraction can be expressed by: Ihkl = S · Mhkl · Lθ · Pθ · Aθ · Thkl · |Fhkl|2

[5.4]

where S is the scale factor. It is used to normalise the observed and calculated intensities, and it depends on wavelength, sample volume, intensity of incoming beam, distance between sample and detector unit cell volume, etc. Mhkl is the multiplicity factor accounting for the presence of multiple symmetrically equivalent points in reciprocal space, Lθ is the Lorentz factor that is defined by the geometry of diffraction, Pθ is the polarization factor accounting for the partial polarization of the scattered electromagnetic wave, Thkl is the preferred orientation factor and the structure factor Fhkl is defined in Eq. (5.2).

© 2008, Woodhead Publishing Limited

120

Solid-state hydrogen storage

As described above, powder diffraction data can be measured with different sources and different geometries. Different levels of information can be obtained, and this will be described in the following.

5.4.1

Phase identification

Every crystalline phase in a sample has a unique powder diffraction pattern determined from the unit cell dimensions and the atomic arrangement within the unit cell. It can be considered a ‘fingerprint’ of the material. Thus, powder diffraction can be used for phase identification by comparing measured data with diffraction diagrams from known phases. The most efficient computer searchable crystallographic database is the PDF-4 from the International Centre for Diffraction Data (ICDD) [3]. It is used by very efficient computerbased search-processes. In 2007 the PDF-4+ database contains information about Bragg-positions and X-ray intensities for more than 450 000 compounds, out of which there are about 107 500 data sets with atomic coordinates. New entries are added every year. The positions of the peaks in the measured pattern have to be determined. This can be done manually, but effective, fast and reliable automatic peak search methods have been developed. The method can obviously be successful only if the phases in the sample are included in the database. However, the database can also help to determine unknown phases if X-ray data exist for another isostructural compound albeit with a different composition.

5.4.2

Powder indexing and extraction of Bragg intensities

The first step in determining a crystal structure is to find the size and the shape of the unit cell. This is normally straightforward for single crystals since the reflections in the three-dimensional reciprocal lattice can be measured individually. However, for powder diffraction it is much more difficult since the experimental data are one-dimensional projections. The result of this is significant overlap of measured Bragg peaks, both for reflections with identical Bragg angles but different positions in reciprocal space, and for reflections with nearly identical Bragg angles where the experimental resolution is not sufficient to distinguish the peaks. To determine the six parameters, a, b, c, α, β and γ (the unit cell axes and the angles between them), could seem to be a reasonable task, but in practice it can be extremely difficult. The difficulties are partly because of systematic errors for example related to reduced accuracy of the determined peak positions and zero shifts and partly because of the sometimes severe overlap of Bragg peaks. Samples with more than one unknown phase make this even more difficult. There is no simple approach to solve this problem. Several methods including trial-and-error approaches

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

121

and different search algorithms are available. Each of these methods has its specific strengths and weaknesses. Some approaches work best for highsymmetry unit cells, others for low-symmetry cases, etc. All approaches use a list of d-values (corresponding to 2θ values for a specific wavelength) as input. However, only well-resolved reflections should be used, and this reduces usually the maximum number to 20–30 reflections. The computer program CRYSFIRE [4] has a common user interface for nine different powder indexing programs. This allows an easy comparison among results from different programs. One of the challenges is to find the smallest unit cell satisfying the input reflections, and a specific figure of merit is used to rank the proposed unit cells. The next step is to determine the space group for the compound. In total there are 230 possible space groups, and it is necessary to know the symmetry and thereby the space group to be able to find the positions of the atoms in the structure [5]. The crystal system is easily determined from the unit cell axes and angles, e.g. it is cubic if all axes are equal and all angles are 90º. In addition, systematic absences (extinctions) among the reflections (meaning position for Bragg peaks with zero intensity), will further limit the number of possible space groups. Lattice centring such as body-centred cubic (BCC) and face-centred cubic (FCC) results in specific extinction rules. For example body-centred space groups will only have intensity for reflections hkl where h + k + l = 2n (the sum is an even number). In addition, the atoms in the asymmetric unit within the unit cells can be linked via symmetry operations such as inversion, rotational axis, mirror planes, screw axis and glide planes. Some of these (e.g. screw axis and glide planes) involve additional extinction rules for specific groups of reflections. A systematic evaluation of possible extinctions is needed to find possible space groups that can be used in the structure determination. The final step before structure determination by traditional methods is extraction of integrated intensities from the powder diffraction pattern (see below). Intensities can nearly always be determined only after deconvolution of partially overlapping Bragg reflections. This can be done for individual peaks or for groups of peaks, but usually full decomposition methods are used. There are basically two techniques: the iterative Le Bail method [6] and the constrained linear least-squares approach developed by Pawley [7].

5.4.3

Structure solution from powder diffraction data

Even when the unit cell, the space group and Bragg intensities have been found, the solution of the crystal structure, i.e. to determine the positions of the atoms, is not straightforward. As mentioned above, this is an inherent problem with X-ray and neutron diffraction since the information about the phases of the structure factor is lost in the experiments (see Eq. (5.2)). Ab

© 2008, Woodhead Publishing Limited

122

Solid-state hydrogen storage

initio methods have been developed to overcome this problem. The reciprocal space techniques direct methods and the Patterson method start from only the observed intensities. Such methods are therefore very efficient for single crystals where the Bragg peaks corresponding to indices hkl are resolved. For powder diffraction this is much more complicated because of the intrinsic overlap of multiple reflections. During the past years these methods have also been adopted for powder diffraction [8]. Both techniques require individual intensities and both the Le Bail and the Pawley methods are useful. The direct methods approach is a technique that handles the crystallographic phase problem using mathematical relationships between intensities of Bragg reflections [9]. Possible solutions for the structure factor phases and thus determination of the position of the individual atoms in the unit cell are proposed from statistical relationships between groups of reflections. The Patterson method is often called the heavy atom method since it is effective to determine atoms with the strongest scattering, and for X-rays these are the heavy atoms. The method was developed by Patterson in 1934 [10]. |Fhkl|2 values are used as coefficients in Fourier transformations, and the resulting so-called Patterson map consists of a distribution of interatomic vectors in the unit cell dominated by the strongest scatters. Another interesting reciprocal space approach is the maximum entropy method. It is based on Bayesian estimation theory [11], and can be used both to solve the phase problem and to determine the relative intensities of overlapping reflections. During the past years global optimisation methods that are applied in direct space have become powerful. They involve assessments of multiple trial crystal structures, and in practice the concept is to minimise disagreements between the observed diffraction data and diffraction data calculated from a structure model. Monte Carlo simulations are usually used in these approaches, and typically they are computationally intensive. However, continuous improvements in the algorithms have significantly increased the potential for such approaches for structure solution based on powder diffraction data. Probably the most popular approach related to metal hydrides is so-called simulated annealing method [12], for example as implemented in the computer program FOX [13]. This is a Monte Carlo method where the sample undergoes virtual heating to certain ‘temperatures’. The starting ‘temperature’ should be high in the beginning to avoid local minima, but during the optimisation the ‘temperature’ can gradually be decreased. A proper cooling scheme is important. In the parallel tempering approach optimisation is performed at several temperatures at the same time, and no cooling scheme has to be chosen before the optimisation.

5.4.4

Structure refinements – the Rietveld method

The final step in a structure solution is to optimise and refine the experimental diffraction pattern and crystallographic parameters against the observed powder

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

123

diffraction data. Before the middle of the 1960s the only method was to use individual integrated intensities, but as it has been shown above, this is inaccurate for powder data because of the overlap problems. For single crystal diffraction this method works well since intensities of large number of Bragg peaks can be measured, and accurate crystallographic data can be obtained by least-square refinements based on the observed and calculated intensities. In 1967 Hugo Rietveld suggested using the whole experimental powder diffraction diagram in the refinements by fitting a calculated diagram to the observed data [14]. This approach is similar to the full pattern decomposition (Pawley and Le Bail algorithms), but in the Rietveld method also refinements of structural parameters, such as the positions of the atoms, are included. It should be emphasised that the Rietveld method is a structure refinement method and not a method to solve unknown structures. An adequate structure model has to be known in advance. The minimising function in the Rietveld method is: n

M = Σ w i ( y iobs – y icalc ) 2 i =1

[5.5]

n is the number of measured data points, y iobs and y icalc are the observed and calculated intensity at point i of the powder diffraction pattern, respectively, and wi is the weight for point given by w i = 1/ σ i2 where σi is the variance. The calculated intensity can be described by (assuming a sample with only one phase): y icalc = bi + Si Σ M hkl ⋅ Li ⋅ Pi ⋅ Ahkl ⋅ Thkl ⋅ | Fhkl | 2 ⋅ Φ (2θ hkl – 2θ i ) hkl

[5.6] bi is the background in point i. It can either be determined manually by interpolation between selected points or described by a polynomial. The sum covers all reflections contributing to Bragg scattering in point i. Si, Li, Pi, Ahkl, Thkl and Fhkl are defined above. Φ is the profile function coming from instrumental and sample broadening in the diffraction pattern. 2θhkl is ideal Bragg angle for reflection hkl and 2θi is the scattering angle corresponding to point i. Several factors influence the shape and width of the measured Bragg peaks, the most important being: collimation of the beam, degree of mosaicity of the monochromator system, detector resolution, for some experimental set-ups the size of the sample in addition to imperfection/ mosaicity in the sample. The resulting profile shape is a convolution of all these components. The measured profiles show both a Gaussian and Lorentzian contribution. The Rietveld method was developed with a pure Gaussian distribution which well fits data from low-resolution neutron powder diffractometer, but with improved resolution, and X-ray diffraction in general, a significant Lorentzian contribution is present. Nowadays the so-called

© 2008, Woodhead Publishing Limited

124

Solid-state hydrogen storage

pseudo-Voigt profile function is the most popular profile function. It is a linear combination of a Gaussian and a Lorentzian function. The peak broadening is measured by the ‘full-width-at-half-maximum’ (FWHM). The Gaussian contribution is given by:

FWHM Gaussian = and the Lorentizan:

U tan 2 θ + V tan θ + W

[5.7]

Y + X tan θ [5.8] cosθ For some measurement geometries there are also introduced asymmetries in profiles in particular at small scattering angles, and this gives additional parameter(s) in the Rietveld refinements. The typical parameters that are refined with the Rietveld method are summarised in Table 5.2. The starting point for the refinements should be an adequate structure model. The Rietveld method can also be used for multiphase samples. Usually neither preferred orientation nor absorption is refined. Refinements of anisotropic displacement parameters require high-resolution data measured to high 2θ angles. The sequence of parameters used in the refinements is important. Several computer programs are available for Rietveld refinements, examples are Fullprof, GSAS and TOPAS-Academic [15–17]. It is possible both to refine data with several phases and simultaneously data measured with different sources (e.g. by X-rays and neutrons) and by different wavelengths. This is in particular important for hydrogen storage materials where X-ray diffraction gives the positions of the heavy atoms while neutron diffraction is important for location of hydrogen atoms in the structure. An important question to answer is when is the fit to the structure model acceptable? First of all, there should not be any unidentified reflections in the powder diagram. A visual method is to compare the observed and calculated diagram. Figure 5.6 shows observed, calculated and difference plots from FWHM Lorentzian =

Table 5.2 Parameters for Rietveld refinements for a single phase compound based on constant wavelength X-ray or neutron diffraction data Scale factor (S) Zero-shift Background (when refined as polynomial) Lattice parameters: in general a, b, c, α, β, γ Profile parameters: U, V, W, X, Y, asymmetry parameter(s) Coordinates of atoms: xj, yj, zj Isotrope displacement parameters: Bj Occupation parameters: Oj Anisotropic displacement paramters: Bij Preferred orientation: T Absorption: A

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

125

3500 Na2LiAID6 PND

3000

Intensity (arb. units)

2500 2000 1500 1000 500 0 350 0 –350

20

40

60

80

100

120

2θ (°)

5.6 Observed intensities (circles) and calculated intensities from Rietveld refinements (upper line) of powder neutron diffraction (PND) data for Na2LiAlD6. The difference between observed and calculated intensity are shown with the bottom line. Reprinted from Brinks et al. [18], copyright (2005), with permission from Elsevier.

Rietveld refinements. The latter is the difference between the observed and calculated intensities for every measured point. Different figures of merits used to evaluate the quality of the refinements are also available. The most widely used are the weight profile R-factor:  n w i ( y iobs – y icalc ) 2  iΣ =1 Rwp =  n  Σ w i ( y iobs ) 2  i =1

    

1/2

[5.9]

And the goodness of fit: n

χ2 =

Σ w i ( yiobs – yicalc ) 2 i =1 n–p

 Rwp  =    Rexp 

2

[5.10]

where p is the number of free least square parameters and Rexp is the socalled expected profile R-factor. An important application of the Rietveld method is determination of relative amounts of different phases in a sample with so-called quantitative phase

© 2008, Woodhead Publishing Limited

126

Solid-state hydrogen storage

analysis. Based on multi-phase refinements the molar-, weight- and volumefractions of the different phases can be calculated from the individual scale factors and the unit cell contents and volumes. Finally, the accurate analyses of the profile parameters and their dependence on the scattering angle 2θ can give information about strain and particle sizes in the sample. The main contribution to these effects will appear as Lorentzian broadening. The first term in Eq. (5.8) describes size broadening and the second term corresponds to the strain contribution to the Lorentzian broadening.

5.5

Examples of structures and results from powder diffraction studies

Structures and results from powder diffraction studies are shown in many places in this book. Here we will briefly describe two recent examples to illustrate the potential and status of powder diffraction for studies of hydride materials: (i) structural determination of Mg(BD4)2 and (ii) in situ experiments and quantitative phase analysis to study the desorption process in different AlD3 phases.

5.5.1

Structure of Mg(BD4)2

Boron-based compounds are promising materials for hydrogen storage, and one example of a high-capacity material is Mg(BH4)2 with a theoretical hydrogen capacity of 14.8 wt%. For this compound it exists as both a lowtemperature phase and a high-temperature phase, α- and β-phases, respectively. The transformation is described by Riktor et al. [19]. Her et al. [20] have used synchrotron X-ray diffraction to solve the structure of both the α- and the β-phase based on a single-phase hydrided sample. However, from X-rays only reasonable positions for the hydrogen atoms can be indicated. Černý et al. [21] have determined the complete low-temperature structure for the deuterided compound using both synchrotron X-ray and neutron diffraction. Their sample was multiphase with approx. 50 wt% of the main phase. A complication with neutron diffraction studies of boron-based compounds is the strong absorption in natural boron, and a double-walled vanadium sample holder was used for the neutron diffraction experiment. It should be added that the 11B isotope has a significantly lower absorption of neutrons and thus an improved signal/noise ratio compared with natural boron. Therefore, if possible, the 11B isotope should be used for neutron diffraction experiments, but in this case it was not possible to synthesise the 11B-rich compound. Both the α- and the β-phases are complex with large unit cells, but the lowtemperature phase is remarkably complicated. The X-ray and neutron diffraction patterns are shown in Fig. 5.7. Černý et al. used FOX [13] to solve the

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

127

Intensity

90 000

0

15

20

25

15 2θ (°) (a)

20

25

Mg(BD4)2 LiCl Li2MgCl4 LiBD4

5

10

Intensity

40 000

Mg(BD4)2 LiCl Li2MgCl4 LiBD4

0 0

25

50

75 2θ (°) (b)

100

125

150

5.7 Observed (circles), calculated (upper line) and difference (bottom) (a) synchrotron X-ray and (b) neutron diffraction patterns for Mg(BD4)2. The Bragg positions for individual phases are shown with ticks. From Cˇerny´ et al. [21] and reproduced with permission. Copyright Wiley-VCH Verlag GmbH Co. KGaA.

© 2008, Woodhead Publishing Limited

128

Solid-state hydrogen storage

structure and TOPAS-Academic [17] for refinements. Her et al. used the direct-method program EXPO [22] for the structure solution and TOPASAcademic for refinements. The space group for the α-phase is P61 with unit cell dimensions: a = 10.318 and c = 36.998 Å (and angles 90, 90 and 120°) (from [21]). The unit cell contains 55 symmetry independent atoms, and in the final Rietveld refinements 186 free parameters were refined. This structure solution is clearly state-of-the-art for structure solution using both highresolution synchrotron X-ray diffraction data and the combination of X-ray and neutron diffraction.

5.5.2

In situ diffraction studies of AlH3

AlH3 with 10.1 wt% hydrogen is one of the promising materials for hydrogen storage. From a structural point of view it is an interesting compound since at least six different modifications exist depending on the synthesis method. At present the structure of four modifications are known: α-, β-, γ- and α ′AlH3. These structures have been determined by combined synchrotron Xray and neutron diffraction using deuterided samples, meaning AlD3 [23– 25]. There are significant differences in the structures. The space groups are: R 3 c , Cmcm, Fd 3 m and Pnnm for α-, α′-, β- and γ-AlD3, respectively. The structures of the α, α ′ and β isomorphs consist of 3D networks of cornersharing AlD6 octahedra, where each hydrogen atom is shared between two octahedra. However, the connectivity of the octahedra is different. For the αmodification the octahedra in the first coordination sphere around each octahedron are not interconnected, whereas for α ′-AlD3 four of the six octahedra are interconnected in pairs and for β-AlD3 all six octahedra are bound to two others. There are two types of AlD6 octahedra in γ-AlD3, one sharing corners only and one sharing both edges and corners. One of the interesting questions is how the different modifications decompose. The decomposition of β- and γ-AlD3 is studied by time-resolved in situ synchrotron X-ray diffraction by Grove et al. [26]. During the experiments the samples were placed in capillaries connected to a vacuum pump and heated at 1 K/ min. Data were collected with a two-dimensional image plate system (MAR345) every second minute. Data converted to one-dimensional plots are shown as a function of the temperature in Fig. 5.8 for the decomposition of both β- and γ-AlD3. To understand the process in detail, quantitative phase analyses were performed for a number of data sets measured at different temperatures during the heating. Figure 5.9 shows the resulting development of phases as a function of temperature. The phase compositions at each temperature are determined from complete Rietveld refinements, and this is a time-consuming process. Figure 5.9(a) shows that β-AlD3 transforms to α-AlD3 before release of hydrogen and decomposition to Al. The transformation from the β to the

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

129

70 140

Al

100

Scan number

T (°C)

60

α

50 40 30 20

60

β

10 0 5

10

15

20 25 2θ (°) (a)

30

35

100 Al

α

Scan number

80

60

40

γ 20

0

10

15

20 2θ (°) (b)

25

30

5.8 Synchrotron X-ray diffraction patterns showing the desorption as a function of temperature for (a) β-AlD3 and (b) γ-AlD3. The heating rate is 1 K/min and a new diffraction pattern is collected every second minute. Reprinted in part with permission from Grove et al. [26]. Copyright (2007) American Chemistry Society.

© 2008, Woodhead Publishing Limited

130

Solid-state hydrogen storage 100

α -AlD3

β -AlD3

80

wt%

60

40

wt%

10 8 6

Al

4 2 0

60 80

20

100 120 140 T (°C)

γ -AlD3

0 40

60

80

100 T (°C) (a)

120

140

160

100

γ -AlD3 80

Al

80 60

wt%

wt%

60

40

40 20 0

20

α -AlD3

100

120 T (°C)

140

0 40

60

80

100 T (°C) (b)

120

140

160

5.9 Relative amount of the different phases during decomposition of (a) β-AlD3 and (b) γ-AlD3. The raw data are plotted in Fig. 5.8 and the relative amounts have been determined by quantitative phase analyses. Reprinted in part with permission from Grove et al. [26]. Copyright (2007) American Chemistry Society.

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

131

α isomorph starts at about 80 °C and the following decomposition of α-AlD3 to Al and D2 at about 120 °C. It is clear from this study that β transforms into α before further decomposition into Al and D2. The transformations are more complicated for γ-AlD3 (see Fig. 5.9(b)). At about 90 °C it starts to transform to α. However, at 110 °C the decomposition rate of γ-AlD3 increases accompanied by the formation of Al. This can be explained by a transformation of the γ-phase directly to Al. Above about 130 °C α-AlD3 also decomposes into Al. This shows therefore a significant difference between the γ and β decomposition. It is also clear from these experiments that α- is more stable than both β- and γ-AlD3. From in situ synchrotron X-ray diffraction studies it has also been shown that α ′- is less stable than α-AlD3 [27]. As described above, the AlD3 isomorphs have quite different structures with different connections between the AlD6 octahedra in the four modifications. The transformations from β-, γ- and α′-AlD3 to α-AlD3 are accompanied by major rearrangements of the AlD6 octahedra. The understanding of this is still unclear, for example if the transformations involve intermediate phases that have not yet been observed.

5.6

Future trends

Powder diffraction techniques have developed enormously since 1990. This is partly due to improved X-ray and neutron sources and instrumentation and partly because of improved methods, algorithms and software for structure solution and refinements of experimental data. The more effective computers are also important. In the new high-flux neutron spallation source, SNS in the United States, the coming J-PARC in Japan (2008) and the planned European Spallation Source (ESS) there will be powder neutron diffractometers with significantly better performance than at the present sources. These sources will allow higher resolution, but probably most important will be the possibilities for in situ experiments on a much shorter time scale than with the present instrumentation. Since neutron diffraction is crucial for studies of hydrogencontaining materials, development of these neutron sources will result in major achievement. Also new synchrotrons and upgrades of the existing third-generation synchrotrons such as ESRF will give new possibilities for more advanced in situ experiments. Diffraction experiments combined with other techniques, such as different spectroscopic methods, will develop in the coming years. At present it is possible to carry out synchrotron X-ray diffraction experiments combined with for example Raman spectroscopy and extended X-ray absorption fine structure (EXAFS) spectroscopy. Such experiments will help to clarify hydrogenation/dehydrogenation reactions involving for example amorphous/ nanocrystalline phases, as for example observed in several borohydrides.

© 2008, Woodhead Publishing Limited

132

Solid-state hydrogen storage

Combined diffraction/spectroscopic experiments will probably also be available at neutron scattering facilities in the future. In the typical diffraction experiments only the position and intensity of the Bragg peaks are used. This gives information about the long-range order. However, also the background contains structural information and then in particular related to short-range order. In total scattering experiments are both the Bragg peaks and the background included in the analyses. Both Xray and neutron diffraction data can be used. Such experiments will give further understanding of atomic arrangements. The first total scattering experiment on a metal hydride was a total scattering experiment on a sample of VD0.7 [28]. Finally, there has been significant progress in software and methods for structure solution and refinements during the last years. Based on the steady improvement in computer performance, it is expected that algorithms and software codes will be further improved and strengthen in the coming years.

5.7

Sources of further information and advice

G. E. Bacon (1975) Neutron Diffraction. Clarendon Press. J. Baruchel, J. L. Hodeau, M. S. Lehmann, J. R. Regnard, C. Schlenker (1993) Neutron and Synchrotron Radiation for Condensed Matter Studies. Theory, Instruments and Methods. Volume 1. Springer Verlag. D. L. Bish, J. E. Post (eds.) (1989) Modern Powder Diffraction. Reviews in Mineralogy, Vol 20. Mineralogical Society of America. P. Coppens (1992) Synchrotron Radiation Crystallography. Academic Press Limited. W. I. David, K. Shankland, L. M. McCusker, C. Bärlocher (eds.) (2006) Structure Determination from Powder Diffraction Data. IUCr Monographs on Crystallography 13, Oxford University Press. A. Furrer (ed.) (1994) Neutron Scattering from Hydrogen in Materials. Proceedings of the Second Summer School on Neutron Scattering, Zuoz, Switzerland. World Scientific Publishing Co. A. Furrer (ed.) (1998) Complementarity between Neutron and Synchrotron X-ray Scattering. Proceedings of the Sixth Summer School on Neutron Scattering, Zuoz, Switzerland. World Scientific Publishing Co. G. Giacovazzo, H. L. Monaco, G. Artioli, D. Viterbo, G. Farrairis, G. Gilli, G. Zanotti, M. Catti (2002) Fundamental of Crystallography. Second Edition. Oxford University Press. C. Hammond (1997) The Basics of Crystallography and Diffraction. IUCr texts on crystallography 3, Oxford University Press. A. D. Krawitz (2001) Introduction to Diffraction in Materials Science and Engineering. John Wiley & Sons Ltd.

© 2008, Woodhead Publishing Limited

Structural characterisation of hydride materials

133

V. K. Percharsky, P. Y. Zavalij (2003) Fundamentals of Powder Diffraction and Structural Characterization of Materials. Kluwer Academic Publishers. G. L. Squieres (1978) Introduction to the Theory of Thermal Neutron Scattering. Cambridge University Press. G. Will (2005) Powder Diffraction: The Rietveld Method and the Two Stage Method to Determine and Refine Crystal Structures from Powder Diffraction Data. Springer Verlag. M. M. Woolfson (1997) An Introduction to X-ray Crystallography. Second Edition. Cambridge University Press. R. A. Young (ed) (1993) The Rietveld Method. IUCr Monographs on Crystallography 5, Oxford University Press.

5.8

References

1. W. Friedrich, P. Knipping, M. Laue, Sitzungber. (Kgl.) Bayerische Akad. Wiss., (1912) pp. 303–322. 2. W. L. Bragg, Proc. Cambridge Phil. Soc. 17 (1913), 43. 3. http://www.icdd.com 4. http://www.ccp14.ac.uk/tutorial/crys 5. T. Hahn (ed.) (2002) International Tables for Crystallography, vol. A. Fifth revised edition, Published for the International Union of Crystallogaphy by Kluwer Academic Publishers. 6. A. Le Bail, H. Duroy, J. L. Fourquet, Mater. Res. Bull. 23 (1988), 447. 7. G. S. Pawley, J. Appl. Crystallogr. 14 (1981), 357. 8. W. I. F. David, K. Shankland, L. B. McCusker, Ch. Baerlocher (eds.) (2002) Structure Determination from Powder Diffraction Data. IUCr Monographs on Crystallography 13, Oxford University Press. 9. J. Karle, H. Hauptman, Acta Crystallogr. 3 (1950), 181. 10. A. L. Patterson, Phys. Rev. B 46 (1934), 372. 11. C. J. Gilmore, Acta Crystallogr. A52 (1996), 561. 12. A. A. Coelho, J. Appl. Crystallogr. 33 (2000), 899. 13. V. Favre-Nicolin, R. Černý, J. Appl. Crystallogr. 35 (2002), 734. 14. H. M. Rietveld, Acta Crystallogr. 22 (1967), 151; 2 (1969), 65. 15. J. Rodriguez-Carvajol, Physica B 192 (1993), 55. 16. A. C. Larson, R. B. Von Dreele (1994) GSAS Report No. LAUR 86-748. Los Alamos National Laboratory, New Mexico, USA. 17. A. A. Coelho (2004) TOPAS-Academic, http://members.optusnet.com.au/alancoelho 18. H. W. Brinks, B. C. Hauback, C. M. Jensen, R. Zidan, J. Alloys Comp. 393 (2005), 27. 19. M. D. Riktor, M. H. Sørby, K. Ch łople, M. Fichtner, F. Buchter, A. Züttel, B. C. Hauback, J. Mater. Chem. 17 (2007), 4939. 20. J.-H. Her, P. W. Stephens, Y. Gao, G. L. Soloveichik, J. Rijssembeek, M. Andrus, J.C. Zhao, Acta Crystallogr. B63 (2007), 561. 21. R. Černý, Y. Filinchuk, H. Hagemann, K. Yvon, Angew. Chem. Int. Ed. 46 (2007), 5765. 22. A. Altomare, M. C. Burla, M. Camalli, B. Carrozzini, G. L. Cascarano, C. Giacovazzo, A. Guagliardi, A. G. G. Moliterni, G. Polidori, R. Rizzi, J. Appl. Crystallogr. 32 (1999), 339.

© 2008, Woodhead Publishing Limited

134

Solid-state hydrogen storage

23. H. W. Brinks, A. Istad-Lem, B. C. Hauback, J. Phys. Chem. B 110 (2006), 25833. 24. H. W. Brinks, W. Langsley, C. M. Jensen, J. Graetz, J. J. Reilly, B. C. Hauback, J. Alloys Comp. 433 (2007), 180. 25. H. W. Brinks, C. Brown, C. M. Jensen, J. Graetz, J. J. Reilly, B. C. Hauback, J. Alloys Comp. 441 (2007), 364. 26. H. Grove, M. H. Sørby, H. W. Brinks, B. C. Hauback, J. Phys. Chem. C 111 (2007), 16693. 27. S. Sartori, S. M. Opalka, O. M. Løvvik, M. Guzik, X. Tang, B. C. Hauback, J. Mater. Chem. 18 (2008), 2361. 28. M. H. Sørby, R. G. Delaplane, A. Mellergård, A. Wännberg, B. C. Hauback, H. Fjellvåg, J. Alloys Comp. 363 (2004), 214.

© 2008, Woodhead Publishing Limited

6 Neutron scattering studies for analysing solid-state hydrogen storage D. K. R O S S, University of Salford, UK

6.1

Introduction

In the search for useful hydrogen storage materials, whether solid state or H2 adsorbed on surfaces, the most immediately relevant data are, of course, the macroscopic thermodynamics and kinetics of the ab(ad)sorption/desorption process. However, if we seek to use scientific reasoning to guide our search, it is clearly necessary to understand the behaviour of hydrogen on the atomic scale both within the material and when entering or leaving it. It is in the latter context that thermal neutron scattering can make a crucial contribution, particularly when linked to ab initio modelling. Neutron scattering is a widely used technique for studying the structure and dynamics of solids and liquids. This is because a thermalised neutron emerging from a moderator at a nuclear reactor or pulsed neutron source has a wavelength that is comparable to interatomic distances and an energy that is comparable to the energies of atomic excitations in the solid. Moreover, we have good techniques for measuring diffraction patterns and energy transfers very accurately. There are many cases where neutron diffraction (ND) has significant advantages over X-ray or electron diffraction while inelastic neutron scattering (INS) often has advantages over IR and Raman measurements, for instance, in the measurement of phonon dispersion curves in single crystal samples. Other areas of application include small angle neutron scattering (SANS) for measuring longer range spatial correlations, where neutrons have many advantages over X-rays and quasi-elastic neutron scattering (QENS), which probes the diffusive motions, particularly of hydrogen, in a unique way. One notable property of the neutron, which we will not be particularly concerned with here, is its magnetic moment which allows a whole array of methods for investigating the magnetic structure and dynamics of solids. General textbooks are available on the subject, such as Squires [1]. A full description of quasi-elastic neutron scattering may be found in the book by Bee [2]. 135 © 2008, Woodhead Publishing Limited

136

Solid-state hydrogen storage

6.2

The neutron scattering method

6.2.1

Advantages of the neutron scattering technique for investigating solid-state hydrogen storage systems

The particularly significant contribution that neutron scattering can make to the study of hydrogen in solids arises from a number of factors: •





• •

• •

The neutron scattering cross-section of the proton (80 × 10–28 m–2 or, colloquially, 80 barns) is between one and two orders of magnitude greater than that of any other isotope so that it is easy to detect hydrogen in quite low concentrations in solids. The hydrogen cross-section is predominantly incoherent (σinc = 79.7 b, σcoh = 1.8 b) for reasons described below and so the measured scattering is dominated by scattering from individual atoms summed over all H atoms present. Incoherent inelastic scattering from polycrystalline solids reproduces the frequency distribution of the hydrogen vibrations in a solid, weighted with the square of the hydrogen vibration amplitudes and inversely with the mass of the hydrogen atom and so normally other elements can be ignored in a solid containing hydrogen. It should be noted that this amplitude-weighted density of vibrational states can also be directly derived from ab initio simulations of the solid’s dynamics and hence provides a direct check on the accuracy of these calculations. Moreover, the enthalpy change on hydriding/dehydriding can be estimated using this density of vibrational states. This technique is particularly useful for electrical conductors because IR and Raman techniques are not available. However, the technique is still very valuable for insulators because the electromagnetic techniques only provide frequencies at the zone centre and the intensities of these features are strongly influenced by the selection rules involved in the electronic response. Incoherent quasi-elastic neutron scattering from diffusing hydrogen atoms allows us to investigate the elementary jumps involved in the tracer diffusion process. On the other hand, the deuteron (2D), has a relatively large cross section which is predominantly coherent (σcoh = 5.6 b and σinc = 1.8 b) and so the diffraction pattern from a crystalline solid contains terms that enable us to locate D in crystal structures through the interference between scattered waves from all pairs of nuclei. H and D have scattering lengths of opposite sign so that there are a whole range of phase contrast techniques that can identify the role of hydrogen, for instance in SANS. A general advantage of neutrons is that they are relatively penetrating so

© 2008, Woodhead Publishing Limited

Neutron scattering studies



137

that samples can be contained in cans that are able to withstand considerable gas pressures and temperatures. This feature enables us to measure the actual process of (de)hydrogenation in situ under working hydrogen pressures and so we can directly investigate processes that determine the kinetics of hydrogen cycling into and out of a hydrogen store. Being the lightest isotope, hydrogen has very pronounced quantum mechanical properties. As the neutron interacts with the nucleus, it directly observes these properties, formally observing the transition state probability between two quantum states of the solid. Of particular importance to us here is the nature of the scattering from para- and ortho-molecular hydrogen because selection rules require that these molecules have antisymmetric total wave functions. Hence we find that para-hydrogen, with anti-parallel spins has only even rotational quantum numbers (J) while ortho-hydrogen, with parallel spins can only have odd values of J. Moreover, the scattering lengths combine in such a way that para-hydrogen only scatters significantly via a spin-flip process involving a J = 0 to 1 transition while ortho-hydrogen has large cross-sections for both spinflip and non-spin-flip scattering. Also, the energy difference between the lowest energy para-state (J = 0) and the lowest ortho state (J = 1) for free hydrogen molecules is 14.7 meV, which is easily measured with high accuracy. These features make it instantly possible to identify whether hydrogen is present in its molecular form and to probe the trapping states with progressively less trapping energy as hydrogen is added. The basic theory of neutron scattering from molecular hydrogen has been described by Young and Koppel [3].

The application of neutron scattering to understanding the role of hydrogen in solids has been described in various general reviews [4,5]. Specific applications to intermetallic compounds are described by Richter et al. [6]. Its use in vibrational spectroscopic studies in chemistry, biology, materials science and catalysis are described by Mitchell et al. [7]. For a review of the basic properties of metal–hydrogen systems, we would refer the reader to the book by Fukai [8].

6.3

Studies of light metal hydrides

Because of the need to store hydrogen in a system with a low mass so that it can approach the US Department of Energy (DoE) gravimetric target of 6– 9% of the storage system mass, we will be interested in applying the technique to complex metal hydrides such as the alanates, borohydrides and amides and to intermetallic hydrides based on magnesium. In these novel systems, various interesting features can be investigated such as the existence of continuous phase transitions as have been observed in in situ diffraction

© 2008, Woodhead Publishing Limited

138

Solid-state hydrogen storage

studies of the lithium amide system [9,10] and in similar investigations into the hydrogen absorption/desorption process in alanates. Borohydrides are also of interest but from our point of view suffer from the disadvantage that samples would have to be made from the 11B isotope because of the strong neutron capture cross-section of 10B.

6.4

Studies of molecular hydrogen trapping in porous materials

The alternative proposed method of storing hydrogen in a solid matrix involves the physisorption of H2 on a high surface area material. The original interest in this approach was based on some very promising reports of high mass uptakes in single-walled carbon nanotubes [11] and carbon nanofibres [12]. Although these results seem to have been affected by residual water in the vacuum system, there have since been a wide range of experiments on a variety of carbon frameworks, zeolites, ice clathrates, metal oxide frameworks (MOFs), etc. While the normal van der Waals adsorption energy is rather weak (~ 40 meV/atom) and would require that the adsorber was cooled to 80 K to get a reasonable uptake, recent work suggests that it may be possible to enhance this adsorption energy with suitable catalysts, possibly involving spillover effects. The work to be described here (Section 6.7) reported below shows how neutron inelastic scattering can be used to characterise the potential energy surface at different trapping sites in a solid.

6.5

The basic theory of neutron scattering

The neutron is an uncharged nucleon with a mass essentially the same as a proton, a spin of 1/2 and a magnetic moment of –1.913 nuclear bohr magnetons. As we can neglect consideration of the magnetic interaction, being only interested here in the behaviour of hydrogen, we can confine our attention to the neutron–nucleus interaction. Full discussion of the theory can be found elsewhere [1,2] so the present treatment is considerably simplified with emphasis on physical principles rather than mathematical rigour. The probability of a neutron interacting with a nucleus is given in terms of the microscopic cross-section, σ, defined as the rate of interaction/nucleon in unit incident neutron flux (one neutron/unit area/second). Of the possible types of interaction, we are only interested in scattering processes, for which the cross-section is σs. In general, scattering events can result in changes to both the direction and the energy of the neutron so we define a double differential cross-section, d2σ(E0 – E′, θ)/dE′dΩ to be the rate of scattering from an initial energy E0 to a unit range about the final energy E′ and through a scattering angle, θ, into unit solid angle, Ω. Assuming that the nucleus does not change its internal energy level, the scattering process would be a simple

© 2008, Woodhead Publishing Limited

Neutron scattering studies

139

two-body collision until the neutron energy is lowered to near thermal energies where the binding of the target atom in the solid becomes significant and hence influences the scattering process. It is, of course, this near thermal energy range (0–1 eV) that is of interest to us here. From the definitions above, we can write: ∞

∫ ∫ ∞



d 2 σ( E0 – E ′ , θ ) d Ω dE ′ = dE ′ d Ω





d σ( E0 , θ ) d Ω = σ s ( E0 ) dΩ [6.1]

For a fixed nucleus at the origin, this quantity is easily derived in quantum mechanical terms. If we represent the incident flux as a plane wave, we can write ψ(r) = exp (ik0 · r), where k0 is the incident wave vector and hk = mv is the momentum of neutron (with m = neutron mass and v = neutron velocity). On this basis, the neutron probability density in the incident beam, ψψ*, is unity so that the incident flux is v neutrons/unit area/s. We can now represent the scattered wave as an expanding spherical wave centred on the origin which can be written ψ′(r) = (–a/r) exp(ik′ · r). Here, k′ is the scattered wave vector and a is known as the neutron scattering length. The – sign is conventionally inserted here because, for most isotopes, the ‘compound nucleus’ formed from the neutron and the target nucleus is far from a stable nuclear state so we get ‘potential scattering’ with a 180° phase change in the scattering process. The scattering length a obviously depends on the nuclear properties (and is different for every isotope). In particular, if the nucleus has a finite spin I, the neutron–nucleus interaction is spin-dependent and so we need to define a+ and a– for parallel and anti-parallel orientations of the neutron spin relative to the nuclear spin. Here the total spin of the compound nucleus, J+/– = (I +/– 1/2), needs to be defined because it determines the statistical weight of the a+ and a– states. It should be noted that if the compound nucleus has an energy anywhere near the energy of a stable state, the values of a can vary quite widely and indeed can be either positive or negative as here there is strong coupling between the incident neutron and the target nucleus. The proton is a good example of this because in the anti-parallel state, the compound nucleus formed is very close to the ground state of the deuteron (spin zero!). This is why a– is negative and so much larger than for any other nucleus. It also follows that the scattering cross-section for an isolated fixed nucleus will be dσ/dΩ = a2 = σ/4π. We can now consider the scattering from a set of N nuclei rigidly fixed at positions ri (so that the neutron does not exchange energy with the target nucleus). If we observe the scattered waves a long way from the scattering nuclei, we can sum them and then define the scattered flux by multiplying the summed wave function by its complex conjugate and hence we obtain [1] an expression for the angular cross-section defined/nucleon:

© 2008, Woodhead Publishing Limited

140

Solid-state hydrogen storage

dσ/dΩ = (1/N)| ∑i ai exp(iQ · ri) |2 = (1/N) ∑ij aiaj exp [iQ · (ri – rj)]

[6.2]

Here, Q (= k′ – k0) can be thought of as the wave vector transport as hQ is the momentum transferred in the collision. The matrix on the right-hand side of the equation can be simplified by first averaging over the diagonal terms, yielding and then over the non-diagonal terms, where (remembering that the values of ai and aj are completely uncorrelated with the position vectors, rij) we get 2 ∑ij(i≠j) exp[iQ · (ri – rj)]. Now by reinserting diagonal terms 2 into the summation and subtracting the same quantity from the leading term, we can write: dσ/dΩ = [ – 2] + (1/N) 2 ∑ij exp[iQ · (ri – rj)]

[6.3]

Here the first term, [ – 2], is independent of the angle of scattering. It is just the sum of the scattering from each nucleus taken independently and is called the incoherent scattering. Note that this term is zero if a has the same value for all nuclei. If we integrate over a solid angle, we get an expression for the incoherent scattering cross-section, σinc = 4π [ – 2]. Also, the second term, (1/N) 2 ∑ij exp[iQ · (ri – rj)] is called the coherent scattering and we can similarly write σcoh = 4π 2. This term depends on the positions of all the atoms present taken in pairs and defines the diffraction pattern for the solid. It is analogous to X-ray diffraction except that, because the scattering is from the nucleus, there is no atomic form factor. This last expression, (6.3), has been derived for fixed nuclei but the separation process into incoherent and coherent terms carries over into all kinds of neutron scattering theory. Moreover, it is particularly important for hydrogen. Remembering that the number of quantum states associated with a compound nuclear spin of J will be 2J + 1, we find that the average scattering length taken over the two spin states of hydrogen can be written:

=

(2 J + + 1) a + + (2 J – + 1) a – (2 J + + 1) + (2 J – + 1)

[6.4]

Substituting J+ = (I + 1/2) and J– = (I – 1/2), for the case of the proton with I = 1/2, we find that = 3/4 a+ + 1/4 a–. As mentioned above, a– is large and negative while a+ is positive and not quite a third of the value of a– so that is small and negative. Numerically, this means that the incoherent crosssection is 79.7 b while the coherent cross-section σcoh (= (4π 2)) is only 1.8 b and so the proton is a strongly incoherent scatterer. On the other hand, the deuteron turns out to have σcoh = 5.6 b and σinc = 2.0 b. Hence, to use neutron diffraction to identify the position of hydrogen in a crystal structure, it is more or less essential to prepare the deuterated version of the material, because, for H, the incoherent scattering gives an intense flat background to

© 2008, Woodhead Publishing Limited

Neutron scattering studies

141

the diffraction pattern. Also, as is negative for hydrogen and positive for deuterium, it is possible to make an isotopic mixture that has exactly zero. Scattering from tritium is also of interest in some cases. It has a total scattering cross-section of 3.1 b (σcoh = 2.89 b, σinc = 0.14 b). Tabulations of the values of cross-sections and scattering lengths can be found in many places [13,14]. To complete our discussion of these general features of neutron scattering, we will formally extend the quantum mechanical treatment introduced above to include energy transfers, i.e. by assuming that the atoms in our set are held together by a set of mutual interactions. The initial state is therefore described by a particular set of quantum numbers which define the Hamiltonian or total energy of the system. As a result of an inelastic collision, any of these numbers can be changed by one or more, each integer change being associated with a quantum of energy and an equal amount of energy will be transferred to or from the neutron. This process is described by Fermi’s Golden Rule [1], which may be written: d 2 σ (E 0 – E ′, 0) = Σ i Σ j Σ n pn Σ n ′ ( k ′ / k 0 ) d E ′ dΩ

∫ ψ ( r ) a a δ [ r – ( r – r )] n

i

j

× ψ *n ′ ( r ) exp[ i ( Q ⋅ r )] d r δ ( E 0 – E ′ – E n ′ + E n )

i

j

[6.5]

In this expression, the scattering system is assumed to be in thermal equilibrium and so the cross-section has to be averaged over each possible initial state of the system n, weighted by the appropriate Bose–Einstein thermal occupation number pn and then summed over all final states n′. The delta function in r describes the short range neutron–nucleus interaction and the energy ensures energy conservation. k0 and k′ are the neutron wave vectors for the incoming and outgoing neutron respectively, and the wavefunction for the nth quantum state is ψn(r). As before Q = k′ – k0. The full usefulness of this equation may be found in the standard textbooks [1]. It will be seen that, if we separate out the parts of the equation that involve the neutron wave function, the rest of the equation is only a function of Q and ω. Van Hove [15] pointed out that this was a general property of neutron scattering cross-sections. Thus the expression for the cross-section can be reduced to the form of the scattering function which is a function of only two variables (in contrast to the three variables required to define the inelastic scattering cross-section). This relationship can be written:

d 2 σ ( E0 – E ′, θ )  σ   k ′  = S ( Q, ω )  4π   k0  d E ′d Ω

[6.6]

where S(Q, ω) is known as the scattering function, again with coherent and incoherent versions.

© 2008, Woodhead Publishing Limited

142

Solid-state hydrogen storage

6.6

Theory of inelastic neutron scattering

6.6.1

Theory of inelastic neutron scattering from solids

As described above, inelastic neutron scattering measures the energy that the system can exchange with the neutron and the probability of each exchange (see equation 6.5). Given that we are interested only in hydrogen, we can initially restrict ourselves to incoherent scattering, i.e. those terms in equation (6.5) for which i = j, where the initial and final quantum states both refer to the same nucleus. Actually, we can set this problem up in two different ways – either as the wavefunctions of individual protons at occupied sites in the lattice or in terms of phonons in a periodic lattice. The first approach works best for isolated protons which do not interact strongly with neighbouring atoms, but the method is easily adapted to the case where the proton is situated in an anharmonic potential. The second approach works well if the material has a periodic stoichiometric structure but is limited to harmonic interactions. Alternatively, we can use the deuteride and measure the coherent inelastic scattering. Here, we would normally adopt the second approach using a single crystal sample. Being a coherent process, we obtain information on the relative motion of pairs of atoms which can be described in terms of phonon dispersion curves. This is, of course, the most direct way of investigating the interactions between pairs of atoms in the crystal structure. The most direct experimental approach is to use a triple axis spectrometer on a steady state (reactor) source, the method pioneered by Brockhouse and Stewart [16]. Another approach that works for polycrystals, and has become possible through the extensive modelling software now available, is to calculate the spherically averaged (polycrystalline) cross-section for direct comparison with the experimental data and to adjust the model to improve the fit [17]. Incoherent inelastic scattering from a proton in a harmonic potential well: the Einstein oscillator model The simplest analytic model for an isolated proton in a lattice assumes that it is situated in a potential well centred on an interstitial site. This model is particularly appropriate to protons in a transition metal lattice, where the electron from the hydrogen atom can be accommodated in the d-band of the metal, but is also applicable to many other cases as well – e.g. to molecular hydrogen trapped in ion-exchanged zeolites (see Section 6.8.2 below). The model assumes that there are no interactions between neighbouring hydrogen atoms and that there is little coupling with the lattice modes. This implies that M/mp >> 1 where mp is the mass of the proton and M is the mass of the lattice atom. In transition metals, with face-centred cubic (FCC), body-centred cubic (BCC) or hexagonal close packed (HCP) lattices, the proton normally sits on either octahedral or tetrahedral sites. In more complex intermetallic

© 2008, Woodhead Publishing Limited

Neutron scattering studies

143

lattices, such as Laves phases or AB5 alloys, the site geometries can be much more complex. However, the shape of the potential well can be expanded using Taylor’s theorem. Now the first term will be a constant, the second will be zero if the origin is taken at the centre of the well and the third term is the harmonic (parabolic) term which can be separated into independent terms in the x, y and z directions ( V ( r ) = 1 / 2 m p ( ω 2x x 2 + ω y y 2 + ω 2z z 2 )) . If the higher terms in the expansion can be neglected, the wave function of the proton can be represented as the product of three independent one-dimensional Hermite polynomials where the corresponding energy levels are equally spaced with energies, En = (n + 1/2)hω. Fermi’s Golden Rule (Eq. 6.5) can now be used to calculate the cross-sections explicitly (see Elsaesser et al. [23]). The result is: S(Q, ω) = exp[–2W(Q)] exp (hω/2kT) × Σ ∞k , l , m – ∞ I k ( s k ) I l ( s l ) I m ( s m ) δ ( hω – k hω k – lhω l – mhω m ) [6.7] where k, l, m represent the change in the existing quantum numbers in each of the Cartesian directions, subject to the condition that the initial and final quantum numbers must lie between 0 and ∞. In the equation, In(s) are the modified Bessel functions where

 hQi2   hω i  si =   csch  2 kT   2mp ω i 

[6.8]

and Qi is the component of Q in the ith Cartesian direction. In equation (6.7), 2W(Q) is the Debye Waller factor in the direction, Q, which can be written: 2 W ( Q ) = Σ i (< ni > + 1 / 2 )( h / m p ω i ) Qi2 = Σ i < ui2 > Qi2

[6.9]

where is the thermal average phonon occupation number in the ith direction and < ui2 > is the mean square displacement of the proton from its mean position in the ith direction. This equation implies that if the wave vector, Q, has a component in a particular Cartesian direction i, then there is a finite probability that the proton can be transferred to any energy level in the ith direction that is consistent with energy conservation. Consideration of the modified Bessel functions show that a term involving n quanta being transferred will vary as Q2n times a form factor that is the Fourier transform of the product of the initial and the complex conjugate of the final wavefunctions of the proton. It will be noted that if the parabolic potential is known, then the cross-sections can be calculated explicitly with no selection rules that depend on the electronic structure – as is required when dealing with IR and Raman measurements.

© 2008, Woodhead Publishing Limited

144

Solid-state hydrogen storage

Perturbation analysis of anharmonic and anisotropic effects in inelastic neutron scattering from H in solids Because the proton wavefunction is not localised, it samples the potential energy surface in a volume close to the centre of the interstitial site. Hydrogen, being the lightest nucleus, samples the potential further from the centre of the site than any other nucleon (with the exception of the µ+ meson). Its wavefunction is therefore more subject to the effects of the higher terms of the Taylor expansion than any other scatterer. The allowed set of such terms depends on the symmetry of the interstitial site. For octahedral sites in a cubic system, the potential can be written – for terms up to the quartic – as: V(x, y, z) = c2(x2+ y2 + z2) + c4(x4+ y4+ z4) + c22(x2y2 + y2z2 + z2x2)

[6.10]

From this equation, we can obtain a reasonable estimate of the wavefunctions and energy levels to be expected for this potential using perturbation theory [19]. The energy levels are given by the expression: Eklm = (hω0/2) + β(j2 + j + 1/2) + γ[(2k + 1)(2l + 1) + (2l + 1)(2m + 1) + (2m + 1)(2k + 1)]

[6.11]

with j = k + l + m. From this, we can derive a set of energy transfers: ε100 = ε010 = ε001 = hω0 + 2β + 4γ (fundamental terms) ε200 = ε020 = ε002 = hω0 + 6β + 8γ (first harmonic) ε110 = ε011 = ε101 = 2hω0 + 4β + 12γ (combination vibration) where these parameters are related to the parameters defining the potential by ω0 = √2c2/m, β = 3hc4/4mc2, γ = h2 c22/c2m Here, the notation ε110 implies that the Cartesian operators have been raised by unity in the x and y directions and this implies that Q has to have finite components in the x and y directions. This result implies that the energy levels are no longer evenly spaced and that the intensity of the corresponding inelastic peak will vary strongly with the direction of Q relative to the crystal axes. This model works well for α-Pd/H [20]. Here the respective energy levels were observed at 69, 138 and 156 meV, yielding the following values for the parameters: hω0 = 50 meV, β = 9.5 meV and γ = 0. It should be noted that these parameters imply rather too large a perturbation for perturbation theory to work reliably. Comparison with ab initio calculations Over the past decade or so, ab initio calculations of lattice dynamics using density functional theory (DFT) have become increasingly accurate. In these

© 2008, Woodhead Publishing Limited

Neutron scattering studies

145

calculations, the total energy of an assembly of atoms is calculated using DFT to represent the effects of electron–electron correlations. Then the nuclear coordinates are relaxed so as to minimise the total energy. As long as the initial configuration of the atomic positions is sufficiently accurate, the relaxed structure should coincide with the real system. The process of relaxing to the minimum energy effectively provides force constant data for dynamical calculations for direct comparison with neutron scattering data. The situation is somewhat more complicated when we consider proton dynamics because the quantum properties of the proton have also to be calculated. This is done in practice using the Born–Oppenheimer approximation, namely that motions on different timescales can be separated. Specifically, this means that we can solve the Schrödinger equation for the electrons over a whole range of different proton positions so that the variation of the total energy of the system with proton position yields a potential energy surface within which to calculate the wavefunctions and energy states of the proton. This method, sometimes referred to as the ‘frozen phonon’ method, was pioneered by Ho et al. for the NbH1.0 system [21,22]. The calculation was done for a periodic lattice and the protons were all moved in phase with each other but in anti-phase with the metal atoms (the zone centre phonons). The various excited states of the proton were calculated and agreed well with experiment. The approach was extended to the Pd-H system by Elsaesser et al. [23] and to a single crystal Pd-H0.85, by Kemali et al. [24] where full ab initio calculations of the inelastic scattering cross-section were compared with experimental neutron scattering measurements. Having determined the full three-dimensional wavefunctions of the proton, the full inelastic cross-section was calculated using equation (6.5). For a harmonic spherically symmetric wavefunction, the energy levels would be fully degenerate and the scattering would be spherically symmetric. However, for an anharmonic system such as Pd-H, the (1,1,0) level is split from the (2,0,0) level which is itself split into a singly degenerate state, |C> and two degenerate states, |A> and |B>. These energy levels can be easily identified using their directional dependence: the (1,1,0) level gives maximum scattering in the (110) and (111) directions and is zero in the (100) direction, while the (2,0,0) level gives a maximum in the (100) direction (because Q lies entirely in one Cartesian direction, only one Cartesian wavefunction can change its energy level). Similarly, the (111) level has its maximum in the (111) direction. The experimental energy levels agree pretty well with the calculated values (Fig. 6.1). Incoherent scattering from systems with interacting protons The discussion above has been concerned with systems in which the protons do not interact significantly with each other so that we can analyse the scattering on the basis of scattering from individual protons in a fixed potential

© 2008, Woodhead Publishing Limited

146

Solid-state hydrogen storage 0.10

S(Q, ω)

0.00 [111]

0.06 0.04

[110]

0.02 [001]

0.00 0

40

80

120

160 200 240 280 Energy transfer (meV)

320

360 400

6.1 The experimental and calculated incoherent scattering functions S(Q,ω), against the energy transfer (points represent the data, lines represent the theory). For clarity, the spectra for the [110] and the [111] directions have been shifted vertically. The calculated spectra have been convolute with the experimental resolution function [24].

energy well. However, for crystalline systems in which there is significant interaction, we have to analyse the scattering in terms of the phonon dispersion curves. In this approach, which is limited to harmonic potentials, we describe the motion of the atoms in terms of their periodic displacements in the crystal, i.e. plane waves with wave vectors q. The full theory can be found in the standard textbooks [1,18]. By using the periodicity of the lattice, the equations of motion for a given phonon yield the ‘dynamical matrix’. This matrix is of dimension 3n, where n is the number of atoms in the unit cell and the 3 represents the number of Cartesian axes. Each element, i, j, in this matrix is determined by the force constant relating the relative displacements of two atoms, each in a particular Cartesian direction. This matrix equation can be solved, yielding 3n eigenvalues, ωs(q), which are the frequencies of the 3n solutions for the particular q used. For each ωs(q), there is an eigenvector, representing the corresponding displacements of the n atoms in the unit cell in each Cartesian direction. These displacements are in general complex numbers, implying that the phase of the displacements of a given atom may differ from the phase of the phonon solution assumed. Now we can plot the 3n ‘dispersion curves’, ωs(q), as a function of q. We can also derive the corresponding neutron scattering cross-section for the creation or annihilation of a phonon which now depends on satisfying both energy conservation and momentum conservation, where the phonon has an energy hω and a momentum hq: hω = E0 – E′ and Q = k0 – k′ = ␶ + q

© 2008, Woodhead Publishing Limited

[6.12]

Neutron scattering studies

147

Here ␶ is a reciprocal lattice vector. The intensity of the scattering at a particular value of q is determined from the relative vibration amplitudes of the atoms taken in pairs over the unit cell and weighted by the scattering lengths of each atom. The classical method of using neutrons to measure these dispersion curves involves the use of single crystal samples and a triple axis spectrometer [1,16]. However, we are mainly interested in samples where the scattering will be dominated by hydrogen and, as we have seen, the coherent scattering of hydrogen is small compared with the incoherent cross-section. Hence the inelastic scattering will be dominated by the self (diagonal) term in the full cross-section. All the information about the conservation of momentum disappears and we are left with a hydrogen vibrational amplitude-weighted contribution from each phonon mode in the unit cell. Of course, it would be possible to use deuterium but if we did, we would still have the problem of making single crystals of hydrides (deuterides). As most hydrides are formed in the solid state by a first-order transition involving a significant lattice parameter change (with the notable exception of palladium [24] and a couple of other systems), it is therefore generally impossible to get adequate single crystals. We are therefore forced to use hydrogen and analyse the incoherent scattering. Here we can show [1] that the incoherent inelastic scattering cross-section can be written: 2 d 2 σ inc = k ′ [< n ( ω ) + 1>] Q Θ ( ω ) 2ω dω dΩ 4 π k0

[6.13]

where we have replaced (E′ – E0) by hω, is the thermal population of modes of frequency, ω, and Θ(ω) is the amplitude-weighted density of states given by: Θ ( ω ) = f ( ω ) Σ ρ ( σ ρinc / M p ) exp (– Q 2 < uρ2 >) Σ j | eρs ( ω ) | 2 [6.14] Here f(ω) is the normalised number of phonon modes of frequency ω summed over all phonon modes, s, and wave vectors, q, in the Brillouin zone of the crystal, eρs ( ω ) is the displacement of the ρth hydrogen atom, of mass Mp, in the sth mode summed over all the phonons of frequency ω, and < uρ2 > is the mean square amplitude of displacement of the ρth hydrogen averaged over all modes. For the case of a binary hydride with two atoms/unit cell, there are six modes. The first three of these are ‘acoustic’, that is to say that the metal atom and the hydrogen atom are roughly in phase with each other and that both have about the same amplitude of vibration. The other three modes, however, are ‘optical’, i.e they are roughly in antiphase but with an immobile centre of gravity. This means that the square of the H vibration amplitude increases in proportion to the mass of the metal atom. Thus, the scattering from the optical modes will normally outweigh that from the acoustic modes

© 2008, Woodhead Publishing Limited

148

Solid-state hydrogen storage

by more than an order of magnitude. Further, if there is no interaction between adjacent H atoms, the frequency range within a given dispersion curve will vary as Mp/Mm where Mm is the mass of the metal atom. For heavy metals, this range is small and so the scattering appears as a rather narrow intense peak. However, for lighter metals, particularly where H–H interactions can be significant, the spectral shape can become much more complex. However, with the recent development of ab initio calculations, it has become possible to calculate the inelastic scattering cross-section directly for comparison with neutron scattering data. Examples of such measurements will be given below.

6.6.2

Inelastic neutron scattering from molecular hydrogen

An exception to the above considerations of the incoherent scattering from hydrogen comes in the case of molecular hydrogen where, owing to the light mass of the atoms and the weak interaction with the surroundings, the overall wavefunction of the molecule has to be anti-symmetric. Thus, if the nuclear spins are anti-parallel (para-hydrogen), the angular momentum, J, must be even – whereas, for parallel spins (ortho-hydrogen), it must be odd. The energies of the rotational states for a free hydrogen molecule are given by: EJm = EJ = BJ(J + 1)

[6.15]

where B = h 2 /2 µd e2 is the rotational constant, = 7.35 meV for H2. Here, µ is the reduced mass for the molecule, de is the equilibrium separation of the H nuclei (0.0741 nm) and mh is the component of the angular momentum parallel to the quantisation axis where m = –1,0,1 for the J = 1 level. It is obvious that this association of the relative nuclear spin states with the angular momentum state is contrary to our assumption above that the spin direction (and hence the scattering length) of a given proton is independent of its surroundings, and this has a profound effect on the neutron scattering from molecular hydrogen. The full theory of this process has been given by Young and Koppel [3]. These authors showed that the scattering can be divided up into terms corresponding to transitions between different J states – if J′–J is even, then the neutron spin stays unchanged but if J′–J is odd, then the neutron has exchanged spin with one of the nuclei and the nuclear spin of the molecule is changed from odd to even (ortho to para) or vice versa. Also, of course, if J changes, the neutron has to lose or gain the corresponding quanta of rotational energy. Thus, the cross-sections can be written:

© 2008, Woodhead Publishing Limited

Neutron scattering studies

149

d 2 σ J ,J ′ = ( k / k 0 ) FJ – J ′ ( Q ) exp [–2 W ( Q )] δ ( E – E ′ + E j – E j ′ ) d Ω dE′ ∗ Smol(Q, ω)

[6.16]

where ∗ represents a convolution, the FJ–J′ (Q) are the form factors for the rotational transitions and Smol(Q, ω) is the scattering function for the molecule centre of mass. Explicitly, the first few rotational transition form factors can be written: F0–0 = 4(σcoh/4π) j 02 F0–1 = 12(σinc/4π) j12 F1–0 = (4/3)(σinc/4π) j12 F1–1 = 4[(σcoh/4π) + (2/3)(σinc/4π)] (2 j 22 + j 02 )

[6.17]

Here σcoh and σinc are as defined above for the hydrogen nucleus and j0, j1 and j2 are the zero, first and second spherical Bessel functions with argument Qde/2. Remembering that σinc = 79.7 b and σcoh = 1.8 b, it is clear that for para-hydrogen in its ground state, the scattering will be dominated by the J = 0 to J′ = 1 (or J = 0 to J′ = 3, 5…) neutron energy loss transition. As the J = 0 state has zero rotational energy and the J = 1 has an energy J(J + 1)B = 14.7 meV, the neutron energy loss cross-section will be dominated by a delta function at 14.7 meV. The above expressions for the rotational form factors are based on the assumption that the molecule is situated in a spherical potential so that the different m states are degenerate. However, if the molecule is trapped in a perturbed potential well, the degeneracy of the m levels will be lifted. Thus, for ellipsoidal perturbation of the J = 1 state, we expect to find the m = ±1 levels split from the m = 0 state. The energy level diagram for this situation can be found in, for instance, Mitchell et al. [7]. Simple perturbation theory would suggest that the mean energy should remain fixed at 14.7 meV. If the mean energy decreases, we can interpret this to mean that the H–H distance (2de) has increased owing to interaction with the trapping medium. As can be seen in equation (6.16), this rotational transition part of the cross-section has to be convoluted with the scattering function for the molecule centre of mass. If the molecule is rigidly bound, Smol (Q, ω) = δ(ω). If it is free to recoil, its cross-section will be as for scattering from a perfect gas. The assumption here is that the initial state of the molecule has a defined momentum taken from a Maxwell–Boltzmann distribution and that all final states are available for the recoiling molecule. In this case, the molecular scattering function can be written:

© 2008, Woodhead Publishing Limited

150

Solid-state hydrogen storage

 ( hω – h 2 Q 2 / 4 M ) 2  SPG ( Q, ω ) = ( M / π h 2 Q 2 k B T )1/2 exp  – ( h 2 Q 2 k B T / M )   [6.18] where M is the mass of the molecule, kB is Boltzmann’s constant and T is an effective gas temperature. It will be noted that this has a Gaussian shape, with its mean value displaced to the average recoil energy, h2Q2/4M. Note that here Q is a scalar because the scattering is isotropic. If, on the other hand, the molecule is trapped in a harmonic potential well, the simple harmonic oscillator scattering function SSHO(Q, ω) will take the usual form [18] and the neutron will scatter the molecule into one of a series of equally spaced energy levels, the spacing being proportional to the steepness of the parabolic well. At low temperatures there will be a probability that it starts and finishes in its ground state. Here, the cross-section for elastic scattering will vary as exp(–Q2)δ(ω) where is the mean square displacement of the molecule in the potential. However, if the molecule is trapped in a sufficiently shallow potential well it ends up in one of a number of closely spaced energy levels. Summing the scattering over these closely spaced states will give a continuous distribution that will approximate to the perfect gas distribution. The total scattering function for this system can be well represented by: SSHO (Q, ω) = exp(–Q2)δ(ω) + [1 – exp(–Q2)] SPG(Q, ω) [6.19] Now, this expression can be substituted into (6.16) to yield the full expression including both a rotational transition combined with the molecular centre of gravity response. In particular, if the molecule starts on the para-ground state (J = 0), the scattering will be dominated by the transition to the J′ = 1 (ortho) state. Now, performing the convolution integral, the scattering will consist of a peak at J(J + 1)B (= 14.7 meV) due to elastic scattering from the molecule and a broad Gaussian peak centred on J(J + 1)B + h2Q2/4M. Similar terms will appear at larger energy transfers for J′ = 3,5,… etc. Because the transition rate between the ortho- and para- states is low except in the presence of magnetic catalysts, we can equilibrate the gas at a given temperature in the presence of a magnetic catalyst. For low temperatures, in the presence of a catalyst, the system will become virtually entirely para while for high temperatures it will tend to 25% para and 75% ortho. We can therefore do neutron scattering experiments with different predetermined ratios of the two forms so that both cross-sections can be extracted.

6.6.3

The theory of quasi-elastic neutron scattering

The treatment above refers to the probability of the neutron causing a transition of the scattering system from one fixed quantum state to another. In contrast

© 2008, Woodhead Publishing Limited

Neutron scattering studies

151

to this, quasi-elastic scattering describes a situation where the scattering nucleus is free to diffuse around in space and this causes a kind of Doppler broadening, the final neutron velocity (energy) acquiring a continuous range of values symmetrically around the original velocity (energy). Quasi-elastic scattering therefore provides us with a very direct technique for studying diffusion of an atomic species in either a liquid or a solid phase. Given that, again, hydrogen is easily studied, it is a potentially useful way of investigating dynamic mechanisms in hydrogen storage systems. The basic formulation of this problem was given by Van Hove [25] in the form of his space–time correlation functions, Gs(r, t) and G(r, t). He showed that the scattering functions, as defined above, for a diffusing system are given by the Fourier transformation of these correlation functions in time and space. Incoherent scattering is linked to the self-correlation function, Gs(r, t) which provides a full definition of tracer diffusion while coherent scattering is the double Fourier transform of the full correlation function which is similarly related to chemical or Fick’s law diffusion. Formally the equations can be written: S inc ( Q, ω ) = (1/2 π )

∫∫ G ( r , t ) exp [i ( Q ⋅ r + ωt ) dr dt

S coh ( Q, ω ) = (1/2 π )

∫∫ G ( r , t ) exp [i ( Q ⋅ r + ωt ) d r dt

s

[6.20]

Here, Gs(r, t), the self-correlation function, is defined to be the probability of finding a nucleus at position r at time t if that nucleus was at the origin at t = 0. Similarly, G(r, t) is the probability of finding an atom at position r at time t, if any atom was at the origin at t = 0. Strictly speaking, the scattering functions are not symmetric in ω because the probabilities of neutron energy gain is related to energy loss by the Boltzmann factor (detailed balance), exp(–hω/kT), but so long as hω 12 MPa. Also Chambers et al. [56] claimed that tubular, platelet and herringbone forms of carbon nanofibre were capable of adsorbing in excess of 11%, 45% and 67% respectively at room temperature and 12 MPa. While these estimates proved to be rather high, probably because of difficulties in removing all the water vapour from the apparatus and of estimating the purity of the samples, nanotubes have proved to be a very useful model system for studying the energy of adsorption of hydrogen molecules on sites of well-determined geometry. The first INS measurements on H2 adsorbed on SWNT were reported by Brown et al. [57] who observed a broad peak at an energy transfer of 14.5 meV having a width of about 2.4 meV which was about twice the instrumental resolution (1.1 meV) and was independent of temperature. The desorption of H2 with increasing temperature suggested a trapping energy of around 60 meV. The peak was attributed to H2 trapped on the convex external surface. Subsequently Ren and Price [58] observed a similar broad peak but with a width that increased linearly with temperature from 4.2 and 35 K. From their data, they concluded that H2 was physisorbed in the interstitial tunnels in the SWNT bundles. Subsequently, a similar sample was examined in some detail by Georgiev et al. [59,60] using the TOSCA inverse geometry spectrometer at ISIS, Rutherford Appleton Laboratory, UK (Fig. 6.5). This instrument provides probably the best available resolution in the 15 meV energy transfer range and, having good statistics, was capable of observing the splitting of the rotational peak. The forward and backward detector banks on this instrument give two distinct Q values. The sample was held at 20 K and was loaded with incremental amounts of hydrogen so that if different sites are present giving different spectra, the relative occupation of the two sites can be determined as a function of the total H2 adsorbed. For low coverage, these data clearly show the peak to be split into two components, the higher energy peak at about 15.1 meV having about half the area of the lower energy peak at about 13.5 meV. This suggests that the higher energy peak corresponds to the J = 1, m = 0 state while the lower energy peak corresponds to the J = 1, m = ±1 state. This is as would be expected for a molecule trapped in a potential well with a fairly narrow minimum in the direction normal to a surface but quite wide parallel to the surface. As the coverage is increased, a new broad peak appears first as a single peak at 14.5 meV and then, above 100% coverage, as a broader peak that requires two Gaussian peaks centred on 14.2 and 14.6 meV to fit it properly. The temperature dependence of the desorption suggests that the split peak has a trapping energy of around 80 meV while the single peaks

© 2008, Woodhead Publishing Limited

Neutron scattering studies

161

S(Q, ω)

(002)

0.5

0

100

1.0

1.5

200 300 400 Neutron energy loss (cm–1) (a)

3.5

500

600

Empty carbolex 30%, 17 K 50%, 17 K 80%, 30 K 100%, 17 K 144%, 17 K

3.0 2.5

S(Q, ω)

2.0

2.0 1.5 1.0 0.5 0.0 0

10

11

12

13 14 15 16 Energy loss (meV) (b)

17

18

19

20

6.5 Inelastic neutron scattering measurements on H2 adsorbed on single walled carbon nanotubes measured at 20 K on the TOSCA spectrometer at ISIS, Rutherford Appleton Laboratory, UK. (a) The full energy spectrum with the diffraction pattern measured from the nanotube sample as an inset. (b) The energy spectra on an expanded scale showing the rotational peak for a series of different hydrogen coverages. (c) The variation of the areas of fitted Gaussian peak with H2 surface coverage [59, 60] (see page 162).

© 2008, Woodhead Publishing Limited

162

Solid-state hydrogen storage 2.2

E = 13.5meV E = 14.2 meV E = 14.6 meV E = 15.1 meV

2.0

Peak integral intensity

1.8 1.6 1.4 1.2 1.0

Rotationally free H2

J = 0,m = 0 to J = 1,m = ±1 transition of the perturbed adsorbate

Perpendicular vibration?

0.8 0.6 0.4

J = 0,m = 0 to J = 1,m = 0 transition

0.2 0.0 –0.2 –20

0

20

40 60 80 100 Surface coverage (%) (c)

120

140

160

6.5 (Continued)

have a much lower trapping energy of around 40 meV. The authors concluded that the more strongly trapping site was along the groove sites on the surfaces of the bundles while the single peak at 14.5 meV was due to trapping on the convex surfaces. They also suggested that above 100% coverage, a second layer forms over the convex surface, which is responsible for the new 14.6 meV peak, while the first layer, now being more strongly trapped, shows its peak at 14.1 meV. It will be noted that the weighted mean of the split peak is at 14.0 meV, significantly lower than the free molecule value (14.7 meV). This suggests that the H–H distance has been increased by the interaction with the surface. We may note that Raman scattering studies of H2 on SWNTs [61] showed slight shifts in the stretch frequencies (Q band), showing one component with an upshift relative to the gas molecule of around 0.23 meV associated with a trapping site and a downshift of –0.13 meV associated with a convex surface site. The neutron scattering data yields other valuable information. By fitting the Young and Koppel model [3] to the recoil humps, the data give accurate values of the effective Maxwell–Boltzmann temperatures of the trapped molecules while a comparison of the forward and backward spectra yield values of the Debye–Waller factor and hence of the mean square displacement of the molecule in its site. As expected, when the mean square displacement decreases, the effective temperature increases. Thus, as the H2 coverage increases, there is firstly a small increase in the effective temperature from 130 K as molecules in the groove site are pushed closer together. Then, as molecules start to trap on the convex surface, the average temperature begins

© 2008, Woodhead Publishing Limited

Neutron scattering studies

163

to decrease, reaching 80 K at 140% coverage. Presumably at even higher coverage, the average temperature would asymptotically approach the physical temperature of the surface. Very similar results were reported by Schimmel et al. [62], except that these authors found that the higher energy peak observed at 14.72 meV has almost twice the area of the low energy component observed at 13.63 meV. They also concluded that the split peak is due to H2 adsorbed on groove sites. Also, Liu et al. have reported measurements on H2 adsorbed in boron-substituted SWNTs [63]. These studies, with the lower-resolution FANS spectrometer at NIST, showed the same constituent peaks at 13.5 and 15.1 meV with little difference due to the boron loading, possibly because only a few per cent of B atoms were incorporated in the SWNT structure. The experiment also compared the spectra from arc and laser-produced samples, the former having somewhat broader peaks, attributed to a greater range of tube diameters and other defects in this sample. Other high surface area carbon systems have since been investigated. Georgiev et al. [64] observed H2 on activated carbons derived from pine wood. Here, at low coverages, a pure recoil spectrum is observed, suggesting completely free recoil on a graphene surface. At higher coverage, a broadened peak appears and this is attributed to H2 in nanoscale slit pores where the adsorption energies on both surfaces augment each other. Schimmel et al. [65] reached similar conclusions from measurements on activated charcoal, carbon nanofibres and SWNT samples using a lower-resolution spectrometer at the Delft reactor.

6.8.2

Hydrogen trapped in zeolites

Zeolites are compounds of Si, Al and O which consist of a framework containing a regular array of cavities. The framework has a net negative charge which is balanced by the ‘exchange’ cations which are held electrostatically within the cavities and can therefore be exchanged with other cations using an appropriate aqueous solution. Zeolites have also been investigated as possible hydrogen storage media [66,67]. For these systems, adsorption isotherms measured at 80 K absorb up to 2.2% by weight in the case of Ca-X. Actually, the amount adsorbed in different zeolites depends on the size of the cavities and the exchange ion. Experimental data have been reviewed and compared with molecular dynamics simulations by Vitillo et al. [68]. Although the amounts of hydrogen adsorbed are not very promising for practical applications, they have proved to be useful test bed systems for understanding the nature of the H2 adsorption process. Thus, for instance, data due to Jhung et al. [69] for H2 adsorption in Y zeolite exchanged with H, Na, K and Cs at 80 K, imply that the heat of adsorption is in the range 5.7– 6.6 kJ/mole for these systems, showing a tendency to increase with the

© 2008, Woodhead Publishing Limited

164

Solid-state hydrogen storage

S(Q, ω)/arb. units

electrostatic field within the zeolite and also to decrease as the amount of hydrogen adsorbed increases. Inelastic neutron measurements have been performed on TOSCA for H2 on zeolite-X having Na+, Ca2+ and Zn2+ exchange ions (Fig. 6.6) [70,71]. Measurements were performed on para-H2 in zeolite X exchanged with Na+, Ca+ and Zn+. These spectra show a new feature of the inelastic scattering, namely the series of uniformly spaced peaks above the complex peak derived from the perturbed rotational states around 14.7 meV. It is clear that these peaks represent the vibrational states of the H2 molecules which are convoluted with the rotational scattering as shown in Eq. 6.16. Also, clearly, a set of equally spaced levels suggests that the molecules are located in rather well-defined parabolic potentials. It will also be noted that the spacing and therefore the steepness of the parabola increases in the sequence Na+, Ca2+, Zn2+. Indeed, the spacing is proportional to the polarising potential of the

0

20 40 60 Neutron energy loss (meV)

80

6.6 The INS spectra of para-H2 in (from top to bottom) zeolites Na–X, Ca–X, Zn–X. Also shown are the spectra decomposed into Gaussians. The Zn–X and the Na–X spectra had their backgrounds removed using a smooth function. These spectra are all measured at low H2 uptakes of around 0.15 wt% [71].

© 2008, Woodhead Publishing Limited

Neutron scattering studies

165

cations, given by Z/r, where Z is the charge on the cation and r is the ionic radius. Recently, remarkable results have been reported on Cu exchanged ZSM5 [72–74]. Firstly, Ramirez-Cuesta and Mitchel [72] reported a split rotational level at approximately 12 and 14 meV, the 12 meV peak decreasing in energy as the H2 loading was increased. The reduction in the mean energy of the rotational level was attributed to a small increase in the H–H distance, giving rise to a reduced rotational constant, an indication that there is a significant electronic interaction with the Cu ion. Subsequently, Georgiev et al. [73] reported INS measurements on the IN5 spectrometer at the ILL, showing complex tunnelling peaks in the energy range up to 2 meV energy transfer. Here, the sample had been carefully outgassed in a vacuum in order to convert the Cu ions to their monovalent state. It is not clear whether the observed scattering is from ortho- or para-hydrogen. Unfortunately, the TOSCA and IN5 energy transfer ranges do not coincide so it is difficult to establish whether both sets of peaks at ~2 and 12–14 meV coexist. However, it is clear that the trapping energy for the hydrogen is remarkably high for Cu+. This strong interaction also shows up in the IR measurements where the H2 stretch frequencies in this system are at 3070 and 3125cm–1 compared with 4161cm–1 in the unperturbed molecule. In contrast, samples exchanged with alkali metals show downward shifts of this frequency of only 50–90 cm–1 [73], supporting the idea that the interaction with Cu+ is dramatically higher. To confirm this, in subsequent observations on hydrogen adsorption measured volumetrically, Georgiev et al. [74] found heats of adsorption of 75 kJ/mole at low coverages decreasing to 40 kJ/mole at higher coverages (0.4 H2/Cu atomic). These values are remarkably an order of magnitude higher than values reported for other zeolite systems [75], suggesting that this is an example of non-dissociative chemisorption. However, as mentioned before, although the heat of adsorption is close to practical values, the gravimetric adsorption is too low to be useful.

6.8.3

Hydrogen trapped in ice clathrates

Recently there has been considerable interest in the use of ice clathrates as hydrogen stores. Under kilobar pressures, it was possible to get (32 + x) H2 molecules into the clathrate cubic structure II containing 136 H2O molecules based on there being one H2 in each of the 16 small cages and four (x = 16) in each of the 8 larger cages up to a decomposition temperature of 180 K [76]. At higher temperatures, it is necessary to use a larger guest molecule such as tetrahydrofuran (THF) in the larger cage to stabilise the structure up to its melting point and at 50 bar [77]. In this case, only the smaller cage is occupied by H2 molecules with one molecule/cage. This latter system has been studied with INS with interesting results [78]. To minimise the scattering

© 2008, Woodhead Publishing Limited

166

Solid-state hydrogen storage

from the host, both the ice and the THF were deuterated. Again, the results could be interpreted in terms of the convolution of the perturbed J = 0 to J = 1 rotational level with the energy levels of the H2 molecule in the small cage. However, in contrast to the zeolite case above [72], the potential well is very anharmonic, which is not surprising given the rigid cage structure. The authors modelled this potential by adding model potentials between the molecule of H2 and the host structure over the 512 nearest D2O molecules and then solving Schröedinger’s equation for H2 in the resulting flat-bottomed potential well. The agreement of the experimental peaks with this model is remarkably good.

6.8.4

Molecular hydrogen storage in metal oxide frameworks

Metal–organic framework compounds (MOFs) constitute a relatively new class of porous compounds, which employ molecular building units with predefined functions and geometries which are used to assemble periodic solids containing a regular array of pores of controllable volume with interconnecting tubes with controllable diameters. Thus, for instance, frameworks build by linking octahedral Zn4O(O2C–)6 groups yields some of the highest porosities known [79]. The metal oxides that link the organic clusters also create trapping sites for hydrogen that could in principle provide the molecular chemisorption that is necessary to yield the right kind of trapping energy. Of the huge number of MOFs now known, only a small fraction have been checked for hydrogen uptake [80 and references therein] and of these only a few have been investigated using INS. The majority of these measurements have been made with the QENS spectrometer at IPNS at the Argonne National Laboratory. In these experiments the sample was kept at low temperatures but the hydrogen was not converted to the para state before admission to the sample capsule so it is presumed that it contained about 75% of ortho-hydrogen. The first MOF to be measured (MOF-5 = IRMOF1) [81] shows peaks at 10.3 and 12.1 meV at low loadings which are attributed to the J = 0–1 transitions on two different sites, the former being attributed to a site at the Zn atom and the latter to sites on the benzenedicarboxylate (BCD) linking molecules. The 10.3 meV band is presumed to correspond to the more strongly adsorbing site because at higher loadings, the 12.1 meV peak increases in intensity. At higher loadings still, the 12.1 meV line seems to break into four components, suggesting a number of different sites on the linkers. The model for the J = 1 energy levels involved assumes that the energy is lower for an H2 orientation parallel to the plane (m = ±1) as compared to one normal to the plane (m = 0) [82]. This suggests that the peaks observed correspond to transitions from the J = 0 level to the J = 1, m = ±1 level and that therefore there should be

© 2008, Woodhead Publishing Limited

Neutron scattering studies

167

a further peak for the J = 1, m = 0 level beyond the range of the measurement (>18 meV). An alternative explanation would be that there is little orientational dependence but that the rotational level is considerably lowered by a significant increase in the H–H distance. However, given the presence of ortho-H2, we might also expect to see the transition between the J = 1, m = ±1 level to the J = 1, m = 0 level so the interpretation must remain tentative. In a subsequent paper [80], the measurements were extended to IRMOF-8, IRMOF-11 and IRMOF-177. The measured spectra for these compounds behave in a similar way to IRMOF-1, in that, at low loading, there are two single peaks corresponding to two sites, which the authors attribute to (a) sites adjacent to the Zn atom (stronger binding – peak at lower energy) and (b) sites on the organic linker (weaker binding, higher energy and with relatively higher intensity at higher loading). For higher loadings, the spectra become even more complex than in IRMOF-1, as is to be expected for these somewhat less symmetric systems.

6.9

Quasi-elastic scattering measurements on hydrogen diffusing in hydrides

6.9.1

Quasi-elastic scattering studies of hydrogen in Laves phases

The most notable series of measurements of quasi-elastic neutron scattering in intermetallic hydrides in recent years has been that by Skripov and coworkers on hydrogen diffusion in cubic Laves phases (C15-type). In these systems, quasi-elastic neutron scattering measurements show that the hydrogen has two types of jump motion with different characteristic frequencies [83]. The faster jump motion corresponds to localised jump motions while the slower motion gives rise to long range diffusion. This corresponds to the case discussed on page 153. If only the localised motion were present, there would be an elastic structure factor with a Q-dependence determined by the Fourier transform of the H–H correlation function at long time. The remainder of the peak intensity would be quasi-elastic with an energy width determined by the inverse of the mean time between jumps on the hexagon. If the jumps were between two sites, the width of the quasi-elastic peak would be independent of Q. For larger numbers of sites, it will in general be made up of Q-dependent proportions of several Lorentzian broadenings of fixed widths, as determined by the localised site geometry. Now, if we introduce a slower long-range diffusion, then the localised S(Q, ω) is convoluted with the appropriate version of the Chudley–Elliott model. In practice, this means that the elastic peak becomes broadened by an amount that increases with Q2 at low Q, thus defining the tracer diffusion coefficient for hydrogen in the system. In most of the compounds examined, of composition AB2Hx, at low

© 2008, Woodhead Publishing Limited

168

Solid-state hydrogen storage

and intermediate hydrogen concentrations (up to x~ 2.5) H occupies only the tetrahedral g sites, with A2B2 coordination, while the tetrahedral e sites (with AB3 coordination) only begin to fill at higher H concentrations. The g sublattice consists of hexagons lying perpendicular to the direction. Each g site has three nearest neighbours, two g-sites on the same hexagon at distance r1 and one on an adjacent hexagon at distance r2. The ratio r2/r1 depends on the actual H site, which, in turn depends on the relative ratio of the metallic radii, rA and rB. For most cases investigated, [84–89], rA/rB ≤ 1.25 and here r1 ≤ r2 and, in consequence, there is a fast localised diffusion around the g sites on the hexagon and long-range diffusion that depends on the slower jumps between hexagons . However, for the case YMn2 [90], rA/rB = 1.425 and hence r2 < r1 and so the localised diffusion is between g sites on adjoining hexagons and long-range diffusion results from jumps around a given hexagon.

6.9.2

Quasi-elastic scattering from hydrogen in alanates

Now, turning to quasi-elastic scattering measurements on complex hydrides which offer the possibility of higher gravimetric H contents, we find that the measurements are rather difficult because the jump rate of the hydrogen atoms is either too slow or, only a small fraction of the atoms take part in the motion (vacancy or defect diffusion). Moreover, the material is extremely sensitive to atmospheric moisture, which creates practical difficulties. To date, the most interesting measurement is that due to Vegge and co-workers [91–93] on sodium alanates, NaAlH4 and Na3AlH6. For these materials, it has been found that activation using TiCl3 considerably accelerates the hydrogen absorption/desorption rate [94]. The samples were measured using the backscattering spectrometer at KFA Jülich, Germany. For both compounds, with or without TiCl3, it was possible to detect any quasi-elastic scattering only at quite high temperatures. For NaAlH4, doped or undoped, there was no detectable quasi-elastic scattering below 315 °C. In fact, only around 0.5% of the hydrogen was mobile at 390 °C in the undoped sample. For Na3AlH6, at 350 °C, only the doped sample showed any broadening (2% of atoms) while at 390 °C, the undoped sample showed significant quasi-elastic scattering (13%) with clear indication of the Chudley–Elliott form with jump lengths of around 2.8 Å. Taken in conjunction with the ab initio calculations, it is inferred that for NaAlH4, there is a localised motion involving the transfer of an H atom from an (AlH4)– ion to a faulted AlH3 cluster. Activation energies for further jumps are significantly higher than this localised motion. As expected, the QENS shows a broadening that is more or less independent of Q, confirming a localised motion. For the Na3AlH6 case, as mentioned, the doped sample broadening displayed substantial Chudley–Elliott model behaviour, which is consistent with the ab

© 2008, Woodhead Publishing Limited

Neutron scattering studies

169

initio calculations which showed that, assuming that the Ti atom was on an Al site, there is a lower barrier to long-range diffusion (~ 0.36 eV) but for only a few per cent of the hydrogens. These results are fully consistent with the rather low rates of H2 evolution even for finely divided material at high temperatures.

6.9.3

Quasi-elastic broadening due to H2 diffusion on surfaces

Finally, in this section, we should note that QENS can be used to study the diffusion of H2 on surfaces. This has been demonstrated by Fernandez-Alonso et al. [95] who adsorbed hydrogen direct from room temperature (i.e. 75% ortho-H2) onto a sample of nanohorns and observed the quasi-elastic scattering using IRIS at ISIS. This spectrometer provides 9 µeV resolution in the quasielastic region but also allows one to make inelastic measurements around the rotational level at 14.7 meV with a resolution of about 70 µeV. The quasielastic measurements were possible because the sample contained ortho-H2. Measurements were made at 25 and 15 K. The quasi-elastic intensity decreased with increasing Q as expected owing to the size of the H2 molecule and the mean square displacement of the molecule on the surface. The broadening was fitted with a jump diffusion model, yielding values of the tracer diffusion coefficient of 0.96 Å2 ps–1 at 15 K and 6.5 Å2 ps–1at 25 K with corresponding jump lengths of 5.4 Å at 15 K and 6.81 Å at 25 K. This fit was obtained by allowing for an elastic contribution to the peak with intensity about four times the quasi-elastic peak. This was attributed to an immobile fraction although some extra elastic component might be expected as a result of averaging over the orientation of Q relative to the surface normal. In principle, because of the convolution in equation (6.16), the inelastic peak around 14.7 meV should also be quasi-elastically broadened due to the motion of the molecule centre of mass but the data recorded for this peak region were measured at 1.5 K. It showed similar substructure due to splitting of the J = 1 rotational line seen for H2 on nanotubes [59].

6.10

Conclusions

In this chapter, we have attempted to show the valuable role that neutrons can play in developing our understanding of how hydrogen behaves in the new kinds of hydrogen absorbers that are now being investigated. For the moment, fairly simple experiments are being undertaken. Increasingly, however, experimentalists will need to do more sophisticated experiments such as varying the ortho–para ratio in the H2 being adsorbed. Experiments will also be used to monitor changes in the structure of absorbers as hydrogen is cycled in and out for the many hundreds of cycles that would be needed in

© 2008, Woodhead Publishing Limited

170

Solid-state hydrogen storage

practice. The better our understanding of the behaviour of these materials subjected to these variables, the easier it will be to adjust the composition or other physical properties to enhance performance. It would be difficult to exaggerate the importance of finding materials that will operate successfully as a fuel system in a hydrogen-fuelled car and neutron scattering will inevitably play an important role in this effort.

6.11

References

1. G.L. Squires, Introduction to the Theory of Thermal Neutron Scattering, CUP, Cambridge 1978. 2. M. Bee, Quasi-elastic Neutron Scattering, Institute of Physics, Bristol 1988. 3. J.A. Young and J.U. Koppel, Phys. Rev. A, 135 (1964) 603. 4. T. Springer and D. Richter, in Methods of Experimental Physics: Neutron Scattering, Vol 28b, p. 131 Academic Press, NewYork 1987. 5. D.K. Ross, Neutron scattering studies of metal-hydrogen systems, in Topics in Applied Physics: 73 Hydrogen in Metals III: Properties and Applications, ed. H. Wipf, Springer Verlag, Berlin 1997. 6. D. Richter, R. Hemplemann and R.C. Bowman, in Hydrogen in Intermetallic Compounds II, ed. L. Schlapbach, Topics in Applied Physics Vol 67, Springer, Berlin 1992. 7. P.C.H. Mitchell, S.F. Parker, A.J. Ramirez-Cuesta and J. Tomkinson, Vibrational Spectroscopy with Neutrons with Applications in Chemistry, Biology, Materials Science and Catalysis, World Scientific Publishing and Imperial College Press, Singapore, and London 2005. 8. Y. Fukai, The Metal Hydrogen System: Basic Bulk Properties, Springer Series: Materials Science Vol 21, Springer, Berlin 1993. 9. E. Weidner, D.J. Bull, I.L. Shabalin, S.G. Keens, M.T.F. Telling and D.K. Ross, http://dx.doi.org/10.1016/j.cplett.2007.07.010 10. W.I.F. David, M.O. Jones, D.H. Gregory, C.M. Jewell, S.R. Johnson, A. Walton and P.P. Edwards, J. Am. Chem. Soc., 129 (2007) 1594. 11. A.C. Dillon, K.M. Jones, T.A. Bekkedahl, D.S. Bethune and M.J. Heben, Nature, 386 (1997) 377–9. 12. A. Chambers, C. Park, R.T.K Baker and N.M. Rodriguez, J. Phys. Chem. B, 102 (1998) 4253–6. 13. V.F. Sears, Neutron News, 3, (1992) 26. 14. National Centre for Neutron Research, National Institute for Science and Technology, USA. See website http://www.ncnr.nist.gov/resources/n-lengths/ http:// neutrons.ornl.gov/ 15. L. Van Hove, Phys. Rev., 95 (1954) 249. 16. B.N. Brockhouse and A.T. Stewart, Phys. Rev., 100 (1955) 756. 17. D.L. Roach, J.D. Gale and D.K. Ross, Neutron News, 18 (2007) 21–3. 18. V.F. Turchin, Slow Neutrons, IPST, Jerusalem 1965. 19. J. Eckert, J.A. Goldstone, D. Tonks and D. Richter, Phys. Rev. B, 27 (1983) 1980. 20. J.J. Rush, J.M. Rowe and D. Richter, Z. Phys. B., 55 (1984) 283. 21. K.-M. Ho, H.-J. Tao and X.-Y. Zhu, Phys. Rev. Letts., 53 (1984) 1586. 22. H.-J. Tao, K.-M. Ho and X.-Y. Zhu, Phys. Rev. B, 34 (1986) 8394.

© 2008, Woodhead Publishing Limited

Neutron scattering studies

171

23. C. Elsaesser, K.M. Ho, C.T. Chan, M. Faehnle, J. Phys. Condens. Matter, 4 (1992) 5189; 4 (1992) 5207. 24. M. Kemali, J.E. Totolici, D.K. Ross and I. Morrison, Phys. Rev. Letts., 84 (2000) 1531–34. 25. L. Van Hove, Phys. Rev. 110 (1954) 249. 26. C.T. Chudley and R.J. Elliott, Proc. Phys. Soc., 77 (1960) 353. 27. W. Gissler and H. Rother, Physica, 50 (1970) 380. 28. A.V. Skripov, J.C. Cook, T.J. Udovic, M.A. Gonzalez, R. Hemplemann and V.N. Kozhanov, J. Phys. Condens. Matter, 15 (2003) 3555–66. 29. J.M. Rowe, J.J. Rush, L.A. de Graaf, G.A. Ferguson, Phys. Rev. Letts., 29 (1972) 1250. 30. C.J. Carlile and D.K. Ross, Solid State Commun., 15 (1974) 1923. 31. M.J. Gillan, J. Phys. C., 19 (1986) 6169. 32. G.H. Yinyard, Phys. Rev. 110 (1958) 999. 33. Y. Li and G. Wahnstrom, Phys Rev. B, 46 (1992) 14528. 34. M.J. Rowe, K. Skold, H.E. Flotow and J.J. Rush, J. Phys. Chem. Solids, 32 (1971) 41. 35. A.V. Skripov, L.S. Voyevodina and R. Hempelmann, Phys. Rev. B, 73 (2006) 014302. 36. I.S. Anderson, J.E. Bonnet, A. Heidemann, D.K. Ross and S.K.P. Wilson, J. LessCommon Metals, 101 (1984) 405. 37. D.K. Ross and D.L.T. Wilson, Neutron Inelastic Scattering, IAEA, Vienna, Vol. 1, (1977) 383. 38. J.D. Barnes, J. Chem. Phys., 58 (1973) 5193. 39. P.L. Hall and D.K. Ross, Mol. Phys., 36 (1978) 1549–54; 42 (1981) 673–82. 40. M. Eisenstadt and A.C. Redford, Phys. Rev., 132 (1963) 635. 41. D.G. Westalke, J. Alloys Compounds, 90 (1983) 251–73. 42. D. Fruchart, A. Rouault, C.B. Shoemaker and D.P. Shoemaker, J. Less-Common Met, 73 (1980) 363–8. 43. J.F. Fernandez, M. Kemali, D.K. Ross and C. Sanchez, J. Phys. Condens. Matter, 11 (1999) 10353–73. 44. B. Bogdanovic and M. Schwickardi, J Alloys Compounds, 253–254 (1997) 1–9. 45. J. Íñiguez, T. Yildrim, T.J. Udovic, M. Sulic and C.M. Jensen, Phys. Rev. B, 70 (2004) 060101(R). 46. Q.J. Fu, A.J. Ramirez-Cuesta and S.C. Tsang, J. Phys. Chem. B, 110 (2006) 711– 715. 47. J.R. Santisteban, G.J. Cuello, J. Dawidowski, A. Fainstein, H.A. Peretti, A. Ivanov and F.J. Bermejo, Phys. Rev. B, 62 (2000) 37–40. 48. H.G. Schimmel, M.R. Johnson, G.J. Kearley, A.J. Ramirez-Cuesta, H. Huot and F.M. Mulder, Mat. Sci. Eng. B, 108 (2004) 38–41. 49. H.G. Schimmel, M.R. Johnson, G.J. Kearley, A.J. Ramirez-Cuesta, J. Huot and F.M. Mulder, J. Alloys Compounds, 393 (2005) 1–4. 50. K. Chlopek, C. Frommen, A. Leon, O. Zabara and M. Fichtner, J. Materials Chem., 17 (2007) 3496–503. 51. J. Tomkinson and T.C. Waddington, J. Chem. Soc. Faraday Trans., 72 (1976) 528– 538. 52. D.G. Allis and B.S. Hudson, Chem. Phys. Letts, 385 (2004) 166–72. 53. M.R. Hartmann, J.J. Rush, T.J. Udovic, R.C. Bowman and S.-J. Hwang, J. Solid State Chem., 180 (2007) 1298–305. 54. A.C. Dillon et al., Nature, 386 (1997) 377.

© 2008, Woodhead Publishing Limited

172

Solid-state hydrogen storage

55. Y. Ye et al., App. Phys. Lett., 74 (1997) 2307. 56. A. Chambers, C. Park, R.T.K. Baker and N.M. Rodriguez, J. Phys Chem. B, 122 (1998), 4253. 57. C.M. Brown, T. Yildirim, D.A. Neumann, M.J. Heben, T. Gennett, A.C. Dillon, J.L. Alleman and J.E. Fischer, Chem. Phys. Letts., 329 (2000) 311. 58. Y. Ren and D.L. Price, Appl. Phys. Lett., 79 (2001) 3684. 59. P.A. Georgiev, D.K. Ross, A. De Monte, U. Montaretto-Marullo, R.A.H. Edwards, A.J. Ramirez-Cuesta and D. Colognesi, J. Phys: Condens. Matter, 16 (2004) L73. 60. P.A. Georgiev, D.K. Ross, A. De Monte, U. Montaretto-Marullo, R.A.H. Edwards, A.J. Ramirez-Cuesta, M.A. Adams and D. Colognesi, Carbon, 43 (2005) 895. 61. K.A. Williams, B.K. Pradhan, P.C. Eklund, M.K. Kostov and M.W. Cole, Phys. Rev. Lett., 88 (2002) doi 165502. 62. H.G. Schimmel, G.J. Kearley and F.M. Mulder, Chem. Phys. Chem., 5 (2004) 1053–4. 63. Y. Liu, C.M. Brown, J.L. Blackburn, D.A. Neumann, T. Gennett, L. Simpson, P. Parilla, A.C. Dillon and M.J. Heben, J. Alloys Comp., 446–447 (2007) 368–72. 64. P.A. Georgiev, D.K. Ross, P. Albers and A.J. Ramirez-Cuesta, Carbon, 44 (2006) 2724. 65. H.G. Schimmel, G. Nijkamp, G.J. Kearley, A. Rivera, K.P. de Jong and F.M. Mulder, Mat. Sci. Eng. B, 108 (2004) 124–9. 66. H.W. Langmi, A. Walton, M.M. Al-Mamouri, S.R. Johnson, D. Book, J.D. Speight, P.P. Edwards, I. Gameson, P.A. Anderson and I.R. Harris, J. Alloys Comp., 356–357 (2003) 710–15. 67. H.W. Langmi, D. Book, A. Walton, S.R. Johnson, M.M. Al-Mamouri, J.D. Speight, P.P. Edwards, I. Harris and P.A. Anderson, J. Alloys Comp., 404–406 (2005) 637–42. 68. J.G. Vitillo, G. Ricchiardi, G. Spoto and A. Zecchina, Phys. Chem. and Chem. Phys., 7 (2005) 3948–54. 69. S.H. Jhung, S.J. Lee, J.W. Yoon, D.P. Kim and J.-S. Chang, Int. J. Hydrogen Energy, 32 (2007) 4233–7. 70. A.J. Ramirez-Cuesta, P.C.H. Mitchell, D.K. Ross, P.A. Georgiev, P.A. Anderson, H.W. Langmi and D. Book, J. Mater. Chem., 17 (2007) 2533–9. 71. A.J. Ramirez-Cuesta, P.C.H. Mitchell, D.K. Ross, P.A. Georgiev, P.A. Anderson, H.W. Langmi and D. Book, J. Alloys Comp., 446–447 (2007) 393–6. 72. A.J. Ramirez-Cuesta and P.C.H. Mitchell, Catalysis Today, 120 (2007) 368–73. 73. P.A. Georgiev, A. Albinati, B.L. Mojet, J. Ollivier and J. Eckert, J. Am. Chem. Soc., 129 (2007) 8086–7. 74. P.A. Georgiev, A. Albinati and J. Eckert, Chem. Phys. Letts., 449 (2007) 182–5. 75. F.J. Torres, J.G. Vitillo, B. Civalleri, G. Ricchiardi and A. Zecchina, J. Phys. Chem. C, 111 (2007) 2505–13. 76. K.A. Lokshin, Y. Zhao, D. He, W.L. Mao, H-K. Mao, R.J. Hendry, M.V. Lobanov and M. Greenblatt, Phys. Rev. Letts., 93 (2004) Article 125503. 77. L.J. Floresse, C.J. Peters, J. Schoonman, K.C. Hester, C.A. Koh, S.F. Dec, K.N. Marsh and E.D. Sloan, Science, 306 (2004) 469–71. 78. L. Ulivi, M. Celli, A. Giannasi, A.J. Ramirez-Cuesta, D.J. Bull and M. Zoppi, Phys. Rev. B, 76 (2007) 161401(R). 79. H. Li, M. Eddaoudi, M. O’Keeffe and O.M. Yaghi, Nature, 402 (1999) 276–9. 80. J.L.C. Rowsell, J. Eckert and O.M. Yaghi, J. Am. Chem. Soc., 127 (2005) 14904–10. 81. N.L. Rosi, J. Eckert, M. Eddaoudi, D.T. Vodak, J. Kim, M. O’Keeffe and O.M. Yaghi, Science, 300 (2003) 1127–9.

© 2008, Woodhead Publishing Limited

Neutron scattering studies

173

82. D. White and E.N. Lassettre J. Chem. Phys., 32 (1960) 72. 83. A.V. Skripov, Defect Diffus. Forum, 224–225 (2003) 75. 84. A.V. Skripov, J.C. Cook, D.S. Sibirtsev, C. Karmonik and R. Hempelmann, J. Phys. Condens. Matter, 10 (1998) 1787. 85. A.V. Skripov, M. Pionke, O. Randl and R. Hempelmann, J. Phys. Condens. Matter, 11 (1999) 1489. 86. A.V. Skripov, J.C. Cook, C. Karmonik and V.N. Kozhanov, Phys. Rev. B, 60 (1999) 7238. 87. A.V. Skripov, J.C. Cook, T.J. Udovic and V.N. Kozhanov, Phys. Rev. B, 62 (2000) 14099. 88. D.J. Bull, D.P. Broom and D.K. Ross, Chem. Phys., 292 (2003) 153. 89. A.V. Skripov, M.A. Gonzalez and R. Hempelmann, J. Phys. Condens. Matter, 18 (2006) 7249–56. 90 A.V. Skripov, M.A. Gonzalez and R. Hempelmann, J. Phys. Condens. Matter, 18 (2006) 7249–56. 91. Q. Shi, J. Voss, H.S. Jacobsen, K. Lefmann, M. Zamponi and T. Vegge, J. Alloys Comp., 446–447 (2007) 469–473. 92. T.R. Jensen, A. Andreasen and T. Vegge, J. Phys. Chem., 111 (2007) 3886–92. 93. T. Vegge, Phys. Chem. Chem. Phys., 8 (2006) 4853–61. 94. B. Bogdanovic and M. Schwickardi, J. Alloys Comp., 253–254 (1997) 1. 95. F. Fernandez-Alonso, F.J. Bermejo, C. Cabrillo, R.O. Loutfy, V. Leon and M.L. Saboungi, Phys. Rev. Lett., 98 (2007) 215503.

© 2008, Woodhead Publishing Limited

7 Reliably measuring hydrogen uptake in storage materials E. MacA. G R A Y, Griffith University, Australia

7.1

Introduction

The most fundamental driver of hydrogen storage research is the need to achieve performance that makes the technology viable. The goal of operating automobiles from an on-board hydrogen storage tank has led to a focus on the reversible capacity of the storage system, which must respect constraints on available mass and volume. For this and other reasons, the need to authoritatively measure the hydrogen uptake of a new storage material in the laboratory is paramount. The long-standing controversy over the hydrogen storage properties of nanostructured carbons can almost certainly be ascribed in part to the special problems associated with measuring small samples with high surface area, prepared by nanoscience routes: experience with metals and complexes prepared by metallurgy does not carry over completely to the low-density materials of current interest. Thus the new world of nanostructured materials has brought new challenges requiring careful design of both apparatus and experiment to produce reliable results for hydrogen uptake. The technical aspects of measuring hydrogen uptake are addressed in this chapter, including methodology and apparatus. Because the storage context implies end-use requiring hydrogen gas, electrochemical techniques are not covered. The most direct and authoritative approaches to measuring hydrogen uptake are discussed, rather than indirect measures such as lattice parameter or resistivity that require calibration of the concentration scale. The focus is on the most common techniques, which are based on measurement of the hydrogen pressure in a closed system (manometry) or of the mass of the sample and its loading of absorbed or adsorbed hydrogen (gravimetry). A problem affecting both classes of technique, that of the influence of inaccurate knowledge of the volume occupied by the sample itself on the reliability of the data, is given prominence. A new approach to measuring hydrogen uptake, the variable-volume hydrogenator, is proposed as the best solution to this all-pervading problem. 174 © 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

175

Generally the hydrogen concentration is represented here by the hydrogento-host atomic ratio: H = nH = mH / M H X nX mX / MX

[7.1]

where X denotes the host, nH and MH are the number of absorbed moles and molar mass of hydrogen, and nX and MX are the number of moles of host atoms in the sample and their molar mass. The mass fraction of hydrogen in the sample is:

fH =

mH = mX + mH

H/ X M H/ X + X MH

[7.2]

Unless specified, ‘absorption’ generally includes adsorption and ‘hydrogen’ generally includes protium (1H or H and H2) and deuterium (2H or D and D2).

7.2

Compressibilities of hydrogen and deuterium

A direct measurement of hydrogen uptake by a material requires knowledge of the density of the surrounding hydrogen gas. This requirement is obviously intrinsic to those techniques in which the number of moles of hydrogen taken up by the sample is calculated from a change in pressure or volume. Even where the measurement is of a change in the mass of hydrogen in the sample or in the sample cell, the hydrogen gas density is needed to apportion hydrogen between the sample and the surroundings, or to correct for buoyancy if the sample is weighed. Considering the absolute accuracies to which pressure, volume, temperature and mass may be practically measured, hydrogen may safely be taken to be ideal above room temperature and at pressures below a few bar. At higher pressures or lower temperatures, the need to account for its non-ideality becomes apparent. Perhaps the simplest approach, and the one used exclusively here, is to define the compressibility, Z, of the gas by: pVm =Z RT

[7.3]

where Vm is the molar volume. Thus Z = 1 refers to an ideal gas. The density of the gas is then: ρ=

Mp ZRT

[7.4]

where M is the molar mass (formula weight) of the gas. The origin of the non-ideality is interactions between the gas atoms or molecules, which may

© 2008, Woodhead Publishing Limited

176

Solid-state hydrogen storage

be crudely modelled by the modified equation of state for a fluid of van der Waals:  a   p + 2  ( Vm – b ) = RT  Vm 

[7.5]

where a is a measure of the attractive part of the interaction and b is a measure of the excluded volume of the gas, i.e. of the repulsive core of the interaction. While a and b may be represented by constants in a sufficiently restricted space of experimental parameters, their dependence on p and T makes the general application of Eq. (7.5) impractical without modification. The same comments apply to virial equations of low order. Tabulated data of good accuracy (say better than 0.1%) for some or all of Vm, ρ and Z are available in a wide range of pressures and temperatures for protium and, to a lesser extent, deuterium. A comprehensive tabulated source is Vargaftik (1975). Other sources are discussed by Hemmes et al. (1986). The most recent, most comprehensive and generally most accurate source is contained in the commercially available National Institute of Standards and Technology (USA) Reference Fluid Thermodynamic and Transport Properties database (REFPROP, URL http://www.nist.gov/srd/nist23.htm), in which are embedded advanced equations of state for protium (Jacobsen et al., 2007) and deuterium (McCarty, 1989). From a practical perspective, access to software that calculates Vm, ρ or Z as a function of p and T greatly facilitates the calculation of hydrogen uptake, especially if it can be incorporated in or called by a spreadsheet that performs the hydrogen uptake calculations for an entire experiment. The REFPROP package satisfies these criteria, albeit at the cost of some complexity owing to its very comprehensive capabilities. Alternatively, a calculator is available for just the compressibilities of protium and deuterium, based on numerical solution of the modified van der Waals equations developed for protium by Hemmes et al. (1986) and for deuterium by McLennan and Gray (2004). This calculator reproduces the tabulated data of Michels et al. (1959) and is of comparable accuracy to REFPROP for 273 < T < 423 K and p < 1000 bar. The accuracy is limited at cryogenic temperatures by the uncertainty in the original data in Michels et al. (1959). This calculator is available free from the author on request. Figure 7.1 demonstrates the importance of using compressibility in the calculation of hydrogen uptake at even modest pressures by contrasting the calculated absorption isotherms for the LaNi5–D2 system at room temperature with and without correction for the deuterium compressibility. Even at pressures of a few bar the error is significant over the breadth of the absorption plateau. Once pressures of several hundred bar are reached, the uncorrected isotherm is completely misleading.

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

177

700

Pressure (bar)

600 500 400 300 200 100 0 0.0

0.2

0.4

0.6

0.8 H/M

1.0

1.2

1.4

1.6

7.1 Isochoric isotherms for the LaNi5–D2 system recorded at 25 °C with (filled circles) and without (open circles) correction applied for the compressibility of the deuterium gas. Even at a pressure of a few bar, the error is accumulating noticeably, becoming extreme at pressures of hundreds of bar.

7.3

Measurement regimes

The most widely used representations of hydrogen uptake characteristics are pressure-composition isotherms (pcT), notionally isobaric (constant-pressure) temperature scans, of which thermal desorption spectroscopy (TDS) is an example, and notionally isoplethic (constant-concentration) temperature scans, the latter usually being displayed on a van’t Hoff diagram for the purpose of determining the enthalpy and entropy changes associated with the reaction. The isochoric (constant-volume) regime corresponds to the constraint of a closed system in which a fixed total amount of hydrogen is exchanged between the gas atmosphere and the sample. Hydrogen is added to the system in aliquots and then the system is closed while the sample and gas react under isochoric conditions. These regimes are contrasted in the pcT diagram in Fig. 7.2, in which paths from an isotherm at temperature T1 to an isotherm at T2 arrive at the same point, P, in the pressure–composition space from very different starting points, according to the constraint. Pressure–composition isotherms may be constructed under an isochoric constraint by adding hydrogen aliquots to the system and measuring the hydrogen absorption or desorption stepwise, or under an isobaric constraint by actively controlling the pressure over the sample to a constant value while the sample absorbs or desorbs.

© 2008, Woodhead Publishing Limited

178

Solid-state hydrogen storage

Isopleth

psys

Isochore

Isobar P

T1 T2

H/X

7.2 Delineation of isochoric, isobaric and isoplethic approaches to the same point P on a pressure–composition isotherm at temperature T2, starting from an isotherm at a higher temperature T1.

A point that has not been directly addressed in the literature is whether the outcomes of a change in hydrogen content of the sample via an isochoric, an isobaric or an isoplethic path are equivalent. The only evidence known to the author suggests that they are not equivalent. An isobaric measurement of the Pd–H2 system by Benham and Ross (1989) produced a quite different isotherm from that familiar from isochoric measurements. In particular, an isobaric study of the effects of small changes in hydrogen pressure on the hydrogen content of LaNi5 by Gray (1992) suggests that isochoric measurements produce an isotherm at an apparently lower pressure than the real isotherm pressure measured isobarically, i.e. at constant chemical potential. Figure 7.3 shows the approach to equilibrium via an isochoric path along with actual gravimetric data for the LaNi5 system published by Gray (1992). Considering first the experimental data, when the system pressure was slowly increased then decreased by a few kPa, the sample did not switch abruptly from conversion from the α to the β phase to desorption of a fixed proportion of the α and β phases: a smooth corner is apparent where the hydrogen content of the sample initially continues to increase while the gas pressure decreases. Assuming that the isobaric measurement best represents the true state of the sample and that the state of the sample does not depend on the measurement regime, changing to isochoric conditions would force the system to approach apparent equilibrium along the isochore, stopping when the isochore is tangent to the real pcT curve. At least in the case of LaNi5, then, isochoric isotherms

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

179

26 24 Isochore 22

Relative pressure (kPa)

20

Isobar

e ic isoth

rm

18 16

Isocho

er ric isoth

m

14 12 10 8 6 0.72

0.74

0.76

0.78 H/M

0.80

0.82

0.84

7.3 Isochoric approach to equilibrium compared to a real isobaric isotherm for the LaNi5–H2 system in which the gas pressure was increased then decreased by a few kPa under active pressure control to within a few Pa. The point at which the isochore is tangent to the isobaric curve defines the isochoric isotherm, which therefore lies below the isobaric isotherm. The experimental data are from Gray (1992).

appear to actually record points at which the sample is close to pure desorption of the dilute and concentrated phases, rather than absorption. A fourth regime worthy of consideration is constant molar flow, particularly for investigating the kinetics of hydrogen absorption and desorption. In this case a flow controller could be used to control and measure the flow of gas to the sample cell (but see also Section 7.7) with the pressure in the sample cell used as a measure of the chemical potential required to force the hydrogenation reaction at the measured rate.

7.4

Measurement techniques

There are numerous techniques available for measuring the amount of hydrogen absorbed or desorbed by the sample under the desired constraint(s). These generally involve measurements of some combination of hydrogen pressure, component volumes, hydrogen flow and sample mass in a gas handling

© 2008, Woodhead Publishing Limited

180

Solid-state hydrogen storage

system consisting of a hydrogen source, a hydrogen sink, a gas manifold and a sample cell. Because flow-based measurements are more difficult for small samples, and practically rather limited in the range of available pressure ratings for mass-flow transducers, we will focus on measurements based on pressure (manometric) and sample weight (gravimetric). The constant-volume system is ubiquitous, but we will also mention a variable-volume approach in Section 7.6, particularly in connection with the problem of sensitivity to the volume (density) of the sample itself (Section 7.5). A manometric system of constant volume is known as a Sieverts (or colloquially pcT) apparatus and is universally operated under isochoric conditions, although isobaric conditions are possible with modification, as explained below.

7.4.1

Sieverts technique

Figure 7.4 shows a generic Sieverts hydrogenator. The measurement of hydrogen uptake is made step-wise. Suppose that, at the end of the k–1th step, a pressure psys, the system pressure, of hydrogen is present throughout the hydrogenator. The gas in the reference volume is at temperature Tref and the gas in the sample cell is at Tcell. The connecting valve, S, is closed to isolate the sample cell, which has an empty volume Vcell. A new pressure pref at temperature Tref is established in the reference volume, Vref. S is then opened and, after a suitable settling time, a new value of psys is measured, along with new values of Tref and Tcell. Following Blach and Gray (2007), the number of moles of hydrogen atoms absorbed or desorbed by the sample in the kth step, ∆n Hk is then calculated from the change in the pressure measured when S is opened:

Pressure transducer

P

Vacuum pump Hydrogen source

S

Vref

Vcell

7.4 Minimal Sieverts (isochoric) hydrogenator. To record an absorption isotherm, gas prepared at measured pressure and temperature in Vref is admitted to the sample cell via valve S. The absorbed amount is calculated from the final pressure in the system and the temperatures of the gas in Vref and Vcell using Eq. (7.6).

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials k k   psys p ref ∆n Hk = 2   – k k k k k k   Z ( p ref , Tref ) RTref Z ( psys , Tref ) RTref

181

  Vref 

k k –1    psys psys – –  ( Vcell – VX )  k k k k –1 –1 –1 k k   Z ( psys , Tcell ) RTcell Z ( psys , Tcell ) RTcell 

[7.6]

where Z is again the compressibility and VX = mX/ρX. Note that the mass of the sample, mX, and its density, ρX, depend in general on its hydrogen content. Equation (7.6) simply accounts for all the gas in the system. The total change in the hydrogen content of the sample after N steps is: N

n HN = Σ ∆ n Hk

[7.7]

k =1

As the measurement relies on changes in the pressure in the system owing to absorption or desorption of hydrogen by the sample, a figure of merit which relates to the system resolution and accuracy is a helpful design parameter. Consider the evolution of the pressure in the hydrogenator during the kth step, and momentarily omit the index k for convenience. Once the connecting valve S has been opened and the ideally instantaneous change in pressure owing to the larger volume sampled by the pressure transducer has occurred, the system is isochoral (constant volume) and the number of moles of hydrogen contained as gas and in the sample is constant: Vj 2 p Σ + n H = constant sys j Z ( psys , T j ) T j R

[7.8]

where the sum runs over all the volumes in the hydrogenator, Vref and Vcell in the simplest case under consideration. The change in the hydrogen content of the sample is therefore reflected in a change in the system pressure which is given, according to Eq. (7.8), by: Vj ∆ n H = – 2 ∆ psys Σ j Z ( psys , T j ) T j R

[7.9]

Using the definition of H/X, the isochoric constraint may then be expressed in terms of the time evolution of the system pressure as the hydrogen is absorbed or desorbed during the kth step according to the kinetics of the sample:

( )

k k psys ( t ) = psys (0) – s k ∆ H X

[7.10]

where psys(0) is the system pressure immediately after the valve S has been opened, before any change in nH has occurred, and –sk is the slope of the isochore for the kth step:

© 2008, Woodhead Publishing Limited

182

Solid-state hydrogen storage

sk =

nX R Vj 2Σ j Z ( pk , T k) T k sys j j

[7.11]

sk indicates the sensitivity of the system to changes in H/X, as measured by changes in the system pressure, and so helps to quantify the performance of the hydrogenator. Figure 7.5 shows the actual locus of the system during an isochoric absorption step. If we ignore Z and assume that the measurement is isothermal, sk is a constant. The inclusion of Z bends the isochore, increasing the hydrogenator sensitivity when Z > 1, with potentially beneficial effect when measuring at high hydrogen pressures. Blach and Gray (2007) propose a figure of merit, η, for the sensitivity of the Sieverts hydrogenator:

psys

k p sys(0)

Isochore slope –sk k p sys k –1 p sys

nXk –1

nXk

H/X

7.5 Real locus of the system in pressure–composition space during an isochoric step from the k−1th to the kth point on the recorded isotherm (Eqns 7.10, 7.11). The excursion of the pressure in the k occurs when the valve S (Fig. 7.4) is opened hydrogenator to p sys(0) instantaneously. The system then approaches equilibrium according to the kinetics of the sample and the thermal relaxation time of the sample/cell sub-system. The pressure excursion is lessened if S is opened slowly, so that absorption commences while psys is still rising. The slope of the isochore is constant only if the compressibility is constant.

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

η=

sk δp

183

[7.12]

where δp is the usable absolute resolution of the pressure transducer, and recommend η ≥ 100 as a rule of thumb to obtain data of high quality, assuming that other errors (in the volumes, temperatures and compressibilities in Eq (7.6) do not limit the hydrogenator performance. While the universal practice is to join the end-points of each isochoric step to make an isotherm, two caveats must be acknowledged: (i) the system undergoes excursions to pressures that may be far from the isotherm and only approaches the drawn isotherm as the free energy drive approaches zero owing to the falling (absorption) or rising (desorption) hydrogen pressure; (ii) the value of ∆nH calculated for a given step is a global average value for the entire sample and is not representative of the local change in hydrogen content unless the sample is homogeneous over a sufficiently small length scale (where ‘homogeneous’ comprises a multiphase system with spatially invariant phase proportions). Large-aliquot effect These intrinsic features of the isochoric technique can combine to adversely affect the quality of the measurement of a sample that exhibits pressure hysteresis owing to the so-called ‘large-aliquot effect’ reported by Park and Flanagan (1983) and incorrectly ascribed to high interface velocity. This effect arises when a large step in hydrogen concentration (high ∆nH) is made in a system with low total volume, necessitating a big excursion in psys (Eq. 7.9). Pons and Dantzer (1993) predicted from combined heat- and masstransfer modelling that temperature gradients in a two-phase system exhibiting pressure hysteresis cause a severe spatial inhomogeneity of the phase proportions. Gray et al. (1994) explained the large-aliquot effect as a consequence of the pressure excursion (∆psys) when a transforming twophase hydride system does so in a range of pressures, meaning that the plateaux in the absorption and desorption pressure–composition isotherms are sloping, owing for instance to a distribution of interatomic potentials at H sites. Considering a step across the absorption plateau, a large increase in system pressure may provide sufficient free energy drive to transform all parts of the sample. As isochoric absorption begins it simultaneously heats the absorbing regions and reduces the pressure. psys thus falls below the transformation pressure of some parts of the sample and these may actually go into desorption. The result is that while the sample globally follows the isochore, local regions are far from it. This effect is greatly exacerbated by poor heat sinking of the sample, which promotes severe temperature gradients owing to the enthalpy of the transformation, as predicted by Pons and Dantzer

© 2008, Woodhead Publishing Limited

184

Solid-state hydrogen storage

(1993). Kisi and Gray (1995) demonstrated the occurrence of inhomogeneous phase proportions with X-ray and neutron diffraction. Figure 7.6 shows an extreme example of the large-aliquot effect in LaNi5. The comparison is between two neutron powder diffraction patterns from the same sample recorded at very nearly the same value of H/X, with the difference that one was arrived at in small pseudo-isobaric steps and the other in a single, large isochoric step requiring a very high excursion of the system pressure. The lattice parameters of LaNi5Dx depend strongly on the phase proportions and state (absorption or desorption) of the transforming α and β phases (Kisi et al. 1994) (plus γ phase if present), and the diffraction pattern is an average over the volume (several cm3) of the large, poorly heat-sinked sample. Therefore only a homogeneous sample will yield a well-defined diffraction pattern. In Fig. 7.6 the more sharply defined pattern (dotted line) implies that the phase proportions are uniform throughout the sample. The second pattern (solid line) is unintelligible beyond the fact that the phase proportions are far from uniform. The corresponding system pressure was half-way between the absorption and desorption plateau, confirming the mixture of absorbing and desorbing regions. 50 000

Al

Neutrons/ms

40 000

30 000

20 000

10 000 95

100

105 Time of flight (ms)

110

115

7.6 Portions of neutron powder diffraction patterns recorded on the high resolution powder diffractometer (HRPD) instrument at ISIS (UK) from LaNi5 charged with deuterium in situ to approx. D/M = 0 to 0.6 in the α + β two-phase region. Dotted line: after multiple pseudoisobaric absorption steps. Solid line: after a single isochoric absorption step from D/M = 0. The latter data are uninterpretable except that they obviously represent regions of sample with widely distributed lattice parameters. The highest peaks come from the aluminium sample cell and demonstrate that the only difference between the two measurements is the state of the sample.

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

185

While the large-aliquot effect was explored using measurements on LaNi5– H2, it is just as relevant to materials of current interest based on Li complexes, as these exhibit pressure hysteresis, have high enthalpies of transformation, poor thermal conductivity owing to small particle size and sloping pressure plateaux (e.g. Luo and Rönnebro, 2005). Isobaric operation of a Sieverts hydrogenator Figure 7.7 shows an in-principle solution to the problem of the large-aliquot effect through a simple modification to the Sieverts isochoric hydrogenator to permit isobaric operation. The principle of determining H/X remains the same as in Eq. (7.6), but the pressure in the sample cell is controlled at a constant value by feeding the hydrogen to or from the sample through a needle valve (NV) under automatic control. Figure 7.8 depicts the locus of the system during a pseudo-isobaric absorption step in which the pressure in the sample cell is controlled at pset until the isobar crosses the isochore, at which point NV is fully open and the final hydrogen content of the sample can be calculated in the usual way. The addition of a second pressure transducer on the sample cell is necessary for pressure control and also allows H/X to be calculated at all times, albeit with degraded accuracy since two pressure readings are involved. Because the pressure over the sample never rises above (absorption) or sinks below (desorption) the set value during an ideally

Pref

Pcell NV

Vacuum pump

S

Hydrogen source

Vref

Vcell

7.7 Sieverts hydrogenator modified from the standard configuration (Fig. 7.4) for isobaric operation by the addition of a needle valve (NV) and a second pressure transducer. The needle valve feeds gas to or from the sample under feedback control such that Pcell is constant. In addition to providing the feedback signal, the second pressure transducer allows the hydrogen content of the sample to be calculated at all times.

© 2008, Woodhead Publishing Limited

186

Solid-state hydrogen storage

psys

Isobar k p set k p sys

k –1 p sys

nXk –1

nXk

H/X

7.8 Real locus of the system in pressure–composition space during a pseudo-isobaric step from the k–1th to the kth point on the recorded isotherm in the modified Sieverts hydrogenator. The excursion of the k by the controlled pressure in the hydrogenator is limited to p set opening of the needle valve (NV in Fig. 7.7).

isobaric step, temperature gradients in the sample cannot cause compositional inhomogeneity (at least, not with the normal exothermic absorption and endothermic desorption). Of course the initial pressure in the reference volume must be chosen judiciously relative to the set pressure for the isobar, or else the isobar may meet the global system isochore well before it meets the isotherm. Even so, the potential for inhomogeneity is greatly reduced in this pseudo-isobaric approach to equilibrium. Isobaric isotherms have rarely been recorded, perhaps owing to the extra complexity of the apparatus, despite the physical good sense of measuring isotherms under a constraint of constant chemical potential. An example from a gravimetric apparatus is given by Benham and Ross (1989). If the isotherm has very low slope then a pure isobaric approach cannot obtain data points in the two-phase region, but a mixed isobaric/isochoric approach will do this successfully and restrict the excursion of the pressure over the sample. It is perhaps a moot point that the large-aliquot effect should not occur anyway without a sloping plateau, but the different results found for Pd–H2 by Benham and Ross using the isobaric rather than the usual isochoric conditions make further exploration of isobaric methodologies worthwhile.

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

187

While isobaric conditions will prevent compositional inhomogeneity where the plateaux are sloping, the free energy drive is low owing to the small pressure excursion, stretching the kinetics in time significantly compared with the isochoric approach, in which the free energy drive remains relatively high until the isochore approaches the isotherm. Isobaric conditions also have a place in measuring the true kinetics of absorption and desorption by sweeping the hydrogen content of the sample through its full range in a single step under constant chemical potential.

7.4.2

Gravimetric technique

Figure 7.9 shows a generic gravimetric measuring system in which the gas atmosphere is managed by a manifold essentially the same as the basic Sieverts hydrogenator in Fig. 7.4. The entire balance mechanism is pressurised and the tare weight is thus surrounded by the same pressure of hydrogen gas as the sample. This is a significant point of differentiation between this more traditional configuration and that in which the sample cell is isolated from the balance mechanism by a magnetic suspension, so that the balance mechanism remains at one atmosphere. The single-sided balance is a particular case of that depicted in Fig. 7.9, so the mathematical formalism by which the

Gas / vacuum

T

S

7.9 Minimal gravimetric hydrogenator, based on a symmetric balance with both the sample (S) and the tare weight (T) suspended in the hydrogen gas.

© 2008, Woodhead Publishing Limited

188

Solid-state hydrogen storage

hydrogen content of the sample is calculated is developed for the more general case. The principle of the measurement is to determine the mass of adsorbed or absorbed hydrogen, mH, by weighing the sample in hydrogen. The sample mass, ms = mX + mH, is approximately balanced by a tare mass, mt. Every component of mass m and density ρ immersed in the gas experiences a buoyancy force equal to the weight of gas at density ρg displaced by its volume v. The total force on the balance may then be written as: F=g

ρ

 g, j  Σ k ( m j – v j ρg, j ) = g j = Σ k m j 1 – j =s,t,… s,t,… ρj  

[7.13]

where s refers to the sample side of the balance, t to the tare side and k = +1 for components on the sample side or –1 for components on the tare side. ρg,j is the density of the gas in the region of the jth component. The term in parentheses expresses the buoyancy effect on every component. The central problem of the gravimetric technique is to extract ms from Eq. (7.13). Clearly it is advantageous to use materials of the same density on the sample and tare sides of the balance to minimise the resultant buoyancy contribution to Eq. (7.13) through subtraction of terms with k = ±1. For the most general case of a beam balance with tare weight and sample suspended in the same working gas at different temperatures, Eq. (7.13) becomes: F = m – m – Mp  v s ( Ts , p ) – v t ( Tt )  s t g R  Z ( Ts , p ) Ts Z ( Tt , p ) Tt  + ∆m b –

v bt ( Tt )  Mp  v bs ( Ts ) – R  Z ( Ts , p ) Ts Z ( Tt , p ) Tt 

+ ∆m h –

Mp R

+ ∆m 0 –

Mp ∆v 0 RZ ( T0 , p ) T0

 v hs ( Ths ) v ht ( Tht )   Z (T , p) T – Z (T , p) T  hs hs ht ht   [7.14]

where M is the formula weight of the working gas, vs is the volume of the sample that is not accessible to the working gas, vt is the volume of the tare weight, which is assumed to be impervious to hydrogen, vbs and vbt are the volumes of the buckets (if any) on the sample and tare sides of the balance, ∆mb is the difference in the masses of the buckets, vhs and vht are the volumes of the hangdown (suspension) components, ∆mh is the difference in the masses of the hangdown components on the sample and tare sides, Ths and Tht are effective temperatures for the hangdown components, which typically traverse zones in a range of temperatures between Ts and T0, the temperature at the balance beam, ∆v0 is the difference in the effective volumes of the two

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

189

sides of the balance beam and ∆m0 is the difference in the effective masses of the two sides of the balance beam. Equation (7.14) assumes that the balance beam is isothermal and that all components except the sample are impervious to the working gas. Several cases of Eq. (7.14) correspond to commonly encountered practical gravimetric systems and are considered in Section 7.5.2. Gravimetric measurement regimes A gravimetric system is subject to many of the same concerns expressed about the Sieverts technique. The gas may be introduced to the sample region in pure isochoric or isobaric modes, although the pressure shock of strict isochoric operation may damage the balance mechanism. A more likely scenario is that to protect the balance mechanism, gas flows to the balance via a control valve that sustains a pressure difference between the gas handling system and the sample environment. If the sample (say) absorbs hydrogen very quickly, then the actual operating regime may be constant flow, but the pressure over a very slowly absorbing sample will quickly exceed the equilibrium absorption pressure, making the regime effectively isochoric. This variability may cause a degree of irreproducibility in the measured isotherms, whose position may depend on the measuring regime (Fig. 7.3). For this reason it is desirable that the gas flow to and from the balance be actively controlled in a reproducible way, such as to ensure that the pressure over the sample changes monotonically throughout the execution of an isotherm, or in small isochoric steps, or in pseudo-isobaric steps (Fig. 7.8). Depending on the means by which the hydrogen gas pressure in the balance is changed, gravimetry is also susceptible to the large-aliquot effect. The author’s experience is that while the sample temperature excursion owing to a change in the hydrogen content may be larger than in a Sieverts apparatus, owing to the rather high thermal resistance of the gas between the sample and the walls of the pressure cell, the temperature gradient within the sample is relatively small for the same reason, thus mitigating the large-aliquot effect.

7.5

System characterisation

An accurate model of the system consisting of the apparatus and the sample is essential for extracting reliable values for the uptake of hydrogen from Eq. (7.6) (Sieverts technique) or Eq. (7.14) (gravimetric technique).

7.5.1

Sieverts apparatus

Calibration of the volume of each component of the Sieverts hydrogenator is fundamental to its accuracy. The system model must include the temperature

© 2008, Woodhead Publishing Limited

190

Solid-state hydrogen storage

profile of the system. For this reason it is very desirable that the manifold and reference volume(s) be temperature controlled and isothermal. While measuring the temperature of the gas in each major volume is straightforward, the sample cell and its pipework (to the right of S in Fig. 7.4) require careful modelling because the sample temperature may be very different from that of the rest of the hydrogenator. The characterisation of an isothermal hydrogenator, such as one totally immersed in a temperature-controlled bath, is very straightforward. For instance, an accurately known calibration volume may be connected to the hydrogenator, filled with a measured pressure of calibrating gas, and then sequentially opened to the previously evacuated hydrogenator component volumes. These are then calculated based on the fixed quantity of gas in the system. The compressibility of the calibrating gas must of course be accounted for. The same methodology is the basis for characterisation of the system when the temperatures of the reference volume and sample differ. A likely scenario is that Vref is known and Vcell or Vcell – VX (the cell dead volume) in Eq. (7.6) is not known. Beginning with the sample cell evacuated and a pressure p0 of calibrating gas in Vref, the number of moles of gas in the system is fixed at n0 = (p0Vref)/[Z(p0, T 0)RT 0], where T 0 is the initial temperature of the reference volume, by closing the system. This gas is then admitted to the sample cell via S and the system pressure settles to a value psys(Tref, Ts) that depends on the temperature distribution in the sample cell and its associated pipework. Supposing that all components to the right of S in Fig. 7.4 are isothermal at temperature Ts, then the cell volume (or dead volume) may be straightforwardly calculated from n0 =

psys ( Tref , Ts )  Vref Vcell   Z ( psys , Tref ) Tref + Z ( psys , Ts ) Ts  R  

[7.15]

A step in temperature occurring at the valve S between Tref and Ts is an unrealistic assumption. Two simple thermal models are now considered, followed by a variation with application in high-pressure systems. Equivalent volume model e If in Eq. (7.15) Vcell/Ts is replaced by Vcell ( Ts )/ Tref then an effective cell e volume Vcell may be calculated:

  n 0 RTref Vref e Vcell =  –  Z ( psys , Ts )  psys ( Tref , Ts ) Z ( psys , Tref ) 

[7.16]

This volume depends on the sample temperature through psys and on Tref. If Ts is scanned through the range of working temperatures then a representation of the effective cell volume as a function of Ts is obtained. The weakness of

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

191

this approach is that Tref must be reasonably constant throughout the measurement of hydrogen uptake based on this equivalent volume, say to within about 1 K. It is otherwise robust, as long as the compressibility of the calibrating gas is included in the calculation so that the equivalent volume does not become dependent on pressure as well as sample temperature. Divided volume model The simplest possible thermal model of the real temperature distribution in the sample cell and its associated pipework is a step function in temperature at a position which should be determinable through a suitable calibration procedure. Figure 7.10 shows the sample cell divided into volumes V1 at temperature Tref and V2 at the temperature of the sample, Ts, with the constraint that Vcell = V1 + V2 is the real volume of the system closed off by the valve S. The basis of the calibration in this divided-volume approach is to start with the sample temperature at some base value and change Ts through the full working range while recording the system pressure. Once again, n0 = (p0Vref)/[Z(p0, T 0)RT 0] moles of calibrating gas are introduced to the reference volume. When this gas is admitted to the sample cell via S, the system pressure settles to a value psys(Tref, Ts) that may easily

S

V1

Vcell = V1 + V2

V2

7.10 Divided-volume model of the temperature distribution in the sample cell and associated pipework. V1 is assumed to be at the same temperature as the reference volume. V2 is assumed to be at the same temperature as the sample.

© 2008, Woodhead Publishing Limited

192

Solid-state hydrogen storage

be shown to depend on the temperature of the region around the sample through:

Z ( psys , Tref ) Tref Rn 0 Z ( psys , Tref ) Tref V2 – Vref = V1 + Z ( psys , Ts ) Ts psys ( Tref , Ts )

[7.17]

If Ts is scanned and a graph is constructed of the left side of Eq. (7.17) against [Z(psys, Tref)Tref]/[Z(psys, Ts)Ts], the intercept will be V1 and the slope will be V2. The divided-volume model is generally more robust than the equivalentvolume model but nevertheless has some weaknesses. First, Tref must be constant during the calibration procedure. Second, the position at which the step function in temperature between Tref and Ts is assumed to occur must be independent of Ts, so that V1 and V2 are constants. If the sample cell is at Ts and the transition between V1 and V2 occurs in the connecting pipework, this condition is fairly easy to satisfy. Third, to accurately delineate V1 and V2, the coefficient of V2 in Eq. (7.17) must be a fairly strong function of temperature, either through a wide range of scanned values for Ts or through a rapidly varying compressibility. Volume calibration based on compressibility A related approach that relies on a non-linear variation of compressibility with temperature is useful in high-pressure Sieverts systems and can be used, in fact is best used, with hydrogen as the calibrating gas. Re-casting Eq. (7.17) so that psys is a function of the measured variables, psys ( Tref , Ts ) =

Vref Z ( psys , Tref ) Tref

Rn 0 V1 V2 + + Z ( psys , Tref ) Tref Z ( psys , Ts ) Ts [7.18]

the right side of Eq. (7.18) is fitted to the measured pressure with V1 and V2 as fitting parameters. This approach does not require Tref to be constant and makes no assumption about the compressibility but still assumes that V1 and V2 are constant in the temperature range scanned.

7.5.2

Gravimetric apparatus

Equation (7.13) shows that the accuracy of the value of the sample mass depends absolutely on accurate knowledge of the volumes or densities of the suspended components. Methodologies for measuring these in situ are now considered for three common realisations of a gravimetric system.

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

193

Two-sided balance with symmetric temperature distribution If the system is perfectly balanced by all components of the balance itself having the same masses and volumes on the sample and tare sides, only the difference between the sample and tare weights contributes to the total measured force. For this reason it is very advantageous to maintain the temperature environments around the sample and tare weight as nearly identical as possible, by, for example, a two-tube furnace for measurements above room temperature. For the realistic case of symmetric temperature but asymmetric masses, Eq. (7.14) reduces to

Mp F =m –m – [ v ( T , p ) – v t ( Ts )] s t g Z ( Ts , p ) RTs s s + ∆m b –

Mp [ v ( T ) – v bt ( Ts )] Z ( Ts , p ) RTs bs s

+ ∆m h –

Mp [ v hs ( Ths ) – v ht ( Ths )] Z ( Ths , p ) RThs

+ ∆ m0 –

Mp ∆ v0 Z ( T0 , p ) RT0

[7.19]

While the mass imbalances between the sample and tare sides of the balance (∆m0, ∆mh, ∆mb) are constant and can usually be electronically tared (offset) when the sample is loaded, the volume imbalances require quantification. This can be accomplished by first stripping back the balance suspension to the bare beam and recording the resultant force as a function of gas pressure with the entire balance at or near T0, thus allowing the value of ∆v0 to be extracted by fitting the data to the last line of Eq. (7.19). The suspension components pose a more complicated problem, owing to the temperature distributions that may exist along their lengths. If the suspension is light and of dense material, the buoyancy forces on the suspension components will be very small, probably negligible, and changes in their volumes with sample temperature may certainly be neglected. Thus a measurement made as a function of gas pressure with the hangdown components in place and the entire system at T0 will yield a sufficiently reliable value for ∆v h ( Th ) ≈ ∆vh(T0) = vhs(T0) – vht(T0) if one is required. If the suspension consists of, say, thick quartz filaments, then a measurement of the resultant balance force at roughly constant pressure in a wide range of temperatures may be fitted to the third line of Eq. (7.19) to obtain effective values for ∆vh, e.g. referred to Ts by using Th = Ts in Eq. (7.19). Now adding the sample and tare buckets and again recording the resultant force as a function of gas pressure at Ts = T0 = Tt will allow the value of ∆vb(T0) = vbs(T0) – vbt(T0) to be

© 2008, Woodhead Publishing Limited

194

Solid-state hydrogen storage

extracted. Depending on the volumes of the buckets, it may suffice to use a thermal expansion correction to calculate ∆vb(Ts) for different sample temperatures, or a series of measurements at different temperatures can be made if necessary. The foregoing procedure accounts for all terms in Eq. (7.19) except those in the first line. While the characterisation of the balance itself can be performed with the usual working gas, although a relatively dense gas may be used advantageously to amplify the buoyancy effects, the volume difference between the sample and the tare weight must be determined with a gas that does not interact with the sample (or tare weight, obviously). Assuming that such a gas is available (but see Section 7.6.1), finally recording the resultant force as a function of gas pressure in the full range of sample temperatures to be used allows values for vs(Ts) – vt(Ts) to be extracted, including any hydrostatic pressure effect on the sample. The effect of hydrogen absorption on the mass of the sample is now able to be extracted. Two-sided balance with asymmetric temperature distribution In some circumstances it may be impractical to achieve symmetry between the sample and tare sides of a two-sided balance, e.g. if the sample is in a cryogenic environment that cannot readily be duplicated on the tare side. Equation (7.14) applies to this scenario. While measurements of ∆vh(T0) and ∆vb(T0) may be made as described in the preceding sub-section, separated values for vbs(Ts), vbt(Tt), v hs ( Ts ) and v ht ( Tt ) are required. This can be accomplished by measurements of the resultant force at roughly constant high pressure in the full range of sample temperatures to which the second and third lines of Eq. (7.14) are separately fitted subject to the constraints ∆vh(T0) = vhs(T0) – vht(T0) and ∆vb(T0) = vbs(T0) – vbt(T0). The same procedure is then applied to individually determining values for vs(Ts) and vt(Tt). Single-sided balance In a single-sided system the tare-side components are isolated from the sample side, typically by a magnetic suspension of the sample and associated components across the wall of the pressure cell. This arrangement has the advantage of also isolating the balance mechanism from the working gas and so generally facilitating a higher working pressure than is common among double-sided systems. On the down side, there is no first-order cancellation of buoyancy forces between the sample and tare sides of the balance, so the masses and volumes of the suspended components must be measured with the greatest care. For this scenario Eq. (7.14) reduces to

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

F = m – Mp  v s ( Ts , p ) s g R  Z ( Ts , p ) Ts

+ mh –

195

 – m + m – Mp  v b ( Ts )  0 b  R  Z ( Ts , p ) Ts 

Mp  v h ( Th )  R  Z ( Th , p ) Th 

[7.20]

where m0 accounts for all the components outside the sample cell. As the weight and volume of the magnetic suspension components inside the sample cell are necessarily much greater than in the case of the double-sided balance, it is in principle necessary to remove the sample and sample bucket and record the resultant force at a high pressure in the full range of sample temperatures, so that a set of values can be obtained to represent vh, e.g. referred to Ts by using Th = Ts in Eq. (7.20). It is again probably advantageous to do this with a dense gas in the sample cell to amplify the buoyancy contribution. The sample bucket may now be added and an analogous set of measurements performed to obtain ∆vb(Ts). Finally, the effective sample volume may be measured with a gas towards which the sample is inert to complete the model of this system.

7.6

The sample volume problem

Whereas the need to measure temperature and pressure accurately is obvious and widely accepted, the problem of accounting for the sample volume turns out to be more subtle and potentially much more detrimental to the measurement of H/X, as recently explained in detail by Blach and Gray (2007) for the Sieverts technique. As an example, Figure 7.11 shows the disastrous effect of a ±25% change in the assumed density of a sample of potassium-intercalated graphite on the calculated isotherm for deuterium absorption by the Sieverts technique. The sample volume problem has been pointed out before (Beeri et al., 1998) in the context of high-pressure measurements, but Blach and Gray (2007) showed that the problem also occurs in an isochoric manometric apparatus at pressures of a few bar, particularly with low-density materials. Sircar (2001) gives an example of comparably wrong measurements of the Gibbsian surface excess (the total adsorbed amount less the amount of gas that would occupy the volume of the adsorbed phase) of N2 on alumina with a surface area of only 120 m2 g–1, owing to an inadequate model of the interaction of the calibrating gas (He) with the sample during measurements of its density. The Sieverts technique is sensitive to the density of the sample because the volume occupied by the sample must be subtracted from the volume of the empty sample cell in order to calculate H/X. The gravimetric technique is intrinsically sensitive to the volume or density of the sample through the buoyancy force on it. Knowledge of the sample volume is also required to

© 2008, Woodhead Publishing Limited

196

Solid-state hydrogen storage 700 600

Pressure (bar)

500 400 300 200 100 0 0.00

0.02

0.04

0.06 H/X

0.08

0.10

0.12

0.14

7.11 Effect of a ±25% change in assumed sample density on the apparent hydrogen uptake of C24K at room temperature, measured in a system of insufficient volume relative to the volume of sample (after Blach and Gray, 2007). Circles: ρ = 2.0 g/cm3, inverted triangles: ρ = 1.5 g/cm3, upright triangles: ρ = 2.5 g/cm3.

construct isotherms of excess hydrogen uptake (the excess over the amount of gas that would be present anyway in the absence of the sample) and, at a deeper level of investigation, to obtain information about the adsorbed phase itself via determination of the Gibbsian surface excess. The difficulty then arises of knowing the volume occupied by the sample with sufficient accuracy throughout the course of the measurement of hydrogen uptake. Taking a certain number of moles of sample atoms, corresponding to a particular achievable hydrogen uptake via H/X, the volume of the sample increases as its density decreases. As low-density samples are generally those with greatest promise for hydrogen storage at acceptable hydrogen density by mass, the sample problem is thus most significant for the very materials of greatest current interest. Whatever approach is taken to characterising a gravimetric system, comparison with deuterium uptake by the same sample is a valuable check on the measured uptake of protium.

7.6.1

Is the calibrating gas inert?

The most common way of defining the volume of the sample is to measure the effective volume of the loaded sample cell (Sieverts technique) or the sample itself (gravimetric technique) using an inert gas. The validity of this

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

197

procedure is undoubted where the sample has a three-dimensional morphology with low surface area and its density does not change owing to hydrogen absorption. In reality, even He is reported to adsorb detectably onto lowdimensional active materials such as activated carbon and zeolites (Malbrunot et al., 1997), single-walled carbon nanotubes at 300 °C (Haas et al., 2005), and even relatively inert silicalite (Gumma and Talu, 2003), raising serious doubts about the reliability of this approach unless great care is taken with the determination of the skeleton density (that corresponding to the actual volume occupied by sample atoms) of the sample. When the surface area of the sample is a significant contributor to its hydrogen uptake, it is necessary to assume that the sample is equally accessible to the calibrating and working gases. As their sizes are comparable, this is less of a problem with helium and hydrogen than with, say, helium and methane. When the sample density changes with hydrogen concentration, the problem of uncertain density cannot be solved by calibration at H/X = 0 and the system should be designed to minimise its effect as far as possible. While materials that absorb H rather than adsorb H2 are expected to exhibit the greatest effects of hydrogen uptake on density, even adsorption systems may not be immune, as Dreisbach, et al. (2002) reported a swelling of activated carbon owing to He uptake. Correcting for the sample density becomes an even more crucial prerequisite when adsorption isotherms are interpreted on a model of gas plus surface adsorbed phase plus the skeleton of sample itself (e.g. Dreisbach et al., 2002). Sircar (2001) describes methodologies for both manometric and gravimetric measurements by which the Gibbsian surface excess may be determined using a simple model of the low-pressure adsorption of the calibrating gas. The assumption that the calibrating gas does not adsorb on the sample at the highest working temperature is required, however. Gumma and Talu (2003) further discuss the concept and experimental difficulty of defining the Gibbs surface that encloses the impenetrable atomic skeleton of the sample and separates it from the adsorbed phase. The ‘impenetrable volume’ of a sample is of course the volume occupied by its electrons, so even this volume in principle may vary with the strength of the interaction between the sample and adsorbent. The approach of Gumma and Talu is more realistic than that of Sircar, in that no assumption is made about the degree of adsorption of the calibrating gas and the parameters of its adsorption isotherm are fitted to experimental data in the limit of low pressure. The model of Dreisbach et al., which is based on an assumed analytical isotherm is also worthy of consideration. A critical comparison of the latest approaches to the analysis of the same data is required. While the methodologies of Sircar (2001), Dreisbach et al. (2002) and Gumma and Talu (2003) are aimed at determining the Gibbsian surface

© 2008, Woodhead Publishing Limited

198

Solid-state hydrogen storage

excess adsorption, they apply equally to the determination of hydrogen uptake as an end in itself, as the problem of the interaction of the calibrating gas with the sample is common to both aims. An aspect of the inertness of the calibrating gas that receives little attention is its purity. Aside from the potential to poison the surface of the sample against hydrogen uptake, an impure calibrating gas may cause problems in the characterisation of a gravimetric system owing to preferential, reversible adsorption of heavy, reactive impurities such as O2. This gettering of the calibrating gas by the sample is much less of a problem in a manometric apparatus. In general, however, the purity of the calibrating gas should be about as good as that of the hydrogen.

7.6.2

Sieverts technique

The premise of the Sieverts technique is that accurate values for all the parameters in Eq. (7.6) are known at the end of the kth step, most fundamentally the volumes that constitute the system. Equation (7.6), which has been written so as to expose the dependence of the calculation on the volume of sample, shows that uncertain knowledge of the sample volume (via its mass and density) and cell volume affect the calculation as if an error had occurred in the calibration of the hydrogenator. As pointed out in the preceding section, expansion of the sample owing to absorption needs to be allowed for by calculation or measurement, which might not be feasible, or, preferably, by designing out the sensitivity to it. Partially differentiating Eq. (7.6) with respect to density shows that (a) the effect of a change in density on the calculated hydrogen uptake depends on ρ–2, confirming the increased sensitivity to low but uncertain sample densities, and (b) the dependence on the actual instantaneous conditions of p, V and Z is complicated and not amenable to analytic analysis. Blach and Gray (2007) therefore employed simulation to explore the consequences of inaccurately known sample density and, for comparison, compressibility and volumes. The most significant differences between the various experimental set-ups tested were in the ratio of the reference volume, Vref, to the empty volume of the sample cell, Vcell, and in the fraction of the cell volume actually occupied by sample. A first intuition, that insensitivity to the sample density will be conferred by making the sample cell volume as small a fraction of the system volume as possible, is wrong. Likewise, a small sample in a cell which is itself a small fraction of the total system volume is not completely effective. The best outcome was obtained with comparable reference and cell volumes and a sample which occupied a small fraction of the system volume. This may be rationalised as follows. The vulnerability of the Sieverts technique is that it relies on calculating the quantity of hydrogen exchanged with the sample in

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

199

the kth step by the subtraction of two relatively large numbers, which are the amounts of hydrogen in the system before and after the kth step. If the cell k in Eq. (7.6), will not be much volume is very small, the system pressure, psys k different from the pressure in the reference volume, p ref , that initiated it. If Vref is fairly large, the first term in Eq. (7.6) will then be of moderate magnitude. However, the change in system pressure between the kth and (k + 1)th steps may be large if the isotherm contains a small number of points, and this difference amplifies the volume term containing the difference between the empty cell and the sample volumes, making it also of moderate magnitude. Thus there is the potential for a large effect on the calculated quantity of hydrogen exchanged with the sample owing to an error in the volume occupied by the sample, implicating the sample density. The approach of making both Vref and Vcell large compared with the notional volume of sample needs to be balanced against the figure of merit for the system (Eq. 7.12), which will degrade as the system volume becomes too large unless a pressure transducer of increased accuracy is employed. It is not possible to precisely define the optimum ratio of volumes because of the complicated dependence of the systematic error in H/X caused by a density error on p, T and Z: there is no single set-up that is optimum for all conditions of pressure and temperature, but it is certainly possible to greatly lessen the sensitivity of the results to uncertain sample density by following the rulesof-thumb proposed by Blach and Gray (2007) for the Sieverts technique: (i) ensure that the reference volume and the empty cell volume are (a) both large compared with the volume notionally occupied by the sample, by a factor of at least 100 in each case, and (b) ideally about equal; (ii) ensure a figure of merit for the hydrogenator (Eq. 7.12) of at least 100. A further important point made by Blach and Gray (2007) is that measuring a dense sample such as LaNi5 as a de facto standard to check system performance can be misleading: a perfectly satisfactory measurement of LaNi5 does not guarantee good performance with samples of much lower density.

7.6.3

Gravimetric technique

As in the case of the Sieverts technique, the appearance of the sample density on the bottom line in Eq. (7.13) implies a ρ–2 sensitivity of the result to sample density and increased buoyancy forces with samples of low density. Equation (7.14) shows that the best outcome is with matched sample and tare-weight densities as well as matched masses, thus cancelling the buoyancy force on the sample. At the opposite extreme, the single-sided balance (or a double-sided balance with grossly mismatched sample and tare-weight densities) requires careful calibration with an inert gas as discussed in Section 7.6.1.

© 2008, Woodhead Publishing Limited

200

Solid-state hydrogen storage

When the density of the sample varies with hydrogen uptake, Eq. (7.14) makes it clear that, because the measured value scales with sample mass and volume in the same way, nothing can be done with the design of the apparatus to lessen the effect of the changing sample density. The only recourse is to independent measurements by diffraction, for instance, or to the very detailed interpretation of a suite of data on a hopefully realistic physical model, such as that of Dreisbach et al. (2002). Independent measurements may be the only way to reliably know that the sample density is not constant anyway. It seems likely that there are published studies in which this effect has led to the wrong interpretation of the results.

7.7

The variable-volume hydrogenator

The convenience, robustness and low entry cost of a manometric hydrogenator ensure that the Sieverts apparatus is the most popular approach to measuring hydrogen uptake. The sample volume problem and the basic isochoric measurement regime are significant disadvantages. The gravimetric technique is intrinsically more flexible, owing to the decoupling of the measurement of hydrogen uptake from the choice of isochoric, isobaric, etc. measurement regimes. The weaknesses of gravimetry are the need to accurately model the volumes of all suspended components and the great difficulty of accounting for the effect of a sample density that depends on hydrogen uptake. A new manometric approach is proposed here, intermediate in cost and complexity between a manometric and a gravimetric system, in which Vref is variable.

7.7.1

Variable-volume technique

Figure 7.12 shows a minimal variable-volume system. Vref, or some substantial fraction of it, is a volume that depends on the position of a piston, ideally realised with a computer-operated motor drive: Vref (x) = V0 – αx

[7.21]

where x ∈ [0, xmax] is the position of the piston, V0 is the constant part of Vref, αxmax is the displacement volume of the variable portion and where α = – ∂Vref/∂x. Increasing x thus corresponds to compression of the gas in Vref. Within limits imposed by the size of the variable volume compared to the total system volume including Vcell, the control of x may be based on any desired constraint (through feedback) to realise conditions of constant volume (x = constant), constant pressure, constant concentration or constant molar flow. An important advantage of the variable-volume technique over the Sieverts technique becomes apparent by re-casting Eq. (7.6) for the variable-volume scenario:

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials Pressure transducer

201

P

Vacuum pump

S

Hydrogen source

V0

Vcell

Vref = V0 – α x

0

x

xmax

7.12 Minimal variable-volume hydrogenator, constructed from a basic Sieverts apparatus by including a variable component in Vref, such as a motor-driven piston, with displacement volume αxmax.

0 0   psys Vref psys ( x ) Vref ( x )  ∆n H ( x ) = 2   –  0 0 0 Z ( p , T ) RT ( x ) Z p T RT ( , ) sys ref ref   sys ref ref 

0 psys psys ( x )    + –  ( Vcell – VX )  0 0 0 ( , ) ( ) Z p T RT x sys cell cell   Z ( psys , Tcell ) RTcell 

[7.22] where the ‘0’ superscript refers to the initial state at the start of the compression or decompression sequence in which the piston is driven in or out. The sample cell is represented in Eq. (7.22) by a single volume at temperature Tcell for clarity, but the application of a model of the temperature distribution as discussed in Section 7.5.1 is a straightforward extension. Compared with the traditional isochoric approach, it is as if a single hydrogenation or dehydrogenation step were executed. Thus the difference between the final and initial number of moles of H in the sample is computed by direct reference 0 , rather than from a cumulative stepwise sequence. to the initial state, p = psys Compared with Eq. (7.6), removal of the reference to the system pressure at the end of the preceding step removes the mechanism referred to in Section 7.6.1 for amplifying the effect of the sample volume on the calculation. As the number of isochoral steps is greatly decreased to possibly one per isotherm, the cumulative errors intrinsic to the Sieverts technique are also largely eliminated.

© 2008, Woodhead Publishing Limited

202

Solid-state hydrogen storage

7.7.2

Characterisation of the variable-volume apparatus

The basis of the calibration of the volumes in the hydrogenator is as discussed in Section 7.5.1. The availability of the variable reference volume greatly facilitates calibration of any other volume in the system, particularly if it is under computer control. Vref is first calibrated at enough piston positions x to accurately measure ∂Vref/∂x. The unknown volume to be calibrated, Vu, which can include the constant part of Vref, is connected and an initial pressure p0 is established. The piston is then driven in and the pressure and all relevant temperatures are recorded against x. The amount of gas in the system is constant, so assuming for clarity that all of Vref is at the same temperature and the compression sequence starts at x = 0:   V0 Vu p0  + 0 0 0 0 0 0   Z ( psys , Tref ) RTref Z ( psys , Tu ) RTu  V0 – α x Vu   – = psys ( x )   Z ( p , T ) RT ( x ) Z ( p , T ) RT ( x ) sys sys u u ref ref  

[7.23]

which may be solved for psys(x) and fitted to the data to obtain both V0 and Vu if Tref and Tu are very different. Alternatively, if Tu = Tref then Eq. (7.23) reduces to a simple form that can be graphed to find V0 + Vu from the slope: V + Vu   p 0 / Z ( p 0, T 0 )  1– x=  0  α   p ( x )/ Z ( p, T) 

[7.24]

In summary, the variable-volume technique promises to confer some valuable advantages: • • • • •

Great flexibility – constant volume, pressure, composition or molar flow. Mitigation of the sample volume problem. Improved overall accuracy by avoiding cumulative errors. Ability to take many data points without compromising accuracy through error accumulation. Facilitation of automated many-point volume calibrations.

At the time of writing a variable-volume hydrogenator for pressures up to 345 bar is being constructed at Griffith University. Results will be published in due course.

7.8

Summary and conclusions

The possibilities and problems of measuring the uptake of hydrogen using manometric and gravimetric apparatus have been considered. The universal dependence of the measurement on the density of the gas around the sample

© 2008, Woodhead Publishing Limited

Reliably measuring hydrogen uptake in storage materials

203

means that some measure of gas density beyond the ideal-gas approximation should almost always be used. The possible operating regimes (isochoric, isobaric, isoplethic, constant flow) were described. Detailed models of the Sieverts manometric hydrogenator and a generic gravimetric system based on a microbalance were presented and the characterisation of the system of hydrogenator plus sample was discussed. The problems caused by the invariable need to accurately know the effective volume of the sample were considered in some detail. While calibration with an inert gas is the in-principle solution to this problem, even helium adsorbs measurably onto many materials with large surface area, and the volume of the sample may change with hydrogen content, which calibration with He cannot accommodate. Criteria by which a Sieverts apparatus can be designed to avoid the sample-volume problem by making it insensitive to the sample volume or density were presented. The sample-volume problem is much tougher in the case of gravimetry, with no apparent design-based route available to avoid the absolute need for accurate measurements of the impenetrable volume or skeleton density of the sample. A new approach to measuring hydrogen uptake, that of the variable-volume hydrogenator, was presented and predicted to be much less susceptible to the sample-volume problem, to have excellent flexibility in choice of operating regime and to offer generally improved accuracy over the Sieverts technique.

7.9

Acknowledgements

It is a pleasure to acknowledge the colleagues with whom the work on technique and methodology has been carried out over a number of years: Tomasz Blach, Craig Buckley, Erich Kisi, Keith McLennan, Jim Webb. The data for Fig. 7.1 were kindly provided by Tomasz Blach. Thanks are due to Jim Webb for critically reading the manuscript.

7.10

References

Beeri O, Cohen D, Gavra Z, Johnson J R and Mintz M H (1998), J Alloys Cmpnds, 267, 113–120. Benham M J and Ross D K (1989), Z Phys Chem N F, 163, S25–S32. Blach T B and Gray E MacA (2007), J Alloys Cmpnds, 446–447, 692–697. Dreisbach F, Lösch H W and Harting P (2002), Adsorption, 8, 95–109. Gray E MacA (1992), J Alloys Cmpnds, 190, 49–56. Gray EMacA, Buckley CE and Kisi EH (1994), J Alloys Cmpnds, 215, 201–211. Gumma S and Talu O (2003), Adsorption, 9, 17–28. Haas M K, Zielinski J M, Dantsin G, Coe C G, Pez G P and Cooper A C (2005), J Mater Res, 20, 3214–3223. Hemmes H, Driessen A and Griessen R (1986), J Phys C: Solid State Phys, 19, 3571– 3585. Kisi E H and Gray E MacA (1995), J Alloys Cmpnds, 217, 112–117.

© 2008, Woodhead Publishing Limited

204

Solid-state hydrogen storage

Kisi E H, Gray E MacA and Kennedy S J (1994), J Alloys Cmpnds, 216, 123–129. Jacobsen R T Leachman J W and Lemmon E W (2007), Int J Thermophys, 28, 758–772. Luo W and Rönnebro E (2005), J Alloys Cmpnds, 404–406, 392–395. Malbrunot P, Vidal D, Vermesse J, Chahine R and Bose T K (1997), Langmuir, 13, 539– 544. McCarty R D (1989), Correlations for the Thermophysical Properties of Deuterium, Boulder, CO, National Institute of Standards and Technology. McLennan K G and Gray, E MacA (2004), Meas Sci Technol, 15, 211–215. Michels A, de Graaff W, Wassenaar T, Levelt J M H and Louwerse P (1959), Physica, 25, 25–42. Park C N and Flanagan T B (1983), J Less-Common Met, 94, L1–L4. Pons M and Dantzer P (1993), Z Phys Chem N F, 183, S225–S234. Sircar S (2001), AIChE J, 47, 1169–1176. Vargaftik N B (1975), Tables on the Thermophysical Properties of Liquids and Gases, New York, Wiley.

© 2008, Woodhead Publishing Limited

8 Modelling of carbon-based materials for hydrogen storage J. Í Ñ I G U E Z, Institut de Ciencia de Materials de Barcelona (ICMAB-CSIC), Spain

8.1

Introduction

The possibility of using carbon-based materials for hydrogen storage attracts great interest, for a variety of reasons. Most importantly, carbon is a cheap and light storage medium, ideal from the point of view of the gravimetric requirements associated with the use of fuel cells in automobiles. In addition, as required for a hypothetical large-scale commercialization, the industry has great expertise in the production of carbons, with considerable control of their micro and nano-structures. Unfortunately, and in spite of much research effort, the performance of carbon-based materials for hydrogen storage is poor, as even the best systems are far from what is required for applications. The difficulty that needs to be overcome is well known and of a very fundamental nature: the dominating carbon–hydrogen interactions are not appropriate for the purpose of storing and releasing hydrogen in the ideal fuel-cell working conditions. Indeed, in the usual materials, which range from activated carbons to nanotubes, most carbon atoms are in an sp2 configuration (see Fig. 8.1) and bind to three nearest neighbours through σ and π bonds, as in graphite. In such conditions, the carbons interact with molecular hydrogen through weak van der Waals forces, which result in binding energies of less than 0.1 eV and are insufficient to retain the hydrogen at ambient pressure and temperature. On the other hand, as illustrated in Fig. 8.2, atomic hydrogen tends to form strong chemical bonds with the carbon atoms, with binding energies of several electronvolts; as a consequence, significant heating of several hundreds of degrees Celsius is required to release the hydrogen atoms, which renders this storage strategy impractical. The possibility of using carbon-based materials for hydrogen storage thus depends on our ability to tune the dominating interactions with the H2 molecules so that their magnitude lies midway between the above-mentioned cases of physisorption (i.e, van der Waals) and chemisorption. Ideally, one would like to have binding energies between 0.2 and 0.8 eV, which would make it 205 © 2008, Woodhead Publishing Limited

206

Solid-state hydrogen storage

C

sp3

sp2

C

C

C C

σ

C

π C (a)

(b)

C (c)

8.1 The binding orbitals of carbons atoms (i.e, one 2s and three 2p, neglecting spin degeneracy) can hybridize in various ways that favour different types of atomic coordinations. Most importantly, the so-called sp3 hybridization leads to four symmetry-related orbitals that point along tetrahedral directions (a). Such a hybridization favours a three-dimensional coordination of the carbon atoms, as it occurs in the diamond crystal or the methane molecule. (b) In the case of the sp2 hybridization, one 2p orbital remains unhybridized. Such an orbital arrangement leads to planar structures, as in graphite, and is the most relevant to this chapter. (c) The formation of a double bond between two sp2 carbon atoms. The double bond involves a couple of sp2 orbitals, which lead to σ bonds, as well as the two unhybridized 2p orbitals, which form π bonds. Finally (not shown in the figure), so-called sp hybridization may also occur; sp carbons tend to form one-dimensional structures. H C

sp2 C C

C (a)

C

C

C

C

C

~sp3 C

C C

(b)

8.2 (a) The bonding between the carbons in a graphene sheet. All the atoms are in the sp2 hybridization state, and all the corresponding C—C bonds have a mixed character with both σ and π contributions. (b) The partial hybridization shift, from sp2 to sp3, that occurs when a hydrogen atom approaches a carbon of the graphene sheet. Such a change in the electronic structure of the carbon atom permits the formation of a relatively strong C—H bond, with an associated binding energy of several electronvolts. It also implies significant structural changes; for example, the corresponding C—C distances increase by about 0.1 Å.5

possible to retain the hydrogen at operating conditions without application of significant hydrogen pressure, and to release it upon moderate heating that might be provided by a working fuel cell directly. One route towards such an interaction engineering consists in modifying the nanostructure of the material in a way that significantly affects the bonding between carbons.

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

207

For example, carbon atoms in a highly curved graphene sheet should display an electronic configuration with some sp3, as opposed to sp2, character which might result in a relatively strong interaction with hydrogen molecules; also, one may expect stronger interactions associated with defects (e.g. vacancies) in the carbon network. A second possibility would involve the insertion in the carbon structure of heteroatoms that might act as binding sites or charge the carbon network to make it more reactive. Both lines of work are currently receiving much attention; however, while small improvements on the stateof-the-art are frequently reported, a material that is apt for real applications is yet to be discovered. In conclusion, the key issue at this time is to identify a way to bind individual hydrogen molecules with the appropriate strength to a carbonbased structure. It should be noted that this is a rather fundamental problem that pertains to the interactions among a relatively small number of atoms and, thus, is ideally suited for investigation using quantum-mechanical ab initio simulation methods. Indeed, such computational tools offer the possibility of exploring in detail new ideas and thus determine their viability. What is more, the contribution of the simulation works may be particularly important in a field like this one, where it is often the case that many hydrogen-binding sites or mechanisms coexist in real samples, and it may be difficult to characterize each of them individually. (The initial reports of very large hydrogen uptakes by various carbon nanostructures at room temperature,1,2 which turned out to be misleading, constitute a good example of such problems.) Reliable simulations could (and should) play the role of guiding and motivating the experimental work, focusing the efforts on realizing the most promising predictions. The goal of this chapter is to offer an overview of what can be expected from the application of modern computer simulation methods to the study of the fundamental hydrogen interactions in carbon-based materials. It is not my purpose to review the enormous body of literature in an exhaustive way; I just refer to a limited number of works that in my opinion are representative and allow me to illustrate the main points. In Section 8.2 I briefly describe the status of the methods, and their application, in what regards the study of the classical physisorption and chemisorption cases. The main thrust of the chapter is developed in Section 8.3, which is devoted to the description of recent theoretical predictions that carbon-supported light transition metals may constitute ideal binding sites for hydrogen molecules. In Section 8.4 I conclude by discussing the current status and prospects of the field.

8.2

Hydrogen interactions with carbons: physisorption and chemisorption

A hydrogen molecule interacts with the usual carbonaceous structures via weak van der Waals (vdW) forces, binding or physisorbing with an energy

© 2008, Woodhead Publishing Limited

208

Solid-state hydrogen storage

that is experimentally known to range from 0.04 to 0.12 eV, the largest value corresponding to the regime of low hydrogen coverage, where the H2—H2 repulsion is not significant.3,4 A hydrogen atom, on the other hand, reacts chemically with one carbon atom, forcing an orbital hybridization change from sp2 to sp3; the resulting binding or chemisorption energy is calculated (and widely accepted) to be between 2 and 3 eV.5,6 These two situations are reasonably well understood and characterized. The greatest research efforts have been devoted to the physisorption case, which is much closer to the energy range required for practical storage applications.

8.2.1

Physisorption

Most of the work done to date is based on classical effective potentials that describe with sufficient accuracy the relevant interatomic and intermolecular forces, in particular the vdW couplings between hydrogen molecules and the various carbon structures considered (mainly graphite and nanotubes). An important point to note is that, since the electrons are not explicitly treated in such models, the computational cost of the simulations is relatively small, which makes it possible to perform, for realistic model systems composed of hundreds of atoms, sophisticated statistical calculations in the grand-canonical ensemble, and thus compute the amount of hydrogen that can be stored at particular temperature and pressure conditions, and even consider the quantum effects associated to the hydrogen molecules. Useful reviews of the most important results and literature can be found in Hirscher and Becher,7 Meregalli and Parrinello8 and Simonyan and Johnson.9 Various groups worked on the problem employing different effective potentials and simulation techniques, and reached remarkably consistent conclusions. In essence, all the carbon structures considered were found to store a relatively small amount of hydrogen, and that amount was predicted to decrease dramatically as temperature increases. The case of graphitic nanofibres10,11 constitutes an illustrative case: for example, in the idealized case of a 10 Å interplanar distance and a hydrogen pressure of 100 atm, a promising storage capacity of 7–8 wt% was calculated at 77 K, but that value falls to 1.2–1.5 wt% when the temperature rises to 300 K. Note that, at the time they were published, these and similar results were in clear contradiction with the most popular experiments, which reported enormous hydrogen uptakes by nanotubes and nanofibres (up to 67 wt% on graphitic nanofibres2). The experimental situation was eventually clarified,7,12–14 and the correctness of the simulations results, which essentially reflect the small magnitude of the underlying vdW interactions, is now widely accepted. The investigations based on effective potentials suffer from severe limitations. Most importantly, they are restricted to situations in which the

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

209

hydrogen molecules do not dissociate, and tell us nothing about the probability that such a dissociation occurs. Possible variations in the electronic structure of the carbon system (e.g. associated to changes in the curvature of a graphene sheet), and the ensuing changes in the carbon–hydrogen interactions, are not captured either. Such phenomena can be addressed using first-principles methods in which the electronic problem is explicitly treated. Indeed, there is a growing body of literature discussing physisorption problems from first principles; what follows is a brief and critical overview. Currently, the dominant and most efficient first-principles scheme is density functional theory15,16 in its two most popular realizations: the so-called local density and generalized gradient approximations (LDA16 and GGA17, respectively). The status of DFT calculations regarding physisorption problems is an awkward one: both the LDA and the GGA are known to give an essentially wrong description of the vdW forces (e.g. the asymptotic behaviour of 1/r6 is not captured by any of them), but the LDA accidentally renders an attractive interaction of the right magnitude for the systems of interest here18 (the GGA, which is a better theory, predicts a repulsive interaction19). Several groups have thus used LDA methods to study the behaviour of hydrogen molecules ‘physisorbed’ on carbon structures, the works by Alonso et al. arguably being the most interesting ones. While such LDA results have to be considered with caution, some of their conclusions can, in my opinion, be taken as reliable. Most remarkably, static LDA calculations render a high activation energy barrier, of more than 2 eV, for the dissociation of the hydrogen molecule deposited on a nanotube surface,20 which strongly suggest that most of the hydrogen bound to these carbonaceous systems remains molecular, thus supporting the validity of the effective potential studies mentioned above. On the other hand, LDA studies including thermal effects21 suggest that the dissociation of H2 (or, at least, a binding state with some degree of chemisorption) may occur at room temperature, as the atomic vibrations of the carbon structure enhance the interactions with the hydrogen molecule. LDA has also been used22 to predict the maximum H2 coverage that can be attained on a nanotube surface, and to conclude that the electronic character (metallic or semiconducting) of a nanotube does not affect significantly the nature of the carbon–H2 interaction. Recently, other groups have proposed alternative, equally questionable, DFT approaches employing hybrid functionals23 that seem to render results essentially in agreement with those obtained with the LDA. It should be noted here that there are more approximate ways to perform DFT calculations that correctly describe the vdW forces. For example, several groups have recently proposed semi-empirical schemes to correct the usual DFT functionals so that the dispersion forces are modelled properly.24,25 Also, a non-local DFT has been developed26 that incorporates the dispersion forces in a truly ab initio way; this theoretical breakthrough leads to a very

© 2008, Woodhead Publishing Limited

210

Solid-state hydrogen storage

promising simulation scheme, but further work is required to make the calculations affordable from a computational point of view. To the best of my knowledge, none of these methods has so far been used to study problems of hydrogen storage in carbons. Finally, it should be pointed out that there exist some applications of very accurate, and computationally demanding, quantum chemistry and quantum Monte Carlo methods to investigate this type of problems. For example, Gordillo et al.27 reported a diffusion Monte Carlo study of the adsorption of hydrogen in (5,5) carbon nanotubes; they computed binding energies of about 0.15 eV. Also, several groups4,28,29 have published quantum chemical calculations of H2 on graphitic platelets and graphene, obtaining binding energies in the range 0.05–0.10 eV. In addition, it has been found that the curvature of the carbon sheet significantly enhances the attractive interaction with the hydrogen molecule;4 while the structural models considered in these works might be too simple to be realistic, such results clearly point at an interesting possibility that deserves further investigation. To conclude, it is worth noting that these very accurate theoretical results can be used to construct effective models like those mentioned at the beginning of this section, but whose parameters would now be obtained ab initio. One could thus construct atomistic potentials that describe correctly the dependence of the carbon– hydrogen interactions on the curvature of the carbon structure considered, etc. The recent work of Sun et al.30 is a good example of such an approach.

8.2.2

Chemisorption

The adequacy of the usual ab initio methods, including DFT, to study the nature and energetics of chemical bonding is well established and, thus, in this case we do not suffer from the problems described in the previous section on physisorption. One can typically assume that LDA overestimates binding energies by several tenths of electronvolt, while GGA usually reproduces experimental values to within 0.1 or 0.2 eV.17,31 The particular bonds we are interested in, C—C and C—H, are not specially difficult to treat and GGA has been convincingly shown to be reliable.5 Indeed, as far as I know, most of the detailed information we have about the chemisorption of hydrogen atoms in the carbons of interest here come from GGA calculations. As illustrated in Fig. 8.2, individual hydrogen atoms form very stable chemical bonds with the carbon atoms in graphitic structures, including nanotubes. The H—C bond implies a change in the orbital hybridization of the carbon atom, which acquires some sp3 character, and the ensuing structural distortion of the carbon lattice.5 The associated binding energies range from 2 to 3 eV. Hence, breaking such bonds requires significant heating, up to 600 °C,32 which makes chemisorption an impractical mechanism for hydrogen storage applications.

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

211

Hirscher and Becher7 and Meregalli and Parrinella8 give a fair account of the most relevant theoretical studies of hydrogen chemisorption on carbons, and summarize the issues that are typically discussed (e.g., the maximum number of atoms that can be chemisorbed, the differences between chemisorption on the inner and outer surfaces of nanotubes) While there is some interest to such works, they are not discussed in this chapter, since the focus of the book is on hydrogen storage.

8.3

Predictions for hydrogen storage in carbon nanostructures coated with light transition metals

The inclusion of heteroatoms in the carbon structure constitutes the best studied and, in my opinion, most promising alternative to obtain a carbonaceous material where the hydrogen binding energies lie in the 0.2–0.8 eV range. Much of the recent interest in this strategy comes from an experimental report by Chen et al.,33 who claimed a hydrogen storage of up to 20 wt% in Li and K doped carbon nanotubes. While this result was soon shown to be incorrect,34,35 it motivated a number of theoretical works that led to very interesting conclusions: it was shown that alkali metals deposited on nanotubes and graphene can enhance, by charging the carbon network, the interactions with the hydrogen molecules. (See representative references by Froudakis36 and Cabria et al.37; the latter includes a brief summary of, and references to, the theoretical literature on the problem.) Recent work38 on the stable Li12C60 molecule has further shown that the Li atoms can themselves act as a binding site for hydrogen molecules, with average binding energies of about 0.08 eV per H2, thus improving slightly over the physisorption values. The possibility of using heteroatoms as binding sites was introduced simultaneously by the NIST (Yildirim’s) and NREL (Heben’s) groups. Both performed GGA calculations of nanotubes39 and fullerenes40,41 coated with 3d transition metals, and discovered that such idealized systems would constitute a perfect hydrogen storage medium. The rest of this section is devoted to the description of the such findings, discussing in detail the nature of the metal—H2 interactions that make such systems so appropriate for storage purposes. I also comment on current attempts to realize such materials.

8.3.1

Theoretical predictions of an ideal hydrogen storage medium

The main results of Yildirim and Ciraci39, Zhao et al.40 and Yildirim et al.41 are easily summarized. The simulations showed that a light transition metal

© 2008, Woodhead Publishing Limited

212

Solid-state hydrogen storage

atom chemisorbed on a carbon nanostructure, such as a nanotube or a fullerene, can bind several hydrogen molecules, up to four or five. The first molecule binds strongly, with an energy of more than 1 eV, and the additional molecules bind with energies of about 0.5 eV per molecule. Typically, it is found that no activation energy barrier needs to be overcome for the metal—H2 bond to form. The most stable situations found correspond to non-dissociated, but significantly stretched, hydrogen molecules; yet, solutions in which the H2 molecules dissociate are quasi-degenerate in energy with the most stable ones. Thus, at ambient conditions all such situations will coexist in a dynamic way, as explicitly demonstrated by short ab initio molecular dynamics simulations at room temperature.42 The considered carbon nanostructures are stable for dense transition metal coverages and, remarkably, the binding between the metal atoms and the H2 molecules is considerably affected only in the limit of very high coverages. For example, for the case of Ti,39 the metal—H2 binding energy of interest is about 0.5 eV for an isolated metal atom deposited on a (8,0) tube; such energy reduces to about 0.4 eV in a dense-coverage situation corresponding to the chemical formula C8TiH8, and to 0.2 eV for the chemical formula C4TiH8. The calculations also show that the binding energy of the Ti atoms to the carbon structure is rather insensitive to the level of coverage, remaining above 2 eV in all the cases considered. The calculations thus predict that such materials have the potential to store up to a 6–7 wt% of hydrogen, reversibly, at ambient conditions. Zhao et al.40 and Yildirim et al.41 consider the trends across the 3d transition metal family. In essence, it was found that the performance of Sc, Ti, V and Cr is very similar as regards the metal—H2 binding energy, although the chemisorption of the metal on the fullerene is weaker as the atomic number increases. Heavier 3d transition metals seem more problematic for this concept to work, as the binding of the metal atoms to the fullerenes is not energetically favourable41 and the metal—H2 binding energies, computed for C5H5M–Hn complexes, where M is the metal atom, are too large to be practical.40

8.3.2

The metal—H2 binding mechanism

Interestingly, there is little new to this metal—H2 binding mechanism that displays such a convenient energetics for storage purposes. The issue is discussed at some length in the original paper by Yildirim and Ciraci,39 who provided a detailed analysis of the electronic structure of the Ti–4H2 complex adsorbed on a (8,0) nanotube. Here we summarize their discussion. The first thing to note is that Ti-4s electrons are almost completely absent from the computed binding states, and Mulliken population analysis suggests they are promoted into Ti-3d orbitals. Secondly, the binding state just below the Fermi energy of the system displays a marked Ti-3d character, with

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

213

significant C-2p and H-1s contributions. This is clearly reflecting a charge transfer from the Ti atom to empty C-2p and H-1s orbitals; more specifically, to the antibonding orbitals of the 4H2 complex and the C6 ring on which the Ti atom is adsorbed. The third relevant observation pertains to the bonding orbitals of the 4H2 complex, which are found to lie in the energy range from 10 to 6 eV below the Fermi energy. By integrating the density of states associated with such orbitals, Yildirim and Ciraci found that they are about 0.4 electrons short from the nominal occupancies one would expect, and everything indicates the absent electrons have been transferred to empty Ti3d orbitals. Such a peculiar charge sharing between the Ti and H atoms is strongly reminiscent of the back-donation of the Dewar–Chatt–Duncanson model,43–45 which is known to be the interaction mechanism in the so-called metal—H2 Kubas complexes46,47 illustrated in Fig. 8.3. The connection becomes even more obvious when one notes that the calculated H—H distance of the molecules in the Ti–4H2 complex is about 0.84 Å, i.e. significantly longer than the equilibrium bond length of an isolated hydrogen molecule (i.e. 0.74 Å); that is exactly what is found in the case of the so-called agostic hydrogens of a Kubas complex. Finally, in what regards the maximum number of hydrogen molecules that a particular metal atom can bind, the simulations results are compatible with the so-called 18-electron rule,48 which is wellknown in coordination chemistry; this issue has recently been studied in great detail by Jena and co-workers.49 Hence, the phenomenon discovered by Yildirim and Ciraci39, Zhao et 40 al. and Yildirim et al.41 was nothing new to the community of chemists working on organometallic complexes. In retrospect, it is amusing to note it took so long to realize that such agostic hydrogens, which display a bond H

H e–

e–

M

e–

e– H

(a)

M

H

(b)

8.3 Schematic representation of the charge back-donation mechanism that is responsible for the binding between an H2 molecule and a metal atom (denoted by M in the figure) discussed in the text. (a) The charge transfer from the occupied (binding) H2 molecular orbital to appropriately oriented, unoccupied d orbitals of the metal atom. (b) The charge transfer from occupied d orbitals of the metal, with the appropriate symmetry or orientation, to the empty (anti-bonding) H2 molecular orbitals. As a result of such processes, bonding M—H2 orbitals form and the molecular H2 bond weakens, with a concomitant increase in the H—H distance.

© 2008, Woodhead Publishing Limited

214

Solid-state hydrogen storage

length that can be up to 20% longer than that of an isolated hydrogen molecule, might be interesting for storage purposes as they bind to their ligand (i.e. the metal atom) in a way that clearly seems to stand somewhere between physisorption and chemisorption. Thus, as pointed out by Zhao et al.,40 what is new about the theoretical work at NIST and NREL is the demonstration that metal atoms adsorbed on carbonaceous structures can bind several of these agostic hydrogens, and the realization that the computed binding energies have exactly the magnitude required for storage applications. More recent work from the NREL group,50 in which they compare DFT and quantum Monte Carlo results for binding energies in systems that are similar to those discussed above, shows that the DFT binding energies reported39–41 can be expected to be accurate to within 0.1 eV. This is not a surprise, as the bonding in Kubas complexes is, in essence, of a mixed ionic– covalent character and such situations are described with acceptable accuracy within DFT. The NIST group has further investigated the nature of the metal–H2 bonding by computing the associated vibrational spectrum.51 They have found a wealth of distinctive features that could constitute a fingerprint of such Ti–nH2 complexes in a successful experimental realization of this type of system. For example, the agostic hydrogens are found to have characteristic stretchingmode frequencies ranging from about 2850 to about 3700 cm–1 (depending on the hydrogen load and configuration) in contrast with the 4400 cm–1 frequency that would be associated to an isolated H2 molecule; peculiar Ti— H2 and Ti—H stretching frequencies, markedly different from those of the normal mode bands in a titanium hydride, were also identified.

8.3.3

Further theoretical investigations

The works summarized above have led to a flurry of activity, on both the theoretical and experimental fronts. On the simulation side, there has been some effort to discuss the difficulties foreseeable in the actual experimental realization of the proposed hydrogen-storage concept. Most remarkably, Jena and co-workers have argued,52 based on first-principles results, that clustering of the metal atoms is likely to occur for dense coverages of the carbon surfaces. While their study only considers Ti on fullerenes, and does not take into account the actual kinetics of the segregation and cluster formation, it does seem clear that such a clustering may pose a serious difficulty in the preparation and eventual use of such materials. It should also be noted that short molecular dynamics studies by Yildirim42 for temperatures up to 850 K show no Ti diffusion on sparsely covered fullerenes; yet, while encouraging such works do not preclude the possibility of clustering. There is also considerable activity devoted to exploring extensions and variations of the original ideas; for example, there is ongoing work on boron–

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

215

nitride nanostructures,53,54 titanium–carbon nanoparticles,55,56 etc. While valuable, such contributions have not led to any significant advance from the hydrogen storage perspective, and will not be discussed here. Interestingly, the two continuations that might prove most relevant in the long term have come from the teams that proposed the new concept themselves. The NREL team has investigated50 fullerenes doped with light elements (e.g. B and Be) and found that the substitutional atoms can act as appropriate binding sites for hydrogen. While the gravimetric capacity is not remarkable (each B or Be atom binds a single hydrogen molecule), the binding energy is in the right range from 0.2 to 0.6 eV per molecule. Such a convenient interaction relies on a nonclassical three-centre mechanism; it remains to be seen whether it can be made useful for practical purposes. Finally, Ciraci, Yildirim and co-workers have recently demonstrated57 that transition-metal— ethylene complexes retain the storage properties of the fullerene-supported metals, reaching gravimetric storage capacities of 14 wt%. The authors propose that such hydrogen-storing molecules could be incorporated into nanoporous carbons to avoid dimerization and polymerization problems during the hydrogen charging and discharging cycles.

8.3.4

Experimental attempts at realizing the theoretical predictions

The experimental realization of the new storage concept is far from trivial; to the difficulties mentioned above one should add contamination problems during sample preparation and others. Yet some encouraging progress is being made. To my knowledge, the most promising results so far have been published by J.-K. Kang’s group: they studied58 multi-walled carbon nanotubes with Ni nanoparticles dispersed on them, and found that hydrogen amounting to the 2.8 wt% of the material is released in the range of 340–520 K. Further, from their experiments they estimated a hydrogen desorption activation energy of about 0.32 eV per H2 molecule. Such a value is slightly above the 0.26 eV that the same authors compute59 for a hydrogen molecule bound to an individual Ni atom chemisorbed on a nanotube surface; the difference in binding energies is consistent with the fact that in their actual samples they have Ni aggregates, rather than isolated atoms, which should bind the hydrogens more strongly. Thus, Kang’s results clearly suggest that the predicted Dewar–Chatt–Duncanson interaction is at work in their samples. Further investigations to characterize that metal–hydrogen binding, and independent confirmation of the results, are required to determine whether the theoretical predictions have indeed been realized. Other experimental efforts have led to interesting results. Most remarkably, Fischer et al. have synthesized carbide-derived carbons, using TiC and ZrC as precursors, in an attempt to obtain a large amount of accessible metal

© 2008, Woodhead Publishing Limited

216

Solid-state hydrogen storage

atoms on the walls of the pores in the structure.60 The authors find an enhancement in the storage capacity of the metal-containing carbons, but it is unclear yet whether the underlying metal–H2 interaction is the one discussed in this section

8.4

Conclusions and future trends

The theoretical studies described in this chapter constitute representative examples of how the available first-principles methods can be used for accurate studies of carbon nanostructures for hydrogen storage. The limit of weak van der Waals interactions is clearly the most challenging one, but even for that case there are promising techniques, such as the recently introduced empirically corrected DFT schemes,24,25 that are at our disposal. In any case, from the hydrogen-storage perspective, the most interesting phenomena involve a significant charge transfer or orbital hybridization, and for such situations the usual DFT approaches, specially the GGA’s, have a demonstrated accuracy and predictive power. At present, there are several research lines where the first-principles methods can be of considerable help. Section 8.3 is devoted to the possibilities that I consider most promising, which involve the insertion in the carbon structure of heteroatoms that can act as appropriate binding sites. The existing predictions indicate that couplings of the right magnitude can be obtained, and now the theoretical focus should be on the simulation of such mechanisms in more realistic carbons (e.g. graphitic nanofibres, activated carbons, carbide-derived carbons). In particular, we face the challenge of identifying metal species and carbon systems that do not suffer from difficulties associated to metal segregation and clustering, that are resistant to oxidation and other contamination problems, etc. In other words, the theory should make the transition from the proof-of-concept calculations for ideal systems to tackling directly the materials that can be experimentally realized. That is the spirit of our recent work on metal-intercalated graphitic materials.65 Other possibilities to improve the hydrogen storage capacity of carbons should receive further theoretical consideration. We have already mentioned the predictions of an enhancement of the carbon—H2 interactions by curving graphene sheets4 or working with graphite platelets.28 If such a stronger interaction can be induced without causing the dissociation of the H2 molecules (which would lead to the undesired chemisorption of the hydrogen atoms), this would indeed be a very promising strategy; further work is needed to produce predictions that are able to attract the interest of the experimental groups. Some workers have suggested it might be possible to take advantage of defects in the carbon network, as these could provide appropriate binding sites for the hydrogen atoms or molecules.61 The few theoretical investigations have so far considered how the chemisorption of individual hydrogen atoms

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

217

is modified in the presence of, for example, Stone–Wales defects in a graphite sheet.62 However, inducing relatively small variations in the chemisorption energies of hydrogen atoms does not offer much hope for storage applications. In my opinion, the focus should rather be on the possibility that hydrogen molecules bind without dissociation to the defect sites. Such a non-dissociative binding, if it occurs, is likely to be stronger than the usual physisorption, which is exactly what we are seeking. Finally, I should mention here the recently proposed possibility of enhancing hydrogen storage in nanostructured carbons by spillover.63 The energetics of the hypothetical path for a reversible storage process was recently discussed,64 based on first-principles calculations, by the same authors who had previously performed the experiments. According to them, the possibility of releasing atomic hydrogen at moderate temperatures relies on the existence of easy desorption sites in the interior of singlewalled carbon nanotubes. Further simulation and experimental work is required to determine the validity and potential of such an intriguing scheme.

8.5

Sources of further information and advice

The review paper of Hirscher and Becher7 constitutes a carefully written, excellent introduction to the early theoretical work in the field. If anything, it lacks references to some more technical, less well-known, contributions that I have included here (specially in the subsection on physisorption) and which, to my knowledge, had not been collected in any previous review. As for the theoretical methods, one reference that addresses the most popular quantum-simulation scheme (density functional theory), focusing on its application to many materials-science problems, is: R.M. Martin, Electronic Structure: Basic Theory and Practical Methods, Cambridge University Press, 2004; reprinted in 2005. Finally, Kubas complexes are discussed in detail in Ref. 47 which also includes references to the relevant literature.

8.6

References

1. A.C. Dillon, K.M. Jones, T.A. Bekkedahl, C.H. Kiang, D.S. Bethune and M.J. Heben, ‘Storage of hydrogen in single-walled carbon nanotubes’, Nature, 386, 377 (1997). 2. A. Chambers, C. Park, R. Terry, K. Baker and N.M. Rodriguez, ‘Hydrogen storage in graphitic nanofibers’, The Journal of Physical Chemistry B, 102, 4253 (1998). 3. E.L. Pace and A.R. Siebert, ‘Heat of adsorption of parahydrogen and orthodeuterium on graphon’, Journal of Physical Chemistry, 63, 1398 (1959). 4. Y. Okamoto and Y. Miyamoto, ‘Ab initio investigation of physisorption of molecular hydrogen on planar and curved graphenes’, Journal of Physical Chemistry B, 105, 3470 (2001). 5. S.M. Lee and Y.H. Lee, ‘Hydrogen storage in single-walled carbon nanotubes’, Applied Physics Letters, 76, 2877 (2000).

© 2008, Woodhead Publishing Limited

218

Solid-state hydrogen storage

6. O. Gülseren, T. Yildirim and S. Ciraci, ‘Tunable adsorption on carbon nanotubes’, Physical Review Letters, 87, 116802 (2001). 7. M. Hirscher and M. Becher, ‘Hydrogen storage in carbon nanotubes’, Journal of Nanoscience and Nanotechnology, 3, 3 (2003). 8. V. Meregalli and M. Parrinello, ‘Review of theoretical calculations of hydrogen storage in carbon-based materials’, Applied Physics A, 72, 143 (2001). 9. V.V. Simonyan and J.K. Johnson, ‘Hydrogen storage in carbon nanotubes and graphitic nanofibers’, Journal of Alloys and Compounds, 330–332, 659 (2002). 10. M. Rzepka, P. Lamp and M.A. de la Casa-Lillo, ‘Physisorption of hydrogen on microporous carbon and carbon nanotubes’, Journal of Physical Chemistry B, 102, 10894 (1998). 11. Q. Wang and J.K. Johnson, ‘Computer simulations of hydrogen adsorption on graphitic nanofibers’, Journal of Physical Chemistry B, 103, 277 (1999). 12. A.C. Dillon and M.J. Heben, ‘Hydrogen storage using carbon nanotubes: past, present and future’, Applied Physics A, 72, 133 (2001). 13. A. Züttel, P. Sudan, Ph. Mauron, T. Kiyobayashi, Ch. Emmenegger and L. Schlapbach, ‘Hydrogen storage in carbon nanostructures’, International Journal of Hydrogen Energy, 27, 203 (2002). 14. B. Panella, M. Hirscher and S. Roth, ‘Hydrogen adsorption in different carbon nanostructures’, Carbon, 43, 2209 (2005). 15. P. Hohenberg and W. Kohn, ‘Inhomogeneous electron gas’, Physical Review, 136, B864 (1964). 16. W. Kohn and L.J. Sham, ‘Self-consistent equations including exchange and correlation effects’, Physical Review, 140, A1133 (1965). 17. J.P. Perdew, K. Burke and M. Ernzerhof, ‘Generalized Gradient Approximation made simple’, Physical Review Letters, 77, 3865 (1996). 18. J.S. Arellano, L.M. Molina, A. Rubio and J.A. Alonso, ‘Density functional study of adsorption of molecular hydrogen on graphene layers’, Journal of Chemical Physics, 112, 8114 (2000). 19. K. Tada, S. Furuya and K. Watanabe, ‘Ab initio study of hydrogen adsorption to single-walled carbon nanotubes’, Physical Review B, 63, 155405 (2001). 20. J.A. Alonso, J.S. Arellano, L.M. Molina, A. Rubio and M.J. López, ‘Interaction of molecular and atomic hydrogen with single-wall carbon nanotubes’, IEEE Transactions on Nanotechnology, 3, 304 (2004). 21. H. Cheng, G.P. Pez and A.C. Cooper, ‘Mechanism of hydrogen sorption in singlewalled carbon nanotubes’, Journal of the American Chemical Society, 123, 4845 (2001). 22. I. Cabria, M.J. López and J.A. Alonso, ‘Density functional study of molecular hydrogen coverage on carbon nanotubes’, Computational Materials Science, 35, 238 (2006). 23. G. Mpourmpakis, G.E. Froudakis, G.P. Lithoxoos and J. Samios, ‘Effect of curvature and chirality for hydrogen storage in single-walled carbon nanotubes: a combined ab initio and Monte Carlo investigation’, The Journal of Chemical Physics, 126, 144704 (2007). 24. X. Wu, M.C. Vargas, S. Nayak, V. Lotrich and G. Scoles, ‘Towards extending the applicability of density functional theory to weakly bound systems’, Journal of Chemical Physics, 115, 8748 (2001). 25. Q. Wu and W. Yang, ‘Empirical correction to density functional theory for van der Waals interactions’, Journal of Chemical Physics, 116, 515 (2002). 26. D.C. Langreth, M. Dion, H. Rydberg, E. Schröder, P. Hyldgaard and B.I. Lundqvist,

© 2008, Woodhead Publishing Limited

Modelling of carbon-based materials for hydrogen storage

27. 28. 29.

30. 31.

32.

33.

34. 35.

36. 37.

38. 39.

40. 41.

42. 43. 44.

45.

219

‘Van der Waals Density Functional Theory with applications’, International Journal of Quantum Chemistry, 101, 599 (2005). M.C. Gordillo, J. Boronat and J. Casulleras, ‘Isotopic effects of hydrogen adsorption in carbon nanotubes’, Physical Review B, 65, 014503 (2001). T. Heine, L. Zhechkov and G. Seifert, ‘Hydrogen storage by physisorption on nanostructure graphite platelets’, Physical Chemistry Chemical Physics, 6, 980 (2004). A. Ferre-Vilaplana, ‘Numerical treatment discussion and ab initio computational reinvestigation of physisorption of molecular hydrogen on graphene’, The Journal of Chemical Physics, 122, 104709 (2005). D.Y. Sun, J.W. Liu, X.G. Gong and Z.-F. Liu, ‘Empirical potential for the interaction between molecular hydrogen and graphite’, Physical Review B, 75, 075424 (2007). X.J. Kong, C.T. Chan, K.M. Ho, and Y.Y. Ye, ‘Cohesive properties of crystalline solids by the generalized gradient approximation’, Physical Review B, 42, 9357 (1990). A. Nikitin, H. Ogasawara, D. Mann, R. Denecke, Z. Zhang, H. Dai, K. Cho and A. Nilsson, ‘Hydrogenation of single-walled carbon nanotubes’, Physical Review Letters, 95, 225507 (2005). P. Chen, X. Wu, J. Lin and K.L. Tan, ‘High H2 uptake by alkali-doped carbon nanotubes under ambient pressure and moderate temperatures’, Science, 285, 91 (1999). R.T. Yang, ‘Hydrogen storage in alkali-doped carbon nanotubes revisited’, Carbon, 38, 623 (2000). F.E. Pinkerton, B.G. Wicke, C.H. Olk, G.G. Tibbetts, G.P. Meisner, M.S. Meyer and J.F. Herbst, ‘Thermogravimetric measurement of hydrogen absorption in alkalimodified carbon materials’, Journal of Physical Chemistry, 104, 9460 (2000). G.E. Froudakis, ‘Why alkali-metal-doped carbon nanotubes possess high hydrogen uptake’, Nano Letters, 1, 531 (2001). I. Cabria, M.J. López and J.A. Alonso, ‘Enhancement of hydrogen physisorption on graphene and carbon nanotubes by Li doping’, The Journal of Chemical Physics, 123, 204721 (2005). Q. Sun, P. Jena, Q. Wang and M. Marquez, ‘First-principles study of hydrogen storage in Li12C60’, Journal of the American Chemical Society, 128, 9741 (2006). T. Yildirim and S. Ciraci, ‘Titanium-decorated carbon nanotubes as a potential high-capacity hydrogen storage medium’, Physical Review Letters, 94, 175501 (2005). Y. Zhao, Y.-H. Kim, A.C. Dillon, M.J. Heben and S.B. Zhang, ‘Hydrogen storage in novel organometallic buckyballs’, Physical Review Letters, 94, 155504 (2005). T. Yildirim, J. Íñiguez and S. Ciraci, ‘Molecular and dissociative adsorption of multiple hydrogen molecules on transition metal decorated C60’, Physical Review B, 72, 153403 (2005). T. Yildirim, private communication. M.J.S. Dewar, ‘A review of the pi-complex theory’, Bulletin de la Société Chimique de France, 18, C71 (1951). J. Chatt and L.A. Duncanson, ‘Olefin co-ordination compounds. Part III. Infra-red spectra and structure: attempted preparation of acetylene complexes’, Journal of the Chemical Society, (Oct.) 2939 (1953). J.-Y. Saillard and R. Hoffmann, ‘C—H and H—H activation in transition metal complexes and on surfaces’, Journal of the American Chemical Society, 106, 2006 (1984).

© 2008, Woodhead Publishing Limited

220

Solid-state hydrogen storage

46. G.J. Kubas, ‘Molecular hydrogen complexes: coordination of a sigma bond to transition metals’, Accounts of Chemistry Research, 21, 120 (1988). 47. G.J. Kubas, ‘Metal–dihydrogen and sigma-bond coordination: the consummate extension of the Dewar–Chatt–Duncanson model for metal–olefin pi bonding’, Journal of Organometallic Chemistry, 635, 37 (2001). 48. J.E. Huheey, E.A. Keiter and R.L. Keiter, Inorganic Chemistry: Principles of Structure and Reactivity, (Harper Collins College Publishers, New York, 1993). 49. B. Kiran, A.K. Kandalam and P. Jena, ‘Hydrogen storage and the 18-electron rule’, The Journal of Chemical Physics, 124, 224703 (2006). 50. Y.-H. Kim, Y. Zhao, A. Williamson, M.J. Heben and S.B. Zhang, ‘Nondissociative adsorption of H2 molecules in light-element-doped fullerenes’, Physical Review Letters, 96, 016102 (2006). 51. J. Íñiguez, W. Zhou and T. Yildirim, ‘Vibrational properties of TiHn complexes adsorbed on carbon nanostructures’, Chemical Physics Letters, 444, 140 (2007). 52. Q. Sun, Q. Wang, P. Jena and Y. Kawazos, ‘Clustering of Ti on a C60 surface and its effect on hydrogen storage’, Journal of the American Chemical Society, 127, 14582 (2005). 53. S.A. Shevlin and Z.X. Guo, ‘Transition-metal-doping-enhanced hydrogen storage in boron nitride systems’, Applied Physics Letters, 89, 153104 (2006). 54. X. Wu, J.L. Yang and X.C. Zeng, ‘Adsorption of hydrogen molecules on the platinumdoped boron nitride nanotubes’, The Journal of Chemical Physics, 125, 044704 (2006). 55. N. Akman, E. Durgun, T. Yildirim and S. Ciraci, ‘Hydrogen storage capacity of titanium met-cars’, Journal of Physics: Condensed Matter, 18, 9509 (2006). 56. Y. Zhao, A.C. Dillon, Y.H. Kim, M.J. Heben and S.B. Zhang, ‘Self-catalyzed hydrogenation and dihydrogenation adsorption on titanium carbide nanoparticles’, Chemical Physics Letters, 425, 273 (2006). 57. E. Durgun, S. Ciraci, W. Zhou and T. Yildirim, ‘Transition-metal-ethylene complexes as high-capacity hydrogen-storage media’, Physical Review Letters, 97, 226102 (2006). 58. H.-S. Kim, H. Lee, K.-S. Han, J.-H. Kim, M.-S. Song, M.-S. Park, J.-Y. Lee and J.K. Kang, ‘Hydrogen storage in Ni nanoparticle-dispersed multiwalled carbon nanotubes’, Journal of Physical Chemistry B, 109, 8983 (2005). 59. J.W. Lee, H.S. Kim, J.Y. Lee and J.K. Kang, ‘Hydrogen storage and desorption properties of Ni-dispersed carbon nanotubes’, Applied Physics Letters, 88, 143126 (2006). 60. J.E. Fischer, Y. Gogotsi, T. Yildirim et al., unpublished. (A brief summary of the results can be found at www.hydrogen.energy.gov/pdfs/progress06/iv_c_4_fischer.pdf.) 61. S. Orimo, A. Züttel, L. Schlapbach, G. Majer, T. Fukunaga and H. Fujii, ‘Hydrogen interaction with carbon nanostructures: current situation and future prospects’, Journal of Alloys and Compounds, 356–357, 716 (2003). 62. S. Letardi, M. Celino, F. Cleri and V. Rosato, ‘Atomic hydrogen adsorption on a Stone-Wales defect in graphite’, Surface Science, 496, 33 (2002). 63. A.J. Lachawiec, Q. Gongshin and R.T. Yang, ‘Hydrogen storage in nanostructured carbons by spillover: bridge-building enhancement’, Langmuir, 21, 11418 (2005). 64. F.H. Yang, A.J. Lachawiec and R.T. Yang, ‘Adsorption of spillover hydrogen atoms on single-wall carbon nanotubes’, Journal of Physical Chemistry B, 110, 6236 (2006). 65. M. Cobain and J. Iñiguez, ‘First-principles investigation of hydrogen storage in metal-intercalated graphitic materials, Journal of Physics: Condensed Matter, 20, 285212 (2008),

© 2008, Woodhead Publishing Limited

Part III Physically bound hydrogen storage

221 © 2008, Woodhead Publishing Limited

9 Storage of hydrogen in zeolites P. A. A N D E R S O N, University of Birmingham, UK

9.1

Introduction

The Swedish mineralogist Baron Axel Crønstedt is credited with the discovery in 1756 of the class of aluminosilicate materials known as zeolites (Crønstedt, 1756). The name – literally ‘boiling stone’ [Greek zeos (to boil) and lithos (stone)] – derives from his observation that on heating the mineral stilbite with a blowtorch it hissed and bubbled as steam was evolved. Since then approximately 40 naturally occurring zeolites have been discovered and over 130 entirely synthetic structures have been prepared (Baerlocher et al., 2001), many of which have no natural counterpart. A defining feature of zeolites is their regular intracrystalline network of pores and channels of subnanometre dimensions that, depending on precise composition, result in internal surface areas of up to c. 1000 m2/g. The general formula for a zeolite is: (Mn+)x/n[(AlO2)x(SiO2)y]x–· wH2O where cations M of valence n balance the anionic charge on the framework, resulting from the presence of trivalent aluminium. The silicon to aluminium ratio of a zeolite is crucially important in determining its chemical properties. An empirical rule due to Löwenstein (1954) forbids the presence of Al–O– Al linkages in the structure, thus requiring the Si/Al ratio to be greater than, or equal to, 1. The charge-balancing cations required in aluminosilicates are normally alkali metal or alkaline-earth metal ions, such as Na+, K+, Ca2+ and Ba2+, in naturally occurring zeolites; however, synthetic zeolites frequently contain both inorganic and organic cations. Zeolites have a number of important industrial uses (Dyer, 1988). The size of the zeolite channels and cavities means that, along with the cations, intracrystalline water is readily accommodated, which has the effect of making the cations highly mobile and thus exchangeable, leading to the use of zeolites as ion-exchangers, e.g. in water softeners and in the clean-up of radioactive waste. In addition, a large range of different metal cations may readily be 223 © 2008, Woodhead Publishing Limited

224

Solid-state hydrogen storage

incorporated into zeolites through ion-exchange from aqueous solution. In most cases water can be reversibly removed with little or no change to the lattice structure, leaving the exchangeable cations coordinated to a number of the oxygen atoms of the framework, which bear a net negative charge and are usually located within the same aluminosilicate ring. Dehydrated zeolites exhibit a strong affinity for water resulting in their widespread use as desiccants. Alternatively, a wide range of other adsorbate molecules may be absorbed into the pores. Zeolites are also known as ‘molecular sieves’, a term first used by McBain (1932), which arises from their ability selectively to absorb molecules of the correct dimensions to fit into their cavities or channels. This property has contributed to their large-scale commercial use as heterogeneous catalysts and in gas separation processes. The structure of a zeolite framework is dependent on how the tetrahedral building units, TO4 (where T = Si or Al), are connected. Many different structures are known (Baerlocher et al., 2001), all of which have characteristic topological features. For instance, linear or pseudo-linear channels may be formed with diameters ranging between 4.2 and 7.4 Å, or cages with diameters of 6.4–12 Å and window sizes from 2.3 Å (6-ring) to 13 Å (18-ring). The frameworks are designated by a three letter code indicating their structure type. As it is possible for two or more zeolites of different composition to have the same framework topology but different names, it is helpful to specify the structure type code in brackets. Zeolites are often classified into families according to which type of rings and cages, also known as ‘secondary building units’, they contain (see Fig. 9.1). The truncated octahedron known as the sodalite (or β-) cage, for example, can be connected in different ways (see Fig. 9.2) to form structures exhibiting diverse properties such as SOD (as seen in the mineral sodalite), LTA (in zeolite A) and FAU (zeolites X and Y). The commercial importance and diverse uses of zeolites have long prompted attempts to synthesise related compounds. Many framework materials containing atoms other than silicon and aluminium are now known. These materials, known as ‘zeotypes’, are not zeolites in the strictest sense as they are not aluminosilicates. Examples include the aluminophosphates, commonly known as AlPOs (Wilson et al., 1982), silicoaluminophosphates – SAPOs (Lok et al., 1984) – and gallophosphates – GaPOs (Parise, 1985).

9.2

Hydrogen encapsulation at high temperatures

In contrast to current interest in the use of high surface area solids such as zeolites for the physisorption of hydrogen under cryogenic conditions, the earliest work in the storage of hydrogen in zeolites was concerned with a completely different phenomenon. At room temperature and below hydrogen molecules do not enter certain zeolite cages, but at elevated temperatures

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

S4R

S6R

S8R

D4R

D6R

D8R

β -cage

α -cage

225

Supercage (FAU)

9.1 Some common secondary building units showing oxygen atoms (light) and silicon or aluminium atoms (dark).

and pressures, in many cases hydrogen may be forced into the cages, with enhanced vibrational motion assisting access through the apertures. Upon returning to ambient conditions, hydrogen may remain trapped inside the cages, only to be released on subsequent reheating. The phenomenon was first examined by Fraenkel and Shabtai (1977) in a pioneering study on ion-exchanged zeolite A (LTA) that established the critical influence of the zeolite exchangeable cations on the amount of hydrogen encapsulated. In experiments carried out at temperatures between 200 and 400 °C, in the pressure range 20–920 bar, they found the amount of hydrogen encapsulated first to increase with ionic radius for Na+ and K+ before decreasing for Rb+ and Cs+. These observations were interpreted as resulting from the interplay of two factors as the cation radius increased: the decrease in effective pore size and consequent increase in encapsulate stability; and the decrease in available pore volume within the zeolite. Despite the slightly lower capacities RbA and CsA showed more promise for hydrogen encapsulation as CsA was found to lose only 7% of its encapsulated hydrogen after 5 days at 25 °C, whereas KA lost 46% of its hydrogen after 45 hours. On loading at 300 °C and 917 bar, a hydrogen encapsulation capacity of 65 cm3 g–1 (at standard temperature and pressure) was achieved for CsA (c. 0.6 wt%), which was regarded as quite promising when compared with other hydrogen storage materials of the day. In a subsequent detailed analysis of the desorption of hydrogen encapsulated in zeolite A, Fraenkel (1981) reported that a combination of Cs+ and Na+ cations was most efficient in trapping hydrogen. Two different diffusion pathways for hydrogen were proposed based on the passage of hydrogen

© 2008, Woodhead Publishing Limited

226

Solid-state hydrogen storage

from one α-cage to another or from an α-cage to a β-cage and vice versa. The activation energy for diffusion was found to depend on the size of the exchangeable cations and, in addition, a marked increase was observed in NaCsA when the number of Cs+ ions per unit cell was increased. It was proposed that in NaA the 8-rings that control diffusion between neighbouring α-cages (see Fig. 9.2) may be regarded as ‘open’ as the small sodium cations do not hinder the diffusion of hydrogen. In this zeolite hydrogen can be trapped effectively only in the smaller β-cages and capacity is therefore low. On ion-exchange it is these sodium ions occupying the 8-ring sites that are replaced first with caesium, and a large increase in capacity was observed at around 2.5 Cs+ ions per unit cell (Pm 3 m ) as α-cage encapsulation commenced (Fraenkel, 1981). On account of inhomogeneities in the cation distribution within the α-cages, the 8-ring windows were said to be ‘conditionally open’. Conditionally opened windows are formed as a result of electrostatic distortion leading to exchangeable cation displacement. When NaA is ion-exchanged with Cs+ ions, cation replacement is believed to occur first at 8-ring sites until these become fully occupied by three Cs+ ions. Meanwhile, there are eight 6-ring site Na+ ions surrounding the 8-ring site in perfect symmetry together with one Na+ ion at a site opposite the D4R. The Na+ ion at this site may to some extent displace the exchangeable ion at the 8-ring site in order to minimise electrostatic repulsion, thus providing a conditionally opened window.

Sodalite (or β-) cage 6-ring (S6R) β-cage D4R

SOD structure (sodalite)

8-ring (S8R)

D6R

α-cage LTA structure (zeolite A)

FAU structure (zeolites X & Y)

9.2 Zeolite structures based on the sodalite cage structural unit (the vertices of the rings and polyhedra are occupied by Si or Al atoms and the lines represent the oxygen bridges).

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

227

When the fourth Cs+ ion goes into the structure it resides in the β-cage in exchange for a Na+ ion of the 6-ring site. As a result the 8-ring site-occupying Cs+ ion returns to its position shutting the window permanently. Further Cs+ exchange occurs exclusively at 6-ring sites. As the number of Cs+ ions per unit cell increased from 2.5 to 4, an increase in thermal stability of the encapsulate was observed as the conditionally open windows became fully closed, but a decrease in hydrogen capacity, attributed by Fraenkel to a reduction in effective pore volume. In studies on the pressure and temperature dependence of hydrogen encapsulation in NaA and Cs2.5Na9.5A, a linear dependence of hydrogen capacity on applied pressure was found (Heo et al., 1991; Yoon and Heo, 1992; Yoon, 1993). A capacity of 0.04 wt% H2 was observed in the former at 153 bar and 350 °C, and 0.17 wt% was achieved in the latter and the same temperature at 129 bar. The difference was consistent with Fraenkel’s earlier suggestion that encapsulation occurred only in the β-cages in NaA. It was also noted that, in theory, higher capacities should be achievable at lower temperatures and the observation that this did not occur in practice was ascribed to an inability of hydrogen molecules to pass efficiently through the cage windows at lower temperatures. Weitkamp et al. (1995) investigated the storage capacity of a range of zeolites of different structure and composition at pressures of 25−100 bar and temperatures between 20 and 300 °C. The zeolites, including A, sodalite (SOD), X and Y (FAU), sigma-1 (DDR), ZK5 (KFI), rho (RHO) and the silicoaluminophosphate zeolite A analogue SAPO42 (LTA), were chosen on the basis that they share certain structural features such as α- and β-cages. Hydrogen capacities were found to be strongly dependent on framework structure and composition and, in general, were much lower than those observed by Fraenkel and co-workers. In contrast to previous work, Weitkamp et al., concluded that hydrogen was primarily trapped in small cages such as βcages rather than larger α-cages and accordingly the highest storage capacity of 0.08 wt%, after loading at 100 bar and 300 °C, was observed in sodalite, which consists entirely of β-cages. Krishnan et al. (1996) demonstrated the importance of cations in hydrogen encapsulation in zeolite rho. Up to 0.124 wt% H2 was encapsulated at only 1 bar and 200 °C in CdCs-rho, but no evidence of hydrogen uptake was observed in cation free H-rho. Temperature programmed desorption (TPD) measurements showed that hydrogen trapped in CdCs-rho was released in three separate peaks at 107, 298 and 340 °C, which were attributed to different encapsulation sites.

9.3

Low-temperature physisorption

Although the physisorption of hydrogen on zeolites has been studied for 40 years or more (Basmadjian, 1960), early work focused on separation phenomena

© 2008, Woodhead Publishing Limited

228

Solid-state hydrogen storage

and relatively little attention was paid at first to adsorption capacities. In their study on the use of molecular hydrogen as a probe for acidity testing in zeolites, Makarova et al. (1994) measured the adsorption of H2 at 77 K on a number of hydrogen-exchanged zeolites. They found that the total adsorption capacity at a given pressure was an inverse function of pore size, increasing in the order HY (FAU) < H-ZSM5 (MFI) < H-mordenite. The highest observed hydrogen uptake was c. 0.7 wt% at 0.65 bar in H-mordenite, but fitting of the data to the Dubinin equation predicted that considerably higher capacities should be possible at higher pressures in HY. The same year, Ruthven and Farooq (1994) reported that CaNaA (LTA) adsorbed 0.44 wt% at 77 K and 0.1 bar. Kazansky et al. (1998a) studied hydrogen adsorption at 77 K on a range of isostructural zeolites: NaY and three samples of NaX (FAU) with different Si/Al ratios. The amount of hydrogen adsorbed was found to increase in the sequence NaY (Si/Al = 2.4; 56 Na+ per unit cell) < NaX (Si/Al = 1.4; 80 Na+) < NaX (Si/Al = 1.2; 87 Na+) < NaX (Si/Al = 1.05; 94 Na+) with adsorption corresponding to c. 1.2 wt% reported for NaX (Si/Al = 1.05) at 0.6 bar. The close relationship between the amount of adsorbed hydrogen and the number of sodium ions in the zeolite structure indicated that the cations act as adsorption sites, but the authors also found that the number of hydrogen molecules adsorbed per sodium cation increased with decreasing Si/Al ratio. The strength of the interaction between the cations and H2 is expected to decrease with increasing framework basicity (decreasing Si/Al ratio) as a result of the reduced effective positive charge on the sodium cations, so this observation cannot be explained through interactions with the sodium cations alone. It was suggested that the strength of hydrogen adsorption could be enhanced through the interaction of hydrogen with the oxygen atoms of the zeolite framework and that the strength of this interaction increased with increasing framework basicity. A detailed analysis of the physisorption of molecular hydrogen isotopes on NaA in the temperature range 40–120 K at pressures up to 0.8 bar, also using the Dubinin model, was carried out by Stéphanie-Victoire et al. (1998). At a given temperature and pressure the amount of hydrogen adsorbed increased in the order H2 < HD < D2, reflecting the decrease in liquid phase volume per molecule in the same sequence. A hydrogen uptake of nearly 16 molecules of H2 (17 for D2) per primitive unit cell (Pm 3 m ) was observed at 40 K and 0.8 bar. This corresponds to a capacity of 1.9 wt% H2, but significantly the value at 77 K under the same pressure was around two-thirds of that at 40 K, indicating that the zeolite was far from saturated at liquid nitrogen temperatures. The isosteric heats of adsorption (Qst) were strongly dependent on coverage, falling steadily from 10.7 kJ mol–1 at loading levels of 0.5 H2 molecules per unit cell to 6.2 kJ mol–1 at 10 H2 per unit cell, still significantly higher than the heat of liquefaction of H2 (0.9 kJ mol–1). The good agreement with the

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

229

Dubinin model was taken by the authors as evidence that hydrogen displays a fluid behaviour when adsorbed in NaA and an estimate of accessible pore volume led them to infer that all three molecules could enter the smaller βcages. Nijkamp et al. (2001) performed hydrogen adsorption measurements on a wide range of carbonaceous and silica/alumina-based materials, including zeolites L (LTL), ZSM5 and ferrierite (FER), at 77 K and pressures up to 1 bar. Upon analysis of the results, they concluded that adsorption of hydrogen under these conditions was dependent on the presence of a large volume of micropores with a suitable diameter. A capacity of 0.72 wt% was recorded for zeolite ZSM5 and 0.54 wt% for the mesoporous silica MCM 41. Forster et al. (2003) examined hydrogen adsorption in the zeolitic nickel phosphates VSB1 and VSB5 at 77 K. Much stronger adsorption found in the latter was held to be related to the presence of coordinatively unsaturated framework Ni2+ cations, but storage capacities remained well below 1 wt% at pressures up to 1 bar. Jhung et al. (2005) measured adsorption capacities in a range of porous materials including zeolites, SAPOs and AlPO5 (AFI), finding a maximum capacity at 77 K and 1 bar of 160 ml g–1 (c. 1.4 wt%) in SAPO34 (CHA). In contrast to the more usual vacuum adsorption experiments, Yuvaraj et al. (2003) studied the replacement of nitrogen by hydrogen in Na-mordenite (MOR) and K L (LTL), on heating from 173 to 423 K in a flow of 10% hydrogen in nitrogen after initial cooling to 143 K in nitrogen. In Na-mordenite the replacement was found to take place in a number of stages occurring at different temperatures: replacement of gaseous nitrogen in the main channels (185 K) and in the side pockets (240 K); replacement of nitrogen said to be ‘condensed’ around Na+ cations, in the channels (220 K) and in the side pockets (270 K); and desorption of nitrogen bound at Na+–Oδ– pairs (300, 350 and 390 K). At the temperatures studied, however, the amount of hydrogen present in both zeolites remained low ( 1000 cm–1, have been reported for coppercontaining ZSM5 (Solans-Monfort et al., 2004; Serykh and Kazansky, 2004; Kazansky and Serykh, 2004c; Spoto et al., 2004; Kazansky and Pidko, 2005). Bands observed at 3079 and 3130 cm-1 after adsorption of hydrogen on Cu(I)-ZSM5, prepared through reaction of CuCl vapour with H-ZSM5 were assigned to hydrogen adsorbed on Cu+ ions coordinated to two and three framework oxygens, respectively (Solans-Monfort et al., 2004; Spoto et al., 2004). The magnitude of these interactions was ascribed to strong synergistic bonding between the cations and hydrogen molecules in a side-on configuration, involving electron donation from the H2 σ orbital and back donation from the 3dπ Cu+ orbitals and resulting in three-centre Cu+–H2 covalent bonding

© 2008, Woodhead Publishing Limited

238

Solid-state hydrogen storage

that significantly weakens the H–H bond (see Fig. 9.4). Since the work of Kubas and Ryan (1984), this mode of bonding between transition metals and hydrogen molecules, forming so-called ‘T-shaped’ complexes, is well established and results in a highly red shifted but relatively weak IR band, as the H–H stretch is not accompanied by strong dipole oscillations in this configuration. The strength of the interaction (estimated at 60–80 kJ mol–1) is more than five times greater than that observed in Na-ZSM5 (11 kJ mol–1) and must be regarded as chemisorption rather than physisorption. The Cu+(η2-H2) complexes formed were found to be stable at room temperature and subambient pressure. Serykh and Kazansky (2004) used D2 and H2-D2 mixtures to confirm that similar bands at 3082 and 3130 cm–1 observed in CuH-ZSM5 prepared through reduction of Cu2+ ion-exchanged NH4–ZSM5 were indeed due to strongly adsorbed hydrogen. Detailed volumetric measurements confirmed that less than half the Cu+ ions present were apparently able to adsorb hydrogen strongly, leading the authors to conclude that the adsorption sites might be associated with dimeric Cu+ species, though this was not confirmed by later work (Kazansky and Pidko, 2005). For the first time bands were also observed that were ascribed to vibrations of the Cu–H2 bonds. Berlier et al., (2006) reported the spectroscopic characterisation of Fesilicalite (MFI) after thermal treatment through transmission FTIR. The resulting materials were complex, containing both framework and nonframework Fe, the latter present as both Fe2+ and Fe3+ ions, and significant Brønsted acidity. IR bands were observed corresponding to hydrogen adsorbed at Brønsted sites, two separate Fe2+ and two Fe3+ sites, with the bathochromic shifts considerably greater for the Fe 3+…H 2 adducts (∆νHH = –172, –202 cm–1) than for the Fe2+…H2 counterparts (∆νHH = –112, –134 cm–1). Interestingly, the observed red shifts were three times greater than that observed H1

H2 Cu

Al

9.4 Calculated structure of hydrogen adsorbed on a Cu(I)-ZSM5 (MFI) fragment. Adapted from Figure 10 in Kazansky and Pidko (2005).

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

239

in Fe3+-containing H-ZSM5 (Sigl et al., 1997) and evidence that, as in the Cu(I) case, the bonding is not entirely electrostatic in nature was discussed. In a detailed diffuse reflectance and transmission FTIR study on the hydrogen forms of three different zeolites with the chabazite structure (CHA), Regli et al. (2005) confirmed the finding that although hydrogen adsorption increased with increasing strength of the Brønsted acid sites, it decreased with increasing density of the sites (Zecchina et al., 2005a). The reduction was ascribed to hydrogen bonding between acid sites in close proximity to each other, resulting in a situation where the energetic cost of disrupting the hydrogen bonding is higher than the adsorption enthalpy of the OH…H2 adduct. In an important recent development, Otero Areán et al. (2003) applied the method of Paukshtis and Yurchenko (1983) to examine the thermodynamics of hydrogen adsorption on Li-exchanged ZSM5. The method involved the simultaneous measurement of temperature, equilibrium pressure and IR band intensity, over a range of temperatures, and produced values of ∆Ho = –6.5 (±0.5) kJ mol–1 and ∆So = –90 (±0.5) J mol–1 K–1 for the standard adsorption enthalpy and entropy, respectively. The latter is considerably smaller in magnitude than the standard entropy of hydrogen gas, indicating considerable residual freedom of the molecule in the adsorbed state. The application of the method to zeolites and related materials has recently been reviewed by Garrone and Otero Areán (2005). Thermodynamic parameters derived in this way for various zeolites are summarised in Table 9.1. From even this limited data set it is clear that the adsorption enthalpies are strongly dependent both on the nature of the cations and the zeolite framework. Unsurprisingly, the highest adsorption enthalpy reported so far was for the highly polarising divalent Mg2+ cation in magnesium-containing zeolite Y, with an observed red shift in the IR of ∆νHH = –107 cm–1 (Turnes Palomino et al., 2006; Otero Areán et al., 2007). Table 9.1 Thermodynamic parameters for the adsorption of hydrogen in zeolites, derived from variable temperature IR measurements Zeolite

Structure type

o ∆H ads /kJ mol–1

o ∆S ads /J mol–1 K–1

Reference

Li-ZSM5 Na-ZSM5 K-ZSM5 HNaY Na-ferrierite K-ferrierite Li-ferrierite MgNaY NaK-ETS10 (Na+ adsorption)

MFI MFI MFI FAU FER FER FER FAU n/a

–6.5 –10.3 –9.1 –11.7 –6.0 –3.5 –4.1 –18.2

–90 –121 –124

Otero Areán et al. (2003) Otero Areán et al. (2005) Otero Areán et al. (2005) Gribov et al. (2006) Otero Areán et al. (2006) Otero Areán et al. (2006) Nachtigall et al. (2006) Otero Areán et al. (2007a)

© 2008, Woodhead Publishing Limited

–8.7

–78 –57 –57 –136

Ricchiardi et al. (2007)

240

Solid-state hydrogen storage

Ricchiardi et al. (2007) reported the observation by IR of hydrogen adsorption on ETS10. Distinct bands from ortho- and para-H2 in different adsorbed states were observed, and the conversion of ortho–para was measured over a timescale of hours, indicating the presence of a catalysed reaction. Hydrogen adsorption at 20 K was found to occur in three different regimes: at low pressures ordered 1:1 adducts with Na and K ions exposed in the channels of the material were formed, which gradually converted into ordered 2:1 adducts, with further addition of H2 occurring through the formation of a disordered condensed phase (see Fig. 9.5). The ortho–para conversion of hydrogen in chabazite was also studied by Stockmeyer (1991, 1992a) by measuring the neutron transmission of the sample after gradual adsorption of hydrogen in steps of one molecule per unit cell at temperatures between 40 and 70 K. Inelastic neutron scattering (INS) has been used by various groups to examine the dynamics of molecular hydrogen adsorbed in zeolites. The sensitivity of the technique to the motions of hydrogen atoms renders it particularly useful as a probe of low-frequency vibrational modes. Furthermore, as on account of their intrinsic spin neutrons are able to induce transitions involving nuclear spin flips that are generally forbidden in optical spectroscopy, detailed information on the intermolecular potential experienced by a hydrogen molecule can be obtained. In a pioneering study, Stockmeyer (1985) found that the H2 motion in a sample of natural chabazite, predominantly containing calcium ions, was ‘gas-like’ at room temperature but ‘solid-like’ below 200 K with a continuous change between the extremes. At low temperature a broad frequency spectrum was derived containing peaks at 48 and 88 cm–1 (6 and 11 meV), thought to

(a)

(b)

(c)

(d)

9.5 Representation of the adsorption of hydrogen in ETS10 as a function of hydrogen concentration: (a) one H2 molecule per Na+ cation; (b) two molecules per Na+ cation; (c) more than two molecules per Na+ cation; and (d) formation of a condensed phase. Adapted from Figure 3 in Ricchiardi et al. (2007).

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

241

be due to hindered rotational and translational modes of H2 adsorbed on several sites. Braid et al. (1987) observed bands at 72 and 155 cm–1 in ZnNaA at 10 K that were interpreted in terms of hindered rotation of the adsorbed H2, about an axis joining the centre of the bond to a zinc cation, in a six-fold potential well with a calculated barrier height of 29 ± 3 kJ mol–1. The dynamics of hydrogen in CoNaA were investigated by Nicol et al. (1988). Analysis of the data suggested that H2 is bound at the cobalt cations end on, and performs 180° rotations with a barrier height of 5.4–6.3 kJ mol–1. A transition at 123 cm–1 was identified as a vibration of the Co–H2 bond. Interestingly, no change in the INS spectrum was observed after the sample was warmed to room temperature and cooled back to 12 K. The hindered rotations and vibrations of molecular hydrogen, deuterium and deuterium hydride adsorbed into partly and fully calcium-exchanged NaA were studied at low temperatures by Eckert et al. (1996). The INS spectra of the rotational transitions of H2, HD and D2 clearly showed that physisorption occurs on several different sites, even at very low loadings at 5 K, a sign that adsorption strengths do not vary substantially. CaNaA and CaA both exhibited strong rotational bands near 48 cm–1, indicating that the principal adsorption site for H2 in the former is in the vicinity of Ca2+ cations with a rotational barrier of about 4.3 kJ/mol. Sites with lower rotational barriers were observed in CaNaA, but the authors were reluctant on the basis of their data to identify observed bands conclusively with molecules adsorbed on single cations. INS evidence for multiple adsorption sites below 60 K was also found by Fu et al. (1999) in NaX, and Ramirez-Cuesta et al. (2000, 2001) obtained evidence for H2 bound both side-on (η2) and end-on (η1) at 20 K in CoAlPO18 (AEI). Mojet et al. (2001) used INS to study the adsorption of H2 in Fe-ZSM5, prepared through sublimation of FeCl3 onto H-ZSM5. Adsorption at 110 K resulted in strongly bound H2 species, ascribed to chemisorbed molecular hydrogen, with clear evidence for at least two different sites. Forster et al. (2003) also found evidence of strongly bound H2 in the zeolitic nickel phosphate VSB5. Tam et al. (2004) examined the reduction of dehydrated AgA by H2 to produce metallic silver nanoparticles. In this case a low-energy transition at 0.3 meV was ascribed to the formation of a strongly bound chemisorption complex comprising molecular hydrogen bound end-on to a terminal silver of the Ag32+ linear complex cation thought to be formed in dehydrated AgA. The results indicated that the first step in the reduction of the silver is the formation of a Ag32+(H2) complex rather than a dissociative adsorption of hydrogen. The complex was found to be stable up to 200 K at which point reduction commenced. Ramirez-Cuesta and Mitchell (2007) reported INS measurements on the adsorption of para-hydrogen and hydrogen deuteride (HD) on coppercontaining ZSM5 as a function of both loading level and temperature. The

© 2008, Woodhead Publishing Limited

242

Solid-state hydrogen storage

H2 molecule was found to be bound as a single species lying parallel to the surface and remained bound to the surface up to 70 K in a condensed liquidor solid-like state. In high-resolution studies on ion-exchanged zeolite X, Ramirez-Cuesta et al. (2007a,b) measured rotational–vibrational spectra of adsorbed para-hydrogen, where the vibrations were those of the H2 molecule against the binding site. The regularity of the observed vibrational sequences indicated that the molecules were residing in parabolic potential wells and the vibrational frequencies were found to be proportional to the polarising power of the cations (Na+ < Ca2+ < Zn2+), providing experimental confirmation that polarisation of the H2 molecule dominates the interaction of hydrogen with the binding sites. Although the total scattering intensity was proportional to the loading level, the intensity of the vibrational bands did not increase at loading levels above 0.3 wt% H2, indicating that the binding sites were saturated. Additional hydrogen introduced remained unbound or only weakly bound. The translational mobility of H2 and HD in NaA has been studied through quasi-elastic neutron scattering (QENS) by Kahn et al. (1989) and by Cohen De Lara and Kahn (1992a,b). It was found that, at low loading, diffusion of both molecules through the structure was fast in comparison with that of larger molecules such as methane and appears not to be strongly influenced by the framework geometry. At 50 K hydrogen was found to be immobile on the timescale of the experiment with the proportion of mobile molecules increasing steadily with temperature, reaching 100% by 150 K. Stockmeyer (1992b) derived a translational diffusion constant of Dt = 2 × 10–3 cm2 s–1 for hydrogen in Na–mordenite, at T = 130 K and a loading level of 35 H2 molecules per unit cell, from QENS data. Martin et al. (1996) measured the diffusion coefficient of HD molecules at low temperatures in the 1D channels of AlPO4-5 and found it to decrease with increasing loading level. Some mobility was detected even at temperatures as low as 10 K and diffusion was found to be significantly slower in the all-silica zeolite silicalite. In contrast Fu et al. (1999) detected no diffusive motion below 20 K in NaX at high loading levels. Good agreement between hydrogen diffusion measurements in zeolites A and X performed by the QENS and pulsed field gradient (PFG) NMR techniques was found by Bär et al. (1999). In general, diffusion rates were found to increase with temperature above 110 K with both techniques yielding an activation energy for diffusion of c. 2.5 kJ mol–1 in NaA. In NaX, hydrogen diffusivity was also found to increase with loading level, accompanied by a decrease in activation energy from c. 4 to less than 1 kJ mol–1, and this was taken as an indication of the importance of the hydrogen–cation interactions in the diffusion process. Diffusivity was found to decrease with decreasing free apertures of the zeolite pore structure for zeolites X, A and chabazite. The unexpectedly low diffusion rates in silicalite were attributed to the ‘capture’

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

243

of a significant fraction of the hydrogen molecules by the pentasil chains of the framework. Vannice and Neikam (1971) used electron spin resonance (ESR) spectroscopy to investigate hydrogen spillover effects in a range of Pd and Pt metal-containing Y zeolites. Olivier et al. (1980) inferred the presence of Ni+(H2) complexes in zeolite X containing both nickel and calcium ions from electron paramagnetic resonance (EPR/ESR), electron–nuclear double resonance (ENDOR) and diffuse reflectance UV-visible measurements. The formation of the complexes was thought to be governed by the reactions Ni2+ + H2 → Ni0 + 2H+ Ni0 + Ni2+ + 2H2 → 2Ni+(H2) The second of these reactions was apparently reversible at temperatures above 200 °C as long as the Ni0 remained finely dispersed and did not aggregate to form large metal particles. As well as this complex, thought to be located in the β-cages, Kermarec et al. (1982) found evidence of a second complex thought to be located in the α-cages. This work was followed up by Michalik et al. (1984) who reported the formation of Ni+(H2) and Ni+(H2)n (n = 2 or 3) complexes in the α-cages. Although the β-cage complex was found to be stable at room temperature, presumably because hydrogen cannot diffuse through the 6-ring windows connecting the β-cage to the α-cage, the α-cage complexes were found to be stable only in the presence of hydrogen and it was postulated that the Ni+ ions were involved in hydrogen-stabilised dimers. The role of calcium ion exchange in the formation of these complexes was considered to be crucial, as the introduction of nickel to NaX results in most of the Ni2+ ions occupying the hexagonal prisms where they are inaccessible to gas molecules. The strong preference of Ca2+ for the 6-ring sites that block access to the hexagonal prisms from the β-cages forces the nickel cations to occupy more accessible sites. Goldfarb and Kevan (1986) reported the formation of analogous Rh(II) complexes, but the exact nature of metal–hydrogen complexes inferred from ESR measurements remains to be substantiated.

9.6

Theoretical studies and modelling

Theoretical studies on the interaction of hydrogen with zeolites fall into three main categories: statistical/diffusion based models; binding site models; and modelling to aid interpretation of specific experiments/experimental techniques. A large number of such studies have been reported; a brief survey is given below but a comprehensive review is not attempted here. To assist in the analysis of early experimental work on the encapsulation of hydrogen at room temperature and above, Fraenkel (1981) derived a

© 2008, Woodhead Publishing Limited

244

Solid-state hydrogen storage

temperature programmed desorption equation for hydrogen encapsulated in zeolite A (LTA), from which activation energies for diffusion were calculated. Yoon and Heo (1992) and Yoon (1993) treated hydrogen molecules encapsulated in zeolite A as a three-dimensional lattice gas and were able to reproduce the pressure dependence of hydrogen encapsulation in Cs2.5Na9.5A and NaA, respectively. Kim et al. (1997) used potential energy functions to calculate activation energies for the ‘decapsulation’ of several small molecules including hydrogen from different cation-exchanged forms of zeolite A, obtaining good agreement with experimental results. Anderson et al. (1999) performed Monte Carlo simulations on the adsorption of hydrogen in NaA, finding that important features including side-on binding to the Na+ cations and the inaccessibility of the β-cages to hydrogen under ambient conditions were replicated. Van den Berg et al. (2004a) assessed the prospects for hydrogen encapsulation in all-silica frameworks using interatomic potential-based calculations. Zeolites containing 6-ring windows (AFG, AST, FRA, LOS and SOD) were chosen as the diameter of a 6-ring at c. 2.9 Å matches closely the Lennard Jones diameter of H2. Diffusion of hydrogen was found to be fastest for the sodalite framework (SOD), for which a series of more detailed studies were subsequently carried out (van den Berg, 2004b,c,d). Molecular dynamics simulations indicated even faster diffusion in Na-sodalite, with the enhancement ascribed mainly to an increase in the flexibility of the sodalite framework that occurs only if the sodium cations are well ordered within the structure (van den Berg, 2004b). The importance of including full framework flexibility if accurate simulations are to be obtained was also highlighted by this work (van den Berg, 2004c,d). Similar studies on multicage clathrasils with the MTN and LOS structures showed a lower diffusion rate than that in sodalite, but still fast enough at 573 K to attract interest as a potential sizeselective membrane or storage medium (van den Berg et al., 2005a, 2006a). Van den Berg et al. (2005b) also carried out a comparison of the use of periodic density functional theory (DFT) and force-field treatments to model the absorption of hydrogen in all-silica sodalite, finding that the use of wellparameterised force fields provided a more reasonable physical description of such systems. A rather promising estimate (4.8 ± 0.5 wt%) for the thermodynamic maximum theoretical storage capacity of sodalite frameworks was obtained from molecular mechanics calculations, more than a factor of ten higher than experimental values reported so far (van den Berg et al., 2005c; Weitkamp et al., 1995). Isotherms derived from grand canonical Monte Carlo (GCMC) calculations, however, suggested that such high capacities could only be obtained under extremely low temperature and/or extremely high-pressure conditions (van den Berg et al., 2006b). GCMC calculations were also carried out for a range of known (AFG, AST, FRA, GIU, LIO, LOS, LTN, MAR, MSO, RUT, SOD, UOZ) and five hypothetical

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

245

clathrasils, yielding a surprisingly large range of adsorption energies (2.84– 6.90 kJ mol–1) for rather similar all-silica frameworks. The maximum storage capacities for a wider range of zeolite structure types (CHA, DDR, FAU, FER, KFI, LTA, LTL, MEL, MFI, MOR, RHO and SOD), were assessed by Vitillo et al. (2005) using classical atomistic simulations. Capacities in the range (1.50–2.86 wt%) were predicted for the all-silica frameworks, with the highest values obtained for FAU and RHO, and a more modest value of 1.92 wt% for SOD. The nature of hydrogen binding sites in zeolites has also been the focus of considerable interest. Zelenkovskii et al. (1984) and Senchenya and Kazansky (1988) used quantum-chemical calculations of the interaction of H2 with the cluster Al(OH)3 to confirm that hydrogen prefers two point coordination at both an aluminium and an oxygen atom when adsorbed at framework Lewis acid sites (see Fig. 9.6). Garrone et al. (1992) and Senchenya and Kazansky (1994) also studied the interaction of molecular hydrogen with bridging hydroxyl groups in zeolites. Larin and Cohen de Lara (1996) developed a theoretical model that enabled the charge on cations in zeolite A to be estimated from the IR frequency shift (∆νHH) on adsorption of hydrogen. This information is invaluable in calculating zeolite–adsorbate interactions for a wide range of adsorbate molecules. More recently, Solans-Monfort et al. (2004) used periodic ab initio calculations to examine the interaction of H 2 with Cu(I)-chabazite (CHA) and were the first to highlight the unusually strong binding (up to 56 kJ mol–1) between H2 and Cu+, in a side-on configuration, at certain zeolite sites (see Fig. 9.7). Adsorption energies as high as 65 and 87 kJ mol–1 were obtained from DFT cluster calculations by Kazansky and Pidko (2005) for hydrogen adsorbed in the same η2 configuration (see Fig. 9.4) in

(a)

(b)

9.6 Calculated structure of hydrogen adsorbed on aluminosilicate rings: (a) 5-ring; and (b) 6-ring. Adapted from Figure 1 in Torres et al. (2007a).

© 2008, Woodhead Publishing Limited

246

Solid-state hydrogen storage

0.742 2.04

Al

2.12

H

0.80

H

H

1.65

O

Si

9.7 Optimised structure (from periodic ab initio calculations) of hydrogen adsorbed in Cu(I)-chabazite (CHA). Adapted from Figure 5 in Solans-Monfort et al. (2004).

Cu(I)-ZSM5 (MFI). Calculated IR frequencies were found to be in good agreement with measured DRIFT spectra. Otero Areán et al. (2006) and Nachtigall et al. (2006) used periodic DFT calculations to calculate adsorption enthalpies of molecular hydrogen on Na-, K- and Li-ferrierite (FER), finding good agreement with experimental values derived from variable temperature infrared measurements. In all cases hydrogen was found to bind in a side-on (η2) configuration and up to two molecules could be adsorbed on cations located at the intersection of two channels, whereas only one was adsorbed on cations in the channel wall sites (see Fig. 9.8). Ricchiardi et al. (2007) also concluded that more than one H2 molecule could be adsorbed at specific cation sites, in the titanosilicate ETS10, on the basis of both IR measurements and molecular mechanic simulations (see Fig. 9.5). Torres et al. (2007a) studied the interaction of hydrogen with alkali metal ions and magnesium in high-silica zeolites using both cluster and periodic models (CHA structure). The results confirm previous findings that H2 interacts preferentially with both the exchangeable cations and the oxygen atoms of the zeolite framework: the adopted geometries permit the molecule to bind side-on to M+ and end-on to a negatively charged framework oxygen, consistent with the quadrupolar shape of the electrostatic potential of the hydrogen molecule (see Fig. 9.6). The latter interaction was shown to have a significant effect on ∆νHH on hydrogen adsorption, in particular. Although cluster models were found to be useful in understanding the general features of hydrogen adsorption on zeolite exchangeable cations, periodic models are required for

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

(a)

247

(b)

9.8 Structure (from periodic DFT calculations) of hydrogen adsorbed in ferrierite (FER): (a) on an Na+ cation in a channel wall site; and (b) on a K+ cation in a channel intersection site. Adapted from Figure 2 in Otero Areán et al. (2006).

an accurate description of adsorption on a specific zeolite type. The latter were also used to investigate the binding of hydrogen in Mg-chabazite (Torres et al., 2007b). A number of theoretical studies have examined the heterolytic dissociative adsorption of H2 in zinc-containing zeolites, where a hydridic hydrogen remains bound to the zinc cation and a proton is transferred to an oxygen atom of the zeolite framework. Barbosa et al. (2001) used DFT cluster calculations to examine adsorption and dissociation on Zn2+ cations at 4-ring and 5-ring sites, as well as (ZnO)4 clusters and intrazeolite [Zn–O–Zn]2+ dimers. Dissociation was found to be more favourable on the oxygen-containing species and was examined in more detail, using periodic DFT calculations based on the chabazite framework as a model, for [Zn–O–Zn]2+, in which case dissociation was found to result in the formation of both [Zn(OH)]+ and [Zn(H)]+ species (Barbosa and Van Santen, 2003). Also using DFT cluster calculations, Shubin et al. (2003) found that dissociation in Zn,H-ZSM5 was preceded by adsorption of molecular hydrogen at a zinc cation with an adsorption energy of 32.2 kJ mol–1. Benco et al. (2005a) used ab initio periodic DFT calculations to study hydrogen adsorption in Zn-mordenite (MOR), finding that for all Zn2+ sites examined, dissociation of H2 is an exothermic process. Similar calculations on mordenite containing a range of different adsorption sites (including Na+, Cu+, Ag+, Zn2+, Cu2+, Ga3+ and Al3+) resulted in the classification of adsorption sites into three categories (Benco et al., 2005b): (i) weak sites unable to dissociate H2 e.g. Brønsted acid sites, weak Lewis acid sites; (ii) Lewis acid sites capable of dissociating H2 → H+ + H– e.g. Zn2+, Al3+ and Ga3+ in certain sites; and (iii) monovalent cations (Cu+ and Ag+) that bind to H2 with very large adsorption

© 2008, Woodhead Publishing Limited

248

Solid-state hydrogen storage

energies, but for which heterolytic dissociation does not occur as it would result in the production of an energetically unfavourable neutral MH molecule. In a detailed study using periodic DFT, Barbosa and Van Santen (2007) examined the adsorption and dissociation of H2 on Zn2+ cations located in 8ring sites in the channels of three distinct zeolite structures with different Si/ Al ratios: chabazite (5), mordenite (23) and ferrierite (35). Two alternative pathways to dissociation – direct dissociation and dissociation via adsorbed H2 – were identified depending on zeolite structure, Si/Al ratio and, in particular, the distribution of aluminium in the framework. Molecular adsorption was favoured in all three zeolites when the 8-ring contained two next nearest neighbour AlO4 tetrahedra. Hydrogen adsorbed on zinc at this type of site is probably responsible for IR spectra that have been observed experimentally (Kazansky, 2003). Direct dissociation was favoured when the two aluminium atoms were further separated in the ring, but only in the chabazite and mordenite structures. Although the strongest interaction with hydrogen was found for Zn2+ located in an 8-ring in ferrierite containing only one aluminium atom with a second distant in the structure, direct dissociation was not favourable in this zeolite. Clearly, the likelihood of dissociation is not related in a straightforward way to the basicity of the zinc cations and the strength of their interaction with hydrogen. In common with the calculations of Torres et al. (2007a) on alkali metal zeolites, the binding strength of H2 adsorbed on Zn2+ is significantly enhanced through an interaction between one of the hydrogen atoms and an oxygen atom of the ring. It is hard not to consider this interaction as the first step towards heterolytic dissociation, and the inability to form a strong hydrogen bond between the adsorbed H2 and a framework oxygen bound to aluminium, as a precursor to the resultant Brønsted acid site, may explain why molecular adsorption does not necessarily lead to direct dissociation even if the interaction with Zn2+ is strong. Scarano et al. (2006) investigated the energy barriers for diffusion of both hydrogen molecules and atoms in HY through ab initio calculations. The barrier for passage of H atoms was found to be c. 10 kJ mol–1 less than for H2 suggesting that dissociation of hydrogen on palladium metal might permit access of hydrogen to the β-cages of HY, under conditions where they were not directly accessible to molecular hydrogen, via a dissociation–diffusion– recombination pathway.

9.7

Other potential applications of zeolites in a hydrogen energy system

The role of zeolite membranes in hydrogen separation and purification has recently been reviewed by Ockwig and Nenoff (2007). They report that more than 14 zeolite structure types have been employed as H2 selective separation

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

249

membranes, including MFI, LTA, MOR and FAU. Membranes comprising zeolite thin films, often on multilayer porous supports, exhibit high diffusion mobilities for H2 leading to high H2 fluxes, with good separation selectivities for mixtures containing gases such as N2, CH4 and, in particular, CO2 that permit the development of separation technologies for gas streams from steam reforming of methane, currently the most widely used method of hydrogen production. Enhanced H2/CO2 and H2/CH4 separation selectivities in the range 33–59 were reported for silylated MFI and CHA membranes by Hong et al. (2005). It is anticipated that zeolite membranes will be able to compete favourably with metal membranes both in terms of economics and performance. Based on H2/He isotherms measured on CaNaA (LTA), Sood et al. (1992) examined the possibility of concentrating trace amounts of hydrogen isotopes (in particular tritium) in a gas stream through pressure swing adsorption (PSA) at 77 K. Simulation results indicated that a single-pass hydrogen isotope recovery of 50–80% should be achievable, and that hydrogen purity of more than 99% could be obtained from a purge stream containing only 0.1% total hydrogen in helium. The practical feasibility of recovering hydrogen in this way was demonstrated by Ruthven and Farooq (1994) through the recovery of hydrogen at greater than 90% purity and 80% total recovery from a feed containing less than 0.2% hydrogen in helium with CaNaA as the adsorbent. In addition to separation and purification, zeolites have also demonstrated promise in a number of different types of hydrogen production processes. Examples include the use of hydrogen zeolites, notably zeolite Y (FAU), as supports for nickel in the catalytic conversion of methane (Inaba et al., 2002); zeolite catalysts for thermochemical hydrogen production (Momirlan and Veziroglu, 2005); and zeolite-based photocatalyst systems for hydrogen production from solar energy (Dutta and Vaidyalingam 2003; Zhang et al., 2004, Kim and Dutta, 2007).

9.8

Prospects for the use of zeolites in a hydrogen energy system

A defining feature of zeolites is a regular intracrystalline network of pores and channels of subnanometre dimensions that, depending on precise composition, results in internal surface areas up to c. 1000 m2 g–1. Though less than other materials such as high surface area (HSA) carbons and MOFs, much of this difference may be ascribed to the heavier atomic constituents of zeolites (most commonly Al, Si and O). Hydrogen storage capacities as high as 2.55 wt% have been reported for NaX at 77 K and 40 bar by Du and Wu (2006), corresponding to more than 30 kg H2 m–3. Despite optimistic predictions that gravimetric capacities of up to 5 wt% might be thermodynamically

© 2008, Woodhead Publishing Limited

250

Solid-state hydrogen storage

possible in sodalites (van den Berg et al., 2005c), and the fact that in many cases substantial increases in hydrogen storage are often possible with moderate increases in pressure (Regli et al., 2005), it is clear that zeolites are unlikely to provide a realistic solution to the problem of mobile storage, where much higher gravimetric capacities are desirable. In an exciting recent development, however, Yang et al. (2006, 2007) reported volumetric capacities as high as 45 kg H2 m–3 and a gravimetric capacity of 6.9 wt% – the highest reported in a porous material – in zeolite-templated carbons at 77 K and 20 bar. Although many workers have been quick to discount the potential of zeolites for hydrogen storage on account of low gravimetric storage capacities, the observed volumetric capacity is equivalent to a carbon-based material exhibiting a gravimetric capacity of c. 3.3 wt% (the difference again being attributable to the chemical composition of the zeolite materials). Significant advantages of zeolites include low cost, low temperature of activation, good thermal and mechanical stability and the fact that they burn neither in air nor in a hydrogen atmosphere. For these reasons they may ultimately find favour as a cheap and safe medium for physisorption-based storage in high volume stationary stores. From a scientific and a technological point of view, zeolites also possess the considerable advantages that: (a) as crystalline materials they are more amenable to structural characterisation than amorphous porous materials; (b) a wide variety of different framework structures and chemical compositions are available; and (c) their well-established and complex intrapore chemistry provides genuine opportunities for enhancing hydrogen storage capabilities. A key challenge for all physisorption-based hydrogen storage materials is to find ways of increasing the energy of interaction of hydrogen with the storage medium in order to provide a useful hydrogen storage capacity at or closer to room temperature, but unfortunately the dispersive forces that govern the interaction of hydrogen with the internal surfaces of many porous materials are intrinsically weak. The presence of strong Lewis acid sites (in particular in the form of a large number of exposed cations) – and the equally important role of the basic oxygens of the framework – providing higher energy adsorption sites marks zeolites out as a notable exception. Adsorption enthalpies close to the optimum of –15.1 kJ mol–1 suggested by Bhatia and Myers (2006), for operation of physisorption-based stores at around room temperature at pressures between 1.5 and 30 bar, have been observed for hydrogen in magnesium-containing zeolites (Otero Areán et al., 2007); even stronger interactions are apparently possible with monovalent and divalent transition metal cations including Ni+, Cu+, Fe2+, Co2+ and Zn2+. The importance of a deep understanding of, and an ability to manipulate, complex zeolite chemistry involving the precise location of zeolite cations and the distribution of aluminium in the framework, in order to produce sites that interact strongly with hydrogen, is an important recurring theme. The

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

251

idea of attempting to enhance the binding strength of hydrogen within the zeolite pores through the introduction of chemically active guest materials, in metal or in hydride form, is one that has been applied with success to related materials but not yet to zeolites (Gutowska et al., 2005; Shatnawi et al., 2007). The result is composite storage materials that contain extremely fine guest materials with sub-nanometre dimensions, and the possibility of enhanced desorption and absorption kinetics. A number of other features broadly unique to the interaction of hydrogen with zeolites merit further investigation. Significant amounts of hydrogen (up to 0.6 wt%) may be trapped or ‘encapsulated’ in zeolite cages at elevated temperatures (Fraenkel and Shabtai, 1977). Reports of a pore blocking effect in zeolite rho (Langmi et al., 2003, 2005) and hysteresis in the absorption/ desorption of hydrogen at 77 K in MOFs (Zhao et al., 2004) suggest that the cage structure of zeolites could be exploited to enhance storage properties at lower temperatures. Corbin et al. (1993) have shown that xenon can be effectively trapped in zeolite rho in this way. The weakness of the physisorption interaction and lack of access of hydrogen into smaller zeolite cages where it may be encapsulated mean that extremely high hydrogen pressures are required for significant hydrogen uptake at room temperature. The observation of dissociative hydrogen chemisorption, whether on zeolite cations such Zn2+ (Kazansky et al., 2003a), intrazeolite cationic clusters such as Ag +n (Baba et al., 2002), or on zeolite-supported nanoparticles of noble metals such as palladium and platinum (Nishimiya et al., 2001), suggest that diffusion of hydrogen into all the cages of a zeolite framework may be possible under much milder conditions (Scarano et al., 2006). The ability to control hydrogen diffusion in zeolites, through dissociation and recombination rather than thermal activation, may prove crucial in the design of improved hydrogen storage and separation materials. Whatever the method of hydrogen production, there is generally a need for purification and separation and it is in this area that zeolites are currently showing exceptional promise (Ockwig and Nenoff, 2007). Lower energy consumption, greater ease of operation and superior separation performance mark out membranes as a highly promising alternative to PSA or fractional distillation, and the estimated costs of zeolite membranes compare favourably to those of metal membranes. The ability of zeolites to trap hydrogen at subambient temperatures also suggests a possible application in low cost boil off traps associated with cryogenic hydrogen storage.

9.9

Acknowledgements

I would like to thank Matthew Turnbull (School of Chemistry, University of Birmingham) for assistance with assembling material, and the following for the kind provision of figures: Dr Petr Nachtigall (Academy of Sciences of

© 2008, Woodhead Publishing Limited

252

Solid-state hydrogen storage

the Czech Republic); Prof. Carlos Otero Areán and Dr Gemma Turnes Palomino (University of the Balearic Islands); Dr Evgeny Pidko (Eindhoven University of Technology); and Prof. Piero Ugliengo and Drs Gabriele Ricchiardi and Jenny Vitillo (University of Turin). The author is part of the AWM ‘Science Cities’ Hydrogen Energy Project and the EPSRC-funded UKSHEC2 Sustainable Hydrogen Energy Consortium.

9.10

References

Anderson C-R, Coker D F, Eckert J and Bug A L R (1999), ‘Computational study of molecular hydrogen in zeolite Na-A. I. Potential energy surfaces and thermodynamic separation factors for ortho and para hydrogen’, J Chem Phys, 111, 7599. Baba T, Akinaka N, Nomura M and Ono Y (1993), ‘Reversible redox behaviour between silver cations and silver metal particles in hydrated Ag-Y zeolites’, J Chem Soc, Faraday Trans, 89, 595. Baba T, Komatsu N, Sawata H, Yamaguchi Y, Takahashi T, Sugisawa H and Ono Y (1999), ‘1H magic angle spinning NMR evidence for dissociative adsorption of hydrogen on Ag+-exchanged A- and Y-zeolites’, Langmuir, 15, 7894. Baba T, Tohjo Y, Takahashi T, Sawada H and Ono Y (2001), ‘Properties of chemisorbed hydrogen species on Ag-A zeolite partially reduced with hydrogen as studied by 1H MAS NMR’, Catal Today, 66, 81. Baba T, Sawada H, Takahashi T and Abe M (2002), ‘Chemisorption study of hydrogen and methane by 1H MAS NMR and conversion of methane in the presence of ethylene on Ag-Y zeolite’, Appl Catal A: General, 231, 55. Baerlocher C, Meier W M and Olson D H (2001), Atlas of Zeolite Framework Types (5th revised ed.), Amsterdam, Elsevier. Bär N-K, Ernst H, Jobic H and Kärger J (1999), ‘Combined quasi-elastic neutron scattering and NMR study of hydrogen diffusion in zeolites’, Magn Reson Chem, 37, S79. Barbosa L A M M and Van Santen R A (2003), ‘Study of the activation of C–H and H– H chemical bonds by the [ZnOZn]2+ oxycation: influence of the zeolite framework geometry’, J Phys Chem B, 107, 14342. Barbosa L A M M and Van Santen R A (2007), ‘The activation of H2 by zeolitic Zn(II) cations,’ J Phys Chem C, 111, 8337. Barbosa L A M M, Zhidomirov G M and Van Santen R A (2001), ‘Theoretical study of the molecular hydrogen adsorption and dissociation on different Zn(II) active sites of zeolites’, Catal Lett, 77, 55. Basmadjian D (1960), ‘Adsorption equilibria of hydrogen, deuterium, and their mixtures’, Can J Chem, 38, 141. Beck K, Pfeifer H and Staudte B (1993), ‘Diffuse reflectance Fourier-transform infrared studies of hydrogen adsorbed on sodium zeolites’, J Chem Soc, Faraday Trans, 89, 3995. Benco L, Bucko T, Hafner J and Toulhoat H (2005a), ‘Periodic DFT calculations of the stability of Al/Si substitutions and extraframework Zn2+ cations in mordenite and reaction pathway for the dissociation of H2 and CH4’, J Phys Chem B, 109, 20361. Benco L, Bucko T, Hafner J and Toulhoat H (2005b), ‘A density functional theory study of H2 on active sites in mordenite’, J Phys Chem B, 109, 22491. Berlier G, Gribov E, Cocina D, Spoto G and Zecchina A (2006), ‘Probing the Brønsted

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

253

and Lewis acidity of Fe-silicalite by FTIR spectroscopy of H2 adsorbed at 20 K: evidences for the formation of Fe3+/H2 and Fe2+/H2 molecular adducts’, J Catal, 238, 243. Bhatia S K and Myers A L (2006), ‘Optimum conditions for adsorptive storage’, Langmuir, 22, 1688. Bordiga S, Garrone E, Lamberti C, Zecchina A, Otero Areán C, Kazansky V B and Kustov L M (1994), ‘Comparative IR-spectroscopic study of low-temperature H2 and CO adsorption on Na zeolites’, J Chem Soc, Faraday Trans, 90, 3367. Borovkov V Yu, Serykh A I and Kazansky V B (2000), ‘Diffuse reflectance IR spectra of ortho-hydrogen and para-hydrogen adsorbed on zeolite BaX at 77 K and their interpretation in terms of the Stark effect’, Kinet Catal., 41, 787. Braid I J, Howard J, Nicol J M and Tomkinson J (1987), ‘Low wavenumber (0–210 cm–1) modes of hydrogen molecules adsorbed on a partially Zn(II)-exchanged A zeolite’, Zeolites, 7, 214. Cohen de Lara E and Kahn R (1992a), ‘Diffusivity of the deuterated hydrogen molecule HD in NaA zeolite by neutron scattering experiment. Comparison with H2 in NaA’, J Chem Phys, 96, 862. Cohen de Lara E and Kahn R (1992b), ‘Diffusivity of hydrogen and methane molecules in A zeolites: neutron scattering measurements and comparison’, Zeolites, 12, 256. Condon E U (1932), ‘Production of infrared spectra with electric fields’, Phys Rev, 41, 759. Corbin D R, Abrams L, Jones G A, Smith M L, Dybowski C R, Hriljac J A and Parise J B (1993), ‘Entrapment and controlled release of xenon in Cd2+-exchanged zeolite Rho’, J Chem Soc, Chem Commun, 1027. Crawford M F and Dagg I R (1953), ‘Infrared absorption induced by static electric fields’, Phys Rev, 91, 1569. Crønstedt A F (1756), ‘Ron och beskrifning om en oberkant barg art, som kallas zeolites’, Kongl Vetenskap Acad Handl Stockholm, 17, 120. Darkrim F, Aoufi A and Malbrunot P (2000), ‘Hydrogen adsorption in the NaA zeolite: a comparison between numerical simulations and experiments’, J Chem Phys, 112, 5991. Du X M and Wu E D (2006), ‘Physisorption of hydrogen in A, X and ZSM-5 types of zeolites at moderately high pressures’, Chin J Chem Phys, 19, 475. Dutta P K and Vaidyalingam A S (2003), ‘Zeolite-supported ruthenium oxide catalysts for photochemical reduction of water to hydrogen’, Micropor Mesopor Mater, 62, 107. Dyer A (1988), An Introduction to Zeolite Molecular Sieves, Chichester, Wiley. Eckert J, Nicol J M, Howard J and Trouw F R (1996), ‘Adsorption of hydrogen in Caexchanged Na-A zeolites probed by inelastic neutron scattering spectroscopy’, J Phys Chem, 100, 10646. Efstathiou A M, Suib S L and Bennett C O (1990), ‘Encapsulation of molecular hydrogen in zeolites at 1 atm’, J Catal, 123, 456. Förster H and Schuldt M (1977), ‘IR active fundamentals of deuterium, nitrogen, and oxygen in zeolitic matrices’, J Chem Phys, 66, 5237. Förster H and Schuldt M (1978), ‘Molecular motion studies of A2 molecules engaged in zeolites’, J Mol Struct, 47, 339. Förster H and Schuldt M (1980), ‘Studies of mobility and surface interaction of A2 molecules in zeolite A by IR spectroscopy’, J Mol Struct, 61, 361. Förster H and Frede W (1984), ‘Induced IR overtone and fundamental bands of isotopic H2 molecules adsorbed in NaCaA zeolites’, Infrared Phys, 24, 151.

© 2008, Woodhead Publishing Limited

254

Solid-state hydrogen storage

Forster P M, Eckert J, Chang J-S, Park S-E, Ferey G and Cheetham A K (2003), ‘Hydrogen adsorption in nanoporous nickel(II) phosphates’, J Am Chem Soc, 125, 1309. Fraenkel D (1981), ‘Zeolitic encapsulation. Part 1. Hydrogen diffusion in A-type zeolite encapsulates’, J Chem Soc, Faraday Trans 1, 77, 2029. Fraenkel D and Shabtai J (1977), ‘Encapsulation of hydrogen in molecular sieve zeolites’, J Am Chem Soc, 99, 7074. Fu H, Trouw F and Sokol P E (1999), ‘A quasi-elastic and inelastic neutron scattering study of H2 in zeolite’, J Low Temp Phys, 116, 149. Garrone E, Kazansky V B, Kustov L M, Sauer J, Senchenya I N and Ugliengo P (1992), ‘Spectroscopic and ab initio study of the interaction of molecular hydrogen with the isolated silica hydroxyls and related systems’, J Phys Chem, 96, 1040. Garrone E and Otero Areán C (2005), ‘Variable temperature infrared spectroscopy: a convenient tool for studying the thermodynamics of weak solid–gas interactions’, Chem Soc Rev, 34, 846. Goldfarb D and Kevan L (1986), ‘Reduction and oxidation effects on rhodium species in Na–X zeolite studied by electron spin echo and electron spin resonance spectroscopies’, J Phys Chem, 90, 2137. Gribov E N, Cocina D, Spoto G, Bordiga S, Ricchiardi G and Zecchina A (2006), ‘Vibrational and thermodynamic properties of Ar, N2, O2, H2 and CO adsorbed and condensed into (H,Na)-Y zeolite cages as studied by variable temperature IR spectroscopy’, Phys Chem Chem Phys, 8, 1186. Gutowska A, Li L, Shin Y, Wang C M, Li X S, Linehan J C, Smith R S, Kay B D, Schmid B, Shaw W, Gutowski M and Autrey T (2005), ‘Nanoscaffold mediates hydrogen release and the reactivity of ammonia borane’, Angew Chem Int Ed, 44, 3578. Harris R, Book D, Anderson P and Edwards P (2004), ‘Hydrogen storage: the grand challenge’, Fuel Cell Rev, 1, 17. Heo N H, Rho B R, Kim D H and Kim J T (1991), ‘Fuel gas encapsulation in ionexchanged zeolite A: 1, Encapsulation characteristics for H2 and CH4’, Hwahak Konghak, 29, 407. Hirscher M and Panella B (2007), ‘Hydrogen storage in metal–organic frameworks’, Scripta Mater, 56, 809. Hong M, Falconer J L and Noble R D (2005), ‘Modification of zeolite membranes for H2 separation by catalytic cracking of methyldiethoxysilane’, Ind Eng Chem Res, 44, 4035. Inaba M, Murata K, Saito M, Takahara I and Mimura N (2002), ‘Hydrogen production by conversion of methane over nickel-supported USY-type zeolite catalysts’, React Kinet Catal Lett, 77, 109. Jhung S H, Yoon J W, Kim H-K and Chang J-S (2005), ‘Low temperature adsorption of hydrogen on nanoporous materials’, Bull Korean Chem Soc, 26, 1075. Kahn R, Cohen de Lara E and Viennet E (1989), ‘Diffusivity of the hydrogen molecule sorbed in NaA zeolite by a neutron scattering experiment’, J Chem Phys, 91, 5097. Kayiran S B and Darkrim F L (2002), ‘Synthesis and ionic exchanges of zeolites for gas adsorption’, Surf Interface Anal, 34, 100. Kazansky V B (1999), ‘Drift spectra of adsorbed dihydrogen as a molecular probe for alkaline metal ions in faujasites’, J Mol Catal A: Chem, 141, 83, Kazansky V B (2003), ‘Localization of bivalent transition metal ions in high-silica zeolites with the very broad range of Si/Al ratios in the framework probed by low-temperature H2 adsorption’, J Catal, 216, 192. Kazansky V B and Pidko E A (2005), ‘A new insight in the unusual adsorption properties of Cu+ cations in Cu-ZSM-5 zeolite’, Catal Today, 110, 281.

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

255

Kazansky V and Serykh A (2004a), ‘A new charge alternating model of localization of bivalent cations in high silica zeolites with distantly placed aluminum atoms in the framework’, Micropor Mesopor Mater, 70, 151. Kazansky V B and Serykh A I (2004b), ‘Unusual localization of zinc cations in MFI zeolites modified by different ways of preparation’, Phys Chem Chem Phys, 6, 3760. Kazansky V B and Serykh A I (2004c), ‘Unusual forms of molecular hydrogen adsorption by Cu+1 ions in the copper-modified ZSM-5 zeolite’, Catal Lett, 98, 77. Kazansky V B, Kustov L M and Borovkov V Yu (1983), ‘Near infrared diffuse reflectance study of high silica containing zeolites’, Zeolites, 3, 77. Kazansky V B, Borovkov V Yu and Karge H G (1997), ‘Diffuse reflectance IR study of molecular hydrogen and deuterium adsorbed at 77 K on NaA zeolite’, J Chem Soc, Faraday Trans, 93, 1843. Kazansky V B, Borovkov V Yu, Serich A and Karge H G (1998a), ‘Low temperature hydrogen adsorption on sodium forms of faujasites: barometric measurements and drift spectra’, Micropor Mesopor Mater, 22, 251. Kazansky V B, Jentoft F C and Karge H G (1998b), ‘First observation of vibration– rotation drift spectra of para- and ortho-hydrogen adsorbed at 77 K on LiX, NaX and CsX zeolites’, J Chem Soc, Faraday Trans, 94, 1347. Kazansky V B, Borovkov V Yu, Serykh A I, van Santen R A and Stobbelaar P J (1999a), ‘On the role of zinc oxide nanometric clusters in preparation of ZnNaY zeolite by ion exchange’, Phys Chem Chem Phys, 1, 2881. Kazansky V B, Borovkov V Y and Karge H G (1999b), ‘Diffuse reflectance IR study of molecular hydrogen and deuterium adsorbed at 77 K on NaA zeolite. Part 2. Overtone, combination and vibration–rotational modes and thermodesorption from NaA zeolite’, Z Phys Chem, 211, 1. Kazansky V B, Borovkov V Yu, Serykh A I, van Santen R A and Anderson B G (2000), ‘Nature of the sites of dissociative adsorption of dihydrogen and light paraffins in ZnHZSM-5 zeolite prepared by incipient wetness impregnation’, Catal Lett, 66, 39. Kazansky V B, Serykh A I, van Santen R A and Anderson B G (2001), ‘On the nature of the sites of dihydrogen molecular and dissociative adsorption in ZnHZSM-5. II. Effects of sulfidation’, Catal Lett, 74, 55. Kazansky V B, Serykh A I and Bell A T (2002), ‘Siting of Co2+ ions in cobalt-modified high-silica zeolites probed by low-temperature molecular hydrogen adsorption’, Catal Lett, 83, 191. Kazansky V B, Serykh A I, Anderson B G and van Santen R A (2003), ‘The sites of molecular and dissociative hydrogen adsorption in high silica zeolites modified with zinc ions. III DRIFT study of H2 adsorption by the zeolites with different zinc content and Si/Al ratios in the framework’, Catal Lett, 88, 211. Kazansky V B, Subbotina I R, van Santen R A and Hensen E J M (2004), ‘DRIFTS study of the chemical state of modifying gallium ions in reduced Ga/ZSM-5 prepared by impregnation. I. Observation of gallium hydrides and application of CO adsorption as a molecular probe for reduced gallium ions’, J Catal, 227, 263. Kermarec M, Olivier D, Richard M, Che M and Bozon-Verduraz F (1982), ‘Electron paramagnetic resonance and infrared studies of the genesis and reactivity toward carbon monoxide of Ni+ ions in a NiCa–X zeolite’, J Phys Chem, 86, 2818. Khodakov A Yu, Kustov L M, Kazansky V B and Williams C (1992), ‘Infrared spectroscopic study of the interaction of cations in zeolites with simple molecular probes. Part 2. Adsorption and polarization of molecular hydrogen on zeolites containing polyvalent cations’, J Chem Soc, Faraday Trans, 88, 3251.

© 2008, Woodhead Publishing Limited

256

Solid-state hydrogen storage

Kim J S, Hwang K J, Hong S B, No K T (1997), ‘Activation energy for the decapsulation of small molecules from A-type zeolites’, Bull Korean Chem Soc, 18, 280. Kim Y and Dutta P K (2007), ‘An integrated zeolite membrane/RuO2 photocatalyst system for hydrogen production from water’, J Phys Chem C, 111, 10575. Kokes R J (1973), ‘Characterization of adsorbed intermediates on zinc oxide by infrared spectroscopy’, Acc Chem Res, 6, 226. Krishnan V V, Suib S L, Corbin D R, Schwarz S, Jones G E (1996), ‘Encapsulation studies of hydrogen on cadmium exchanged zeolite rho at atmospheric pressure’, Catalysis Today, 31, 199. Kubas G J and Ryan R R (1984), ‘Characterization of the first examples of isolable molecular hydrogen complexes, M(CO)3(PR3)(H2) (M = Mo, W; R = Cy, i-Pr). Evidence for a side-on bonded H2 ligand’, J Am Chem Soc, 106, 451. Kustov L M and Kazansky V B (1991), ‘Infrared spectroscopic study of the interaction of cations in zeolites with simple molecular probes. Part 1. Adsorption of molecular hydrogen on alkaline forms of zeolites as a test for localization sites’, J Chem Soc, Faraday Trans, 87, 2675. Langmi H M, Walton A, Al-Mamouri M M, Johnson S R, Book D, Speight J D, Edwards P P, Gameson I, Anderson P A and Harris I R (2003), ‘Hydrogen adsorption in zeolites A, X, Y and RHO’, J Alloys Compd, 356–357, 710. Langmi H M, Book D, Walton A, Johnson S R, Al-Mamouri M M, Speight J D, Edwards P P, Harris I R and Anderson P A (2005), ‘Hydrogen storage in ion-exchanged zeolites’, J Alloys Compd, 404–406, 637. Larin A V and Cohen de Lara E (1996), ‘Estimate of ionicity of zeolite NaA using the frequency shift values of physisorbed molecular hydrogen’, Mol Phys, 88, 1399. Lok B M, Messina C A, Patton R L, Gajek R J, Cannon T A and Flanigen E M (1984), ‘Silicoaluminophosphate molecular sieves: another new class of microporous crystalline inorganic solids’, J Am Chem Soc, 106, 6092. Löwenstein W (1954), ‘The distribution of aluminium in the tetrahedra of silicates and aluminates’, Am Mineral., 39, 92. Makarova M A, Zholobenko V L, Al-Ghefalli K M, Thompson N E, Dewing J and Dwyer J (1994), ‘Brønsted acid sites in zeolites’, J Chem Soc, Faraday Trans, 90, 1047. Martin C, Coulomb J P and Ferrand M (1996), ‘Direct measurement of the translational mobility of deuterium hydride molecules confined in a model microporous material: AlPO4-5 zeolite’, Europhys Lett, 36, 503. McBain JW (1932), The Sorption of Gases and Vapours by Solids, London, Routledge and Sons. Michalik J, Narayana M and Kevan L (1984), ‘Formation of monovalent nickel in NiCa– X zeolite and its interaction with various inorganic and organic adsorbates. Electron spin resonance studies’, J Phys Chem, 88, 5236. Mojet B L, Eckert J, van Santen R A, Albinati A and Lechner R E (2001), ‘Evidence for chemisorbed molecular hydrogen in Fe-ZSM5 from inelastic neutron scattering’, J Am Chem Soc, 123, 8147. Momirlan M and Veziroglu T N (2005), ‘The properties of hydrogen as fuel tomorrow in sustainable energy system for a cleaner planet’, Int J Hydrogen Energy, 30, 795. Nachtigall P, Garrone P, Turnes Palomino G, Rodríguez Delgado M, Nachtigallová D and Otero Areán C (2006), ‘FTIR spectroscopic and computational studies on hydrogen adsorption on the zeolite Li-FER’, Phys Chem Chem Phys, 8, 2286. Nicol J M, Eckert J and Howard J (1988), ‘Dynamics of molecular hydrogen adsorbed in CoNa–A zeolite’, J Phys Chem, 92, 7117.

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

257

Nijkamp M G, Raaymakers J E M J, van Dillen A J and de Jong K P (2001), ‘Hydrogen storage using physisorption – materials demands’, Appl Phys A, 72, 619. Nishimiya N, Kishi T, Mizushima T, Matsumoto A and Tsutsumi K (2001), ‘Hyperstoichiometric hydrogen occlusion by palladium nanoparticles included in NaY zeolite’, J Alloys Compd, 319, 312. Ockwig N W and Nenoff T M (2007), ‘Membranes for hydrogen separation’, Chem Rev, 107, 4078. Olivier D, Richard M, Che M, Bozon-Verduraz F and Clarkson R B (1980), ‘EPR, ENDOR, and UV–visible study of the nickel–hydrogen interactions in a NiCa–X zeolite’, J Phys Chem, 84, 420. Otero Areán C, Manoilova O V, Bonelli B, Rodríguez Delgado M, Turnes Palomino G and Garrone E (2003), ‘Thermodynamics of hydrogen adsorption on the zeolite LiZSM-5’, Chem Phys Lett, 370, 631. Otero Areán C, Rodríguez Delgado M, Turnes Palomino G, Tomás Rubio M, Tsyganenko N M, Tsyganenko A A and Garrone E (2005), ‘Thermodynamic studies on hydrogen adsorption on the zeolites Na-ZSM-5 and K-ZSM-5’, Micropor Mesopor Mater, 80, 247. Otero Areán C, Turnes Palomino G, Garrone E, Nachtigallová D and Nachtigall P (2006), ‘Combined theoretical and FTIR spectroscopic studies on hydrogen adsorption on the zeolites Na-FER and K-FER’, J Phys Chem B, 110, 395. Otero Areán C, Turnes Palomino G and Llop Carayol M R (2007), ‘Variable temperature FT-IR studies on hydrogen adsorption on the zeolite (Mg,Na)-Y’, Appl Surf Sci, 253, 5701. Parise J B (1985), ‘Some gallium phosphate frameworks related to the aluminium phosphate molecular sieves: X-ray structural characterization of {(PriNH3)[Ga4(PO4)4·OH]}·H2O’, J Chem Soc, Chem Commun, 606. Paukshtis E A and Yurchenko E N (1983), Use of IR-spectroscopy in studies of acid– basic properties of heterogeneous catalysts’, Russ Chem Rev, 52, 242. Ramirez-Cuesta A J and Mitchell P C H (2007), ‘Hydrogen adsorption in a copper ZSM5 zeolite: an inelastic neutron scattering study’, Catal Today, 120, 368. Ramirez-Cuesta A J, Mitchell P C H, Parker S F and Barrett P A (2000), ‘Probing the internal structure of a cobalt aluminophosphate catalyst. An inelastic neutron scattering study of sorbed dihydrogen molecules behaving as one- and two-dimensional rotors’, Chem Commun, 1257. Ramirez-Cuesta A J, Mitchell P C H and Parker S F (2001), ‘An inelastic neutron scattering study of the interaction of dihydrogen with the cobalt site of a cobalt aluminophosphate catalyst. Two-dimensional quantum rotation of adsorbed dihydrogen’, J Mol Catal A: Chem, 167, 217. Ramirez-Cuesta A J, Mitchell P C H, Ross D K, Georgiev P A, Anderson P A, Langmi H W and Book D (2007a), ‘Dihydrogen in zeolite CaX – an inelastic neutron scattering study’, J Alloys Compd, 446–447, 393. Ramirez-Cuesta A J, Mitchell P C H, Ross D K, Georgiev P A, Anderson P A, Langmi H W and Book D (2007b), ‘Dihydrogen in cation-substituted zeolites X – an inelastic scattering study’, J Mater Chem, 17, 2533. Regli L, Zecchina A, Vitillo J G, Cocina D, Spoto G, Lamberti C, Lillerud K P, Olsbye U and Bordiga S (2005), ‘Hydrogen storage in chabazite zeolite frameworks’, Phys Chem Chem Phys, 7, 3197. Ricchiardi G, Vitillo J G, Cocina D, Gribov E N and Zecchina A (2007), ‘Direct observation and modelling of ordered hydrogen adsorption and catalyzed ortho–para conversion on ETS-10 titanosilicate material’, Phys Chem Chem Phys, 9, 2753.

© 2008, Woodhead Publishing Limited

258

Solid-state hydrogen storage

Ruthven D M and Farooq S (1994), ‘Concentration of a trace component by pressure swing adsorption’, Chem Eng Sci, 49, 51. Scarano D, Bordiga S, Lamberti C, Ricchiardi G, Bertarione S and Spoto G (2006), ‘Hydrogen adsorption and spillover effects on H–Y and Pd-containing Y zeolites’, Appl Catal A: General, 307, 3. Senchenya I N and Kazansky V B (1988), ‘Nonempirical quantum-chemical study of the adsorption of molecular hydrogen and the mechanism of its heterolytic dissociation on aluminium oxide’, Kinet Catal, 29, 1158. Senchenya I N and Kazansky V B (1994), ‘A quantum-chemical study of the interaction of molecular hydrogen with bridging hydroxyl groups in zeolites and the reaction of H-D exchange between these species’, Kinet Catal, 35, 61. Serykh A I and Kazansky V B (2004), ‘Unusually strong adsorption of molecular hydrogen on Cu+ sites in copper-modified ZSM-5’, Phys Chem Chem Phys, 6, 5250. Serykh A I, Sokolova N A, Borovkov V Y and Kazansky V B (2000), ‘Study of the state of nickel in ion-exchanged NiNaY zeolite by IR spectroscopic using molecular hydrogen adsorbed at low temperatures’, Kinet Catal, 41, 688. Shatnawi M, Paglia G, Dye J L, Cram K D, Lefenfeld M and Billinge S J L (2007), ‘Structures of alkali metals in silica gel nanopores: new materials for chemical reductions and hydrogen production’, J Am Chem Soc, 129, 1386. Shubin A A, Zhidomirov G M, Kazansky V B and Van Santen R A (2003), ‘DFT cluster modeling of molecular and dissociative hydrogen adsorption on Zn2+ ions with distant placing of aluminium in the framework of high-silica zeolites’, Catal Lett, 90, 137. Sigl M, Ernst S, Weitkamp J and Knözinger H (1997), ‘Characterization of the acid properties of [Al]-, [Ga]- and [Fe]-HZSM-5 by low-temperature FTIR spectroscopy of adsorbed dihydrogen and ethylbenzene disproportionation’, Catal Lett, 45, 27. Solans-Monfort X, Branchadell V, Sodupe M, Zicovich-Wilson M C, Gribov E, Spoto G, Busco C and Ugliengo P (2004), ‘Can Cu+-exchanged zeolites store molecular hydrogen? An ab-initio periodic study compared with low-temperature FTIR’, J Phys Chem B, 108, 8278. Sood S K, Fong C, Kalyanam K M, Busigin A, Kvelton O K and Ruthven D M (1992), ‘A new pressure swing adsorption (PSA) process for recovery of tritium from the ITER solid ceramic breeder helium purge gas’, Fusion Technol, 21, 299. Spoto G, Gribov E, Bordiga S, Lamberti C, Ricchiardi G, Scarano D and Zecchina A (2004), ‘Cu+ (H2) and Na+ (H2) adducts in exchanged ZSM-5 zeolites’, Chem Commun, 2768. Stéphanie-Victoire F and Cohen de Lara E (1998), ‘Adsorption and coadsorption of molecular hydrogen isotopes in zeolites. II. Infrared analyses of H2, HD, and D2 in NaA, J Chem Phys, 109, 6469. Stéphanie-Victoire F, Goulay A-M and Cohen de Lara E (1998), ‘Adsorption and coadsorption of molecular hydrogen isotopes in zeolites. 1. Isotherms of H2, HD, and D2 in NaA by thermomicrogravimetry’, Langmuir, 14, 7255. Stockmeyer R (1985), ‘Continuous fluid–solid transition of hydrogen in chabazite’, Zeolites, 5, 393. Stockmeyer R (1991), ‘In situ observation of the ortho-para hydrogen conversion in chabazite by neutron scattering’, Ber Buns Ges Phys Chem, 95, 458. Stockmeyer R (1992a), ‘In situ observation of the o–p hydrogen conversion rate by neutrons’, Physica B, 180, 508. Stockmeyer R (1992b), ‘Diffusive motion of hydrogen in mordenite studied by quasielectric neutron scattering’, Zeolites, 12, 251.

© 2008, Woodhead Publishing Limited

Storage of hydrogen in zeolites

259

Tagaki H, Hatori H, Soneda Y, Yoshizawa N and Yamada Y (2004), ‘Adsorptive hydrogen storage in carbon and porous materials’, Mater Sci Eng B, 108, 143. Tam C N, Trouw F R and Iton L E (2004), ‘Inelastic neutron scattering study of the activation of molecular hydrogen in silver-exchanged A zeolite: first step in the reduction to metallic silver at low temperature’, J Phys Chem A, 108, 4737. Torres F J, Vitillo J G, Civalleri B, Ricchiardi G and Zecchina A (2007a), ‘Interaction of H2 with alkali-metal-exchanged zeolites: a quantum mechanical study’, J Phys Chem C, 111, 2505. Torres F J, Civalleri B, Terentyev A, Ugliengo P and Pisani C (2007b), ‘Theoretical study of molecular hydrogen adsorption in Mg-exchanged chabazite’, J Phys Chem C, 111, 1871. Turnes Palomino G, Llop Carayol M R and Otero Areán (2006), ‘Hydrogen adsorption on magnesium-exchanged zeolites’, J Mater Chem, 16, 2884. van den Berg A W C, Bromley S T, Jansen J C and Maschmeyer T (2004a), ‘Hydrogen storage and diffusion in 6-ring clathrasils’, Stud Surf Sci Catal, 154, 1838. van den Berg A W C, Bromley S T, Flikkema E and Jansen J C (2004b), ‘Effect of cation distribution on self-diffusion of molecular hydrogen in Na3Al3Si3O12 sodalite: a molecular dynamics study’, J Chem Phys, 121, 10209. van den Berg A W C, Bromley S T, Flikkema E, Wojdel J, Maschmeyer Th and Jansen J C (2004c), ‘Molecular-dynamics analysis of the diffusion of molecular hydrogen in all-silica sodalite’, J Chem Phys, 120, 10285. van den Berg A W C, Bromley S T, Ramsahye N and Maschmeyer T (2004d), ‘Diffusion of molecular hydrogen through porous materials: the importance of framework flexibility’, J Phys Chem B, 108, 5088. van den Berg A W C, Flikkema E, Jansen J C and Bromley S T (2005a), ‘Self-diffusion of molecular hydrogen in clathrasils compared: dodecasil 3C versus sodalite’, J Chem Phys, 122, 204710. van den Berg A W C, Bromley S T, Wojdel J C and Jansen J C (2005b), ‘Molecular hydrogen confined within nanoporous framework materials: comparison of density functional and classical force-field descriptions’, Phys Rev B, 72, 155428. van den Berg A W C, Bromley S T and Jansen J C (2005c), ‘Thermodynamic limits on hydrogen storage in sodalite framework materials: a molecular mechanics investigation’, Micropor Mesopor Mater, 78, 63. van den Berg A W C, Flikkema E, Lems S, Bromley S T and Jansen J C (2006a), ‘Molecular dynamics-based approach to study the anisotropic self-diffusion of molecules in porous materials with multiple cage types: application to H2 in losod’, J Phys Chem B, 110, 501. van den Berg A W C, Bromley S T, Wojdel J C and Jansen J C (2006b), ‘Adsorption isotherms of H2 in microporous materials with the SOD structure: a grand canonical Monte Carlo study’, Micropor Mesopor Mater, 87, 235. Vannice M A and Neikam W C (1971), ‘On the chemical nature of hydrogen spillover’, J Catal, 20, 260. Vitillo J G, Ricchiardi G, Spoto G and Zecchina A (2005), ‘Theoretical maximal storage of hydrogen in zeolitic frameworks’, Phys Chem Chem Phys, 7, 3948. Weitkamp J, Fritz M and Ernst S (1995), ‘Zeolites as media for hydrogen storage’, Int J Hydrogen Energy, 20, 967. Wilson S T, Lok B M, Messina C A, Cannan T R and Flanigen E M (1982), ‘Aluminophosphate molecular sieves: a new class of microporous crystalline inorganic solids’, J Am Chem Soc, 104, 1146.

© 2008, Woodhead Publishing Limited

260

Solid-state hydrogen storage

Yang Z, Xia Y, Sun X and Mokaya R (2006), ‘Preparation and hydrogen storage properties of zeolite-templated carbon materials nanocast via chemical vapor deposition: effect of the zeolite template and nitrogen doping’, J Phys Chem B, 110, 18424. Yang Z, Xia Y and Mokaya R (2007), ‘Enhanced hydrogen storage capacity of high surface area zeolite-like carbon materials’, J Am Chem Soc, 129, 1673. Yoon J-H (1993), ‘Pressure-dependent hydrogen encapsulation in Na12-zeolite A’, J Phys Chem, 97, 6066. Yoon J-H and Heo N H (1992), ‘A study on hydrogen encapsulation in Cs2.5–zeolite A’, J Phys Chem, 96, 4997. Yuvaraj S, Chang T-H and Yeh C-T (2003), ‘Low-temperature-programmed replacement of nitrogen by hydrogen in pores of mordenite’, J Phys Chem B, 107, 4971. Zecchina A, Otero Areán C, Turnes Palomino G, Geobaldo F, Lamberti C, Spoto G and Bordiga S (1999), ‘The vibrational spectroscopy of H2, N2, CO and NO adsorbed on the titanosilicate molecular sieve ETS-10’, Phys Chem Chem Phys, 1, 1649. Zecchina A, Bordiga S, Vitillo J G, Ricchiardi G, Lamberti C, Spoto G, Bjørgen M and Lillerud K P (2005a), ‘Liquid hydrogen in protonic chabazite’, J Am Chem Soc, 127, 6361. Zecchina A, Spoto G and Bordiga S (2005b), ‘Probing the acid sites in confined spaces of microporous materials by vibrational spectroscopy’, Phys Chem Chem Phys, 7, 1627. Zelenkovskii V M, Zhidomirov G M and Kazanskii V B (1984), ‘Quantum-chemical study of molecular hydrogen on a Lewis acidic center’, Zh Fiz Khim, 58, 1788. Zhang X H, Li W Z and Xu H Y (2004), ‘Application of zeolites in photocatalysis’, Prog Chem, 16, 728. Zhao X B, Xiao B, Fletcher A J, Thomas K M, Bradshaw D and Rosseinsky M J (2004), ‘Hysteretic adsorption and desorption of hydrogen by nanoporous metal–organic frameworks’, Science, 306, 1021.

© 2008, Woodhead Publishing Limited

10 Carbon nanostructures for hydrogen storage P. B É N A R D and R. C H A H I N E, Institut de recherche sur l’hydrogène, Canada

10.1

Introduction

Physical adsorption on activated carbons and zeolites is widely used for the separation and purification of gas mixtures.1,2 It also offers the possibility of lowering the storage pressure of compressed gases such as natural gas and hydrogen1,2. For natural gas, physisorption on activated carbon offers a substantial improvement in low-pressure storage density at room temperature. However for hydrogen, the temperature must be lowered to 77 K to achieve comparable performance. At 77 K and 17 bar, 40% of the density of liquid hydrogen can be stored in a given volume filled with activated carbon, or about five times the density of gaseous hydrogen in the same volume under the same conditions. Despite these performance achievements, the storage densities that have been achieved using activated carbons remain below the ambitious objectives3 proposed by the US Department of Energy, which fix a system target of 2 kW h/kg (6 wt%), 1.5 kW h/l, and US$4/kW h for 2010 and of 3 kW h/kg (9 wt%), 2.7 kW h/l, and US$2/kW h for 2015, based on achieving a driving range of 500 km using a hydrogen powered fuel cell engine. Following Dillon et al.’s seminal paper on physisorption of hydrogen on single wall nanotubes,4 several novel carbon-based microporous materials, carbonate materials and even metal–organic frameworks5 have been proposed as potential storage media for hydrogen. These nanomaterials could in principle be optimized for hydrogen storage through various physical and chemical treatments, including metal insertion. This chapter reviews the hydrogen sorption properties of carbon nanostructures and outlines the issues concerning their use as storage media. The first section discusses the adsorption of hydrogen in the context of solidstate storage of hydrogen in materials. This is followed by a brief overview of the concepts relevant to the adsorption of supercritical gases in nanoporous adsorbents and of the measurement and characterization of their sorption properties. The adsorption of hydrogen on activated carbons, carbon nanotubes, carbon nanofibers and carbides is then discussed from the point of view of 261 © 2008, Woodhead Publishing Limited

262

Solid-state hydrogen storage

experiments and simulations. The use of nanocarbons as physisorbents for hydrogen is also compared with metal–organic frameworks. The final section discusses outlooks for carbon-based sorbents as storage media for hydrogen, notably through the use of the spillover mechanism and metal insertion.

10.2

Storage of hydrogen in solids

Hydrogen is currently stored in vehicles as a gas in high-pressure cylinders or as a liquid at 20 K in cryogenic reservoirs. The maximum storage densities that have been achieved so far using these storage technologies are 1.2 kW h/l (volumetric) and 1.7 kW h/kg (gravimetric) for liquid hydrogen (which could be pushed to 18% weight using light materials6). The system density of compression storage units falls short of the 2010 targets3 and it is difficult to foresee further improvement that could achieve the 2015 targets3. Even at 1000 bar, the energy density of compressed hydrogen at room temperature represents less than one-third of the energy density currently available in gasoline (6 kW h/l). As to liquid hydrogen storage, although it could come close and even surpass the 2010 targets, the volumetric density of a liquid hydrogen storage system is physically limited by the density of liquid hydrogen (to 2.3 kW h/l at 1 atmosphere), below the 2015 DOE objective. Meeting the 2015 targets will require going beyond conventional hydrogen storage. The storage density can be increased by exploiting the interatomic and intermolecular forces between hydrogen and other atoms. A given volume of water, for instance, contains more hydrogen than the same volume of liquid hydrogen. Hydrogen could be stored by chemical binding to other elements in chemical hydrides (such as organic liquids, for instance), inside a solid metallic substrate by forming a metal hydride, or, depending on the nature and the strength of the molecular force fields involved, by chemisorption or physisorption on large surface area materials such as activated carbons or metal–organic frameworks. The characteristic binding energy of hydrogen is a key factor in the determination of the thermodynamics of materials-based reservoir, most notably as it pertains to charging and discharging hydrogen, as well as the boil-off rate from cryogenic systems. Hydrogen binding in chemical hydride binding energies is typically greater than 100 kJ/mol, whereas it ranges from 50 to 100 kJ/mol in metal hydrides, and lies below 10 kJ/mol when physisorbed on a high surface area solid. The binding energy will determine the heat load required to release hydrogen from storage. Materials with hydrogen binding energies larger than 50 kJ/mol require more power to release 5 kg of hydrogen than boiling the equivalent mass of water7. However, materials with binding energies less than 10 kJ/mol require cryogenic operation to achieve acceptable storage densities. The range 10–50 kJ/mol is being promoted as best suited for storage applications. Storage by sorption processes on the surface of materials has been the

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

263

topic of intense interest for hydrogen storage applications because the relatively low binding energy allows for an easier release of hydrogen from confinement, and the weight of the materials is relatively low compared with metal hydrides. Activated carbons, in particular, have been proposed as simple and easily available adsorbents for hydrogen for cryogenic storage at liquid nitrogen temperature. The possibility of larger binding energies, which would allow operation at higher temperatures, has fueled interest in other carbon nanostructures such as nanotubes, nanofibers, and carbides, whose hydrogen storage properties may be optimized through various physical and chemical processes. In addition to the characteristic binding energy of the hydrogen molecule with the adsorbent EA, the surface area available for adsorption processes SA (specific surface area) and the bulk density d of the adsorbent are critical parameters for an adsorption-based storage strategy. The bulk density of the storing materials determines the weight and volume of the storage unit. Both parameters can be combined into the average surface available per unit volume of the adsorbent, which should be maximized to optimize the storage density.

10.3

Carbon nanostructures and hydrogen storage

The wide variety of carbon structures arise from the hybridization of the partially filled sp2 orbitals of the carbon atom. Carbon atoms can bind covalently with three, five and even seven neighboring carbon atoms, leading to the possible formation of pentagonal and heptagonal faces. These faces allow the existence of complex solids and macromolecules, such as activated carbon, graphite, diamond, carbon nanofibers and fullerenes such as C60 buckyballs, single-wall nanotubes (SWNT) and multiwall nanotubes (MWNT) (Fig. 10.1). This versatility, in addition to the low atomic weight of carbon and its propensity to bind chemically or physically other atoms or molecules makes it eminently suitable as the major structural component of storage media for hydrogen. In two dimensions, carbon atoms tend to form graphene planes, in which carbon atoms (separated by a distance of 1.41 Å) are arranged hexagonally. Graphite is an ordered set of graphene planes held together by van der Waals forces. Activated carbons, on the other hand, are sets of randomly arranged graphitic planes of various sizes, which usually form a highly porous threedimensional disordered structure, characterized by a pore size distribution. Pores are usually classified as micropores (or nanopores), whose size is of the order of a few molecular diameters (width < 2 nm), mesopores (width 2– 50 nm) and macropores (width > 50 nm) (Fig. 10.2). Microporous activated carbons are the only activated carbons of interest to gas storage applications. SWNTs are tubular carbon nanostructures (Fig. 10.3). They have a diameter in the order of 1 nm and can reach hundreds of micrometers in length. They tend to self-organize in bundles of hundreds of units,8 usually in hexagonal

© 2008, Woodhead Publishing Limited

264

Solid-state hydrogen storage

10.1 Carbon nanostructures. From top to bottom: graphene layers (slit pores), C60 fullerene and single wall nanotube.

10.2 SEM image of a sample of the activated carbon IRH-4DD with a specific surface area of 2000 m2/g. The macropores (largest pores) and the mesopores (smaller pores) can be seen. Observing the micropores would require better resolution.

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

265

10.3 SEM (top) and TEM (bottom) images of a single wall nanotube sample.

arrangement. An SWNT can be obtained from a graphene layer by performing a rectangular cut such that the edges and the vertices along the edges on each side of the cut can be put into correspondence. The SWNT is formed by wrapping the cut along its axis and by capping the resulting cylinder with two hemispheres of a buckyball of appropriate diameter. Depending on the cut, the resulting SWNT will be classified as armchair, zigzag or chiral. Multiwall nanostructures are macromolecular arrangements of two or more SWNTs inserted one into the other. The adsorption process on bundled single wall nanotubes is governed by four adsorption sites: the internal sites within the individual SWNTs, the interstices between the nanotubes within the bundle, the grooves between pairs of SWNTs at their surface, and the external surface sites (Fig. 10.4). Carbon nanofibers are filamentous cylindrical or conical structures formed of various arrangements of stacked graphene sheets (as cones or cups), with a diameter ranging from one to several hundred nanometers, and lengths ranging from 1 µm to several millimeters.9 Nanostructured carbon foams

© 2008, Woodhead Publishing Limited

266

Solid-state hydrogen storage y 15

Peripheral site

Tube site

10

5

0

–5

–10

–15 –15

Interstitial site –10

–5

x 0

5

10

15

10.4 Adsorption sites in bundles of single wall nanotubes. From E. Mélançon and P. Bénard, Langmuir 20, 7852–7859 (2004).

based on SWNT networks have also been put forward, offering high specific surfaces and better binding than SWNT bundles.10 Other carbon-based structures have been proposed for hydrogen adsorption. High surface area carbide-derived carbons can be produced by the selective chemical etching of carbides, leaving behind an amorphous carbon with high specific surface area (2000 m2/g) and a large open pore volume (80%). This approach offers the possibility of tuning the pore structure to specific applications by selection of the starting carbide and using various control processes. These structures are basically a disordered arrangement of bent graphene layers with a narrow size distribution.11 Carbon aerogels are tridimensional networks of interconnected nanometer-sized carbon particles which form a porous structure that can be synthesized inexpensively. They exhibit a large specific surface area and their bulk properties such as density, pore size, pore volume, and specific surface can be controlled. Metal nanoparticles can be incorporated into the aerogels during synthesis, and they could be used as scaffolds for metal hydrides. Metal–organic frameworks12,13 are solids related to carbon nanostructures through the fact that they consist of transition metals bridged by carbon ligands. Their large specific surfaces (1000–6000 m2/g) make them prime candidates for physisorption-based storage applications. In addition, they may be tailored to specific storage applications by changing ligands, transition metals or by doping.

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

10.4

267

Supercritical adsorption in nanoporous materials

Adsorption can be defined as a process that results in the increase of the density of a gas or a solute (the adsorbate) in the vicinity of the surface of a substrate (the adsorbent) due to molecular interactions between the adsorbate and the adsorbent. The density of the adsorbate far from the surface is defined as its bulk value, and corresponds to the density of the free adsorbate under the same thermodynamic conditions. For gases, adsorption can be divided into two regimes: supercritical and subcritical adsorption. The former occurs above the critical temperature of the adsorbent and as such exhibits no gas–liquid phase transition. The latter is observed below the critical temperature of the adsorbent, and its isotherm is characterized by a steep increase of the adsorbed density close to the saturation pressure, associated with the formation of a liquid layer on the surface of the adsorbate. In practice, supercritical adsorption is the only regime usually considered for hydrogen storage due to the low critical temperature of hydrogen (33.18 K) and the objective of operating an adsorption-based storage system as close to room temperature as possible. The adsorption process is characterized by the adsorbed gas density, na, which is the amount of adsorbate molecules within the adsorption volume Va, defined as the volume where the density of the adsorbate is larger than the density of the bulk gas (ρg) under the same thermodynamic conditions of pressure and temperature. The adsorbed gas density is usually measured per unit of mass of the adsorbent. The two quantities na and Va are not readily accessible experimentally.14 The pore volume (Vpore) accounts for the entire porous structure of the adsorbent, which we define here as including intergranular spaces and its external surface. The pore volume can be divided into two contributions: the adsorption volume Va and the pore dead volume of the adsorbent Vdead, in which the adsorbate molecules are negligibly subjected to the molecular force field of the adsorbent, and which consist of large pores and interstices. For an ideal microporous adsorbent, Vpore ≅ Va, because the pore structure is fully characterized by molecular length scales. The pore volume is usually measured using helium, based on the assumption that He molecules interact very weakly with most adsorbents. Within this approximation, Va for helium is zero and Vpore ≅ Vdead. The excess adsorbed density (nex) is defined as the difference between the total amount of adsorbate molecules (Na) found within the volume VVoid of a measurement cell containing the adsorbent at a given temperature and pressure, minus the number of molecules Ng that would have been found in the void volume in the absence of adsorbate–adsorbent interactions. The volume VVoid is the sum of the pore volume of the adsorbent Vpore and the volume of the section of the cell which contains no adsorbent (VEmpty). The number Ng is

© 2008, Woodhead Publishing Limited

268

Solid-state hydrogen storage

obtained from the product of the density of the bulk gas (obtained from the equation of state of the adsorbate) at the same temperature and pressure by the void volume. The excess number of adsorbate molecules is thus given by: Nex = Na – ρgVVoid

[10.1]

The excess adsorbed density is obtained by dividing this expression by the mass M of the adsorbent in the cell:

n ex =

N ex M

[10.2]

Note that Na/M is not equal to the absolute adsorbed density na since it also includes molecules in the volume VEmpty. In terms of the local density, Eq. (10.2) can be written:15 n ex = 1/ M



( ρ ( r ) – ρg ) dr

[10.3]

VVoid

The number of adsorbate molecules inside VEmpty is ρgVEmpty and the void volume is VVoid = VEmpty + Vpore. Eq. (10.1) can be expressed in terms of the absolute adsorption by the following equation: nex = na – ρgvpore

[10.4]

without any reference to the empty volume of the measurement cell, where na =

N a – ρg VEmpty M

[10.5]

is defined as the absolute adsorbed density. This quantity is a monotonically increasing function of pressure. It vanishes rapidly with increasing temperature on a scale determined by an energy representative of the adsorbate–adsorbent interactions. The quantity vpore = Vpore/M is the pore volume per unit mass of the adsorbent. The excess density, defined by Eq. (10.4), is the quantity obtained through the volumetric method of adsorption measurements. It is comparable to the absolute density na when the bulk gas density is much smaller than nex (at high enough temperature and low pressure). The excess density differs appreciably from na at high pressure and low temperature.15,16 At a given temperature above the critical temperature of the adsorbate, the absolute adsorbed density will increase at a slower pace than the bulk gas as a function of pressure when the pores become sufficiently filled with adsorbate molecules. The excess density isotherms, which are a function of the difference of these two quantities, thus exhibit a maximum as a function of pressure. The quantity of hydrogen (in moles) stored in an adsorption-based storage unit of total Volume VSys can be estimated using the following expression:

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

Nstored = Na + ρgVEmpty = (nex + ρgvpore)M + ρgVEmpty

269

[10.6]

The stored volumetric density is thus ( n ex + ρg v pore ) M ρg VEmpty N stored = + VSys VSys VSys

[10.7]

assuming a system volume independent of the amount adsorbed. The gravimetric density stored is defined by x stored =

N stored M Sys + N stored m H 2

[10.8]

where MSys is the total mass of the storage system and m H 2 is the molar mass of molecular hydrogen. The molar densities (10.7) and (10.8) can be expressed as mass stored per system volume or per system mass by multiplying them by the molar mass of the adsorbate. The excess density multiplied by the molar mass of the adsorbate is often used as an estimate of the gravimetric density of the system instead of (10.7), because Vs and MSys are only known after designing the system. This assumes MSys = M and neglects the contribution of the bulk phase in the pore volume to the total amount of gas in the storage unit, as well as the contribution of the stored hydrogen to the total mass of the filled storage system. The differential enthalpy of adsorption, also called isosteric heat of adsorption, is often used to characterize the energetics of adsorption phenomena. It depends upon the properties of the adsorbent and the adsorbate17 and reflects the magnitude of gas–solid interactions. It can be measured by calorimetry or calculated from adsorption isotherms at constant coverage using the expression:15,18

 ∂ ln( P )  ∆h = – R   ∂(1/ T )  na

[10.9]

The differential enthalpy of adsorption is thus the slope of the adsorption isotheres, which are curves of the natural log of pressure as a function of inverse temperature at constant adsorbed density. Monson and Myers have derived a formulation of the thermodynamics of adsorption in terms of excess variables, which yields a similar expression for the excess isosteric heat of adsorption evaluated at constant excess density. The excess differential enthalpy, however, exhibits a singularity close to the maximum of the excess isotherms because the excess density becomes a multivaluate function of pressure.15

10.5

Theory

An efficient model for the excess adsorption isotherms of microporous adsorbents should rely on a minimum number of parameters with a clear

© 2008, Woodhead Publishing Limited

270

Solid-state hydrogen storage

physical interpretation. Only isotherm models based on physisorption, caused by van der Waals interactions between the adsorbate molecules and the adsorbents, will be considered in this section.

10.5.1 Virial expansion In the low-pressure limit the excess density can be expressed as a virial series expansion:19 nex = BAS(p/kT) + CAAS(p/kT)2 + DAAS(p/kT)3 + …

[10.10]

The excess adsorbed density at constant temperature is thus expected to vary linearly with pressure when the latter is close to zero (Henry’s law). The first non-zero coefficient (BAS), also called the second virial coefficient, is fully r determined by the interaction V ( r ) between a single molecule and the surface:

BAS = 1 M

r

 – VkT( r ) –  e



 r 1  dr 

[10.11]

where M is the mass of the adsorbent. Knowledge of the adsorbent–adsorbate interactions can therefore allow a direct estimate of the energy of adsorption. The virial coefficient BAS can be obtained experimentally by finding the intercept of a plot of n/p as a function of p. It can be expressed as a universal scaling function in terms of the rescaled inverse temperature (Fig. 10.5). 12 10

In (BAS/α z0)

8 6 4 2 0 –2 0

2

4

6

8 x = εs /kT

10

12

14

10.5 Normalized virial coefficient BAS as a function of reduced inverse temperature. The white circles refer to hydrogen on AX-21. The black circles refer to methane adsorption data on the activated carbon CNS-201. εs and z0 are the energy and characteristic distance Lennard-Jones parameters (respectively) for a planar potential, and α is a characteristic molecular area. From P. Bénard and R. Chahine, Hydrogen Energy Progress XII, p 1121 (1998).

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

271

r The interaction potential V ( r ) between an adsorbate molecule and the adsorbent atoms can be written as: r r r V ( r ) = Σ i U ( r – Ri ) r r where r and Ri are the positions of the adsorbate molecule and the ith adsorbent atom. r The intermolecular potential U ( r ) is often modeled using the Lennard– Jones potential:  σ 6 r σ 12  V ( r ) = 4ε   –    [10.12] r r    r where σ is the distance where V ( r ) = 0 and ε is the well depth of the potential. Modeling of nanoporous activated carbons such as AX-21™ often assumes that the pores are formed from two parallel graphite planes, the so-called slit-pore model.20 The molecular force field of the adsorbent is obtained from the superposition of two Lennard–Jones planar potentials describing two parallel planes separated by a distance d which should be representative of the pore size distribution. Assuming an average uncorrugated planar interaction potential between the adsorbent planes and an adsorbate molecule, the second virial coefficient is then given by:

BAS =2 SA σ



d*/2

0

–1 [ V ( y + d*/2)+V ( y – d*/2)]  e kT – 

1 dy 

[10.13]

where d* = d/σ, y is a dimensonless integration variable and SA is the specific surface, which is the area of the adsorbent accessible to adsorbate molecules per unit mass of the adsorbent. The planar potential represents an averaged potential from a continuous (non-corrugated) distribution of carbon atoms on a surface. It could be described, for instance, by a 10–4 Lennard–Jones planar potential:  2 1 10 1 4 V ( y) = 2ε s    –    y  y  5  

[10.14]

where εs = πθεσ2, θ is the surface density of a graphene plane (0.38 atoms per Å2) and y is a dimensionless variable representing the vertical distance to the plane. If the values ε = 30.5 K and σ = 3.19 Å are assumed, and the interlayer distance and the specific surface are treated as adjustable parameters, the values d = 8.1 Å for the interlayer distance and 2900 m2/Å for the specific surface area for hydrogen sorption on the activated carbon AX-21 are obtained. Although the estimated specific surface is close to the Brunauer–Emmett– Teller (BET) estimate (2800), density functional theory (DFT) analysis of AX-21 shows a pore size distribution peaked around 12.5 Å, which is far

© 2008, Woodhead Publishing Limited

272

Solid-state hydrogen storage

from the value of the interlayer distance obtained. The second virial coefficient of SWNTs has also been studied using a cylindrical potential in the continuum approximation.21 10 4   1  1  V *( r *, R*) = 3  21  M11 ( r */ R*) –  M 5 ( r */ R*)   R*    32  R* 

[10.15] where r* = r/σ, R* = R/σ and V*(r*, R*) = V(r*σ, R*σ)/εs are respectively the reduced distance from the axis, the reduced radius of the SWNT and the reduced potential, with Mn ( x ) =



π

0

dφ (1 + x – 2 x cos φ ) n /2 2

[10.16]

10.5.2 Langmuir model The Langmuir isotherm is a local model of the absolute adsorption based on monolayer filling of non-interacting molecules. As such, it represents a model of supercritical adsorption on surfaces. It predicts a monotonically increasing function of pressure, which saturates asymptotically as a function of pressure to a value nm: na = n m

B  CP , C = 10 exp ( A / R ) exp  – RT 1 + CP   p

[10.17]

where A and B are constants and P is the pressure.15 It can be put into the following form:

p p = 1 + n Cn m nm

[10.18]

from which the parameters nm and C can be determined from experimental isotherms. The saturation density nm and the coefficient C can be obtained from a simple linear fit of the experimental data to (10.18). The excess adsorption version of this isotherm requires the knowledge of the pore volume: n ex =

n m ( CP ) – ρg Vpore 1 + CP

[10.19]

A comparison of experimental adsorption data of methane on silicate at 25 °C fits to both absolute and total Langmuir isotherms shows that both can describe the experimental results at low pressure for a comparable pressure range. However both fail at high pressure15 whereas molecular simulations

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

273

correctly predict the experimental data. The validity of the Langmuir model is limited to low-pressure, high-temperature adsorption of gases in the supercritical regime. As such, it is of limited usefulness to the description of adsorption over the wide temperature and pressure ranges relevant to storage applications.

10.5.3 Self-consistent Ono–Kondo approach Aranovitch et al. developed self-consistent equations describing the adsorption process on carbon slit-pores in the supercritical regime based on a statistical physics lattice approach.22,23 As is the case for the Langmuir isotherm, this approach yields Henry’s law in the low-pressure limit. This approach does not, however, describe the porous structure of the carbon. These effects are assumed small and are incorporated into the fitting parameters of the model. The model relies on the calculation of the density of adsorbate molecules in N discrete layers sandwiched by two graphene planes representing a slit pore. Adjacent molecules interact with an Ising site potential E. Molecules adjacent to the two graphene planes are subjected to a uniform potential EA. The coverage of the ith layer, defined by the relative density xi = ni/ns (where ns is the molar density of a completely filled adsorption layer) is obtained from self-consistent equations22,23 with the boundary condition x1 = xN. The excess adsorption isotherm is defined as: N

n ex = A Σ x i – Vpore ρg i =1

[10.20]

The model has been fitted to adsorption isotherms of hydrogen on AX-21 over the temperature range 25–298 K and the pressure range 0–6 MPa, yielding the following values for the model parameters: Ea = –3.871 kJ/mol, E = 0.519 kJ/mol and ns = 74.0 mol/l (saturation density of a layer, which is consistently larger than the liquid density). The saturation constant A has to be fitted to the function: A(T) = A0 – A1 exp (–B/T) with A0 = 30.426 mol/l, A1 = 12.623 mol/l and B = 71.042 K. The pore volume Vporewas fitted to a value of 1.62 l/kg.

10.5.4 Dubinin pore-filling approach The adsorption isotherm of microporous adsorbents have often been modeled by the Dubinin–Astakhov model. In this approach, which is based on a socalled pore-filling of an adsorbent by a subcritical gas (T < Tc), the total adsorbed density is expressed as: nA = n0 exp [–(A/E)m]

[10.21]

where n0 is the saturated density in the micropores of the adsorbent, m is a

© 2008, Woodhead Publishing Limited

274

Solid-state hydrogen storage

structural heterogeneity parameter, E is a characteristic energy of the average gas–solid interaction potential in the pores,24 and where A is the adsorption potential, defined as follows: A = RT ln (PA/P)

[10.22]

where R is the universal gas constant, PA is a constant which corresponds to the saturation pressure at the temperature T. The limits m = 1 and m = 2 reduce to the Freundlich and the Dubinin–Raduskevitch isotherms,25 respectively. In the supercritical regime, PA cannot be identified with pseudo-saturation vapour. It can be treated as a fitting parameter. Amankwah and Schwarz, for instance, determine Ps using:26 Ps = (T/TC)γPC

[10.23]

where Tc and Pc are the critical temperature and pressure of the adsorbate. Dubinin27 originally proposed γ = 2. The parameter γ is fitted in such a way that the characteristic curves of nA as function of A are independent of temperature. This approach has been used to characterize the adsorption properties of SWNTs, using H2, CO2 and N2 as adsorbate.28

10.6

Adsorption of hydrogen on activated carbons and carbon nanostructures

10.6.1 Activated carbons Activated carbons have been effectively proposed and tested as a transportation storage medium for methane and natural gas at room temperature. For hydrogen, however, acceptable excess adsorbed densities for storage applications require low-temperature operation and highly nanoporous activated carbons, with high specific surface area and high bulk density. Figure 10.6 shows the excess adsorption isotherms of hydrogen on the activated carbon AX-21, obtained from volumetric measurements over the supercritical temperature range of 35–300 K. This activated carbon has a specific surface of about 2800 m2/g and a bulk density of about 0.3 g/cm3. The micropore volume is estimated at 1.06 cm3/g. The isotherms are fully reversible and present no hysteresis. At 77 K, the maximum excess density adsorbed observed is about 54 g/kg at 35 bar. The contribution of the excess density to hydrogen storage is 5.4 wt%, relative to the mass of activated carbon. To obtain the full storage capacity, the contribution of the compressed bulk phase in the pore volume of the adsorbent must be added. The pore volume of the adsorbent can be estimated from the ratio of the density of activated carbon to graphite (assuming graphite corresponds to the compact limit of activated carbons and that the entire pore volume of the adsorbent is accessible to adsorbate molecules):

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

275

60

298 K 273 K 253 K 233 K 213 K 193 K 173 K 153 K 133 K 113 K 93 K 77 K

Excess adsorption (g/kg)

50

40

30

20

10

0

0

2

4 Pressure (MPa) (a)

6

8

120 77 K 60 K 45K 40 K 35 K 30 K 295K

Excess adsorption (g/kg)

100

80

60

40

20

0 0

2

4 Pressure (MPa) (b)

6

10.6 (a) Hydrogen excess adsorption in gram H2 per kilogram of activated carbon from 77 to 295 K. Data from P. Bénard and R. Chahine, Hydrogen Energy Progress XII, p 1121 (1998). (b) Lowtemperature isotherms from 30 to 77 K compared to the room temperature isotherm. Data from J. Michelsen, R. Chahine, A. Tessier and D. Cossement (unpublished, 2007).

© 2008, Woodhead Publishing Limited

276

Solid-state hydrogen storage

d AC   Vpore =  1 –  Vadsorbent d  graphite 

[10.24]

where dAC is the density of the activated carbon and dgraphite, the density of graphite (2.2 g/cm3). Dividing (10.24) by the mass of the adsorbent yields the gravimetric pore volume. For AX-21 this approximation yields a total pore volume of 2.88 cm3/g. The micropore volume of AX-21 thus represents 37% of the total pore volume. The mesopore and macropore volumes represent no more than 10% of the total pore volume, the rest of the pore volume accounting for the contribution of the intergranular porous structure and the external surface of the adsorbent. The pore volume obtained by curve fitting from equation (10.20) lies between the measured micropore volume and the estimate obtained from (10.24). The contribution of the pores to the gravimetric density is obtained by multiplying equation (10.23) by the bulk gas density of the adsorbent. This yields a contribution of: d AC  1  x bulk =  1 – ρ d graphite  d AC g 

[10.25]

to the gravimetric storage density. At 77 K and 35 bar, this yields xbulk = 0.033 or 3.3%. The net storage density is therefore 8.7% relative to the mass of the adsorbent. The high-pressure, low-temperature adsorbed density of hydrogen on activated carbon seems to correlate linearly with the micropore volume and the specific surface area of the adsorbent.29 The density of hydrogen inside the micropores can be estimated from the ratio of the excess density adsorbed to the micropore volume, to which must be added the contribution of the bulk phase in the micropores. This yields a value of 62 mg/cm3, suggesting that at 77 K and 35 bar, the density of hydrogen inside the micropore is already close to that of liquid hydrogen. Similar values (from 61 to 71 mg/ cm3) are obtained for other activated carbon adsorbents with different pore volumes and specific surfaces.29 As can be seen from Fig. 10.6, the storage capacity is extremely sensitive to temperature. The maximum density that can be stored as a cryoadsorbed gas occurs close the critical point. At 35 K and 0.6 bar, the contribution of the excess adsorbed density to storage is 9.5 wt%. The sorption properties of activated carbons have extensively been characterized in terms of the Dubinin model (see Poirier et al.30 and Table 10.1 below). Owing to the amorphous structure of activated carbons, grand canonical Monte Carlo simulations of hydrogen adsorption on activated carbons are difficult to perform. In practice, a random porous structure of a specific activated carbon coherent with the pore size distribution could be created

© 2008, Woodhead Publishing Limited

Carbon nanostructures for hydrogen storage

277

Table 10.1 Dubinin–Ashtakov parameters for hydrogen adsorption on single wall nanotubes (SWNTs-Pr and SWNTs-700) and on the activated carbon IRH-33 Sample

n0 (cm3/g)

E (J/mol)

m

γ

SWNTs-Pr SWNTs-700 IRH-33

0.33 0.40 0.80

3560 3400 3150

1.54 1.29 1.90

2.48 1.92 2.08

using the reverse Monte Carlo method which can yield three-dimensional particle configurations consistent with measured structural parameters, such as the measured structure factor or the pair correlation function g(r).31 Most hydrogen adsorption simulations on activated carbon have been performed assuming a graphite slit pore approximation. Wang and Johnson32 used a path integral Monte Carlo method. The hydrogen–hydrogen interactions were modeled using the Silvera–Goldman potential33 which was found to reproduce accurately the properties of molecular hydrogen over a wide range of pressure and temperature in path integral grand canonical simulations of hydrogen gas.34 The Crowell–Brown (uncorrugated) interaction potential was used to simulate the adsorbate–adsorbent interaction in the slit pores. Reasonable agreement could be obtained with experimental adsorption data of hydrogen on AX-21. Classical grand canonical Monte Carlo simulations of hydrogen in slit pores using standard Lennard–Jones potentials for the adsorbate– adsorbate and adsorbent–adsorbate interactions seem to indicate that a maximum of about 8 wt% excess density of hydrogen at 77 K and 40 bar can be expected with an optimal spacing of the carbon layers of about 20 Å.35

10.6.2 Single-wall nanotubes As discussed earlier, the cylindrical geometry of nanotubes is expected to lead to deeper adsorbate–adsorbent interaction potential well inside small diameter SWNT. In a bundle, the interstitial sites could be even more favourable due to the overlap of the molecular force fields of each SWNT. However, the small pore volume associated with these sites along with the small diameter of these pores makes their relative contribution to the overall adsorbed density small. SWNTs were first proposed as storage media for hydrogen by Dillon et al.4, using temperature programmed desorption (TPD) measurements on small samples (1 mg). The soot used in the measurements contained 0.1–0.2 wt% SWNT bundles of 7–14 nanotubes of approximately 12 Å in diameter. The hydrogen sorption properties of the SWNT samples were investigated for the pressure and temperature ranges of 0.03–0.4 bar and 80–500 K. The authors identified physisorption as the leading mechanism responsible for hydrogen

© 2008, Woodhead Publishing Limited

278

Solid-state hydrogen storage

adsorption and estimated the characteristic energy of adsorption activation energy measured to be 19.6 kJ/mol compared with 5.54 kJ/mol for the activated carbon AX-21.36 By extrapolating their results to room temperature, the authors expected storage capacities of 5–10 wt% of hydrogen on pure SWNTs. Owing to the small size of the samples, the small concentration of SWNTs within the samples and the presence of other impurities, these results have since been found to be unreliable. Volumetric measurements of the adsorption isotherms of hydrogen on SWNTs show that they are type 1 reversible isotherms similar to activated carbons (Fig. 10.7). Although a substantial dispersion in the experimental values of the adsorbed density was observed, improved measurements have been shown to be reliable.30 The small amount of SWNTs typically available for adsorption experiments requires that adsorbed density measurements be performed with sensitive instruments.37 The purity of the hydrogen used in the sorption experiments can also be an issue. In addition, the synthesis and preparation of the samples can influence substantially adsorption on SWNTs. Notably, thermal treatments can enhance the adsorption properties of SWNTs38 by removing guest species blocking access to adsorption sites. Current measurements show that under ambient conditions, pure SWNTs adsorb little hydrogen (10 Å) in MIL-100 and MIL-101 are not necessarily favourable or optimised for H2 storage. The Sc(III) material [Sc2(L)3] (L2– = 1,4-benzenedicarboxylate) has been shown to store 1.5 wt% of H2 at 1 atm at 77 K (Perles et al., 2005).

11.2.6 Non-transition metal carboxylates A MOF containing the tri-nuclear cluster M3O unit has been very recently obtained via the synthesis of the highly porous [In3O(L)1.5(H2O)3](H2O)3(NO3) (H4L = 3,3′,5,5′-azobenzenetetracarboxylic acid) (Liu et al., 2007). In this material three In cations are bridged by six 4-connected planar carboxylate ligands to afford 6-connected [In3O(O2R)6] moieties to form an overall squareoctahedral topology (O’Keeffe et al., 2000). The desolvated framework has a Langmuir surface area of 1417 m2/g and a pore volume of 0.50 cm3/g, with a H2 uptake of 2.61 wt% at 78 K and 1.2 bar. This is a high uptake considering the pressure, but potential saturation adsorption is limited due to its relatively

© 2008, Woodhead Publishing Limited

Metal–organic framework materials for hydrogen storage

301

small pore volume. Recently, new porous materials derived from carboxylate ligands and lighter Mg(II) (Dincă and Long, 2005; Rood et al., 2006; Senkovska and Kaskel, 2006) and Al(III) (Férey et al., 2003; Loiseau et al., 2006) centres have been reported.

11.2.7 Metal–organic frameworks (MOFs) based upon pyridine-carboxylate linkers Another approach in preparing porous materials is to combine two types of linker ligands within the same structure. Thus, bimetallic tetracarboxylate units M2(O2CR)4 can be combined with linear ditopic amine linkers that bind at the two axial sites of the paddlewheel (Fig. 11.7a). In these materials, the square planar 4-connected bimetallic nodes are, therefore, bound by linear linkers L defining a 4 × 4 two-dimensional grid, and these sheets are then pillared by linker P to form a porous 3-D framework. A series of Zn(II) complexes have been synthesised using this methodology (Chun et al., 2005; Lee et al., 2005; Chen et al., 2006b) and they all have, except C6 and C7 (Fig. 11.7), stable frameworks on removal of crystallised solvent. C1 has the highest BET surface area (1450 m2/g) and the others have similar BET surface areas around 1000 m2/g. The H2 adsorption curves confirm that the pendant arms on the linker L can help to take up more H2 at low pressure, but they also limit the potential of these materials in terms of total H2 capacity. The material C1 has the largest pore volume and therefore shows good potential for H2 storage at higher pressure. A further approach involves incorporation of both pyridyl and carboxylate donors within the same angular ligand. Thus, the pyridyl group can act as a terminal group replacing solvent molecules in the metal cluster node (Horike et al., 2006; Jia et al., 2006; Lin et al., 2006b; Humphrey et al., 2007). This methodology can also, in principle, produce framework materials of especially high connectivity with the pyridyl and carboxylate linkers limiting interpenetration when the length of the linker is increased (C.L. Chen et al., 2005; Ma et al., 2005). The 12-connected frameworks [Ni3(OH)(BPPC)3]·8DMF·10H2O and [Fe3(O)(BPPC)3]·3DMF·17H2O have been constructed using the tri-branched linker pyridine-3,5-bis(phenyl-4carboxylate) (BPPC2–) bound to the tricapped trigonal prismatic polyhedral metal nodes [Ni3(OH)] or [Fe3O], respectively (Jia et al., 2007). The overall topology for these highly unusual 12-connected frameworks is 31844256, and desolvation affords highly porous materials with large surface areas (1553 m2/g for Ni complex and 1200 m2/g for Fe complex). Their H2 adsorption capacities are proportional to their surface areas with the Ni complex adsorbing 4.15 wt% H2 at 77 K and 20 bar, with 3.05 wt% uptake for the Fe complex under the same conditions. Another successful approach to preparing porous materials has involved

© 2008, Woodhead Publishing Limited

302

Solid-state hydrogen storage L

P N

C1 Me

P O L

L

M

O O O

O O

C2 L

Me Me Me

Me M

O

Me

L

C3 Me

O

P

Me

L

P N

N N

C6

N

N N

N N

C7

N

C4

N F

F

F

F

C5

Me

Me

C8 Me

Me

N N

N N

N

(a)

C1 C2 C3 C4 C5 C6

Vp (cm3/g)

BET surface area (m2/g)

H2 uptake (1 bar, 77 K)

0.75 0.59 0.50 0.52 0.57 0.62

1450 1100 920 1000 1070 1120

2.01 2.08 1.85 1.70 1.78 1.68

(b)

11.7 Views of (a) framework formation in which different mixed carboxylate pyridyl linkers bind to a bimetallic node through its equatorial and axial directions, respectively, and (b) gas adsorption data for selected frameworks.

the linking of Ni(II) amine macrocyclic units with bridging carboxylic linkers (Lee and Suh, 2004).

11.2.8 Mn and Cu tetrazolates and related ligands 1,3,5-Benzenetristetrazole is a tri-branched linker incorporating N-donors that can bridge two metal centres (Fig. 11.8a). This is analogous to the carboxylate group, which can also stabilise multinuclear metal nodes. Reaction of 1,3,5-benzenetristetrazole (BTT) and MnCl 2 in DMF affords [Mn(DMF)6]3[(Mn4Cl)(BTT)8(H2O)12]2·42DMF·11H2O·20CH3OH which incorporates an 8-connected [Mn4Cl]7– node linked by 3-connected BTT ligands (Dincă et al., 2006). The fundamental building unit for the structure is the truncated octahedral as shown in Fig. 11.8(b) with octahedral sharing

© 2008, Woodhead Publishing Limited

Metal–organic framework materials for hydrogen storage Mn

Mn N Mn

303

N

N

N N

N N

Mn

N

N

N

N

N

Mn

Mn (a)

(b)

11.8 Views of (a) BTT linker and its connection to metal ions; (b) the crystal structure of sodalite-type Mn-BTT (or Cu-BTT) framework (Dinca˘ et al., 2007).

square faces to generate a cubic sodalite structure. In the solvated material, the encapsulated Mn(II) cations are coordinated by organic solvents, DMF and methanol. To generate a porous material, the DMF was exchanged with methanol followed by desolvation in vacuo to afford a material with partially desolvated Mn(II) and exposed coordination sites. The material has a BET surface area of 2100 m2/g and can adsorb 6.9 wt% H2 at 90 bar and 77 K. Significantly, the excess volumetric adsorption reaches 60 g/l of H2 which is well above the 45 g/l of DOE 2010 criteria, and only 11 g/l lower than the density of liquid hydrogen (71 g/l at 1 bar at 20 K). This raises a possibility of replacing current liquid H2 storage systems with a liquid N2 coolant at moderate pressures. Using CuCl2 in the same reaction produces a product that is isostructural with the Mn material (Dincă et al., 2007). The Cu complex can be completely desolvated under reduced pressure and heating to 120 °C. Significantly, the Cu material has a lower BET surface area (1710 m2/g) than its Mn analogue, and accordingly, adsorbs less H2 5.7 wt% at 90 bar at 77 K. Materials based on Mn(II)-formate (Dybtsev et al., 2004) and -carboxylates (Moon et al., 2006) and on Cu and Co (Ma and Zhon, 2006; Sun et al., 2006; Ma et al., 2007a) and Pd centres (Navarro et al., 2006) linked to bridging linkers have also been found to show H2 storage capabilities.

11.3

Interactions of H2 with metal–organic frameworks: experiments and modelling

Gas storage experiments with MOFs are typically conducted at 77 K, a temperature that is well above the critical temperature for H2. The weak

© 2008, Woodhead Publishing Limited

304

Solid-state hydrogen storage

binding of H2 to the inner wall of MOFs means that the molecules are dynamically disordered in the framework, and this makes crystallographic determination of H positions problematic (Rowsell et al., 2005b,c). Potential sites for interaction of H2 with the MOF interior include the metal centre(s), carboxylate functions and the aromatic rings (Yildirim and Hartman, 2005; Spencer et al., 2006). Analysis of the interaction of H2 within MOFs also comes from calculation of the heat of adsorption ∆Hads of H2 using the Clausius–Clapeyron equation (Jaroniek, 1988) or by a virial-type expression (Czepirski and Jagiello, 1989) to fit isotherm data at two temperatures. Inelastic neutron scattering (INS), X-ray scattering, Raman spectroscopy and neutron powder diffraction at low temperatures on D2-loaded MOFs can also probe the possible H2 adsorption sites in MOFs.

11.3.1 Interactions of hydrogen with exposed metal sites It has been postulated that exposed metal sites in MOFs attract and bind H2 molecules. Much effort has been placed into studies to confirm the existence of such interactions and to estimate their strength. Kaye and Long (2005) have reported studies on a series of dehydrated Prussian Blue analogues M 3[Co(CN)6]2 (M = Mn, Fe, Co, Ni, Cu and Zn) and on IRMOF-1 [Zn4O(BDC)3]. The heats of adsorption ∆Hads for the Prussian Blue analogues are in the region of 6.5–7.0 kJ/mol, except for Mn3[Co(CN)6]2 which has a low value of 5.3 kJ/mol. These values are significantly higher than for IRMOF1 for which ∆Hads = 4.7 kJ/mol. The dehydrated frameworks M3[Co(CN)6]2 do have exposed metal sites which are considered to be advantageous in attracting and binding H2 within the framework, and this parallels molecular chemistry where there are many examples of H2 binding to Mn complexes and other transition metal species (Kubas et al., 1984; Sweany and Watzke, 1997; Toupadakis et al., 1998). However, a high-energy X-ray scattering experiment on H2-loaded samples of Mn3[Co(CN)6]2 and time-of-flight neutron powder diffraction on D2-loaded samples of Mn3[Co(CN)6]2 have afforded no evidence for such binding interactions of the adsorbed H2 or D2 with the accessible metal sites (Chapman et al., 2006). These experiments were carried out at 30 and 77 K, respectively, using a differential pair distribution function (PDF) to estimate the position of H2 molecules. The most plausible position for the H2 molecules in the Mn3[Co(CN)6]2 framework, based upon these studies, is about a single disordered site at the centre of the pore, which maximises van der Waals interaction with the pore surface, which is lined with highly polarisable cyanide groups. At very low temperature (5 K), the positions of D2 molecules in the MOFs are relatively static, which allows use of neutron powder diffraction techniques to determine the positions of D2 molecules. The Rietveld profile analysis of

© 2008, Woodhead Publishing Limited

Metal–organic framework materials for hydrogen storage

305

neutron powder diffraction data of D2-loaded samples of Cu3[Co(CN)6]2 reveal two possible adsorption sites (Hartman et al., 2006), one at the centre of the pore, the (1/4, 1/4, 1/4) site, which is consistent with the PDF study on Mn3[Co(CN)6]2 (Chapman et al., 2006). The other is situated on (x, 0, 0) (x ≈ 0.183) sites, which results from [Co(CN)6]3– vacancies, suggesting a D2– Cu distance of 3.16 Å. The study suggested that at low loading, the (1/4, 1/4, 1 /4) site is the preferred adsorption site, and the weak D2–Cu interaction accounts for only ~25% hydrogen adsorption at high loading. Stronger binding of D2 to exposed metal sites in [Mn]3[(Mn4Cl)(BTT)8] (Section 11.2.8) has been established by neutron powder diffraction at 3.5 K (Dincă et al., 2006). A Mn(II)–D distance of 2.27 Å suggests that D2 molecules interact significantly with exposed Mn(II) sites, but this distance is significantly longer than distances reported for σ-bonded η2-H2 complexes of group 6 metal ions (M–H c. 1.7–2.0 Å ) (Kubas, 2001). Neutron powder diffraction also reveals that once the exposed Mn(II) sites are saturated, D2 molecules are sited at the surface of cavities created by the tetrazolate linkers. This D2–Mn binding appears to be consistent with the high heat of adsorption of H2 observed for [Mn]3[(Mn4Cl)(BTT)8]. In the low H2 coverage region, ∆Hads = 10.1 kJ/mol, higher than the observed value of 5.9 kJ/mol for Mn3[Co(CN)6]2. In the high H2 coverage region, however, the heats of adsorption ∆H ads of H 2 in [Mn] 3 [(Mn 4 Cl)(BTT) 8 ] (5.7 kJ/mol) and Mn3[Co(CN)6]2 (5.3 kJ/mol) are very similar. We should note that the neutron scattering experiments were conducted at 3.5 K when the D2 molecules are condensed inside the micropores of the [Mn]3[(Mn4Cl)(BTT)8] framework. Although these experiments confirm the binding of H2 molecules with the exposed Mn(II) sites in [Mn]3[(Mn4Cl)(BTT)8] at 3.5 K, it is still difficult to speculate upon the strength of interaction of H2 at exposed Mn(II) at 77 K (Dincă et al., 2006). Neutron powder diffraction studies on the Cu analogue of [Mn]3[(Mn4Cl)(BTT)8] gave (Dincă et al., 2007) a Cu–D2 distance of 2.47 Å. Compared with the data for [Mn]3[(Mn4Cl)(BTT)8], the longer Cu–D2 distance is in agreement with the slightly lower heat of adsorption of 9.5 kJ/mol for the Cu material. At H2 loading of 1.6 wt% the heats of adsorption for the Mn and Cu materials are the same (6.0 kJ/mol) suggesting adsorption of H2 at high loading is controlled primarily by pore size and shape. The binding of H2 to exposed metal sites has also been probed for [Cu3(BTC)2]∞ (HKUST-1 in Section 11.2.4). Rietveld analysis of the neutron powder diffraction data for D2-loaded samples of [Cu3(BTC)2] ∞ reveals that there are six possible adsorption sites, the most favourable being the unsaturated axial site of the binuclear Cu fragment (Peterson et al., 2006). The Cu–D2 distance of 2.39 Å suggests significant interaction between H2 and Cu(II), but this is much weaker than that observed for molecular σ-bonded η2-H2 complexes. Interestingly, the heat of adsorption of H2 binding in HKUST-1

© 2008, Woodhead Publishing Limited

306

Solid-state hydrogen storage

is 6.8 kJ/mol at low coverage, significantly lower than the 9.5 kJ/mol observed ( for [Cu]3[(Cu4Cl)(BTT)8] (Dinca, 2007) even though the Cu–D2 distance is longer in the latter material. A possible interpretation of this observation is that the Cu–D2 interaction is not a covalent bond but rather is highly constrained by the shape and size of the cavity of the MOF. Additionally, values of ∆Hads of 6.0 kJ/mol at 1.9 wt% H2 loading suggests no significant difference in binding preference between the six possible adsorption sites, as determined by neutron diffraction studies. Exposed metal sites are also present in the desolvated In-based complex, [In3O(ABTC)1.5](NO3) (H4ABTC = 3,3′,5,5′-azobenzenetetracarboxylic acid) (Liu et al., 2007). In contrast to the above Cu and Mn systems, INS studies on [In3O(ABTC)1.5](NO3) indicate occupation of multi-sites in the MOF. This is also in agreement with the analysis of ∆Hads from H2 isotherms for this material, which showed a nearly constant value (6.5 kJ/mol) at loadings below 1.8 wt%. In this case with In(III) present at the node, open metal sites do not necessarily out-perform other sites in the MOF structure, and the geometry of the pores and framework topology are key factors in determining strong interaction between H2 and the porous host. Recently, new porous materials derived from carboxylate ligands and lighter Mg(II) centres have been reported (Dincă and Long, 2005; Senkovska and Kaskel, 2006) and these have been shown to have high isosteric heats of absorption of 7.0– 9.5 kJ/mol reflecting the increased van der Waals contact area associated with a small pore and possible interactions with the Mg(II) centre.

11.3.2 Interactions of hydrogen with metal–organic frameworks (MOFs) without exposed metal sites The Zn-based IRMOF series of materials has been intensively studied. The Raman spectra of H2- and D2-adsorbed IRMOF-1 confirm substantial shifts of 10 cm–1 and Q-branch line-broadening of H–H and D–D vibrations for the adsorbed gases compared with free gases (Centrone et al., 2005). The shifts suggest the presence of prevailing attractive forces, and line-broadening indicate the existence of multiple adsorption sites. INS studies at 15 K for IRMOF-1, IRMOF-8, IRMOF-11 and MOF-177 reveal that the most favorable adsorption site is close to carboxylate groups (Rosi et al., 2003; Rowsell et al., 2005b), with the Zn4O cluster being a less attractive site and the positions around aryl groups appearing to show the weakest interactions with H2 molecules. Interestingly, the differences in the heats of adsorption between these different sites vary upon framework structure even though they incorporate very similar building blocks (Rowsell and Yaghi, 2006). Thus, for IRMOF-1 the heat of adsorption varies slightly from 4.8 kJ/mol at zero coverage, decreasing to 4.2 kJ/mol at 0.9 wt% H2 adsorption. For the

© 2008, Woodhead Publishing Limited

Metal–organic framework materials for hydrogen storage

307

double-interpenetrating framework IRMOF-11, the heat of adsorption at low coverage is 9.1 kJ/mol and drops off considerably with the amount adsorbed, reducing to 5.1 kJ/mol at 1.2 wt% H2 adsorption. These data indicate that the strong binding sites in the Zn-based MOFs are largely due to the overlap of the attractive potential of proximal surfaces. In summary, although there is evidence supporting the binding of H2 at exposed metal sites in MOFs below 5 K, the presence of such binding is still in doubt at 77 K. Also, the binding of H2 to the exposed metal sites can be affected by the framework structure and the choice of metal centre. The variations in heat of adsorption of H2 on different MOFs suggest that the topology and structure of the framework material intimately affect the adsorption mechanism. This serves to underline the importance of exploring new framework materials and topologies to rationalise and understand the mechanism underpinning H2 uptake and storage in MOFs.

11.3.3 Modelling Owing to the structural similarities of the IRMOF series and the extensive reports on their gas adsorption properties, most modelling work reported thus far has focused on this series of materials. Grand canonical Monte Carlo (GCMC) simulation of H2 adsorption has produced acceptable agreement with experimental results, particularly for the low pressure ( 150 bar for about 1000 hours. Joubert et al. [116] studied the effect of amount of Sn added on the lattice parameters of LaNi5. However, the MmNi5 hydride counterpart with Sn alloying showed a significant loss in hydrogen capacity after cycling; see Balasubrmaniam [117]. Kim and Lee [75] studied cycling behavior of MmNi4.5Fe0.85, and MmNi4.5Al0.85 and reported good retention of hydriding properties and also showed reproportionation after 3000 thermal cycles. Chandra and Lynch [111] reported thermal aging effects on LaNi4.25Al0.75 for nuclear applications, and found no disproportionation when aged at 200 °C with intrinsic thermal aging at a hydrogen pressure of 19 atm for ~210 hours; the pressure actually dropped by 2 atm during aging, indicating annealing of the hydride. The isotherms (taken at 116 °C) were very similar but with no disproportionation. Vacuum aging of this intermetallic also did not lead to any disproportionation; in fact the plateau pressures reduced slightly [111]; a 5 year study at the Savannah River Company, at Aiken, SC, on this sample showed a severely sloped plateau, with very broad X-ray diffraction Bragg peaks after 202 days, and was worse after that [111]. In addition, there are several studies performed in Percheron-Guegan’s [29, 72] and Latroche’s [73, 118] group at CNRS in Paris, R.C. Bowman Jr at Jet Propulsion Laboratory CalTech, Pasadena, California, Frank Lynch at HCI, Bailey, Colorado, Jai Young Lee’s group at KIST, Seoul, S. Korea, and others.

12.4.2 Decrepitation of hydrides – disproportionation aspects The process of hydrogenation and dehydrogenation of a metal lattice causes lattice expansion and contraction, and in general leads to the development of fine powders called decrepitation. This is one of the most common phenomena that occur in metal hydrides. The starting materials to make hydrides are often coarse particles each in order of several mm in size. Each of these particle sizes, of course, has many grain boundaries in the micrometer range. Once the hydrogen molecules interact with the surface, a catalytic reaction generates H+ ions which move relatively fast through the grain boundaries (at low temperatures), and slower diffusivities through the volume of the grain. After repeated cycling, disintegration of the particles occurs [5], which is very common in hydrides in elemental as well as in intermetallic hydrides or alloys; the degree depends on the number of cycles and the plastic behavior of the alloy or metal. In the case of LaNi5 the particles are slowly attritioned

© 2008, Woodhead Publishing Limited

338

Solid-state hydrogen storage

to a minimum size after which there is very little reduction in size. This may be due to micro-plasticity exhibited in intermetallics, which are generally brittle. Many of the elemental hydrides (e.g. vanadium hydrides) have a metallic character, and plastic deformation occurs during the hydriding that accommodates the strain, thus preventing significant decrepitation [20]. For an example of pressure cycling of classical CaCu5, LaNiAl0.3, and hydrides, see Goodell [119], which shows the effect of pressure cycling for LaNi5Al0.3 and CaNi5 (Fig. 12.14). It can be seen that stability of LaNi5Al0.3 and also MmNi4.5Al0.5 hydrides (not in this figure) appear to have virtually no disproportionation; possibly due to Al ordering in Ni sub-lattice [75]. Figure 12.14 also shows that the hydriding properties are almost recovered by reproportionation.

12.4.3 Intrinsic thermal cycling elemental vanadium alloyed with 0.5 at.%C hydrides One of the interesting elemental hydrides that have more than one plateau is that of vanadium metal. Note that binary CaNi5 intermetallic hydride also has two plateaus [97]. Vanadium hydrides have been proposed to be used in Joule Thomson sorption pump for NASA’s surveillance satellites [121], using VHx and LaNi4.3Sn0.24 hydride sorbent beds. Besides, the vanadium hydrides have the highest volumetric capacity amongst the elemental hydrides, and can be hydrided at room temperature. There are three different hydride phases formed during hydriding of V metal: V2H (β1), VH (β2), and VH2 (γ). Early work on thermodynamic properties of vanadium was done by Fagerstroem et al. [122]. Later, Reilly and Wiswall [123], Fujita et al. [124], Griffiths et al. [125], Schober [126], Veleckis and Edwards [127], and Rummel [128] reported on the low- and high-pressure hydriding characteristics of pure vanadium. It is the high-pressure mixed-phase region between the β and γ phases that is suitable for the space-related applications mentioned above [9, 129–132]. The stabilization of the γ phase to facilitate hydrogen desorption and absorption between γ↔β2 was extensively studied by adding alloying elements [123, 126, 128, 133]. Further studies were performed by Yukawa et al. [134, 135] on stabilizing both the β and γ phases by adding alloying elements (V-1 mol%M and V-3 mol%M, where M is a 3d-, 4d-, or 5dtransition metal, such as Ti, Cr, Mn, Fe, Co, and Ni. Details of low-pressure hydrides are shown in Chandra et al. [137]. An example of the α ↔ β1 and β2 ↔ γ plateau continuity as a single plot in hydriding vanadium is shown in Fig. 12.15 [28]. Note that data were not collected between 0.1 and 31.6 kPa, so a dotted line is drawn to suggest the continuity of the isotherm to the data in the high-pressure region. The formation of α phase, existence of an α ↔ β1 plateau, and single β1 and β2 phase regions are all consistent with the features reported by Griffiths et al. [125]

© 2008, Woodhead Publishing Limited

Cycling at 85 °C 0 → 2068 kPa (300 psia) 30 min cycle period

Pressure (atm) H2 or D2

0.6

LaNi5

0.4 CaNi5 0.2 Reproportioning treatments 0

500

1000 Cycle number (a)

1500

5000

20

2000

10

1000

5

500

2

200

1.0

100

.5

50

.2

20

.10

10

.05

O2

.02 .01

5

H2

0

1

2

3 4 5 x in CaNi5Hx (b)

6

2 7

12.14 Comparison of hydrogen capacity loss as function number of pressure cycles for LaNi4.7Al0.3, LaNi5.2, and CaNi5, before and after pressure cycling at 85 °C between 0 and 2068 kPa (300 pounds/square inch absolute, psia) [14]. This shows a significant loss in CaNi5 but this loss is recovered after reproportionation. (b) Reference isotherm of CaNi5 (taken at 25 °C) showing three different plateaus due to isotope [120].

Intermetallics for hydrogen storage

Hydrogen transfer (H/M)

0.8

50

Pressure (kPa) H2 or D2

LaNi4.7Al0.3

0

10,000

100

1.0

339

© 2008, Woodhead Publishing Limited

340

Solid-state hydrogen storage 7

γ

Isotherms taken at room temperature (298 K) β2 + γ

6

280 000 Pa

5

VH0.8 (BCT)

4

VH2 (FCC)

β2

V H z

log p (Pa)

3

y x

y

2

z

x

1 V (BCC)

0 –1

y z

–2

α + β1

–3 –4

Pure V (absorption)* Pure V (desorption) V-0.5% C (absorption) V-0.5% C (desorption) 4000 cycles (absorption) 4000 cycles (desorption)

β1

x

0.001 Pa 0

0.5

1 H/V

1.5

2

12.15 Low- and high-pressure isotherms of pure vanadium [125]*, and high-pressure isotherms of V–0.5at.%C at 298 K showing the α, α + β1, α1 + β2, and β2 + γ phase regions [28]. The structures of VH2 and others shown above are made by using data from Maeland et al. [138, 139].

for pure vanadium hydrides. The high-pressure data of Luo et al. [137] did agree in terms of plateau pressures, but their maximum H/M of 1.62 is quite low, as compared to a value here of H/M = 1.95 at 298 K; the maximum H/ M was also low for highly strained V–0.5%C alloy. The maximum H/M value for pure vanadium of Chandra et al. [28] agreed well with the data of Reilly and Wiswall (H/M ~2 at 313 K) [123]. The high-pressure isotherms of V–0.5%C taken at 298 K showed a β1 ↔ γ plateau at approximately 230 kPa (Fig. 12.15). Small changes in plateau pressure are attributed to the presence of carbon in the lattice. In Fig. 12.16, we also see the effect of prolonged thermal cycling between two hydride phases β2 ↔ γ of V–0.5%C for 4000 times on V–0.5%C hydride. It should be noted that the cycling was performed in a P–T intrinsic test apparatus and the isotherms were obtained in a separate experiment in a Sievert’s apparatus. It is interesting to note that desorption plateau pressures (under all conditions) are nearly the same, but the successive absorption plateau pressures are relatively higher as a function of number of cycles. It is suggested that dislocation motion during hydriding may contribute to this effect, considering that there is a large volume change during cycling between the β2 ↔ γ phases. Further discussion on the effect of thermal

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

341

NH 0.47 4400

ABS 20

2.0

ABS 1000

4000

DES 20

DES 778

DES 1000

3600

ABS 1345

ABS 1633

ABS 4000

DES 1345

DES 1633

DES 4000

3200

p (Pa) × 10–3

0.6

2800 2400 2000 1600 1200 800 400 0 0.7

0.9

1.1

1.3 1.5 r = H/M

1.7

1.9

2.1

12.16 Isotherms showing the effect of thermal cycling on the plateau absorption and desorption pressures measured at 298 K. Absorption data for 778 cycles is not available. Taken from Chandra et al. [28, 140].

cycling, crystal structure and domain size, and microstrain, 1/2, effects before and after are discussed by Chandra et al. [28] and Sharma [30].

12.5

Extrinsic testing of intermetallic hydrides

12.5.1 Extrinsic gaseous impurity effects on LaNi5 and Fe-Ti Relatively few studies have been performed on extrinsic cycling of classical metal hydrides. Sandrock and Goodel [141] reported on early extrinsic tests performed in the 1980s. They described the effect of impurities on the contemporary classical AB5 and AB hydrides by gaseous CO, CO2, CH4, C2H4, N2, NH3, H2S and CH3SH (methyl mercaptan). In these experiments thin (1.5 mm) compressed disks were used to evaluate cycle life. These alloy disks were first activated at room temperature using purified hydrogen. The majority of the impurity tests were performed at 25 and 85 °C in these experiments. A (dynamic) isotherm was obtained before and after the impurity test by pressure cycled between the plateau pressures [142]; although details are described in the reference it is important to mention that cycle time was 30 minutes with 15 minutes each for absorption and desorption (using high-

© 2008, Woodhead Publishing Limited

342

Solid-state hydrogen storage

purity hydrogen gas premixed with impurity gas in ppm levels before testing), and hydrogen absorption vs. time profiles were obtained, and shown in Fig. 12.17. Impurity tests on LaNi5 and Fe-Ti alloy with H2S and CH3SH and CO impurity gases exhibit a rapid loss of hydrogen capacity, and considered to have been poisoned [142].

Hydrogen absorption (H/M)

1.0 Cycle no. = 0

0.8

1 0.6

2

0.4 3 0.2 5 0

4

H2 + 300ppm CH3SH

6 0

1

2 3 Time (minutes) (a)

4

1.0

Hydrogen absorption (H/M)

Cycle no. = 0 1

0.8

2 4

0.6

10 0.4

20

40

0.2 H2 + 300ppm NH3 0

0

1

2 Time (minutes) (b)

3

4

12.17 Hydrogen absorption as a function of time showing losses after extrinsically cycling with mercaptan (CH3SH) and NH3 gas impurities in H2. From Sandrock and Goodell [141].

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

343

Tests on ppm levels of oxygen in hydrogen revealed significant loss in capacity due to oxidation and generally this effect is irreversible [141]. In another case Sandrock and Goodell tested the effect of ppm levels of NH3, CO2, and CO which slowed down the kinetics of absorption/desorption; this effect was referred to as ‘retardation.’ Nitrogen and CH4 did not have any effects upon cycling on these two hydrides. The Sandrock–Goodell damage model was developed for hydrogen transfer capacity using ppm levels of impurity gas in hydrogen for any class of material [142]: φ) [ ] [ ] exp –  [(C –( NAqr C )/100]

∆ H =∆ H M M

0

o

+m

  

where ∆[H/M]0 = initial hydrogen transfer capacity, N = number of cycles, C = impurity concentration, Co = threshold concentration for damage, φ = damage function or hydridable metal atoms consumed/impurity atom, q = [H/M]0 stoichiometric ratio for an intermetallic, r = number of active impurity atoms/impurity molecules, A = (0.001)(1–m) and m = concentration dependence. For example, LaNi5 and TiFe0.85Mn0.15 tests performed with CO at 25 °C, showed φ = 1733 (m = 1) and 1320 (m = 1), whereas NH3 at 25 °C showed φ = 8.6 (m = 0.5) and 12.6 (m = 0.5), and O2 at 25 °C showed φ = 1.4 (m = 0.5) and 169 (m = 1). For further details, refer to Sandrock and Goodell [141].

12.5.2 Cycling test and response to oxygen as a minor impurity degradation behavior of AB5 hydrides Short- and long-term extrinsic tests on cyclic life of LaNi5 were reported by Goodell [119]. Short-term tests (30 cycles) showed an initial rapid decline up to 3 cycles and then a slow recovery of hydrogen transfer up to 30 cycles. The change is attributed to the absorption kinetics rather than changes in ultimate capacity, i.e. the reaction slowed down; minor variation in alloy composition did not change the trend. The dip in the curve is due to availability of oxygen, or more cyclic dependence. Tests run at 85 °C showed remarkably good hydrogen transfer without a decrease in hydrogen transfer, as compared with that of 25 °C tests up to 30 cycles [94]; similar results were found with CO gas. However beyond 30 cycles, this hydrogen transfer decreased rapidly (almost linearly) from 30 to 100 cycles at 85 °C. McHenry [143] observed severe degradation in LaNi5 hydriding properties with H2–0.5%O2 at ~25 °C. Alloy surface poisoning with gaseous impurities in hydrogen has a profound effect on the hydriding properties and is highly dependent on the surface structure of the alloys or compounds [144]; thus surface poisoning or contamination of metal hydrides is a very important research issue. Surface

© 2008, Woodhead Publishing Limited

344

Solid-state hydrogen storage

poisoning may be initiated with small amounts of impurities in hydrogen and is a rather complex problem. Examples of surface contamination, with 300 ppm of H2O, O2, and CO in H2 gas, for Fe0.85Mn0.15Ti and LaNi5 hydrides are shown in Fig. 12.18 [143]. This impurity, although small in quantity, has profound effects on the surface catalytic properties in long-term cyclic use. It appears that surface poisoning inhibits the H2→2H catalytic dissociation, which is essential for hydrogen absorption, and perhaps affects hydrogen diffusion in the structure. The resistance to degradation of surface properties

Hydrogen absorption (H/M)

0.8 H2O 0.6

O2

0.4

CO

0.2

0

0

10

20

30

20

30

Cycle number (a)

Hydrogen absorption (H/M)

1.0 0.8 H2O O2

0.6 CO 0.4 0.2 0

0

10 Cycle number (b)

12.18 Extrinsic cyclic response of (a) Fe0.85Mn0.15Ti, 0.5 (hydrogen cycle at 25 °C, 69 → 634 → 69 kPa) and (b) LaNi5 (0.5 hour cycle at 25 °C, 69 → 276 → 69 kPa.) using impure hydrogen high temperature and pressure aging at nearly isothermal conditions. Figures from Sandrock and Goodell [144].

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

345

due to attack by the impurity needs to be evaluated for different alloys. In LaNi5 and Fe-Ti type hydrides, there is propensity to form La2O3, La(OH)3, or TiOx when trace oxygen gas is adsorbed on the surface. The heat of formation of La2O3 is very high (–1277 kJ/mol) as compared with that of TiO2 (–913 kJ/mol). The nickel (or iron) and lanthanum atoms on the surface are oxidized during the initial cycles. The O2 and H2O surface-contaminated Fe0.85Mn0.15Ti hydrides do not recover but the LaNi5 hydride does recover after about 10 cycles. The carbon monoxide contamination is very severe in the case of both LaNi5 and Fe0.85Mn0.15Ti alloys (Fig. 12.18). There is a steep decay in the hydrogen transfer property as seen in Fig. 12.14. It is postulated that a Ni(CO)4 layer forms due to reaction with CO gas [53]. To summarize, the AB5 hydrides (LaNi5) are resistant to extrinsic impurities such as O2 and H2O, but not CO gas. The (Fe, Mn)Ti are generally not resistant to any of these impurities; they are slightly resistant to CO gas.

12.6

Extrinsic cycling of complex hydrides

Sandrock et al. [145] pointed out that the high impurity level of the hydrogen released from complex hydrides during desorption can cause problems during cycling. Impurity effects in hydrogen gas itself, as well as evolved gases for prolonged cycling in PEM fuel cell application are detrimental; especially hydrocarbons that lead to the formation of CO. Sandrock et al. [145] reported Dry-TiCl3 catalyzed Na-Al hydrides did not evolve any impurity gas during reactions. Gross et al. [146] and Chen et al. [147] reported that the hydrogen capacities were lowered to 3 wt% using liquid catalysts (both absorption and desorption). They also performed X-ray diffraction measurements and showed that the low hydrogen capacity was mainly due to an inability to completely recharge to a composition of 100% NaAlH4. Full desorption of hydrogen to NaH and Al was also shown, but mechanisms need attention. In general, fresh charges of hydrogen (with ppm or ppb impurity) at every cycle will cause degradation of hydrides and eventual loss of hydrogen absorption capacity.

12.6.1 Pressure extrinsic cycling studies on imide/amide system An example of Li amide–imide pressure cycling data is presented, although there are detailed presentations in this book on thermodynamics and structure of these materials by other authors. Chandra et al. [148] determined the effect of long-term pressure cycling between (1, 56, 163, 501, and 1100 cycles) for the Li–N–H system using industrial hydrogen [hydrogen min% (v/v) 99%, water ~ 32 ppm, O2 ~ 10 ppm, N2 ~ 400 ppm, Total hydrocarbons:

© 2008, Woodhead Publishing Limited

346

Solid-state hydrogen storage

10 ppm, CO2 ~ 10 ppm, CO ~ 10 ppm, argon may be present, reads as oxygen]. Figure 12.19 shows isotherms (between Li2NH and LiNH2) taken after a certain number of cycles; a loss of ~2.53 out of ~5.6 wt% hydrogen was observed after 1100 pressure cycles. It can be noted that the capacity drops to ~25% of its initial capacity after the first 500 cycles, but subsequently there is little further loss in capacity. The material was subjected to one hour hydriding–dehydriding cycles (30 minutes each) with maximum pressure of 2 bar and vacuum by introducing fresh industrial hydrogen at every cycle. X-ray diffraction analyses and isotherms (after cycling in desorbed condition) contaminated with impure hydrogen are also shown in Fig. 12.19. The phases in desorbed condition showed mainly Li2NH, LiH, and Li2O phases. The Xray diffraction (XRD) pattern showed nearly complete reversal from LiNH2 → Li2NH after cycling with expected LiH phase and impurity phase (Li2O). After 1100 pressure cycles, the LiH phase increases from 11 wt% to 58 wt%, and the Li2NH phase decreases from 82 wt% to 13 wt%. The decrease in amount of Li2NH can be observed in the Bragg peak at 30.5° of the XRD patterns. The impurity Li2O phase was increased from 7 wt% to 29 wt% after 1100 cycles. The increased amount of the Li2O phase is indicated by the increase in Bragg peak intensity at 56° of the XRD patterns. The decreases in phase concentrations of Li2NH (imide) phases is due to the formation of Li2O and stable LiH; a very small amount of LiOH phase is observed. In another study, thermal aging of Li2NH with 100 ppm of CO in UHP hydrogen also used 100 ppm CO for approximately 240 hours at 12 bar and 325 °C. The isotherms obtained before and after aging did not show any significant change. It does appear that the behavior of these complex hydrides is quite different from those of the classical hydrides, as the degraded oxides are not readily reduced back to the parent phases under reasonable thermodynamic conditions. Similarly, we performed extrinsic tests using 100 ppm CH4 in hydrogen and found that there was a decrease in hydrogen capacity of ~1.5 wt.%, but the kinetics (not shown here) were not affected significantly. The methane impurity tests were obtained by pressure cycling 40 times at 225°C, shown in Fig. 12.20 [149].

12.7

Conclusions

This chapter summarizes important contributions on the AB5/AB and elemental hydrides. Examples of microalloying in LaNi5- and FeTi-based intermetallic hydrides have been emphasized, and there has been some discussion on elemental hydrides. Classification of testing methods for durability of alloys/ intermetallics were presented, examples of intrinsic and extrinsic tests that include cycling and aging have been reported for AB5/AB and other alloys. More recent results on extrinsic tests on lightweight complex hydrides (imideamide) that were cycled between Li2NH/LiNH2 phases were also reported.

© 2008, Woodhead Publishing Limited

@LiOH

250 2.0

0

2

3 wt% H (a)

4

5

6

50

+ Li2NH + Li2NH + Li2NH + Li2NH ++ LiH ++ LiH ++ LiH ++ LiH

163 cycles #Li2O

501 cycles

1100 cycles

+ Li2NH ++ LiH

++ LiH

#Li2O #Li2O

56 cycles

#Li2O

++ LiH ++ LiH ++ LiH

1 cycle

#Li2O

+ Li2NH + Li2NH + Li2NH + Li2NH

60

++ LiH

+ Li2NH

#Li2O

#Li2O #Li2O #Li2O #Li2O

+ Li2NH + Li2NH + Li2NH + Li2NH

++ LiH

+ Li2NH

40

++ LiH

+ Li2NH + Li2NH

30

++ LiH

#Li2O #Li2O #Li2O #Li2O + Li NH 2 + Li2NH + Li NH + Li2NH 2 ++ LiH ++ LiH ++ LiH ++ LiH #Li2O #Li2O #Li2O #Li2O

+ Li2NH

#Li2O + Li2NH ++ LiH #Li2O

1

1000 500 0

@

+ Li2NH

0

Intensity (counts)

P (bar)

Li2NH starting material After 56 cycles After 163 cycles After 501 cycles After 1100 cycles

0.1

40

70

2θ (°) (b)

© 2008, Woodhead Publishing Limited

347

12.19 (a) Absorption isotherms obtained after non-equilibrium pressure cycling for 1, 56, 163, 501, 1100 times with impure hydrogen (ppm levels of O2, H2O and others). (b) Corresponding ex-situ X-ray diffraction taken from the sample after each isotherm was obtained in desorbed condition. Data taken from Chandra et al. [148].

Intermetallics for hydrogen storage

1

1000 500 0 1000 500 0 1000 500 0 1000 500 0

+ Li2NH

30

++ LiH

0.5

10

++ LiH

0.0

H: Li–N–H 1.0 1.5

348

Solid-state hydrogen storage 100

Extrinsic pressure After cycling cycling Li2NH s LiNH2 Before cycling

P (bar)

10

1

LiNH2 Li2NH

0.1

0.01

0

1

2

Li N HH Li N

3 4 wt% H

5

6

7

12.20 Isotherms of Li2NH–LiNH2 taken at 255 °C after 40 extrinsic pressure cycles are shown. See Fig. 12.10(b)–(d) for pressure cycling procedure. This cycling was performed using 100 ppm CH4 mixed with UHP hydrogen. Note that hydriding kinetics did not change, but the hydrogen capacity reduced ~2 wt% hydrogen. Molecular models made using Crystal Maker software from the data of T. Noritake et al. [149] for Li2NH, and K. Miwa et al. [150]. The isotherms are taken from Chandra et al. [148].

In general, a reasonable amount of intrinsic data is available, but extrinsic data are lacking; more research is needed for the hydrides to be used in practical applications.

12.8

Acknowledgements

I would like to acknowledge the US DOE Metal Hydride Center of Excellence and Sandia National Laboratory for their support for our hydride research program. I want to gratefully acknowledge the valuable support of Gary Sandrock and P. Goodell for this chapter and over the years. I would also like to thank Frank Lynch, Jim Reilly, Louis Schlapbach, Ricardo Schwarz, Joseph Wermer, and Bob Bowman for their collaborations and helpful suggestions over the years. I gratefully thank Wen-Ming Chien, Raja Chellappa, Josh Lamb, Mike Coleman, A. Sharma, Sandeep Bagchi, and Steve Lambert of my hydrogen group at the University of Nevada, Reno for their contributions to this chapter.

12.9

References

1. Bogdanovic, B. and Schwickardi, M., Ti-doped alkali metal aluminium hydrides as potential novel reversible hydrogen storage materials, Journal of Alloys and Compounds, 1997, 253: p. 1–9.

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

349

2. Hoffman, K.C., et al., Metal hydride storage for mobile and stationary applications, International Journal of Hydrogen Energy, 1976, 1(2): p. 133–151. 3. Chandra, D., Reilly, J.J. and Chellappa, R., Metal hydrides for vehicular applications: the state of the art, Journal of Metals, 2006, 58(2). 4. Fukai, Y., The Metal-Hydrogen System – Basic Bulk Properties, 1993, Berlin: Springer-Verlag. 5. Schlapbach, L., ed. Hydrogen in Intermetallic Compounds II: Surface and Dynamic Properties, Applications, Topics in Applied Physics. 1992, Springer: Berlin. 6. Bowman, R.C., Development of metal hydride beds for sorption cryocoolers in space applications, Journal of Alloys and Compounds, 2003, 356: p. 789–793. 7. Panasonic, Available from: http://www.panasonic.com/industrial/battery/oem/chem/ nicmet. 8. Coleman, M., et al. in Advanced Materials for Energy Conversion II, TMS Annual Meeting, 2004. Charlotte, NC. 9. Lynch, F.E., Metal hydride practical applications, Journal of the Less-Common Metals, 1991, 172: p. 943–958. 10. Aoki, K., Yanagitani, A. and Masumoto, T., Crystalline to amorphous transformation in laves phase GDFE2 induced by hydrogen absorption, Applied Physics Letters, 1988, 52(25): p. 2122–2123. 11. Gingl, F., Vogt, T. and Akiba, E., Trigonal SrAl2H2: the first Zintl phase hydride, Journal of Alloys and Compounds, 2000. 306(1–2): p. 127–132. 12. Chen, Y., et al., Structure and hydrogen storage property of ball-milled LiNH2/ MgH2 mixture, International Journal of Hydrogen Energy, 2006, 31(9): p. 1236– 1240. 13. Shang, C.X., et al., Mechanical alloying and electronic simulations of (MgH2 + M) systems (M=Al, Ti, Fe, Ni, Cu and Nb) for hydrogen storage, International Journal of Hydrogen Energy, 2004, 29(1): p. 73–80. 14. Varin, R.A., et al., Formation and environmental stability of nanocrystalline and amorphous hydrides in the 2Mg–Fe mixture processed by controlled reactive mechanical alloying (CRMA), Journal of Alloys and Compounds, 2004, 373(1–2): p. 270–286. 15. Mueller, W.M., Blackledge, J.P. and Libowitz, G.G., Metal Hydrides, 1968, New York: Academic. 16. Schlapbach, L., ed. Hydrogen in Intermetallic Compounds I: Electronic, Thermodynamic, and Crystallographic Properties, Preparation, Topics in Applied Physics. Vol. 63. 1988, Springer Verlag: Berlin. 17. Sandrock, G., A panoramic overview of hydrogen storage alloys from a gas reaction point of view, Journal of Alloys and Compounds, 1999, 295: p. 877–888. 18. Yvon, K., Complex transition-metal hydrides, Chimia, 1998, 52(10): p. 613–619. 19. Libowitz, G.G., The Solid State Chemistry of Binary Metal Hydrides, 1965, New York: Benjamin Press. 20. Lambert, S.W., et al., Investigation of hydriding properties of LaNi4.8Sn0.2, LaNi4.27Sn0.24 and La0.9Gd0.1Ni5 after thermal cycling and aging, Journal of Alloys and Compounds, 1992, 187: p. 113–135. 21. Sandrock, G. and Thomas, G., The IEA/DOE/SNL on-line hydride databases, Applied Physics A – Materials Science & Processing, 2001, 72(2): p. 153–155. 22. Luo, W. and Ronnebro, E., Towards a viable hydrogen storage system for transportation application, Journal of Alloys and Compounds, 2005. 404: p. 392– 395.

© 2008, Woodhead Publishing Limited

350

Solid-state hydrogen storage

23. Sandrock, G. and Thomas, G., IEA/DOE/SNL Hydride Databases, available from: http://hydpark.ca.sandia.gov. 24. Sandrock, G., et al., Accelerated thermal decomposition of AlH3 for hydrogenfueled vehicles, Applied Physics A – Materials Science & Processing, 2005, 80(4): p. 687–690. 25. Sandrock, G., State of the art review of hydrogen storage in reversible metal hydrides for military fuel cell applications, in Final Report to ONR, Arlington, VA 1992. 26. Sandrock, G., Hydrogen–metal systems, in Hydrogen Energy System – Production and Utilization of Hydrogen and Future Aspects, 1995, Kluwer Academic: Dordrecht. p. 135. 27. Ohoyama, K., et al., Revised crystal structure model of Li2NH by neutron powder diffraction, Journal of the Physical Society of Japan, 2005. 74(1): p. 483–487. 28. Chandra, D., Sharma, A., Chellappa, R., Cathey, W.N., Lynch, F.E., Bowman Jr, R.C., Wermer, J.R. and Paglieri, S.N., Hydriding and structural characteristics of thermally cycled and cold worked V-0.5 at.% alloy, Journal of Alloys and Compounds, 2008, 452: p. 312–324. 29. Percheron-Guegan, A., Lartigue, C. and Achard, J.C., Correlations between the structural properties, the stability and the hydrogen content of substituted LaNi5 compounds, Journal of the Less Common Metals, 1985. 109(2): p. 287–309. 30. Sharma, A., Effect of Thermal Cycling and Cold-work on V0.995C0.005 Hydrides, 1992, University of Nevada, Reno: Reno, Nevada. 31. Finholt, A.E., Bond, A.C. and Schlesinger, H.I., Lithium aluminum hydride, aluminum hydride and lithium gallium hydride, and some of their applications to organic and inorganic chemistry, Journal of the American Chemical Society, 1947, 69: p. 1199. 32. Brower, F.M., et al., Preparation and properties of aluminum-hydride, Journal of the American Chemical Society, 1976, 98(9): p. 2450–2453. 33. Graetz, J. and Reilly, J.J., Decomposition kinetics of the AlH3 polymorphs, Journal of Physical Chemistry B, 2005, 109(47): p. 22181–22185. 34. Graetz, J. and Reilly, J.J., Nanoscale energy storage materials produced by hydrogendriven metallurgical reactions, Advanced Engineering Materials, 2005, 7(7): p. 597–601. 35. Libowitz, G.G., The nature and properties of transition metal hydrides, Journal of Nuclear Materials, 1960, 2: p. 1–11. 36. Flanagan, T.B. and Clewley, J.D. Hysteresis in metal-hydrides, Journal of the Less-Common Metals, 1982, 83(1): p. 127–141. 37. Libowitz, G.G., Hayes, H.F. and Gibb, T.R.P., The system zirconium–nickel and hydrogen, Journal of Physical Chemistry, 1958, 62: p. 76. 38. Hoffman, K.C., et al., BNL-20168. 1975, Brookhaven National Laboratory: Upton, New York. 39. Orimo, S., Fujii, H.G. and Yoshino, T., Reactive mechanical grinding of ZrNi under various partial pressures of hydrogen, Journal of Alloys and Compounds, 1995, 1995: p. 287–294. 40. Buschow, K.H.J., Thermal stability and magnetic properties of Zr(1–x)Fex Alloys, Journal of Less-Common Metals, 1981, 79: p. 243–253. 41. van Vucht, H.N., Kuijpers, F.A. and Bruning, H.C.A.M. Philips Research Reports, 25, 1970. 42. Reilly, J. and Wiswall, R.H., Formations and properties of iron titanium hydride, Inorganic Chemistry, 1974, 13: p. 218.

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

351

43. Buchner, H., Energiespeicherung in Metallhydriden, 1982, Springer: Vienna. 44. Buschow, K.H.J., Hydrogen absorption in intermetallic compounds, in Handbook of the Physics and Chemistry of Rare Earths, K.A. Geschneidner and L. Eyring, Editors. 1984, Elsevier: Lausanne. p. 1–104. 45. Garg, S.C. and McClaine, A.W., Metal Hydrides for Energy Storage Applications, Tech. Note N-1393. 1975, Navy Civil Engineering Laboratory: Port Hueneme, CA. 46. Newkirk, H.W., Hydrogen Storage by Binary and Ternary Intermetallics for Energy Applications – A Review, Report URCL-52110. 1976, Lawrence Livermore Laboratory: Livermore, CA. 47. Reilly, J.J., Metal Hydrides as Hydrogen Storage Media and Their Applications, in Hydrogen: Its Technology and Implications, Vol. II, K.E. Cox and K.D. Williamson Jr., Editors. 1977, CRC: Cleveland. p. 13–48. 48. Reilly, J.J., Metal hydride technology, Zeitschrift fur Physikalische ChemieWiesbaden, 1979, 117: p. 155–184. 49. Reilly, J.J. and Sandrock, G.D., Hydrogen storage in metal hydrides, Scientific American, 1980, 242(2): p. 118–129. 50. Sandrock, G., State-of-the-art Review of Hydrogen Storage in Reversible Metal Hydrides for Military Fuel Cell Applications, Final Report for ONR Contract NOO014-97-M-0001, NTIS Order No. AD-A328 073/2INZ. 51. Sandrock, G.D., Hydrogen Storage, in Hydrogen Energy, Rpt. 199. 1981, Royal Swedish Academy of Engineering Sciences: Stockholm. p. 67–151. 52. Sandrock, G.D. and Huston, E.L., How metals store hydrogen, Chemtech, 1981, 11(12): p. 754–762. 53. Snape, E. and Lynch, F.E., Metal-hydrides make hydrogen accessible – II. Chemtech, 1980, 10(12): p. 768–773. 54. Suda, S., Recent advancements of metal hydride heat pump developments, Chemical Engineering (Japan), 1983, 18: p. 1–29. 55. Suda, S., ed., Hydrogen Storage Alloys – Devlopments and Applications to Energy Systems, 1984, Ohyou Gijuto Shuppan: Tokyo. 56. Swisher, J.H. and Johnson, E.D., Hydrides versus competing options for storing hydrogen in energy-systems, Journal of the Less-Common Metals, 1980, 74(2): p. 301–320. 57. van Mal, H.H., Stability of Ternary Hydrides and Some Applications, 1976, Technishe Hogeschool Delft. 58. Wenzl, H., Properties and applications of metal hydrides in energy conversion systems, International Metals Review, 1982, 27: p. 140. 59. Wiswall, R.H., Hydrogen Storage, in Hydrogen in Metals II: Application Oriented Properties, G. Alefeld and Volkl, J., Editors, Springer: Berlin, 1978. p. 201. 60. Snape, E., Huston, E.L. and Sandrock, G.D., Development of solar-hydrogen systems using metal hydrides, in 2nd Miami Int. Conference on Alternative Energy Sources. 1981. 61. Buchner, H., The Hydrogen/Hydride Energy Concept, in Hydrides for Energy Storage, Anderson A.F. and Maeland, A.J. Editors, 1978, Pergamon: Oxford, p. 569–599. 62. Alefeld, G., Introduction, in Hydrogen in Metals II, Alefeld G. and Volkl, J. Editors. 1977, Heidelberg: Berlin. p. 2–10. 63. Anderson, A.F. and Maeland, A.J., eds. Hydrides for Energy Storage, 1978, Pergamon: Oxford.

© 2008, Woodhead Publishing Limited

352

Solid-state hydrogen storage

64. Huston, E.L., Liquid and Solid Storage of Hydrogen, in 5th World Hydrogen Energy Conference, 1984, Pergamon, Oxford. 65. Luo, S., et al., Thermodynamic and degradation studies of LaNi4.8Sn0.2-H using isotherms and calorimetry, Journal of Alloys and Compounds, 1995, 231(1–2): p. 473–478. 66. Luo, S., et al., Thermodynamic studies of the LaNi5–xSnx–H system from x = 0 to 0.5, Journal of Alloys and Compounds, 1995, 231(1–2): p. 467–472. 67. Suda, S. Recent development of hydride energy systems in Japan, in 5th World Hydrogen Energy Conference, 1984: Pergamon, Oxford. 68. Wang, X.L. and Suda, S., Reaction-kinetics of hydrogen metal hydride systems, International Journal of Hydrogen Energy, 1990, 15(8): p. 569–577. 69. Luo, S., et al., Split plateaux in the LaNi5–H system and the effect of Sn substitution on splitting, Journal of Alloys and Compounds, 1997, 253: p. 226–231. 70. Bowman, R.C., et al., The effect of tin on the degradation of LaNi5–ySny metalhydrides during thermal cycling, Journal of Alloys and Compounds, 1995, 217(2): p. 185–192. 71. Cerny, R., et al., Anisotropic diffraction peak broadening and dislocation substructure in hydrogen-cycled LaNi5 and substitutional derivatives, Journal of Applied Crystallography, 2000, 33(4): p. 997–1005. 72. Percheron-Guegan, A. and Lartigue, C., Materials Science Forum, 1988, 31(125). 73. Latroche, M., Paul-Boncour, V. and Percheron-Guegan, A., Structural instability in R1–xNi2 Compounds and their hydrides (R = Y, rare-earth), Zeitschrift fur Physikalische Chemie – International Journal of Research in Physical Chemistry & Chemical Physics, 1993, 179: p. 261–268. 74. Han, J.I. and Lee, J.Y., An investigation of the intrinsic degradation mechanism of LaNi5 by thermal-desorption technique, International Journal of Hydrogen Energy, 1988, 13(9): p. 577–581. 75. Kim, G.H. and Lee J.Y., The changes of hydrogenation properties induced by thermal cyclings in MmNi4.5Al0.5 and MmNi4.15Fe0.85, Journal of the Less-Common Metals, 1987, 132(1): p. 123–132. 76. Uchida, H., Terao, K. and Huang, Y.C., Current problems in the development and application of hydrogen storage materials, Zeitschrift fur Physikalische Chemie Neue Folge, 1989, 164: p. 1275–1284. 77. Yvon, K. and Fisher, P., Crystal structure and magnetic structures of ternary metal hydride: a comprehensive review, in Hydrogen in Intermetallic Compounds I: Electronic, Thermodynamic, and Crystallographic Properties, Preparation, L. Schlapbach, Editor. 1988, Springer Verlag: Berlin. p. 88–138. 78. Gupta, M. and Schlapbach, L., Electronic properties, in Hydrogen in Intermetallic Compounds I: Electronic, Thermodynamic, and Crystallographic Properties, Preparation., L. Schlapbach, Editor. 1988, Springer Verlag: Berlin. p. 139–217. 79. Flanagan, T.B. and Oates, W.A., Thermodynamics of intermetallic compound– hydrogen systems, in Hydrogen in Intermetallic Compounds I: Electronic, Thermodynamic, and Crystallographic Properties, Preparation, L. Schlapbach, Editor. 1988, Springer Verlag: Berlin. p. 49–85. 80. Archad, J.C., et al. in Proceedings of the 2nd World Congress. 1978, Zurich, Switzerland. 81. Mendelsohn, M., Gruen, D. and Dwight, A., Group 3A and 4A substituted AB5 hydrides, Inorganic Chemistry, 1979, 18(12): p. 3343–3345. 82. Mendelsohn, M.H., Gruen, D.M. and Dwight, A.E., LaNi5–xAlx is a versatile alloy system for metal hydride applications, Nature, 1977. 269(5623): p. 45–47.

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

353

83. Mendelsohn, M.H., Gruen, D.M. and Dwight, A.E., Effect on hydrogen decomposition pressures of group IIIA and IVA element substitutions for Ni in LaNi5 alloys, Materials Research Bulletin, 1978, 13(11): p. 1221–1224. 84. Mendelsohn, M.H., Gruen, D.M. and Dwight, A.E., Effect of aluminum additions on the structural and hydrogen absorption properties of AB5 alloys with particular reference to the LaNi5–xAlx ternary alloy system, Journal of the Less-Common Metals, 1979, 63(2): p. 193–207. 85. Sandrock, G.D. in Proceedings of the 2nd World Congress, 1978. Zurich, Switzerland. 86. van Mal, H.H., Buschow, K.H.J. and Miedema, A.R., Hydrogen absorption in LaNi5 and related compounds: experimental observations and their explanation, Journal of Less-Common Metals, 1974, 35: p. 65–76. 87. Trzeciak, M.J., Dilthey, D.F. and Mallet, M.W., Battelle Menorial Report BMI 1112, 1956: Columbus, Ohio. 88. Beck, R. and Mueller, W.M., Investigation of Hydriding Characteristics of Intermetallic Compounds, in Summary Report, AEC Contract AT(33-3)-3. 1962, Denver Research Institute: Denver, CO. 89. Pebler, A. and Gulbransen, E.A., Thermochemical and structural aspects of the reaction of hydrogen with alloys and intermetallic compounds of zirconium, Electrochemical Technology, 1966, 4: p. 211–215. 90. Pebler, A. and Gulbransen, E.A., Equilibrium studies on the system ZrCr2-H2, ZrV2-H2, ZrMo2-H2 between 0 and 900 °C, Transactions of the Metallurgical Society of AIME, 1967, 239, p. 1593–1596. 91. Sandrock, G. and Thomas, G., Compilation of IEA/DOE/SNL hydride databases, in Report IEA/H2/TR-97. 1997, International Energy Agency, Paris. 92. Jacob, I. and Shaltiel, D., Hydrogen absorption in Zr(AlxN(1–x), B=Fe, Co) Laves phase compounds, Solid State Communications, 1978, 27: p. 1750180. 93. Jacob, I. and Shaltiel, D., Hydrogen sorption properties of some AB2 Laves phase compounds, Journal of the Less-Common Met., 1979, 65: p. 117–128. 94. Reilly, J. and Wiswall, R.H., Hydrogen storage and purification systems III, in Report BNL 21322. 1976, Brookhaven National Laboratory. 95. Bernauer, O., et al., Fundamentals and properties of some Ti/Mn based Laves phase hydrides, International Journal of Hydrogen Energy, 1989, 14: p. 187–200. 96. Reilly, J.J. and Wiswall, R.H., BNL Report 16546. 1972, Brookhaven National Lab. p. 1. 97. Sandrock, G.D., et al., On the disproportionation of intermetallic hydrides, Zeitschrift fur Physikalische Chemie Neue Folge, 1989, 164: p. 1285–1290. 98. Ahn, S.T., Kim, Y.G. and Lee, J.Y., Formation of the amorphous phase in Zr2Al by hydrogen absorption and the effects of titanium substitution on the amorphization behavior, Journal of Alloys and Compounds, 1992, 186(1): p. 45–52. 99. Aoki, K., Yamamoto, T. and Masumoto, T., Hydrogen induced amorphization in RNi2 Laves phases, Scripta Metallurgica, 1987, 21(1): p. 27–31. 100. Kim, Y.G. and Lee, J.Y., Hydrogen-induced transformation to an amorphous state in the Laves phases Ce(Ru, M)2 (M Equivalent-to Fe, Co, Ni), Journal of Alloys and Compounds, 1993, 191(2): p. 243–249. 101. Cohen, R.L. and West, K.W., Intrinsic cycling degradation in LaNi5 and annealing procedures for re-forming the material, Journal of the Less-Common Metals, 1983, 95(1): p. 17–23. 102. Lundin, C.E. and Lynch, F.E., Solid State Hydrogen Storage Materials for Application to Energy Needs, 1975, First Annual Technical Report, AFOSR Contract F4462074-C-0020, University of Denver.

© 2008, Woodhead Publishing Limited

354

Solid-state hydrogen storage

103. Lundin, C.E. and Lynch, F.E., Modification of hydriding properties of AB5 type hexagonal alloys through manganese substitution, in International Conference on Alternate Energy Sources, T.N. Veziroglu, Editor. 1978: University of Miami. 104. Cohen, R.L., West, K.W. and Wernick, J.H., Degradation of LaNi5 by temperatureinduced cycling, Journal of the Less-Common Metals, 1980, 73(2): p. 273–279. 105. Cohen, R.L., West, K.W. and Wernick, J.H., Degradation of LaNi5 hydrogenabsorbing material by cycling, Journal of the Less-Common Metals, 1980, 70(2): p. 229–241. 106. Goodell, P.D. and Rudman, P.S., Hydriding and dehydriding rates of the LaNi5-H system, Journal of the Less Common Metals, 1983, 89(1): p. 117–125. 107. Gamo, T., et al., Life properties of Ti-Mn alloy hydrides and their hydrogen purification effect, Journal of the Less-Common Metals, 1983. 89(2): p. 495–504. 108. Park, J.M. and Lee, J.Y., The intrinsic degradation phenomena of LaNi5 and LaNi4.7Al0.3 by temperature induced hydrogen absorption–desorption cycling, Materials Research Bulletin, 1987, 22(4): p. 455–465. 109. Benham, M.J. and Ross, D.K., Experimental-determination of absorption–desorption isotherms by computer-controlled gravimetric analysis, Zeitschrift fur Physikalische Chemie Neue Folge, 1989, 163: p. 25–32. 110. Josephy, Y., Bershadsky, E. and Ron, M., Investigation of LaNi5 upon prolonged cycling in International Symposium on Metal–Hydrogen Systems, Sept. 2–7 1990. Banff, Alberta, Canada. 111. Chandra, D. and Lynch, F.E., in Rare Earths Proceeding, Bautista R.G. and Wong, M.M., Editors, 1989, TMS Publication: Warrendale, PA p. 83–98. 112. Dantzer, P., Static, dynamic and cycling studies on hydrogen in the intermetallics LaNi5 and LaNi4.77Al0.22, Journal of the Less-Common Metals, 1987, 131: p. 349– 363. 113. Marmaro, R.W. and Lynch, F.E., Investigation of long term stability, in metal hydrides, in Report NAS9-18175. 1991, Hydrogen Consultants, Inc.: Littleton. 114. Bowman Jr, R.C., Luo, C.H. and Ahn, C.C., The effect of tin on the degradation of LaNi(5–y)Sny hydrides during thermal cycling, Journal of Alloys and Compounds, 1995, 217: p. 185–192. 115. Bowman Jr., R.C., et al., Degradation behavior of LaNi5–xSnHx (x = 0.20–0.25) at elevated temperatures, Journal of Alloys and Compounds, 2002. 330–332: p. 271– 275. 116. Joubert, J.-M., Latroche, M., Cerny, R., Bowman Jr, R.C., Percheron-Guéqan, A. and Yuon K., Crystallography study of LaNi5–x Sn0.2 ≤ x ≤ 0.5 compounds and their hydrides, Journal of Alloys and Compounds, 1999, 293–295: p. 124–129. 117. Balasubrmaniam, R., Mungole, M.N. and Rai, K.N., Hydriding properties of MmNi5 system with aluminum, manganese and tin substitutions, Journal of Alloys and Compounds, 1993, 196: p. 63–70. 118. Latroche, M., et al., Structural studies of LaNi 4 CoD 6.11 and LaNi3.55Mn0.4Al0.3Co0.75D5.57 by means of neutron powder diffraction, Journal of Alloys and Compounds, 1995, 218(1): p. 64–72. 119. Goodell, P.D., Stability of rechargeable hydriding alloys during extended cycling, Journal of the Less-Common Metals, 1984, 99(1): p. 1–14. 120. Sandrock, G.D., et al., Hydrides and deuterides of CaNi5, Materials Research Bulletin, 1982, 17: p. 887–894. 121. Bowman Jr, R.C., Freeman, B.D. and Ryba, E.L., International Absorption Heat Pump Conference ASME, 1993, 31, p. 265.

© 2008, Woodhead Publishing Limited

Intermetallics for hydrogen storage

355

122. Fagerstroem, C.-H., Manchester, F.D. and Pitre, J.M., H-V (hydrogen–vanadium), in Phase Diagrams of Binary Hydrogen Alloys, F.D. Manchester, Editor, 2000, ASM International: Materials Park, Ohio, p. 273–292. 123. Reilly, J.J. and Wiswall, R.H.J., The higher hydrides of vanadium and niobium, Inorganic Chemistry, 1970, 9(7): p. 1678–1682. 124. Fujita, K., Huang, Y.C. and Tada, M., The studies on the equilibria of Ta-H, Nb-H, and V-H systems, Journal of Japanese Institute of Metals, 1979, 43(7): p. 601– 610. 125. Griffiths, R., Pryde, J. and Righini Brand, A., Phase diagram and thermodynamic data for the hydrogen/vanadium system, Journal of the Chemical Society Faraday Transactions I, 1972, 68: p. 2344–2349. 126. Schober, T., Vanadium-, niobium- and tantalum-Hydrogen, Solid State Phenomena, 1996, 49–50: p. 357–422. 127. Veleckis, E. and Edwards, R.K., Thermodynamic properties in the systems vanadium– hydrogen, niobium–hydrogen, and tantalum–hydrogen, Journal of Physical Chemistry, 1969, 73(3): p. 683–692. 128. Rummel, W., Selective absorption of hydrogen isotopes by vanadium and nickeltitanium, Siemens Forschungs-Und Entwicklungsberichte – Siemens Research and Development Reports, 1981, 10(6): p. 371–378. 129. Bowman Jr, R.C., Ryba, E.L. and Freeman, B.D., Characterization of prototype sorption cryo coolers for periodic formation of liquid and solid hydrogen, Advances in Cryogenic Engineering, 1994, 39: p. 1499. 130. Bowman, R.C., Freeman, B.D. and Phillips, J.R., Evaluation of metal hydride compressors for applications in Joule–Thomson cryocoolers, Cryogenics, 1992, 32(2): p. 127–137. 131. Bowman, R.C., et al., Performance testing of a vanadium hydride compressor, Zeitschrift Fur Physikalische Chemie – International Journal of Research in Physical Chemistry & Chemical Physics, 1994, 183: p. 245–250. 132. Cantrell, J.S., et al., Studies of phase compositions and hydrogen diffusion in VHX, Zeitschrift Fur Physikalische Chemie-International Journal of Research in Physical Chemistry & Chemical Physics, 1993, 181: p. 83–88. 133. Meuffels, P., KFA Julich, Report No. 2081, 1986. 134. Yukawa, H., et al., Alloying effects on the stability of vanadium hydrides, Journal of alloys and compounds, 2002, 330–332: p. 105–109. 135. Yukawa, H., et al., Alloying effects on the hydriding properties of vanadium at low hydrogen pressures, Journal of Alloys and Compounds, 2002. 337(1–2): p. 264– 268. 136. Chandra, D., et al., Vanadium Hydrides at Low and High Pressure, Final Report to Tritium Science and Engineering Group, 2006, Los Alamos National Laboratory. 137. Luo, W.F., Clewley, J.D. and Flanagan, T.B., Thermodynamics and isotope effects of the vanadium–hydrogen system using differential heat conduction calorimetry, Journal of Chemical Physics, 1990, 93(9): p. 6710–6722. 138. Maeland, A.J., Investigation of vanadium-hydrogen system by X-ray diffraction techniques, Journal of Physical Chemistry, 1964, 68: p. 2197. 139. Maeland, A.J., Gibb Jr, T.R.P. and Schumacker, D.P., A novel hydride of vanadium, Journal of American Chemical Society, 1961, 83: p. 3728. 140. Lamb, J., Chandra, D., Coleman, M., Sharma, A., Cuthey, W.N., Wermer, J.R., Paglieri, S.N., Bowman Jr, R.C. and Lynch, F.E. Low and high pressure hydriding of V–0.5art% C alloy, submitted to Journal of Nuclear Materials.

© 2008, Woodhead Publishing Limited

356

Solid-state hydrogen storage

141. Sandrock, G. and Goodell, P.D., Cyclic life of metal hydrides with impure hydrogen: overview and engineering considerations, Journal of the Less-Common Metals, 1984, 104: p. 159–173. 142. Huston, E.L., Development of a Commercial Metal Hydride Process for Hydrogen Recovery, Rept BNL 35440, 1984, Brookhaven National Laboratory: Upton, NY. 143. McHenry, E., A new design for a low-pressure Ni–H2 Cell, in Fall Meeting of the Electrochemical Society. 1977. 144. Sandrock, G.D. and Goodell, P.D., Surface poisoning of LaNi5, FeTi and (Fe, Mn)Ti BY O2, CO and H2O, Journal of the Less-Common Metals, 1980, 73(1): p. 161–168. 145. Sandrock, G.D., et al., Engineering considerations in the use of catalyzed sodium alanates for hydrogen storage, Proceeding of the 2000 Hydrogen Program Review, NREL/CP-570-28890. 2000. 146. Gross, K.J., et al., In-situ X-ray diffraction study of the decomposition of NaAlH4, Journal of Alloys and Compounds, 2000, 297(1–2): p. 270–281. 147. Chen, P., et al., Interaction of hydrogen with metal nitrides and imides, Nature, 2002, 420(6913): p. 302–304. 148. Chandra, D., Chien, W. and Lamb, J., Effect of Gaseous Impurities on Long-Term Thermal Cycling, Final Report DOE FY 2007, 2007, DOE Hydrogen Program. p. 91–99. 149. Noritake, T., et al., Crystal structure and charge density analysis of Li2NH by synchrotron X-ray diffraction, Journal of Alloys and Compounds, 2005, 393(1–2): p. 264–268. 150. Miwa, K., et al., First-principles study on lithium amide for hydrogen storage, Physical Review B (Condensed Matter and Materials Physics), 2005, 71(19): p. 195109–195116.

© 2008, Woodhead Publishing Limited

13 Magnesium hydride for hydrogen storage D. G R A N T, University of Nottingham, UK

13.1

Introduction

Magnesium is used primarily to fabricate lightweight structural alloys but its low density and reactivity also make this metal hydride attractive for hydrogen storage applications with a high reversible energy density of 9 MJ kg–1 and a hydrogen capacity of 7.7 wt%. This has attracted substantial interest to the hydrogen storage community for half a century. Therefore this chapter will introduce the background to magnesium and magnesium hydride in terms of its history, production, structure and properties in Section 13.2. Section 13.3 will introduce the thermodynamics of the hydrogenation and dehydrogenation of MgH2. The key drawbacks to commercialisation of magnesium hydrogen stores such as store temperature, kinetics and susceptibility to contamination are also discussed in Section 13.3. Approaches to combat these disadvantages are ball milling (Section 13.4), metal catalysts (Section 13.5) and metal oxide catalysts (Section 13.6). Models exploring the kinetic mechanisms of these systems are introduced in Section 13.7. Finally combinations of catalysts and the catalytic and synergistic properties of mixed systems are highlighted in Section 13.8 in conclusions and future trends.

13.2

Background to magnesium and magnesium hydride

The name magnesium originates from Magnesia, a city in Greece where large deposits of magnesium carbonate were discovered in ancient times. Magnesium is relatively plentiful, making up 2.7% of the Earth’s crust. It occurs in three isotopes: 24Mg(79%), 25Mg(10%) and 26Mg(11%). Its high reactivity means that magnesium is not found as a metal but as compounds and ores. There are commercial amounts of magnesium ores in most countries. Its success in manufacturing methods such as in die cast processing combined with the improved corrosion resistance of the latest Mg alloys and the variety of applications exploiting its structural strength to weight ratio has resulted 357 © 2008, Woodhead Publishing Limited

358

Solid-state hydrogen storage

in a global production that has doubled since the late 1980s to over 400 000 tonnes per annum. Magnesium production dates back to 1808 when Sir Humphry Davy found that magnesium oxide was the oxide of a previously unknown metal. He was the first to isolate a small quantity of magnesium from magnesium oxide. Twenty years later Antoine Bussy obtained purer and larger amounts of the metal by fusing magnesium chloride with metallic potassium vapour to get the metallic form and is credited with the discovery of magnesium, while in 1833 Michael Faraday electrolysed dehydrated liquid magnesium chloride to form liquid magnesium and chlorine gas. Commercial production started in Germany in 1886 after Robert Bunsen demonstrated electrolysis of the fused chloride in 1852. Lloyd Montgomery Pidgeon built the first industrial metallothermic magnesium extraction plant in Canada based on early German patents, in which dolomite was reduced with ferrosilicon under vacuum. Today two main processes are used to produce metallic magnesium. One is based on the electrolysis of fused anhydrous magnesium chloride (MgCl2) derived from magnesite, brine or seawater, and accounts for about 80% of the output, while the other is the thermal reduction of magnesium oxide (MgO) by ferrosilicon derived from carbonate ores. Another method recently being developed uses electrolysis of fused anhydrous MgCl2 derived from serpentine ores. Magnesium hydride synthesis goes back almost 100 years with the pyrolysis of ethylmagnesium halides (Jolibois, 1912) to give a mixture of magnesium hydride, magnesium halide and ethylene: 2MgXC2H5 → MgH2 + MgX2 + 2C2H4 However pure magnesium hydride was obtained in 1951 by the pyrolysis of diethylmagnesium at c. 400 K in vacuum: Mg(C2H5)2 → MgH2 + 2C2H4 This tends to produce an off-white finely divided solid that is pyrophoric. Other methods include the treatment of diethylmagnesium with diborane or lithium aluminium hydride: 3MgEt2 + B2H6 → 3MgH2 + 2BEt3 More recently reacting Mg with hydrogen at 200 bar at 850 K with a catalyst gives magnesium hydride in a larger form which is not pyrophoric: Mg + H2 → MgH2 The commercial production of light metal hydrides is of significant importance for future hydrogen storage application development (Eigen et al., 2007). Since the process temperature must be significantly below the equilibrium pressure and the melting temperature of Mg is 925 K then

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

359

magnesium is hydrogenated in the solid state which can take days (Bouaricha et al., 2001) or longer. Reactive milling is commonly investigated and complete hydrogenation of Mg can be in as little as a few hours (Huot et al., 1995), but more often mill times are of the order of several days (e.g. de Castro et al., 2004a; Gutfleisch et al., 2005) to provide suitable mixing of the catalyst and the required microstructure and size of the powders to enable effective hydriding/dehydriding kinetics of a potential hydrogen store. Developing milling methods to economically produce large quantities of metal hydrides including magnesium hydride is an important challenge (Eigen et al., 2007).

13.2.1 Properties and structure A quick look at all the binary hydrides will reveal that apart from V most are outside the idealised hydrogen store conditions of having a dissociation pressure 1–10 bar at temperatures between room temperature and 400 K. This has been known for some time (Libowitz, 1965, Sandrock, 1999), Fig. 13.1, and has led to approaches in which strong hydride-forming elements have been combined with weak hydriding elements to form alloys (many of 1000

600

T (°C) 400 300 200

100

50

25

10000 NiH

Dissociation pressure (bar)

1000 100 M g H2

10 1 0.1

PdH0.6

ThH2 YH2

Th4H15

0.01 0.001

VH2

TiH2 ZrH2 CaH2 LaH2 0.5

UH3 NaH

1

1.5

2 T –1 (10–3 K–1)

2.5

3

3.5

13.1 Dissociation pressures for elemental hydrides. The ideal pressure and temperature window for a practical hydride store for a transport application using a polymer electrolyte membrane (PEM) fuel cell is indicated by the grey box.

© 2008, Woodhead Publishing Limited

360

Solid-state hydrogen storage

them intermetallic) or complexes based on borohydrides and aluminates in an attempt to obtain the target range of dissociation pressure and temperatures. This oversimplified rule of mixtures approach to alloys and complex mixtures has met with partial success. However, magnesium hydride still holds interest for many researchers due to its low atomic weight, high hydrogen storage capacity, low metallic cost and simplicity of structure despite its slow kinetics and storage operating temperature of 600 K. Attempts to ameliorate and tackle these drawbacks are discussed in Sections 13.3–13.6, but it is useful, before then to consider the properties of magnesium and its hydride that have kept the imagination and faith in a simple system which has some unique properties even among its fellow saline hydrides. Magnesium is a silvery white metal of atomic mass 24.305 and has an ignition temperature of 918 °C in dry air which decreases with increasing moisture content. Oxidation of Mg at room temperature and 423 K for 15 min results in an oxide layer of less than 2 nm (Fournier et al., 2002). The rate of oxidation dry and in moist air, forming Mg(OH)2, is logarithmic at 600 K. Magnesium reacts with gaseous chlorine to form magnesium chloride and with nitrogen at c. 800 K to form Mg3N2. Magnesium in its finely divided form is readily ignitable and can ignite spontaneously in the presence of water or cutting fluids containing fatty acids. The hydrogen produced is an explosion hazard in addition to fire. Magnesium has a hexagonal crystal structure, space group P63/mmc with lattice parameters of a = 0.321 nm, c = 0.521 nm at 293 K, resulting in a density of 1.74 × 103 kg m–3. Magnesium hydride is an off-white colour and, when finely divided, is pyrophoric in air at room temperature. It has a desorption temperature of c. 600 K at 1 bar H2 which, although much lower than many other binary hydrides (see Fig. 13.1), is thought to be too high for a practical device (Bérubé et al., 2007), with estimates suggesting a third of the hydrogen in a store would be required to maintain the store temperature required. Uptake of hydrogen gas starts with the adsorption and dissociation of hydrogen molecules on sites on the magnesium surface followed by diffusion and hydride formation. During hydrogenation of magnesium, as H is introduced into the hexagonally close-packed (HCP) Mg metal lattice, the H atoms initially occupy tetrahedral interstitial sites, forming the α-phase, a solid solution of H in Mg, for H concentrations up to a maximum of c. 9 at% near the melting temperature of c. 650 °C (San-Martin and Manchester, 1987). Upon further addition of H, the β-MgH2 phase is formed. β-MgH2 has a body centre tetragonal lattice of rutile type, space group P42/mnm, with lattice parameters a = 0.452 nm and c = 0.302 nm (Noritake et al., 2002) and a density 1.42 × 103 kg m–3. The Mg atoms are octahedrally coordinated to six H atoms, while the H atoms are coordinated to three Mg atoms in a planar coordination (Fig. 13.2). Crucially β-MgH2 is known to transform into the metastable γ-phase of

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

361

c a

b

13.2 Crystal structure of β-MgH2.

MgH2 (San-Martin and Manchester, 1987) when subjected to high compressive stress (7–8 GPa) which is of relevance to ball mill preparations; see Section 13.4. γ-MgH2 has an orthorhombic structure analogous to α-PbO2, with space group Pbcn and lattice parameters a = 0.453 nm, b = 0.544 nm, and c = 0.493 nm (Semenenko et al., 1978; Bastide et al., 1980). The packing and coordination of the Mg and H atoms are not affected by the transformation, but the H octahedra surrounding each Mg atom are deformed, as the straight octahedral chains take on a zigzag form. The saline or salt-like hydrides include the binary hydrides of all the alkali metals and the alkaline earth metals from calcium through to barium (Libowitz, 1965). The crystal lattices of all these hydrides tend to have metal cations and hydrogen anions. Therefore the alkali metals and the alkaline earth metals have similar properties of hardness, brittleness, optical properties and crystal structure as their corresponding halides. The hydrolysis of these ionic hydrides yields hydrogen at the anode. Magnesium hydride is often included in the saline hydrides owing to the similarities of many of its properties. However it shows the least electronegative difference, with the degree of ionic character of the bond between the metal and hydrogen decreasing from Ce to Li in the alkali metal series and from Ba to Ca in the alkaline earth series. Magnesium hydride, although classified as a saline hydride, is therefore the least ionic and is generally accepted to have a mixture of ionic and covalent bonding. The Mg and H atoms are ionised inequivalently with magnesium at Mg1.91+ while hydrogen is weakly ionised

© 2008, Woodhead Publishing Limited

362

Solid-state hydrogen storage

H–0.26 (Noritake et al., 2002). Magnesium hydride is the only saline hydride that has a density less than its metal.

13.3

Thermodynamics and hydride mechanisms

The formation of magnesium hydride by Mg(s) + H2(g) → MgH2(s) is an exothermic reaction and reliable thermodynamic properties for the formation of magnesium hydride have been obtained calorimetrically since the middle of the 20th century: for example ∆Hf = –74.5 kJ mol–1 H2, ∆Sf = –135 J K–1 mol–1 H2 (Stampfer et al., 1960) is comparable to a recent study by Bogdanovic et al. (1999) in which ∆H f = –72.9 kJ mol –1 H 2 , ∆Sf = –132.2 J K–1 mol–1 H2. Equilibrium enthalpy and entropy values of the hydriding reaction can also be calculated from the experimental equilibrium isotherms and with care can give reasonable estimates of ∆Hf and ∆Sf such as –70 kJ mol–1 H2 and –126 J K–1 mol–1 H2 (±1%) respectively (Vigeholm et al., 1983). The value of ∆Hf for magnesium hydride is approximately half that of the other alkaline earth metals, which is consistent with its containing covalent bonds and having an intermediate character between ionic and covalent bonding as stated above. However, the magnitude of the enthalpy is still relatively high at –74.5 kJ mol–1 H2, resulting in a temperature of 551 K for a 1 bar plateau pressure which is still too high compared with the ideal –40 kJ mol–1 H2, needed for a corresponding temperature of only 300 K. Note this can be readily calculated from the van’t Hoff equation assuming that most binary hydride entropy changes during decomposition is due to the evolution of the gas such that ∆S = –130.7 J K–1 mol–1 H2 (Grochala and Edwards, 2004). As stated in Section 13.2, maintaining a magnesium hydride store at 550–600 K is inefficient and there are significant technological implications such as heat management during hydriding and dehydriding. Thus a desirable range of bond strengths between 10 and 60 kJ mol–1 H2 is often targeted by researchers (Bérubé et al., 2007), arguing physisorption is too weak below 10 kJ mol–1 H2 while chemisorption is too high above 60 kJ mol–1 H2. Therefore, to conclude this section, the high heat of formation of magnesium hydride provides a major drawback to using the material for a hydrogen store owing to the high operating temperatures required. However, another major problem is the slow hydriding and dehydriding kinetics. The absorption mechanism of MgH2 at 600 K is said to initiate through the hydride forming at the Mg grain surface, and that its rate of formation is controlled by the density of nucleation sites (Belkbir et al., 1981). The reaction would then stop when a hydride layer covered the surface of the magnesium essentially preventing fast diffusion to the core of the particle, slowing down the kinetics of the β phase transition considerably (Bérubé et al., 2007). Therefore nano-

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

363

sized Mg particles are required for a practical store material. In decomposition, the dehydrogenation process involves the initial formation of α-Mg at the surface of a particle, followed by the formation and growth of Mg nuclei. Finally, the Mg phase is formed, accompanied by a contraction in particle volume and the formation of cracks, aiding diffusion through the particle (Bohmhammel et al., 1998). This leads to slow kinetics even at elevated temperatures. A number of approaches have been adopted to speed up the kinetics. A list of modified magnesium sorption properties has been tabulated by Sakintuna et al. (2007), most of which involve ball milling, which is the topic of the next section, followed by utilising metal/metal alloys catalysts (Section 13.5) or metal oxide catalysts (Section 13.6). Further models relating to the kinetic mechanisms in all these systems are explored in Section 13.7.

13.4

Ball milling to improve hydrogen sorption behaviour

One of the most common approaches to improve the hydrogen sorption behaviour of magnesium hydride is to use ball milling. Ball mill types range from vibratory or shaker-style mill to planetary-style mill. The different operation and design of these mills influence the impact energies and temperatures subjected to the mill material (Rajamani et al., 2000; Takacs and McHenry, 2006). For example, local ball temperatures have been found to be below 400 K in shaker-type ball mills, while temperatures above 500 K have been reported for planetary-style mills. In addition, different milling vial and ball materials and sizes, as well as ball to powder weight ratios, all have an effect on the particle size, degree of agglomeration, presence of contaminants and structure of the resulting powders. Initially one would think that the approach is quite simple. For example, untreated pure magnesium hydride, with an average particle size of tens of micrometres, typically desorbs in an hour at 623 K. By ball milling, the particle size and agglomeration of particles are reduced and the surface roughness increases. This leads to an increased surface area, such as from 1 to 10 m2 g–1, the introduction of defects, increased nucleation sites and a reduction in the diffusion path length for hydrogen leaving the hydride. Such an approach can therefore reduce the desorption time significantly, for example 10 min at 623 K (Huot et al., 1999). However the exact mechanism of H2 desorption from ball milled MgH2 has many possibilities. For example the rate-limiting step in the desorption of hydrogen from MgH2 has been proposed to be either the diffusion pathways of hydrogen through the magnesium grains formed on the surface of the particle (Selvam et al., 1986), or the recombination of H2 molecules on the particle surface prior to desorption (Barkhordarian et al., 2006). In both cases, a reduction in particle size resulting from ball milling would be expected to enhance the rate of hydrogen desorption,

© 2008, Woodhead Publishing Limited

364

Solid-state hydrogen storage

either by reducing diffusion path lengths for H2, or by increased surface area available for recombination of H2 molecules. Ball milling magnesium hydride does more than reduce the particle size and increase the surface area, defects and internal stress. The milled hydride often reveals a phase change from β-MgH2 to γ-MgH2 due to the stress; see Section 13.2. In the example above, Huot et al. showed that after only 2 h of ball milling the milled sample contained 74 wt% β-MgH2, 18 wt% γ-MgH2, and 8 wt% MgO. Whether fully or partially transformed to γ-MgH2 the metastable crystal structure usually recovers to β-MgH2 in the first cycle depending on the cycling parameters. To illustrate this conversion from γMgH2 to β-MgH2, Fig. 13.3 shows in situ X-ray diffraction (XRD) data of the dehydrogenation of ball milled MgH2 during heating from room temperature to 873 K in 1 bar of He. Note the Mg metal starts to form above 600 K and the metastable γ-phase is still present until the transformation from hydride to metal is complete at c. 800 K. The β-MgH2 crystallite size increases from 15 to 53 nm at 723 K before transforming fully to Mg. Ball milling can also reduce the magnesium hydride decomposition temperature as well as desorption time. A good illustration of this is provided by Varin et al. (2006) who analysed samples of MgH2 milled for time periods

Intensity (a.u.)

Mg

β-MgH2 Mg

Mg β-MgH2

28

(°C

re

ra

300

pe

26

30

32 2θ

34

Te m

24

γ-MgH2

tu

400

γ-MgH2

200 36

38

13.3 A typical in situ series of X-ray diffractogram (XRD) data for milled MgH2 heated from 150 °C to 600 °C at 10 °C min–1 under He. A scan from 24 to 38 in 2θ was taken every 50 °C, (Croston, 2007).

© 2008, Woodhead Publishing Limited

)

600 500

Magnesium hydride for hydrogen storage

365

from 0.25 to 100 h in a planetary-style mill and measured the decomposition onset temperature using a differential scanning calorimeter (DSC) compared with the particle size measured by scanning electron microscopy (SEM). They found a near linear reduction of the decomposition onset temperature from c. 650 K to 610 K by reducing particle sizes from 2000 to 500 nm. Extrapolation of the curve would indicate another possible 10 K reduction but this is still c. 50 K above the equilibrium temperature predicted from the thermodymanics. It can be argued that utilising the γ-MgH2 to destabilise the remaining phase β-MgH2 would allow the desorption onset temperature to be lowered further (Varin et al., 2006). However, this would apply only to the first dehydrogenation cycle and therefore has little practical benefit for a cyclable hydrogen store. However, theoretically there is a critical threshold of particle size below which the desorption onset temperature decreases more rapidly, predicting that reducing MgH2 particle size to the nanometre scale could achieve hydrogen desorption at temperatures as low as 400 K (Varin et al., 2006). In addition, modelling using ab initio and density functional theory (DFT) calculations has shown the destabilisation temperature of MgH2 reduces with decreasing cluster size, and that a crystallite size of 0.9 nm could result in a desorption temperature of below 500 K (Wagemans et al., 2005). To date, however, ball milling has achieved only a lowering of the MgH2 desorption temperature to c. 610 K above the thermodynamic equilibrium limit of the bulk material. Milling contamination has led to further developments in ball milling. MgO is readily formed on the surface of magnesium and since ball milling produces new surfaces and typically a factor 10 increase in surface area then, not unsurprisingly, a significant amount, 2–10 wt%, of MgO is often observed in ball milled samples (Fournier et al., 2002; Varin et al., 2006) due to oxygen contamination and pickup in the mill. This is an obvious concern as it affects the potential hydrogen store efficiency. One recent approach is to enhance H2 dissociation close to the Mg/MgH2 surface, and to reduce the detrimental MgO oxide formation on the surface of Mg particles (Shang and Guo, 2004) by milling Mg or MgH2 with graphite (Huot et al., 2003b; Imamura et al., 2003; Shang and Guo, 2004) and multiwalled carbon nanotubes (MWNT) (Chen et al., 2004) in argon or a hydrogen atmosphere. The results showed the absorption and desorption rates on the first cycle were significantly faster at 13 min than for a commercial MgH2 sample, >1 h. However the effect did not continue for subsequent cycles, suggesting that the graphite acts only to enhance the initial activation of the Mg (Huot et al., 2003b) by preventing the formation of the detrimental oxide layer at the Mg surface. The above example introduces a derivative of ball milling called reactive mechanical alloying (RMA). RMA in a hydrogen atmosphere (Bobet et al., 2003a; Gennari et al., 2001; Song et al., 2004; Terzieva et al., 1995; Tessier and Akiba, 1999) rather than an inert Ar atmosphere has some merit. In

© 2008, Woodhead Publishing Limited

366

Solid-state hydrogen storage

RMA, particles of Mg are ball milled in an atmosphere of H2 gas, resulting in a mixture of γ and β-MgH2 of small particle size. This appears to reduce the formation of the detrimental oxide layer, due to the accelerated hydride formation over the surface of the metal, though care must be taken comparing different mills and their different vessel leak rates. In addition RMA has beneficial effects when used with additives. For example, Bobet et al. (2001) studied RMA of magnesium with 10 wt% Co, Ni, or Fe, in an attempt to identify the effects of RMA on the hydriding properties. The embrittlement of the Mg caused by the hydrogen gas during milling aided a reduction in particle size, while the elemental additions increased the amount of MgH2 formed. Co and Fe were found to be particularly effective at catalysing hydride formation, possibly due to the lower stability of Mg–Co and Mg–Fe phases compared with more stable Mg–Ni compounds. This use of additives is a method adopted often in combination with ball milling and is the subject of the next sections on metal and alloy additives (13.5), and metal oxide catalysts (13.6), to magnesium hydride.

13.5

Metal and alloy additives

As stated above, milling produces nanoscaled materials with reduced diffusion distances to improve the kinetics and increase surface area, which favours dissociation of hydrogen atoms by offering a large number of dissociation sites and allowing fast gaseous diffusion to the centre of the particle. By combining this approach with a catalyst, a significant improvement in the kinetics can be made by allowing electron transfer and hydrogen dissociation through a mechanism called spillover (Mitchell et al., 2003) in which hydrogen molecules dissociate on the catalyst surface. Hydrogen atoms either remain on the catalyst or diffuse to the catalyst support or native surface oxide layer aiding diffusion into the magnesium metal. Other advantages of additives are improved resistance to contaminants such as oxygen by allowing this reduced activation barrier route though an oxide layer. Combining a catalyst with ball milling reduces the quantity of catalyst required as the catalyst is dispersed at the nano-scale level throughout the structure and may remove the need for an activation step. A typical example of the ball milled and catalyst approach is Pd with MgH2 (Hjort et al., 1996; Yoshimura et al., 2004; Zaluska et al., 1999; Zaluski et al., 1997; von Zeppelin et al., 2002). For example additions of up to 1 wt% Pd nanoparticles were added to the MgH2 powder during ball milling in a shaker-style ball mill, resulting in dispersed nanoparticles of Pd on the MgH2 surface (Zaluska et al., 1999). Hydrogen absorption and desorption rates were measured from 500 to 600 K, with an absorption pressure of 10 bar H2, and a desorption pressure of 1 bar H2. Rates were significantly improved by the presence of the Pd, with full absorption of hydrogen

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

367

(c. 6 wt%) occurring in less than 40 min, compared with 120 min for ball milled MgH2 of a similar particle size of 30 nm with no added Pd. The strong dissociative effect of Pd on H2 at the Mg surface overcame the detrimental effect of surface oxidation, and eliminated the need for an activation process. Other metal and alloy additives studied include transition metals such as Ni (Hanada et al., 2005), Co (Hanada et al., 2005), Ti (Charbonnier et al., 2004; Liang et al., 1999a), Fe (Hanada et al., 2005), V (Charbonnier et al., 2004; Dehouche et al., 2000; Liang et al., 1999b), and Nb (Huot et al., 2002, 2003a; Pelletier et al., 2001), as well as intermetallics such as LaNi5 and FeTi (Liang et al., 1998). Of these, vanadium and titanium have shown promise. For example Liang et al. (1996b), formed a nanocomposite of βMgH2 + γ-MgH2 + VH0.81 by milling together MgH2 powder with 5 at% (c. 9.7 wt%) vanadium. Sorption kinetics were studied after 20 h of milling and complete desorption of hydrogen occurred in c. 3 min at 573 K, as opposed to c. 30 min for just 2 wt% desorption from pure MgH2. This was accompanied with a reduction in activation energy for the desorption process from 120 to 64 kJ mol–1; and a re-absorption of 2 wt% H2 in 17 min, at room temperature. Ti metal has been shown to behave in a similar way to V, with 5 min reported for complete hydrogen desorption from MgH2 at 573 K for a sample of MgH2 ball milled with 9 wt% Ti (Liang et al., 1999a). Nanocomposite hydrogen storage materials have also been formed by milling magnesium hydride with intermetallics of Mg2Ni, FeTi, LaNi5 (Liang et al., 1998), PdFe (von Zeppelin et al., 2002) or MmNi5–x (CoAlMn)x (Zhu et al., 2002). The intermetallic phases, with their faster hydriding kinetics, are used to catalyse the hydriding of the magnesium. It is important to maintain a small particle size, optimising the ratio of the components to avoid losing too much wt% efficiency, and to choose an intermetallic with a high dissociation efficiency for H2. Effectively spillover occurs because of the fast adsorption of hydrogen at or near the magnesium surface by the intermetallic phase flooding the metal with hydrogen atoms, reducing the activation energy by diminishing the effect of the oxide barrier at the metal surface. Liang et al. found that a 1:1 weight ratio mixture of LaNi5 and Mg formed a Mg + LaHx + Mg2Ni composite on hydrogen cycling at 573 K. The resulting composite was found to release 2.5 wt% H2 in 8 min at room temperature and at a pressure of 15 bar H2. A similar approach has been adopted with RMA with additives such as Co (Bobet et al., 2001, 2003a), V (Huot et al., 2003b), Ge (Gennari et al., 2002), Ni (Bobet et al., 2001; Tessier and Akiba, 2000), Nb (de Castro et al., 2004b), and Fe (Bobet et al., 2001; Huot et al., 2003b). Of these, a mixture of Mg + 5 at% V + graphite was hydrided after an hour of milling at 300 °C, 4 bar H2 (Huot et al., 2003b). Huot et al. also demonstrated that RMA for 5 h at 523 K with graphite and no catalyst was also very effective. For example

© 2008, Woodhead Publishing Limited

368

Solid-state hydrogen storage

desorption into 7 kPa at 623 K took only 12 min while a milled sample with no graphite took 80 min under the same conditions.

13.6

Metal oxide catalysts

Metal oxide catalysts are both a catalyst and a very effective milling aid that can create defects in the magnesium hydride structure due to their ceramic nature and ability to be produced to less than 10 nm in size. Owing to the potential loss of capacity, much work is focused on low wt% additions which are dispersed as effectively as possible. Typical approaches have been made with commercially available oxides which are often micrometre sized powders. Examples include Sc2O3 (Oelerich et al., 2001), TiO2 (Oelerich et al., 2001; Wang et al., 2000), V2O5 (Oelerich et al., 2001), Nb2O5 (Barkhordarian et al., 2003, 2004; Fatay et al., 2005; Friedrichs et al., 2006), Cr2O3 (Bobet et al., 2003b; Dehouche et al., 2002; Oelerich et al., 2001), Mn2O3 (Oelerich et al., 2001), Fe2O3 (Song et al., 2006), CuO (Oelerich et al., 2001), CoO (Song et al., 2004), Al2O3 (Oelerich et al., 2001) and SiO2 (Oelerich et al., 2001), via ball milling or RMA. Oelerich et al. (2001) found that additions of only 0.05 wt% oxide to MgH2 powder, followed by milling for 120 h resulted in no phase changes, but did notably enhance the adsorption and desorption kinetics at 573 K. The most effective oxides appear to be those that have metals that are multivalent, which tend to enhance the kinetics, and those that had only a single oxidation state had no effect. This is exemplified in the work of Croston in an investigative study of a range of binary and ternary metal oxides of aluminium, silicon, titanium and zirconium, as well as Pd-modified TiO2 samples. The prepared oxides had grain sizes from 5 to 250 nm and were ball milled with MgH2 for 20 h. Figure 13.4 shows the variation in desorption onset temperature with different combinations of metal oxide catalysts from this study (Croston, 2007). The oxide structure may also play a role in providing additional sites for hydrogen dissociation. For example the RMA of Mg with amorphous Cr2O3 obtained from a supercritical fluid method offered improvements in hydrogen absorption measured at 573 K, with desorption of more than 50% of the hydrogen occurring in 30 min, compared with 45 min for the mixture with crystalline Cr2O3. (Bobet et al., 2003b). In ball milled MgH2–0.2 mol% Cr2O3 after 1000 absorption–desorption cycles at 573 K, desorption rates decreased by a factor of four. These effects were attributed to the microstructural coarsening of the MgH2 from 21 to 84 nm (Dehouche et al., 2002), implying fewer defects in the crystal structure while the storage capacity increased from 5.9 wt to 6.4 wt%. Another argument relating to the benefit of oxide additives is that the presence of brittle oxides, when combined with ball milling, simply causes structural refinement to help break down the hydride into smaller particles.

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

369

Güvendiren et al. (2004) studied the effects of Al2O3 additives on the average particulate size of MgH2. After milling MgH2 with 5 wt% Al2O3 for 3 h under argon, the average crystallite size of the β-MgH2 phase had decreased from 31.2 to 20.4 nm. The most successful metal oxide additive reported is Nb2O5 (Barkhordarian et al. 2003, 2004, 2006; Fatay et al., 2005, Friedrichs et al., 2006). Additive loadings ranging from 0.4 to 17 wt% Nb2O5 have been examined. Full desorption from a MgH2 – 4 wt% Nb2O5 mixture occurred in 1.5 min at 573 K (Barkhordarian et al., 2006) although the mixture had to be milled for 100 h to achieve this. Milling for only 20 h doubled the desorption time while milling for only 2 h resulted in an order of magnitude increase in desorption time at c. 15 min. Nb2O5 is acting as a catalyst for chemisorption at the sample surface, resulting in a decrease in activation energy for hydrogen desorption from 120 to 62 kJ mol–1. One mechanism proposed by Oelerich is related to the high defect density introduced on the metal oxide particles due to the high energy ball milling. However the link to the multivalent possibilities of Nb, V and Ti nanostructured oxides and their distribution is more compelling. Neither Barkhordarian et al. nor Fatay et al., in similar studies, found evidence of any structural transformation of the Nb2O5 phase, both suggesting that the Nb2O5 acts as the catalyst for H2 desorption. However, Friedrichs et al. used X-ray photoelectron spectroscopy (XPS) analysis and showed a reduction of the Nb species after cycling, and evidence of a MgNb2O3.67 phase was observed by XRD after cycling of ball milled MgH2–Nb2O5. The data suggest that Nb2O5 is not the active catalyst in this case, and they propose a reaction mechanism involving niobium oxide species of lower oxidation state forming a network of ‘pathways’ for hydrogen diffusion within the MgH2 phase. Further support for the importance of variable oxidation state of the metal oxide is provided by titanium and vanadium oxide additive studies on MgH2. In a study of Mg–20 wt% rutile TiO2 prepared by RMA in a H2 atmosphere, it was found that complete H2 desorption occurred in 9 min at 350 °C in 0.1 MPa H2 (Wang et al., 2000). In the survey of a range of oxide materials by Oelerich et al. (2001), a 0.03 wt% TiO2 addition to ball milled MgH2 for 20 h resulted in a material which took only 7 min for complete desorption at 573 K in vacuum, while 0.06 wt% V2O5 addition under the same conditions showed complete desorption in 5 min which is also an oxide of a multivalent metal. Peshev et al. (1989) suggested that Ti of a lower oxidation state than +4 can be caused by the high local temperatures arising during milling. The Gibb’s free energies of oxidation of Ti, Nb and Mg are such that it is probable a reduction in the metal oxide catalyst by the magnesium that occurs during dehydrogenation at elevated temperatures, as observed in Nb2O5 (Friedrichs et al., 2006) and for TiO2 milled with MgH2 (Croston,

© 2008, Woodhead Publishing Limited

370

Solid-state hydrogen storage

2007). Both show similar reductions in activation energies for hydrogen desorption. Before getting over-excited about the importance of the electronic structure of the metal oxide catalyst it is worth noting the other mechanisms that may affect the kinetics of MgH2 such as the metal oxide catalysts acting as a milling aid, their high defect density, size and surface area effect, crystal structure and availability of sites for OH groups. It is therefore a complex system to which there is a temptation to oversimplify. Figure 13.4 helps to clarify some of the important mechanisms in this complex situation and assess their relative importance. For the active metal oxides the onset temperature is clearly reduced with reduced metal oxide particle size. This is likely to be due to the increase in catalyst surface area and dispersion of the nanocrystals within the matrix rather than a refinement of the MgH2 grain size due to improved milling aid. This is elegantly shown for 20 h milled MgH2 subsequently mixed for just 15 min with either micrometre sized or 15 nm sized Nb2O5 catalyst (Friedrichs et al., 2006). They report a factor 10 reduction in desorption times using the nano- rather than micrometre sized particles with mixing by milling for only 15 min; however, the nonmilled physically mixed control with nanoparticles reduced the desorption time by only c. 20%. Returning to Fig. 13.4 the y-axis shows titania particles after different calcination temperatures, providing comparison between amorphous titania particles dried at 70 °C (343 K) and calcined at 400 °C (673 K), in which case the TiO2 is primarily anatase, and calcined at 800 °C (1073 K), in which case the TiO2 is primarily rutile. Figure 13.4 shows the effectiveness of the anatase crystal structure of TiO2 compared to the rutile form in aiding the dehydrogenation of MgH2, and thus illustrates the dangers of looking at a single mechanism. We would consider the reduction mechanism of rutile and anatase TiO2 to be very similar, but we might expect that rutile is more readily reduced than anatase, owing to a weaker binding energy between Ti and O in the rutile phase (Rekoske and Barteau, 1997). This would suggest that the rutile TiO2 should in fact be more effective than the anatase in enhancing the dehydrogenation of MgH2. The reason this is not the case is due to the surface area of oxides. Rutile TiO2 may reduce more readily than anatase when measured per unit surface area, but the much lower surface area of the rutile samples means that on a similar mass basis, the anatase, having more exposed surface per unit mass, loses more oxygen and is thus more effective. Another complicating factor is the presence of physisorbed water and OH groups on the surface of the oxide additives affecting dehydrogenation of MgH2. Water reacts readily with MgH2 via the spontaneous reaction (Grosjean et al., 2006): MgH2 + 2H2O → Mg(OH)2 +2H2 ∆H = –277 kJ mol–1

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

371

Dehydrogenation onset temperature (°C) 180 10Al90Si70 10Al90Si120 10Al90Si400 10Al90Si800 50Al50Si70 50Al50Si120 50Al50Si400 50Al50Si800 90Al10Si70 90Al10Si120 90Al10Si400 90Al10Si800 Si70 Si120 Si400 Si800 Al70 Al120 Al400 Al800 Zr70 Zr120 Zr400 Zr800 10Ti90Si70 10Ti90Si120 10Ti90Si400 10Ti90Si800 50Ti50Si70 50Ti50Si120 50Ti50Si400 50Ti50Si800 90Ti10Si70 90Ti10Si120 90Ti10Si400 90Ti10Si800 10Ti90Al70 10Ti90Al120 10Ti90Al400 10Ti90Al800 50Ti50Al70 50Ti50Al120 50Ti50Al400 50Ti50Al800 90Ti10Al70 90Ti10Al120 90Ti10Al400 90Ti10Al800 Ti70 Ti120 Ti400 Ti800 Ti-5Pd Ti-1Pd 1Pd

230

280

330

344

254

297 305

279

308 313 308

295

261

347 346 330

325 321

340

277

288 283 293

268

325 307 337

291 310

205 205

0

361 360 360 367 362 367 365 359 359 368 372

325

298

263

367 362 363 354 369 366 367 358

293 292

273

257

380

358 360

200

400

600

800

1000

1200

1400

Oxide surface area (m2 g–1)

13.4 Data showing dehydrogenation onset temperature as determined by thermal gravitational analysis (TGA) and oxide surface area for MgH2 milled for 20 h with 20 wt% of various oxides (Croston, 2007). Nomenclature of y-axis labels refers to the wt% ratio of oxides followed by the calcination temperature of the oxide particle in °C. The Ti–5Pd and Ti–1Pd refer to 5 and 1 wt% Pd distributed on the 20 wt% titania. Unmarked row represents milled MgH2 with no additive. © 2008, Woodhead Publishing Limited

372

Solid-state hydrogen storage

This effect depends on the preparation and storage of the metal oxides. Therefore it is important to realise that more than one mechanism may be in operation with the metal oxides other than their catalytic behaviour, which leads us on to possible kinetic models in the following section.

13.7

Kinetic models of hydrogen absorption

A number of kinetic models are available to attempt to explain the potential mechanisms of hydrogen absorption and desorption from MgH2 (Barkhordarian et al., 2006; Mintz and Zeiri, 1995). The equations for the models are: SC1: α = kt 1

JMA3: [– ln (1 – α )] 3 = kt 1

CV3: 1 – [1 – α ] 3 = kt

1

JMA2: [– ln (1 – α )] 2 = kt 1

CV2: 1 – [1 – α ] 2 = kt

[ ]

2 CVD3: 1 – 2α – [1 – α ] 3 = kt 3

where α represents the transformed fraction, k is the reaction constant and t is the time. SC1 refers to a surface-controlled reaction where chemisorption of hydrogen is the rate-limiting step. JMA3 and JMA2 are Johnson–Mehl– Avrami models which represent models of nucleation and growth at random points in the bulk (3D nucleation) and surface (2D nucleation) in the material, illustrated in the schematic in Fig. 13.5. In the JMA models, the rate-limiting step is the constant velocity of the moving interface between the metal and the hydride; hydrogen diffusion is fast, and nucleation rate is not time dependent. CV3 (3D) and CV2 (2D) are equations for contracting volume models in which nucleation starts at the surface of the sample. This nucleation zone is thin compared with the particle size, and the kinetics are fast compared with the growth kinetics into the bulk. CV3 and CV2 also assume hydrogen diffusion is not rate-limiting. For hydrogen diffusion as a rate limiting step the equation is CVD3 (Barkhordarian et al., 2006). The two-dimension growth models CV2 and JMA2 assume that one dimension is kinetically restricted. These models have been applied to adsorption and desorption of MgH2 mixed with Nb2O5 and TiO2 catalysts (Croston, 2007; Friedrichs et al., 2006). With Nb2O5 the absorption behaviour follows CVD3 diffusion-controlled contracting volume model. However desorption with catalyst levels below 1 mol% (9 wt%) milled for 20 h showed surface control SC by chemisorption as the rate-limiting step or CV2 growth occurs at higher catalyst levels. The amount of catalyst required to switch to the CV2 model decreased with increasing milling time. The kinetics of metal catalysts, such as MgH2 ball mixed with 5 at% V, tend to follow the JMA2 model at low temperatures

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

373

13.5 Schematic of two different phase growth models. The transformed phases are the shaded areas. The left-hand side represents JMA3, which is three-dimensional growth of existing nuclei with constant interface velocity between the Mg and MgH2. The right-hand side represents CV2 contracting volume model of two-dimensional growth with constant interface velocity between Mg and MgH2.

with the rate-limiting step being nucleation and 2D growth below 523 K (low temperature and low plateau pressure leading to low driving force). However at higher temperatures the driving force is greater and the catalyst particles allow recombination sufficiently fast such that the rate-limiting step now is the motion of the Mg to MgH2 interface and hence the kinetics followed the CV2 model (Liang et al., 2000). Annealing the same MgH2 and catalyst samples can lead to diffusion-controlled or surface-controlled ratelimiting steps, depending on a low or high driving force. Milled MgH2 without catalyst over a wide range of pressures and temperatures tends to follow a JMA3 model with growth of nuclei with constant Mg to MgH2 interface velocity (Liang et al., 2000), although this was interpreted as a form of decreasing nucleation rate and 2D growth of nuclei JMA2. Using a 20 wt% TiO2 catalyst at 573 K also follows a JMA2 model (Croston, 2007), unlike Nb2O5 catalyst. The desorption of hydrogen from this sample therefore appeared to nucleate at points on the surface and within the bulk of the hydride, with kinetics controlled by the two-dimensional growth of Mg nuclei along the grain boundaries. These results suggest that the presence of the active Ti species provided active sites within the grain boundaries for the

© 2008, Woodhead Publishing Limited

374

Solid-state hydrogen storage

nucleation of Mg in the same way as observed for nanodispersed Ni (Hanada et al., 2005) or similar to the V catalyst system for low driving force, thus following the JMA model and growth of nuclei with constant interface velocity. These examples illustrate that the desorption kinetics are related to the pressure, temperature and microstructure of the MgH2 system as well as the catalyst used. The microstructure varying with milling time, catalyst and any post-annealing. Thus the rate-limiting steps and activation energies are easily varied even with one system. For example, activation energies can be drastically reduced from c. 120 kJ mol –1 for the chemisorption of uncatalysed nanocrystalline MgH2, produced by milling, to values approaching 60 kJ mol–1 depending on catalyst and concentration (Liang et al., 2000).

13.8

Conclusions and future work

The thermodynamics of the hydriding and dehydriding of MgH2 requiring high temperatures for a dissociation pressure of 1 bar, and the slow kinetics of the system are the two main problems to the exploitation of a MgH2 hydrogen store. We have seen that, by utilising milling and catalysts, dramatic reduction in the kinetics can be achieved down to a few minutes which may be commercially acceptable. Unfortunately there is a commensurate loss in hydrogen storage capacity depending on which catalyst and how much is used. Useful kinetic models can help indicate whether chemisorption, nuclei growth or contracting volume mechanisms dominate. Future development will likely look at combinations of catalysts targeting these rate-limiting steps which vary depending on material structure and conditions of temperature and pressure. For example, Fig. 13.4 shows the beneficial effect of just 1 wt% Pd dispersed onto nano-particles of titania. This combination of catalysts resulted in an extremely effective method of increasing the kinetics and reducing the onset temperature. This was achieved by reducing the activation energy by over a half to 59 kJ mol–1 via a spillover mechanism in which the Pd aids dissociation of H2 while the titania provides a high surface area of dissociative sites. The importance of reduced oxide states as illustrated by the multivalent oxides and in particular Nb2O5, V2O5 and TiO2 will also continue to attract attention in this system into understanding the key mechanisms that affect the kinetics. Despite this progress on kinetics the thermodynamics on MgH2 remains a problem. One approach is to reduce the heat of formation by destabilising the system by using additives that form compounds and alloys in either or both the hydrogenated or dehydrogenated states. This not only puts the hydride closer to the optimum window of dissociation pressure and temperature for a practical device, (see Fig. 13.1), but the lower heat of formation reduces the thermal management demands on a hydrogen store. Such thermal management problems have significant impact on applicability and store

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

375

designs for different applications. The early work in this area dates back to the late 1960s (Reilly and Wiswall, 1967, 1968) on Cu and Ni additions. Mg2NiH4 is a good example of an alloy that is formed in both the hydrogenation and dehydrogenation states, thus it is only the stabilisation energy that contributes to lowering the dehydrogenation energy. This results in an enthalpy of hydride formation that is still too high at –65 kJ mol–1 H2, close to that of MgH2, and a dissociation pressure of 1 bar at 520 K. If, on the other hand, stabilisation is too weak, such as in the case of a MgAl alloy which forms on dehydrogenation and the Al reforms and segregates during hydrogenation (Zaluska et al., 2001), then the enthalpy of dehydrogenation will change only slightly. More recently elegant work by Vajo et al. (2004) with a Mg and Si mixture showed that Si forms a strong bond only in the dehydrogenation state, resulting in the dehydrogenation enthalpy reducing to 36 kJ mol–1 H2. Although this is encouraging this system has the drawback of reducing the hydrogen storage capacity to 5 wt% and unfortunately the kinetics of this system at low temperatures are too slow to rehydrogenate, equilibrium at 1 bar being 300 K, while increasing the temperature to above 430 K to improve the kinetics may require pressures over 100 bar. So what holds for the future of MgH2? Clearly destabilisation without losing capacity has led to a whole host of complex light metal hydride combinations which are the subject of Chapter 17. Of these systems considerable interest has come from MgH2 mixed with LiBH4 or other borohydrides. A stoichiometric 1:2 molar ratio has shown promise (Vajo et al., 2005) with added catalysts such as TiCl3 followed by more detailed studies (Bosenberg et al., 2007) following the two-step reaction on desorption MgH2 + 2LiBH4 ↔ Mg + 2LiBH4 + H2 ↔ MgB2 + 2LiH + 4H2 However, this reaction is kinetically restricted and proceeds only at elevated temperatures. However rehydrogenation of LiBH4 and MgH2 are found to occur simultaneously at lower temperatures of c. 450 K at 50 bar hydrogen pressure compared with LiBH4 on its own requiring 900 K and 350 bar to rehydrogenate from LiH. The mutual catalytic/synergistic effect in such systems is intriguing and other molar compositions such as 3.3:1 MgH2 to LiBH4 (Yu et al., 2006) have led to exploring the destabilising effect of Mg on LiBH4 (Mao et al., 2007) and the decomposition of MgH2 as the LiBH4 decomposes to LiH and the influence of LiMg alloys. The approach again is essentially to look at potential catalysts and additives to improve the kinetics and destabilisation, such as recent exciting developments using MgB 2 (Barkhordarian et al., 2007). To conclude it would seem that whether on its own, or part of a complex hydride system or even as a catalyst Mg still has a lot to give in developing future metal hydride stores.

© 2008, Woodhead Publishing Limited

376

13.9

Solid-state hydrogen storage

References

Barkhordarian, G., Klassen, T. and Bormann, R. (2003) Fast hydrogen sorption kinetics of nanocrystalline Mg using Nb2O5 as catalyst, Scripta Materialia, 49, 213–217. Barkhordarian, G., Klassen, T. and Bormann, R. (2004) Effect of Nb2O5 content on hydrogen reaction kinetics of Mg, Journal of Alloys and Compounds, 364, 242– 246. Barkhordarian, G., Klassen, T. and Bormann, R. (2006) Kinetic investigation of the effect of milling time on the hydrogen sorption reaction of magnesium catalyzed with different Nb2O5 contents, Journal of Alloys and Compounds, 407, 249–255. Barkhordarian, G., Klassen, T., Dornheim, M. and Bormann, R. (2007) Unexpected kinetic effect of MgB2 in reactive hydride composites containing complex borohydrides, Journal of Alloys and Compounds, 440, L18–L21. Bastide, J. P., Bonnetot, B., Letoffe, J. M. and Claudy, P. (1980) Polymorphism of magnesium hydride under high pressure, Materials Research Bulletin, 15, 1215–1224. Belkbir, L., Joly, E. and Gerard, N. (1981) Comparative-study of the formationdecomposition mechanisms and kinetics in LaNi5 and magnesium reversible hydrides, International Journal of Hydrogen Energy, 6, 285–294. Bérubé, B., Radtke, G., Dresselhaus, M. and Chen, G. (2007) Size effects on the hydrogen storage properties of nanostructured metal hydrides: a review, International Journal of Energy Research, 31, 637–663. Bobet, J. L., Akiba, E. and Darriet, B. (2001) Study of Mg–M (M = Co, Ni and Fe) mixture elaborated by reactive mechanical alloying: hydrogen sorption properties, International Journal of Hydrogen Energy, 26, 493–501. Bobet, J. L., Chevalier, B., Song, M. Y. and Darriet, B. (2003a) Improvements of hydrogen storage properties of Mg-based mixtures elaborated by reactive mechanical milling, Journal of Alloys and Compounds, 356, 570–574. Bobet, J. L., Desmoulins-Krawiec, S., Grigorova, E., Cansell, F. and Chevalier, B. (2003b) Addition of nanosized Cr2O3 to magnesium for improvement of the hydrogen sorption properties, Journal of Alloys and Compounds, 351, 217–221. Bogdanovic, B., Bohmhammel, K., Christ, B., Reiser, A., Schlichte, K., Vehlen, R. and Wolf, U. (1999) Thermodynamic investigation of the magnesium–hydrogen system, Journal of Alloys and Compounds, 282, 84–92. Bohmhammel, K., Christ, B. and Wolf, G. (1998) Kinetic investigations on the basis of isothermal DSC measurements of hydrogenation and dehydrogenation of magnesium hydride, Thermochimica Acta, 310, 167–171. Bosenberg, U., Doppiu, S., Mosegaard, L., Barkhordarian, G., Eigen, N., Borgschulte, A., Jensen, T. R., Cerenius, Y., Gutfleisch, O., Klassen, T., Dornheim, M. and Bormann, R. (2007) Hydrogen sorption properties of MgH2–LiBH4 composites, Acta Materialia, 55, 3951–3958. Bouaricha, S., Dodelet, J. P., Guay, D., Huot, J. and Schulz, R. (2001) Activation characteristics of graphite modified hydrogen absorbing materials, Journal of Alloys and Compounds, 325, 245–251. de Castro, J. F. R., Santos, S. F., Costa, A. L. M., Yavari, A. R., Botta F, W. J. and Ishikawa, T. T. (2004a) Structural characterization and dehydrogenation behavior of Mg-5at.%Nb nano-composite processed by reactive milling, Journal of Alloys and Compounds, 376, 251–256. de Castro, J. F. R., Santos, S. F., Costa, A. L. M., Yavari, A. R., Botta F., W. J. and Ishikawa, T. T. (2004b) Structural characterisation and dehydrogenation behaviour of

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

377

Mg-5at.%Nb nano-composite processed by reactive milling, Journal of Alloys and Compounds, 376, 251–256. Charbonnier, J., de Rango, P., Fruchart, D., Miraglia, S., Pontonnier, L., Rivoirard, S., Skryabina, N. and Vulliet, P. (2004) Hydrogenation of transition element additives (Ti, V) during ball milling of magnesium hydride, Journal of Alloys and Compounds, 383, 205–208. Chen, D., Chen, L., Liu, S., Ma, C. X., Chen, D. M. and Wang, L. B. (2004) Microstructure and hydrogen storage property of Mg/MWNTs composites, Journal of Alloys and Compounds, 372, 231–237. Croston, D. (2007) The effect of metal oxide additives on MgH2, University of Nottingham. Dehouche, Z., Djaozandry, R., Huot, J., Boily, S., Goyette, J., Bose, T. K. and Schulz, R. (2000) Influence of cycling on the thermodynamic and structure properties of nanocrystalline magnesium based hydride, Journal of Alloys and Compounds, 305, 264–271. Dehouche, Z., Klassen, T., Oelerich, W., Goyette, J., Bose, T. K. and Schulz, R. (2002) Cycling and thermal stability of nanostructured MgH2-Cr2O3 composite for hydrogen storage. Journal of Alloys and Compounds, 347, 319–323. Eigen, N., Keller, C., Dornheim, M., Klassen, T. and Bormann, R. (2007) Industrial production of light metal hydrides for hydrogen storage, Scripta Materialia, 56, 847– 851. Fatay, D., Revesz, A. and Spassov, T. (2005) Particle size and catalytic effect on the dehydriding of MgH2, Journal of Alloys and Compounds, 399, 237–241. Fournier, V., Marcus, P. and Olefjord, I. (2002) Oxidation of magnesium, Surface and Interface Analysis, 34, 494–497. Friedrichs, O., Sanchez-Lopez, J. C., Lopez-Cartes, C., Klassen, T., Bormann, R. and Fernandez, A. (2006) Nb2O5 ‘pathway effect’ on hydrogen sorption in Mg, Journal of Physical Chemistry B, 110, 7845–7850. Gennari, F. C., Castro, F. J. and Urretavizcaya, G. (2001) Hydrogen desorption behavior from magnesium hydrides synthesized by reactive mechanical alloying, Journal of Alloys and Compounds, 321, 46–53. Gennari, F. C., Castro, F. J., Urretavizcaya, G. and Meyer, G. (2002) Catalytic effect of Ge on hydrogen desorption from MgH2, Journal of Alloys and Compounds, 334, 277–284. Grochala, W. and Edwards, P. P. (2004) Thermal decomposition of the non-interstitial hydrides for the storage and production of hydrogen, Chemical Reviews, 104, 1283– 1315. Grosjean, M. H., Zidoune, M., Roue, L. and Huot, J. Y. (2006) Hydrogen production via hydrolysis reaction from ball-milled Mg-based materials, International Journal of Hydrogen Energy, 31, 109–119. Gutfleisch, O., Dal Toe, S., Herrich, M., Handstein, A. and Pratt, A. (2005) Hydrogen sorption properties of Mg-1 wt.% Ni-0.2 wt.% Pd prepared by reactive milling, Journal of Alloys and Compounds, 404–406, 413–416. Güvendiren, M., Bayboru, E. and Ozturk, T. (2004) Effects of additives on mechanical milling and hydrogenation of magnesium powders, International Journal of Hydrogen Energy, 29, 491–496. Hanada, N., Ichikawa, T. and Fujii, H. (2005) Catalytic effect of nanoparticle 3d-transition metals on hydrogen storage properties in magnesium hydride MgH2 prepared by mechanical milling, Journal of Physical Chemistry B, 109, 7188–7194. Hjort, P., Krozer, A. and Kasemo, B. (1996) Hydrogen sorption kinetics in partly oxidized Mg films, Journal of Alloys and Compounds, 237, 74–80.

© 2008, Woodhead Publishing Limited

378

Solid-state hydrogen storage

Huot, J., Akiba, E. and Takada, T. (1995) Mechanical alloying of Mg–Ni compounds under hydrogen and inert atmosphere, Journal of Alloys and Compounds, 231, 815– 819. Huot, J., Liang, G., Boily, S., van Neste, A. and Schulz, R. (1999) Structural study and hydrogen sorption kinetics of ball-milled magnesium hydride, Journal of Alloys and Compounds, 295, 495–500. Huot, J., Pelletier, J. F., Liang, G., Sutton, M. and Schulz, R. (2002) Structure of nanocomposite metal hydrides, Journal of Alloys and Compounds, 330, 727–731. Huot, J., Pelletier, J. F., Lurio, L. B., Sutton, M. and Schulz, R. (2003a) Investigation of dehydrogenation mechanism of MgH2-Nb nanocomposites, Journal of Alloys and Compounds, 348, 319–324. Huot, J., Tremblay, M. L. and Schulz, R. (2003b) Synthesis of nanocrystalline hydrogen storage materials, Journal of Alloys and Compounds, 356, 603–607. Imamura, H., Kusuhara, M., Minami, S., Matsumoto, M., Masanari, K., Sakata, Y., Itoh, K. and Fukunaga, T. (2003) Carbon nanocomposites synthesized by high-energy mechanical milling of graphite and magnesium for hydrogen storage, Acta Materialia, 51, 6407–6414. Jolibois, P. (1912) Study of pure magnesium and magnesium hydrides, Comptes Rendues Academie Science Paris, 155, 353–355. Liang, G., Boily, S., Huot, J., van Neste, A. and Schulz, R. (1998) Hydrogen absorption properties of a mechanically milled Mg–50wt%LaNi5 composite, Journal of Alloys and Compounds, 268, 302–307. Liang, G., Huot, J., Boily, S., van Neste, A. and Schulz, R. (1999a) Catalytic effect of transition metals on hydrogen sorption in nanocrystalline ball milled MgH2–Tm (Tm=Ti, V, Mn, Fe and Ni) systems, Journal of Alloys and Compounds, 292, 247–252. Liang, G., Huot, J., Boily, S., van Neste, A. and Schulz, R. (1999b) Hydrogen storage properties of the mechanically milled MgH2–V nanocomposite, Journal of Alloys and Compounds, 291, 295–299. Liang, G., Huot, J., Boily, S. and Schulz, R. (2000) Hydrogen desorption kinetics of a mechanically milled MgH2+5at.%V nanocomposite, Journal of Alloys and Compounds, 305, 239–245. Libowitz, G. G. (1965) The Solid-State Chemistry of Binary Metal Hydrides, New York, W.A. Benjamin, Inc. Mao, J. F., Wu, Z., Chen, T. J., Weng, B. C., Xu, N. X., Huang, T. S., Guo, Z. P., Liu, H. K., Grant, D. M., Walker, G. S. and Yu, X. B. (2007) Improved hydrogen storage of LiBH4 catalyzed magnesium, Journal of Physical Chemistry C, 111, 12495–12498. Mintz, M. H. and Zeiri, Y. (1995) Hydriding kinetics of powders, Journal of Alloys and Compounds, 216, 159–175. Mitchell, P. C. H., Ramirez-Cuesta, A. J., Parker, S. F., Tomkinson, J. and Thompsett, D. (2003) Hydrogen spillover on carbon-supported metal catalysts studied by inelastic neutron scattering. Surface vibrational states and hydrogen riding modes, Journal Physical Chemistry B, 107, 6838–6845. Noritake, T., Aoki, M., Towata, S., Seno, Y., Hirose, Y., Nishibori, E., Takata, M. and Sakata, M. (2002) Chemical bonding of hydrogen in MgH2, Applied Physics Letters, 81, 2008–2010. Oelerich, W., Klassen, T. and Bormann, R. (2001) Metal oxides as catalysts for improved hydrogen sorption in nanocrystalline Mg-based materials, Journal of Alloys and Compounds, 315, 237–242.

© 2008, Woodhead Publishing Limited

Magnesium hydride for hydrogen storage

379

Pelletier, J. F., Huot, J., Sutton, M., Schulz, R., Sandy, A. R., Lurio, L. B. and Mochrie, S. G. J. (2001) Hydrogen desorption mechanism in MgH2–Nb nanocomposites, Physical Review B, 6305, art. no.-052103. Peshev, P., Khrussanova, M., Chakarov, D., Terzieva, M. and Marinova, T. (1989) Surfacecomposition of Mg–TiO2 mixtures for hydrogen storage, prepared by different methods, Materials Research Bulletin, 24, 207–212. Rajamani, R. K., Songfack, P. and Mishra, B. K. (2000) Impact energy spectra of tumbling mills, Powder Technology, 108, 116–121. Reilly, J. J. and Wiswall, R. H. (1967) Reaction of hydrogen with alloys of magnesium and copper, Inorganic Chemistry, 6, 2220–2223. Reilly, J. J. and Wiswall, R. H. (1968) Reaction of hydrogen with alloys of magnesium and nickel and the formation of Mg2NiH4, Inorganic Chemistry, 7, 2254–2256. Rekoske, J. E. and Barteau, M. A. (1997) Isothermal reduction kinetics of titanium dioxide-based materials, Journal of Physical Chemistry B, 101, 1113–1124. Sakintuna, B., Lamari-Darkrim, F. and Hirscher, M. (2007) Metal hydride materials for solid hydrogen storage: a review, International Journal of Hydrogen Energy, 32, 1121–1140. Sandrock, G. (1999) A panoramic overview of hydrogen storage alloys from a gas reaction point of view, Journal of Alloys and Compounds, 295, 877–888. San-Martin, A. and Manchester, F. D. (1987) The H–Mg (hydrogen–magnesium) system, Bulletin of Alloy Phase Diagrams, 8, 431–437. Selvam, P., Viswanathan, B., Swamy, C. S. and Srinivasan, V. (1986) Magnesium and magnesium alloy hydrides, International Journal of Hydrogen Energy, 11, 169–192. Semenenko, K. N., Verbestkii, V. N., Kalashnikov, Y. A., Temofeeva, N. V. and Ioffe, M. I. (1978) Phase transitions of metal hydrides under condition of superhigh pressure, Vest. Mosk. Univ. Ser. 2, Khim., 19, 718–722. Shang, C. X. and Guo, Z. X. (2004) Effect of carbon on hydrogen desorption and absorption of mechanically milled MgH2, Journal of Power Sources, 129, 73–80. Song, M., Lee, D. and Kwon, I. (2004) Influences on the H-2-sorption properties of Mg of Co (with various sizes) and CoO addition by reactive grinding and their thermodynamic stabilities, Metals and Materials International, 10, 69–75. Song, M., Kwon, I., Kwon, S., Park, C., Park, H. and Bae, J. S. (2006) Preparation of hydrogen-storage alloy Mg–10 wt% Fe 2O 3 under various milling conditions, International Journal of Hydrogen Energy, 31, 43–47. Stampfer, J. F., Holley, C. E. and Suttle, J. F. (1960) The magnesium hydrogen system, Journal of the American Chemical Society, 82, 3504–3508. Takacs, L. and McHenry, J. S. (2006) Temperature of the milling balls in shaker and planetary mills, Journal of Materials Science, 41, 5246–5249. Terzieva, M., Khrussanova, M., Peshev, P. and Radev, D. (1995) Hydriding and dehydriding characteristics of mixtures with a high magnesium content obtained by sintering and mechanical alloying, International Journal of Hydrogen Energy, 20, 53–58. Tessier, P. and Akiba, E. (1999) Catalysed reactive milling, Journal of Alloys and Compounds, 295, 400–402. Tessier, P. and Akiba, E. (2000) Decomposition of nickel-doped magnesium hydride prepared by reactive mechanical alloying, Journal of Alloys and Compounds, 302, 215–217. Vajo, J. J., Mertens, F., Ahn, C. C., Bowman, R. C. and Fultz, B. (2004) Altering hydrogen storage properties by hydride destabilization through alloy formation: LiH and MgH2 destabilized with Si. Journal of Physical Chemistry B, 108, 13977–13983.

© 2008, Woodhead Publishing Limited

380

Solid-state hydrogen storage

Vajo, J. J., Skeith, S. L. and Mertens, F. (2005) Reversible storage of hydrogen in destabilized LiBH4, Journal of Physical Chemistry B, 109, 3719–3722. Varin, R. A., Czujko, T. and Wronski, Z. (2006) Particle size, grain size and gammaMgH2 effects on the desorption properties of nanocrystalline commercial magnesium hydride processed by controlled mechanical milling, Nanotechnology, 17, 3856– 3865. Vigeholm, B., Kjoller, J., Larsen, B. and Pedersen, A. S. (1983) Formation and decomposition of magnesium hydride, Journal of the Less-Common Metals, 89, 135–144. Wagemans, R. W. P., van Lenthe, J. H., de Jongh, P. E., van Dillen, A. J. and de Jong, K. P. (2005) Hydrogen storage in magnesium clusters: quantum chemical study, Journal of the American Chemical Society, 127, 16675–16680. Wang, P., Wang, A. M., Zhang, H. F., Ding, B. Z. and Hu, Z. Q. (2000) Hydrogenation characteristics of Mg–TiO2 (rutile) composite, Journal of Alloys and Compounds, 313, 218–223. Yoshimura, K., Yamada, Y. and Okada, M. (2004) Hydrogenation of Pd capped Mg thin films at room temperature, Surface Science, 566, 751–754. Yu, X. B., Grant, D. M. and Walker, G. S. (2006) A new dehydrogenation mechanism for reversible multicomponent borohydride systems – the role of Li–Mg alloys, Chemical Communications, 3906–3908. Zaluska, A., Zaluski, L. and Ström-Olsen, J. O. (1999) Nanocrystalline magnesium for hydrogen storage, Journal of Alloys and Compounds, 288, 217–225. Zaluska, A., Zaluski, L. and Ström-Olsen, J. O. (2001) Structure, catalysis and atomic reactions on the nano-scale: a systematic approach to metal hydrides for hydrogen storage, Applied Physics A: Materials Science and Processing, 72, 157–165. Zaluski, L., Zaluska, A. and Ström-Olsen, J. O. (1997) Nanocrystalline metal hydrides, Journal of Alloys and Compounds, 253, 70–79. Zaluski, L., Zaluska, A., Tessier, P., Ström-Olsen, J. O. and Schulz, R. (1995) Catalytic effect of Pd on hydrogen absorption in mechanically alloyed Mg2Ni, LaNi5 and FeTi, Journal of Alloys and Compounds, 217, 295–300. von Zeppelin, F., Reule, H. and Hirscher, M. (2002) Hydrogen desorption kinetics of nanostructured MgH2 composite materials, Journal of Alloys and Compounds, 330, 723–726. Zhu, M., Gao, Y., Che, X. Z., Yang, Y. Q. and Chung, C. (2002) Hydriding kinetics of nano-phase composite hydrogen storage alloys prepared by mechanical alloying of Mg and MmNi(5–x)(CoAlMn)(x), Journal of Alloys and Compounds, 330, 708–713.

© 2008, Woodhead Publishing Limited

14 Alanates as hydrogen storage materials C. J E N S E N, University of Hawaii at Manoa, Hawaii, Y. W A N G and M. Y. C H O U, Georgia Institute of Technology, USA

14.1

Introduction

Group I and II salts of [AlH4]– (alanates) have recently received considerable attention as potential hydrogen storage materials. These materials are currently referred to as ‘complex hydrides’, although alanates contain anionic metal complexes. These materials have high hydrogen gravimetric densities (Table 14.1) and are, in most cases, commercially available. Thus, a priori, they would seem to be viable candidates for application as practical, on-board hydrogen storage materials. Many of these complex hydrides have, in fact, been utilized in ‘one-pass’ hydrogen storage systems in which hydrogen is evolved from the hydride upon contact with water. However, the hydrolysis reactions are highly irreversible and could not serve as the basis for rechargeable hydrogen storage systems. The thermodynamics of the direct, reversible dehydrogenation of some complex hydrides lies within the limits that are Table 14.1 Basic material properties of alanates Material

Density (g/mol)

Density (g/cm3)

Hydrogen (w%)

Hydrogen (kg/m3)

Tm (°C)

∆H f0 (kJ/mol)

LiAlH4 NaAlH4 KAlH4 Mg(AlH4)2 Ca(AlH4)2

37.95 54.00 70.11 86.33 102.10

0.917 1.28 – – –

10.54 7.41 5.71 9.27 7.84

– – 53.2 72.3 70.4

190d 178 – – >230d

–119 –113 – – –

d

Decomposition temperature. Sources: SCI Finder Scholar ‘Hazardous Substances Data Bank’ data are provided by the National Library of Medicine (US); CRC Handbook of Chemistry and Physics, 83rd ed., 2002–2003; H. Jacobs, E. von Osten, Zeit. Naturfors. B-A J. Chem. Sci. 1976, 385; H. Jacobs, Zeit. Anorg. Allg. Chem. 1976, 427(1), 8; O.M. Løvvik, O. Swang, S.M. Opalka, J. Mater. Res. 2005, 20, 3199; O.M. Løvvik, P.N. Molin. AIP Conf. Proc. 2006, 837, 85.

381 © 2008, Woodhead Publishing Limited

382

Solid-state hydrogen storage

required for a practical, on-board hydrogen carrier. All of these materials are, however, plagued by high kinetic barriers to dehydrogenation and/or rehydrogenation in the solid state. Traditionally, it was thought that it would be impossible to reduce the barrier heights to an extent that would give reaction rates that even approached those required for vehicular applications. Thus, until recently, complex hydrides were not considered as candidates for application as rechargeable hydrogen carriers. This situation was changed by Bogdanović and Schwickardi. Their pioneering studies demonstrated that upon doping with selected titanium compounds, the dehydriding of anionic aluminum hydrides could be kinetically enhanced and rendered reversible under moderate conditions in the solid state.1 This breakthrough has led to a worldwide effort to develop doped alanates as practical hydrogen storage materials. Any effort to develop complex hydrides as practical hydrogen storage materials requires knowledge of their atomic structure and the thermodynamics of their fundamental dehydrogenation and rehydrogenation reaction chemistry. This chapter provides a summary of information and an overview of the progress that has been made towards the utilization of alanates as on-board hydrogen carriers. We have focused on materials with high practical potential and have excluded those with properties that clearly preclude practicality. One attractive feature of alanates is that lithium and sodium salts are readily available commercially. Magnesium alanate can be readily prepared with sodium alanate and magnesium hydride via a metathesis reaction.2 The mixed metal alanate, Na2LiAlH6, is prepared through ball milling of sodium hydride, lithium hydride, and sodium alanate.3 Potassium alanate can be prepared by the direct synthesis of potassium hydride and aluminum under high temperature and pressure.4

14.2

Atomic structure of alanates

14.2.1 NaAlH4 Lauher et al.5 determined the atomic structure of NaAlH4 through a singlecrystal X-ray diffraction study in 1979. Refinement of their data in space group I41/a showed the compound to consist of isolated [AlH4]– tetrahedra in which the Na atoms are surrounded by eight [AlH4]– tetrahedra in a distorted square antiprismatic geometry. Their results gave an Al–H bond length of 1.532(.07) Å. These findings were significantly shorter than the Al–H bond distances that were previously determined from a single crystal X-ray study of LiAlH 64 (average value of 1.548(.17) Å). Bel’skii et al.7 noted that this was inconsistent with the implications of the infrared spectra of the compounds as the Al–H stretching frequency of NaAlH4 is observed at a lower frequency than that of LiAlH4 (1680 and 1710 cm–1, respectively).

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

383

A second single crystal study generated data that converged to give an Al– H distance of 1.61(.04) Å7 that was in agreement with the IR data. However, it was noted that full resolution of the issue of the Al–H bond distance required neutron diffraction data as X-ray diffraction data tends to give erroneously short metal–hydrogen distances and very large uncertainties in the determination of hydrogen coordinates. The structure of NaAlD4 has been determined by powder neutron diffraction data8 at 8 and 295 K. Selected interatomic distances and bond angles are given in Table 14.2. The atomic structure was found to be made up of isolated [AlD4]– tetrahedra surrounded by sodium atoms (Fig. 14.1). The shortest Al–Al separations were 3.737(.01) and 3.779(.01) Å at 8 and 295 K, respectively. The two unique Na–D bond distances were nearly equal, 2.403(.02) and 2.405(.02) Å at 8 K and 2.431(.02) and 2.439(.02) Å at 295 K. The Al–D distances were found to be 1.626(.02) and 1.627(.02) Å at 8 and 295 K, respectively. Previously, X-ray data by Bel’skii et al.7 reported a shorter and much more uncertain Al–D distance of 1.61(.04) Å. Upon cooling from 295 K to 8 K, the Al–D distances showed no significant change. The two unique D–Al–D bond angles in the [AlD4]– tetrahedron were reported to be 107.32° and 113.86° at 295 K.

14.2.2 LiAlH4 The crystal structure of LiAlH4 was originally determined by Sklar and Post6 through an X-ray diffraction study. Hauback et al.9 carried out a more detailed atomic structure determination of LiAlD4 based on combined powder neutron and X-ray diffraction studies. The compound crystallized in the space group P21/c. The atomic structure was found to consist of isolated [AlD4]– tetrahedra surrounded by lithium atoms (Fig. 14.2). Selected interatomic distances and bond angles are given in Table 14.2. The minimum Al–Al distance between tetrahedra was 3.754(.01) Å at 295 K. The Al–D distances averaged 1.619(.07) Å at 295 K, which are longer than the distances ranging from 1.516 to 1.578 Å that were deduced from the X-ray structure determination.6 The D–Al–D angles of LiAlD4 were found to vary by less than 1.5° from the angles of a perfect tetrahedron. The Li–D distances ranged from 1.831(.06) to 1.978(.08) Å at 295 K and from 1.841(.09) to 1.978(.12) Å at 8 K.

14.2.3 KAlH4 The crystal structure of KAlD4 has been determined by Hauback et al.10 KAlD4 has a BaSO4-type structure with space group Pnma. The structure (Fig. 14.3) consists of isolated [AlD4]– tetrahedra in which potassium atoms are surrounded by seven of the tetrahedra (ten D atoms total). Selected interatomic distances and bond angles are given in Table 14.2. The average

© 2008, Woodhead Publishing Limited

384

Solid-state hydrogen storage

Table 14.2 Selected inter-atomic distances (Å) and angles (°) in various crystal structures at 8 and 295 K. Estimated standard deviations in parentheses Material/reference

Atoms

295 K

8K

NaAlD4 Hauback et al.8

Al-D (×4) Na-D (×4) (×4) Al-Na (×4) (×4) D-Al-D (×4) (×2)

1.626(.02) 2.431(.02) 2.439(.01) 3.544(.01) 3.779(.01) 107.32(1) 113.86(1)

1.627(.02) 2.403(.02) 2.405(.02) 3.521(.01) 3.737(.01) 107.30(1) 113.90(1)

LiAlD4 Hauback et al.9

Al-D1 Al-D2 Al-D3 Al-D4 Li-D1 Li-D2

D1-Al-D2 D1-Al-D3 D1-Al-D4 D2-Al-D3 D2-Al-D4 D3-Al-D4

1.605(.06) 1.633(.05) 1.623(.05) 1.603(.07) 1.920(.08) 1.936(.07) 1.978(.08) 1.831(.06) 1.909(.08) 3.214(.06) 3.234(.05) 3.260(.04) 3.328(.06) 3.415(.06) 110.1(3) 109.2(3) 110.8(3) 108.7(3) 107.1(3) 111.0(3)

1.625(.06) 1.621(.05) 1.645(.05) 1.596(.06) 1.896(.12) 1.932(.09) 1.978(.12) 1.841(.09) 1.870(.10) 3.200(.09) 3.232(.12) 3.265(.11) 3.285(.10) 3.401(.12) 109.0(3) 108.2(3) 111.0(3) 108.9(3) 108.4(4) 111.3(3)

KAlD4 Hauback et al.10

Al-D1 (1×) Al-D2 (1×) Al-D3 (2×) K-D1 (1×) K-D2 (1×) K-D2 (2×) K-D3 (2×) K-D3 (2×) K-D3 (2×) D3-Al-D3 D3-Al-D2 D3-Al-D1 D2-Al-D1

1.546(.13) 1.669(.13) 1.629(.08) 2.596(.14) 2.833(.15) 3.182(.06) 2.840(.10) 2.883(.11) 2.980(.12) 106.7(23) 106.5(18) 111.2(18) 114.56(72)

1.589(.07) 1.659(.07) 1.638(.04) 2.627(.07) 2.772(.08) 3.137(.03) 2.753(.06) 2.841(.06) 2.955(.06) 106.39(12) 107.39(10) 111.05(9) 113.3(4)

Na3AlD6 Rönnebro et al.12

Al-D1 Al-D2 Al-D3 Na1-D1 Na1-D2 Na1-D3 Na2-D1

1.746 1.758 1.770 2.268 2.226 2.261 2.267

Li-D3 Li-D4 Li-Al

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

385

Table 14.2 (Continued) Material/reference

Atoms Na2-D2

Na2-D3

D1-Al-D2 D1-Al-D3 D2-Al-D3 Li3AlD6 Brinks and Hauback13

Al1-D1 Al2-D2 Li-D1

Li-D2

D1-Al1-D1

D2-Al2-D2

295 K

8K

2.391 2.307 2.669 2.699 2.299 2.627 2.766 88.9 89.0 89.4 1.754(.04) 1.734(.04) 1.892(.10) 2.009(.12) 2.120(.12) 1.948(.08) 1.987(.09) 2.051(.14) 86.98(15) 93.02(15) 180.00(22) 87.36(13) 92.64(13) 180.00(22)

Na2LiAlD6 Brinks et al.15

Al-D Li-D D-Al-D

1.760(.03) 1.933(.03) 90.00(–), 180.00(–)

Mg(AlH4)2 Fossdal et al.11

Al-H1 Al-H2 Mg-H2 Mg-Al H1-Al-H2 H2-Al-H2

1.561(.12) 1.672(.04) 1.833(.07) 3.459(.01) 110.1(3) 108.79(15)

1.606(.10) 1.634(.04) 1.870(.06) 3.482(.04) 113.5(2) 105.24(13)

Al–D distance was 1.631 Å at 8 K and 1.618 Å at 295 K. The minimum Al– Al distance between the tetrahedra was 4.052 Å at 295 K. Also, D–Al–D bond angles were close to ideal and ranged from 106.4 to 113.3° at 8 K and 106.2 to 114.6° at 295 K. In addition, the minimum K–D distance was 2.596 Å at 295 K (larger than the Na–D distance of NaAlD4 and the Li–D distance of LiAlD4). It is noteworthy that variations in the crystal structures of MAlH4 compounds (M = Li, Na, K) arise from the differences in the size of the alkali cations of Li+, Na+, and K+, which result in coordination numbers of 5, 8, and 10, respectively.

© 2008, Woodhead Publishing Limited

386

Solid-state hydrogen storage

14.1 Crystal structure of NaAlD4. Complex anion, [AlD4]– tetrahedra, is linked via Na shown as black sphere. Reprinted with permission from Hauback et al.8 Copyright 2003, Elsevier.

14.2 Crystal structure of LiAlD4. Complex anion, [AlD4]– tetrahedra, is linked via Li shown as black spheres. Reprinted with permission from Hauback et al.9 Copyright 2002, Elsevier.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

387

14.3 Crystal structure of KAlD4. Complex anion, [AlD4]– tetrahedra, is linked via K shown as black sphere. Reprinted with permission from Hauback et al.10 Copyright 2005, Elsevier.

Al Mg H

b c (a) Al Mg H

b a (b)

14.4 Crystal structure of Mg(AlH4)2. Complex anion, [AlH4]– tetrahedra, is linked via Mg shown as black sphere. Views along (a) a-axis and (b) c-axis. Reprinted with permission from Fossdal et al.11 Copyright 2005, Elsevier.

© 2008, Woodhead Publishing Limited

388

Solid-state hydrogen storage

14.2.4 Mg(AlH4)2 The structure of magnesium alanate was determined by Fossdal et al.11 through a combination of X-ray and neutron diffraction. The space group was confirmed to be P3m1. The structure consists of a sheet-like arrangement consisting of [AlH4]– tetrahedra surrounded by six Mg atoms in a distorted MgH6 octahedral geometry (Fig. 14.4). Selected interatomic distances and bond angles are shown in Table 14.2. The Al–H distances ranged from 1.606(.10) to 1.634(.04) Å at 8 K and 156.1(.12) to 167.2(.04) Å at 295 K. These distances are in the same range as those found for lithium, sodium, and potassium alanate.

14.2.5 Na3AlH6 and Li3AlH6 In 2002, Rönnebro et al.12 explored the perovskite-related structure of Na3AlH6. These studies determined the positions of the hydrogen atoms from neutron diffraction data of a deuterated sample. It was found that the best fit data was for the monoclinic space group P21/n (no. 14). The structure was a distorted face-centered cubic (FCC) structure of [AlD6]3– units with sodium in all of the octahedral and tetrahedral sites. The complex anions were found to be distorted [AlH6]3– octahedra. Selected interatomic distances and bond angles are reported in Table 14.2. The Al–D distances were 1.746, 1.758, and 1.770 Å,

14.5 Crystal structure of Li3AlD6. Complex anion, [AlD6]– octahedra, linked via Li. Reprinted with permission from Brinks and Hauback.13 Copyright 2003, Elsevier.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

389

which are longer than the Al–D distances for NaAlD4 (1.626(.02) Å). The D–Al–D bond angles ranged from 88.9 to 89.4°. The Na–D distances ranged from 2.226 to 2.766 Å. The structure of Li3AlD6 was determined by Brinks and Hauback through X-ray and neutron diffraction.13 The space group was found to be R3 and the structure (Fig. 14.5) consisted of isolated [AlD6]3– octahedra connected by six-coordinated Li atoms (each Li atom is coordinated to two corners and two edges of the octahedra, six hydrogen atoms total). The structure is described as a distorted body-centered cubic (BCC) of [AlD6]3– units with half the tetrahedral sites filled with Li.14 Selected interatomic distances and bond angles are given in Table 14.2. The Al–D distances were 1.734 and 1.754 Å, which are comparable to the 1.758 Å (average), found for the sodium analog.12 The D–Al–D bond angles were in the range of 86.98 to 93.02°. The Li–D distances ranged from 1.892 to 2.120 Å and the shortest Al–Al distance was 4.757 Å. It is important to note that Li has a coordination number of only 6 in Li3AlD6, while the Na has coordination numbers of 6 and 8 in Na3AlH6.12 This is due to the differences in the size of the alkali metal ions.

14.6 Crystal structure of Na2LiAlD6 at 22 °C showing alternating AlD6 (dark) and LiD6 (light) octahedra in all directions with Na in interstitial 12-coordinated sites. Reprinted with permission from Brinks et al.15 Copyright 2005, Elsevier.

© 2008, Woodhead Publishing Limited

390

Solid-state hydrogen storage

14.2.6 Na2LiAlH6 The structure of Na2LiAlD6 (Fig. 14.6) was determined by Brinks et al.15 The structure was well defined in a cubic unit-cell with a space group of Fm 3 m. The compound had a perovskite-type structure best described as a distorted FCC of [AlD6]3– units with Li in the octahedral sites and Na in the tetrahedral sites. Selected interatomic distances are given in Table 14.2. The Al–D bond distances were found to be 1.760(.03) Å. The Al–D bond distances of the [AlD6]3– octahedra were significantly shorter in size than octahedra of Li3AlD6 (average of 1.774 Å)13 and longer than the octahedra of Na3AlD6 (1.756 Å).12 The AlD6 octahedra of Na2LiAlD6 are ideal having equal Al–D distances and ideal angles. The shortest Al–Al distance was 5.222 Å.

14.3

Dehydrogenation and rehydrogenation reactions in alanates

14.3.1 Undoped alanate Dehydrogenation reactions Alkali metal alanates undergo dehydrogenation in the 200–300 °C temperature range (Li, 201 °C; Na, 265 °C; K, 290 °C) to give aluminum metal and the corresponding alkali metal hydrides (see Eq. 14.1). NaAlH4 → NaH + Al + 3/2H2

[14.1]

Further dehydrogenation of the binary metal hydrides occurs only at temperatures in excess of 400 °C. In 1952, Garner and Haycock found that the dehydrogenation of LiAlH4 in the 100–150 °C temperature range results in elimination of only 50% of the hydrogen content of the hydride rather than the 75% observed at 200 °C. They concluded that dehydrogenation to LiH and aluminum must be a two-step process involving an intermediate ‘LiAlH2’ phase.16 The dehydrogenation of NaAlH4 was extensively studied by Ashby and Kobetz.17 They found that controlled heating at 210–220 °C for 3 h evolved 3.7 wt% hydrogen to give Na3AlH6. This work established that the first step of the dehydrogenation proceeds according to Eq. (14.2) and that further elimination of hydrogen to give aluminum and NaH occurs through a separate reaction (seen in Eq. 14.3) that takes place at ~250 °C. 3NaAlH4 → Na3AlH6 + 2Al + 3H2

[14.2]

Na3AlH6 → 3NaH + Al + 3/2H2

[14.3]

The occurrence of these two distinct reactions during the dehydrogenation process has been verified in subsequent studies of the thermolysis of LiAlH4, NaAlH4, KAlH4 by a variety of techniques including: differential thermal Xray diffraction (XRD),18 thermal gravimetric analysis (TGA),19,20 hydrogen

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

391

pressure–composition (P–C) isotherm (PCT) studies,21,22 and most recently through in situ XRD study.23 The thermal dehydrogenation of LiAlH4 was studied by Block and Gray in 1965 through differential scanning calorimetry (DSC).24 This study confirmed the multistep dehydrogenation pathway and provided the first detailed information about the thermodynamic parameters of this process. They found that LiAlH4 undergoes a phase transition at 160–177 °C before undergoing an initial dehydrogenation reaction to give Li3AlH6, as seen in Eq. (14.4) at 187–218 °C. This first dehydrogenation process was determined to be exothermic and with a ∆H of –10 kJ/mol H2. LiAlH4 → 1/3Li3AlH6 + 2/3Al + H2

[14.4]

Since the entropy of hydrogenation must be negative, this is a non-spontaneous process under all conditions. A second dehydrogenation reaction, seen in Eq. (14.5), was observed to occur at 228–282 °C. Li3AlH6 → 3LiH + Al + 3/2H2

[14.5]

This reaction was found to be endothermic with a ∆H of 25 kJ/mol H2. Finally, the dehydrogenation of LiH was observed in the 370–483 °C temperature range. The enthalpy of the dehydrogenation of LiH(s) has subsequently been determined to be 140 kJ/mol.25 The thermodynamics of the dehydrogenation of NaAlH4 to NaH and Al was first studied by Dymova and Bakum in 1969 through differential thermal XRD.18 This was followed by more detailed TGA and DSC studies by Claudy and co-workers.26,27 The latter study revealed that in addition to the chemical transformations seen in equations (14.1) and (14.2), the thermal decomposition of the hydride also involves the melting of NaAlH4 and the conversion of pseudo-cubic, α-Na3AlH6 to FCC, β-Na3AlH6. Consideration of the enthalpic values determined in these studies for: the Na3AlH6 phase change (1.8 kJ/ mol); the dehydrogenation of β-Na3AlH6 (13.8 kJ/mol), and the direct dehydrogenation of NaAlH4 to NaH and Al (56.5 kJ/mol), allows the calculation of a ∆H of 40.9 kJ/mol for the solid-state dehydrogenation of NaAlH4 to Na3AlH6. A similar value of ∆H for this process (36.0 kJ/mol) can be derived for the solid-state dehydrogenation by combining the heat of fusion of NaAlH4 (23.2 kJ/mol) determined by Claudy et al. with the ∆H for the dehydrogenation of liquid NaAlH4 to Na3AlH6 (12.8 kJ/mol).22 PCT studies of sodium aluminum hydride were first carried out by Dymova et al.22 The observation of two plateaus in these studies clearly confirmed the occurrence of two independent dehydrogenation reactions. At 210 °C, hydrogen plateau pressures of 15.4 and 2.1 MPa were found for the first and second reaction, respectively. The potential of solid NaAlH4 as practical on-board hydrogen storage material was obscured for many years because of the high kinetic barriers to the dehydrogenation reactions and their reverse. Also, pure sodium aluminum

© 2008, Woodhead Publishing Limited

392

Solid-state hydrogen storage

hydride melts at 183 °C. Since dehydrogenation occurs at higher temperatures, only the unacceptably high plateau pressures associated with the much lower ∆Hdehyd of liquid NaAlH4 could be observed. The initial dehydrogenation of KAlH4 to the hexa-hydride occurs at a much higher temperature (300 °C) than the analogous process for NaAlH4. Morioka et al.28 have pointed out that this might be the result of differing thermodynamic stabilities associated the parent alanates as the formation energies of KH and NaH are nearly equal (–56.3 and –56.3 kJ/mol, respectively) but those of NaAlH4 and KAlH4 are quite different (–155 and –183.7 kJ/mol, respectively). In contrast to the alkali metals, studies of the dehydrogenation of Mg(AlH4)2 by Fichtner et al.2 showed that at 163 °C, the dehydrogenation of the hydride proceeds according to Eq. (14.6) without the involvement of [AlH6]3– as an intermediate phase. Mg(AlH4)2 → MgH2 + 2Al + 3H2

[14.6]

The resulting MgH2 undergoes further dehydrogenation at 287 °C to produce magnesium metal that will react with aluminum at 400 °C to give Al3Mg. The dehydrogenation of Mg(AlH4)2 is apparently exothermic and thus its direct rehydrogenation precluded by thermodynamic constraints. Rehydrogenation reactions The microreverse of the dehydrogenation of NaAlH4 was first accomplished by Clasen29 who found that NaAlH4 can be obtained in quantitative yields from tetrahydrofuran (THF) solutions of sodium hydride and activated aluminum powder at 150 °C and 13.6 MPa of hydrogen pressure in the presence of a catalytic amount of triethylaluminum. The solvent-mediated, direct synthesis was improved to the point of commercial viability by Ashby et al.30 Later, Dymova et al.4 achieved the hydrogenation of the sodium system without solvent. Complete conversion to product was achieved at 270 °C under 17.5 MPa of hydrogen pressure in 2–3 h. The same group also reported the direct synthesis of KAlH4 under analogous conditions. However, it was recently discovered that KH/Al can be hydrogenated to KAlH4 under only 1.0 MPa of hydrogen in the 250–330 °C temperature range.28 Apparently, the greater thermodynamic stability of KAlH4, compared with NaAlH4 discussed above, makes the enthalpy of hydrogenation significantly more exothermic and thus lowers the pressure required for hydrogenation. Unlike sodium and potassium alanate, the enthalpy of hydrogenation of Li3AlH6 to LiAlH4 is endothermic by 9 kJ/mol H2. Since the entropy of hydrogenation must be negative, this is a non-spontaneous process under all conditions and thus direct hydrogenation cannot be achieved. However, the hydrogenation of LiH/Al to Li3AlH6 is weakly exothermic and thus could, in

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

393

principle, be accomplished under very high pressure or at moderate pressures while also subjecting the material to some other means of overcoming the unfavorable entropic driving force. Recently, Wang et al.31 reported evidence of the cyclic dehydrogenation and rehydrogenation between LiAlH4·4THF and Li3AlH6, LiH, and Al. The dehydrogenation of half-cycles evolved about 4.0 wt% hydrogen below 130 °C. The decomposed products were rehydrogenated in THF. As Clasen29 and Ashby et al.30 found for the sodium system, the formation of a THF adduct circumvents the unfavorable thermodynamics associated with the formation of the pure alanate. However, the presence of the organic adduct introduces the additional problems of: (1) a requisite side process to remove the adduct prior to dehydrogenation; and (2) hydrogen contamination that must be overcome if the process is to be utilized as a practical method. The direct rehydrogenation of adduct-free Mg(AlH4)2 is also apparently precluded by thermodynamic constraints.

14.3.2 Doped NaAlH4 Titanium compounds have long been known to catalyze the dehydrogenation of complex aluminum hydrides in solution. This effect was first reported by Wieberg et al. in 1951 who observed titanium catalyzed dehydrogenation of LiAlH4 in a diethyl ether suspension.32 Also, as discussed above, the commercial preparation of NaAlH4 is accomplished through the catalyzed reaction of sodium hydride and activated aluminum powder in THF suspension.30 The observation of a seemingly related catalytic effect of titanium for the dehydrogenation of Mg2Cl3AlH4 in solution,33 prompted Bogdanović and Schwickardi to begin a systematic study of the catalytic effects of titanium in the reversible dehydrogenation of complex aluminum hydrides that was extended to the solid state. In 1995 they disclosed34 that doping of NaAlH4 with a few mole percent of titanium significantly enhanced the kinetics of the dehydrogenation and rehydrogenation processes. Remarkably, NaAlH4 was found to undergo dehydrogenation in the solid state at 150 °C.1 The conditions required for rehydrogenation were also dramatically reduced to 5 h at 170 °C and 15.2 MPa. Thus, this discovery provided a means of removing the ‘insurmountable’ kinetic obstacle to harness the long-recognized favorable thermodynamics of the reversible dehydrogenation of NaAlH4 in the solid state. These findings suggested that sodium aluminum hydride could function as a rechargeable hydride under moderate conditions and might be developed for application as on-board hydrogen storage material. However, the hydrogen capacities of their materials were found to quickly diminish upon cycling. Following the initial, only 4.2 wt% of the 5.6 wt% that was eliminated could be restored under the moderate conditions employed in these studies. The hydrogen capacity was further diminished to 3.8 wt% after the second dehydriding cycle and reduced to only 3.1 wt% after 31 cycles.1 Additionally,

© 2008, Woodhead Publishing Limited

394

Solid-state hydrogen storage

the dehydrogenation/rehydrogenation kinetics of these materials were inadequate for practical application as an on-board hydrogen storage material. The original Bogdanović materials were prepared by evaporation of suspensions of NaAlH4 in diethyl ether solutions of the soluble titanium compounds, titanium tetra-n-butoxide, Ti(OBun)4, and beta-titanium trichloride, β-TiCl3. It was subsequently found that catalytic enhancement of NaAlH4 also occurs upon mechanically mixing the titanium catalyst precursors with the aluminum hydride host.35 The materials resulting from this method of preparation have kinetic and cycling properties that are much closer to those required for a practical hydrogen storage medium.35,36 The kinetic improvement is illustrated in Fig. 14.7, which compares the thermal programmed desorption (TPD) spectrum of undoped NaAlH4 with hydride that was doped by both methods. It can be seen that the onset of rapid kinetics for the first dehydrogenation process occurs at ~150 °C for hydride that is doped by the original method of Bogdanovic´ and at a much lower, ~120 °C, temperature for NaAlH4 that is doped by mechanically milling. The discontinuity that occurs after elimination of ~3.5 wt% hydrogen reflects the independence of the two dehydrogenation reactions. It is evident from Fig. 14.7 that the two titanium doped materials differ primarily in their effectiveness in catalyzing the first dehydrogenation reaction. A more detailed kinetic study was carried out in which the release of hydrogen from samples of the homogenized material was monitored at constant temperatures.38 As seen in Fig. 14.8, the material undergoes rapid dehydrogenation at 130 °C and proceeds at an appreciable rate even at 80 °C. 6 5

H/M (wt%)

4 3 2

Ti doped (mechanically milled) Ti doped (Bogdanovic´)

1

Undoped

0 0

50

100 150 200 Temperature (°C)

250

300

14.7 Thermal desorption (heating rate of 2 °C/min) from NaAlH4. Reprinted with permission from Jensen and Gross.37 Copyright 2001, Springer.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

395

2.0 180 °C

1.8 1.6

Pressure (atm)

1.4 1.2 1.0 0.8

130 °C

0.6 0.4

80 °C

0.2 0

0

5

10 Time (min)

15

20

14.8 Progression of dehydrogenation upon heating from NaAlH4 mechanically doped with 2 mol% Ti(OBun)4 over the course of time at constant temperatures. Reprinted with permission from Jensen and Gross.37 Copyright 2001, Springer.

Doping through mechanical milling of NaAlH4 with dopant precursors is not only a more effective means of charging the hydride with catalyst but also activates the material through reduction of the average particle size.35,39 It has, in fact, been demonstrated40 that ball milling alone improves the dehydrogenating kinetics of undoped NaAlH4. However, the introduction of catalysts enhances the kinetics far beyond those that are achieved by reducing the particle size. A number of studies have compared the effectiveness of different titanium dopant precursors.39,41–43 In addition to TiCl3, Ti(OBun)4, TiCl4, and TiBr4, TiF3 has been proposed by several groups to be a highly effective, and perhaps a superior, dopant precursor.44 Gross et al. have also shown that kinetic effects matching those achieved with the Ti(III) precursors are attained upon extended cycling of hydride doped with other titanium compounds that appear to be less effective dopant precursors during the initial cycles of dehydrogenation/rehydrogenation.45 Recently, it was found that even untreated, commercial available Ti powder can act as an effective dopant precursor upon prolonged mechanical milling.46 However, in order to obtain a material with kinetics and stable hydrogen cycling capacities that match those of hydride doped with Ti(III) or Ti(IV) precursors, it is necessary to mill the Ti powder with mixtures of NaH/Al rather than NaAlH4. Figure 14.9 compares

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage

2.5

1 3.5 2 3 4-6 3.0 7 8 2.5

2.0

2.0

1.5

1.5

3.5

}

H-amount desorbed (wt%)

3.0

1–10

1.0

NaH+Al+4mol%Ti

1.0

NaAlH4+4mol%Ti

0.5

0.5

0.0

0.0 0

2

4

6

8

10

12 0 2 Time (h)

4

6

8

10

H-amount desorbed (wt%)

396

12

14.9 Comparison of the cycling dehydrogenation profiles between NaH + Al + 4 mol% Ti and NaAlH4 + 4 mol% Ti. Both samples were prepared by mechanical milling under a H2 atmosphere for 10 h. Reprinted from Wang and Jensen.46 Copyright 2004, American Chemical Society.

the cycling dehydrogenation performance of materials that were both prepared through milling of the hydride with Ti powder under a H2 atmosphere for 10 h. No signs of degradation in either hydrogen capacity or dehydrogenation kinetics were seen for the material prepared from NaH/Al. Unfortunately, the potential for an improved hydrogen capacity upon elimination of the doping by-products resulting from the Ti(III) precursors was not realized as only 3.3 wt% hydrogen was obtained from dehydrogenation at 150 °C. The low hydrogen cycling capacity is, at least partly, due to the very slow kinetics of the second dehydrogenation step, which are inferior to those of materials prepared from the Ti(III) or Ti(IV) precursors. Outstanding kinetic enhancement of hydrogen cycling kinetics as well as hydrogen cycling capacities have been observed for NaAlH4 doped with Ti·0.5THF nanoparticles.47,48 However, the practical application of the nanoclusters is precluded by the difficulty and cost of their preparation. Additionally, the kinetic enhancement of the materials doped with the nanoparticles has been found to steadily diminish upon hydrogen cycling.48 Comparative studies of other metal halides as dopant precursors for treating NaAlH4 have shown that similar levels of kinetic enhancement of the reversible dehydrogenation can be achieved upon doping with chlorides of zirconium, vanadium, and several lanthanides.36,39 Lower levels of catalytic activity have been reported to occur in hydride that was charged with FeCl2 and

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

397

NiCl2,39 as well as with carbon.40 It was recently reported that ball milling the hydride with 2 mol% ScCl3, CeCl3, or PrCl3 gives rise to a material with rehydrogenation kinetics and storage capacities that are significantly greater than those of the Ti-doped hydride.49 In addition, the rehydrogenation of the scandium doped system was found to require lower pressures than the titanium doped material. Although these dopants were found to reduce rehydrogenation times by a factor of ~2 at 10.0 MPa and a factor of ~10 at 5.0 MPa, the high kinetic performance upon cycling at the lower pressure was found to persist only with the cerium doped material. Wang et al.50 studied Sc-doped NaAlH4 at higher dopant loadings and observed fast hydrogenation kinetics and minimal hydrogen capacity losses over cycling. It has been found that titanium and zirconium catalysts are quite compatible and can exert their differing catalytic roles in concert to optimize the dehydrogenation/re-hydrogenation behavior of NaAlH4.39 Bogdanovic´ et al.39 also found that catalytic activity can be optimized through charging the hydride with dual dopants. They discovered a synergistic effect in hydride that is doped with a combination of Ti(OBun)4 and Fe(OEt)2. The resulting materials exhibit dehydrogenation and rehydrogenation kinetics that are superior to those observed for the hydride that is charged with either of the dopants alone. Mechanistic studies In an effort to develop improved catalysts, researchers have recently been investigating a fundamental understanding of the chemical nature of the titanium dopants and their mechanism of action. This began with the determination of the chemical nature of the titanium species that are formed upon mechanical milling of NaAlH4 with the dopant precursors through synchrotron X-ray and neutron diffraction as well as transmission electron microscopy (TEM), scanning electron microscopy (SEM), and electron paramagnetic resonance (EPR) spectroscopy. In addition to kinetic studies, insight into the mechanism of action of the dopants was gained through studies of the destabilization of hydrogen in NaAlH4 by the dopants through infrared, nuclear magnetic resonance (NMR), and anelastic spectroscopy. Brinks et al.51 completed synchrotron X-ray and neutron diffraction studies of NaAlH4, Na3AlH6, and NaH/Al doped with 2 mol% of Ti additives. The results showed that directly after ball milling there were no signs of any Ticontaining phases. However, after several cycles of dehydrogenation and rehydrogenation, a shoulder appears on the high-angle side of the Al reflections (Fig. 14.10) that is interpreted as the FCC solid solution with the approximate composition of Al0.93Ti0.07. In the hope of elucidating the changes occurring in the structural environment of the hydrogen atom upon doping the hydride, neutron diffraction structure

© 2008, Woodhead Publishing Limited

398

Solid-state hydrogen storage Al

Intensity (arb units)

30000

20000 Al

10000 Al1–xTix

0 12.0

12.5

14.0

14.5

2θ(°)

14.10 Comparison of the Al peaks observed by synchrotron X-ray diffraction for NaAlH4 doped with 2 mol% TiCl3 before and after seven cycles of dehydrogenation/rehydrogenation.

determination of NaAlD4 was carried out by Hauback et al.8 In contrast to the X-ray structure determination of NaAlH4 (completed by Belskii et al.7 in 1983) in which only the atomic position of the Na and Al could be located, the atomic positions of the deuterium atoms can be reliably located by neutron diffraction. Owing to the inherent problem of incoherent neutron scattering by 1H, it was necessary to develop a method for the synthesis of high-purity NaAlD4. This was accomplished through the reaction of LiAlD4 with NaF in the presence of an aluminum alkyl catalyst. Neutron diffraction data was collected at the Institute for Energy Technology (Norway) from a sample of NaAlD4 that was prepared through this method. Final refinement of the data gave the structure of the hydride seen in Fig. 14.1. The deuterium atoms were well located giving a structure with two unique Al–D distances of 1.627(2) and 1.626(2) Å, and two unique D–Al–D angles of 107.30(1)° and 113.90(1)°. Unfortunately, no significant structural differences were found upon refinement of the neutron diffraction data that was collected from a sample of the doped hydride. Andrei et al.52 conducted TEM and SEM studies (both with energydispersive spectroscopic X-ray analysis) of NaAlH4 doped with 2 mol% TiF3. After samples of NaAlH4 doped with 2 mol% TiF3 had undergone 15 cycles of dehydrogenation/rehydrogenation, X-ray analysis showed that Ti had been incorporated into an Al containing phase. However, analysis of samples of the doped hydride immediately following ball-milling indicates that the

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

399

distribution of Ti is quite uneven. As seen in Fig. 14.11, a variety of methods, including selected-area diffraction and high-resolution imaging confirmed that most of the Ti was present as unreacted TiF3. Further confirmation and information about the Ti-dopant has been provided through studies of doped NaAlH4 by EPR spectroscopy conducted by Kuba et al.53 The EPR spectra were obtained for samples of Ti-doped NaAlH4 that were subjected to different numbers of cycles of dehydrogenation/ rehydrogenation. As seen in Fig. 14.12, Ti was observed to evolve from its initial Ti(III) state through a series of Ti(0) species during the first five cycles. Although the conversion of Ti(III) to Ti(0) occurs much more readily for TiCl3 doped samples than those prepared with TiF3, in both cases the evolution of Ti follows the same sequence that involves three distinguishable Ti(0) species and ends in the predominance of the same single Ti(0) species. The spectra of samples of NaAlH4 containing 2 mol% of cubic Al3Ti and Ti powder are distinctly different from any of those observed for the Ti(0) species that arise during the hydrogen cycling of the hydride. A relatively minor change was observed in the hydrogen cycling kinetics during the first 10 cycles of dehydrogenation/rehydrogenation. This finding, in conjunction with the electron microscopy studies, strongly suggests that the hydrogen cycling kinetics are unaffected whether a Ti(III) to Ti(0) species predominates. Thus, the profoundly enhanced hydrogen cycling kinetics of Ti-doped NaAlH4 are apparently due to a minority Ti species and the majority of the Ti is in a resting state. Wide-line solid-state 1H NMR spectra of a balled milled sample of undoped and doped NaAlH4 have also been obtained. As seen in Fig. 14.13, the spectra of both materials contain a broad and narrow component. The narrow feature has a chemical shift of c. 2 ppm and an unusually narrow linewidth of c. 1–2 kHz while the broad feature has a more typical linewidth of c. 40 kHz. Using inversion-recovery experiments, the T1 of the narrow and broad features were determined to be 32 ms and 25 s, respectively. The former is particularly striking, because this short T1 is atypical for protons in the solid state. The relative intensity of the narrow component was unchanged upon heating a ball-milled sample under dynamic vacuum at 100 °C for 24 h. Thus it would seem unlikely that the narrow feature is associated with residual organic solvent as this treatment would remove the majority of this type of impurity. Additionally, the infrared spectrum of this material was devoid of any absorptions in the C–H stretching region. Furthermore, the relative intensity of the narrow component increases as the hydride is doped or subjected to ball milling for longer periods of time. Obviously, this should not increase the level of organic solvent in the material. Also the 2H NMR spectrum of NaAID4 that is prepared in protio-solvents also contains an analogous sharp component thus precluding the possibility that this peak is due to residual solvent or an impurity in the hydride. We have also eliminated the possibility

© 2008, Woodhead Publishing Limited

400

Solid-state hydrogen storage

TiF3

TiF3

TiF3

TiF3

10 µm

(a)

Ti F

C Ti

0

Ti

Na

1

2

3

4

5

Cu

6

7 8 (b)

C 9

10

11

12

13

14

15

14.11 (a) SEM image taken with backscattered electrons (BSE). (b) Energy dispersive X-ray spectrum of a particle observed in uncycled NaAlH4 doped with 2 mol% TiF3 shows strong Ti and F peaks.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

0 1000

3000 2000

0 1000

5000 4000 (a)

3000 2000

0 1000

8000

6000

7000 8000

5000 4000 (b)

3000 2000

7000 6000

5000 4000 (c)

7000 6000

8000

–2000 0 1000 3000 5000 7000 –1000 2000 4000 6000 8000 (d)

0 1000

3000 2000

0 1000

5000 4000 (e)

3000 2000

7000 6000

5000 4000 (f)

401

8000

7000 6000

8000

–2000 0 1000 3000 5000 7000 –1000 2000 4000 6000 8000 (g)

0 1000

3000 5000 7000 2000 4000 6000 8000 (h)

14.12 EPR spectrum of: (a) 2 mol% TiF3 doped NaAlH4, uncycled; (b) 2 mol% TiF3 doped NaAlH4, after 3 cycles; (c) 2 mol% TiF3 doped NaAlH4, after 5 cycles; (d) 2 mol% TiF3 doped NaAlH4, after 10 cycles; (e) 2 mol% TiCl3 doped NaAlH4, uncycled, (f) 2 mol% TiCl3 doped NaAlH4, after 3 cycles; (g) 2 mol% TiCl3 doped NaAlH4, after 5 cycles; (h) 2 mol% TiCl3 doped NaAlH4, after 10 cycles (y axes, arbitrary units, x-axes, scaled to G-values (kHz)).

© 2008, Woodhead Publishing Limited

402

Solid-state hydrogen storage

NaAlH4 Narrow broad 23% 77%

Ti-doped NaAlH4 Narrow broad 35% 65%

400

200

0

–200

–400

ppm

1

14.13 Solid-state H NMR spectra of NaAlH4 with and without doping with 2 mol% Ti(OnBu)4.

that the sharp component is free H2 gas as its spectrum consists of a resonance that is much narrower and has a T1 of 13 ms. The remarkably short T1 of the narrow feature suggests that the second population hydrogen is metal-bound hydrogen which is not in a discrete, rigid [AlH4]– environment and that a significant population of hydrogen in NaAlH4 is highly mobile at even ambient temperature. We have also observed in the spectra of samples of the hydride that are ball milled and/or doped with titanium that the T1 of the broad component of the spectrum decreased by several orders of magnitude, clearly indicating a marked perturbation of hydrogen throughout the bulk material. Evidence that Ti-doping perturbs Al–H bonding throughout the bulk of the hydride was also obtained from infrared spectroscopy by Gomes et al.54 Figure 14.14 shows that, upon doping the hydride, the infrared absorption corresponding to the Al–H asymmetric stretching mode was seen to shift by ~15 cm–1 to higher frequency while that of the H–Al–H asymmetric bending mode shifted by ~20 cm–1 to lower frequency. A mobile population of hydrogen has also been observed by anelastic spectroscopy measurements that were carried out by Palombo et al.55 Heating NaAlH4 doped with 2 mol% TiCl3 to 436 K introduces a thermally activated relaxation process with a frequency of 1 kHz at 70 K. This denotes the formation of a point defect with a very high mobility (~5 × 103 jumps/s at 70 K). The relaxation involves the reorientation of H around Ti.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

403

The studies concerning the mechanism of action of NaAlH4 doped with 2 mol% TiF3 all indicate a change in Ti from a Ti(III) species (TiF3) to an Alassociated Ti(0) species. This change is shown to occur within the first few cycles of dehydrogenation/rehydrogenation. Only a relatively minor change is observed in the hydrogen cycling kinetics whether a Ti(III) to Ti(0) species predominates. Therefore, studies strongly suggest that the enhanced hydrogen cycling kinetics of Ti-doped NaAlH4 are due to a minority Ti species and that the majority of the Ti is in a resting state. Clearly, the effort to improve the dehydrogenation and rehydrogenation kinetics of complex hydrides would greatly aid the understanding of the mechanism of action of the dopants. Initial approaches to gain such insight into the Ti-doped NaAlH4 were to characterize the active Ti species. However, described above, it has been found that the active species must be a small minority of the Ti present in the material. Thus, the identification and characterization of the true active species will be difficult, if not impossible. On the other hand, it has been discovered that Ti doping has the effect of generating hydrogen-containing point defects in the hydride. It is evident that the generation of these defects is the key to enhancing the dehydrogenation/rehydrogenation kinetics of complex hydrides and thus it is imperative that they be characterized.

Absorbance

Pure NaAlH4 Milled 20 min Doped with 2 mol% Ti(OBu)4(*)

ν3

710

Combination bands

ν4

1000

1500 Infrared frequences (cm–1)

2000

14.14 Comparison of the infrared spectra of doped and undoped NaAlH4. (Gomes et al.54).

© 2008, Woodhead Publishing Limited

404

Solid-state hydrogen storage

14.3.3 Other doped alanates Bogdanović and Schwickardi found that the kinetic enhancement upon Tidoping can be extended to the reversible dehydrogenation of LiNa2AlH6 to LiH, 2 NaH, and Al.1 They found the mixed alkali metal alanate to have significantly lower plateau pressure at 211 °C and thus a higher ∆Hdehy than Na3AlH6. Fossdal et al.56 conducted PCT measurements over the 170–250 °C temperature range and, from the van’t Hoff plot of their data, determined ∆Hdehy to be 56 kJ/mol. Graetz et al.57 observed that the undoped mixed alanate will also undergo reversible dehydrogenation. They found that while doping 4 mol% TiCl3 results in a highly pronounced kinetic enhancement, it has no effect on the thermodynamic properties of the hydride. Titanium has also been found to enhance the dehydrogenation kinetics of LiAlH4.58 Balema et al.58 found that LiAlH4 could be transformed into Li3AlH6 and Al at room temperature upon the addition of 3 mol% TiCl4 upon 5 minutes of mechanical milling. The dehydrogenation of Li3AlH6 to LiH and Al was found by Chen et al.59 to occur at temperatures as low as 100 °C upon doping the hydride with 2 mol% TiCl3. Blanchard et al.60 reported that addition of a few mol% VCl3 can also reduce the thermal decomposition temperature of LiAlH4 by 60 °C. Chen et al. also reported XRD evidence59 for partial rehydrogenation of the resulting mixture of LiH/Al to an uncharacterized ‘intermediate phase’. However, attempts to reproduce this result were unsuccessful. Recently, Ritter and co-workers31 confirmed the findings of the earlier studies of Tidoped lithium alanate. These investigators also reported that the kinetics of the formation of the THF adduct of LiAlH4 from the ball milling of THF solutions of dehydrogenated material were improved by the addition of 0.5 mol% of a titanium catalyst. Fichtner et al.2 studied the kinetic effects of Mg(AlH4)2 doped with 2 mol% TiCl3 and mechanically milled for up to 100 minutes. The studies showed that the peak decomposition temperature was reduced in the presence of the titanium dopant. The starting point of the first decomposition step shifted to lower temperatures; however, complete dehydrogenation to MgH2 still requires heating to ~200 °C.

14.4

Density-functional calculations of alkali and alkaline-earth alanates

Density functional theory (DFT)61,62 has become a standard theoretical approach in the exploration of new solid-state hydrogen storage materials. Intense computational work has been attributed to the understanding of the promising alkali and alkali-earth alanate systems. Numerous theoretical studies, in particular, have focused on the systems of tetrahydridoalumate

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

405

NaAlH 4 , 63–99 LiAlH 4 , 63,66,67,74,75,77,80,81,83,94,100–107 KAlH 4 , 63,66,67,83,108 Ca(AlH 4), 109–113 and Mg(AlH4) 2; 80,81,111,114–119 hexahydridoalumate of N a 3A l H 6, 57,59,60,62,63,65,66,68,74,75,77,80,81,83,85,89,91,114,120–123 L3AlH6,63,66,80,81,83,101,124 K3AlH6;66,83,101,120,123 pentahydridoalumate of MgAlH5,125,126 CaAlH5;104,106,110,112,120,125–127 and extended to several mixed alanates such as K2NaAlH6.128,129 The objective of the theoretical work has been not only to be able to explain the peculiar observed properties among various known alanate systems, but, more importantly, to explore and design new alanates with improved thermal properties. This section presents an overview of progress achieved from the powerful DFT calculations in studying a broad range of alanate properties. We will start with the structural properties, followed by electronic and chemical bonding properties, reaction energetics, vibrational properties, mechanism of catalyst, and finally methods to predict new reaction pathways by the destabilization technique.

14.4.1 Crystal structures of alanates Crystal structure is an essential piece of information to understand the alanate systems. DFT calculation is a proven tool to provide structural understanding of known hydride materials. Great successes have also been reported in making accurate predications for experimentally unknown structures using DFT calculations. One is a simulated annealing procedure63, where, starting from a guessed structure, the molecular dynamics is used to simulate heating and cooling until the minimum energy structure is obtained. The other is a brute-force low-energy structure search108,110 with input structures having the same stoichiometry found in the International Crystal Structures Database (ICSD). Full geometrical optimizations on the atomic positions or cell parameters are carried out for each of possible initial guesses. The theoretically investigated alanates are listed in Table 14.3. For LiAlH4, NaAlH4, KAlH4, Li3AlH6, Na3AlH6, Mg(AlH4)2, and CaAlH5 the structural parameters are available experimentally, as presented in the previous section. The deviation in the calculated structural parameters from experimental values for these compounds is within 3%. In the case of CaAlH5, KAlH4 the crystal structures are first predicted by DFT calculations and then confirmed by subsequent diffraction experiments. The structural parameters for Ca(AlH4)2 are the theoretical predictions and no experimental results are available. For the tetrahydride alanates, it consists of a packing of tetrahedra AlH4 and the alkali or alkaline-earth metal atoms. The minimum Al–H distances are found to have similar values of 1.60, 1.62, 1.61, 1.64, and 1.63 Å in Mg(AlH4)2, LiAlH4, Ca(AlH4)2, NaAlH4, and KAlH4, respectively. The H– Al–H angle also varies little, showing that the tetrahedra of AlH4 complex

© 2008, Woodhead Publishing Limited

406

Solid-state hydrogen storage

are slightly distorted. This suggests that the AlH4 tetrahedra are stable units with nearly the same geometry, insensitive to the local environment.

14.4.2 Electronic structures and chemical bonding The nature of complicated chemical bonding in various hydride alanates has been explained with the help of a combination of calculated quantities such as partial density of states (PDOS), electronic band structure, total and difference charge-density distribution, electron-localization function (ELF), and Kulliken130 population analysis. It was found that the general common feature found among the alkali metal (Li, Na, K) and alkaline-earth metal (Mg, Ca) alanates in all types of hydrides of AlH4, AlH6 and AlH5 is the presence of dominant ionic bonding between the metal and the subunit of the Al with H. The interaction between Al and H within the subunit, however, comprises both ionic and covalent character, which decreases when one goes from Li to K due to the increased ionic radius. The calculated charge transfer clearly shows that electrons are transferred from alkali/alkalineearth metal and Al to the H sites, but the charge transfer from Al is found to be anisotropic, indicating the presence of ionic–covalent bonding between Al and H. The calculated ELF is extremely small between metal and H as well as small at the metal sites with spherically symmetric distribution, confirming bonding interaction between metal and H is purely ionic. The ELF distribution at the H site is not spherically symmetric and it is polarized towards Al atoms, indicating the presence of directional bonding between Al and H. Mulliken charges reflect nearly pure ionic picture with Li+ and H–. The overlap population between Li+ and H– is also close to zero, as expected for an ionic compound. All the alanates composed of AlH4 are found to be insulator with the band gap value67,70,100 of the order of 5 eV, which are known to be underestimated from the DFT-based calculations. The quasi-particle corrections65 found a value of about 6.9 eV. The valence band of these compounds consists of two well-separated regions. The lower energy region of valence band contains mainly Al-s and H-s electrons. The top of the valence band is dominated by H-s states as well as Al-p states which are energetically degenerate, implying the presence of covalent bonding between Al and H. Weak DOS at the alkali metal site indicates the ionic bonding between alkali and AlH4 unit.

14.4.3 Stability and thermodynamics of decomposition Electronic structure studies identified many similarities in the chemical binding among different alkali and alkaline-earth aluminum hydrides, however, with no clear signs in explaining significant differences in the decomposition

© 2008, Woodhead Publishing Limited

Table 14.3 Optimized ground-state (T = 0) structural parameters from DFT calculations ref

a Å

b Å

c Å

Al–H min Å

Al–H max Å

M–H Å

H–Al–H (°)

Egap eV

LiAlH4 ( P 2 1/ c )

[67, 102]

4.837

7.809

7.825

1.62

1.65

1.85

107.8

5.0

NaAlH4 ( I 4 1/ a )

[10] [30] [67, 29] [50] [67] [65] [66] [63] [56] [57] [67, 43]

4.9965 4.979 5.0004 9.009

4.9965 4.979 5.0004 5.767

11.0828 11.103 11.1141 7.399

5.18 5.23 5.232 13.37 13.408 8.017

5.18 5.23 5.232 9.28 9.562 8.017

5.98 6.04 6.025 8.91 8.947 9.4622

1.631 1.64 1.64 1.63 1.601 1.60 1.60 1.608 1.62 1.74

1.631 1.64 1.65 1.64 1.619 1.62 1.63 1.633 1.63 1.75

2.395 2.41 2.717 2.70 1.886 1.89 1.90 2.196

106.8

5.04 4.80 5.0 5.05 5.2 4.1 4.3 4.3 4.0

1.90

87.1

3.5

[30] [67, 70] [69] [67, 70]

5.349 5.385 5.347 6.025

5.349 5.51 5.526 6.102

7.707 7.719 7.678 8.582

1.764 1.76 1.75 1.78

1.784 1.77 1.77 1.79

2.283 2.22 2.227 2.58

89.8

2.73 3.0 3.3 3.0

[72]

4.7499

8.8127

6.6281

1.68

1.78

1.86

81.72

2.48

[57]

8.334

6.927

9.709

1.69

1.84

KAlH4 (Pnma) Mg(AlH4)2 (P-3m/1) Ca(AlH4)2 (Pbac) Li3AlH6 (R-3) Na3AlH6 ( P 2 1/ n ) K3AlH6 ( P 2 1/ n ) MgAlH5 ( P 2 1/ c ) CaAlH5 ( P 2 1/ n )

107.5 107.7 105.7 105.8

89.0

Alanates as hydrogen storage materials

Alanate

407

© 2008, Woodhead Publishing Limited

408

Solid-state hydrogen storage

behavior observed. For instance, as discussed in an earlier section, Na and K alanates exhibit reversible hydrogenation while Li and Mg alanates do not. The thermodynamic stability and reaction pathway, hence, has been investigated by computing enthalpy changes for the concerned reactions. The change of the entropy of formation HF and reaction HR at 0 K is defined as follows, respectively: ∆HF = E (MlAlmHn) – lE (M) – mE(Al) – n E(H2) 2

[14.7]

∆HR = ∑E (products) – ∑E (reactants)

[14.8]

and where E is the total energy of the individual bulk materials of interest or gaseous hydrogen molecule obtained from DFT calculations. The DFT predicted ∆HF and ∆HR for the reactions at T = 0 K are listed in Table 14.4. A more negative formation enthalpy is indicative of a more stable phase, relative to other candidate structures. Likewise, a more negative reaction enthalpy represents an increased favorability for the release of H2. There is one important difference that Li3AlH6 has a much more negative formation enthalpy than the heavier hexahydrides. It is too stable, which may make the reverse reaction thermodynamically unfavorable in the Li alanates. The lower Table 14.4 DFT predicted enthalpies of formation and reaction for the alanates at 0 K in unit of kJ/mol H2 Alanates

∆H F

Ref.

NaAlH4

–54.94 [67]

Na3AlH6

–69.9

[67]

LiAlH4

–55.5

[67]

Li3AlH6

–102.8

[67]

KAlH4

–70.0

K3AlH6

Reaction equation

∆H R

Ref.

NaAlH4 → NaH + Al + 3/2H2 NaAlH4 → 1/3(Na3AlH6 + 2Al + 3H2) 1/3Na3AlH6 → NaH + 1/3Al + 1/2H2

58.3 34.5 23.8

[4] [4] [4]

LiAlH4 → LiH + Al + 3/2H2 LiAlH4 → 1/3(Li3AlH6 + 2Al + 3H2) 1/3Li3AlH6 → LiH + 1/3Al + 1/2H2

22.5 9.2 13.3

[4] [4] [4]

[67]

KAlH4 → KH + Al + 3/2H2 KAlH4 → 1/3(K3AlH6 + 2Al + 3H2)

93.4

[4]

–78.5

[67]

1/3K3AlH6 → KH + 1/3Al + 1/2H2

Mg(AlH4)2

–21.1

[58]

Mg(AlH4)2 → MgH2 + 2Al + 3H2

–6.2

[58]

Ca(AlH4)2

–59.4

[58]

Ca(AlH4)2 → CaH2 + 2Al + 3H2

–20.7 –12.5 –4.8 29.8 172.1

[58] [56] [57] [57] [57]

CaAlH5

© 2008, Woodhead Publishing Limited

Ca(AlH4)2 → CaAlH5 + Al + 3/2H2 CaAlH5 → CaH2 + Al + 3/2H2 CaH2 → Ca + H2

Alanates as hydrogen storage materials

409

stability of the lithium is due to the small size of the lithium ion which constrains the crystal stability. The agreement of the calculated and measured values listed in the previous section is, in general, good. The difference may lie partly in the fact that the intermediate species are actually less welldefined experimentally. It is interesting to see in Table 14.4 that the decomposition of Ca alanate is significantly more endothermic than that of Mg alanate. The formation enthalpy of Ca alanate is similar to that of NaAlH4, which may explain the observed reversible hydrogenation in Ca alanate at reasonable conditions. On the other hand, the magnesium alanate being thermodynamically too unstable could explain the difficulty in hydrogenating Mg alanate from the gas phase.

14.4.4 Vibrational properties of alanates Valuable insight into the decomposition process and hydrogen evolution can be gained through lattice dynamics studies. The comprehensive vibrational properties of NaAlH477,99, Ca(AlH4)2110 have been investigated using the first-principle approach based on DFT and linear response theory. The importance of low-energy vibrations, dominated by the motion of the metal and metal–hydrogen subunit, for the melting process of undoped sodium alanate is reported.

14.4.5 Mechanism of the catalytic effects The critical role played by the Ti catalyst in helping hydrogen cycling in alanates has been an on-going research subject for theoreticians. Ti-doped NaAlH4 system has been most studied as a prototypical system for materials of similar structures. The exact nature of the titanium catalyst action and the location of the Ti atoms still remain unclear; however, several different theories have been elucidated to explain the observed reaction dynamics. Three doping hypotheses have been applied to study the catalytic effects with DFT simulations. One is to search for evidence of whether Ti exists as a bulk dopant on the sodium alanate lattice70,73,74,76,78,79,82,84,91,92,122 and if Ti-related vacancies lower the diffusion barrier of hydrogen. Another model examines the changes of surface states due to the addition of Ti to the NaAlH4 surfaces66,69,72,75,81,87,93–95 instead of inside the bulk. The third model proposes that the titanium atoms are neither inside the bulk nor on the surface of NaAlH4, but forming a separate phase on a bulk Al surface96,97,131,132 as a catalytically active TiAl3 type of species. Theoretical studies of Ti-doped NaAlH4 using a standard bulk cohesive energy76, or the product side of the reaction equations as the reference state,72 show that Ti on the surface or substituted into the lattice is unstable compared with Ti in Ti bulk, with substitution of Al by Ti being the most favorable. On

© 2008, Woodhead Publishing Limited

410

Solid-state hydrogen storage

the other hand, substitution of lattice Na by Ti is found to be stable75,92 if gas-phase atoms are used as the reference state. This discrepancy was reported purely due to the choice of different reference energy states.73 Similar to NaAlH4, some calculations121,122 found that Ti is unstable and substitutes Al atom in the Na3AlH6 lattice. One stable configuration84 is reported where Ti substitutes Al and also brings in an interstitial H at H-rich and Al-poor condition. The possible effect of point defect on the hydrogen diffusion and desorption in bulk alanates has also been studied.67,73,79,82,121,122 It was found73,122 that the vacancies promote hydrogen desorption better than Ti substituting as impurity in solids. In undoped alanates with the presence of hydrogen vacancy, the barrier82,121 for long-range hydrogen diffusion in NaAlH4 (Na3AlH6) is 0.31 (0.75) eV and for localized motion is 0.44 (0.41) eV. In addition to Ti metal, a calculation85 on doping other metal atoms to NaAlH4 lattice indicated that Cr and Fe are more effective than Ti in desorbing hydrogen. The NaAlH4 (001) surface has been shown to be the most stable surface of the different crystal faces.88,95 An interstitial site beneath the NaAlH4 (001)93,94 is the most favorable location for Ti, where it forms a complex structure of TiAl3H12. The hydrogen desorption energy is significantly reduced as compared with that from the undoped surface. Another spin-polarized calculation87 on vacancy-mediated H2 desorption demonstrated that the presence of Na vacancy on NaAlH4 (001) surface weakens two Al–H bonds, by which the desorption kinetics of H2 is accelerated. It was found95 that the addition of atomic Ti is thermodynamically viable on both (001) and (110) surfaces only if simultaneous formation of Na vacancies. A third model is that the catalyst acts as a hydrogen dissociation– recombination site near the surface. At the dehydrogenation, the catalyst kinetically facilitates the release of AlH3 from solid-state alanates. At the hydrogenation, the catalyst helps the absorption of hydrogen and formation of AlH3 on Al surface.96,97,131 The role is to assist in the dissociation or recombination of hydrogen and structural decomposition by massive mass transfer of aluminum via the mobile AlH3. Calculations found stable TiAl3 formed near the Al surface with Ti occupying subsurface sites. This improves hydrogen kinetic properties on Al surface. The dissociation energy barrier is greatly reduced by 0.6 eV. Overall, the catalytic ability of a transition metal dopant is found to be correlated to its tendency to form bonds with Al atoms.

14.4.6 Destabilization As the known stable complex hydrides lack sorption kinetics, it is important to identify possible metastable phases. It has long been known that the thermodynamics of dehydrogenation reactions can be changed by mixing

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

411

additives to form new compounds which are energetically favorable than the ones without additives. The basic principle is to destabilize the original material. Some success has been reported113,133,134 using DFT calculations to efficiently screen potential reactions of mixed compounds which have the acceptable enthalpy in the range 30–58 kJ/mol. The procedure is brute force by first identifying a set of N structures of compounds in crystallographic databases with the correct chemical formula and the correct ionic coordination. Then DFT calculations are carried out for the selected material in all N structures. The lowest energy structure is determined after relaxing all structural degrees of freedom. Finally reaction enthalpies for the destabilized metal hydrides are calculated. DFT calculation has proven to be a valuable tool both in modifying and optimizing the known hydrogen storage material, and predicting new resources, even though their crystal structures are not fully known experimentally. Electronic structures are powerful in understanding the chemical bonding characteristics; however, the formation and reaction energies are essential in exploring the thermodynamic stabilities among different alanates.

14.5

Future trends

Although there are instances of transition metal doped alanates that have hydrogen cycling capacities, thermodynamic properties, or kinetic performances that are suitable for on-board vehicular hydrogen storage (Table 14.5), no material meets all three criteria. Sodium alanate, however, has a combination that most closely approaches those that are requisite of the long-sought, ‘holy grail’ material. Overall, a number of thermodynamic, kinetic, capacity, and safety issues cloud the future practical application of this class of materials.

14.5.1 Thermodynamic considerations The dehydrogenation of both LiAlH4 and Mg(AlH4)2 is irreversible. Thus thermodynamic constraints would allow only NaAlH4 to be utilized as an on-board hydrogen storage material that could be recharged by direct hydrogenation. However, LiAlH4 and Mg(AlH4)2 might find application in a system in which they were utilized as ‘chemical hydrides’ involving their regeneration through an off-board chemical process. A van’t Hoff plot of the data obtained from PCT studies of Ti-doped NaAlH4 is presented in Fig. 14.15. From the data,43 the enthalpy of the dehydrogenation of NaAlH4(s) to Na3AlH6 and Al has been determined to be 37 kJ/mol H2. This value is in line with the predictions of the studies discussed earlier. In accordance with this value, the temperature required for an equilibrium hydrogen pressure of 0.1 MPa has been determined138 as 33 °C

© 2008, Woodhead Publishing Limited

412

Table 14.5 Hydrogen storage properties of alanates Hydrogen (wt%) (ideal)

Hydrogen (wt%) (Obs.) ———————————— 1st dehyd. Re-hyd.

Condition ∆Hdehyd ———————————————— (kJ/mol H2) 1st Dehyd. Re-hyd. Temp. (°C) Exp. Temp. (°C) (Press. (MPa))

Ref.

LiAlH4 = LiH + Al + 3/2H2 LiAlH4 = 1/3Li3AlH6 + 2/3Al + H2 LiAlH4 = 1/3Li3AlH6 + 2/3Al + H2 (with Ti dopant) Li3AlH6 = 3LiH + Al + 3/2H2 Li3AlH6 = 3LiH + Al + 3/2H2 (with Ti dopant) NaAlH4 = NaH + Al + 3/2H2

8.0 5.3 5.3

8.0 5.3 5.3

– – –

201 187–218 25

– – –

5.8 –9.1 –

16 16, 24 135

5.6 5.6

5.6 5.5

– –

228–282 100–120

– –

27.0 –

16, 24 136

5.6

5.6

5.6

265

27.0 (17.5)

56.5

NaAlH4 = NaH + Al + 3/2H2 (with Ti dopant) NaAlH4 = 1/3α-Na3AlH6 +2/3Al + H2 NaAlH4 = 1/3α-Na3AlH6 +2/3Al + H2 (with Ti dopant) β-Na3AlH6 = 3NaH + Al + 3/2H2

5.6

5.0

3.5–4.3

160

56.5

3.7 3.7

3.7 3.7

– 3.5

210–220 90–150

12.0–15.0 (11.5) – 12.0 (11.5)

3.0

3.0



250



46.8

β-Na3AlH6 = 3NaH + Al + 3/2H2 (with Ti dopant) Na2LiAlH6 = 2NaH + LiH + Al + 3/2H2 Na2LiAlH6 = 2NaH + LiH + Al + 3/2H2 (with Ti dopant) KAlH4 = KH + Al + 3/2H2

3.0

3.0

2.9

100–150

12.0 (11.5)

47.0

3.5 3.5

3.2 3.0

2.8–3.2 2.6–3.0

170–250 170–250

– –

53.0 53.0

4, 17, 22, 26, 27 33, 39 63, 137 17, 22 39, 63 137 17, 22, 26, 27 33, 36, 39 57, 58 57, 58

4.3

3.5

2.6–3.7

290

~86.0

28

6.9 6.9

6.9 6.9

– –

163–285 140–200

25.0–33.0 (0.10) – –

– –

2 2

Mg(AlH4)2 = MgH2 + Al + 2H2 Mg(AlH4)2 = MgH2 + Al + 2H2 (with Ti dopant)

© 2008, Woodhead Publishing Limited

36.0 37.0, 40.9

Solid-state hydrogen storage

Reaction

Alanates as hydrogen storage materials

413

and highly practical hydrogen plateau pressures of 0.2 and 0.7 MPa have been found at 60 and 80 °C, respectively.43 Unfortunately, the equilibrium hydrogen plateau pressures of the Na3AlH6/NaH + Al + H2 equilibrium in the 70–100 °C temperature range are insufficient for utilization in a polymer electrolyte membrane (PEM) fuel cell system in which heat of the steam exhaust would be used to drive hydrogen release for the storage material. As seen in Fig. 14.15, a temperature of 110 °C is required to attain a plateau pressure of 0.1 MPa. Thus unless a system has an additional means of heating the storage material, the potential hydrogen storage capacity of sodium alanate is limited to 3.6 wt%. As discussed in the previous section, Na2LiAlH6 and KAlH4 have even lower plateau pressures than Na3AlH6 and thus offer no storage potential in this type of system.

14.5.2 Kinetic considerations Ti-doped LiAlH4, NaAlH4, and Mg(AlH4)2 all undergo dehydrogenation at appreciable rates at temperatures at or below 100 °C, suggesting their possible

220 200

175 150 125

Temperature [°C] 100 80 60

40

20

1000 Bogdanovic´ et al.39 Sandia Nat. Lab.37

180 °C

Pressure [atm]

100

NaAlH4

Na3AlH6 10 33 °C

NaH 1 110 °C

0.1

2

2.5

3

3.5

1000/T

14.15 Van’t Hoff plot of equilibrium pressures as a function of temperature for the NaAlH4 ⇔ 1/3(α-Na3AlH6) + 2/3Al + H2 and the αNa3AlH6 ⇔ 3NaH + Al + 3/2H2 reactions. Samples doped with 2 mol% each of Ti(OBun)4 and Zr(OPri)4. Reprinted with permission from Jensen and Gross.37 Copyright 2001, Springer.

© 2008, Woodhead Publishing Limited

414

Solid-state hydrogen storage

application as hydrogen carries for on-board PEM fuel cells. The recent finding of further kinetic enhancement of NaAlH4 through Sc-doping demonstrates that even faster rates of dehydrogenation can be achieved. However, it should be noted that scandium is expensive and may not be the best option for practical on-board hydrogen storage. The kinetics of rehydrogenation for Ti-doped NaAlH4, the only candidate material for on-board recharging, is a much more daunting challenge to any effort for its practical development. As advances have been made in improving the kinetics of rehydrogenation, the problem of managing the heat generated from this highly exothermic process has come to light.139 Clearly, a very high-performance heat exchanger will be required to adequately remove large amounts of heat that would be discharged during the short (~5 minute) time that is considered to be acceptable for practical on-board re-fueling.

14.5.3 Hydrogen cycling capacity performance The long-term cycling of mixtures of NaH/Al that were doped with 2 wt% TiF3 has been studied through 100 cycles (hydrogenation 150 °C, 11.4 MPa; dehydrogenation 160 °C).140 A release of 3.4 wt% of hydrogen was obtained in the first dehydrogenation half-cycle. Upon cycling, the storage capacity showed a steady increase to the 4.0 wt% limit that was achieved by the fourth dehydrogenation half-cycle. This capacity remained through 40 cycles before the onset of a steady decrease resulting in a loss of capacity from 4.0 to 3.5 wt% by the end of the 100th cycle. Synchrotron powder X-ray diffraction and Rietveld profile fitting analysis of Ti-doped (NaH + Al) after 100 cycles showed the presence of additional phases, which are apparently linked to the observed diminishing dehydrogenation characteristics. Decreases in dehydrogenation kinetics and the total amount of released hydrogen over a number of cycles parallel the appearance of the intermediate phase, Na3AlH6, in the prolonged cycling samples. This indicates a reduction in the effectiveness of the Ti-dopant for the hydrogenation of Na3AlH6 to NaAlH4. Clearly, development methods for stabilizing the hydrogen storage capacity upon long-term cycling would be necessary before this material could be used for practical applications. Moreover, the 3.5–4.0 wt% cycling capacity falls far short of the current US Department of Energy capacity target that is thought to be necessary for a practically viable, on-board hydrogen storage material.

14.5.4 Safety The main safety concern associated with the alanates is their high reactivity with water. The heat of the hydrolysis reactions can drive the thermal

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

415

decomposition of the hydrides. However, the pressure rise associated with introduction of small amounts of contaminant water into the system during recharging would not constitute a significant fraction of the pressure rating of any credible container. However, contact of this hydride system with a larger amount of water upon accidental tank rupture in a wet environment is clearly a safety concern. This hazard must be eliminated either through system engineering or chemical modification of the hydride.

14.6

Conclusions

In view of the thermal management issue associated with on-board recharging of Ti-doped sodium alanate and the irreversibility of the dehydrogenation of LiAlH4 and Mg(AlH4)2, it appears that a practical hydrogen storage system based on alanates would most likely require off-board recharging. Currently, only sodium alanate can be directly charged under hydrogen pressure and, thus, until new, highly economical chemical processes for the regeneration of alanates are developed, it is the only member of this class of materials which is a plausible candidate for practical hydrogen storage for vehicular applications. Despite the problems discussed above, Ti-doped sodium alanate to date stands as the hydrogen storage material with the best combination of material properties that are suited for vehicular hydrogen storage. However, its cycling hydrogen capacity is only 3–4 wt%. Thus, barring the discovery of the ‘holy grail’ hydrogen storage material, the practical application of these materials will await a time when there is public acceptance of vehicles that are lighter weight and/or capable of cruise ranges that are significantly shorter than 300 miles.

14.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

References B. Bogdanović, M. Schwickardi. J. Alloy Comp. 1997, 1–9, 253. M. Fichtner, O. Fuhr, O. Kircher. J. Alloy Comp. 2003, 356–357, 418. J. Huot, S. Boily, V. Guther, R. Schulz. J. Alloy Comp. 1999, 283, 304. T.N. Dymova, N.G. Eliseeva, S.I. Bakum, Y.M. Dergachev. Dokl. Akad. Nauk SSSR 1974, 215, 1369. J.W. Lauher, D. Dougherty, P.J. Herley. Acta Cryst. 1979, B35, 1454. N. Sklar and B. Post. Inorg. Chem. 1967, 6, 699. V.K. Bel’skii, B.M. Bulychev, A.V. Golubeva. Russ. J. Inorg. Chem. 1983, 28, 1528. B.C. Hauback, H.W. Brinks, C.M. Jensen, K. Murphy, A.J. Maeland. J. Alloy Comp. 2003, 358, 142. B.C. Hauback, H.W. Brinks, H. Fjellvåg. J. Alloy Comp. 2002, 346, 184. B.C. Hauback, H.W. Brinks, R.H. Heyn, R. Blom, H. Fjellvåg. J. Alloy. Comp. 2005, 394, 35. A. Fossdal, H.W. Brinks, M. Fichtner, B.C. Hauback. J. Alloy Comp. 2005, 387, 47.

© 2008, Woodhead Publishing Limited

416

Solid-state hydrogen storage

12. E. Rönnebro, D. Noréus, K. Kadir, A. Reiser, B. Bogdanović. J. Alloy Comp. 2000, 299, 101. 13. H.W. Brinks, B.C. Hauback. J. Alloy Comp. 2003, 354, 143. 14. H.W. Brinks, B.C. Hauback, D. Blanchard, C.M. Jensen, M. Fichtner, H. Fjelvåg. Adv. Mat. Eng. Con. II. TMS. 2004, 197. 15. H.W. Brinks, B.C. Hauback, C.M. Jensen, R. Zidan. J. Alloy Comp. 2005, 392, 27. 16. W.E. Garner, E.W. Haycock. Proc. Roy. Soc. (London) 1952, A211, 335. 17. E.C. Ashby, P. Kobetz. Inorg. Chem. 1966, 5, 1615. 18. T.N. Dymova, S.I. Bakum. Russ. J. Inorg. Chem. 1969, 14, 1683. 19. V.A. Kuznetsov, N.D. Golubeva, K.N. Semenenko. Russ. J. Inorg. Chem. 1974, 19, 669. 20. J.A. Dilts, E.C. Ashby. Inorg. Chem. 1972, 11, 1230. 21. T. Kiyobayashi, S.S. Srinivasan, D. Sun, C.M. Jensen. J. Phys. Chem. A. 2003, 107, 7671. 22. T.N. Dymova, Y.M. Dergachev, V.A. Sokolov, N.A. Grechanaya. Dokl. Akad. Nauk SSSR 1975, 224, 591. Engl. Ed. 1976, 556. 23. K.J. Gross, S. Guthrie, S. Takara, G. Thomas. J. Alloy Comp. 2000, 297, 270. 24. J. Block, A.P. Gray. Inorg. Chem. 1965, 4(3), 304. 25. M.W. Chase, C.A. Davies, J.R. Downey, D.J. Frurip, R.A. McDonald, A.N. Syverud. JANAF Thermochem. Tab. (Third Ed.), J. Phys. Chem. Ref. Data, Suppl. 1. 1985, 14, 1. 26. P. Claudy, B. Bonnetot, G. Chahine, J.M. Letoffe. Thermochim. Acta 1980, 38, 75. 27. J.-P. Bastide, B. Bonnetot, J.M. Letoffe, P. Claudy. Mater. Res. Bull. 1981, 16, 91. 28. H. Morioka, K. Kakizaki, S. Chung, A. Yamada. J. Alloy Comp. 2003, 353, 318. 29. H. Clasen. Angew. Chem. 1961, 73, 322. 30. E.C. Ashby, G.J. Brendel, H.E. Redman. Inorg. Chem. 1963, 2, 499. 31. J. Wang, A.D. Ebner, J.A. Ritter. J. Am. Chem. Soc. 2006, 128, 5949. 32. E. Wieberg, R. Bauer, M. Schmidt, R. Uson. Z. Naturfor. 1951, 6b, 393. 33. B. Bogdanovic´, M. Schwickardi. J. Serb Chem. Soc. 1989, 54, 579. 34. German patent application 195 26 434.7, 1995. 35. C.M. Jensen, R. Zidan, N. Mariels, A. Hee, C. Hagen. Int. J. Hyd. En. 1999, 23, 461. 36. R.A. Zidan, S. Takara, A.G. Hee, C.M. Jensen. J. Alloy Comp. 1999, 285, 119. 37. C.M. Jensen and K.J. Gross. Appl. Phys. A 2001, 72, 213. 38. C.M. Jensen, D. Sun, K. Murphy, C. Wang, K. Kumashiro, W. Neimczura. Proc. of 2002 U.S. DOE Hydrogen and Fuel Cells Annual Program/Lab R&D Review. May 6–10, 2002. Golden, CO. 39. B. Bogdanović, R.A. Brand, A. Marjanovic, M. Schwickardi, J. Tölle. J. Alloy Comp. 2000, 302, 36. 40. A. Zaluska, L. Zaluski, J.O. Ström-Olsen. J. Alloy Comp. 2000, 298, 125. 41. B. Bogdanović, M. Schwickardi. German Patent DE 19526434A1, 1995. 42. D.L. Anton. J. Alloy Comp. 2003, 356, 400. 43. E.H. Majzoub, K.J. Gross. J. Alloy Comp. 2003, 356–357, 363. 44. S.S. Srinivasan, H.W. Brinks, B.C. Hauback, D. Sun, C.M. Jensen. J. Alloy Comp. 2004, 377, 283.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials

417

45. K.J. Gross, E.H. Majzoub, S.W. Spangler. J. Alloy Comp. 2003, 356–357, 423. 46. P. Wang, C.M. Jensen. J. Phys. Chem. B. 2004, 108, 15827. 47. B. Bogdanović, M. Felderhoff, S. Kaskel, A. Pommerin, K. Schlichte, F. Schüth. Adv. Mater. 2003, 15, 1012–1015. 48. M. Fichtner, O. Fuhr, O. Kircher, J. Rothe. Nanotechnology 2003, 14, 778–785. 49. B. Bogdanović, M. Felderhoff, A. Pommerin, F. Schüth, N. Spielkamp. Adv. Mater. 2006, 18, 1198. 50. T. Wang, J. Wang, A.D. Ebner, J.A. Ritter. J. Am. Chem. Soc. 2006, 128, 949. 51. H.W. Brinks, C.M. Jensen, S.S. Srinivasan, B.C. Hauback, D. Blanchard, K. Murphy. J. Alloys Comp. 2004, 376, 215. 52. C.M. Andrei, J. Walmsley, H.W. Brinks, R. Homestad, C.M. Jensen, B.C. Hauback. Appl. Phys. A 2005, 80, 709. 53. M.T. Kuba, S.S. Eaton, C. Morales, C.M. Jensen. J. Mater. Res. 2005, 20, 3265. 54. G. Gomes, G. Renaudin, H. Haqemann, K. Yvon, M.P. Salic, C.M. Jensen. J. Alloys Comp. 2005, 390, 305. 55. O. Palombo, R. Cantelli, A. Paolone, S.S. Srinivasan, C.M. Jensen. J. Phys. Chem. B 2005, 109, 1168. 56 A. Fossdal, H.W. Brinks, J.E. Fonneløp, B.C. Hauback. J. Alloy Comp. 2005, 397, 135. 57. J. Graetz, Y. Lee, J.J. Reilly, S. Park, T. Vogt. Phys. Rev. B 2005, 71, 184115. 58. V.P. Balema, K.W. Dennis, V.K. Pecharsky. Chem. Comm. 2000, 1665. 59. J. Chen, N. Kuriyama, Q. Xu, H.T. Takeshita, T. Sakai. J. Phys. Chem. B 2001, 105, 11214. 60. D. Blanchard, H.W. Brinks, B.C. Hauback, P. Norby. Mat. Sci. Eng. B 2004, 108(1–2), 54. 61. W. Kohn, L.J. Sham. Phys. Rev. B 1965, 140. 62. P. Hohenberg, W. Kohn. Phys. Rev. B 1964, 136. 63. S.-C. Chung, H. Morioka. J. Alloys Compounds 2004, 372(1–2), 92–96. 64. A. Aguayo, and D.J. Singh. Phys. Rev. B 2004, 69(15), 155103. 65. A. Peles, et al. Complex hydrides. Phys. Rev. B 2004, 70(16), 165105. 66. O.M. Løvvik, O. Swang, S.M. Opalka. J. Mater. Res. 2005, 20(12), 15. 67. P. Ravindran et al. J. Power Sources 2006, 159(1), 88–99. 68. S.M. Opalka, D.L. Anton. J. Alloys Compounds 2003, 356–357, 486–9. 69. P. Vajeeston et al. Appl. Phys. Lett. 2003, 82(14), 2257. 70. O.M. Lovvik, S.M. Opalka. Appl. Phys. Lett. 2006, 88(16), 161917. 71. J. Liu, and Q. Ge. J. Alloys Compounds 2007, 446–447, 31, 267. 72. A. Marashdeh et al. Chem. Phys. Lett. 2006, 426(1–3), 180. 73. C.M. Araujo et al. Phys. Rev. B 2005, 72(16), 165101. 74. C.M. Araujo et al. Appl. Phys. Lett. 2005, 86(25), 251913. 75. J. Iniguez, T. Yildirim. Appl. Phys. Lett. 2005, 86(10), 103109. 76. O.M. Lovvik, S.M. Opalka. Phys. Rev. B 2005, 71(5), 054103. 77. E.H. Majzoub, K.F. McCarty, V. Ozolins. Phys. Rev. B 2005, 71(2), 10. 78. A. Marashdeh et al. J. Phys. Chem. C, 2007, 111(23), 8206. 79. A. Peles, C.G. Van de Walle. J. Alloys Compounds 2007, 446–447, 459. 80. V.A. Popa, G.A. de Wijs, G. Brocks. Phys. Rev. B 2007, 75(3), 035204. 81. M.J.V. Setten, et al. Phys. Rev. B 2007, 75(3), 035204. 82. Q. Shi et al. J. Alloys Compounds 2007, 446–447, 469. 83. M.E. Arroyo y de Dompablo, G. Ceder J. Alloys Compounds 2004, 364(1–2), 6. 84. K. Bai, P. Wu, Appl. Phys. Lett. 2006, 89(20), 201904.

© 2008, Woodhead Publishing Limited

418 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111.

112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129.

Solid-state hydrogen storage A. Blomqvist et al. Appl. Phys. Lett. 2007, 90(14), 141904. A.J. Du, S.C. Smith, G.Q. Lu. Phys. Rev. B 2006, 74(19), 193405. A.J. Du, S.C. Smith, G.Q. Lu. Appl. Phys. Lett. 2007, 90(14), 143119. T.J. Frankcombe, and O.M. Lovvik. J. Phys. Chem. B 2006, 110(1), 622. X. Ke, I. Tanaka. Phys. Rev. B 2005, 71(2), 024117. R.S. Kumar et al. Phys. Rev. B 2007, 75(17), 174110. E.-K. Lee, Y.W. Cho, J.K. Yoon. J. Alloys Compounds 2006, 416(1–2), 245. J. Iniguez et al. Phys. Rev. B 2004, 70(6), 060101. J. Liu, Q. Ge. J. Phys. Chem. B 2006, 110(51), 25863. J. Liu, Q. Ge. Chem. Commun 2006, 1822. T. Vegge, Phys Chem. Chem. Phys. 2006, 8(42), 4853. S. Chaudhuri, J.T. Muckerman. J. Phys. Chem. B 2005, 109(15), 6952. S. Chaudhuri et al. J. Am. Chem. Soc. 2006, 128(35), 11404. B. Bogdanovic´ et al. Scripta Materialia, 2007, 56(10), 813. A. Peles, M.Y. Chou. Phys. Rev. B 2006, 73(18), 184302. P. Vajeeston et al. Phys. Rev. B 2003, 68(21), 212101. O.M. Lovvik et al. Phys. Rev. B 2004, 69(13), 134117. O.M. Lovvik. J. Alloys Compounds 2004, 373(1–2): 28. O.M. Lovvik, J. Alloys Compounds 2003, 356–357, 178. Y. Song, R. Singh, Z.X. Guo, J. Phys. Chem. B 2006. 110(13), 6906. A.V. Talyzin, B. Sundqvist. Phys. Rev. B 2004, 70(18) 180101 P. Vajeeston et al. Phys. Rev. B 2003, 68(21), 212101. J.K., Kang et al. J. Chem. Phys. 2004, 121(21), 10623. P. Vajeeston, P. Ravindran, A. Kjekshus, H. Fjellvåg. J. Alloys Compounds 2003, 363, L7–L11. O.M. Lovvik. Phys. Rev. B 2005, 71(14), 144111. C. Wolverton, V. Ozolins. Phys. Rev. B 2007, 75(6), 064101. O.M. Lovvik, P.N. Molin. In Hydrogen in Matter: A Collection from the Papers Presented at the Second International Symposium on Hydrogen in Matter (ISOHIM). 2006, Uppsala (Sweden): AIP. C.W.A.V. Ozoliņš. Phys. Rev. B 2007, 75, 064101. S.V. Alapati, J.K. Johnson, D.S. Sholl. J. Phys. Chem. B, 2006, 110(17), 8769. M. Fichtner, O. Fuhr, O. Kircher, J. Alloys Compounds 2003, 356–357, 418. M. Fichtner et al. Inorg. Chem. 2003, 42(22), 7060. Z.F. Hou. J. Power Sources, 2006, 159(1), 111. M. Palumbo et al. Calphad 31, 2007, 457. O.M. Lovvik, P.N. Molin, Phys. Rev. B 2005, 72(7), 073201. M.J. van Setten et al. Phys Rev. B 2005, 72(7) 073107. P. Vajeeston et al. Phys. Rev. B 2005, 71(9), 092103. J. Voss et al. J. Phys. Chem. B, 2007, 111(15), 3886. S. Li, P. Jena, R. Ahuja. Phys. Rev. B, 2006, 73(21), 214107. O.M. Lovvik, O. Swang. J. Alloys Compounds, 2005, 404–406, 757. P. Vajeeston et al, Phys. Rev. B 2004, 69, 020104. A. Klaveness et al. Phys. Rev. B 2006, 73(9), 094122. A. Klaveness et al. J. Alloys Compounds 2007, 433(1–2), 225. C. Weidenthaler, T.J. Frankcombe, M. Felderhoff. Inorg. Chem. 2006, 45(10), 3849. E. Ronnebro, E.H. Majzoub. J. Phys. Chem. B 2006, 110(51), 25686. M.H. Sorby et al. J. Alloys Compounds 2006, 415(1–2), 284.

© 2008, Woodhead Publishing Limited

Alanates as hydrogen storage materials 130. 131. 132. 133. 134. 135. 136. 137. 138. 139.

140.

419

R.S. Mulliken J. Chem. Phys. 1955, 23. Y. Wang, M.-Y. Chou. Unpublished work. R.T. Walters, J.H. Scogin. J. Alloys Compounds 2006, 421(1–2), 54. S.V. Alapati, J.K. Johnson, D.S. Sholl, J. Alloys Compounds 2007, 446–447, 23. S.V. Alapati, J.K. Johnson, D.S. Sholl, J. Phys. Chem. C, 2007, 111(4), 1584. V.P. Balema, K.W. Dennis, U.K. Pecharsky. Chem. Commun. 2000, 1665. J. Chen, N. Kuriyama, Q. Hu, H.T. Takeshita, T. Sakai. J. Phys. Chem. B 2001, 105, 11214. T. Kiyobayashi, S.S. Srinivasan, P. Sun, C.M. Jensen. J. Phys. Chem. A 2003, 107, 7671. G. Sandrock, K. Gross, G. Thomas, C. Jensen, D. Meeker, S. Takara. J. Alloys Comp. 2002, 330–332, 696. D.L. Anton, M.A. Mosher, S.M. Opalka. Proc. of 2003 U.S. DOE Hydrogen Fuel Cells and Infrastructure Technologies Program Review. May 19–22, 2003. Berkeley, CA. S.S. Srinivasan, H.W. Brinks, B.C. Hauback, D. Sun, C.M. Jensen. J. Alloy Comp. 2004, 377, 283.

© 2008, Woodhead Publishing Limited

15 Borohydrides as hydrogen storage materials Y. N A K A M O R I and S. O R I M O, Tohoku University, Japan

15.1

Introduction

Borohydrides belong to a class of materials with the highest gravimetric hydrogen densities. There are many types of borohydrides depending on the composition of M, B and H, e.g. M(BH4)n, M(B3H8)n, M(B6H6)n, M(B9H9)n and M(B 12 H 12 ) n . 1 Here, we focus on M(BH 4 ) n , tetrahydridoborate/ tetrahydroborate. The element M formed M(BH4)n is shown in Fig. 15.1 and Table 15.1.3–5 The study of M(BH4)n is important area of for research in the use of complex hydrides as hydrogen storage materials from below reason; M(BH4)n has hydrogen-rich anion (BH4)– which is similar to (AH4)– in M(AlH4)n (alanate). Moreover, (M(BH4)n has a higher gravimetric hydrogen density than M(AlH4)n due to B being lighter than Al (see middle line in Fig. 15.1). On the other hand, M(BH4)n does not release hydrogen by the same reaction pathways of alanate, e.g. Na(AlH 4) to Na 3AlH 6. Therefore, confirmation of similarity and differency between M(BH4)n and M(AlH4)n is important. In this chapter, basic properties of M(BH4)n are summarized. In Sections 15.2 and 15.3, synthesis methods and clarified structures of M(BH4)n are presented. The general reaction of dehydrogenation and rehydrogenation, thermodynamics and kinetics properties are shown in Section 15.4. Finally, current potentiality and outlook for hydrogen storage are summarized in Section 15.5. The earlier works on the borohydrides have also been summarized in previous reviews.3–8,16–19

15.2

Synthesis of borohydrides

15.2.1 Alkali borohydrides Alkali borohydrides (tetrahydridoborate/tetrahydroborate), MBH4 (M = Li, Na and K) are white powders that are used for effective reducing agent in organic and inorganic chemistry. LiBH4 is extremely hydroscopic, being rapidly decomposed by moist air. KBH4 hydrolyzes slowly on exposure to 420 © 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials 1

2

Li 18.5 1.0 Na 10.7 0.9 K 7.5 0.8 Rb 4.0 0.8 Cs 2.7 0.7

Be 20.8 1.5 Mg 14.9 1.2 Ca 11.6 1.0 Sr 6.9 1.0 Ba 4.8 0.9

3

4

Ti Sc III 13.5 IV13.1 15.0 1.3 1.5 Y Zr 9.1 10.7 1.2 1.4 Hf Ln 6.8 1.3

5

6

7

8

9

10

11

12

V Cr [Mn] [Fe] [Co] Ni [Cu]* Zn 12.7 9.9 9.5 9.4 9.1 9.1 5.1 4.4 1.6 1.6 1.5 1.8 1.8 1.8 1.9 1.6 Nb [Ag]* [Cd] 8.8 (Mo) (Ru) (Rh) (Pd) 3.3 2.9 1.6 1.9 1.7 [Au] Hg (W) (Ir) 1.9 1.8 2.4 1.9

421

13

14

Al 16.9 1.5 [Ga] 10.6 1.6 [In] 7.6 1.7 Tl 4.9 1.8

Ge 12.2 1.8 [Sn] 9.1 1.8

Tm 5.7

Yb 5.6

Ac

Ln

La 6.6

Ce 6.6

Nd 6.4

1.1-1.2 1.1-1.2

Ac

Th 5.5 1.3

1.1-1.2

Pa 5.6 1.5

U III 4.3 IV 5.4 1.7

Sm 6.2

Eu 6.2

Gd 6.0

Tb 5.9

Dy 5.8

Ho 5.8

Er 5.7

Lu 5.5

1.1-1.2 1.1-1.21.1-1.2 1.1-1.2 1.1-1.21.1-1.2 1.1-1.2 1.1-1.2 1.1-1.2 1.1-1.2

Np Pu 5.4 5.4 1.3 1.3

15.1 Elements M formed M(BH4)n, their gravimetric hydrogen densities in the unit of mass% (middle) and the Pauling electronegativity χP of M (bottom).2 Asterisks indicate compounds stabilized at room temperature by co-ordination with ligands. Brackets indicate compounds which have been reported to be unstable at room temperature but may be isolated at lower temperature.3–5

air (1.7% in 48 days).19 It is therefore better to handle and to store MBH4 (M = Li, Na and K) in an inert gas atmosphere. The first report of pure alkali metal borohydride appeared in 1940 by Schlesinger and Brown20 who synthesized LiBH4 by the reaction of ethyllithium with B2H6. The direct reaction of the corresponding metal/metal hydrides with diborane in etheral solvents under suitable conditions produces high yields of the borohydrides:3,4,21–24 2MH + B2H6 → 2MBH4 (M = Li, Na, K)

[15.1]

Industrially, NaBH4 is an important starting material for the production of other borohydrides. Thus far, over 100 methods of preparation for NaBH4 have been described. Industrial production of NaBH4 is produced by the reaction of extremely fine sodium hydride with trimethyl borate in high boiling hydrocarbon oil at 523–553 K:25,26 4NaH + B(OCH3) 3 → NaBH4 + 3NaOCH3

© 2008, Woodhead Publishing Limited

[15.2]

422

Table 15.1 Density of tetrahydroborate, for crystal structures reported elsewhere Density (g/mol)

Density* (g/cm3)

Hydrogen density (mass%)

LiBH4 NaBH4 KBH4 RbBH4 CsBH4 Be(BH4)2 Mg(BH4)2

16949-15-8 16940-66-2 13762-51-1 20346-99-0 – 17440-85-6 16903-37-0

21.78 37.83 53.94 100.31 147.75 38.70 53.99

18.5 10.7 7.5 4.0 2.7 20.8 14.9

122.1 114.5 87.8 76.8 65.3 146.0 147.4

Ca(BH4)2 Al(BH4)3

17068-95-0 16962-07-5

69.76 71.51

11.6 16.9

(124.1) 133.5

Zr(BH4)4

12370-59-1

150.6

0.66 1.07 1.17 1.92 2.42 0.702 (0.61) 0.989 (0.78 LT) (0.76 HT) (1.07) 0.79 (0.72 LT) (0.69 HT) 1.179 (1.24)

10.7

126.2

Hf(BH4)4

53608-70-1

237.6

1.65 (2.00)

6.8

112.2

Hydrogen density (kg/m3)

T m† (K)

‡ ∆H boro (kJ/mol)

541 778 858 873d 873d 398 593d

–194 –191 –229

533d 208.5 317.5b 301.7 396b 302 396b

–302w –131wo

* Density in brackets shows the calculated value from the crystal structure. † Melting temperature Td. d and b represent decomposition and boiling temperatures, respectively.6–10 ‡ Heat of formation ∆Hboro of M(BH4)n. w and wo show predicted ∆H by first-principles calculation with and without zero point energy.11–15 LT = low temperature, HT = high temperature phase. © 2008, Woodhead Publishing Limited

–398wo –376wo

Solid-state hydrogen storage

CAS No.

Borohydrides as hydrogen storage materials

423

Another process developed by Bayer uses finely ground borosilicate glass, sodium and hydrogen:27 Na2B4O7 + 7SiO2 + 16Na + 8H 2 → 4NaBH4 + 7Na2SiO3

[15.3]

NaBH4 is extracted from the borohydride–silicate mixture with liquid ammonia under pressure. LiBH4 and KBH4 are produced on an industrial scale by the following metathesis reaction:28,29 NaBH4 + LiX → LiBH4 + NaX (X = Cl, Br)

[15.4]

NaBH4 + KOH → KBH4 + NaOH

[15.5]

The borohydride is more soluble whereas NaX/NaOH is less soluble to appropriate solvent. For example, the solubility of LiBH4 is 44 g/100 g ethanol at 298 K, while that of NaBr is small (2.2 g/100 g ethanol). Therefore borohydride can be extracted from the mixed products in Eqs (15.4) and (15.5).

15.2.2 Alkaline-earth borohydrides, trivalent and tetravalent borohydrides Alkaline-earth borohydrides, trivalent and tetravalent borohydrides, M(BH4)n (M = alkaline-earth, Groups 3~14 in the periodic table, n = valence of M) are synthesized by direct metal hydrides with diborane similar to (15.1) or the metathetical reaction as follows: nLiBH4 + MXn → M(BH4)n + nLiX

[15.6]

nNaBH4 + MXn → M(BH4)n + nNaX

[15.7]

Here, X = halogen, such as F, Cl, Br, I. These metathetical reactions occur with certain elements of M, Groups 2 (magnesium, etc.),30–36 Groups 4 (titanium, etc.),9 Groups 13 (aluminum, etc.),37 other transition metals,38,39 rare-earth40 and actinide metal elements.41 The elements known to form M(BH4)n are shown in Fig. 15.1.3–5 Recently, M(BH4)n (M = Mg, Ca~Mn, Zn, Al, Y, Zr and Hf; n = 2–4) were systematically synthesized by mechanical milling following equations (15.6) and (15.7).11–13 The reactions of Eqs 15.6) and (15.7) occur by mechanical milling and/or in solution. The starting material of LiBH4 (15.6) is of great advantage to a mechanical milling process in solid state due to similar ionic radius between Li+ and Mn+ comparing to Na+.12 In the case of solvent reaction, the starting materials should be selected by considering the solvilities of starting metals and products. Sometimes Al(BH4)3 is used42 as a starting material instead of LiBH4 or NaBH4 in the synthesis reactions of Eqs (15.6) or (15.7). Although various reports confirm the progressions of metathetic reactions of (15.6) and (15.7),43 the preparation of pure M(BH4)n without nLiX/nNaX

© 2008, Woodhead Publishing Limited

424

Solid-state hydrogen storage

and solvate is less easy. Usually, since the solubility of M(BH4)n is higher than that of precipitated nLiX/nNaX, the precipitation (nLiX/nNaX) can be removed from solution by filtration. The solvent is then removed from M(BH4)n solution by evacuation at an appropriate temperature. The solubility of starting materials, MBH4 (M = alkali metal) is summarized in Table 15.2.19 In protic solvents, solvolysis occurs for NaBH4, and hydrogen is released as follows: NaBH4 + 4ROH → NaB(ROH)4 + 4H2

[15.8]

Here, ROH is for example, methanol or ethanol, etc. Because NaBH4 decomposes 80% after 1 h in methanol at 273 K, 6% 1 h in ethanol,18 excess NaBH4 was used for the synthesis reaction. For example, Konoplev and Bakulina44 have synthesized Mg(BH4)2 by reaction of NaBH4 and MgCl2 in molar ratio 2.7~3.8:1 in diethyl ether (2:1 stoichiometry). After progression of reaction (15.7), precipitated 2NaCl was filtered, then Mg(BH4)2 ether solution was evacuated at 453 K in order to remove diethyl ether. However, in the case of decomposition temperature of M(BH4)n is lower than the temperature of the solution removed, the preparation of solvent-free M(BH4)n is difficult. Methods of direct synthesis for Mg(BH4)2 and Ca(BH4)2 by heating triethylamine adduct of the borohydride have been reported by Ch łopek et al.45 and Riktor et al.46 in which the previous synthesis methods of Mg(BH4)2 were also described in detail. M(BH4)n (M = Zr, Hf9 and Al42) with a significant vapor pressure at temperatures below which the product thermally decomposes, can be purified by considering the vapor pressure and temperature. The vapor pressures of Zr(BH4)4, Hf(BH4)4 and Al(BH4)3 are shown in Table 15.3. For example, pure Zr(BH4)4 was isolated from the products of reaction (15.6), i.e. mixture of Zr(BH4)4 and 4LiCl, because Zr(BH4)4 sublimed whereas 4LiCl remained in the container by evacuation using a rotary pump (under approximately Table 15.2 Solubility of MBH4 (M = alkali metal) in various solvents (g/100 g solvent at 298 K) Solvent Water Methanol Ethanol Isopropylamine Diethyl ether THF Diglyme Toluene Ammonia DMF a f

Tb of solvent (K) 373.0 310.7 351.5 307.0 309.0 338.0 435 384 239.7 426

LiBH4 29.0 dec. 44 8c 4.3 22.5

a

insol. 77

NaBH4

KBH4

55 16.4b 4.0b 6.0d insol. 0.1e 5.5 insol. 104 18.0f

19.3 0.7 0.25

273 K, slow decomposition. b293 K, decomposition. c283 K. d301 K. e293 K. Dangerous decomposition possible at higher temperature.19

© 2008, Woodhead Publishing Limited

insol. insol. insol. insol. 20 15f

Borohydrides as hydrogen storage materials

425

Table 15.3 Vapor pressure of Zr(BH4)4, Hf(BH4)4 and Al(BH4)3,9,42 Temperature (K)

273 283 293 303 313 323

Pressure (mmHg) Hf(BH4)2

Zr(BH4)2

Al(BH4)3

2.2 4.7 14.9 21.0 36.0 56.4

1.8 4.2 15.0 20.3 33.2 52.2

119.5

10–3 mmHg) at room temperature, below the decomposition temperature of Zr(BH4)4. The sublimed Zr(BH4)4 was trapped in the cold trap at the liquid nitrogen temperature (77 K).10

15.2.3 Multi-cation borohydrides Multi-cation borohydrides, MM′(BH4)n (M = alkaline metals, M′ = other metals) have been synthesized by the following reactions:47 MCln + mLiBH4 → MLim–n(BH4)m + nLiCl MCln + mNaBH4 → MNam–n(BH4)m + nNaCl

[15.9] [15.10]

Thus far, LiZn(BH4)3,16 NaZn(BH4)3, Li2Cd(BH4)4, LiFe(BH4)3, LiNi(BH4)3, [2KBH4]3Zn(BH4)2,16,48 Li2Ni(BH4)4,16,49 etc., have been reported. There are many kinds of multi-cation borohydrides with coordinated solvents.50 Moreover, multi-anion compounds, such as Li3BN2H8 (i.e. Li3(BH4)(NH2)2),51–56 Li 4 BN 3 H 10 (i.e. Li 4 (BH 4 )(NH 2 ) 3 ), [Ni(NH 3 ) 6 ](BH 4 ) 2 , 57 and 58–60 have been reported. Those borohydrides might be [Al(NH3)6](BH4)3 useful for controlling the thermodynamic stabilities, similar to the conventional ‘alloying’ method for hydrogen storage materials.47

15.3

Structure of borohydrides

15.3.1 Crystal structure Alkali borohydrides, MBH4 (M = Li, Na, K, Rb and Cs) The crystal structures of alkali borohydrides MBH4 (M = Li, Na, K, Rb and Cs) are summarized in Fig. 15.2 and Table 15.4. The crystal structure of LiBH4 was reported to be space group of Pcmn at room temperature.65 However, the structure was revised to be Pnma at room temperature by means of recent synchrotron X-ray powder diffraction measurement.61,66 LiBH 4 transforms from orthorhombic Pnma to hexagonal P6 3mc at approximately 380 K, at which (BH4)– tetrahedra are reoriented along the c-

© 2008, Woodhead Publishing Limited

426

Solid-state hydrogen storage

c c b a

LiBH4 Pnma (No. 62)

c a a NaBD4 < 190 K KBD4 < 197 K P42/nmc (No. 137)

a a LiBH4 > 380 K P63mc (No. 186)

a a

a (Occupancy of H/D is 0.5.)

NaBH4 KBD4 RbBD4 CsBD4 Fm-3m (No. 225)

15.2 Crystal structures of alkali borohydrides MBH4 (M = Li, Na, K, Rb and Cs). Large, middle, small circles show alkali, boron, hydrogen atoms, respectively.

axis. The detail of the hexagonal structure as a high-temperature phase is still under discussion.67 Moreover, the structure at high pressure up to 4.5 GPa has been reported in the literature.68,69 MBH4 (M = Na,62,63,70 K,64,71 Rb64 and Cs64) shows NaCl-type structure at ambient conditions. The lattice constants and distances between M and B increase with increasing ionic radius of M, i.e. increasing of atomic number of M (see Table 15.4). However, the distances between B and H, i.e. size of (BH4)– tetrahedra, are almost the same. NaBH4 and KBH4 show structure transition at low temperature; they crystallized in tetragonal P42/nmc below 190 K and 70–65 K, respectively.62,71,72 NaBH4 shows the structure transition to tetragonal P421C with lattice constants a = 4.0864(1) Å, c = 5.5966(7) Å and orthorhombic Pnma with a = 7.33890(1) Å, c = 5.6334(5) Å at 6.3 and 8.9 GPa, respectively.72 No structural transition was observed in RbBH4 and CsBH4 down to 1.5 K. Alkaline-earth borohydrides, M(BH4)2 (M = Be, Mg and Ca) The crystal structures of alkaline-earth borohydrides M(BH4)2 (M = Be, Mg and Ca) are summarized in Fig. 15.3 and Table 15.5. The structure of Be(BH4)2 was investigated by single crystal X-ray diffraction study. The crystal structure is tetragonal with space group of I41cd with lattice constants a = 13.62(1) and c = 9.10(1) Å.35

© 2008, Woodhead Publishing Limited

Table 15.4 Crystal structure of alkali tetrahydroborate MBH4 (M = Li, Na, K, Rb and Cs) LiBH4 RT

LiBH4 HT

NaBD4 LT

NaBH4 RT

KBD4 LT

KBD4 RT

RbBD4 RT

CsBD4 RT

Structure

Orthorhombic

Hexagonal

Tetragonal

Cubic

Tetragonal

Cubic

Cubic

Cubic

Space group

Pnma No. 62

P63mc No. 186

P42/nmc No. 137

Fm-3m No. 225

P42/nmc No. 137

Fm-3m No. 225

Fm-3m No. 225

Fm-3m No. 225

Lattice constants (Å)

a = 7.178 58(4) a = 4.276 31(5) b = 4.436 86(2) c = 6.948 44(4) c = 6.803 21(4)

a = 4.332 c = 5.949

a = 6.1506(6) a = 4.6836(1) c = 6.5707(2)

a = 6.7074(1) a = 7.0156(1)

a = 7.4061(2)

M–B distances 2.521(2) (Å)

2.4963(10)

2.9474(5)

3.0753(3)

3.311 81(4)

3.353 70(4)

3.507 80(4)

3.703 05(9)

B–H distances (Å)

1.03(2) 1.250(13) 1.286(12)× 2

1.27(2) 1.29(2) 1.293(9)× 2

1.2217(3)× 2 1.2217(8)× 2

1.17078(6)× 4 1.205(2)× 4

1.195(2)× 4

1.207(2)× 4

1.216(2)× 4

Temperature

R.T. (< 380 K)

408 K

10 K ( 180~195 K Pna21 (No. 33)

Al(BH4)3 (α) < 180~195 K C2/c (No. 15)

a a a

Zr(BH4)4 Hf(BH4)4 P4-3m (No. 215)

15.4 Crystal structures of trivalent and tetravalent borohydrides, M(BH4)3 (M = Al) and M(BH4)4 (M = Zr and Hf). Large, middle, and small circles show M, boron, hydrogen atoms, respectively.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

431

Table 15.6 Crystal structure of trivalent and tetravalent borodydrides, Al(BH4)3 and M(BH4)4 (M = Zr and Hf) Borohydride

Al(BH4)3 LT(α)

Al(BH4)3 HT(β)

Zr(BH4)4 < RT

Hf(BH4)4 < RT

Structure

Monoclinic

Orthorhombic

Cubic

Cubic

Space group

C2/c No. 15

Pna21 No. 33

P4-3m No. 215

P4-3m No. 215

Lattice constants (Å)

a = 21.917 b = 5.986 c = 21.787 β = 111.90°

a = 18.021 b = 6.138 c = 6.199

a = 5.86

a = 5.827(4)

M–B distances

2.084 63

2.052 70

2.3354(7)

2.281(11)

(Å) B–H distances (Å)

1.166 23 1.169 80 1.230 91 1.242 85

1.148 03 1.229 56 1.234 91 1.163 29

1.209 86(0) 1.272 04(0)× 3

1.15(3) 1.23(2)× 3

113 K

110 K

81

82

83

Temperature

T < 180–195 K

Ref.

80

15.4 and Table 15.6. Both phases are made up of discrete molecular Al(BH4)3 units in which three boron atoms form a planar AlB3 skeleton and are bonded to an Al atom by two bridging hydrogen atoms. This molecular-like structure might be originated in low melting temperature. Recently, the structures of trivalent borohydrides with rare-earth, R(BH4)3 have been investigated.84 The crystal structures of Zr(BH4)476,82 and Hf(BH4)483 were determined by means of single crystal X-ray structural analyses. Because the melting points of Zr(BH4)4 and Hf(BH4)4 are 301.7 and 302 K, respectively, the measurements were conducted at 110–113 K. They crystallize in the cubic symmetry with space group of P4-3m. The crystal structures and parameters are shown in Fig. 15.4 and Table 15.6. In these structures, four (BH4)– complexes are tetrahedrally arranged around Zr4+/Hf4+ atoms, where three of four hydrogen atoms in a (BH4)– complex form a bridging bond with Zr4+/ Hf4+ as shown in Fig. 15.4. Since Hf(BH4)4 with P4-3m symmetry has a soft mode theoretically,11 it might have a structure transition. Through freezing the soft mode, the structure with reduced symmetry of P23 has been theoretically predicted.

15.3.2 Electronic structure Alkali borohyrides, MBH4 (M = Li, Na, K, Rb and Cs) The electronic structures of borohydrides MBH4 were investigated by firstprinciples calculations. Figure 15.5 depicts the total and partial densities of states and valence charge contour plot for orthorhombic LiBH4.85 The electronic

© 2008, Woodhead Publishing Limited

432 40

Solid-state hydrogen storage Total

30 20 10

DOS (states/eV unit-cell)

0

H

s p

Li

1 H 0

B

s p

B

6

H

3 0

Li H

Li H

s

B (b)

6 3 0 –8 –6

–4 –2 0 2 4 Energy (eV) (a)

6

8

10

15.5 (a) Total and partial densities of states and (b) valence charge contour plot for orthorhombic LiBH4 in the (010) plane.85 Reprinted with permission from Miwa et al.85 Copyright 2004 American Physical Society.

structure is nonmetallic with the calculated energy gap of 6.8–7.0 eV.86 Because there is little contribution of Li orbitals to the occupied states, Li atoms are thought to be ionized as Li+ cations. The occupied states split into two peaks: the low-energy states are composed of B-2s and H-1s orbitals and the highenergy states consist of B-2p and H-1s orbitals. A boron atom constructs sp3 hybrids and forms covalent bonds with surrounding four H atoms. The charge from the extra electron needed to form these bonds is compensated by a Li+ cation. This character is also confirmed experimentally by synchrotron Xray diffraction measurement and maximum entropy method (MEM) analysis. The stable structures except for orthorhombic phase67,87,88 and surface electronic structure of LiBH489,90 were also investigated theoretically. The calculated total densities of state (DOSs) for MBH4 (M = Li, Na, K, Rb and Cs) are displayed in Fig. 15.6.86 All these compounds have finite energy gaps between the valence and conduction bands. Hence, they are proper insulators with estimated band gap between 5.5 (CsBH4) and 7.0 eV (LiBH4). The large band gap is a feature which the MBH4 series has in common with the MAlH4 and MGaH4 series (all members of the latter series have band gaps of ~5 eV92). The calculated band gap of the MAlH4 and

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

433

EF 4

CsBH4

2 0 4

RbBH4

DOS (states eV1)

2 0 4

KBH4

2 0 4

NaBH4

2 0 4

LiBH4

2 0

–8

–6

–4

–2

0 2 Energy (eV)

4

5

8

15.6 Calculated ground-state total DOS for MBH4 (M = Li, Na, K, Rb and Cs). Fermi level is set at zero energy and marked by the vertical dotted line; occupied states are shaded.86 Reprinted with permission from Vajeeston et al.86 Copyright 2005 Elsevier.

MGaH4 series appears to be invariant whereas that for MBH4 series varies almost linearly with the size of M+. Also the width of the bands (in particular the valence band) varies nearly linearly with the size of M+. The bandwidth narrowing from LiBH4 to CsBH4 reflects the enhanced M–B and M–H bond strength along the series.86 Alkaline-earth borohydrides, trivalent and tetravalent borohydrides, M(BH4)n (n ≥ 2) Figure 15.7(a) depicts the densities of states for borohydrides M(BH4)n (n ≥ 2) and those for MBH4 for reference.11,93 The electronic structures are nonmetallic with relatively large energy gaps of 2.9–6.8 eV. The occupied states mainly consist of B-sp3 hybrids and H-1s orbitals as was observed in LiBH4. The contribution of metal atoms to the valence states are little except for the filled 3d states found in Zn(BH4)2. These figures support an ionic picture for the interaction between metal cations and (BH4)– anions.

© 2008, Woodhead Publishing Limited

434

Solid-state hydrogen storage

9

LiBH4

6 3 0

CuBH4

6 3 0

NaBH4

6 3

DOS (states/eV formula unit)

0

KBH4

12 6 0

Zr(BH4)4

24 12 0

Hf(HB4)4

24 12 0

Mg(BH4)2

12 6 0

Zn(BH4)2

12 6 0

Sc(BH4)3

18

DOS (states/eV f.u.)

9 0 –10 –8 –6 –4 –2 0 2 4 Energy (eV) (a) 36

6

8 10

6

8 10

27 18 9 0 –10 –8 –6 –4

–2 0 2 4 Energy (eV) (b)

© 2008, Woodhead Publishing Limited

15.7 (a) Total and partial densities of states DOS for M(BH4)n (n ≥ 2) and those for MBH4 for reference. The light grey parts indicate the partial DOS of boron and hydrogen atoms, and the dark grey parts show the partial DOS of metal atoms. The origins of the energies are set to be the top of valence states except for CuBH4, where the energy is shifted by +3.5 eV for clear comparison.11 Reprinted with permission from Nakamori et al.11 Copyright 2006 American Physical Society. (b) Electronic density of states for α-Al(BH4)3. The origin of energy is set to be the top of valence states. The energy positions of the occupied states for molecular Al(BH4)3 are indicated by solid circles for comparison purpose.93 Reprinted with permission from Miwa et al.93 Copyright 2007 Elsevier.

Borohydrides as hydrogen storage materials

435

Figure 15.7(b) depicts the electronic density of states for α-Al(BH4)3.93 The energy gap was to be 6.0 eV. The five occupied states consist of several sharp peaks whose energy positions correspond well to those of molecular Al(BH4)3. The similar correspondence can be found for β-Al(BH4)3. These also suggest the weak interaction between Al(BH4)3 molecules in the solid phases. The electronic structures indicate that the charge transfer from metal cations Mn+ to (BH4)– anions is a key feature for the stability of borohydrides, M(BH4)n.

15.4

Dehydrogenation and rehydrogenation reactions

15.4.1 General reactions Alkali borohydrides Overall dehydrogenation reactions of alkali borohydrides are described as follows: MBH4 → MH + B + 3/2H2

[15.11]

The dehydrogenation reaction of LiBH4 liberates three of the four hydrogen (13.8 mass%) at above melting temperature (541 K) and decomposes into LiH and B.94 Owing to the high stability of LiH, its dehydrogenation reaction proceeds at temperatures above 900–950 K.91 Therefore, it is usually not accessible in technical applications. The similar dehydrogenation pathways were observed for M = Na and K, and their dehydrogenation properties are summarized in Table 15.7. It should be noted that the dehydrogenation temperatures of alkali borohydrides increase with increasing the atomic number of M. The systematic understanding and control of thermodynamic stabilities will be explained in Section 15.4.2. Here, we focus on the dehydrogenation and hydrogenation reactions of LiBH4 having higher gravimetric hydrogen density and lower dehydrogenation temperature among alkali borohydrides. The thermal dehydrogenation spectrum of LiBH4 exhibits four endothermic peaks. The peaks are attributed to a polymorphic transformation around 383 K, melting at around 540–553 K, the dehydrogenation (50% of the hydrogen was released around 763 K), and the dehydrogenation of three of the four hydrogen atoms at 953 K, at which only the third peak (dehydrogenation reaction) is pressure dependent. As shown in Fig. 15.8, the dehydrogenation process of LiBH4 at very low heating rate (0.5 K/min) exhibits three distinct dehydrogenation peaks. This is an indication that the dehydrogenation mechanism involves several intermediate steps. Theoretical predictions have suggested that Li2B12H12107 or LiBH15 exist as an intermediate phase during the dehydrogenation reaction of LiBH4.

© 2008, Woodhead Publishing Limited

436

Hydrogen (mass%) —————————————————— Ideal Obs Obs (dehyd) (rehyd)

Condition: temp. Td(K) ∆H d* (Pressure (MPa)) (kJ/mol H2) —————————————–1st dehyd Rehyd

References

LiBH4 → LiH + B + 3/2H2 NaBH4 → NaH + B + 3/2H2 KBH4 → KH + B + 3/2H2 Mg(BH4)2 → MgH2 + 2B + 3H2

13.9 8.0 5.6 11.2

13.5 – – ~11

5–10 – – 2.4

453–773 838 857 570–673

873–923 (7–35) – – 543 (40)

6–8,66,67,85,110,111 6–8 6–8 95–99,114

Ca(BH4)2 → 2/3CaH2 + 1/3CaB6 + 10/3H2 Ca(BH4)2 → CaH2 + 2B + 3H2

9.6

9.4

7.7

650

8.7

9.0



620

Reaction

* ∆Hd is the enthalpy change of the dehydriding reaction. Calculated value. The others are experimental values.

+

© 2008, Woodhead Publishing Limited

673 (70)

52–76 66 92+ 39.5, 57, 40+ 32+

14,96



40+

100

Solid-state hydrogen storage

Table 15.7 Hydrogen storage properties of M(BH4)n (M = alkali and alkaline metals)

Borohydrides as hydrogen storage materials

437

T (K) 400

500

600

4

Differential for 0.5 K/min

15

3 10

2

5

H/LiBH4

m(H2) (mass%)

700

1 0.5 1

2

4

6 K/min

0 100

200

300 T (°C)

400

0 500

15.8 Integrated thermal decomposition from LiBH4 measured at various heating rates of 0.5–6 K/min. The differential curve for the lowest heating rate of 0.5 K/min is given together with the hypothetical compositions at the peak maxima. The inset shows a possible structure of an intermediate phase from Ohba et al.107 The large, middle, and small spheres denote Li, B, and H atoms, respectively. Reprinted with permission from Züttel et al.111. Copyright 2007 Elsevier. Inset reprinted with permission from Ohba et al.107. Copyright 2006 American Physical Society.

Experimentally, ex situ Raman spectroscopy has delivered some evidence for the existence of Li2B12H12 composed of Li+ and (B12H12)2– ions at above 700 K, as an intermediate phase.108 Recent results of in situ synchrotron radiation powder X-ray diffraction (PXD) and solid state magic angle spinning nuclear magnetic resonance (CP/MAS NMR) indicated that two new phases were observed during dehydrogenation of LiBH4: phase I with hexagonal symmetry and phase II with orthorhombic symmetry in the temperature range ~473–573 K, and ~573–673 K, respectively109 (These phases might be due to moisture contamination and reaction with the sample vessel. The probable causes for these extra phases is being investigated). Using intermediate phase106 having appropriate stability might be useful for controlling the enthalpy change for dehydrogenation reaction, as similar to the other complex hydrides such as alanate (NaAlH4 → Na3AlH6) and amide (LiNH2 → Li2NH) (see Chapters 14 and 16, respectively). After the dehydrogenation reaction of LiBH4, the end-products, lithium hydride (LiH) and boron, were rehydrogenated at 873 K under 35 MPa for 12 hours110 or at 1000 K under 15 MPa111 for over 10 hours to form LiBH4. The rehydrogenation reaction was confirmed by Raman spectroscopy and PXD measurement. The modified lithium borohydrides, LiBH475% + TiO225%

© 2008, Woodhead Publishing Limited

438

Solid-state hydrogen storage

and LiBH4 + 0.2MgCl2 + 0.1TiCl3, were hydrogenated 8 mass% and 5 mass% of hydrogen respectively at 873 K and 7 MPa, respectively.112 The result is in agreement with the claim in Goerrig’s patent about the direct formation of LiBH4 from the elements.113 Alkaline-earth borohydrides The dehydrogenation properties of alkaline-earth borohydrides M(BH4)2 are also shown in Table 15.7. Mg(BH4)2 was dehydrogenated according to the following reaction:44,45,73,74,95–98,114 Mg(BH4)2 → MgH2 + 2B + 3H2

[15.12]

The thermal desorption spectrum of Mg(BH4)2 prepared in the solvent exhibited three endothermic peaks.30 The peaks were reported to be attributed to a polymorphic transformation at 453 K (see Section 15.3.1), melting and dehydrogenation reaction simultaneously at approximately 586 K (corresponding to Eq. (15.12): 11.2 mass%), and dehydrogenation reaction of MgH2 + 2B into Mg + 2B+H2 (3.3 mass%) at 683 K. The existence of intermediate phase 99 or amorphous MgH 2 98 was observed during dehydrogenation reaction of Mg(BH4)2. The dehydrogenation reaction for Ca(BH4)2 was predicted as follows:14 Ca(BH4)2 → 2/3CaH2 + 1/3CaB6 +10/3H2

[15.13]

The theoretical gravimetric density of the effective hydrogen is 9.6 mass%. The thermogravimetric analysis for prepared Ca(BH4)2 shows that the total mass loss up to 800 K is 9.2 mass%, which supports the overall reaction of Eq. 15.13.14 On the other hand, the in situ X-ray diffraction analysis indicated that the dehydrogenation reaction of Ca(BH4)2 is as follows: Ca(BH4)2 → CaH2 + 2B +3H2

[15.14] 100

in which 8.7 mass% of hydrogen was released theoretically. The existence of an unknown intermediate phase during the dehydrogenation reaction was also suggested.45,46 Detailed investigations are required to clarify the reaction pathways for the dehydrogenation reaction of Ca(BH4)2. For both Mg(BH4)2 and Ca(BH4)2, partial rehydrogenation reactions were reported to proceed at above 543 K under 40 MPa for 48 hours99 and at above 623 K under 10 MPa for 2 weeks, 101 respectively. For example, the rehydrogenated reaction for the sample of dehydrogenated Mg(BH4)2 proceeded at above 543 K under 40 MPa for 48 hours,99 at which approximately 6.1 mass% of hydrogen (including 3.7 mass% from Mg to MgH 2) was rehydrogenated. The reported rehydrogenation reaction was not fully reversible, but appropriate intermediate phase such as MgB12H12 might play important role for rehydrogenating reaction.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

439

Trivalent and tetravalent borohydrides It has been reported that CuBH4, Zn(BH4)2 and Cd(BH4)2 decompose as follows:102 CuBH4 → CuH + 1/2B2H6 (at 273 K)

or

[15.15]

Zn(BH4)2 → Zn + 2B + 4H2 (at 358 K)

[15.16a]

Zn(BH4)2 → Zn + B2H6 + H2

[15.16b]

Cd(BH4)2 → Cd + 2B + 4H2 (at 298 K)

[15.17]

Assuming Zn(BH4)2 and Cd(BH4)2 release only hydrogen, 8.5 mass% and 5.7 mass% of hydrogen were released, respectively. However, it has been reported that Zn(BH4)2 also releases diborane11–13,103–105 and the decomposition reaction proceeds slowly at room temperature. Mn(BH4)2 also decomposed to diborane.11–13,115 Attempts to reduce diborane have been made by addition of chlorides, hydrides and transition metals.104,105,115 Al(BH4)3, Ti(BH4)3, Zr(BH4)4 and Hf(BH4)4 are sublimate borohydrides with low melting temperatures. The melting and boiling temperatures are shown in Table 15.8, and the vapor pressure is shown in Table 15.3. It has been reported that these M(BH4)n decompose very slowly to form hydrogen and non-volatile compounds of indefinite composition. The decomposition reaction of Al(BH4)3 was reported to be as follows:11–13,119 2Al(BH4)3 s Al2B4H18 + B2H6

[15.18]

Ti(BH4)3 decomposed completely after several days at room temperature.9 Some 13–14% of Ti(BH4)3 decomposed to form diborane and the rest decomposed to form hydrogen. Table 15.8 Melting and boiling temperatures of M(BH4)n

LiBH4 NaBH4 KBH4 RbBH4 CsBH4 Be(BH4)2 Mg(BH4)2 Al(BH4)3 Zr(BH4)2 Hf(BH4)2

Melting temperature (K)

Boiling temperature (extrapolated) (K)

541 778 858

653* 838* 857* 873* 873* 364.3 593* 317.5 396 391

593 208.5 301.7 302

* Decomposition temperature.

© 2008, Woodhead Publishing Limited

440

Solid-state hydrogen storage

The decomposition of Zr(BH4)4 and Hf(BH4)4 has been shown to take place principally in the vapor phase. 9 It has also been reported that Ti(BH4)3, Zr(BH4)4 and Hf(BH4)4 have been used for making diboride as follows:116–118 Zr(BH4)4 → ZrB2 + B2H6 + 5H2 (at 1173 K)

[15.19]

On the other hand, it has been reported that Zr(BH4)4 releases only hydrogen,11–13 which is inconsistent with Eq. (15.19). Further investigation is necessary to clarify the reaction pathway of these borohydrides M(BH4)n (n ≥ 3). Another important research for dehydrogenation and rehydrogenation is for multicomponent (or mixed) system. For example, the dehydrogenation and rehydrogenation reactions of complex hydride–metal hydride e.g. M(BH4)n–M′Hm have been recently investigated: 2LiBH4 + MgH2 s 2LiH + MgB2 + 4H2

[15.20]

2NaBH4 + MgH2 s 2NaH + MgB2 + 4H2

[15.21]

Ca(BH4)2 + MgH2 s CaH2 + MgB2 + 4H2

[15.22]

in which 11.4 mass%, 7.8 mass% and 8.3 mass% of hydrogen can be stored, respectively.120–123 The reactions are reversible. Details are shown in Chapter 17.

15.4.2 Controlling thermodynamics The thermodynamic stabilities for the series of metal borohydrides M(BH4)n (M = Li, Na, K, Mg, Ca, Sc, Zr, Hf, Cu, Zn and Al; n = 1–4) have been systematically investigated by first-principles calculations.11–15,67 The heat of formation of M(BH4)n ∆Hboro were estimated from the difference of the total energies between the left- and right-hand sides of Eq. (15.23): 1/nM + B + 2H2 → 1/nM(BH4)n

[15.23]

In order to compare the stability of borohydrides composed of the metal cations M which have different valencies, ∆Hboro values were normalized by the number of BH4 tetrahedra in the formula unit. As shown in Section 15.3.2, the bonding character between Mn+ and (BH4)– in M(BH4)n is ionic, and the charge transfer from Mn+ to (BH4)– is responsible for the stability of M(BH 4)n. The ability of the charge transfer can be measured by the electronegativity. The predicted heat of formation, ∆Hboro, as a function of the Pauling electronegativity of M,2 χP, is plotted in Fig. 15.9.11–13,15 A good correlation between ∆Hboro and χP can be found. Therefore the Pauling electronegativity is a useful indicator to estimate thermodynamic stability of borohydrides. Assuming a linear relation, the least square fitting yields:15

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials 100 50

441

CuBH4

Observed structure Assumed structure

Al(BH4)3 Hf(BH4)4

–50

Mg(BH4)2 –100 –150

Zr(BH4)4 Sc(BH4)3 100

NaBH4

–200 –250 0.6

∆Hd (kJ/mol H2)

∆Hboro (kJ/mol BH4)

Zn(BH4)2 0

Ca(BH4)2 LiBH4

KBH4

0.8

1.0

1.2

1.4

M(BH4)n M=K Na Li 50 Ca Mg Hf Zr Sc 0 1.0 1.5 2.0 1.6

1.8

2.0

χP

15.9 Predicted heat of formation, ∆Hboro, as a function of the Pauling electronegativity of M, χP . Inset shows the enthalpy change of the dehydrogenation reaction ∆Hd as a function of χP. Adapted with permission from Nakamori et al.11 and Miwa et al.93 Copyright 2006 American Physical Society and 2007 Elsevier.

∆Hboro = 253.6 χP – 398.0

[15.24]

with an absolute mean error of 9.6 kJ/mol BH4. The Pauling electronegativities of M are shown at the bottom of Fig. 15.1. To estimate the enthalpy changes of the dehydrogenation reaction of borohydride ∆Hd, not only the stability of borohydrides (i.e. Fig. 15.9 and Eqs. (15.23 and 15.24)), but also the stability of products should be taken into account. The reported dehydrogenation equations shown in Section 15.4.1 were used for estimating ∆Hd. However, the dehydrogenation reactions of some borohydrides have not yet been reported. In these cases, the dehydrogenation reaction was assumed as following equation: [15.25] M (BH 4 ) n → MH m + nB + 4 n – m H 2 2 In the case of unknown MHm, direct decomposition into elements was assumed. Then, the enthalpy change of the dehydrogenation reaction ∆Hd was estimated using predicted ∆Hboro and reported values of ∆Hproduct:124 ∆Hd = ∆Hboro – ∆Hproduct

[15.26]

The inset of Fig. 15.9 shows the ∆Hd as a function of Pauling electronegativity of M, χP. Interestingly, a correlation between ∆Hd and χP was also observed. Furthermore, it is expected that M(BH4)n with χP ≥ 1.5 are thermodynamically unstable.

© 2008, Woodhead Publishing Limited

442

Solid-state hydrogen storage

The dehydrogenation properties were experimentally investigated by thermal desorption analysis during the heating process for milling samples.11–13 The dehydrogenation temperature Td is defined as the temperature of the first peak. In Fig. 15.10, Td values are plotted as a function of the Pauling electronegativity χP. A good correlation between dehydrogenation temperature Td and the Pauling electronegativity χP is also observed experimentally, indicating that Td can be approximately estimated by considering χP as an indicator. (The same correlation between Td and χP has been also reported for M(AlH4)n.96) The inset of Fig. 15.10 shows the observed Td as a function of the predicted ∆Hd, in which good correlation is observed. However, observed Td (closed circles) of M(BH4)n in the thermal decomposition experiment during heating process are higher than those predicted from ∆Hd (closed triangles), probably due to poor reaction kinetics of M(BH4)n. The component of released gas is also important. M(BH4)n for M = Al, Mn and Zn release diborane in addition to hydrogen.11–13 On the other hand, the other M(BH4)n release only hydrogen, though the released diborane from M = V and Cr undeniable because of the smaller amount of desorbed gas comparing to accuracy of apparatus. Therefore, it was concluded that M(BH4)n 1000

1000

M(BH4)n M = Na Li 800

Td (K)

Ca

600

Sc Ti V

200 –20 Y

Zr Cr Ca

Li

Na

0 20 40 60 ∆Hd (kJ/molH2)

80

Mg

500

Sc Zr

400 300

Ca Mg

Y

600 400

700

Li Na

800

Td (K)

900

Ti* 0.8

1.2

V*,Cr* 1.6

2.0

χp

15.10 Dehydrogenation temperature Td as a function of the Pauling electronegativity χp. The inset shows the correlation between observed Td (closed circles) and estimated enthalpy change for the dehydrogenation reaction ∆Hd. The predicted Td at 0.1MPa hydrogen (closed triangles) are plotted for M = Li, Na and Ca, which are calculated using van’t Hoff equation ln(Peq/P0) = (∆Hd/RT) – (∆S/R), predicted ∆Hd and ∆S = 130 J/molK. ∗ = Small amount of released gas. Adapted with permission from Nakamori. et al.11–13 Copyright 2006 American Physical Society and 2007 Elsevier.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

443

with χP ≥ 1.5 is unstable for hydrogen storage, which is consistent with theoretical prediction. (On the other hand, some reports have state that even stable borohdyride such as LiBH4 and Mg(BH4)2 release borane.45 More detaileds investigations are required to clarify this issue.) The dehydrogenated temperature Td of MM′(BH4)n is also related to the average χp value of MM′.47 The appropriate combination of cations might be an effective method to adjust the thermodynamic stability of metal borohydrides, similar to the conventional ‘alloying’ method for hydrogen storage alloys. Recently, ∆Hd was also controlled by changing the reaction pathway using appropriate composite of M(BH4)n – M(AlH4)m, M(BH4)n – M(NH2)m and M(BH4)n – MHm. Details are given in Chapter 17.

15.4.3 Promoting kinetics In the dehydrogenation and rehydrogenation reactions of borohydrides, not only hydrogen but also other elements should be diffused, as shown in Fig. 15.11. This is not the same as conventional metal hydride alloy in which diffusion of hydrogen and expansion of lattice for alloy is dominant. Generally, the fast diffusion of elements is very difficult in the solid state at much lower than the melting temperature. Therefore, the observed Td during heating process seems to become higher than those predicted, as shown in the inset of Fig. 15.10. Moreover, higher hydrogen pressure (generally more than 10 MPa) is required for rehydrogenation of M(BH4)n than that predicted from ∆Hd (Fig. 15.12). This is why a high temperature is required for diffusion of elements, at which high hydrogen pressure should be applied for rehydrogenation. Therefore, solving the kinetic problem is one of the important areas of research in the use of complex hydrides as hydrogen storage materials. In order to improve the reaction kinetics, ball milling and additive effects (dopant) were investigated. The addition of SiO2 to LiBH4, lowers the dehydrogenated temperature by approximately 100 K, which is, according to recent studies of the isotherms, due to a catalytic effect and not a thermodynamic effect.125 Au and co-workers126,127 have reported the effect of addition of metal, chloride and/or oxide to LiBH4. The most effective material was LiBH4 + 0.2MgCl2 + 0.1TiCl3 which starts releasing 5 mass% of hydrogen LiBH4



LiH

+

B

+

3/2H2

Li Amorphous B H

B

Li H

H2

15.11 Schematic drawing of dehydriding and hydriding reactions of borohydride, LiBH4.

© 2008, Woodhead Publishing Limited

444

Solid-state hydrogen storage 500

Temp. (K) 400 350 300

250

1000 LaNi5

Peq (MPa)

100

10

1

0.1

LaBH4

NaBH4

0.01 1.0

2.0

3.0 1/Temp. (K–1)

4.0 × 10–3

15.12 Van’t Hoff plots of MBH4, which are calculated using the van’t Hoff equation ln(peq/P0) = (∆Hdes/RT) – (∆S/R), predicting ∆Hd and ∆S = 130 kJ/mol K. That of LaNi5H6 is shown for reference.

at 333 K and can be rehydrogenated to 4.5 wt% at 873 K and 7 MPa. Li et al. have reported that TiCl3 is effective for Mg(BH4)2 among various Ti compounds as an additive.114 Barkhordarian et al. have reported that the kinetic barriers for the formation of LiBH4, NaBH4 and Ca(BH4)2 are drastically reduced when MgB2 is used instead of B as the starting material for the hydrogenation reaction, in which reaction enthalpies are reduced by about 10 kJ/mol H2 at the same time.123 Although the effects of the kinetics and thermodynamics of additives to M(BH4)n should be considered separately, the additives are effective for borohydrides M(BH4)n that are similar to alanates M(AlH4)n. Another method for improving the reaction kinetics is the enhancement of atomic diffusion by using an electromagnetic field such as microwaves. Recently it was reported that the LiBH4 hexagonal phase at a high temperature above approximately 380 K was heated to 1000 K rapidly, and more than 13 mass% of hydrogen was released by microwave irradiation128,129 due to Li super-ionic conductivity.130 In order to achieve phase transition and for the dehydrogenation reaction to proceed for shorter time, a microwave absorber, e.g. B and C, was added to LiBH4. The composite releases more than 6 mass% of hydrogen by microwave irradiation for less than 3.5 min.131 Because microwaves can penetrate nonmetallic materials such as M(BH4)n and M(AlH4), the microwave heating process might be useful for complex hydrides having low thermal conductivity as a trigger for dehydrogenation.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

15.5

445

Future trends

Borohydrides M(BH4)n belong to a class of materials with the highest gravimetric hydrogen densities. In this chapter, the dehydrogenation and hydrogenation reaction of borohydrides M(BH4)n by controlling temperature and pressure were presented. The crystal structures of some M(BH4)n have been reported, though those of other M(BH4)n and intermediate phases have being investigated. The determination of those structures and finding the correlation between structures and hydrogen storage properties are important for future development. Thermodynamic stability for M(BH4)n can be systematically predicted by considering the electronegativity of M. Based on the correlation, one can expect that M(BH4)n with appropriate stability and high hydrogen density. The investigation of M(BH4)n with appropriate stability (di- ~ tetra-valence M) is of great importance in the development of hydrogen storage material. To achieve fast reaction kinetics is another important research direction. For this purpose, finding an appropriate additive (dopant) and using a new heating system such as an electromagnetic field might be useful. Also using an intermediate phase and considering a low melting temperature might provide a useful way for promoting the reaction kinetics and achieving the reversible reaction under moderate conditions.

15.6

Acknowledgements

The authors would like to acknowledge gratefully co-workers and colleagues from within and beyond their research groups that have been involved with the group’s hydrogen storage related research over the last several years, but most notably Drs H.-W. Li, T. Sato, Ms N. Warifune, EMPA, Prof. A. Züttel, and Toyota R&D Labs, Dr K. Miwa, Ms N. Ohba, Dr M. Aoki, Mr T. Noritake, and Dr S. Towata under NEDO project.

15.7

References

1. Wegner P A, Ed. (1973) Boron Hydride Chemistry, Academic Press, New York, Chapter 12, Nido-Heteroboranes. 2. Pauling L (1960), The Nature of the Chemical Bonds, 3rd ed. Cornell University Press, New York, 19. 3. James B D, Wallbridge M G H (1959) Prog. Inorg. Chem., 1, 99. 4. Marks T J, Kolb J R (1977) Chem. Rev., 77, 26. 5. Grochala W, Edwards P P (2004) Chem. Rev., 104, 1283. 6. SCI Finder Scholar: ‘Hazardous Substances Data Bank’, National Library of Medicine (US). 7. CRC Handbook of Chemistry and Physics, 83rd ed., 2002–2003. 8. Orimo S, Jensen C M, Nakamori Y, Eliseo J, Züttel A (2007) Chem. Rev., 107, 4111.

© 2008, Woodhead Publishing Limited

446

Solid-state hydrogen storage

9. Hoekstra H R, Katz J J (1949) J. Am. Chem. Soc., 71, 2488. 10. Volkov V V, Myakishev K G, Yugov S I (1975) J. Appl. Chem. USSR, 48, 2184. 11. Nakamori Y, Miwa K, Li H-W, Ninomiya A, Ohba N, Towata S, Züttel A, Orimo S (2006) Phys. Rev. B, 74, 045126. 12. Nakamori Y, Li H-W, Kikuchi K, Aoki M, Miwa K, Towata S, Orimo S (2007) J. Alloys Comp., 446–447, 296. 13. Nakamori Y, Li H-W, Matsuo M, Miwa K, Towata S, Orimo S, J. Phys. Chem. Solids, doi: 10.1016/j.jpcs.2008.04.017. 14. Miwa K, Aoki M, Noritake T, Ohba N, Nakamori Y, Towata S, Züttel A, Orimo S (2006) Phys. Rev. B, 74, 155122. 15. Kang J K, Kim S Y, Han Y S, Muller R P, Goddard III W A (2005) Appl. Phys. Lett., 87, 111904. 16. Nöth H (1961) Angew Chem., 73, 371. 17. Semenenko K N, Kravchenko O V, Polyakova V N (1973) Russ. Chem. Rev., 42, 1. 18. James B D, Wallbridge M G H (1970) Prog. Inorg. Chem., 11, 99. 19. Hawkins S, Ravenscroft M, Schulz G (1985) Ullmann’s Encyclopedia of Industrial Chemistry vol. A13, Edited by Elvers B, VCH, Weinheim, Deerfield Beach, FL. 20. Schlesinger H J, Brown H C (1940) J. Amer. Chem. Soc., 62, 3429. 21. Schlesinger H J, Brown H C, Hoekstra H R, Rapp L R (1954) J. Amer. Chem. Soc., 75, 199. 22. Brown H C, Mead E J, Subba B C (1955) J. Amer. Chem. Soc., 77, 6209. 23. Chamberlain D L (Apr. 10, 1962) US Patent, 3,029,128. 24. Nöth H, Beyer H (1960) Chem. Ber., 93, 928. 25. Schlesinger H J, Brown H C (1953) J. Am. Chem. Soc., 75, 186. 26. Schlesinger H J, Brown H C (1945) US Patent, 2,534,533. 27. Büchner W, Niederprüm H (1977) Pure Appl. Chem., 49, 733. 28. Brown H C, Choi Y M, Narasimhan S (1981) Inorg. Chem., 20, 4456. 29. Fedor W S, Banus M D, Ingalls D P (1957) Ind. Eng. Chem., 49, 1664. 30. Konoplev V N, Bakulina V M (1971) Rus. Chem. Bull., 20, 136. 31. Wiberg E, Bauer R (1950) Z. Naturforsch, 5b, 397. 32. Wiberg E, Bauer R (1952) Z. Naturforsch, 7b, 58. 33. Wiberg E, Nöth H, Hartwimmer R (1955) Z. Naturforsch, 10b, 292. 34. Burg A B, Schlesinger H I (1940) J. Am. Chem. Soc., 62, 3425. 35. Marynick D S, Lipscomb W N (1972) Inorg. Chem., 11, 820. 36. Varin R A, Chiu Ch, Wronski Z S, (2008) J. Alloys Comp., 462, 201. 37. Schlesinger H I, Brown H C, Hyde E K (1953) J. Am. Chem. Soc., 75, 209. 38. Monnier G (1955) Bull. Soc. Chim. Fr., 1138. 39. Aubry J, Monnier G (1955) Bull. Soc. Chim. Fr., 482. 40. Mirsaidov U, Badalov A, Khikmatov M, Nazarov K, Marufi V K (1991) Russ. J. Phys. Chem., 65, 1543. 41. Banks R H, Edelstein N M, Rietz R R, Templeton D H, Zalkin A (1978) J. Am. Chem. Soc., 100, 1957. 42. Schlesinger H I, Brown H C, Finholt A E (1953) J. Amer. Chem. Soc., 75, 209. 43. Klingen T (1964) Inorg. Chem, 3, 1058. 44. Konoplev V N, Bakulina V M (1971) Rus. Chem. Bull., 20, 136. 45. Ch łopek K, Frommen C, Léon A, Zabara O, Fichtner M (2007) J. Mater. Chem., 17, 3496. 46. Riktor M D, Sørby M H, Ch łopek K, Fichtner M, Buchter F, Züttel A, Hauback B C (2007) J. Mater. Chem., 17, 4939.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials

447

47. Li H-W, Orimo S, Nakamori Y, Miwa K, Ohba N, Towata S, Züttel A (2007) J. Alloys Comp., 446–447, 315. 48. Hagenmuller P, Rault M (1959) Compt. Rend., 248, 2758. 49. Wiberg E, Henle W (1952) Z. Naturforsch, 7B, 582. 50. Makhaev V D, Borisov A P, Lobkovskii E B (1984) Rus. J. Inorg. Chem., 29, 666. 51. Pinkerton F E, Wicke B G (2004) Ind. Phys., 10, 20. 52. Pinkerton F E, Meyer M S, Meisner G P, Balogh M P (2006) J. Phys. Chem. B, 110, 7967. 53. Meisner G P, Scullin M L, Michael P, Balogh M P, Pinkerton F E, Meyer M S (2006) J. Phys. Chem. B, 110, 4186. 54. Aoki M, Miwa K, Noritake T, Kitahara G, Nakamori Y, Orimo S, Towata S (2005) Appl. Phys. A, 80, 1409. 55. Nakamori Y, Ninomiya A, Kitahara G, Aoki M, Noritake T, Miwa K, Kojima Y, Orimo S, (2006) J. Power Sources, 155, 447. 56. Noritake T, Aoki M, Towata S, Nakamori Y, Orimo S, (2006) Appl. Phys. A, 83, 277. 57. Mikheeva V I, Maltseva N N, Sterlyadkina J K (1965) Russ. J. Inorg. Chem., 10, 1292. 58. Bird P H, Wallbridge M G H (1967) J. Chem. Soc. A, 664. 59. Fauroux J C, Teichner S J (1967) Russ. J. Inorg. Chem., 10, 4053. 60. Plešek J, Hermanec S (1966) Collect. Czeck. Chem. Commun., 31, 3845. 61. Soluilé J-Ph, Ranaudin G, Cermy R, Yvon K (2002) J. Alloys Comp., 346, 200. 62. Fischer P, Züttel A (2004) Mater. Sci. Forum, 433–444, 287. 63. Kumar R S, Cornelius A L (2005) Appl. Phys. Lett., 87, 261916. 64. Renaudin G, Gomes S, Hagemann H, Keller H, Yvon K (2004) J. Alloys Comp., 375, 98. 65. Harris P M, Meibohm E P (1947) J. Am. Chem. Soc., 69, 1231. 66. Züttel A, Rentsch S, Fischer P, Wenger P, Sudan P, Mauron Ph, Emmenegger Ch (2003) J. Alloys Comp., 356, 515. 67. Lodziana Z, Vegge T, (2004) Phys. Rev. Lett., 93, 145501. 68. Pistorius C W F T, Phys Z (1974) Chem. Neue. Folge., 88, 253. 69. Talyzina A V, Anderssona O, Sundqvista B, Kurnosovb A, Dubrovinskyb L (2007) J. Solid State Chem., 180, 510. 70. Soldate A M (1947) J. Am. Chem. Soc., 69, 987. 71. Abrahams S C, Kalnajs K (1954) J Chem. Phys, 22, 434. 72. Davis R L, Kennard C H L (1985) J. Solid State Chem., 59, 393. 73. Černý R, Filinchuk Y, Hagemann H, Yvon K (2007) Angew. Chem. Int. Ed., 46, 1. 74. Her J H, Stephens P W, Gao Y, Soloveichik G L, Rijssenbeek J, Andrus M, Zhao J C (2007) Acta Cryst. B, 63, 561. 75. Vajeeston P, Ravindran P, Kjekshus A, Fjellvåg H (2006) Appl. Phys. Lett., 89, 071906. 76. Kedrova N S, Mal’tseva N N (1977) Russ. J. Inorg. Chem., 22, 973. 77. Li H-W, Kikuchi K, Nakamori Y, Ohba N, Miwa K, Towata S, Orimo S (2008) Acta Cryst., 56, 1342. 78. Semenenko K N, Kravchen O V, Lobkovsk E B (1972) J. Struct. Chem. (USSR), 13, 540. 79. Aldridge S, Blake A J, Downs A J, Gould R O, Parsons S, Pulham C R (1997) J. Chem. Soc. Dalton Trans., 1007. The full list of the structural parameters for α-

© 2008, Woodhead Publishing Limited

448

80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98.

99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109.

110.

Solid-state hydrogen storage and β-Al(BH4)3 is available at the Cambridge Crystallographic Data Center (www.ccdc.cam.ac.uk) with the deposition number 186/404. Cambridge Crystallographic Data Center: CCDC No. 230830. Cambridge Crystallographic Data Center: CCDC No. 230829. Bird P, Churchill M (1967) Chem. Com., 74, 403. Broach R W, Chuang I S, Marks T J, Williams J M (1983) Inorg. Chem., 22, 1081. Sato T, Miwa K, Nakamori Y, Ohoyama K, Li H-W, Noritake T, Aoki M, Towata S, Orimo S (2007) Phys. Rev. B, communication. Miwa K, Ohba N, Towata S, Nakamori Y, Orimo S (2004) Phys. Rev. B, 69, 245120. Vajeeston P, Ravindran P, Kjekshus A, Fjellvåg H (2005) J. Alloys Comp., 387, 97. Frankcombe T J, Kroes G-J (2006) Phys. Rev. B, 73, 174302. Frankcombe T J, Kroes G-J, Züttel A (2005) Chem. Phys. Lett., 405, 73. Ge Q (2004) J. Phys. Chem. A, 108, 8682. Du A J, Smith S C, Lu G Q (2006) Phys. Rev. B, 74, 193405. Montgomery C D (1973) Nucl. Eng. Des., 25, 309. Vajeeston P, Ravindran P, Vidya R, Kjekshus A, Fjellvåg H (2004) Cryst. Growth Des., 4, 471. Miwa K, Ohba N, Towata S, Nakamori Y, Züttel A, Orimo S (2007) J. Alloys Comp., 446–447, 310. Stasinevich D S, Egorenko G A (1968) Russ. J. Inorg. Chem., 13, 341. Konoplev V N, Bakulina V M (1971) Izv. Akad. Nauk. SSR Ser. Khim., 1, 159. Matsunaga T, Buchter F, Miwa K, Towata S, Orimo S, Züttel A (2008) Renewable Energy, 33, 193. Matsunaga T, Buchter F, Mauron P, Bielman M, Nakamori Y, Orimo S, Ohba N, Miwa K, Towata S, Züttel A, (2008) J. Alloys Comp., 459, 583. Zhao J-C, Andrus M, Cui J, Gao Y, Lemmon J, Raber T, Rijssenbeek J, Rubinsztajn G, Soloveichik G, Lightweight Intermetallics for Hydrogen Storage, US DOE, Washington DC, 2006, http://www.hydrogen.energy.gov/pdfs/progress06/ iv_a_4c_zhao.pdf. Li H-W, Kikuchi K, Nakamori Y, Miwa K, Towata S, Orimo S (2008) Acta Materialia, 56, 1342. Kim J-H, Jin S-A, Shim J-H, Cho Y W, (2008) J. Alloys Comp., 461, L20. Ronnebro E et al., MRS Fall Meeting, 2007, Boston. Wiberg E (1953) Angew. Chem., 65, 16. Jeon E, Cho Y (2006) J. Alloys Comp., 422, 273. Srinivasan S, Escobar D, Jurczyk M, Goswami Y, Stefanakos E, (2008) J. Alloys Comp., 462, 294. Nakagawa T, Ichikawa T, Kojima Y, Fujii H (2007) Mater. Trans., 48, 556. Kolmogorov A N, Drautz R, Pettifor D G (2007) Phys. Rev. B, 76, 184102. Ohba N, Miwa K, Aoki M, Noritake T, Towata S, Nakamori Y, Orimo S (2006) Phys. Rev. B, 74, 075110. Orimo S, Nakamori Y, Ohba N, Miwa K, Aoki M, Towata S, Züttel A (2006) Appl. Phys. Lett., 89, 21920. Mosegaard L, Møller B, Jørgensen J E, Bösenberg U, Dornheim M, Hanson J C, Cerenius Y, Walker G, Jakobsen H J, Besenbacher F, Jensen T R (2007) J. Alloys Comp., 446, 301. Orimo S, Nakamori Y, Kitahara G, Miwa K, Ohba N, Towata S, Züttel A (2005) J. Alloys Comp., 404–406, 427.

© 2008, Woodhead Publishing Limited

Borohydrides as hydrogen storage materials 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121.

122. 123. 124. 125. 126. 127. 128. 129. 130. 131.

449

Züttel A, Borgschulte A, Orimo S (2007) Scripta Materialia, 56, 823. Au M, Spencer W, Jurgensen A, (2008) J. Alloys Comp., 462, 303. Goerrig D (Dec. 27, 1958) German Patent 1,077,644. Li H-W, Kikuchi K, Nakamori Y, Miwa K, Towata S, Orimo S (2007) Scripta Materialia, 57, 679. G. Sevra et al., International Symposium on Metal–Hydrogen Systems, 2006, Hawaii. Telle J, Roucka R, Tsong I S T, Ritter C, Crozier P A, Chizmeshya A V G, Kouvetakis J (2003) Appl. Phys. Lett., 82, 2398. Rice G W, Woodin R L (1988) Commun. Amer. Ceramic Soc., 71, C181. Hough R L (1966) AIAA J., 4, 107. Ogg R A, Ray J D (1955) Diss. Faraday Soc., 19, 293. Vajo J J, Skeith S L, Mertens F (2005) J. Phys. Chem. B, 109, 3719. Bösenberg U, Doppiu S, Mosegaard L, Barkhordarian G, Eigen N, Borgschulte A, Jensen T R, Cerenius Y, Gutfleisch O, Klassen T, Dornheim M, Bormann R (2007) Acta Materialia, 55, 3951. Dornheim M, Doppiu S, Barkhordarian G, Boesenberg U, Klassen T, Gutfleisch O, Bormann R (2007) Scripta Materialia, 56, 841. Barkhordarian G, Klassen T, Dornheim M, Bormann R (2007) J. Alloys Comp., 440, L18. Manchester F D, Ed. (2000) Phase Diagrams of Binary Hydrogen Alloys, ASM International, http://hydpark.ca.sandia.gov/MaterialsFrame2.html Mauron P, Züttel A. unpublished. Au M, Jurgensen A (2006) J. Phys. Chem. B, 110, 7062. Au M, Jurgensen A, Zeigler K (2006) J. Phys. Chem. B, 110, 26482. Nakamori Y, Orimo S, Tsutaoka T (2006) Appl. Phys. Lett., 82, 112104. Nakamori Y, Matsuo M, Yamada K, Tsutaoka T, Orimo S (2007) J. Alloys Comp., 446–447, 698. Matsuo M, Nakamori Y, Maekawa H, Takamura H, Orimo S, Tsutaoka T (2007) Appl. Phys. Lett., 91, 224103. Matsuo M, Nakamori Y, Yamada K, Orimo S (2006) Appl. Phys. Lett., 90, 232907.

© 2008, Woodhead Publishing Limited

16 Imides and amides as hydrogen storage materials D. H. G R E G O R Y, University of Glasgow, UK

16.1

Introduction

A genuine breakthrough in hydrogen storage materials research, by Chen et al. at the University of Singapore, occurred in 2002 [1]. Not only was the Li–N–H system groundbreaking in terms of creating a reversible storage system that readily met the gravimetric and volumetric criteria of the US Department of Energy (DoE), but here was a revolutionary new system of ‘chemical hydrides’ that was neither based solely around metals and alloys nor depended on high surface area non-metals for the physisorption of hydrogen. In fact, the system was composed of both metallic and non-metallic elements in chemically distinct phases that were interconvertible by the making and breaking of non-metal (N)–hydrogen bonds. The ramifications of this discovery were immediate, but it soon became apparent that many of the challenges that had confronted the previous generations of complex hydrides – reversibility, kinetics, operating (desorption) temperatures – would be similarly challenging for the new nitrides, imides and amides. The following sections catalogue how swiftly the field has moved and ultimately where it may move to if these materials are to be viable in terms of automotive and mobile applications. First, however, this chapter considers the underpinning chemistry of the nitrides, imides and amides and a history of hydrogen interaction and inclusion that perhaps surprisingly, significantly predates this decade’s exciting developments.

16.2

The lithium–nitrogen–hydrogen system

Alkali metal nitride chemistry is effectively confined to that of lithium and sodium [2], and dominated by lithium nitride, Li3N, a compound that has long fascinated solid-state chemists since early reports of its synthesis and characterisation [3–5]. In the ternary phase system, Li–N–H, lithium also forms three stoichiometric ternary hydrogen-containing compounds, the imide, Li2NH, the amide, LiNH2, and the nitride hydride, Li4NH. The first 450 © 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

451

two contain protonic hydrogen H+ while the last contains both nominal N3– and H–. Li4NH forms from the reaction of Li3N with lithium hydride [6] and decomposes to the imide, Li2NH plus either LiH or Li3N when heated at c. 400 °C in hydrogen or nitrogen respectively. However, it is the imide and amide that, to date, have proved significant in the process of hydrogen storage in compounds containing both lithium and nitrogen. The following parts of this section will discuss the history and the chemistry of the main protagonists of the Li–N–H storage system; Li3N, Li2NH and LiNH2.

16.2.1 Lithium nitride and hydrogen: a historical perspective Lithium nitride, Li3N, exists in three forms. The alpha polymorph (α-Li3N) is stable at room temperature and pressure and is synthesised from the elements at elevated temperature; in actuality, this reaction can begin to occur under ambient conditions (Eq. 16.1): 6Li + N2 → 2Li3N

[16.1]

β-Li3N (with the Na3As structure) is obtained from the α-form at moderate pressure (4.2 kbar at 300 K) [7–9] whereas the gamma polymorph (γ-Li3N; isostructural with Li3Bi) [10] transforms completely from β-Li3N between 35–45 GPa and remains stable up to 200 GPa [11]. The first structural determination of α-Li3N was achieved by Zintl and Brauer [5] and the structure redetermined by Rabenau and Schultz [12] several decades later (Fig. 16.1). The unique structure consists of hexagonal layers of planar lithium hexagons centred by nitrogen. Each NLi6 hexagon in the [Li2N] layer is capped above and below the ab plane by a further lithium ion, connecting the layers in the third dimension (along c). Hence, each nitrogen atom is coordinated by a total of eight lithium atoms in hexagonal bipyramidal geometry. Lithium atoms within [Li2N] planes (Li(2)) are in a trigonal planar coordination geometry with N, whereas those between planes (Li(1)) are linearly coordinated to nitrogen. α-Li3N is an indirect band gap semiconductor (c. 2.1 eV) and an Li+ ion conductor with exceptionally fast Li+ ion transport. The latter phenomenon arises from cationic vacancies within the [Li2N] planes (considered in another context below). Over the decades following its initial characterisation, significant progress was made in elucidating links between the unique structure of α-Li3N and its observed physical properties. Aspects of this research are described in much more detail in other reviews [2, 13, 14]. The history of the reaction chemistry of lithium nitride with hydrogen extends back to the early part of the 20th century and a number of works performed by Dafert and Miklauz [15–17]. Initially the action of hydrogen

© 2008, Woodhead Publishing Limited

452

Solid-state hydrogen storage

Li(2) N Li(1)

16.1 Polyhedral representation of the structure of α-Li3N showing the vertex linking of NLi8 hexagonal bipyramids along the c-axis. The small spheres represent Li, the polyhedra are centred by N atoms.

on lithium nitride at 220–250 °C was found to produce ‘Li3NH4’ which subsequently decomposed to ‘Li3NH2’ and hydrogen at higher temperatures (or could be obtained directly from Li3N + H2 above 700 °C). Subsequent investigations by Ruff and Georges and the above authors [18, 17] then revealed ‘Li3NH4’ and ‘Li3NH2’ to be mixtures of LiNH2 + 2LiH and Li2NH + LiH respectively; the median and end-points of the Li–N–H storage cycle developed c. 90 years later [1]. The first substantial evidence for hydrogen inclusion in lithium nitride itself came at the end of the 1970s from both diffraction and spectroscopy. X-Ray diffraction (XRD) demonstrated that the lithium position within the [Li2N] planes is underoccupied by 1–2 %. [19, 20] and similar levels of lithium vacancies were confirmed later by powder neutron diffraction (PND) [21]. The likely charge balancing species for these vacancies is H+and a proposed disordered doping of small amounts of imide, (NH)2– for N3– on the nitrogen sites in the [Li2N] plane. This is evident in both ‘pure’ and intentionally H-doped Li3N via infrared (IR) spectroscopy [22–24]. These studies also indicate a correlation of increasing ionic conductivity with increasing H content (and therefore (NH)2– concentration) and hence a possible relationship between Li+ and H+ mobility. 7Li solid-state nuclear magnetic resonance (NMR) evidence has lent further weight to these suggestions [20, 25–27]. The best structural model to date for hydrogen (deuterium) doped lithium nitride is for Li2.95ND0.02 containing disordered ND2– anions and retaining the parent space group P6/mmm (Fig. 16.2) [28].

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

453

Li(2) D

Li(1)

N

16.2 Structure of Li3–x–yNDy (y ≈ 0.02). The deuterium doped nitride retains the space group of Li3N (P6/mmm) with D (small stippled spheres) disordered in the 6j position (0.19(4)% occupied) [28]. Lithium atoms are depicted as small striped spheres, N as larger spheres.

16.2.2 Lithium imide and lithium amide Lithium imide, Li2NH, forms via the decomposition of lithium amide at high temperature under vacuum (Eq. 16.2) or by the action of gaseous ammonia on lithium hydride (Eq. 16.3) [29, 30]: 2LiNH2 → Li2NH + NH3

[16.2]

2LiH + NH3 → Li2NH + 2H2

[16.3]

The antifluorite structure of the imide – isotypic to Lithia (Li2O) – was determined in 1951 by Juza and Opp [31]. This original description defines a simple cubic unit cell (space group Fm3m) with suggested orientationally disordered (NH)2– groups (Fig. 16.3a), a model which has been challenged more recently in the wake of the increased interest in the Li–N–H system [32–34]. These newer models suggest a number of possibilities. Initially, two alternative cubic solutions were proposed in space groups Fm3m (derived from Juza and Opp’s model) and F 43 m , both containing disordered (NH)2– [32, 33]. In the former, hydrogen occupies 1/12 of the 48h (x, x, 0) sites whereas in the latter it occupies 1/4 of the 16e (x, x, x) sites. Subsequently, refinement against sub-ambient PND data showed that the data could be fit to a doubled cubic unit cell in space group Fd 3 m with hydrogen fully occupying the 32e (x, x, x) site (i.e. an ordered distribution of hydrogen in the imide units; Fig. 16.3b [34]. Moreover, a disordered antifluorite phase was not observed below 350 K and a reversible phase transition was identified by differential scanning calorimetry (DSC). In fact, this order– disorder transition corresponds to one found in differential thermal analysis

© 2008, Woodhead Publishing Limited

454

Solid-state hydrogen storage

N H N Li

Li

c a b (a)

(b)

16.3 Postulated structures of Li2NH: (a) the simple cubic anti-fluorite structure with orientationally disordered (NH)2– groups (H atoms not shown); (b) the ‘doubled’ anti-fluorite ‘low-temperature’ structure with an ordered distribution of hydrogen in the imide units.

(DTA) work performed over three decades earlier (Tdis = 353(3)K) and, although at this time the structure of the phases on either side of the transition were not determined [35], combinations of 7Li and 1H solid-state NMR experiments by Haigh et al. clearly showed different characteristics of the imide above and below this transition temperature [36]. Above Tdis the Li+ ions are highly mobile and the protons rotate freely whereas at much lower temperatures (below Tdis), the Li+ ions do not diffuse readily and the hydrogens are immobile such that the N–H bond is oriented along the face diagonal of the anti-fluorite unit cell. Further evidence for the existence of (NH)2– disorder emanates from infrared spectroscopy, where in stark contrast to the sharp bands observed for lithium amide, LiNH2, the signature N–H stretches at 3180 cm–1 and 3250 cm–1 in lithium imide show significant broadening [37]. Alternative ordered models for Li2NH arise from density functional theory (DFT) calculations. These approaches lead almost unanimously to lower symmetry, predominantly orthorhombic, models with lower calculated ground state energies [34, 38–40]. Currently, the lowest energy structure is one in orthorhombic space group Pbca (a = 5.12 Å, b = 10.51 Å, c = 5.27 Å) in which the N, H and two crystallographically distinct Li atoms occupy 8c general (x, y, z) positions [40]. The lithium atoms in this model have moved significantly from their ideal anti-fluorite positions and the N–H units adopt preferred orientations with respect to their nearest neighbours. The authors of this and the previous studies, however, are quick to acknowledge that lower ground state energy structures may yet exist. Further, they also acknowledge that the experimentally observed ambient temperature structure may be metastable (i.e. kinetically stable). This, in fact, is not unexpected given the way that the materials are prepared and then characterised. Nevertheless, the calculated enthalpies of formation for the orthorhombic

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

455

ordered structures are in reasonable agreement with that determined experimentally (e.g. ∆Hf,298 = –184.1 kJ mol–1 from [39] vs. ∆Hf,298 = –222 kJ mol–1 experimentally [41]). The distinctions between structures and the correlation to ∆Hf (and more specifically the link to N–H orientation and ordering), of course, become of major importance when evaluating and potentially modifying the thermodynamics of the nitride–imide–amide system (with respect to storage) as will be seen in Section 16.2.4. The preparation of lithium amide was first reported by Titherley [42]. It is synthesised by the exothermic reaction (∆H = –130 kJ mol–1) between lithium metal and gaseous or liquid ammonia (Eq. 16.4) [29, 30]: 2Li + 2NH3 → 2LiNH2 + H2

[16.4]

The resulting ionic compound is a white solid which melts between 373 and 375 °C (and decomposes to the imide and ammonia above this upper limit as noted above, eventually decomposing to lithium nitride). Experimentally, the standard enthalpy of formation for lithium amide, ∆Hf, is –176 kJ mol–1 [41]. By contrast to lithium imide, the crystal chemistry of lithium amide is relatively non-contentious. Originally solved from single crystal XRD data by Jacobs and Juza in 1972 (after a partial solution, in which hydrogens were not located, 21 years earlier [31]), the structure of lithium amide is an ordered fluorite derivative crystallising in tetragonal space group I 4 with nitrogen in a distorted cubic close packed arrangement and lithium in tetrahedral interstices (Fig. 16.4) [43].

N

Li(3)

Li(2)

Li(1)

16.4 Tetragonal structure of lithium amide. Nitrogen (light spheres) is arranged in a distorted cubic close packed arrangement with lithium (dark spheres) in tetrahedral interstices.

© 2008, Woodhead Publishing Limited

456

Solid-state hydrogen storage

Although the more complete (1972) solution defines three lithium sites (2a, 2c, 4f) as opposed to two (4e, 4f) from the partial model, 7Li solid-state NMR spectra show resonances corresponding to two inequivalent lithium environments [36]. Lithium amide exhibits two sharp bands in its IR spectrum corresponding to N–H stretches at 3260 cm–1 (symmetric) and 3315 cm–1 (asymmetric), respectively [37]. A subsequent study of the deuterated amide, LiND2, by PND defined N–D(H) bond lengths that were approximately 0.2 Å longer than those previously reported by XRD [44] and more recent PND studies confirmed the longer N–H(D) distances (of c. 0.97 Å and 0.98 Å) [45, 46]. One consequence of the newer neutron models is an increased calculated band gap (3.48 vs 3.2 eV) [45, 47]. An enthalpy of formation for LiNH2 was not calculated, but there are major implications in terms of the increased N–H bond lengths and doping strategies that might seek to exploit destabilising the amide bonds for hydrogen release. A point of interest concerns the large displacement parameters observed for the lithium and hydrogen (deuterium) positions [46]. Similarly large values were found in the original PND investigation [44] which would suggest both that the effect is a real one and perhaps one related to lithium mobility and hydrogen (deuterium) disorder respectively.

16.2.3 Hydrogen storage in the Li–N–H system The possibility of including or substituting hydrogen within lithium nitride became of more than academic interest with the revelation in 2002 that a process was plausible which both allowed for extremely competitively high levels of hydrogen to be stored and for this storage to take place reversibly. That such a process had not been considered as viable previously probably owed a great deal to prior knowledge of lithium amide decomposition and the fact that normally such a process evolves gaseous ammonia (Eq. 16.2). In fact, the synergy between components of the Li–N–H storage system was to prove crucial to its success and in the subsequent design of more complex amide and imide storage materials. The storage process uncovered by Chen et al. closely follows that originally observed by Dafert and Miklauz [15–17] and Ruff and Georges [18] almost a century earlier and involves the two-step process shown in Eq. (16.5) [1]: Li3N + 2H2 ↔ Li2NH + LiH + H2 ↔ LiNH2 + 2LiH

[16.5]

The theoretical gravimetric capacity of this system is thus 11.5 wt% for 4 moles of H atoms to 1 mole of starting Li3N. In practice, the amount of stored hydrogen from nitride through to amide was found to be somewhat less, with uptake of 9.3 wt% and c. 10 wt% from gravimetric and volumetric measurements respectively. Taking the standard enthalpies of formation [41], the overall heat of the reaction in Eq. (16.5) was calculated to be

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

457

–161 kJ mol–1. In fact, only the second step, cycling between imide and amide, is readily reversible and hydrogen corresponding to desorption from imide to nitride could be removed only above 320 °C at 10–5 mbar. Hence, the second step of the Li–N–H reaction cycle can be used to reversibly store hydrogen below 300 °C and by reacting lithium imide with hydrogen as shown in Eq. (16.6): Li2NH + H2 ↔ LiNH2 + LiH

[16.6]

it becomes possible to store theoretically c. 7.0 wt% of hydrogen (with favourable thermodynamics, ∆H = –45 kJ mol–1). In fact, it was found from pressure–composition (P–C) isothermal measurements to be possible to store c. 6.5 wt% of hydrogen reversibly in lithium imide at 255 °C and above. Many of the studies that followed the initial report of reversible storage in the Li–N–H system followed two key reaction steps: (1) the desorption step where lithium amide combines with lithium hydride to release hydrogen and (2) the absorption step where Li3N initially uptakes hydrogen from gaseous H2. Considering the former first, the interplay between LiH and LiNH2 was considered in a drive to understand why a mixture of these components should begin to release H2 at 150 °C when individually the former releases H2 only at 500 °C and the latter normally decomposes to imide and ammonia (Eq. 16.5). Chen et al. considered the mechanism of this process via isotopic exchange experiments with LiNH2 and LiD [48]. A mechanism was tested whereby it was proposed that Hδ+ from the amide and H(D)δ– from the hydride might directly combine to form H2 (i.e. predominantly labelled HD when using LiD if the premise was correct). In fact, the evolved gas contained not only HD but also D2 and, predominantly, H2. This suggested that some D was retained in the solid phase (within the imide) following desorption and therefore that a solid–solid mechanism could not account for the isotopic distribution in the desorbed gas. The first evidence that the reaction might be gas mediated then followed in the work of Ichikawa et al. who observed that contrary to earlier experiments, the desorption reaction between LiNH2 and LiH leads to the evolution of significant (and readily detectable) amounts of ammonia when the reactants are hand ground and reacted in a 1:1 ratio [49]. Interestingly, however, the quantity of evolved ammonia decreases radically when either (a) LiNH2 and LiH are reacted in a 1:2 ratio (as they are in the Li3N–Li2NH–LiNH2 cycle) or (b) the hydride and amide are ball milled (i.e. intimately mixed). Shortly afterwards, similar observations were made by Hu and Ruckenstein who noted that rapid reactions occur between LiH and gaseous NH3 such that effectively the hydride ‘captures’ the gaseous ammonia [50]. This was demonstrated in an elegant but simple experiment in which two samples made up of layers of 1:2 molar ratio LiNH2 and LiH powders were prepared; one with LiH as the initial layer, the other with LiNH2 so-arranged [51].

© 2008, Woodhead Publishing Limited

458

Solid-state hydrogen storage

Both were heated in a reactor with a helium carrier gas passing through the layers sequentially and connected to a mass spectrometer with a fast response inlet capillary system to detect NH3. The results showed that negligible ammonia was detected in the evolved gas mixture when the initial layer was LiNH2 regardless of the contact time (to a minimum of 2.4 ms) of the carrier gas with the two-layer samples. By contrast, the reverse layer arrangement led to significant amounts of NH3 detected in the evolved gas. Hence, it was proposed that the hydrogen desorption process is a two-step one (Eqs 16.7, 16.8): 2LiNH2 → Li2NH + NH3 NH3 + LiH → LiNH2 + H2

[16.7] [16.8]

in which the second is extremely rapid and hence the first is rate-determining. The premise of a two-step reaction in which the amide decomposition was essentially rate-determining was corroborated by Ichikawa et al. using a similar two-layer approach and thermal desorption mass spectrometry (TDMS) [52]. They propose a mechanism in which there is successive amide decomposition immediately followed by reaction of evolved ammonia with remaining LiH (summarised as in Eq. 16.9): LiH + LiNH2 → 1/2LiH + (1/2LiH + 1/2NH3) + 1/2Li2NH → 1/2LiH + 1/2LiNH2 + 1/2Li2NH + 1/2H2 → 1n LiH + 1n LiNH 2 + Σ nk =1 1k Li 2 NH 2 2 2 + Σ nk =1 1k H 2 2 → Li2NH + H2

[16.9]

Ichikawa et al. also follow some of their previous work in discussing the action of possible catalysts in this process, but this is considered in more detail under Section 16.2.4. Desorption in the Li–N–H system, decomposition kinetics of lithium amide and predictions of the lifetime of the system (104 min at operating temperatures of 200 °C) were assessed by combinations of in situ XRD and thermogravimetric measurements performed over relatively low temperatures and hydrogen pressures [53, 54]. Under these conditions uptake kinetics were slow and some slow evolution of ammonia was observed after desorption and rehydrogenation (amounting to approximately 2% of the sample) [53]. A revealing study of the origin and extent of ammonia evolution under equilibrium conditions was provided by Hino et al. who used in situ Raman spectroscopy to quantify the NH3 partial pressure on reacting LiH and LiNH2 in 1:1 and 2:1 ratios [55]. Under these conditions the partial pressure of emitted NH3 is c. 0.1% of that of H2 above 275 °C.

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

459

Nonetheless, this concentration, as might be expected in a closed system, would be sufficient to poison a polymer electrolyte fuel cell (PEFC) and hence evolved ammonia would likely have to be trapped upstream in a working system using these materials. Final likely confirmation of a twostep process for amide desorption was provided by Ichikawa et al. who used isotopic labelling (using LiD and ND3 in addition to the hydrogen-containing equivalents) in conjunction with TPDMS, thermogravimetric analysis (TGA) and Fourier transform IR (FTIR) spectroscopy. The study proposed four amide decomposition models and revealed that the isotopic distribution favoured a mechanism by which a ratio of 9:12:4 H2:D2:HD in the LiNH2 + LiD reaction was obtained. The proposed mechanism that would yield such a distribution could only be one involving mediating ammonia (and its rapid reaction with LiH(D)) and the authors propose a theoretical molecular intermediate ‘LiNH4’ that allows this reaction step to occur. The concept of a rapid reaction of a hydride with ammonia is one that has been exploited in the design of more complex metal–nitrogen–hydrogen storage systems and is described in more detail in Section 16.4. Building on the mechanistic concepts that Ichikawa et al. and Hu and Ruckenstein had established, David et al. proceeded to investigate the structural and compositional consequences of cycling in the Li–N–H system with the intention of revealing more regarding the way the reactions proceed in the solid state [56]. Using ex situ synchrotron XRD and probing the imide– amide step of the storage cycle, it was demonstrated that both hydrogenation and dehydrogenation occur through non-stoichiometric intermediates based on the (higher temperature, simple, disordered) cubic, anti-fluorite structure of Li2NH. The important implications were not only that the evidence further supported the ammonia-mediated model, but also that the creation of defects (cation vacancies), and hence the mobility of Li+ and H+, were absolutely central to the uptake and release process in the reversible section of the Li– N–H cycle. It is the creation and removal of Frenkel-defect based vacancies in the anti-fluorite lattice(s) (allowing exchange of Li+ and H+) which drives the reaction mechanistically in the solid state. There has been considerably less work reported concerning the first step of the Li–N–H sorption cycle whereby Li3N absorbs hydrogen (Eq. 16.10), probably because the challenge of making this process reversible under reasonable working conditions is seen as much greater (∆H = 116 kJ mol–1 for the forward reaction): Li3N + H2 → Li2NH + LiH

[16.10]

Initial work by Hu and Ruckenstein appeared to show that contrary to Chen et al.’s findings, H2 did not begin to be taken up by Li3N until 150 °C, but that this initial uptake temperature was not dependent on the partial pressure of hydrogen [50]. Inevitably, the uptake kinetics improved with increasing

© 2008, Woodhead Publishing Limited

460

Solid-state hydrogen storage

H2 partial pressure, as one might expect. Pre-treating the nitride under argon at temperatures of 400 and 500 °C only resulted in deactivation of the material to hydrogenation, probably, the authors suggest, via sintering. The same authors went on to consider the effects of particle size and additives in more detail subsequently and these and other studies are discussed in Section 16.2.4. Work by Meisner et al. largely corroborates that of Hu and Ruckenstein above [53]. Meisner et al.’s gravimetric measurements show that hydrogenation of Li3N begins at c. 175 °C (the discrepancies in temperature among Chen et al., Hu and Ruckenstein and Mesiner et al. reflect not only on the differing methods used to measure uptake but also probable difference in the nature of the starting materials) and that the kinetics of uptake in the first stage of the cycle are very slow (c. 4.5 wt% in ~1000 min). Given the dual challenges of prohibitively slow uptake kinetics for nitride to imide conversion and the high desorption temperatures to return the amide to the imide for hydrogen release, strategies for the Li–N–H system turned to the modification of particle size and the introduction of additives (e.g. potential catalysts). These aspects are covered in the next section.

16.2.4 The effect of additives and particle size on hydrogen sorption Considering first the uptake of hydrogen by Li3N (Eq. 16.10), Hu and Ruckenstein followed their initial observation of reduced uptake in pretreated lithium nitride [50] by assessing the effects of pre-heating in more detail using temperature programmed hydrogenation (TPH) and scanning electron microscopy (SEM) [57]. In apparent contradiction to many other studies they found that the initial uptake in untreated Li3N was fast (5.5 wt% of H2 in c. 3 min; 230 °C and 7 bar H2) but that 3200 min were needed to absorb a further 2.5 wt%. By contrast, Li3N preheated under vacuum at 400 °C (for 4.5 h) showed almost the opposite behaviour, requiring 3 min to absorb 3 wt% of H2 but absorbing a further 6.5 wt% in 300 min. Preheating to 500 °C deactivated the nitride to the extent that only 2 wt% of hydrogen was taken up over 2620 min. SEM images revealed that the 400 °C pretreated nitride had a larger particle size than the untreated samples as might be expected from sintering. Further, pre-treatment increased the initial uptake temperature by c. 100 °C. The authors suggest that 400 °C pre-treatment improves the cycleability of the nitride in subsequent uptake–release. Ultimately from fewer than six cycles this is difficult to assess, and there is no indication as to the identity of the starting material prior to or following heat treatment. Similar deactivation effects, in fact, were observed by the same authors when high temperatures (400 °C) are used to dehydrogenate Li2NH to return to the nitride [58]. Subsequent rehydrogenation results in uptake of only 0.4 wt% at 230 °C / 7 atm H2.

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

461

Interestingly, an alternative ‘pre-treatment’ of exposing Li3N to air prior to uptake has effects which might be regarded as counter-intuitive [59]. Following 30 min air exposure and vacuum heating at 230 °C, the treated Li3N material was hydrogenated under 7 atm H2 for ≥ 48 h at 230 °C and dehydrogenated under vacuum at 280 °C for 24 h. Although the identity of the mixed material prior to hydrogenation is not reported, following hydrogenation and dehydrogenation the samples contained mixtures of LiNH2 + Li2O + LiH and Li2NH + Li2O + LiH respectively as might be expected. The uptake kinetics of the Li–N–O–(H) mixture was rapid, taking up c. 5 wt% H2 in 3 min at 180–198 °C, and appeared superior to Li3N subjected to similar hydrogenation–dehydrogenation regimes without Li2O mixing, yet there appears to be some contradiction with the later results described above [57] where heat pre-treatment alone (without Lithia pre-mixing) gives rise to similar rapid uptake kinetics. Nevertheless, the authors suggest that Li2O may act as both a ‘stabiliser’ (reducing exothermic sintering effects and increasing cycleability) and catalyst, although apparently as they state: ‘It is worth noting that the exposure time to wet air is critical. Both too long and too short exposure times are not satisfactory’. Given that Li3N reacts spontaneously with moist air under ambient conditions to produce the hydroxide, LiOH, and ammonia (Section 16.2.1) this is undoubtedly true for prolonged exposure. Approaches to modifying materials for the second (imide–amide) stage of the Li–N–H cycle have been varied. Hu and Ruckenstein considered the effect of preparative route on the storage performance of Li2NH by contrasting the usual method of amide decomposition (with or without hydride addition, equations 16.7 and 16.8) with both the action of ammonia on lithium (Eq. 16.4) followed by dehydrogenation [60] and the reaction of Li3N with LiNH2 (Eq. 16.11) [61, 62]: Li3N + LiNH2 → 2Li2NH

[16.11] –1

The above reaction is exothermic (∆H = –77 kJ mol ) and occurs within 10 min at 230 °C. The authors propose that by analogy to the amide decomposition in the presence of lithium hydride, the amide–nitride reaction is a two step process which is ammonia mediated. In this case, Li3N (as opposed to LiH) acts as the trap for gaseous ammonia to form the imide in a fast reaction (Eq. 16.12): 2Li3N + NH3 → 3Li2NH

[16.12]

The removal of NH3 by Li3N shifts the amide decomposition reaction (Eq. 16.2) to the right overcoming the problems of low equilibrium pressure of NH3 in the direct amide decomposition (i.e. when Li3N or LiH are not present) [54]. Whereas imide prepared via the lithium ammoniation/ dehydrogenation route yielded initially slow uptake kinetics (attributed to

© 2008, Woodhead Publishing Limited

462

Solid-state hydrogen storage

amide sintering), that prepared via the Li3N/LiNH2 route not only displayed fast uptake kinetics (5.4 wt% H in 10 min), but also consistently yielded an exceptional gravimetric capacity of 6.8 wt% H over several cycles (only marginally below the theoretical maximum of 6.85 wt%). The material thus appears to outperform imide produced by any other route; a significant discovery and one which now merits further investigation. Drawing on methods and observations from earlier hydride storage systems, many of the initial approaches to improving the desorption of hydrogen from the amide in the second stage of the Li–N–H cycle focused on two complementary approaches: ball milling to reduce particle size and addition of potential catalysts to reduce desorption temperature. These approaches were first adopted in the Li–N–H system by Ichikawa et al. who used each to try to optimise desorption in the amide–imide reaction (Eqs 16.7 and 16.8) [49]. In these initial investigations 1 mol% each of Ni, Fe, Co and TiCl3 were added to LiNH2/LiH mixtures and TDMS was used to follow the desorption with temperature. In ball milled mixtures with no added catalysts, Ichikawa et al. observed that first ammonia evolution was reduced significantly compared to samples mixed manually and second that H2 was released over a temperature range of 180–400 °C. However, when catalysts were added to the ball milled mixture, not only was NH3 evolution significantly diminished, but also the desorption temperatures (Tdes) were drastically reduced. The effect was most prominent for TiCl3 where 5.5 wt% H was released as H2 between 150 and 250 °C. Later work by many of the same authors focused on the TiCl3catalysed desorption process and how the form of the perceived active part of the catalyst might contribute to the H2 release [63, 64]. The studies revealed that in addition to TiCl3, nanoparticles of Ti metal or TiO2 were also effective in reducing Tdes, although neither of these additives outperformed TiCl3. It was suggested that the trichloride and dioxide might react with LiH in the ball milled mixture to produce the corresponding lithium salt + Ti metal and hydrogen. However, although these reactions undeniably release hydrogen (at low temperature and hence might reduce Tdes) it is less clear how this affects the bulk of the amide sample catalytically and how the catalytic action might be sustainable over multiple cycles given the reactions to LiCl or Li2O respectively are likely to be irreversible under sorption/desorption conditions. By contrast to the addition of Ti species, although Mn metal and MnO2 and V metal and V2O5 were observed to reduce the decomposition of LiNH2 alone, when mixed with the amide and LiH, these additives had no discernible effect on the H2 desorption kinetics or Tdes [65]. A number of studies have considered the various effects of ball milling without additives. In fact, Kojima et al. have demonstrated that it is possible to store hydrogen in lithium nitride (and also magnesium nitride, Mg3N2, calcium nitride, Ca3N2, and others) without heating simply by ball milling the nitride under 1 MPa H2 at room temperature [66, 67]. Following 20 h of

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

463

milling, lithium nitride was partially converted into a mixture of imide, amide and hydride. LiNH2 could be similarly synthesised by ball milling LiH under NH3 for 2 h [68]. The amide so produced decomposed above 230 °C. An independent study of the effects of milling on lithium amide decomposition (again without the presence of LiH), reported that the onset temperature of amide decomposition was reduced (from 120 °C to room temperature) by milling [69]. The authors attribute this to mechanical activation via particle size reduction, although decomposition in situ within the mill was not investigated. Increased milling times lead to progressive reductions in the activation energy for decomposition (from 244 kJ mol–1 in untreated amide to 138 kJ mol–1 in amide milled for 180 min). At about the same time as the work by Ichikawa et al. [49] had revealed that adding TiCl3 to LiNH2/LiH reduced Tdes and improved desorption kinetics, Nakamori and Orimo were pioneering an alternative ‘additive’ strategy by which the additives were not so much acting as catalysts but as dopants [70, 71]. In fact, these studies were to pave the way for a whole new branch of research and, with it, new families of complex imides/amides and other composite systems. The details of these materials will be discussed in Section 16.4 and elsewhere within this volume, but the concept employed was that by introducing certain dopants of various electronegativity it might be possible to destabilise the N–H bond by effectively exploiting the so-called inductive effect [72]. A secondary effect on the doped materials might also then be the introduction of defects (vacancies) in order to maintain charge balance. The initial results were extremely promising in that by doping Li3N with 10 at.% Mg, Tdes could be reduced by 50 K. The reasons for this would be revealed as the mixtures and those containing higher levels of Mg dopants were investigated further (Section 16.4.1). Following the practical demonstration of the possibilities regarding N–H destabilisation, a number of theoretical studies were conducted to evaluate the potential for other additives and dopants to act in a similar way in the Li– N–H (and other) system(s). A first theoretical treatment by DFT methods concurred with experiment and supported the premise that partially replacing Li by more electronegative Mg in LiNH2 destabilised the (NH)2– anion [73]. Another study of Mg doping in LiNH2 (where one Li atom was replaced in a 2 × 1 × 1 amide supercell by Mg) revealed that both the Li–N and N–H bonding would be weakened as a result (and also that doping switches the amide from an insulator to a metal) [74]. Following this initial finding, replacement of dopant Mg by Na, K or Al suggested that only doping by Al is likely to destabilise the N–H bonds significantly to reduce Tdes. A separate DFT study (utilising optimally localised Wannier functions as a measure of the electron density distribution around bonding atoms) also found that Mg doping destabilised the N–H bonds in doped lithium amide, but, by contrast, found that K substitution had a similar (albeit much weaker) destabilising

© 2008, Woodhead Publishing Limited

464

Solid-state hydrogen storage

effect [75]. However, the authors also note that the energy cost in replacing Li by Mg is high and a more energetically favourable approach to reduce the non-metal–H bond strength would be to (partially) replace N by an element with a weaker covalent bond to H. A similar conclusion might be drawn from pseudopotential calculations for a range of complex hydrides from alanates to amides [76].

16.3

The imides and amides of the group 2 elements

Perhaps without the same degree of impact, simultaneous to the report of reversible hydrogen storage in the Li–N–H system, Chen et al. reported in the same paper that it was also possible to store hydrogen in the Ca–N–H system by converting calcium nitride hydride, Ca2NH, to calcium imide, CaNH [1] (Eq. 16.13). A short time later they extended their investigations of the system by reporting the reaction of Ca3N2 with hydrogen to yield ideally the imide and calcium hydride (Eq. 16.14) [77]: Ca2NH + H2 ↔ CaNH + CaH2

[16.13]

Ca3N2 + 2H2 → 2CaNH + CaH2

[16.14]

In fact the hydrogenation of the nitride was not observed to go to completion and Chen et al. tentatively identified the reaction process as that in Eq. (16.15), with the major product being an ‘imide-like’ phase: Ca3N2 + X/2H2 → CamN2Hn + (3–m) CaH2

[16.15]

where X = 6 + n – 2m and m and n are dependent on the degree of hydrogenation. The initial storage material, Ca2NH, was synthesised by dehydrogenation of a 1:1 mixture of Ca3N2 + CaH2 at 550 °C. Theoretically, 2.1 wt% of H can be stored in the nitride hydride and experimentally the maximum reversible capacity at 550 °C was close to this (c. 1.9 wt%). The calculated standard enthalpy of absorption for Ca2NH (∆Habs = –88.7 kJ mol–1) is somewhat more exothermic for the nitride hydride than for lithium imide (∆Habs = –66.1 kJ mol–1). The reaction between calcium nitride and hydrogen was, in fact, first reported by Dafert and Miklauz in 1909 who identified the product as ‘Ca3(NH2)2’ which in turn decomposed to CaNH and CaH2 by photolysis under sunlight [78]. Chen et al. observed that the nitride began to uptake H2 at 300 °C and although c. 2.36 wt% of hydrogen could be absorbed, the reaction was not completely reversible even at 600 °C (approximately half of the absorbed hydrogen was released; X = 1.8 in Equation 16.15) and Ca3N2 was not observed in the dehydrogenation products by XRD. Moreover, the dehydrogenation products under varying reaction conditions were apparently imide phases of different hydrogen stoichiometry as suggested by Eq. (16.15).

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

465

More recently, Hino et al. explored the Ca–N–H system starting from the amide, Ca(NH2)2 and focusing on the desorption process [79]. First synthesising calcium amide by ball milling CaH2 under ammonia gas, the decomposition product upon heating the amide at 350 °C was found to be calcium imide according to Eq. (16.16): Ca(NH2)2 → CaNH + NH3

[16.16]

However, as the temperature was increased above 500 °C, additional reflections were observed in the XRD profiles and the product(s) could not be identified. On following the decomposition by TDMS it was noted that not only was NH3 evolved in the decomposition (as expected from Eq. 16.16) but also significant amounts of H2 were generated at temperatures too low to be expected from NH3 decomposition. By analogy to the lithium amide–imide system, it was assumed that either unreacted calcium hydride (from the initial reaction to produce amide) was reacting in situ with NH3 (which could be represented by Eq. 16.17 and amounts to a weight loss of 3.5 wt%) or that unreacted CaH2 was reacting in the solid state with generated CaNH to produce the nitride hydride Ca2NH + hydrogen; the reverse reaction (Eq. 16.13; weight loss of 2.1 wt%) to that originally investigated by Chen et al: Ca(NH2)2 + CaH2 → 2CaNH + 2H2

[16.17]

In fact, TDMS of 1:1 mixtures of Ca(NH2)2:CaH2 and CaNH:CaH2 and a 1:3 mixture of Ca(NH 2 ) 2 :CaH 2 compared with the data from amide decomposition suggest both these processes occur in the latter reaction. However, as Chen et al. had suggested previously, only the imide + hydride reaction is reversible [1, 77] and hence the overall reaction is as shown in Eq. (16.18), giving a total release of 4.0 wt%, but only 2.1 wt% of which is reversible on uptake (with the decomposition of the amide being exothermic): Ca(NH2)2 + 3CaH2 → 2CaNH + 2CaH2 + 2H2 ↔ 2Ca2NH + 4H2

[16.18]

Although of academic interest, the Ca–N–H system was too limited in both its gravimetric capacity and reversibility to be a viable one. Replacing Ca by Mg, however, presents a potentially attractive alternative provided the thermodynamics of the desorption reaction are such that, unlike for Ca–N– H, the reverse absorption process to return to the amide is possible. The Mg– N–H system was first investigated by Nakamori et al. in 2004 [80] although the original decision to explore the decomposition of magnesium amide originated more from the materials design concept associated with Mg doping in the Li–N–H system (see Sections 16.2.4 and 16.4.1) than from the replacement of Ca by Mg in group 2 metal systems. The amide, synthesised from MgH2 and ammonia gas, decomposes in two stages to magnesium imide and then magnesium nitride at 630 and 720 K respectively (Eq. 16.19):

© 2008, Woodhead Publishing Limited

466

Solid-state hydrogen storage

3Mg(NH2)2 → 3MgNH + 3NH3 → Mg3N2 + 4NH3

[16.19]

In the presence of MgH2 however, and by analogy with the Li–N–H and Ca– N–H systems, the decomposition should evolve H2 in place of NH3 presumably, also by analogy, via the rapid reaction of the hydride with evolved NH3. Hence by changing the ratio of amide : hydride, two outcomes are potentially possible (Eqs 16.20 and 16.21): Mg(NH2)2 + MgH2 ↔ 2MgNH + 2H2

[16.20]

Mg(NH2)2 + 2MgH2 ↔ Mg3N2 + 4H2

[16.21]

However, the fact that experimentally NH3 evolution continues to be observed on decomposition of the amide with hydride present suggests that unlike the LiNH2 + LiH desorption step, the kinetics of the reaction between MgH2 and NH3 are slow. This was one further motivation for considering the Li–Mg– N–H system (and the reaction between LiH and Mg(NH2)2 as described in Section 16.4.1). Chen et al. subsequently demonstrated that the kinetic deficiencies of the magnesium hydride + ammonia reaction could be at least partially overcome by ball milling the mixtures of amide and hydride [81, 82]. Ball milling had the dual net effect of decreasing the desorption temperature at which hydrogen was first evolved and suppressing NH3 evolution until 200 °C and above. The amount of evolved ammonia was dependent on the milling time but NH3 emission was significant even after 72 h of milling and could never be prevented above 200 °C. Nevertheless, an appreciable amount of H2 could be desorbed from the amide and, from equations (16.20) and (16.21), in principle it should be possible to release 5.1 and 7.9 wt% respectively. In practice, 4.8 wt% and 7.1 wt% could be desorbed, but ultimately the success of the Mg–N–H system is likely to lie with its reversibility. From the calculated ∆Hf for Mg(NH2)2 (–325 kJ mol–1) the reaction enthalpy for the amide to nitride process (Equation 16.21) is only +14 kJ mol–1 and Chen and coworkers postulate that although in principle the reaction is thus reversible, the equilibrium hydrogen pressure at ambient temperature would be too high [82].

16.4

Mixed metal imides and amides

16.4.1 The Li–Mg–N–H system The pseudo-quaternary nitride–imide–amide system involving lithium and magnesium has been the focus of the largest part of hydrogen storage research in nitrogenous materials since the initial report of high capacity sorption in the Li–N–H system [1]. As alluded to in Section 16.2.4, the route into this area was one concentrating initially on low-level doping of Li3N with Mg

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

467

(c. 10 at.%) [70, 71]. The initial strategy of destabilising N–H bonds by utilising small amounts of Mg substituent gave way to ideas of combining LiNH2 with MgH2 in an analogous way to the combination of lithium amide with lithium hydride in the first desorption (reverse) step of the Li–N–H cycle. In fact, the first report of this approach by Luo [83] was almost simultaneous to the experiments by Nakamori and Orimo [70, 71] and in the former study it was proposed that by replacing LiH by MgH2 the desorption temperature of the amide–imide process might be decreased since MgH2 is inherently less stable than its lithium counterpart. This proved to be the case and the system was found to be reversible over nine cycles at 200 °C and 32 bar H2 with a gravimetric capacity of 4.5 wt%. Moreover, the proposed amide/hydride reaction (Equation 16.22) has a reduced reaction enthalpy compared with the equivalent LiNH 2 /LiH reaction (34 kJ mol –1 vs. 51 kJ mol–1 calculated from the van’t Hoff plots), reinforcing the reduced stability strategy of replacing LiH with MgH2: 2LiNH2 + MgH2 ↔ Li2Mg(NH)2 + 2H2

[16.22]

Two independent studies rapidly followed considering the alternative pathway to mixed imides in the Li–Mg–N–H system, namely combining magnesium amide with lithium hydride [84, 85]. The two studies differed in the ratios of starting materials considered. The former took the 1:2 ratio of amide : hydride and by analogy to Eq. (16.22) sought to liberate hydrogen as a by-product of mixed imide formation, viz. Eq. (16.23) (hereafter referred to as the 1:2 reaction based on the Mg(NH2)2 : LiH ratio): Mg(NH2)2 + 2LiH ↔ Li2Mg(NH)2 + 2H2

[16.23]

Experimentally 3.8 of the 4 H atoms could be stored reversibly at 180 °C, 65 bar H2, equating to gravimetric capacity of 5.5 wt% as compared with the theoretical maximum of 5.87 wt%. Two other important outcomes emerged from this investigation, the first being that, following cycling, the hydrogenated phase mixture was consistently magnesium amide and lithium hydride, suggesting these phases are thermodynamically favoured (over MgH2 + LiNH2) and are also the likely phases present after the first cycle in the 2:1 LiNH2 : MgH2 experiments performed by Luo [83]. (This was subsequently confirmed as likely to be correct by Luo and Rönnebro [86]). The second outcome is that this study marks the emergence of the formal theory that the driving force for amide–hydride reactions is the presence of ‘protonic’ and ‘hydridic’ species, Hδ+ and Hδ–. The presence of these species in amide (or imide) and hydride respectively allows an exothermic combination such as that described in Eq. (16.24). In principle therefore, Chen et al. postulated that a vast range of possible Hδ+–Hδ– reactions might exist from which one could design and tune chemical hydride storage systems. This was to prove a major driver in subsequent materials design, notably in multicomponent systems and has

© 2008, Woodhead Publishing Limited

468

Solid-state hydrogen storage

recently been reviewed as a strategy in devising and explaining complex hydride systems more widely [87]. H+ + H– → H2 (∆H = –17.37 eV)

[16.24]

Meanwhile the second of the two major studies of the period focused on an entirely different starting ratio of Mg(NH2)2 to LiH (3:8) and in so doing drove forward a completely new reaction pathway that significantly increased the attainable limit of stored hydrogen in the Li–Mg–N–H system [85]. The reaction (Eq. 16.25; hereafter referred to as the 3:8 reaction based on the Mg(NH2)2 : LiH ratio) yields a reversible capacity of 7 wt% hydrogen and importantly H2 begins to desorb at temperatures as low as 140 °C, and peaks at 190 °C, in TDMS conditions under argon without the use of additives/ catalysts. Under vacuum, the mixture dehydrogenates at 170 °C. The reverse hydrogenation process can be achieved at 200 °C under 3 MPa H2: 3Mg(NH2)2 + 8LiH ↔ Mg3N2 + 4Li2NH + 8H2

[16.25]

Ensuing studies in the Li–Mg–N–H system followed two main pathways which would eventually converge; one investigating the effects of changing the ratio of amide and hydride reactants, the other focusing on developing an understanding of the kinetics and mechanisms of these reactions. Following from their earlier Mg doping studies [70, 71, 73], Orimo and co-workers considered higher ratios of Li3N : Mg3N2 as a two-component system for hydrogenation [88, 89]. Following extensive XRD investigations of different nitride mixture ratios prepared under different temperature regimes, the reversible reaction shown in Eq. (16.26) (representing Li3N–20% Mg3N2; hereafter referred to as the 3:12 reaction based on the Mg(NH2)2 : LiH ratio) was found to yield both a maximum gravimetric capacity (9.1 wt%) and a minimum dehydrogenation temperature. This dehydrogenation temperature onset could be reduced by a further 140 K by ball milling with 1 mol% Ti (by analogy to LiNH2–LiH), although the maximum in the evolved hydrogen occurs at 540 K and gas continues to be desorbed at > 700 K. 4Li3N + Mg3N2 + 12H2 ↔ 3Mg(NH2)2 + 12LiH

[16.26]

Considering each of the Li–Mg–N–H reaction cycles in more detail, as highlighted above, the 1:2 reaction (Eq. 16.23) is unique in yielding a single complex (imide) phase (as opposed to a mixture of Mg and Li phases) on dehydrogenation. The 1:2 reaction, in fact, desorbs in two stages – twothirds of the H2 is evolved at higher pressures (46 bar), and one-third at lower pressures from PCT studies at 493 K – and from DSC begins at 373 K, reaching a maximum at c. 473 K (an overall heat of desorption was measured as 44.1 kJ mol–1 H2) [90]. The activation energy for the 1:2 reaction was calculated to be 102 kJ mol–1 and higher than many hydrides [91]. However, on the basis of equivalent calculations of the Ea for the decomposition of

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

469

magnesium amide to imide and subsequently nitride (134 kJ mol–1), it was believed that the 1:2 reaction was unlikely to proceed via an initial amide decomposition/ammonia mediation step. This proposal was tested further by many of the same authors by measuring the kinetics of the hydrogenation reaction gravimetrically and by conducting a series of isotopic experiments studying the reaction at the (magnesium) amide–imide–(lithium) hydride interface [92]. Revised Ea values of 88.1 and 130 kJ mol–1 for the amide– hydride reaction and magnesium imide decomposition respectively coupled with the results of the H2/D2 studies, reinforced the authors’ opinions that the Mg(NH2)2–LiH reaction proceeds through a coordinated interaction in the solid state and not via an ammonia generation step. The existence of the new ternary imide, Li2Mg(NH)2, as the dehydrogenation product was corroborated (without further characterisation) in two other studies [93, 94], the first of which suggests that hydrogenation of the ternary phase proceeds via an intermediate phase, ‘Li2MgN2H3.2’ shown to contain both imide and amide species by FTIR. This intermediate phase is then proposed to hydrogenate in one further step to magnesium amide and lithium hydride. Varying the Mg(NH2)2 : LiH ratio from the ideal 1:2 stoichiometry either increases the hydrogen desorption temperature (1:3) or leads to ammonia evolution (1:1) and the appearance of the 1:1 mixed imide phase, Li2Mg2(NH)3 [95] as a product (Eq. 16.27) which cannot be reversibly hydrogenated beyond an uptake of 0.5 wt% at or below 210 °C [96]. 2Mg(NH2)2 + 2LiH → Li2Mg2(NH)3 + NH3 + 2H2

[16.27]

By contrast to the 1:2 reaction, however, it was proposed that the equivalent 3:8 reaction was an ammonia-mediated one by analogy to the findings that had been made in the Li–N–H system [85, 97]. Hence, Mg(NH2)2 first decomposes to Mg3N2 + NH3 followed by the reaction of ammonia with LiH to produce Li(NH2) that reacts further with LiH to yield the imide (Eq. 16.28): 3Mg(NH2)2 → Mg3N2 + 4NH3 4NH3 + 4LiH → 4LiNH2 + 4H2 4LiNH2 + 4LiH → 4Li2NH + 4H2

[16.28]

Perhaps more unexpected was the premise that the reverse hydrogenation reaction (of magnesium nitride + lithium imide) is also ammonia-mediated [98, 99]. The first step of this hydrogenation involves the hydrogenation of the imide to lithium hydride and lithium amide and in the second step the amide then decomposes to imide and ammonia. Subsequently in step 3, the evolved ammonia reacts with reactant Mg3N2 to yield the product Mg(NH2)2 while co-produced Li2NH is hydrogenated further to lithium amide + lithium hydride. As part of this study, experiments considering the hydrogenation of

© 2008, Woodhead Publishing Limited

470

Solid-state hydrogen storage

Mg3N2 alone certainly confirm that without Li2NH the uptake reaction will not proceed under 10 MPa H2 at 200 °C and the authors maintain that nanosized (ball-milled) Mg3N2 mixed intimately with the produced amide is essential if the ammonia-mediated hydrogenation is to occur. Many of the same authors of the above studies then went so far as to suggest that all the Li–Mg–N dehydrogenation reactions (1:2, 3:8 and 3:12) could be rationalised in terms of the ammonia-mediated mechanism [100]. Using TDMS, TGA, XRD and SEM, Leng et al. propose that the general ammonia-mediated dehydrogenation reaction scheme (Eq. 16.29) holds irrespective of the initial ratio of Mg(NH2)2 to LiH. Importantly they both confirm that the 3:8 reaction releases H2 below 400 °C while the 3:12 reaction only releases appreciable H2 above 400 °C and also observe that the 1:2 reaction proceeds first through the binary imides Li2NH and MgNH before forming the ternary imide, Li2Mg(NH)2. xLiH + 3Mg(NH2)2 → xLiH + 3MgNH + 3NH3 → (x–3)LiH + 3LiNH2 + 3MgNH + 3H2 (if x ≥ 3) → (x–6)LiH + 3Li2NH + 3MgNH + 6H2 (if x ≥ 6) → (x–8)LiH + 4Li2NH + Mg3N2 + 8H2 (if x ≥ 8) → (x–12)LiH + 4Li3N + Mg3N2 + 12H2 (if x ≥ 12)

[16.29]

The 3:8 reaction was subsequently studied further by in situ XRD and while diffraction does not elucidate the evolution of the Mg-containing phases over the course of the dehydrogenation reaction, there is some (although not unequivocal) evidence that LiNH2 forms as an intermediate phase, which might lend weight to the ammonia-mediated reaction pathway, at least for the 8:3 case [101]. Further investigations of the Li–Mg–N–H system both on H uptake and release are clearly required to resolve the ambiguities of the reaction mechanisms. Arguably such studies may be made easier by an alternative route to the LiH/Mg(NH2)2 hydrogenated components via a synthesis route combining Mg metal and LiNH 2 under vacuum followed by hydrogenation [102]. Finally and interestingly a further study of the 3:12 dehydrogenation reaction considering P–C isotherms under varying (decreasing) H2 pressure coupled with ex situ XRD, shows that at a plateau and subsequent sloping regions of the isotherms, lithium hydride co-exists first with a proposed stoichiometric mixed imide amide, Li4Mg3(NH2)2(NH)4 and then nonstoichiometric Li4+xMg3(NH2)2–x(NH)4+x [103]. Li2Mg(NH)2 and LiH are the final products at the investigated temperatures (≤523 K). The structures of the stoichiometric and non-stoichiometric imide amides are as yet unknown but have been indexed to tetragonal and orthorhombic cells respectively.

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

471

16.4.2 The Li–Ca–N–H, Mg–Ca–N–H and Na–Mg–N–H systems Initial investigation of the Li–Ca–N–H system revealed that despite a lower theoretical and practically observed gravimetric capacity (4.3 wt%) than either Li–Mg–N–H or Li–N–H, the onset of hydrogen desorption occurs at a lower temperature (and reaches a maximum at a temperature below that of the equivalent in the Li–N–H system) as does the onset of absorption [84]. Unfortunately, however, much lower pressures and higher temperatures than those for Li–Mg–N–H are required to desorb an amount of hydrogen approaching the theoretical maximum in the system. By analogy with Eq. 16.22, Li2Ca(NH)2 is the product of the dehydrogenation reaction of LiNH2 with calcium hydride, CaH2 [104]. The mixed metal imide is formed at 300 °C and crystallises with the hexagonal anti-La2O3 structure with calcium and lithium in octahedral and tetrahedral coordination to (NH)2– respectively (Fig. 16.5). Presumably the reverse hydrogenation reaction would yield LiH and Ca(NH2)2, again by analogy with Li–Mg–N–H. Hydrogen is also readily desorbed from 1:2 and 1:1 mixtures of Mg(NH2)2 and CaH2, although for the former ratio, the desorption profile with temperature is a complex one with H2 effectively evolved in three stages, with the lowest temperature (260 °C) stage potentially reversible [105]. The total hydrogen release experimentally amounts to 4.9 wt% and the final products at the end of desorption (at higher temperature) are observed to be the ternary nitride, Mg2CaN2 [106], CaNH and calcium nitride hydride, Ca2NH (Eq. 16.30):

Ca

Li N

c

a

b

16.5 Anti-La2O3 structure of Li2Ca(NH)2. Li+ (smallest spheres) and Ca2+ (light spheres) are coordinated to (NH)2– in a tetrahedral and octahedral geometry respectively. The disordered H positions are not shown.

© 2008, Woodhead Publishing Limited

472

Solid-state hydrogen storage

2Mg(NH2)2 + 4CaH2 → Mg2CaN2 + Ca2NH + CaNH + 7H2 [16.30] Subsequent temperature programmed desorption (TPD) and DSC measurements of the 1:1 reaction revealed that only above 220 °C was any evolved NH3 observed and that hydrogen desorption from samples milled at 12 h generates an enthalpy of 28.2 kJ mol–1. However, despite indications that thermodynamically the system should be reversible at c. 210 °C, the kinetic barrier to rehydrogenation appears to be substantial [107]. Characterisation of the 1:1 phase mixture after prolonged (72 h) ball milling of the starting materials, Mg(NH2)2 and CaH2 proved to be difficult by diffraction methods although it was apparent that a new phase appeared to evolve, following consumption of the starting materials, that did not match to any of the product phases from the 1:2 reaction (Equation 16.30). This phase was tentatively identified as MgCa(NH)2 on the basis of the starting ratio and the similarity of the XRD patterns to those of the binary imides. A later extended X-ray absorption fine structure/X-ray absorption near edge structure (EXAFS/XANES) study combined with TPD evidence (which indicated a different desorption profile for the proposed ternary phase compared with a mixture of CaNH and MgNH) supported the existence of the mixed metal imide with an assumed cubic CaNH-like unit cell, presumably with Ca and Mg disordered across the metal positions of the rock salt structure [108]. If calcium in the Mg–Ca–N–H system is replaced by sodium then the sorption behaviour changes quite markedly [109]. A 1:1 mixture of Mg(NH2)2:NaH begins to emit H2 at c. 120 °C and reaches a maximum at 160 °C. NH3 and N2 begin to evolve at temperatures above 170 °C , but by decreasing the ratio of Mg(NH2)2 : NaH from 1:1 through to 2:3 to 1:2 (i.e. increasing the amount of NaH), the release of these gasses can be significantly suppressed. This, thus, follows an analogous relationship to the ratios of Mg(NH2)2 : LiH in the Li–Mg–N–H system where it is proposed that LiH acts to remove NH3 in an ammonia-mediated mechanism. Although the gravimetric capacities are low (1.75 wt% for 1:1, 2.17 wt% for 2:3, 1.83 wt% for 1:2) the system is remarkably reversible, and the starting materials can be regenerated (under uptake conditions of 140 °C/160 psi H2) after 10 release–uptake cycles. The product of the dehydrogenation appears to be a complex imide phase as is seen in other quaternary systems. The Na–Mg– (NH) phase remains uncharacterised, as yet.

16.5

Future trends and conclusions

The imides and amides of the group 1 and 2 elements clearly offer some huge advantages among ‘chemical hydride’ systems most significantly the possibilities of reversible, high gravimetric and volumetric capacity solid-

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

473

state storage. The materials also offer the benefits of being cheap and relatively easy to produce. In terms of viable applications there are nevertheless a number of practical and fundamental challenges that need to be addressed and overcome. Like many hydrides (and indeed solid-state storage materials) the imides and amides tend to be hygroscopic and, in extreme cases, pyrophoric. This presents engineering as well as chemical challenges, although limited investigations to date suggest that some of these materials can be passivated on air/water exposure without loss of performance (or indeed react synergistically with oxides). A greater challenge from a materials chemistry perspective is to depress the hydrogen desorption temperatures of the amides sufficiently such that they lie within the range of automotive on-board systems. One of the most dramatic effects in the (multicomponent) amide–hydride systems, however, is the cooperative (Hδ+–Hδ–) effect that brings about massive reductions in the desorption temperatures compared with the amide or hydride alone. This facet (unexpected and not yet well understood) is cause for great optimism in terms of materials and reaction design. The discovery of this effect, in fact, has already led to strikingly rapid progress in imide/amide storage materials over the past 5 years and bodes well for further major advances. (In fact the extrapolation of this symbiotic effect beyond amide– hydride systems is a topic in its own right as discussed in Chapter 17.) Likewise as demonstrated by addition of various additives, but as yet relatively unexplored, there should be a major role for catalysis in optimising desorption and sorption kinetics more generally. This latter challenge is one that has been only partially addressed by physical processing (such as ball milling) and could be a key area for advanced synthetic chemistry principles to be applied. What is also extremely encouraging is that the immense drive to produce improved storage materials among imides and amides, and the momentum that this has induced, has led to significant advances of a more fundamental nature. Among many, these include the discovery of new phases (complex imides and amides), the elucidation of structure and phase relations (with temperature, pressure and composition) and real insight into solid-state mechanism (and also particularly at the physical interfaces between solid and gas). In terms of fundamental research, however, there is a vast amount still to be uncovered and, as is almost always the case, this is likely to have considerable impact on future materials and system design.

16.6

Acknowledgements

The author would like to acknowledge gratefully co-workers and colleagues from within and beyond his research group that have been involved with the group’s hydrogen storage related research over the last several years, but

© 2008, Woodhead Publishing Limited

474

Solid-state hydrogen storage

most notably Dr Catherine Jewell and Dr Jinhan Yao. The author would also like to thank the EPSRC for funding via the SUPERGEN initiative and the University of Glasgow and WestCHEM.

16.7 1. 2. 3. 4. 5. 6. 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.

References P. Chen, Z. Xiong, J. Luo, J. Lin, K. L. Tan, Nature 420 (2002), 302. D. H. Gregory, Coord. Chem. Rev. 215/1 (2001), 301. L. Ouvrard, C. R. Acad. Sci. Paris 114 (1892), 120. R. Brill, Z. Kristallogr. 65 (1927), 94. E. Zintl, G. Brauer, Z. Elektrochem. 41 (1935), 102. R. Marx, Z. Anorg. Allg. Chem. 623 (1997), 1912. M. Mali, J. Roos, D. Binkmann, Phys. Rev. B 36 (1987), 3888. H. J. Beister, S. Haag, R. Kniep, K. Strössner, K. Syassen, Angew. Chem. Int. Ed. Engl. 27 (1988), 1101. S. V. Mitrokhina, K. P. Burdina, K. N. Semenenko, Vestn. Mosk. Univ. Seriya 2 Khimiya 31 (1990), 612. H. J. Beister, S. Haag, R. Kniep, K. Strössner, K. Syassen, Angew. Chem. Int. Ed. Engl. 27 (1988), 1101. A. Lazicki, B. Maddox, W. J. Evans, C. S. Yoo, A. K. McMahan, W. E. Pickett, R. T. Scalettar, M. Y. Hu, P. Chow, Phys. Rev. Lett. 95 (2005), 165503. A. Rabenau, H. Schulz, J. Less-Common Metals 50 (1976), 155. U. von Alpen, J. Solid State Chem. 29 (1979), 379. A. Rabenau, Solid State Ionics 6 (1982), 277. F. W. Dafert, R. Miklauz, Monatsch. Chem. 30 (1909), 649. F. W. Dafert, R. Miklauz, Monatsch. Chem. 31 (1910), 981. F. W. Dafert, R. Miklauz, Monatsch. Chem. 33 (1912), 63. O. Ruff, H. Georges, Ber. Deutsch. Chem. Gesell. 44 (1911), 502. H. Schulz, K. H. Thiemann, Acta Crystallogr. Sect. A 35 (1979), 309. K. Differt, R. Messer, J. Phys. C 13 (1980), 717. D. H. Gregory, P. M. O’Meara, A. G. Gordon, J. P. Hodges, S. Short, J. D. Jorgensen, Chem. Mater. 14 (2002), 2063. J. Wahl, Solid State Commun. 29 (1979) 485. A. Hooper, T. Laap, S. Skaarup, Mater. Res. Bull. 14 (1979), 1617. M. F. Bell, A. Breitschwerdt, U. von Alpen, Mater. Res. Bull. 16 (1981), 267. D. Brinkmann, W. Freudenreich, J. Roos, Solid State Commun. 28 (1978), 233. R. Messer, H. Birli, K. Differt, J. Phys. C 14 (1981), 2731. E. Bertold-Schweickert, M. Mali, J. Roos, D. Brinkmann, Phys. Rev. B 30 (1984), 2891. J. O. Thomas, R. Tellgren, Solid State Ionics 5 (1981), 407. W. J. Jolly, Inorganic Chemistry of Nitrogen, 19–46, W. A. Benjamin, Inc. (New York), 1964. E. C. Franklin, The Nitrogen System of Compounds, 53–63, Reinhold (New York), 1935. R. Juza, K. Opp, Z. Anorg. Allg. Chem. 266 (1951), 325. K. Ohoyama, Y. Nakamori, S. Orimo, K. Yamada, J. Phys. Soc. Jap. 74 (2005), 483. T. Noritake, H. Nozaki, M. Aoki, S. Towata, G. Kitahara, Y. Nakamori, S. Orimo, J. Alloys Comp. 393 (2005), 264.

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

475

34. M. P. Ballogh, C. Y. Jones, J. F. Herbst, L. G. Hector Jr, M. Kundrat, J. Alloys Comp. 420 (2006) 326. 35. R. A. Forman, J. Chem. Phys. 55 (1971), 1987. 36. P. J. Haigh, R. A. Forman, R. C. Frisch, J. Chem. Phys. 45 (1966), 812. 37. Y. Kojima, Y. Kawai, J. Alloys Comp., 395 (2005), 236. 38. B. Magyari-Köpe, V. Ozoliņš, C. Wolverton, Phys. Rev. B 73 (2006), 220101. 39. J. F. Herbst, L. G. Hector Jr, Phys. Rev. B 72 (2005), 125120. 40. T. Mueller, G. Ceder, Phys. Rev. B 74 (2006), 134104. 41. R. J. Mayer (ed.), ‘Lithium, Gmelins Handbuch – Der Anorganischen Chemie’, System No. 20 270–273, Verlag Chemie, GmbH, Weinhein/Bergstrasse, 1960. 42. A. W. Titherley, J. Chem. Soc. (1894), 504. 43. H. Jacobs, R. Juza, Z. Anorg. Allg. Chem. 391 (1972), 271. 44. M. Nagib, H. Jacobs, Atomkernenergie 21 (1973), 275. 45. J. B. Yang, X. D. Zhou, Q. Cai, W. J. James, W. B. Yelon, Appl. Phys. Lett. 88 (2006), 041914. 46. M. H. Sørby, Y. Nakamura, H. W. Brinks, T. Ichikawa, S. Hino, H. Fujii, B. C. Hauback, J. Alloys Comp. 428 (2007), 297. 47. K. Miwa, N. Ohba, S. Towata, Y. Nakamori, S. Orimo, Phys. Rev. B 71 (2005), 195109. 48. P. Chen, Z. Xiong, J. Luo, J. Lin, K. L. Tan, J. Phys. Chem. B 107 (2003), 10967. 49. T. Ichikawa, S. Isobe, N. Hanada, H. Fujii, J. Alloys Comp. 365 (2004), 271. 50. Y. H. Hu, E. Ruckenstein, Ind. Eng. Chem. Res. 42 (2003), 5135. 51. Y. H. Hu, E. Ruckenstein, J. Phys. Chem. A 107 (2003), 9737. 52. T. Ichikawa, N. Hanada, S. Isobe, H. Leng, H. Fujii, J. Phys. Chem. B 108 (2004), 7887. 53. G. P. Meisner, F. E. Pinkerton, M. S. Meyer, M. P. Balogh, M. D. Kundrat, J. Alloys Comp. 404–406 (2005), 24. 54. F. E. Pinkerton, J. Alloys Comp. 400 (2005), 76. 55. S. Hino, T. Ichikawa, N. Ogita, M. Udagawa, H. Fujii, Chem. Commun. (2005), 3038. 56. W. I. F. David, M. O. Jones, D. H. Gregory, C. M. Jewell, S. R. Johnson, A. Walton and P. P. Edwards, J. Am. Chem. Soc. 129 (2007), 1594. 57. Y. H. Hu, N. Y. Yu, E. Ruckenstein, Ind. Eng. Chem. Res. 43 (2004), 4174. 58. Y. H. Hu, N. Y. Yu, E. Ruckenstein, Ind. Eng. Chem. Res. 44 (2005), 4304. 59. Y. H. Hu, E. Ruckenstein, Ind. Eng. Chem. Res. 43 (2004), 2464. 60. Y. H. Hu, E. Ruckenstein, Ind. Eng. Chem. Res. 45 (2006), 182. 61. Y. H. Hu, E. Ruckenstein, Ind. Eng. Chem. Res. 44 (2005), 1510. 62. Y. H. Hu, E. Ruckenstein, Ind. Eng. Chem. Res. 45 (2006), 4993. 63. T. Ichikawa, N. Hanada, S. Isobe, H.Y. Leng, H. Fujii, J. Alloys Comp. 404–406 (2005), 435. 64. S. Isobe, T. Ichikawa, N. Hanada, H.Y. Leng, M. Fichtner, O. Fuhr, H. Fujii, J. Alloys Comp. 404–406 (2005), 439. 65. J. H. Yao, C. Shang, K. F. Aguey-Zinsou, Z. X. Guo, J. Alloys Comp. 432 (2007), 277. 66. Y. Kojima, Y. Kawai, Chem. Commun. (2004), 2210. 67. Y. Kojima, Y. Kawai, N. Ohba, J. Power Sources 159 (2006), 81. 68. H. Y. Leng, T. Ichikawa, S. Isobe, S. Hino, N. Hanada, H. Fujii, J. Alloys Comp. 404–406 (2005), 443. 69. T. Markmaitree, R. Ren, L. L. Shaw, J. Phys. Chem. B 110 (2006), 20710.

© 2008, Woodhead Publishing Limited

476 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104.

Solid-state hydrogen storage Y. Nakamori, S. Orimo, Mater. Sci. Eng. B 108 (2004), 48. Y. Nakamori, S. Orimo, J. Alloys Comp. 370 (2004), 271. J. Etourneau, J. Portier, F. Monil, J. Alloys Comp. 188 (1992), 1. S. Orimo, Y. Nakamori, G. Kitahara, K. Miwa, N. Ohba, T. Noritake, S. Towata, Appl. Phys. A 79 (2004), 1765. H. M. Jin, P. Wu, Appl. Phys. Lett. 87 (2005), 181917. C. Zhang, A. Alavi, J. Phys. Chem. B 110 (2006), 7139. M. Yoshino, K. Komiya, Y. Takahashi, Y. Shinzato, H. Yukawa, M. Morinaga, J. Alloys Comp. 404–406 (2005), 185. Z. Xiong, P. Chen, G. Wu, J. Lin, K. L. Tan, J. Mater. Chem. 13 (2003), 1676. F. W. Dafert, R. Miklauz, Monatsh. Chem. 7 (1909), 649. S. Hino, T. Ichikawa, H. Leng, H. Fujii, J. Alloys Comp. 398 (2005), 62. Y. Nakamori, G. Kitahara, S. Orimo, J. Power Sources 138 (2004), 309. J. Hu, Z. Xiong, G. Wu, P. Chen, K. Murata, K. Sakata, J. Power Sources 159 (2006), 120. J. Hu, G. Wu, Y. Liu, Z. Xiong, P. Chen, K. Murata, K. Sakata, G. Wolf, J. Phys. Chem. B 110 (2006), 14688. W. Luo, J. Alloys Comp. 381 (2004), 284. Z. Xiong, G. Wu, J. Hu, P. Chen, Adv. Mater. 16 (2004), 1522. H. Y. Leng, T. Ichikawa, S. Hino, N. Hanada, S. Isobe, H. Fujii, J. Phys. Chem. B 108 (2004), 8763. W. Luo, E. Rönnebro, J. Alloys Comp. 404–406 (2005), 392. J. Lu, Z. Fang, H. Y. Sohn, Inorg. Chem. 45 (2006), 8749. Y. Nakamori, G. Kitahara, K. Miwa, S. Towata, S. Orimo, Appl. Phys. A 80 (2005), 1. Y. Nakamori, G. Kitahara, K. Miwa, N. Ohba, T. Noritake, S. Towata, S. Orimo, J. Alloys Comp. 404–406 (2005), 396. Z. Xiong, J. Hu, G. Wu, P. Chen, W. Luo, K. Gross, J. Wang, J. Alloys Comp. 398 (2005), 235. H. Kawano, Y. Zhu, A. Tanaka, S. Sugimoto, Thermochim Acta 371 (2001), 155. P. Chen, Z. Xiong, L. Yang, G. Wu, W. Luo, J. Phys. Chem. B 110 (2006), 14221. W. Luo, S. Sickafoose, J. Alloys Comp. 407 (2006), 274. Y. Chen, C.-Z. Wu, P. Wang, H.-M. Cheng, Int. J. Hyd. Energy 31 (2006), 1236. R. Juza, E. Eberuis, Naturwiss. 49 (1962), 104. Z. Xiong, G. Wu, J. Hu, P. Chen, W. Luo, J. Wang, J. Alloys Comp. 417 (2006), 190. T. Ichikawa, H. Y. Leng, S. Isobe, N. Hanada, H. Fujii, J. Power Sources 159 (2006), 126. H. Leng, T. Ichikawa, S. Hino, T. Nakagawa, H. Fujii, J. Phys. Chem. B 109 (2005), 10744. T. Ichikawa, K. Tokoyoda, H. Leng, H. Fujii, J. Alloys Comp. 400 (2005), 245. H. Leng, T. Ichikawa, H. Fujii, J. Phys. Chem. B 110 (2006), 12964. T. Nagakawa, T. Ichikawa, R. Iida, H. Leng, N. Takeichi, T. Kiyobayashi, H. Takeshita, H. Fujii, J. Alloys Comp. 430 (2007), 217. K. Okamoto, K. Tokoyoda, T. Ichikawa, H. Fujii, J. Alloys Comp. 432 (2007), 289. M. Aoki, T. Noritake, G. Kitahara, Y. Nakamori, S. Towata, S. Orimo, J. Alloys Comp. 428 (2007), 307. G. Wu, Z. Xiong, T. Liu, Y. Liu, J. Hu, P. Chen, Y. Feng, A. T. S. Wee, Inorg. Chem. 46 (2007), 517.

© 2008, Woodhead Publishing Limited

Imides and amides as hydrogen storage materials

477

105. J. Hu, Z. Xiong, G. Wu, P. Chen, K. Murata, K. Sakata, J. Power Sources 159 (2006), 116. 106. V. Schultz-Coulon, W. Schnick, Z. Naturforsch. B 50 (1995), 619. 107. Y. Liu, J. Hu, Z. Xiong, G. Wu, P. Chen, K. Murata, K. Sakata, J. Alloys Comp. 432 (2007), 298. 108. Y. Liu, T. Liu, Z. Xiong, J. Hu, G. Wu, P. Chen, A. T. S. Wee, P. Yang, K. Murata, K. Sakata, Eur. J. Inorg. Chem. (2006), 4368. 109. Z. Xiong, J. Hu, G. Wu, P. Chen, J. Alloys Comp. 395 (2005), 209.

© 2008, Woodhead Publishing Limited

17 Multicomponent hydrogen storage systems G. W A L K E R, University of Nottingham, UK

17.1

Introduction

The challenge that faces solid-state hydrogen storage materials for automotive applications is the combined need for a high gravimetric storage capacity and a working temperature below 200 °C (volumetric capacity, although important, is only a significant issue for the physisorption systems as most non-physisorption materials have hydrogen capacities greater than 80 g l–1). The dilemma is that the higher gravimetric capacity materials are hydrides of elements from rows 2 and 3 of the periodic table and these tend to have strong ionic or covalent bonds with hydrogen. For example, the binary hydrides of lithium and magnesium have hydrogen contents of 12.7 and 7.6 wt%, respectively. However, these elements are from the left-hand side of the periodic table and have low Pauling electronegativities (0.98 and 1.31, respectively; cf. 2.20 for hydrogen). This leads to thermodynamically very stable ionic hydrides (∆Hf = –90.5 and –75.3 kJ mol–1, respectively). As we move to the right along the periodic table, elements have electronegativities closer to that of hydrogen (e.g. 2.55 for carbon) and form relatively stable covalent hydrides (e.g. methane, ∆Hf = –74.9 kJ mol–1). Even the complex hydrides of boron have high thermodynamic stabilities such as lithium tetrahydridoborate, LiBH4 (∆Hf = –190.8 kJ mol–1). These very negative enthalpies of formation of course mean there is a correspondingly large endotherm for the dehydrogenation of these hydrides back to their elements, resulting in high temperatures required for an equilibrium hydrogen pressure of 1 bar, T(1 bar), because the Gibbs free energy is dependent on the enthalpy of reaction: ∆G = ∆H – T∆S

[17.1]

and the equilibrium constant, Kp, is related to Gibbs free energy by: ∆G = –RT ln Kp 478 © 2008, Woodhead Publishing Limited

[17.2]

Multicomponent hydrogen storage systems

479

where R is the gas constant. Given that Kp for a solid hydride decomposing to produce 1 mole of hydrogen is, Kp = p(H2)

[17.3]

Eq. (17.2) shows us that ∆G will be zero for an equilibrium pressure of 1 atm. Hence Eq. (17.1) rearranges to: T (1bar) =

∆H ∆S

[17.4]

Given that ∆S is very similar for all the solid hydride dehydrogenations, because these reactions normally liberate one mole of gas from a solid, Eq. (17.4) illustrates the importance of the enthalpy of dehydrogenation upon T(1 bar), i.e. a larger dehydrogenation endotherm results in a higher T(1 bar). Research to try to develop systems based on these high-capacity storage materials have investigated means of accelerating the rate of reaction through creating nanoparticulate materials, e.g. ball milled MgH2 (Huot et al., 1999), producing defective structures with smaller diffusion path lengths to increase the diffusion rate of hydrogen. Also a wide variety of catalysts have been investigated which lower the activation energy, Ea, for hydrogenation and dehydrogenation, e.g. for MgH2, platinum group metals (Gutfleisch et al., 2003), transition metals (Hanada et al., 2005) and oxides (Oelerich et al., 2001; Croston, 2007). However, these strategies only change the kinetics of reaction and do not alter the thermodynamics, i.e. T(1 bar) remains unaltered, viz. 280 °C for MgH2 (although there is some interesting theoretical work suggesting nanostructures of MgH2 with dimensions smaller than 3 nm will have a reduced thermodynamic stability; Wagemans et al., 2005). Destabilisation is a strategy pioneered by Reilly and Wiswall back in the 1960s, where the basic concept was to use alloys to make the hydrides less thermodynamically stable (Reilly and Wiswall, 1968). A classic example is Mg2Ni which can be hydrogenated to form Mg2NiH4, but although there is a modest reduction in the thermodynamic stability (T(1 bar) = 255 °C), there is also a significant loss in hydrogen capacity (3.6 wt%). Such systems are examples of destabilisation for a single phase hydride. Multicomponent hydrogen storage systems (also referred in the literature as reactive hydride composites) comprise of more than one phase in the hydrogen loaded state. These can be merely physical mixtures or a composite (i.e. where one phase acts as a host matrix). Both phases react during dehydrogenation, hence one phase is not simply acting as a catalyst, and often the dehydrogenated products consist of more than one phase (although this does not have to be the case). Given the high hydrogen capacities of complex hydrides such as borohydrides, alanates and amides, it is of no surprise that many of the systems studied have a complex hydride as one of the phases, the other phase typically being

© 2008, Woodhead Publishing Limited

480

Solid-state hydrogen storage Y + Z + H2

X + Y + Z + H2 YZ + H2

Enthalpy

X + YZ + H2

∆Ha

∆H b

∆H a

YH2 + Z

∆Hb

XYH2 + Z

17.1 Enthalpy diagram showing the smaller dehydrogenation endotherm for the destabilised multicomponent system. The dehydrogenation product on the right is for formation of a single phase on the left and for two product phases on the right.

a metal hydride (e.g. LiBH4–MgH2), but addition of other complex hydride phases has also been tried (e.g. LiAlH4–LiNH2). This chapter will first explain the theory of destabilisation and then move on to look at specific examples starting with the destabilisation of complex hydrides with separate sections devoted to different classes of destabilisation agents: first metal hydrides, then non-hydride systems (where either an element or alloy which does not hydride or an inorganic salt is added as the destabilisation agent) and finally attempts to use another complex hydride to form a destabilised multicomponent system. The penultimate section summarises the work reported on other multicomponent hydrogen storage systems, predominantly the destabilisation of binary hydrides by other elements, and the chapter is concluded with an outlook of potential future research.

17.2

Thermodynamic destabilisation

The underlying principle to destabilisation is best illustrated using an enthalpy diagram such as that shown in Fig. 17.1. The example on the left would be illustrative of destabilisation of a binary hydride, YH2. Without the addition of a second phase, YH2 will decompose, forming the elements with a change in enthalpy of ∆Ha (i.e. –∆Hf(YH2), where ∆Hf is the enthalpy of formation). Introducing a second phase, Z, to YH2 now allows an alternative reaction to occur where Y can combine with Z to form YZ, which has a lower change in enthalpy, ∆Hb (i.e. ∆Hb = ∆Ha + ∆Hf(YZ)). For example Si can be added to MgH2 (Vajo et al., 2004) to reduce the enthalpy of dehydrogenation via the following reaction:

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

481

2MgH2 + Si → Mg2Si + 2H2 The reactions on the right in Fig. 17.1 give an example common for complex hydrides, where the hydrided material dehydrogenates to form two other product phases X and Y. For example, lithium tetrahydridoborate, LiBH4, dehydrogenates to form the binary hydride of lithium and the elements boron and hydrogen: LiBH4 → LiH + B + 3/2H2 Note that for this example, not all the dehydrogenation products are elements. This is because lithium hydride itself is a very stable compound and requires much higher temperatures to decompose it than one normally employs (> 600 °C). Adding a suitable second phase, Z, to destabilise XYH2 leads to an alternative reaction where Z combines with one of the elements in the hydride forming YZ. For example, MgH2 can be added to LiBH4, resulting in the following reaction (Vajo et al., 2005): 2LiBH4 + MgH2 s 2LiH + MgB2 + 4H2 For both examples the change in enthalpy for the destabilised reaction, ∆Hb, is less than that for the hydride on its own, ∆Ha, and thus the destabilised reaction will have a correspondingly lower T(1bar). To put it another way, at a given temperature the destabilised multicomponent hydride will have a higher plateau pressure than the hydride on its own. One might wonder about the thermal stability of the intermediate compound formed, YZ, with respect to the elements. However, a significant difference is that for XY to decompose back to its elements X and Y there will be a very small change in entropy (assuming all phases are solids). Thus from equation (16.1) we can see that if ∆S is small high temperatures will be required to make ∆G negative (i.e. become a spontaneous reaction). It has been argued that ∆S for most metal hydrides will approximately equal the entropy of the evolved hydrogen, c. 130 J K–1mol–1(H2) (Züttel et al., 2003), but complex hydrides can have a lower ∆S, e.g. 95 J K–1mol– 1 (H2) as can be determined for LiBH4 from standard thermodynamic data (Lide, 2002). Using these two values for ∆S one can estimate with equation (17.4) the ∆H required for T(1 bar) to be less than 200 °C; for metal hydrides it is 61 kJ mol–1(H2) and for some complex hydrides it is 45 kJ mol–1(H2). There are further limitations: if destabilisation results in a negative ∆H (i.e. dehydrogenation becomes exothermic), this has the effect of making the hydrided state thermodynamically unstable and would result in a non-reversible dehydrogenation reaction because hydrogenation would have a positive change in enthalpy and also a negative change in entropy (assuming no gaseous species are formed during hydrogenation) and thus ∆G would always be positive. Thinking about the practicalities of a hydrogen store, it would be

© 2008, Woodhead Publishing Limited

482

Solid-state hydrogen storage

undesirable to use a material that had a hydrogen equilibrium pressure of 200 bar at room temperature. Releasing the hydrogen from the store is not the problem (although heavier cylinders would be needed to cope with the pressure), but being able to provide an adequate over-pressure for a sufficiently fast charging rate starts to become an issue. Perhaps more important, though, are safety concerns during small temperature excursions (either from a hot day or from a fire): merely increasing the temperature of the store by 25 °C would lead to a doubling of the plateau pressure to over 400 bar. Hence, the target range of ∆H for practical systems is 26–61 kJ mol–1(H2) for ∆S = 130 J K–1mol–1(H2) and 15–45 kJ mol–1(H2) for ∆S = 95 J K–1mol–1(H2). This section has primarily looked at the effect of destabilisation on the thermodynamics for the dehydrogenation reaction. It does of course affect the hydrogenation reaction too. Figure 17.1 shows that the enthalpy for hydrogenation is less negative for the destabilised system and hence ∆G will be less negative too (for a given temperature). Hence, these destabilised systems have a reduced thermodynamic driving force for the hydrogenation reaction. One must not ignore the kinetics for these multicomponent systems which involve solid–solid reactions for the dehydrogenation and for most examples for the rehydrogenation too. Diffusion of elements in the solid phase is often slow, leading to poor kinetics, and will only get worse approaching the low temperatures required for many storage applications. During cycling there may be segregation and sintering of the phases, making it difficult to reverse the concerted solid–solid reaction. It is therefore important that the phases remain intimately mixed to reduce diffusion path lengths. Some ideas in the literature on how to minimise such factors are given in the final section (17.7) after discussing the theoretical and experimental results for a variety of multicomponent hydrogen storage systems.

17.3

Complex hydride–metal hydride systems

The most widely researched multicomponent hydride system is that of LiBH4– MgH2. This was first reported by Vajo et al. (2005) who reported that a 2:1 molar mixture of the hydrides would decompose under a hydrogen atmosphere, forming lithium hydride and magnesium boride. Interestingly, decomposing the mixture under a dynamic vacuum did not lead to the formation of magnesium boride and subsequently LiBH4 could not be re-formed when they tried to rehydrogenate at 400 °C and 100 bar. In the literature there are some discrepancies over the change in enthalpy for this destabilised reaction, which is in part due to the fact that the LiBH4 undergoes a phase change from orthorhombic to hexagonal crystal structure at c. 110 °C and the hexagonal form melts at c. 270 °C (see Chapter 15). Both transitions are endothermic and have an effect on the change in enthalpy for the destabilised dehydrogenation reaction as illustrated in Fig. 17.2.

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

483

2Li + Mg + 2B + 5H2 2LiH + Mg + 2B + 4H2

Enthalpy

2LiH + MgB2 + 4H2

2LiBH4(l) + Mg + H2

2LiBH4(l) + MgH2 2LiBH4(hex) + MgH2 2LiBH4(ortho) + MgH2

17.2 Change in enthalpy for the LiBH4–MgH2 system (not to scale).

Most of the thermodynamic data needed to calculate the enthalpy changes shown in Fig. 17.2 can be found in standard tables of thermodynamic data, e.g. the NIST-JANAF database (Chase, 1998), as given in Table 17.1; however for the transitions for LiBH 4 (from orthorhombic→hexagonal and hexagonal→liquid), experimental data from the literature has been used. Using these data the changes in enthalpy and entropy and T(1 bar) values can be calculated, which are given in Table 17.2 along with values reported in the literature. Table 17.1 Enthalpy and entropy data for the individual compounds relevant to the LiBH4–MgH2 multicomponent system Compound

∆Hf (kJ mol–1)

S (J K–1 mol–1)

Reference

H2 B Li Mg LiH MgH2 MgB2 LiBH4

0 0 0 0 –90.500 –75.300 –91.964 –190.800

130.68 5.9 29.12 32.67 20 31.1 35.98 75.9

Lide (2002) Lide (2002) Lide (2002) Lide (2002) Lide (2002) Lide (2002) Chase (1998) Chase (1998)

Transition

∆H (kJ mol–1)

∆S (J K–1 mol–1)

LiBH4 ortho → hex LiBH4 hex → liq

4.38 8.03

11.4 14.5

© 2008, Woodhead Publishing Limited

Price et al. (2008) Price et al. (2008)

484

Solid-state hydrogen storage

Table 17.2 Calculated and measured changes in enthalpy and entropy, and the corresponding T(1bar) values for the dehydrogenation of LiBH4, MgH2 and mixtures of both hydrides, forming LiH and accompanying products Dehydrogenation

∆H ∆S (kJ mol–1 (H2)) (J K–1 mol–1 (H2))

T(1 bar) Reference* (°C)

LiBH4(ortho) LiBH4(l) MgH2 2LiBH4(ortho) + MgH2

66.9 58.6 75.3 46.0 45.8 t

414 459 296 169 168t

97.3 80.1 132.3 104.0

46 t

52.2

169t t

105.6

t

165t

2LiBH4(hex) + MgH2

43.8 49.5t

98.3 95.6t

173

2LiBH4(l) + MgH2

39.8 40.5c

91.0 81.3c

164 225c

Vajo and Olson (2007) Barkhordarian et al. (2007) Cho et al. (2006) Alapati et al. (2007b) Alapati et al. (2007b) Vajo et al. (2005)

c

Calculated from experimental results. Theoretical calculation. * Unreferenced values are calculated from the data in Table 17.1. t

The data in Table 17.2 show that the T(1 bar) for the decomposition of the orthorhombic form of LiBH4 reacting with MgH2 (169 °C) is well above the orthorhombic→hexagonal phase transition (110 °C). Taking into account the change in enthalpy and entropy for this phase transition (4.38 kJ mol–1 and 11.4 J K–1 mol–1 respectively) the calculated T(1bar) for the destabilised reaction with the hexagonal form of LiBH4 is 173 °C. However, the hydrogenation and dehydrogenation reactions were investigated at temperatures in excess of 300 °C, which is above the melting point of LiBH4. Taking into account the change in enthalpy and entropy of fusion, the calculated T(1bar) drops back to 164 °C. Taking these two transitions into account does not make a big difference to the T(1bar) but does significantly alter the change in enthalpy and entropy for the destabilised reaction, 39.8 kJ mol–1(H2) and 91.0 J K– 1 mol–1 (H2) respectively. These latter theoretical values are in much better agreement with those calculated from a van’t Hoff plot of isotherm data measured at and above 315 °C by Vajo et al. (2005), viz. 40.5 kJ mol–1(H2) and 81.3 J K–1 mol–1(H2) respectively. The hydrogenation isotherms measured by Vajo et al. (2005) clearly show that destabilisation occurs. Measuring these isotherms was hampered by slow kinetics, resulting in sloping plateaus and some incomplete isotherm curves. However, if this were not a concerted reaction one would expect to

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

485

see a double plateau, the first from the dehydrogenation of MgH2, which was not the case. The isotherms also had plateaus at much higher pressures than the plateau pressures expected from either MgH2 or LiBH4 decomposing on their own (see Fig. 17.3). Much of the work reported used temperature programmed experiments such as thermal gravimetric analysis with mass spectroscopy of the product gases (TGA-MS). Figure 17.4 shows typical TGA-MS data for the decomposition of a 2:1 mixture under flowing argon (Price et al., 2008). This plot highlights that the decomposition occurs in steps: first the MgH2 decomposes at just above 300 °C and then the LiBH4 decomposes at 415 °C. Similar steps and temperatures are found regardless of whether an inert atmosphere or different pressures of hydrogen are used (Vajo et al., 2005; Pinkerton et al., 2007). For a destabilised reaction one would expect to observe both phases decomposing together; however, given the slow kinetics anecdotally reported for the isotherm data above (Vajo et al., 2005), it is unsurprising that kinetic control takes over with these temperature programmed experiments leading to the two phases decomposing separately. However, this does not prevent the formation of the magnesium boride product during the decomposition of the borohydride in the presence of hydrogen. In situ X-ray diffraction (XRD) has been used to monitor this reaction (Bosenberg et al., 2007) and the results for decomposing under 5 bar hydrogen using a 5 °C min–1 heating ramp clearly demonstrated the dehydrogenation 100

Pressure (bar)

2L 2Li

10

iBH

4 –0

BH

4 –1

+M gH

2

+M gH 2

Exp

t

1

LiB

0.1 0.001

0.0012

LiB H

4 –1

0.0014 0.0016 1/T (K–1)

M H

4–

gH

2

0

0.0018

0.002

17.3 A van’t Hoff plot showing the expected equilibrium pressures for the dehydrogenation of LiBH4 in both its orthorhombic and molten states (–0 and –1 suffix used to identify the respective state), MgH2 and the destabilised system LiBH4–MgH2 with a 2:1 molar ratio. The experimental data reported by Vajo et al. (2005) are also plotted with a line of best fit for comparison.

© 2008, Woodhead Publishing Limited

486

Solid-state hydrogen storage 100

Intensity (a.u.)

Weight (%)

98 96 94 92 90 88

0

100

200

300 400 Temperature (°C)

500

600

17.4 TGA-MS results for the evolution of H2 from LiBH4 + MgH2 sample with a 2:1 molar ratio, milled for 1 h. Experiments used an argon carrier gas and heating rate of 10 °C min–1.

proceeded first by the decomposition of the MgH2, forming the metal, as shown in Fig. 17.5. This was followed by the decomposition of the LiBH4 which (because it was molten and therefore X-ray amorphous) was identified by the simultaneous evolution of lithium hydride and magnesium boride phases, plus a concomitant reduction in intensity of the magnesium metal pattern as this reacted with the available boron. The products of dehydrogenation showed good rehydrogenation kinetics using 2–3 mol.% of a TiCl3 catalyst, the product after cycling twice took under 2 hours to rehydrogenate to a capacity of just over 9 wt% hydrogen at 300 °C and under 100 bar of hydrogen, the hydrogen uptake increasing to 10 wt% when the temperature was increased to 350 °C (Pinkerton et al., 2007). High-pressure differential scanning calorimetry (HP-DSC) under 50 bar hydrogen showed that rehydrogenation begins at 240 °C and in-situ XRD showed that a small amount of both MgH2 and LiBH4 had formed after hydrogenating for over 5 hours at 265 °C under 150 bar of hydrogen. Even though the conversion rate was only small, these are exceptionally low temperatures for the formation of LiBH4. The effect of using non-stoichiometric mixtures of LiBH4–MgH2 was investigated by Yu et al. (2006). It was found that with increasing the amount of MgH2 in the mixture there was a small shift to lower temperatures for the second dehydrogenation step (i.e. that of the LiBH4). However, increasing the MgH2 level decreased the storage capacity for the system. A molar ratio of 1:0.3 was selected as a compromise between lowering of the dehydrogenation temperature whilst maintaining a maximum storage capacity of 9.8 wt%. TGA-MS results shown in Fig. 17.6 clearly show that under an inert atmosphere

© 2008, Woodhead Publishing Limited

0.5

1.0

1.5

4*π sinθ/λ/Å–1 2.0

70 125

A LiBH4 tetragonal

20

Time (min)

3.0

MgH2

LiBH4 orthorhombic 10

2.5

200

B

30

487

T/°C

Multicomponent hydrogen storage systems

270

40

380

Mg MgB2

50

450

0.5

1.0

→ Isothermal condition

?

60

LiH

1.5

2.0

2.5

3.0

17.5 In situ XRD measurement of the dehydrogenation of 2LiBH4 + MgH2 (with 5 at.% titanium isopropoxide catalyst) under 5 bar of hydrogen with a heating rate of 5 °C min–1. A and B mark the orthorhombic – hexagonal phase transition and melting point of LiBH4, respectively and ? corresponds to an unknown phase (Bosenberg et al., 2007).

354 °C

100 98 96

405 °C

94

2.65 wt% (b) 0.85 wt%

200

92 (a) 90

585 °C 100

Weight (%)

Intensity (a.u.)

5.7 wt%

300 400 500 Temperature (°C)

600

700

17.6 (a) TGA and MS results for the evolution of H2 from 0.3 LiBH4 + MgH2 milled for 1 h. (b) TGA results for LiBH4. Both experiments used an argon carrier gas and heating rate of 10 °C min–1 (Yu et al., 2006).

© 2008, Woodhead Publishing Limited

488

Solid-state hydrogen storage

Intensity (a.u.)

there are three dehydrogenation steps. XRD of the end-products surprisingly showed no lithium hydride, i.e. all the hydrogen in the system had been liberated. In situ neutron diffraction (ND) was used to investigate the intermediates during decomposition. Figure 17.7 shows the evolving ND patterns for two experiments: the first experiment was the decomposition of the 1:0.3 mixture in an evacuated sealed vessel, i.e. decomposition under a self-generated deuterium atmosphere; and the second experiment was the decomposition of the same material but under a dynamic vacuum. As the sample was heated for the first experiment (with a self-generated deuterium atmosphere) the first change was the

730 430 600 330

35

40

45

485 50

60

65

m Te

150 25

ra

tu

Intensity (a.u.)

2θ (°)

425 55

pe

re

(°C

330

)

730 25 600

35

485

40 45

50 2θ (°)

425 55

150

60 65

MgD2

LiBD4 (ortho)

LiBD4(hex)

Mg

LiD

m Te

pe

ra

t

e ur

( °C

)

25

Li0.184Mg0.816(alpha)

Li0.3Mg0.7(beta)

Stainless steel

17.7 In situ ND results for the reaction for 0.3 LiBH4 + MgH2 heated in a sealed vessel, top, and under a dynamic vacuum, bottom (Price et al., 2008).

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

489

orthorhombic to hexagonal phase transition for the LiBD4. The diffraction pattern for this phase then disappears as the temperature continues to increase past the melting point for LiBD4. The MgD2 then decomposes at c. 350 °C, producing Mg metal. There is little change in the diffraction pattern as the sample is heated further until the melting point of Mg is reached (650 °C) when all the diffraction patterns are lost. On cooling however, along with the Mg phase reappearing, LiD is also observed suggesting that the LiBD4 did not decompose until temperatures in excess of 650 °C. The increased decomposition temperature for the LiBD4 is due to the relatively high pressure of the self-generated deuterium atmosphere which would have been over 30 bar after the decomposition of the MgD2 phase (cf. 7.4 bar equilibrium pressure for the decomposition of LiBD4 at 650 °C). This, however, is a result of kinetic stabilisation of the LiBD4 as the plateau pressure for the destabilised reaction is above 100 bar. Eventually, though, the thermodynamically favourable decomposition reaction occurs, as evidenced by the formation of LiD. The overall reaction for this decomposition is: MgD2 + 0.30LiBD4 s 0.70Mg + 0.30LiD + 0.30MgB2 + 1.45D2 Undertaking the decomposition under a dynamic vacuum follows the same initial steps, phase transition and melting of LiBD4 followed by the decomposition of MgD2 forming Mg. However, under these conditions a LiD phase is formed as a result of the decomposition of the LiBD4 (i.e. the second step in the TGA-MS data in Fig. 17.5). The LiD phase is not retained and this reflection is gradually lost, concomitant with this decrease in intensity for the LiD line is a shift in the d-spacing for the Mg pattern (the largest change is for the (002) reflection at a 2θ value of 55°, moving to higher 2θ values. This corresponds to the formation of a Mg–Li alloy resulting from the lithium generated by the decomposition of LiD alloying with the magnesium present. The α-Mg–Li alloy has a maximum lithium concentration of 0.184 and as the lithium concentration exceeds that limit a second phase (the βalloy) starts to form (Yu et al., 2006). This can be seen in Fig. 17.7, where there is the appearance of a reflection due to the β-alloy, while the intensity for the α-alloy reflections decreases. The ND results confirmed that the dehydrogenation reactions under an inert atmosphere or vacuum are: •

Step 1 355 ° C MgH 2   → Mg + H 2



Step 2 405 ° C, Mg 0.30LiBH 4   → 0.30LiH + 0.30B + 0.45H 2



Step 3 > 428 ° C Mg + 0.3LiH  → 0.52Mg 0.70 Li 0.30 + 0.78Mg0.816Li0.184

+ 0.15H2 © 2008, Woodhead Publishing Limited

490



Solid-state hydrogen storage

Overall reaction MgH2 + 0.3LiBH4 s 0.52Mg0.7Li0.3 + 0.78Mg0.816Li0.184 + 0.3B + 1.60H2

It would therefore appear that in step 2, the magnesium is acting as a catalyst for the decomposition of LiBH4. It is interesting to note that LiBH4 has also been shown to be a catalyst for the decomposition of MgH2 (Johnson et al., 2005; Mao et al., 2007). These alloy and boron end-products can be rehydrogenated, which has been shown at 400 °C and 100 bar, but no work has yet been done to investigate the rehydrogenation reaction under milder conditions. Interestingly, the tetrahydridoborate was also formed in this reaction although there was no magnesium boride phase present. This would appear to be contrary to the work reported on the 2:1 system which found no reversibility if MgB2 was not formed. For the more Mg-rich system the boron must be in a more reactive state, which may simply be due to the lower concentration of B for this system remaining highly dispersed in the dehydrogenated products, whereas for higher concentrations of B, this is more likely to lead to sintering, resulting in reduced activity for hydrogenation. When MgB2 is present, the rate-limiting step is related to the reaction of the boron in the boride either combining with the LiH and/or in forming the tetrahydridoborate ion. However, there are currently few experiments which have probed this rehydrogenation mechanism. There has been interest in whether other borides would be better coreactants in a destabilised system with LiBH4 or whether MgB2 could improve the kinetics of hydrogenation for other metal borohydrides. Alapati et al. (2006, 2007a,b) have proposed a number of potential destabilised systems, using first-principle calculations to predict the thermodynamics for these dehydrogenations. For alternative borohydride–metal hydride systems, the most promising combinations were LiBH4 with scandium hydride and with calcium hydride, forming scandium boride (ScB2) and calcium boride (CaB6) respectively. The reverse reaction, hydrogenation of metal borides plus metal hydrides, has been investigated along with other potential destabilised systems (Barkhordarian et al., 2007), but it was found that even at 400 °C and under 350 bar of hydrogen, no reaction had occurred between calcium boride and lithium hydride. Hydrogenation of boron carbide (B4C) was also tried unsuccessfully with lithium hydride under the same conditions. Hydrogenation of other binary hydrides (viz. Mg, Na, Ca, Zr, Y, La, Nd and Pr) ball milled with MgB2 to form the respective borohydride of the binary hydride was attempted (Barkhordarian et al., 2007). Only sodium tetrahydridoborate and calcium tetrahydridoborate were successfully formed, both requiring 200 bar of hydrogen and 300 °C. However, the destabilised system with sodium tetrahydridoborate has too high a theoretical T(1 bar) of 351 °C (even though it has a hydrogen capacity of 7.8 wt%). For the calcium tetrahydridoborate

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

491

system there was no T(1bar) calculated because of a lack of thermodynamic data for this complex hydride, but the storage capacity is lower 8.3 wt% and under the conditions above only had a capacity of 4.7 wt% after 48 hours. High-pressure DSC experiments showed that hydrogenation starts at 250 °C (under 140 bar of H2) and dehydrogenation starts at 320 °C (under 3 bar of H2) (Dornheim et al., 2007). Alanates and amides have also been studied. First-principle calculations suggest calcium alanate has many potentially attractive combinations, but calcium alanate itself is likely to be too unstable (Alapati et al., 2007a). There is also a lot of interest in lithium hydride with lithium amide (Chen et al., 2002) and with magnesium amide (Leng et al., 2004), but these are discussed in detail in Chapter 15.

17.4

Complex hydride–non-hydride systems

Phases other than hydrides or complex hydrides can also be potential destabilisation components. In general these are less attractive combinations as they can lead to much lower hydrogen capacities for the destabilised system as one phase stores no hydrogen. Vajo and Olsen (2007) proposed using salts of magnesium rather than using MgH2 which would potentially give the following generic reaction: 2LiBH4 + MgX2 → 2LiX + MgB2 + 4H2 where X is a halide or a chalcogenide. The three salts investigated further were for X = F, S and Se. The respective ∆H for the destabilised dehydrogenations was calculated to be 45 kJ mol–1(H2), 47 kJ mol–1(H2) and 36 kJ mol–1(H2), giving T(1 bar) values of 150, 170 and 70 °C respectively. The authors claim to have managed to partially hydrogenate (>75%) the respective lithium salts ball milled with MgB2, but there are currently no further experimental details in the open literature on the kinetics or reversibility of this system (although the authors mention that on dehydrogenation, Mg metal was detected). Metals and alloys have also been suggested as destabilisation agents for borohydrides. Cho et al. (2006) reported theoretical results for using aluminium to destabilise the tetrahydridoborates of lithium and sodium. 2LiBH4 + Al → 2LiH + AlB2 + 3H2 2NaBH4 + Al → 2NaH + AlB2 + 3H2 The T(1bar) for these reactions were 188 °C for the lithium example and 459 °C for the sodium case. Even though in the latter case the system shows a drop in the T(1 bar) value of 140 °C, it is still too high to warrant serious consideration. However, LiBH4–Al has a T(1 bar) below 200 °C and

© 2008, Woodhead Publishing Limited

Solid-state hydrogen storage

Intensity (a.u)

492

S4 S3 S2 S1 100

200

300 400 Temperature (°C)

500

600

17.8 MS results for evolution of H2 for LiBH4 (S1) and LiBH4–TiO2 sample with mass ratios of 4:1 (S2), 1:1 (S3), and 1:4 (S4) after 1.5 h ball milling. Temperature programmed reactions used an argon carrier gas and a heating rate of 10 °C min–1 (Yu et al., 2008).

a theoretical hydrogen storage capacity of 8.6 wt%, but is yet to be proved experimentally. Alloys have also been investigated with LiBH4, Barkhordarian et al. (2007) showed that magnesium–tin alloys could destabilise LiBH4. The alloy Li7Si2 ball milled with MgB2 could be hydrogenated at 400 °C under 200 bar of hydrogen. This produced the tetrahydridoborate, an alloy with magnesium and tin according to the reverse of the reaction equation below. However, the kinetics for the hydrogenation was found to be slow, with only 2.5 wt% hydrogen capacity after 48 hours (of a potential 6.3 wt% capacity). 28LiBH4 + 7Mg2Sn + Sn s 4Li7Si2 + 14MgB2 + 56H2 Metal oxides have been found to accelerate the decomposition of LiBH4 (Yu et al., 2008), for example ball milling the borohydride with titania was found to significantly lower the temperature of dehydrogenation. The MS results in Fig. 17.8 show that adding a small amount of titania (mass ratio of 4:1) results in a dehydrogenation peak featuring at 320 °C and with greater amounts of titania the dehydrogenation starts at even lower temperatures, e.g. for a mass ratio of 1:4 there is only one peak at 220 °C with an onset temperature at 150 °C. This reaction was investigated by in situ ND, which showed that there was a fast reaction between the hexagonal-LiBD4 and TiO2, that started just above 180 °C and was finished by 220 °C. The only crystalline end product was lithium titanate, LiTiO2. The proposed dehydrogenation reaction is:

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

493

LiBH4 + TiO2 → LiTiO2 + B + 2H2 but this system has only a modest hydrogen capacity of 3.7 wt% because it requires a large amount of the non-hydride phase. Carbon as a destabilising agent has received attention from a number of researchers, not least because there are a number of carbides and borocarbides known and these may be potential dehydrogenation end-products. It was hypothesised that boron carbide, B4C, was a potential end-product as the reaction given below would have a ∆H of 27 kJ mol–1(H2) (Barkhordarian et al., 2007). However no reaction occurred when ball milled LiH and B4C was attempted to be hydrogenated at 400 °C, under 350 bar of hydrogen. 4LiH + B4C + 6H2 → 4LiBH4 + C First-principle calculations suggested LiBH4 plus an equimolar amount of carbon would lead to a favourable destabilisation reaction (with a ∆H of 28.9 kJ mol–1H2) forming lithium borocarbide, LiBC (Alapati et al., 2007b): LiBH4 + C → LiBC + 2H2 However, it has been found that LiBH4 ball milled with carbon nanotubes dehydrogenated to form lithium carbide, Li2C2 (Yu et al., 2007). This was found not to be a reversible reaction as hydrogenating at 400 °C under 100 bar of hydrogen, lithium hydride was the only crystalline product detected by XRD. The authors suggest that carbon is also formed, but as there is no significant change in the XRD pattern for carbon before and after hydrogenation it is highly likely that the carbon has also hydrogenated forming gaseous hydrocarbons (interestingly a similar destabilisation reaction was postulated by Vajo and Olsen, 2007). 2LiBH4 + 2C → Li2C2 + 2B + 4H2 These examples with carbon illustrate how difficult it is to predict the reaction path that a multicomponent hydrogen storage system will follow. A number of ternary phase multicomponent hydrogen storage systems have been proposed (Alapati et al., 2007b), but these will not be discussed as there will be even greater uncertainty that the proposed reaction will occur and the kinetics are likely to be even worse as such systems will require mass transport between three phases rather than two.

17.5

Complex hydride–complex hydride systems

A reversible destabilised complex hydride–complex hydride multicomponent system has yet to be published. It would appear that although the starting phases are two complex hydrides, either these two phases react together to form a single phase complex hydride or the system upon rehydrogenation does not regenerate the starting phases. One such example is the LiBH4–

© 2008, Woodhead Publishing Limited

494

Solid-state hydrogen storage

LiNH2 system, which was proposed as a destabilised system, undergoing the following dehydrogenation reaction (Aoki et al., 2005): LiBH4 + 2LiNH2 → Li3BN2 + 4H2 First-principle calculations predict this reaction to have a favourable dehydrogenation enthalpy of 23 kJ mol–1(H2) (Aoki et al., 2005); however, upon ball milling these two complex hydrides together it was found that a new single phase had been formed. In fact in two other papers published shortly after, this same Li–B–N–H compound was reported (Pinkerton et al., 2005; Chater et al., 2006) and the compound was identified as a mixed borohydride amide salt of lithium, Li4(NH2)3BH4, with a body-centred cubic (BCC) crystal structure, having a lattice parameter a = 1.066 nm (Chater et al., 2006; Filinchuk et al. 2006). It should be noted that the stoichiometry of the mixed complex hydride does not match the borohydride : amide ratio used by Aoki nor Pinkerton and understandably these authors assumed that the formula of the mixed borohydride amide salt was Li3BN2H8. Although Li4(NH2)3BH4 has a high hydrogen content, 11.9 wt%, it unfortunately is not reversible. Another promising multicomponent system is lithium alanate–lithium amide, this was originally investigated not as a destabilised multicomponent system, but as a means of utilising the lithium amide to destabilise the lithium hydride formed from the decomposition of lithium alanate (i.e. following the reactions for the Li–N–H system as reported by Chen et al. (2002)) and hence release all the hydrogen contained within the alanate (Lu and Fang, 2005): 2LiAlH4 + LiNH2 → 2LiH + LiNH2 + Al + 3H2 → LiH + Li2NH + Al + 4H2 → Li3N + Al + 5H2 It was found that hand-milled mixtures started dehydrogenating at c. 80°C and evolved 8.1 wt% hydrogen by 310 °C. The products were identified as LiH, Al and Li2NH, indicating that the final step in the proposed reaction sequence above had not occurred. This three-stage decomposition was confirmed by Nakamori et al. (2006) who also showed that ball milling the mixture led to the partial decomposition of the materials, resulting in only c. 4 wt% hydrogen capacity after just 60 min ball milling. Other investigators have investigated ball milling different molar ratios 1:1 (Xiong et al., 2006) and 1:2 (Xiong et al., 2007) or using the hexahydridoaluminate, Li3AlH6 in a 1:2 ratio with the amide (Kojima et al., 2006). All the ball-milled mixtures were found to be partially dehydrogenated. In the latter reaction it was found the amide LiNH2 had been formed as determined by XRD and Fourier transform infrared (FTIR) (Kojima et al., 2006). This gave a reversible system with a capacity of 3.6 wt%; however, the alanate was never reformed and the cycling reaction was suggested to be:

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

495

3LiNH2 + 4LiH+ AlN → Li3AlN2 + 2Li2NH + 4H2 with the formation of the lithium aluminium nitride, Li3AlN2, leading to a destabilisation of the lithium amide–imide reaction proposed by Chen et al. (2002). For the 1:2 LiAlH4–LiNH2 system similar results were found, with the cycling reaction being similar to that above, but reflecting the smaller amide ratio in the starting mixture. LiNH2 + 2LiH + AlN s Li3AlN2 + 2H2 Although hydrogenation of the 2:1 alanate–amide dehydrogenation products was not attempted, there is no reason to believe this reaction will not also stop at the formation of LiH and LiNH2 like those for the 1:1 and 1:2 ratios.

17.6

Other destabilisation multicomponent systems

Although recent attention has been with destabilising complex hydride phases because these have high hydrogen capacities, some of the first examples of multicomponent destabilised systems were for binary hydrides destabilised by the addition of an element. Magnesium hydride was shown to be destabilised by mixing with aluminium, forming after dehydrogenation a Mg–Al intermetallic, e.g. β-Al3Mg2 and γ-Al12Mg17 (Zaluska et al., 2001). The system was found to be reversible and an equilibrium pressure for a 1:1 Mg– Al system was three times that for MgH2 at 280 °C, but this represents only a modest change in the enthalpy and corresponding T(1bar) values. A greater destabilisation was achieved using silicon as the destabilising agent. The dehydrogenation of this system resulted in the formation of magnesium silicide, Mg2Si, following the reaction equation below (Vajo et al., 2004): 2MgH2 + Si → Mg2Si + H2 This system has a hydrogen storage capacity of 5 wt% and a calculated enthalpy for dehydrogenation of 36.4 kJ mol–1(H2), which gives a T(1bar) value of 20 °C. However, the kinetics were poor, releasing only 7.5 bar of hydrogen at 300 °C after 3.5 h when the equilibrium pressure should be >1000 bar. The formation of Mg2Si was identified by XRD, but the products could not be rehydrogenated under 100 bar as this needed temperatures < 150 °C and the reaction was found to be kinetically limited. Partial rehydrogenation was achieved by ball milling Mg2Si under 50 bar of hydrogen reforming MgH2 and Si. Germanium, Ge, as a destabilisation phase for MgH2 has not been explicitly investigated, although there is a paper on the catalytic effect of Ge ball milled under hydrogen with magnesium (Gennari et al., 2002). During reactive ball milling, it was found that the presence of germanium destabilised the

© 2008, Woodhead Publishing Limited

496

Solid-state hydrogen storage

MgH2, forming Mg2Ge as well as a Ge–MgH2 composite. The latter showed enhanced kinetics, but there was no direct evidence of destabilisation. Unfortunately, because the paper was interested in the catalytic effect of Ge, only dehydrogenation results were reported and no hydrogenation or isotherm data. Thus, there was no information on reversibility nor thermodynamics for the destabilised system forming Mg2Ge. Silicon has also been shown to be a destabilising agent for lithium hydride forming a lithium silicide, Li2.35Si, as shown in the reaction equation below (Vajo et al., 2004). The authors were attempting to form Li4Si as this system would have a hydrogen capacity of 9 wt%, but the reaction proceeded only as far as forming Li2.35Si, giving a storage capacity of 5 wt%. The enthalpy of dehydrogenation was reduced by 70 kJ mol–1(H2) to 120 kJ mol–1(H2) (as deduced from a van’t Hoff plot), corresponding to a T(1 bar) value of 490 °C (cf.910 °C for LiH): 2.35LiH + Si s Li2.35Si + 1.18H2 The temperature of dehydrogenation for lithium hydride has also been shown to be reduced when ball milled with magnesium hydride (Yu et al., 2006). TGA-MS data showed that magnesium hydride decomposed first, but then LiH in the presence of Mg had an onset of dehydrogenation of c. 400 °C (cf.580 °C for ball milled LiH), forming an alloy with the magnesium: 2yLiH + 2xMg s 2MgxLiy + yH2 This is potentially a very versatile system as the α- and β-Mg-Li alloys cover the full range of Mg–Li compositions. It should be noted that the reversible systems so far reported have been for magnesium-rich combinations only, but the T(1 bar) values for this system are expected to still be too high (TGA results indicate T(1 bar) > 400 °C).

17.7

Future trends

Multicomponent hydrogen storage systems offer one of the most promising means of tuning the thermodynamics in order to develop a high hydrogen storage capacity material which operates at reasonably low temperatures. Most high-capacity materials have too high an enthalpy of dehydrogenation, but, as has been shown in this chapter, it is possible to reduce this by incorporating a destabilisation agent. A number of systems have been identified with T(1 bar) values below 200 °C. Unfortunately at such low temperatures these reactions have so far been kinetically limited. By their very nature, multicomponent systems have considerable mass transport issues, diffusing atoms between the two phases for the dehydrogenation reaction and in some cases the hydrogenation. However, there are examples where the diffusion between multiple phases can be relatively quick, for example the Na–Al–H

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

497

system where two phases are hydrogenated to produce sodium alanate (Chapter 14). Catalysts for multicomponent systems are at an early stage of investigation and hold the promise of greatly accelerating these reactions. To direct this research, greater understanding of rate-limiting steps for multicomponent reactions is needed so that an appropriate catalyst can be investigated. In situ studies of the chemical structure need to be supported by electron microscopy experiments in order to probe the changing morphology of the cycled materials, which will be key in better understanding the diffusion processes occurring. Containment has been suggested as a means of minimising phase segregation and sintering during cycling (Vajo et al., 2007). For example, having the multicomponent system loaded within a porous, inert support, thus physically limiting the degree of segregation/sintering possible, thus keeping the phases well mixed and in intimate contact. There are many more potential multicomponent systems to be investigated and the research direction should perhaps turn to investigate systems that are potentially gas-mediated in order to overcome the mass transport issue. One such example already exists from the amide chemistry, e.g. metal amides decomposing to form ammonia which in turn reacts with lithium hydride (Leng et al., 2004), to be discussed in the following chapter. However, although it is difficult to predict the best combination of phases for a multicomponent hydrogen storage system, it is certain that there will be many more exciting and unexpected developments in this field.

17.8

References

Alapati S V, Johnson J K and Sholl D S (2006), Identification of destabilized metal hydrides for hydrogen storage using first principle calculations, J. Phys. Chem. B, 110, 8769–8776. Alapati S V, Johnson J K and Sholl D S (2007a), Predicting reaction equilibria for destabilised metal hydride decomposition reactions for reversible hydrogen storage, J. Phys. Chem. C, 111, 1584–1591. Alapati S V, Johnson J K and Sholl D S (2007b), Using first principles calculations to identify destabilized metal hydride reactions for reversible hydrogen storage, Phys. Chem. Chem. Phys., 9, 1438–1452. Aoki M, Miwa K, Noritake T, Kitahara G, Nakamori Y, Orimo S and Towata S (2005), Destabilization of LiBH4 by mixing with LiNH2, Appl. Phys. A, 80, 1409–1412. Barkhordarian G, Klassen T, Dornheim M and Bormann R (2007), Unexpected kinetic effect of MgB2 in reactive hydride composites containing complex borohydrides, J. Alloys Compounds, 440, L18–L21. Bosenberg Y, Doppiu S, Mosegaard L, Barhordarian G, Eigen N, Borgschulte A, Jensen T R, Cerenius Y, Gutfleisch O, Klassen T, Dornheim M and Bormann R (2007), Hydrogen sorption properties of MgH2–LiBH4 composites, Acta Materialia, 55, 3951–3958. Chase M W, Ed. (1998), NIST-JANAF Thermochemical Tables, 4th ed., New York, American Institute of Physics.

© 2008, Woodhead Publishing Limited

498

Solid-state hydrogen storage

Chater P A, David W I F, Johnson S R, Edwards P P and Anderson P A (2006), Synthesis and crystal structure of Li4BH4(NH2)3, Chem. Comm., 2439–2441. Chen P, Xiong Z T, Luo J Z, Lin J Y and Tan K L (2002), Interaction of hydrogen with metal nitrides and imides, Nature, 420, 302–304. Cho Y W, Shim J H and Lee B J (2006), Thermal destabilization of binary and complex metal hydrides by chemical reaction: a thermodynamic analysis, Computer Coupling Phase Diagrams Thermochem., 30, 65–69. Croston D, 2007, The effect of metal oxide additives on the hydrogen sorption behaviour of magnesium hydride, PhD Thesis, University of Nottingham, Nottingham, UK. Dornheim M, Doppiu S, Barkhordarian G, Boesenberg U, Klassen T, Gutfleisch O and Bormann R (2007), Hydrogen storage in magnesium-based hydrides and hydride composites, Scripta Materialia, 56, 841–846. Filinchuk Y E, Vyon K, Meisner G P, Pinkerton F E and Balogh MP (2006), On the composition and crystal structure of the new quaternary hydride phase Li4BN3H10, Inorg. Chem. 45, 1433–1435. Gennari F C, Castro F J, Urretavizcaya G and Meyer G (2002), Catalytic effect of Ge on hydrogen desorption from MgH2, J. Alloys Compounds, 334, 277–284. Gutfleisch O, Schlorke-de Boer N, Ismail N, Herrich M, Walton A, Speight J, Harris I R, Pratt A S and Züttel A (2003), Hydrogenation properties of nanocrystalline Mg- and Mg2Ni-based compounds modified with platinum group metals (PGMs), J. Alloys Compounds, 356, 598–602. Hanada N, Ichikawa T and Fujii H (2005), Catalytic effect of nanoparticle 3d-transition metals on hydrogen storage properties in magnesium hydride MgH2 prepared by mechanical milling, J. Phys. Chem. B, 109, 7188–7194. Huot J, Liang G, Boily S, van Neste A and Schulz R (1999), Structural study and hydrogen sorption kinetics of ball-milled magnesium hydride, J. Alloys Compounds, 293, 495– 500. Johnson S R, Anderson P A, Edwards P P, Gameson I, Prendergast J W, Al-Mamouri M, Book D, Harris I R, Speight J D and Walton A (2005), Chemical activation of MgH2; a new route to superior hydrogen storage materials, Chem. Comm., 22, 2823–2825. Kojima Y, Matsumoto M, Kawai Y, Haga T, Ohba N, Miwa K, Towata S, Nakamori Y and Orimo S (2006), Hydrogen absorption and desorption by the Li–Al–N–H system, J. Phys. Chem. B, 110, 9632–9636. Leng H Y, Ichikawa T, Hino S, Hanada N, S Isobe N and Fujii H (2004), New metal–N– H system composed of Mg(NH2)2 and LiH for hydrogen storage, J. Phys. Chem. B, 108, 8763–8765. Lide D R, Ed. (2002), CRC Handbook of Chemistry and Physics, 83rd ed., London, CRC Press. Lu J and Fang Z Z (2005), Dehydrogenation of a combined LiAlH4/LiNH2 system, J. Phys. Chem. B, 109, 20830–20834. Mao J F, Wu Z, Chen T J, Weng B C, Xu N X, Huang T S, Guo Z P, Liu H K, Grant D M, Walker G S and Yu X B (2007), Improved hydrogen storage of LiBH4 catalyzed magnesium, J. Phys. Chem. C, 111, 12495–12498. Nakamori Y, Ninomiya A, Kitahara G, Aoki M, Noritake T, Miwa K, Kojima Y and Orimo S (2006), Dehydriding reactions of mixed complex hydrides, J. Power Sources, 155, 447–455. Oelerich W, Klassen T and Borman R (2001), Metal oxides as catalysts for improved hydrogen sorption in nanocrystalline Mg-based materials, J. Alloys Compounds, 315, 237–242.

© 2008, Woodhead Publishing Limited

Multicomponent hydrogen storage systems

499

Pinkerton F E, Meisner P, Meyer M, Balogh M P and Kundrat M D (2005), Hydrogen desorption exceeding ten weight percent from the new quaternary hydride Li3BN2H8, J. Phys. Chem. B, 109, 6–8. Pinkerton F E, Meyer M S, Meisner G P, Balogh M P and Vajo J J (2007), Phase boundaries and reversibility of LiBH4/MgH2 hydrogen storage material, J. Phys. Chem. C, 111, 12881–12885. Price T E C, Grant D M, Telepeni I, Yu X B and Walker G S (2008), The decomposition pathways for LiBD4–MgD2 multicomponent systems investigated by in-situ neutron diffraction, J. Alloys. Compounds, doi: 10.1016/j.jallcom.2008.05.030. Reilly J J and Wiswall R H (1968), The reaction of hydrogen with alloys of magnesium and nickel and the formation of Mg2NiH4, Inorg. Chem., 7, 2254–2256. Vajo J J and Olson G L (2007), Hydrogen storage in destabilized chemical systems, Scripta Materialia, 56, 829–834. Vajo J J, Mertens F, Ahn C C, Bowman R C and Fultz B (2004), Altering hydrogen storage properties by hydride destabilization through alloy formation: LiH and MgH2 destabilized with Si, J. Phys. Chem. B, 108, 13977–13983. Vajo J J, Skeith S L and Mertens F (2005), Reversible storage of hydrogen in destabilized LiBH4, J. Phys. Chem. B, Lett., 109, 3719–3722. Wagemans R W P, van Lenthe J H, de Jongh P E, van Dillen A J and de Jong K P (2005), Hydrogen storage in magnesium clusters: quantum chemical study, J. Amer. Chem. Soc., 127, 16675–16680. Xiong Z, Wu G, Hu J and Chen P (2006), Investigation on chemical reaction between LiAlH4 and LiNH2, J. Power Sources, 159, 167–170. Xiong Z, Wu G, Hu J, Liu Y, Chen P, Luo W and Wang J (2007), Reversible hydrogen storage by a Li–Al–N–H complex, Adv. Func. Mater., 17, 1137–1142. Yu X B, Grant D M and Walker G S (2006), A new dehydrogenation mechanism for reversible multicomponent borohydride systems – the role of Li-Mg alloys, Chem. Comm., 37, 3906–3908. Yu X B, Wu Z, Chen Q R, Li Z L, Weng B C and Huang T S (2007), Improved hydrogen storage properties of LiBH4 destabilized by carbon, Appl. Phys. Lett., 90, 034106. Yu X B, Grant D M and Walker G S (2008), Low temperature dehydrogenation of LiBH4 through destabilization with TiO2, J. Phys. Chem. C, doi: 10.1021/jp800602d. Zaluska A, Zaluska L and Stom-Olsen J O (2001), Structure, catalysis and atomic reactions on the nano-scale: a systematic approach to metal hydrides for hydrogen storage, Appl. Phys. A, 72, 157–165. Züttel A, Wenger P, Rentsch S, Sudan P, Mauron P and Emmenegger C (2003), LiBH4 a new hydrogen storage material, J. Power Sources, 118, 1–7.

© 2008, Woodhead Publishing Limited

18 Organic liquid carriers for hydrogen storage M. I C H I K A WA, Hokkaido University, Japan

18.1

Introduction

With regard to society’s use of hydrogen energy and fuel cells (FC) in the near future, both industry and government are greatly concerned about the infrastructure technology for the storage of hydrogen and its supply to the various business users of fuel cell systems and hydrogen. NEDO recently forecast a demand for hydrogen of over 20 billion cubic meters in Japan up until 2020 for 5 million units of FC vehicles (15% share of the petrol engine car market) and 10 million kW h home power (15% share of the domestic electricity supply) [1]. This huge quantity of hydrogen is potentially available even now from off-site hydrogen factories, e.g. COG (coke–oven–gas), salt/ water electrolysis and oil refineries scattered across Japan. Taking this into account, we propose the ‘hydrogen highway’, a novel infrastructure technology for hydrogen storage and transportation using organic hydrides as organic liquid carriers [2, 3]. Compared with the conventional technology available – high-pressure cylinders, liquid hydrogen and alloy metals – ‘organic hydrides’ such as oil and kerosene have many advantages based on their higher weight and volume hydrogen capacity, safety handling and commercial availability at a lower cost. Firstly, we discuss their unique chemistry and the catalytic reactions of organic hydrides which make them useful for creating a hydrogen delivery infrastructure for fuel cell systems. We have developed new spray-pulsed reactors using organic hydrides for the rapid supply of hydrogen. Furthermore, we may be able to use the conventional infrastructures used for oil to transport organic hydrides by road, rail and sea and supply them using petrol stations and delivery networks without any additional investment in a new infrastructure of pipelines and specific hydrogen storage facilities. The advantages of organic hydrides will have a greater impact once a hydrogen energy-based society becomes a realistic option, and will be the driving force behind rapid fuel cell deployment. In addition, sustainable energy producers such as hydro-, wind and solar power produce a lot of hydrogen by water electrolysis which 500 © 2008, Woodhead Publishing Limited

Organic liquid carriers for hydrogen storage

501

can be efficiently stored and transported by using organic hydrides as liquid carriers. We have also developed a new type of rechargeable direct fuel cell using organic hydrides. Finally, we discuss the hydrogen transportation network using organic hydrides, the ‘hydrogen highway’, to connect off-site hydrogen production plants, e.g. oil refineries and COG farms, to hydrogen delivery networks for FC automotives, domestic power supplies and various electrical appliances, in terms of technical barriers as well as life-cycle assessment (LCA) energy efficiency and economic evaluation.

18.2

Organic hydrides: chemistry and reactions for hydrogen storage and supply

Among the organic chemicals, as shown in Fig. 18.1 and Table 18.1, a series of aromatic hydrocarbons, such as benzene, toluene and naphthalene, as well as heteroaromatics, such as quinone and carbazole, chemically store a higher capacity of hydrogen by the hydrogenation reaction of each C=C (or C=O, C=N) bond with 1 mole of hydrogen molecules transforming to their corresponding hydrogenated products: cycloalkanes such as cyclohexane, methyl-cyclohexane and decalin, and cyclic heteroalkanes [7–10]. p-Quinone and acetone are used for hydrogen storage in their hydrogenation reaction to 1,4-dihydrocyclohexane and 2-propanol in 6 and 3 wt% hydrogen capacity, respectively. Recently, Pez et al. reported [11] that perhydro-N-ethyl carbazole works as a liquid carrier in c. 3 wt% weight hydrogen capacity in

H O

O

Very good: O H

O

O

O

N H

N

Good:

Moderate:

Very difficult (irreversible):

+ 4H2

6.1%

+ 2H2

5.6%

+ 3H2

7.0%

+ 4H2

N

3.8%

N Et

Et

18.1 Comparison of hydrogen storage capacity and reversibility of dehydrogenation and rehydrogenation of various organic compounds.

© 2008, Woodhead Publishing Limited

502

Organic hydrides

Aromatics

Weight capacity (wt% H)

Volume capacity l(H2)/l

Reactivitya (H2 l/min/g Pt)

Toxicity

Liquidityb (bp/mp °C)

Cyclohexane Methylcyclohexane Bicyclo hexyl Decalin 1-Methyl decalin

Benzene Toluene Biphenyl Naphthalene 1-Methyl naphthalene 2-Ethyl naphthalene

7.19 6.16 7.23 7.29

620 530 700 710

15.8 14.3 25.0 4.2

Carcinogen – – –

(81/–6.5) (101/–127) (200/–38) (190/–36)

5.4 1.7 0.1 0.2

6.62

650

3.5



(210/–60)

0.1

6.06

590

2.8



(215/–58)