Light Metals 2024 (The Minerals, Metals & Materials Series) 3031503074, 9783031503078

Citation preview

Edited by SAMUEL WAGSTAFF

The Minerals, Metals & Materials Series

Samuel Wagstaff Editor

Light Metals 2024

123

Editor Samuel Wagstaff Oculatus Consulting Marietta, GA, USA

ISSN 2367-1181 ISSN 2367-1696 (electronic) The Minerals, Metals & Materials Series ISBN 978-3-031-50307-8 ISBN 978-3-031-50308-5 (eBook) https://doi.org/10.1007/978-3-031-50308-5 © The Minerals, Metals & Materials Society 2024 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland Paper in this product is recyclable.

Preface

It is with great pleasure and honor that I present to you the Light Metals 2024 proceedings. This collection of manuscripts serves as a time capsule for later researchers to not only see how we approach technical problems but also what we value and esteem as an industry. The challenges the industry faces today around energy efficiency and sustainability represent difficult obstacles not unlike those faced by our forefathers previously. By contributing to these proceedings, authors have committed themselves to advancing the collective whole of the industry over any individual gain. The problems we face today are more than any one of us can handle individually. The key does not lie in a single technological breakthrough, derivation, or calculation. Instead, as we look at the exciting breakthroughs happening around us today, we can see that it is through cross-industry collaboration that the greatest strides are made. This is largely the magic of TMS. Each year we gather as an industry to discuss contributions we have all made to making things better than they were the year before. The Light Metals 2024 proceedings embody countless hours of work supplied by the volunteer Symposium Organizers, Session Chairs, Reviewers, and TMS staff who labor tirelessly to bring this tome into fruition each year. Of course, this collection of articles would be empty without the contribution of the individual authors, whose work brings us together this year. Special thanks and acknowledgment also go out to the symposium chairs of the five technical symposia assembled for 2024. I would like to offer my personal thanks to Michael Coley, Nabeel Aljallabi, Julien Lauzon-Gauthier, Anne Kvithyld, Tao Wang, Christopher Hutchinson, Sazol Das, and Timothy Langan. Without their invaluable help, none of this would have happened. The consistent guiding efforts of TMS staff have been integral as always this year. Patricia Warren and Trudi Dunlap were there every step of the way making sure that neither deadlines nor quality were allowed to slip. There are exciting times ahead. With unprecedented expansion projects, bold sustainability targets, and expanding markets, I cannot wait to see the new developments about to touch off. I invite you all to read these proceedings with a spirit of growth and collaboration in your heart. As you seek to create partnerships and discuss these proceedings with colleagues, I hope we will continue to see great leaps in science and innovation. Samuel Wagstaff

v

Contents

Part I

Alumina & Bauxite

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hong Peng, James Vaughan, Sicheng Wang, John Vogrin, and Dilini Seneviratne Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Michael Coley, Anthony Greenaway, Alicia Buckley, Khadeen Henry, Jheanell James, and Jason Brown Sugar-Derived Causticization Additives for the Bayer Process . . . . . . . . . . . . . . . Amit Desai, Jun Su An, and LoongYi Tan

3

13

24

Unveiling the Potential: A Paradigm Shift in Energy Reduction at Hindalco Renukoot Alumina Refinery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Paul Gupta and Nitya Nand Roy

29

An Innovative Approach to Smelter-Grade Alumina Calcination Using Renewable Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Wheatland, J. Fu, Y. Xia, M. E. Boot-Handford, and M. Sceats

38

Numerical Simulations for Performance Optimization of Circulating Fluidized Bed Calciner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bharathesh Kumar, Abhishek Seth, Chandrakala Kari, Vilas Tathavadkar, Ashish Mishra, and Prasanta Bose Improving the Operational Availability of Hydro Alunorte Calciners by Proper Refractories Maintenance Management . . . . . . . . . . . . . . . . . . . . . . . . Mariana A. L. Braulio, Thais A. Novais, Thiago Macedo, Veridiano Gomes, Jessika Silva, Thiago Iwanaga, and Victor C. Pandolfelli Bauxite Processing Via Sulfide Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Caspar Stinn, Lucas Marden, Ethan Benderly-Kremen, William Gilstrap, and Antoine Allanore

46

54

64

Study on a New Method of Clean Production of Alumina by Calcification Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ting-an Zhang, Guozhi Lyu, Yan Liu, and Yiyong Wang

74

Comparative Economic Efficiency of Processing High-Potassium Aluminosilicate Raw Materials into Alumina and Related Products . . . . . . . . . . . R. A. Seitenov, V. A. Lipin, S. N. Akhmedov, and V. V. Medvedev

82

vii

viii

Contents

Development of a Hydrometallurgical Process to Obtain High-Purity Alumina Using Bauxite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bárbara da Rocha Pereira, Morgana Rosset, Amilton Barbosa Botelho Junior, and Jorge Alberto Soares Tenório Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . He Xin, Lv Guo-zhi, Zhang Ting-an, Wang Song, and Wang Long

90

99

Sustainable Valorization of Bauxite Residue (“Red Mud”): Exploring the Potential of H2 Reduction for Multi-metal Recovery . . . . . . . . . . . . . . . . . . . . 110 Ganesh Pilla, Tobias Hertel, and Yiannis Pontikes Dealkalinization Effect of Carbon Dioxide in Flue Gas on Bayer Red Mud . . . . . 118 Chaojun Fang, Yihong Jia, Ruixue Lou, Yongping Wang, Xiaowei Deng, and Bo Lv Pilot Study on the Recovery of Iron from High-Iron Red Mud by Vortex Smelting Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 Xiaofei Li, Ting-an Zhang, Guozhi Lv, and Kun Wang Part II

Aluminum Alloys: Development and Manufacturing

New 6xxx Al–Mg–Si Alloy with High Electric Conductivity and Great Bendability for EV Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 Gregor Michael, Josef Berneder, and Roland Lorenz Effects of Alloying Elements Content on Microstructural Properties of AlMgSiCu Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 Osman Halil Çelik, Onuralp Yücel, Senem İsçioğlu, and Mustafa Demirkazık Influence of Feedrate on Microstructure and Hardness of Conventionally Spin-Formed 6061-O Plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Andrew Boddorff, Cecilia Mulvaney, and Wesley Tayon Influence of Copper Addition on the Thermal Stability and Corrosion Behavior of Aluminum 6082 Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 İlyas Artunç Sarı, Görkem Özçelik, İbrahim Bat, Alptug Tanses, and Zeynep Tutku Özen Corrosion Effect on Mechanical Properties of Stamped Al Alloy 6451 for Auto Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 M. Abdul, W. Shen, Lance Ying, and H. Hu Effect of Shot Peening and High-Temperature Shot Peening on the High Cycle Fatigue of 7010-T7452 Aluminum Alloy . . . . . . . . . . . . . . . . . . 171 Abouthaina Sadallah, Hong-Yan Miao, Benoit Changeux, Elie Bitar-Nehme, Apratim Chakraborty, Sylvain Turenne, and Etienne Martin X-Ray Computed Tomography of Fracture Paths in AA7075-T6 Sheet Torn at 200 °C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Danny Nikolai, Philip Noell, and Eric Taleff Utilizing Magnetic Field Annealing to Enhance the Microstructure and Mechanical Properties of 7075 Aluminum Alloy . . . . . . . . . . . . . . . . . . . . . . . 185 Damilola Alewi, Kirk Lemmen, Haluk Karaca, and Paul F. Rottmann

Contents

ix

Investigating the Potential of Secondary Aluminum Cast Alloys Used as Wrought Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Patrick Krall and Stefan Pogatscher Creative Approaches to Long-Term Recycling of Aluminium Scrap Forming AlSiMgMnCu Alloy with Excellent Mechanical and Microstructural Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Safaa El-Nahas, Ahmed S. Aadli, and Hassan M. Salman High-Throughput Compositional Study of 3xxx Al Alloy Using Laser Synthesis and Small-Scale Rolling: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . 206 Qingyu Pan, Monica Kapoor, John Carsley, and Xiaoyuan Lou Enhancing Recycling Efficiency and Critical Raw Material Substitution in 6xxx Alloys’ Production with Respect to Their Extrusion Feasibility and Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 Zeynep Tutku Özen, İlyas Artunç Sarı, İrem Yaren Siyah, Alptuğ Tanses, Görkem Özçelik, Ali Hakan Nurten, Baris Özdemir, and Melih Caylak Influence of Solidification Rate and Impurity Content on 5/7-Crossover Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 Sebastian Samberger, Lukas Stemper, Peter J. Uggowitzer, Ramona Tosone, and Stefan Pogatscher Influence of Increased Fe, Cu, and Zn Concentrations on Phase Formation in Aluminum A356 (AlSi7Mg0.3) Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234 T. Beyer, R. Kleinhans, M. Rosefort, S. Klan, A. Siemund, P. Decker, and W. Volk AMAG CrossAlloy®—Lightweighting the Future by Unconstraint Alloy Design: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 Lukas Stemper, Florian Schmid, and Ramona Tosone Correlation of Thermodynamic Calculations and Mechanical Properties of an Al-Si Cast Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248 A. Siemund, P. Decker, T. Beyer, and M. Rosefort Design and Characterization of Hierarchically-Strengthened, Cast Al-Ce-Ni-Mn-Sc-Zr Alloys for High-Temperature Applications . . . . . . . . . . . . . . . 255 Clement N. Ekaputra, Jovid U. Rakhmonov, Ekin Senvardarli, David Weiss, Jon-Erik Mogonye, and David C. Dunand Investigation Effect of In-Situ Grain Refiners on the 1XXX and 3XXX Twin Roll Casted Aluminum Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 Tuğçe Sezen, Sergen Belit, Fatmanur Özel, Altay Başaran, Bilal Demir, and Sadık Kaan İpek Developing Banding Microstructures in Directional Solidification of Aluminum Metal Matrix Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 Jaime Perez Coronado, Shanmukha Kiran Aramanda, Jonathan Goettsch, Ashwin J. Shahani, and Alan Taub Investigation of Ripple Formation in Aluminum Flat Products Produced by Different Types of Twin Roll Casters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 Fatmanur Özel, Altay Başaran, Sergen Belit, Sadık Kaan İpek, and Bilal Demir Effect of Shot Peening on Fatigue Properties of A20X Fabricated by Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278 Heidar Karimialavijeh, Apratim Chakraborty, Martin Proebstle, and Etienne Martin

x

Investigations on the Solid-State Additive Manufacturing of Al Alloy: Process, Microstructure, and Crystallographic Texture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 Abhishek Pariyar, Evren Yasa, Adrian Sharman, and Dikai Guan Evaluating Three-Point Bending Behavior of Aluminum Extruded Thin Walled Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 Melih Çaylak, Görkem Özçelik, Berat Bayramoğlu, and Tolgahan Çalı Meshfree Process Modeling and Experimental Validation of Friction Riveting of Aluminum 5052 to Aluminum 6061 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298 Lei Li, Mayur Pole, Hrishikesh Das, Sridhar Niverty, Md Reza-E-Rabby, Jorge F. Dos Santos, and Ayoub Soulami Influence of Welding Tool Material and Type of Joint on the Formability of Friction Stir Welded Tailored Blanks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306 M. Bachmann, R. Göbel, K. R. Riedmüller, and M. Liewald Parameters Controlling Drilling and Tapping Characteristics of Aluminum Based Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 H. Barakat, Y. Zedan, A. M. Samuel, V. Songmene, and F. H. Samuel Determining the Corrosion Speed of Welded AA 5005 Alloy with AA5356 Filler Metals According to Weld Rate Using the MIG Welding Technique . . . . . . 326 Hüseyin Müştak, Yusuf Özçetin, and Günhan Bayrak Investigating the Corrosion Performance of EN-AW-8006 Alloy with Mn and Cu Additions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 Ece Harputlu, Cemil Işıksaçan, Mert Günyüz, and Erdem Atar Annealing Behavior of Cold Rolling Sheets of a Continuous Cast Al-1.5Cu Alloy with Potential Application to Low-Cost Auto Forming Parts or Sheets . . . . 341 Xiyu Wen, Yan Jin, and Wei Li Effect of Annealing Process on Recrystallization Structure, Texture, and Precipitates of 1235D Aluminum Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 Wei Tang, Junpeng Pan, Zhisheng Guo, Hongpo Wang, and Zizong Zhu Comparison of Heating Systems for Aluminum Forging . . . . . . . . . . . . . . . . . . . . 356 Nurcan Akduran, Ahmet Asım Eser, Ahmet Umit Cakal, and Mustafa Acarer Effect of Cold Rolling Prior to Homogenization Heat Treatment on the Microstructural Evolution and Mechanical Properties of Twin-Roll Cast 8026 Aluminum Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362 Ahmet Kabil, Hatice Mollaoğlu Altuner, and Onur Meydanoglu Influence of T6 and T7 Heat Treatments on the Mechanical Properties of Rheocast Secondary AlSi7Cu3Mg Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369 Giulio Timelli, Stefano Capuzzi, and Giulia Scampone Heat Treatment of A20X Alloy Manufactured Using Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376 Heidar Karimialavijeh, Apratim Chakraborty, Martin Proebstle, Kentaro Oishi, Jean-Philippe Harvey, and Etienne Martin Thermomechanical and Metallographic Comparison of Twin-Roll CAST 1235, 3003, 8006, and 8079 Alloy Series Used in the Production of Foil Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 Sergen Belit, Tuğçe Sezen, Altay Başaran, Fatmanur Özel, and Sadık Kaan İpek

Contents

Contents

xi

The Effect of Cold Rolling Strain Degree in Corrosion Resistance of Fully Soft Temper Automotive 5182 Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 Dionysios Spathis, John Tsiros, Andreas Mavroudis, and Athanasios Vazdirvanidis Effect of Silicon Content on Solidification Parameters and Microhardness of Al–Si Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 Edgar R. Ibañez, Carlos D. Rodríguez, Paula R. Alonso, and Alicia E. Ares Part III

Aluminum Reduction Technology

Aluminium Carbide and Carbon Dust in Aluminium Electrolysis Cells—A Conceptual Model for Loss in Current Efficiency . . . . . . . . . . . . . . . . . . 411 Asbjørn Solheim A Method of Cell Heat Balance Control to Enable Variable Power Usage by Aluminium Smelters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419 Nick B. Depree, Yashuang Gao, Mark P. Taylor, and John J. J. Chen Computational Simulation of Electromagnetic Fields in an Aluminum Electrolysis Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Ryan M. Soncini A Method for Anode Effect Prediction in Aluminum Electrolysis Cells Based on Multi-scale Time Series Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436 Kejia Qiang, Jie Li, Jinghong Zhang, Jiaqi Li, Ling Ran, and Hongliang Zhang Predicting Electrolyte and Liquidus Temperatures of Aluminium Smelting Cells for Power Modulation Using Dynamic Model . . . . . . . . . . . . . . . . . . . . . . . . 445 Choon-Jie Wong, Jie Bao, Maria Skyllas-Kazacos, Barry Welch, Jing Shi, Nadia Ahli, Amal Aljasmi, Mohamed Mahmoud, and Mustafa Mustafa Construction and Application of Digital Twin in Aluminum Electrolysis . . . . . . . 453 Jiaqi Li, Kejia Qiang, Chunhua Yang, Xiaofang Chen, Jie Li, and Hongliang Zhang Estimation of the Spatial Alumina Concentration of an Aluminium Smelting Cell Using a Huber Function-Based Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . 464 Luning Ma, Choon-Jie Wong, Jie Bao, Maria Skyllas-Kazacos, Jing Shi, Nadia Ahli, Amal Aljasmi, and Mohamed Mahmoud Limits for the Current Efficiency in Hall-Héroult Cells . . . . . . . . . . . . . . . . . . . . . 474 Asbjørn Solheim Numerical Modeling of Anode Changes and Their Effect on Current Distribution and Magnetohydrodynamic Behavior of an Aluminium Reduction Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481 Jinsong Hua, Pascal Beckstein, Eirik Manger, Steinar Kolås, Øyvind Jensen, and Sigvald Marholm Specific Energy Reduction Through Design Modifications at Aditya Aluminium Smelter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 Venkannababu Thalagani, Rajeev Kumar Yadav, Shanmukh Rajgire, Amit Jha, Amit Gupta, Sai Mahati Bottla, Sanjay Pal, Sarthak Mohapatra, Anshu Mangal, Deepak Dash, Anish Das, Madhusmita Sahoo, Kamal Kant Pandey, and Vilas Tathavadkar Thermo-Electrical Analysis of Lying-Bed Patterns During Preheating Phase . . . . 499 Rohini-Nandan Tripathy, Daniel Marceau, Simon-Olivier Tremblay, Duygu Kocaefe, Antoine Godefroy, and Sébastien Charest

xii

New 32-H Metal Tapping Cycle Implementation at ALBRAS . . . . . . . . . . . . . . . . 508 Camila R. Silva, Franciny Lobato, Benedito Z. Silva, Valfredo C. Filho, Michel V. Pena, João P. F. Souza, Marcio N. Souza, Pierre Reny, and Kurt J. Nilsson Re-Usage of Big Butts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514 Andresa O. Menezes, Camila R. Silva, Marcio N. Souza, Valfredo C. Filho, Michel V. Pena, João P. F. Souza, Paulo N. Júnior, Nayary P. Monteiro, and Marcus Brasiliense Amperage Increase Program and Enablers in EGA Al Taweelah DX Technology Potlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520 Vishal Ahmad, Ishaq Alkharusi, Shaikha Al Shehhi, and Almero Austin Eybers Cell Startup and Early Operational Improvements in Albras . . . . . . . . . . . . . . . . 526 Ana Renata M. Nunes, Michel V. Pena, Pierre Reny, George Cardoso, Márcio Souza, and Ana Carolina Guedes Metal Tapping Yoke and Platform Modification for Improved Locking and Unlocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532 Yousuf Al Bastaki, Hassan Al Raqabani, Sajjad Hussain, Adnan Jafar, Sandeep Naik, Shibu Joseph, Sohail Akram, Zamad Hassan, and Muhammad Saleem Restart of AP30 Cells at Boyne Smelters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537 Daniel Whitfield, Murray Ure, Rashmi Jena, Evan Andrews, and Shashidhar Ghatnatti Technology of Reducing Carbon Dust Amount in Aluminum Electrolysis Production with High Lithium Potassium Electrolyte System . . . . . . . . . . . . . . . . 545 Shengzhong Bao, Kaibin Chen, Guanghui Hou, Huaijiang Wang, Yingtao Luo, Xu Shi, Jing Li, Lifen Luo, Fangfang Zhang, and Changlin Li Application of SAMI Energy-Saving and Current-Intensifying Technology in a 330 kA Potline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553 Jinlong Hou, Yafeng Liu, Hongwu Hu, Wei Liu, Xuan Wang, Xi Cao, and Michael Ren Pot Failure Prediction in EGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562 Shaikha Al Shehhi, Satheesh Mani, Jose Blasques, and Yusuf Ahli SMARTCrane, a Fives’ Digital Solution for Aluminium Production Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568 Vincent Delcourt, Clement Pessemesse, Vianney Boyer, Jean-Paul Leroy, and Frederic Moreira Pereira A Review of Challenges and Solutions in Ledge Control and Measurement in Aluminium Electrolysis Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577 Bazoumana Sanogo, Lukas Dion, Sébastien Gaboury, László Kiss, Thomas Roger, Sébastien Guérard, and Jean-François Bilodeau Accurate Measurement of Anode Current in Aluminum Electrolysis: From Ideal to Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586 Yi Meng, Jun Tie, Chun Li, Rentao Zhao, Hongwei Jiang, Xingzu Peng, Hao Xiao, Dongwei Liu, and Jun Lei

Contents

Contents

xiii

Correlation Between Corrosion Rate and Electrochemical Parameters of Anode Process on the Metallic Electrode in Molten Oxyfluorides . . . . . . . . . . . 596 Andrey Yasinskiy, Thomas Jamieson, Kamaljeet Singh, Guðmundur Gunnarsson, Jon Magnússon, Dominic Feldhaus, Roman Düssel, Isabella Gallino, and Bernd Friedrich Direct Production of Al-Mn Alloys During the Electrodeposition of Aluminum in a Laboratory Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602 Omar Awayssa, Geir Martin Haarberg, Gudrun Saevarsdottir, and Rauan Meirbekova Electrowinning of Al-Sc Master Alloys in the LiF-AlF3-Sc2O3 Melts . . . . . . . . . . . 608 Andrey Yasinskiy, Ilya Moiseenko, Dmitriy Varyukhin, Anastasia Saparova, Aleksandr Samoilo, Pavel Yuryev, Youjian Yang, Zhongning Shi, Zhaowen Wang, Peter Polyakov, and Bernd Friedrich Factors Affecting the Performance of Oxygen-Evolving Ni–Fe–Cu Anodes in Low-Temperature Molten Fluoride Electrolyte for Aluminium Electrowinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614 Gudrun Saevarsdottir, Geir Martin Haarberg, and Sai Krishna Padamata Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts: The Riddle of Dissolving Alumina Rafts Solved . . . . . . . . . . . . 621 Jonathan Alarie, László I. Kiss, Lukas Dion, Martin Truchon, Sébastien Guérard, and Jean-François Bilodeau Fundamental Loss of Current Efficiency During Aluminium Electrolysis and Its Correlation with Sodium Content Dissolved in the Aluminium . . . . . . . . . 630 Lukas Dion and Paul Desclaux Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639 David J. Jarvis, Rosanna E. van den Blik-Jarvis, Rosie F. L. Mellor, and Alf Bjørseth Online Monitoring of Metal Oxides in Molten Fluoride Electrolytes . . . . . . . . . . . 648 Wojciech Gebarowski, Samuel Senanu, Arne Petter Ratvik, Ole Kjos, Henrik Gudbrandsen, and Egil Skybakmoen Smelting 4.0: Digital Strategy for Aluminum Production . . . . . . . . . . . . . . . . . . . 655 Ved Prakash Rai and Datta Raju D Study of the Degradation of Ordinary Refractory Bricks in an Aluminum Reduction Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664 Mohamed Hassen Ben Salem, Gervais Soucy, Daniel Marceau, Antoine Godefroy, and Sébastien Charest Cradle-to-Gate Carbon Footprint Assessment of Graphite Cathode for Aluminium Electrolysis Pots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671 T. Carrère, B. Allard, and T. Reek Influence of Low Temperature on the Surface and Morphological Properties of Hydrated Lime in SO2 Desulfurization Reaction . . . . . . . . . . . . . . . 680 Karthikeyan Rajan, Duygu Kocaefe, Yasar Kocaefe, Julie Bureau, Jonathan Bernier, Yoann Robert, and Yves Dargis

xiv

Contents

Preliminary Testing and Simulations of Pot Integrated Abart (PIA) at Alcoa Mosjøen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688 Asbjørn Solheim, Anders Sørhuus, Ole S. Kjos, Håvard Olsen, and Helene M. E. Granlund Sustainability of Different Aluminium Production Technologies . . . . . . . . . . . . . . 696 Samuel Senanu, Mona Hassel, Asbjørn Solheim, and Egil Skybakmoen Thermal Analysis of Operational Events Affecting Electrolysis Cells and Their Local Alumina Dissolution Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 703 Ali Kodfard, Lukas Dion, Thomas Roger, Sébastien Guérard, and Jean-François Bilodeau Industrial Experimental Study on High Efficiency Recovery Technology of Side Waste Heat in Aluminum Electrolytic Cells . . . . . . . . . . . . . . . . . . . . . . . . 714 Yanan Zhang, Yang Zhang, Qiang Yu, Guisheng Liang, Guanghui Hou, Yuechao Guan, and Junqing Wang Regeneration of Aluminum Fluoride From Pure Bath . . . . . . . . . . . . . . . . . . . . . . 723 Brian Zukas and Xiangwen Wang Low Carbon Emission Technology Upgrading Industrial Pilot of 350kA Pots . . . . 730 Tiejun Wang, Yafeng Liu, Guijun Ge, Shimin Qu, Mingzhu Zhou, Hailong Liu, Wei Zhu, Yuanbing Zhu, Hongwu Hu, Xi Cao, and Michael Ren MHD Stability of Aluminium Cells—Cathode Design Effects . . . . . . . . . . . . . . . . 746 Valdis Bojarevics and Marc Dupuis Modelling and Design of the Cathode Block Assembly Using Different Types of Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754 Marc Dupuis, Xianan Liao, and Nyah Ren Cathode Inspection and Repair Procedure Improvements in ALBRAS . . . . . . . . . 766 Ana Carolina A. Guedes, Marvin Bugge, Michel V. Pena, Marcio N. Souza, Ana Renata N. Monteiro, and Adalberto Pastana Combining New and Old Cathode Block Assembly to Increase the Lifetime of Pot at INALUM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773 Kukuh Yudiarto, Rainaldy Harahap, Ade Buandra, Ari Purwanto, and Ferdy Rahadian Design and Trial of Electrical Collector Plate in Cathode Assemblies . . . . . . . . . . 779 Guorong Cao and Hao Zhang Determination of the Air-Gap Distribution at the Cast Iron to Carbon Cathode Interface Using a 3D Scanning Approach . . . . . . . . . . . . . . . . . . . . . . . . 786 Omolbanin Saeidi, Simon-Olivier Tremblay, Daniel Marceau, Antoine Godefroy, and Sébastien Charest Part IV

Electrode Technology for Aluminum Production

Alternative Binder for Carbon Anode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 797 Sheetal Gupta, Dibyendu Ghosh, Bibhuti Sahu, Amit Gupta, and Vilas Tathavadkar

Contents

xv

Effect of Mixing and Pressing Parameters on the Properties of Biopitch-Based Lab-Scale Carbon Anodes for Use in the Hall-Héroult Electrolytic Cell . . . . . . . . 804 Nooshin Baastani, Simon Laliberté-Riverin, Marie-Aimee Tuyizere-Flora, Guillaume Gauvin, Julien Lauzon-Gauthier, Thierry Ollevier, and Houshang Alamdari CFD Modelling of Air Injection Nozzles in Coke Calcination Kilns, Identification of the Best Compromise Between Carbonaceous Deposit Formation and Kiln Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 815 Marie-Josée Dion, Hans Darmstadt, Louis-Michel Malouin, Eric Duplain, Debbie Soriano, and Sunny Huang Estimation of the Coke Calcination Yield by Granulometry Analysis . . . . . . . . . . 822 Hans Darmstadt, Marie-Josée Dion, André Bouchard, and Luc Coté Comparing Handling Degradation of Shaft and Rotary Cokes . . . . . . . . . . . . . . . 828 Howard Childs, Austin Andrian, Barbara Chu, and Barry Sadler Influence of Selective Crushing and Particle Shape of Shaft and Hearth Calcined Anode Coke Components on Blend Bulk Density . . . . . . . . . . . . . . . . . . 834 Howard S. Childs, Barry Sadler, and Barbara Chu Effect of Boron on the Evolution of Petroleum Coke Active Pore Size Under Air Oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 841 Ramzi Ishak, Francois Chevarin, Gaétan Laroche, Donald Ziegler, and Houshang Alamdari Measurement of SO3 in Flue Gas from Anode Baking Furnace . . . . . . . . . . . . . . 850 Ole S. Kjos, Thomas Park Simonsen, and Thor Aarhaug Performance Improvement of the Anode Baking Process in Horizontal Furnaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856 Emmily Fonseca, Marcus Brasiliense, Paulo Teixeira, Leonardo Campos, Fernando Von Schaffelw, Paulo Nogueira, Alexandre Aquino, and Douglas Almeida Study on the Protection Technology Against Aluminized Surface of Anode Stub for Aluminum Electrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865 Shengzhong Bao, Dongsheng Li, Guanghui Hou, Kaibin Chen, Huaijiang Wang, Jing Li, Xu Shi, Dan Liu, Junyi Ma, and Huiyao Wang Part V

Melt Processing, Casting and Recycling

Reverberatory Furnaces Decarbonization—The Case of Hydrogen Combustion: Proof of Concept and First Experimental Results on Borel Furnace . . . . . . . . . . . 873 Louis Piquard, Emilien Clément, Martin Adendorff, Esin Iplik, and Tomas Ekman Decarbonization of Aluminum Reverberatory Furnaces: The Case of Plasma Melting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 881 Juan E. Salazar, Louis Piquard, Simon Vecten, and Emilien Clement Influence of Water Vapor on the Oxidation Behavior of Molten Aluminum Magnesium Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 890 Stefan Tichy, Simon Doppermann, Philip Pucher, Bernd Prillhofer, Stefan Wibner, and Helmut Antrekowitsch Dissolution Rates of Various Manganese Alloying Elements in Aluminium . . . . . . 897 Anne Kvithyld, Sarina Bao, Martin Syvertsen, Arne Petter Ratvik, Kjerstin Ellingsen, Mehdi Maghsoudi, and Kristján Leósson

xvi

Results Achieved with the Application of Optifine High Efficiency Grain Refiner in the Production of AA5182 Can Lid Stock . . . . . . . . . . . . . . . . . . . . . . . 905 John Courtenay, Lei Shi, JunJun Xia, and Zhao Weitiao Efficient Molten Metal Transfer in the Cast House: Introducing a New Thermal Insulation Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 912 Alireza Hekmat, Alain Simard, Bo Jin, and Michael Mastor An Estimation of Scrap Melting Rates by an Inverted Chvorinov Method . . . . . . 922 S. R. Wagstaff, R. B. Wagstaff, and A. Anestis Recovery Considerations in the Pyrometallurgical Recycling of Used Beverage Cans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 929 Theofani Tzevelekou, Malamatenia Koklioti, Athanasia FIampouri, Nikolaos Chamakos, Ioannis Contopoulos, Alexandros Anestis, Grigorios Galeros, Epameinondas Xenos, and Andreas Mavroudis Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 941 Shahid Akhtar, Massoud Hassanabadi, and Ragnhild E. Aune LAlum—Standardization of Launder Systems for Aluminum Casting . . . . . . . . . 952 Michel J. Quintiano and José G. Hernandez Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 962 Are Bergin, Robert Fritzsch, Shahid Akhtar, Lars Arnberg, and Ragnhild E. Aune A PoDFA Benchmarking Study Between Manual and AI-supervised Machine Learning Methods to Evaluate Inclusions in Wrought and Foundry Aluminum Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 970 Pascal Gauthier, Vincent Bilodeau, and John Sosa Automated Metal Cleanliness Analyzer (AMCA): Improving Digital Image Analysis of PoDFA Micrographs by Combining Deterministic Image Segmentation and Unsupervised Machine Learning . . . . . . . . . . . . . . . . . . . . . . . 977 Hannes Zedel, Eystein Vada, Robert Fritzsch, Shahid Akhtar, and Ragnhild E. Aune Elemental Analysis and Classification of Molten Aluminum Alloys by LIBS . . . . . 984 A. Demir, D. K. Ürk, K. Akben, M. Doğan, E. Pehlivan, Ö. Yalçın, M. A. Kıştan, G. Gökçe, and A. Obalı Enhancing Quantification of Inclusions in PoDFA Micrographs Through Integration of Deterministic and Deep Learning Image Analysis Algorithms . . . . 991 Anish K. Nayak, Hannes Zedel, Shahid Akhtar, Robert Fritzsch, and Ragnhild E. Aune Formation Kinetics of TiB2 in Aluminum Melt Studied Using Laser-Induced Breakdown Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 996 Antonio Vazquez Prudencio, Mehdi Maghsoudi, Kristbjorg Anna Thorarinsdottir, and Kristjan Leosson On the Importance of Measurement and Process Uncertainty in Certifying the Quality of Aluminium-Based Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1002 Varuzan M. Kevorkijan, Sandi Žist, and Lucija Skledar Characterization of Aluminum Dross Compositions Using Rietveld XRD Technique, Standardless XRF Method and Carbon Analyzer . . . . . . . . . . . . . . . . 1009 Hussain Al Halwachi

Contents

Contents

xvii

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1015 Bin Zhang and Gary P. Grealy A Passive Approach to Butt Swell Management . . . . . . . . . . . . . . . . . . . . . . . . . . 1026 S. R. Wagstaff, R. B. Wagstaff, B. Opdendries, A. Anestis, S. Pinis, G. Pashos, E. Xenos, and A. Mavroudis Characterization of Cr-Bearing Intermetallics Causing Pinhole Formation in Twin Roll Cast 8079 Aluminum Alloy Thin Foils . . . . . . . . . . . . . . . . . . . . . . . 1033 Yusuf Özçetin, Ali Ulus, Onur Birbaşar, and Feyza Denizli Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll Casting with a Steel/Copper Roll Pair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039 Seval Aksoy Aydın, Ece Harputlu, Hikmet Kayaçetin, Cemil Işıksaçan, and Erdem Atar In Situ Experimental Study of the Nucleation and Growth of Fe-Al Based Intermetallics: An Insight for Designing Next-Generation Recycling Friendly Aluminium Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1048 G. Salloum-Abou-Jaoude, K.-H. Cheong, S. Akamatsu, Ph. Jarry, and S. Bottin-Rousseau Measurement of the Heat Transfer in the Primary Cooling Area of a Laboratory Direct Chill Casting Plant for Alloy Design . . . . . . . . . . . . . . . . . 1055 Andreas Weidinger, Sebastian Samberger, Florian Schmid, and Stefan Pogatscher Influence of Chemistry and Direct Chill (DC) Casting Parameters on the Formation of Altenpohl Zone in 5xxx Alloys . . . . . . . . . . . . . . . . . . . . . . . 1062 Akash Pakanati, Snorre Rist, Thomas Hartmut Ludwig, Eystein Vada, Shiva Talatori, and Jan-Erik Ødegård Mechanisms of Twin-Roll Caster Tips Degradation . . . . . . . . . . . . . . . . . . . . . . . . 1069 Guillaume Girard, François Veillette, and William Roy Revolutionizing Slab Casting: Unveiling the Power of AI and Computer Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1078 Loïc Fracheboud, Julien Valloton, and Frederik Rummens Study of Vertical Fold Formation on Al–Mg Alloys During Direct Chill Casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1085 Marianthi Bouzouni, Theofani Tzevelekou, Spyridon Pinis, Sofia Papadopoulou, and Andreas Mavroudis Liquid Alloy Atomistic Modelling Perspective to Al Alloy Design . . . . . . . . . . . . . 1098 Philippe Jarry, Alaa Fahs, and Noel Jakse Part VI

Scandium Extraction and Use in Aluminum Alloys

The Role of New Aluminium-Scandium Alloys for Emission Reduction in Various Sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1105 Thomas Dorin and Timothy Langan Sc-Containing Al–Si–Mg (6xxx) Alloys for Automotive Extrusions . . . . . . . . . . . . 1111 Avishan Shomali, Timothy Langan, Thomas Wood, and Paul Sanders

xviii

Investigation of the Mechanical Properties of Flat Rolled Products of Aluminium Alloy Al-Mg-Sc Under Various Deformation Processing Modes . . . . . . . . . . . . . . 1120 Alexander Alabin, Sergey Valchuk, Alexander Krokhin, and Dror Shaked Effect of Sc and Zr Microalloying on Grain Structure After Hot Deformation and Brazing in Al–Mn 3xxx Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1125 Alyaa Bakr, Paul Rometsch, and X.-Grant Chen Hot Deformation Behavior and Post-brazing Grain Structure of Dilute Al–(Sc–Zr) Alloys for Brazed Heat Exchangers . . . . . . . . . . . . . . . . . . . . . . . . . . . 1133 Alyaa Bakr, Paul Rometsch, and X.-Grant Chen Investigating the Influence of Iron Content on the Microstructure and Mechanical Properties of a High Strength Al-Alloy for Additive Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1141 Matteo Turani, Walter Jannic, Paulo Davi Borges Esteves, Enrico Tosoratti, Adriaan Spierings, and Markus Bambach How Can Europe Reduce Offshore Dependence of Its Supply Chain for Critical Metals like Scandium, Niobium, Strontium, Magnesium, and Titanium? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1148 Beate Orberger, Henk van der Laan, Carsten Dittrich, Robin Scharfenberg, Edward Peters, Georges Croisé, Pierre Feydi, Carolin Maier, Richard Schneider, Bernd Friedrich, Yashvi Baria, Konstantinos Sakkas, and Christos Georgopoulos Behavior of Yttrium and Other Impurities in the Production of Scandium Oxide from Bauxite Residue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1154 Alexander Suss, Alexander Kozyrev, Natalia Kuznetsova, Alexander Damaskin, Sergey Pishchalnikov, Andrey Panov, Sergey Ordon, and Oleg Milshin Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1167 Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1173

Contents

About the Editor

Samuel Wagstaff is currently a partner at Oculatus Consulting, specializing in aluminum processing and product development. He holds degrees in Mechanical Engineering from Cornell University (B.S.) and Materials Science from MIT (M.S., Sc.D.). Previously as the Lead Scientist at Novelis, he led new product and process development for the entire R&D ecosystem across three continents. Currently, Dr. Wagstaff focuses on increasing profitability and productivities of nonferrous products by process improvement and fundamental research. He has helped to design over 1 million tons of recycle capacity in the aluminum sector and is the author of over 35 patents. Dr. Wagstaff is currently serving as the principal investigator on a $1M ReMADE grant to develop technology to improve the recyclability of organic laden aluminum scrap and is involved in much of the industrial expansion of the aluminum sector in North America. In his private life Sam enjoys being with his wife, hunting, fishing, and scuba diving.

xix

Program Organizers

Alumina & Bauxite Michael Coley is a Lecturer and Section Head of the Applied Chemistry Division, Department of Chemistry, The University of the West Indies, Mona, Jamaica. He has a Ph.D. in Chemistry and teaches Industrial, Analytical and Environmental Chemistry. Dr. Coley’s research interests include Air Quality and Renewable Energy however he is especially involved in Bayer Process Technology research. His areas of focus include alumina precipitation, the management of process impurities, and the development of strategies for processing marginal bauxites. Over the last 20 years, he has worked on several major projects that involve study of Jamaican bauxites and has reported on his work at major conferences and in academic journals. Dr. Coley enjoys mentoring and is the senior Undergraduate Coordinator for his Department. He is married and has two daughters. Aluminum Alloys: Development and Manufacturing Christopher Hutchinson is a Professor in the Department of Materials Science and Engineering at Monash University in Melbourne, Australia, and co-Chair of the Woodside Energy FutureLab at Monash University. He received his Ph.D. from the University of Virginia in the USA, and after several post-doc years at Grenoble Institute of Technology in France, joined the faculty at Monash. His work covers both the relationship between processing and structure, and structure and properties of engineering alloys with a large fraction of his activities on steels, stainless steels, aluminum, and copper and brasses. The coupling of modelling and experiment, wherever possible, is a priority for his approach to problems. Approximately half of his research is funded in association with industry and half from fundamental agencies such as the Australian Research Council. He is an Editor of Acta Materialia and Scripta Materialia.

xxi

xxii

Program Organizers

Sazol Das is an R&D Leader at Novelis Global Research and Technology Center, Kennesaw, GA, USA. He has over 16 years of experience in Product & Process Development, Diffusion Kinetics, Thermodynamic Modeling, Electronic Materials Packaging, and Thin Film Coating. Dr. Das graduated from McGill University, Montreal, QC, Canada with a Doctoral Degree in Materials Engineering. He is a certified agile Scrum Master and Product Owner who enjoys mentoring, empowering, and inspiring young professionals in their career advancement.

Aluminum Reduction Technology Nabeel Al Jallabi has been a Sr. Manager in Process Control and Development for 30 years with Alba. He has worked in different disciplines including Reduction Lines, Process Control and Development, and Operation Support Services. Al Jallabi has played instrumental roles in many projects including Reduction Lines 1–3 conversion from side break to point feed system, Reduction Lines 3 expansion, Reduction Line 5 commissioning, KA increase across Reduction Lines, lean manufacturing along with the McKinsey team, SPL detoxification and recycling project, Reduction Line 6 commissioning, and Reduction Line 5 restart. Al Jallabi has participated in many conferences as an author and delegate including TMS, ICSOBA, and Australasian. Al Jallabi’s educational background includes a BSc in Mechanical Engineering, MCs in Aluminum Smelting Technology, and an MBA. Electrode Technology for Aluminum Production Julien Lauzon-Gauthier is a Sr. research engineer at Alcoa, within the Operational Excellence smelting technology development group. He is responsible for the development and deployment of new technology and process optimization linked to anode manufacturing for Alcoa’s carbon plants. He is a chemical engineer and holds an M.Sc. and Ph.D. from Université Laval in Canada. His postgraduate work focused on multivariate statistics and machine vision applied to the anode manufacturing process. Dr. Lauzon-Gauthier also held different process engineer roles at Alcoa smelters in Canada before joining the technology development team. He has published his work in scientific journals and at the TMS annual meeting and received the TMS Light Metals Division Young Leaders Professional Development Award in 2020 as well as the Electrode Technology best paper award in 2021.

Program Organizers

xxiii

Melt Processing, Casting and Recycling Anne Kvithyld is a senior research scientist at SINTEF, Norway, where she has been since 2007. Her research interests are centered around refining and recycling of metals. She earned her Ph.D. entitled “Thermal Decoating of Aluminium Scrap” in 2003. She has been a Visiting Post.Doc at the Colorado School of Mines, USA, and was the winner of the 2011 Vittorio de Nora Award for Environmental Improvements in Metallurgical Industries. She is the co-author and co-editor of the book Principles of Metal Refining and Recycling published in 2021 and an active member of The Minerals, Metals & Materials Society (TMS).

Tao Wang currently holds the position of Technical Marketing Manager-Atlantic at Rio Tinto focusing on aluminum product development and customer technical support. Before joining Rio Tinto, he served as a chief metallurgist in the steel industry. He has over 10 years of experience in product/ market development and has published more than 20 journal papers and 8 patents related to metal casting, rolling, and heat treatment. Tao holds a Ph.D. degree in metallurgical engineering from The University of Alabama, and a bachelor’s degree in materials science and engineering from Jiao Tong University in China. Scandium Extraction and Use in Aluminum Alloys Timothy Langan began his career at Martin Marietta developing advanced aluminum alloys for space and Marine applications. He received his Ph.D. in Materials Science from Johns Hopkins University. Dr. Langan is currently working with Sunrise Energy Metals to develop applications for scandium that will be produced from its Sunrise Project in New South Wales, Australia. In this role, Dr. Langan is working with industrial partners to guide, develop, and focus research efforts on aluminum-scandium alloys at universities including Deakin University, Michigan Technological University, Monash University, and Chongqing University.

xxiv

Program Organizers

Frank Palm is a Senior Expert (1990–present) in metallic materials and processes focusing on welding and additive manufacturing for Airbus Central Research & Technology (CR&T). He is a dedicated and passionate specialist in the aerospace technology field, including failure analyzes with a strong current focus on metallic additive manufacturing (laser powder bed fusion material and process fundamentals), elaborating the prospects of Sc in Al-alloy applications for more than 25 years. Over the past 30 years, he has initiated, written, contributed, and published nearly 100 presentations and scientific papers. His background includes mentoring, teaching, and accompanying 12 Airbus in-house Ph.D. theses and about 50 diplomas, masters, and bachelor’s theses. He caused more than 30 invention filings and Airbus-related patent applications and, together with Airbus, currently holds more than 15 patents in the field of additive manufacturing, metallic materials technology, and integral aircraft fuselage constructions (e.g., Scalmalloy®—AlMgSc alloy concept for Additive Manufacturing (brand and patent)). He is a member of the German Engineering Community (VDI), the American Society of Materials (ASM), and of the German Welding Society (DVS). In the frame of the publicly hosted German Research Community (IGF) he has served for more than 10 years voluntarily as an appointed project consultant and assessor. Thomas Dorin is a Senior Research Fellow at Deakin University and the leader of the aluminum research group. Dr. Dorin has a distinct specialization in the incorporation of scandium in aluminum alloys. After gaining his Ph.D. from the University of Grenoble in 2013, Dr. Dorin has been instrumental in the development of innovative aluminum alloys, with his work leading to multiple patents and influential research papers. Among his many accolades, he was awarded the esteemed ICAA16 Early Career Researcher award in 2018. Additionally, he was recognized with the Discovery Early Career Researcher Award by the Australian Research Council in 2019. Dr. Dorin was an organizer and co-chair of the TMS symposium titled “Scandium Extraction and Use in Aluminum Alloys” in both TMS2023 and TMS2024. Dr. Dorin has co-authored 47 journal papers, 21 conference papers, 3 book chapters, and holds 4 patents. Furthermore, his work has been cited over 1800 times, with an h-factor of 22.

Program Organizers

xxv

Paul Rometsch obtained his Ph.D. in Materials Engineering at the University of Queensland, Australia, and has published widely in the field of aluminum metallurgy. He has held various academic and industrial positions focused on R&D within the global aluminum industry, including appointments at Comalco Aluminium (Australia); the University of Southampton and the Luxfer Group (UK); Hydro Aluminium Precision Tubing (Europe and China); and Monash University (Australia). Since 2018, Paul is has been a Principal Research Scientist with Rio Tinto Aluminium, where he currently works on aluminum product metallurgy at the Arvida Research and Development Centre in Saguenay, Québec (Canada). Henk van der Laan is a senior consultant and owner of Van der Laan International Consultancy BV (V.I.C.). Henk was born in 1954 in Rotterdam, The Netherlands. Henk has worked 44 years in the aluminum industry of which 34 years at KBM Affilips BV as an international sales manager for specialty alloys and master alloys in the non-ferro industry. The last 15 years he has been active as specialist for critical metals like magnesium, titanium, and scandium. Henk studied Metallurgy in Utrecht and Business Economics in Rotterdam and Industrial Marketing in Arnhem, all in The Netherlands. As a metallurgist he specializes in the grain refining mechanism of aluminum titanium boron, which is the main product of KBM. Besides his passion for aerospace, astronomy, aluminum, and scandium, Henk is an active mountain biker, golfer, and a fan of all kinds of sports like soccer, cycling, and formula 1. Efthymios Balomenos studied Mining and Metallurgical engineering at National Technical University of Athens and received his Ph.D. degree in Thermodynamics in the same school in 2006. Since 2008 he has been working in the Laboratory of Metallurgy as a postdoc researcher focusing on sustainable process development, CO2 mitigation strategies, exergy analysis, and resource utilization efficiency. He is involved in the research management and coordination of several collaborative large scale research projects (ENEXAL, EURARE, SCALE, ENSUREAL, RemovAl, BIORECOVER, AlSiCaL, SISAL PILOT, HARARE, and ReActiv) most of which focus on Bauxite Residue valorization or alumina production. He has more than 80 research publications in journals and conference proceedings with more than 1,000 citations and an h-index of 19. Since 2015, he has cooperated with MYTILINEOS—Metals as a senior consultant on R&D projects. He was a recipient of the TMS Light Metals Subject Award – Alumina & Bauxite in 2017 and since 2022 he has been on ICSOBA’s board of directors.

xxvi

Program Organizers

Muhammad Akbar Rhamdhani is currently the Director of Fluid and Process Dynamics (FPD) Group; and Program Leader of Net Zero Carbon Materials and Processes (Manufacturing Future Research Platform) at Swinburne University of Technology, Australia. Dr. Rhamdhani is a Professor in Extractive Metallurgy and Metals Recycling and obtained his Ph.D. from McMaster University, Canada, in Materials Science and Engineering. He was a teaching-research academic at the Institute of Technology Bandung (ITB) and the University of Queensland, before joining Swinburne. Dr. Rhamdhani was a Visiting Professor at Katholieke Universiteit Leuven Belgium and a Visiting Scientist at CSIRO. He is also currently an Adjunct Professor at the Faculty of Engineering, Universitas Indonesia. Dr. Rhamdhani’s research expertise is on advanced metal/material refining and impurities removal (e.g., in steel, aluminium, magnesium, silicon, nickel, and minerals); development of new processes for metal production; thermodynamics and kinetics of high temperature metal and chemical processes; and physical chemistry of interface. Dr. Rhamdhani’s current research directions include: (1) recycling and recovery of metals from urban resources (e-waste, end-of-life alkaline and lithium ion batteries, solar panel, permanent magnet; (2) decarbonization and hydrogenation of metallurgical processes; (3) pyrometallurgical processes of rare earth elements (REE) minerals; (4) solar metallurgy (the use of concentrated solar thermal energy for minerals and metals processing); and (5) astro metallurgy (extra-terrestrial minerals and metals processing; e.g., metals extraction on Mars and Lunar).

Aluminum Committee 2023–2027

Executive Committee 2023–2024 Chairperson Stephen Broek, Kensington Technology Inc., Ontario, Canada Vice Chairperson Samuel Wagstaff, Oculatus Consulting, Georgia, United States Past Chairperson Dmitry Eskin, Brunel University, Middlesex, United Kingdom Secretary Kristian Etienne Einarsrud, Norwegian University of Science & Technology, Trondheim, Norway JOM Advisor Anne Kvithyld, SINTEF, Trondheim, Norway Light Metals Division Chair Eddie McRae Williams, Arconic, Pennsylvania, United States

Members-At-Large Through 2024 Stephan Broek, Kensington Technology Inc., Ontario, Canada Kristian Etienne Einarsrud, Norwegian University of Science & Technology, Trondheim, Norway Duygu Kocaefe, University of Quebec, Quebec, Canada Johannes Morscheiser, Novelis Koblenz GmbH, Koblenz, Germany Alan Tomsett, Rio Tinto Pacific Operations, Queensland, Australia

Members-At-Large Through 2025 Dmitry Eskin, Brunel University, Middlesex, United Kingdom Linus Perander, Yara International, Sandefjord, Norway Derek Santangelo, Hatch, Quebec, Canada Jason Tessier, Alcoa Corporation, Quebec, Canada

xxvii

xxviii

Aluminum Committee 2023–2027

Members-At-Large Through 2026 Kristian Etienne Einarsrud, Norwegian University of Science & Technology, Trondheim, Norway Mertol Gökelma, Izmir Institute of Technology, Urla, Turkey Martin Iraizoz, Aluar Aluminum, Puerto Madryn, Argentina Julien Lauzon-Gauthier, Alcoa Corporation, Quebec, Canada Pascal Lavoie, Alcoa Corporation, Quebec, Canada Olivier Martin, Rio Tinto, Saint Jean, France Ray Peterson, Real Alloy, Tennessee, United States Andre Phillion, McMaster University, Ontario, Canada Andre-Felipe Schneider, Hatch, Quebec, Canada Eddie McRae Williams, Arconic, Pennsylvania, United States

Members-At-Large Through 2027 Houshang Alamdari, Laval University, Quebec, Canada Evan Andrews, Boyne Smelters Ltd, Queensland, Australia Samuel Awe, Automotive Components Floby AB, Falköping, Sweden Sumit Bahl, Oak Ridge National Laboratory, Tennessee, United States Roy Cahill, Rio Tinto, Australia Mark Dorreen, CSIRO, Victoria, Australia Marc Dupuis, GeniSim Inc., Quebec, Canada Les Edwards, Rain Carbon Inc., Louisiana, United States John Griffin, ACT LLC, New Jersey, United States Halldor Gudmundsson, Nordural—Century Aluminum Company, Iceland Errol Jaeger, The Business Consultants FZ LLC, Florida, United States Keerti Kappagantula, Pacific Northwest National Laboratory, Washington, United States Anne Kvithyld, SINTEF, Trondheim, Norway Julie Levesque, Quebec Metallurgy Center, Quebec, Canada Xinghua Liu, Alcoa Corporation, Pennsylvania, United States Lorentz Petter Lossius, Hydro Aluminium AS, Norway Pierre Marcellin, Rio Tinto, France Aditya Nittala, Pacific Northwest National Laboratory, Washington, United States Eric Nyberg, Kaiser Aluminum, Washington, United States Jitendra Odhrani, Sunpowergen, United Arab Emirates Hong Peng, University of Queensland, Queensland, Australia Katherine Rader, Pacific Northwest National Laboratory, Washington, United States Arne Ratvik, SINTEF, Trondheim, Norway Herve Yves Roustan, Rio Tinto Aluminium Pechiney, France Barry Sadler, Net Carbon Consulting Pty Ltd, Victoria, Australia Gudrun Saevarsdottir, Reykjavik University, Reykjavik, Iceland Benjamin Schuessler, Pacific Northwest National Laboratory, Washington, United States Samuel Wagstaff, Oculatus Consulting, Georgia, United States Tao Wang, Rio Tinto, Illinois, United States David Sydney Wong, Atmolite Consulting Pty Ltd, Queensland, Australia

Part I Alumina & Bauxite

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication Hong Peng, James Vaughan, Sicheng Wang, John Vogrin, and Dilini Seneviratne

Abstract

With increasing amount of high silica bauxites used as Bayer refinery feed, the re-precipitation of dissolved silicates results in greater volumes of desilication product (commonly known as DSP) which corresponds to elevated caustic consumption and issues with bauxite residue neutralisation and storage. Furthermore, incomplete desilication of pregnant Bayer liquor also results in silicate reactor and piping scaling as well as the possibility of contamination of the alumina product. Optimization of silicate management in the Bayer process is therefore a high priority. Understanding the chemistry of silicate leaching and precipitation of silicate in Bayer process underpins potential process improvements. This literature review summarises the chemistry of DSPs, with a focus on chemical-thermodynamics and reaction kinetics. Keywords




Desilication products Solubility Bauxite




Bayer process




Sodalite




Introduction Since the Bayer process was patented in 1888 by Australian chemist Karl Bayer, this technology has been the major industrial process for the producing of alumina from bauxite ores [1, 2]. Even after 135 years, this process remains unchanged for four major key sections: digestion of aluminarich minerals (gibbsite, boehmite, and diaspore) into hot caustic solution, clarification of the insoluble phases (bauxite

residue/red mud), and precipitation of gibbsite and calcination of the gibbsite to alumina. Bauxite is also comprised of iron phases such as goethite and hematite, titanium oxides, clay minerals comprised of quartz (SiO2), and kaolinite (Al2O3⋅2SiO2⋅2H2O) as well as other impurities such as organics [1, 3, 4]. During bauxite digestion, in addition to extraction of aluminium from bauxite ores (Eqs. 1 and 2), there is also the desilication products (DSP) from clay minerals in solution as shown by Eqs. 3 and 4.

Alumina-Bearing Minerals Digestion AlðOHÞ3 ðsÞ þ NaOHðaÞ , NaAlðOHÞ4 ðaÞ gibbsite AlOOHðsÞ þ NaOHðaÞ þ H2 OðaÞ diaspore/Boehmite , NaAlðOHÞ4 ðaÞ

ð1Þ

ð2Þ

The reactive silica in the bauxite, mainly kaolinite, reacts with the sodium hydroxide to form the sodium metasilicate solution. Once sufficiently supersaturated, the silicates then re-precipitate as insoluble sodium aluminosilicate DSP. Based on the different types of X anions and solution temperature, DSP could be hydroxysodalite, Cl-sodalite and cancrinite [1, 5, 6]. While DSP is costly, it plays a beneficial role in the Bayer process in terms of purifying the liquor of impurities such as sulphate, carbonate, and chloride, and therefore a certain amount of DSP is helpful to ensure a high degree of recyclability of the Bayer liquor.

Reactive Silica Dissolution

H. Peng (&)
J. Vaughan
S. Wang
J. Vogrin
D. Seneviratne School of Chemical Engineering, The University of Queensland, Brisbane, QLD 4072, Australia e-mail: [email protected] © The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_1

Al2 O3
2SiO2
2H2 O þ 6NaOH kaolinite , 2Na2 SiO3 þ 2NaAlO2 þ 5H2 O

ð3Þ

3

4

H. Peng et al.

Dissolved Silica Precipitation 6Na2 SiO3 þ 6NaAlO2 þ Na2 X þ 12H2 O DT

,

3ðNa2 O
Al2 O3
2SiO2
nH2 OÞNa2 X þ 12NaOH Bayer  sodalite ð4Þ

X = OH−, Cl−, SO42− and CO32−. The principles of Bayer desilication have been described in many studies [7, 8]. Impurities in bauxite dissolve and accumulate in the Bayer circuit, contributing to an array of process issues thus making it difficult to operate and control [9] with silicate phases being a primary contributor. Silicates such as kaolinite, halloysite, chamosite, and quartz exhibit different solubilities and reaction kinetics [1, 10]. Quartz typically does not significantly dissolve during the extraction of alumina from bauxite, unless the grain structure is fine and the digestion temperature and residence time are high. However, if quartz does dissolve it is especially costly as it leads to additional loss of dissolved aluminium. Halloysite is reactive but tends to be a minor component if present at all. Kaolinite is the most commonly found reactive silica and dissolves readily in caustic soda [5, 11, 12]. In the absence of crystallization seed, most of the reactive silica forms soluble sodium silicate which then precipitates as DSP during the predesilication and digestion stages of the Bayer process [13]. This review summarises the factors that influence thermodynamics solubility and kinetic DSP crystallisation, and mineral phase transformation during Bayer processes. The objective is to bring together the relevant fundamental information about reaction equilibrium and kinetics to enable the optimisation or improvement of predesilication and digestion in terms of economic and environmental outcomes

Thermodynamic Research of Desilication Products Despite several empirical solubility correlations proposed for silicates [14, 15], there is a lack of reliable chemicalthermodynamic data relevant to the Bayer process. This data is essential for predictive solubility modelling of DSPs. Only a single report has been found on the measurement of thermodynamic data for anhydrous sodalite (Na6(AlSiO4)6*NaCl2) by Komada et al. [16]. Using the group contribution method for calculating properties such as enthalpies, free energies of formation, and heat capacities proposed by Mostafa et al. [17, 18], Park and Englezos [19] estimated the equilibrium constant for sodalite solubility as it was not available in the literature. The results showed that the equilibrium constant ln(K) was 88.7 with large

uncertainty ±10.4 with 95% confidence level. The major contribution in the uncertainties was due to the uncertainty in the Gibbs energy of formation of the DSP phase. Later, Moloy et al. [20] reported the formation and hydration enthalpies of the hydroxysodalite family by hydrothermally synthesising the material from a zeolite phase. More recently, Zeng and Li [21] determined the solubility of sodalite in NaOH-NaAl(OH)4 solutions at temperatures between 30 and 75 °C by dissolving synthetic sodalite into solution. By data regression, they reported sodalite enthalpies and Gibbs energies of formation at standard conditions. For cancrinite, Kurdakova et al. [22] estimated the thermodynamic properties of synthetic sodium carbonate cancrinite at 27 °C using the reported thermodynamic data for calcium cancrinite by Liu et al. [23] and Ogorodova et al. [24]. In this section, the solid (synthesised under hydrothermal conditions) is different from the phases that would be formed during the Bayer process. No reports have been identified that include the enthalpies and Gibbs energies of formation for sodalite and cancrinite at Bayer process conditions. Furthermore, there is no reliable report of the heat capacity– temperature relationship for the high-alumina, semicrystalline DSPs forming in the Bayer process.

Empirical Solubility Models The equilibrium concentration of sodium aluminosilicate in sodium aluminate solutions has been the subject of a number of studies. There is a considerable amount of evidence in literature to suggest that the desilication kinetics are directly proportional to the liquor SiO2 supersaturation ratio. Therefore, in order to optimise desilication it is important to accurately predict equilibrium SiO2 values. The supersaturation ratio is defined in Eq. 5 [25]: r¼

½SiO2 t  ½SiO2 eq ½SiO2 eq

ð5Þ

where [SiO2]t is the silicate concentration (expressed as SiO2 equivalent) at time t and [SiO2]eq is the equilibrium silicate concentration. Previous research has led to development of models predicting equilibrium silicate concentration. However, most of these models are empirical in nature and not based on any experimentally determined thermodynamic parameters, which strictly limits their application to a set range of factors such as temperature, solution ionic strength, caustic concentration, and liquor impurities [26]. Models used in industry (which are confidential and therefore not published) are also limited to their specific refinery due to the empirically determined nature [27]. When conditions fall outside their parameterisation, the model predictions become almost meaningless [28]. Fundamental understanding of

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication

silicate precipitation behaviour and accurate prediction of the products could enable novel, efficient methods of desilication to be realised. Table 1 outlines the correlations for SiO2 solubility in open literature and their range of applicability. As these are empirical models, they have expressed the equilibrium concentration of silicate as [SiO2]eq.

A later study by Park and Englezos modelled silicate solubility from a chemical-thermodynamics perspective at conditions applicable to Kraft pulp mills. In their model, a slightly different precipitation reaction was used to give the corresponding equilibrium constant in Eq. 7 [19]: K c ðT Þ ¼

12 a4H2 O ðm12 OH cOH Þ ð7Þ ðm8Na þ c8Na þ Þðm6AlðOHÞ c6AlðOHÞ Þðm6SiO2 c6SiO2 Þ 4

Thermodynamics Solubility Models Outside of empirical correlations, only a few studies have attempted to construct a silica solubility model from a thermodynamics perspective. The earliest thermodynamicbased model found in literature was created by Jamialahmadi and Müller-Steinhagen [10]. They used the equilibrium constant for the precipitation reaction in Eq. 6 to provide a fit to experimental data. K c ðT Þ ¼

5

½NaOH2 ½Na2 SiO3 ½NaAlO2 ½H2 O

ð6Þ

4

3

3

Although both models were reported to have good predictions, they consistently predicted higher values than the measured data, especially at higher temperatures. When considering solubility from thermodynamic first principles, the equilibrium product Kp is dependent exactly on the nature of the chemical species. Given newly understood speciation, the equilibrium constant Kc can be written as Eq. 8 which considers silicate species in terms of H2SiO42− molality—the predominant silicon ion in Bayer process conditions [36]:

Table 1 Correlations for silicate equilibrium solubility Author

Correlation

Range of applicability

Adamson et al. [29]

SiO2 = 2.6  10–5 [Na2O][Al2O3]

Not stated

Smirnov [30]

SiO2 = 2 + 6.6  10–4 [Al2O3]2 – 3.3  10–2 [Al2O3]

T = 298–368 K [NaOH] = 12–378 g/L [Al(OH)3] = 13– 431 g/L

Leiteizen [31]

SiO2 = 1.28  10–5 [Na2CO3][Al2O3] –4

Not stated –3

–6

Cresswell [32]

SiO2 = 0.1587 – 6.058  10 T(°C) – 2.078  10 [Na2CO3]2 + 9.509  10–6 [Na2CO3]T(°C)

[Na2CO3] + 9.156  10

Hewett et al. [33]

SiO2 = 1.44  10–5 [Al2O3]2 + 1.85  10–4 T(°C) + 2.97  10–4 [NaOH]

T = 328–460 K [NaOH] = 50–190 g/L [Al(OH)3] = 69– 275 g/L

Smith et al. [34]

Below 180 °C: SiO2 = 2.67  10–3 [Al2O3] + 1.18  10–4 [Na2CO3] Above 180 °C: SiO2 = 2.67  10–3 [Al2O3] + 1.18  10–4 [Na2CO3] + 2.5  10–3 (T − 100)

T = 373–523 K [NaOH] = 113–226 g/L [Al(OH)3] = 69– 321 g/L

Sizgek and Aguila [35]

SiO2 = 1.82  10–6 [Na2CO3]2 + 3.37  10–13 [Na2CO3]2T3(°C) + 2.08  10–5 [Na2CO3]2(A/ C)3

T = 373–523 K [NaOH] = 121–196 g/L [Al(OH)3] = 97– 302 g/L

Barnes et al. [14]

SiO2 = −2.26  10–3 [Al(OH)3]0.5 + 2.42  10–4 [NaOH] + 6.66  10–4 [Al(OH)3] [NaOH] + 1.47  10–5 T(°C) − 1.77  10–3 (All concentrations in mol/L)

T = 363–493 K [NaOH] = 155–217 g/L [Al(OH)3] = 108– 174 g/L

Where all the concentrations are in g/L of that compound unless otherwise stated

T = 373–513 K [NaOH] = 136 g/L [Al (OH)3] = unknown

6

K c ðT Þ ¼


H. Peng et al.

m6Na þ c6Na þ




12 12  a10 H2 O mOH cOH  
 m6H SiO2 c6H SiO2 m6AlðOHÞ c6AlðOHÞ mNa2 X cNa2 X 2

4

2

4

4

4

ð8Þ where a is the activity, m is the molality, and c is the activity coefficient of the species. Considerable attention must be paid to speciation in Bayer liquors, especially at high ionic strength and temperature. When considering Eq. 8 as the equilibrium constant equation for DSP, good agreement has been observed with reported solubility values at extremely dilute conditions and low temperature, however fails when increasing either of these conditions. Equally and perhaps more important to DSP modelling than the aluminium speciation is the speciation of silicon. Problems associated with Si speciation modelling can be traced back to its origins in geochemical modelling [37]. It is generally accepted that the predominant silicate species at high pH such as in the Bayer process is the monomeric H2SiO42− although other polymeric and aluminosilicate species may exist. According to the speciation diagram in Fig. 1, the H2SiO42− ion accounts for at least 90% of the total silicate species in solution and is a generalised form of silicate ion shown in the equations as it is the dominant species [38, 39]. A systematic study undertaken by Smirnov [30] into aluminate solutions that revealed three distinguishable regions in a characteristic ‘U’-shaped solubility curve as shown in Fig. 2: (I) at low Al(OH)3 concentrations where a decrease in the alumina concentration leads to a considerable increase in solubility of DSP (higher SiO2); (II) an intermediate zone where the solubility is practically independent of Al(OH)3 concentration; and (III) at high Al(OH)3 concentrations the equilibrium silica concentration increases with increasing alumina concentration. Note that at gibbsite precipitation conditions (3.3–3.8 M Al(OH)3), a sharp

decrease in silicate solubility is inevitable as aluminate concentration decreases [40]. As plant liquors typically do not fall below 100 g/L Al2O3, the empirical correlations do not parameterise a U-shaped curve and instead show only a positive relationship as shown in Fig. 2 Region III. Problems arise when considering both Al and Si species in highly alkaline solution as neither Fig. 1 nor 2 adequately describe their co-existence. Pokrovski et al. were among the first group to investigate Al-Si speciation in basic solutions [41]. It found that the species AlSiO(OH)6− and its formation through Eq. 9 accounted for at least 80% of the total dissolved aluminium in the presence of silicate through 27Al nuclear magnetic resonance measurements.  H4 SiO04ðaqÞ þ AlðOHÞ 4 ðaqÞ $ AlSiOðOHÞ6 ðaqÞ þ H2 OðlÞ

ð9Þ Gout et al. later conducted comprehensive Raman spectroscopic investigations on Al-Si complexation in ultrabasic solutions at 20 °C with higher OH− molality than previous studies [38]. Through measuring band intensities, they derived the amounts of complexed Al and Si in solution shown in Fig. 3. The dimer species AlSiO3(OH)43− was suggested to be the dominant complex formed through Eq. 10:  3 H2 SiO2 4 ðaqÞ þ AlðOHÞ4 ðaqÞ $ AlSiO3 ðOHÞ4 ðaqÞ þ H2 OðlÞ

ð10Þ Zeng and Li modelled silicate speciation using OLI’s mixed solvent electrolyte (MSE) [42] and the AlSiO3(OH)43− species proposed by Gout et al. [43]. Their model appeared to fit their low-temperature data well and suggested it to be the predominant anion at Bayer conditions. However, even sophisticated electrolyte solubility calculations such as that often do not always predict the unusually high Al and SiO2

Fig. 1 Distribution of solution silicate species in a 0.001 M Si solution at T = 25 °C a by Raman spectra [38]; b by geochemical modelling software using IUPAC data (solid line) and HSC data (dashed line) [39]

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication 6

I 5

Crystallization Mechanisms and Kinetics of DSPs

III

Boundary

SiO2 (g/L)

4

II

3 2 1 0 0

50

100

150

200

250

300

7

350

Al(OH)3 (g/L)

Fig. 2 The solubility of silicate (as g/L SiO2) in sodium aluminate solution as a function of aluminate concentration (as g/L Al(OH)3). Graph was taken and modified from [38]

The mechanisms of crystallization during Bayer process, including nucleation, growth, DSP metastable phase transformations, and agglomeration, are complicated. Understanding these mechanisms is crucial for increasing the particle size of DSP and enable adequate recycling of otherwise lost sodium and aluminium. As shown in Eq. 11, mass deposition is split into a diffusion step and a first-order reaction step. Here m = mass deposited in time t; A = crystal surface area; c = solute concentrations in the bulk solution; ci = at the interface and c* = equilibrium saturation; kd and kr = deposition/reaction mass transfer coefficients; and g = exponent g, the order of the overall crystal growth process [45, 46]. These reactions can be combined by approximating the overall driving force or concentration difference and introducing an overall crystal growth coefficient KG. dm dm ¼ k d Að c  c i Þ ¼ k r Að c i  c  Þ ! ¼ K G Aðc  c Þg dt dt ð11Þ

Fig. 3 Complexed Al and Si concentration as a function of the product of concentrations of the free silicate and aluminate [38]

concentrations of Bayer liquors and nuclear waste tank concentrates [44]. With the abundance of conflicting species, further studies into the system at elevated temperature are required for accurate modelling. The chemical-thermodynamic approach for estimating solubility can be a powerful predictive tool, provided the model is based on a complete and accurate data set. There is limited data available for certain aqueous species such as aluminate-silicate polyanions and the heat capacity–temperature relationships have not been established for some important components, including sodalite. Reliably relating the thermodynamic predictions to practical solution phase concentrations requires an estimate of the activity coefficient for the aqueous species which is also currently lacking for aqueous silicates. The systems are further complicated when considering the wide range of other solution components and ion substitutions in the DSP.

The nucleation and crystal growth of DSP can be determined from the desilication rate and the supersaturation order. Smith et al. proposed that the order of the reaction (n) depends on the supersaturation ratio, with n equal to 1 at lower ratios and 2 or 3 at higher ratios [34]. However, even when the initial silicate concentration is the same, the seed mass or surface area can also affect the order of the reaction. Ruan et al. suggest that desilication with a third-order or greater dependence on Si supersaturation is due to nucleation, either primary heterogeneous or secondary, while less than third-order dependence is predominantly due to aluminosilicate crystal growth [47, 48]. The nucleation and crystal growth rates of DSP during Bayer process vary with kaolinite dissolution in caustic solutions [11, 12]. As shown in Fig. 4, kaolinite from bauxite completely dissolves into caustic solutions within hours, which is coupled with an early increase of SiO2 concentration in the solutions. The SiO2 concentration starts to decrease after complete dissolution of kaolinite due to precipitation of DSP which consumes the SiO2 component in the solution [49]. Nucleation and growth rate for DSP have historically been studied in terms of changes in silicate supersaturation, but not in terms of fundamental parameters. Furthermore, evidence of growth rate dispersion and growth as a function of particle size has not been investigated for DSP. The reactivity of kaolinite on the precipitation step to form DSP has also been studied. Kotte states that experiments have shown that kaolinite from most bauxites will

8

H. Peng et al.

Fig. 4 Kaolin dissolution and DSP precipitation. Adapted from [49] with data from the model described in [11]

dissolve rapidly (less than 15 min) provided the temperature is kept close to the atmospheric boiling point [49]. However, this is dependent on the particle size and kaolinite form present. Smith et al. similarly found that DSP precipitation rate is limited by the dissolution of kaolin [50]. Kaolinite can also be heat treated to form meta-kaolin, which reacts rapidly forming DSP phases seen under similar conditions using a soluble silicate source [11, 12, 51, 52].

Desilication Products Phase Transformation The DSPs have much lower solubility than the kaolinite in Bayer liquor. Many researchers have investigated the structure of DSP [6, 13, 50, 53–56]. Depending on the solution chemistry and plant operating conditions, various types of DSP can be formed. The silicon-oxygen tetrahedron is a basic structural element of most DSPs. Based on the different types of anion incorporation and solution temperature, the structure of DSP could be either that of a sodalite (Na6(AlSiO4)6*2NaOH*2H2O) or cancrinite (Na6(AlSiO4)6*Na2CO3*2H2O) [1, 5, 6] with molar soda to a silicate ratio (Na2O/SiO2) of 2/3. These two types of DSPs, on the other hand, are known to form as a result of a series of solution-mediated silica-rich phase transformations (e.g., kaolinite) [45, 57–59]. This sequence was explained by Peng et al. in the context of thermodynamic driving forces and the observed crystallization pathway occurs via the Ostwald successive transformation step rule, which was proposed as amorphous sodium aluminosilicate ! zeolite Linde Type A (LTA) ! sodalite ! cancrinite [39, 60, 61]. It suggests that Bayer process heat exchanger scale is in fact possibly made up of all phases including amorphous material, zeolite A, sodalite, and cancrinite, but their proportions will be

dependent on temperature and conversion rates. This phase transformation sequence has received support from several research groups [1, 62, 63] and is regarded as the ‘correct’ phase transformation sequence in synthetic Bayer liquor. Radomirovic et al. [64] and Peng et al. [60] also observed DSP formation beginning with an amorphous phase under Bayer conditions. In most of the studies investigating phase transformation, a soluble silicate source such as sodium metasilicate pentahydrate or waterglass was used. Some studies have concluded that the starting silicate source influences the precipitating phase, implying that the true phase transformation pathway originating from kaolin is still unknown. It is suggested that DSP is initially heterogeneously nucleated on the kaolin and as kaolin further dissolves, these nuclei are released, allowing for secondary nucleation. Vogrin et al. recently concluded that the DSP formed at alumina digestion concentrations directly precipitates as a high-alumina cancrinite-type structure that can remain stable at equilibrium [65]. This was the proposed phase of DSP in bauxite residues compared to heat exchangers. Generally, in the Bayer liquor, truncated octahedral cages (toc units) are formed by crosslinking AlO4 and SiO4 tetrahedral blocks [66] and these are the fundamental building blocks of the different DSP phases after kaolinite dissolution at temperature less than 100 °C [11, 67, 68].

Anion Effects on the Formation of DSPs Anion impurities present in Bayer liquors are known to have a detrimental effect on the efficiency of the Bayer process. These anions originate from partial dissolution of sulphur- or carbonate-containing minerals, organic material, process additives, as well as the diffusion of atmospheric carbon

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication

dioxide [69]. The presence of different anions can impact the phase transformation rates [50, 52, 61, 70, 71] and trace amounts of impurities (*10–3 M) can cause significant effects on the crystal formation process. As Bayer liquor is recirculated in the process there is a tendency for impurity levels to build up, reducing liquor productivity by affecting alumina solubility, yield, liquor density, and viscosity [72]. As a result, they require purging to maintain reasonable steady-state concentration in the process. Although generally considered undesirable, the precipitation of sodium aluminosilicate shows that various anions can become enclathrated within the framework, including Cl−, CO32−, SO42−, etc. DSP is both paradoxically problematic and useful as the inclusion can actually remove significant amounts of impurities from Bayer liquors [73]. It has been reported that inclusions in DSP may account for up to 75% of Na2SO4 exiting the Bayer circuit, while chloride and carbonate are seemingly lower at about 25% [72]. As anions are able to replace the hydroxide ion, the potential to reduce caustic losses by charging sodium salts after digestion has been suggested and patented in the literature [74, 75]. Only a few studies have been conducted into anion incorporation in DSP that are relevant to Bayer process temperatures, caustic and sodium aluminate concentrations [7, 76, 77]. Fundamental studies by Seimiya et al. into desilication in the presence of impurities revealed that the percentage of anions enclathrated in DSP increases with higher synthesis temperatures and longer times [78–80]. Sodalite synthesised with added Na2CO3 contained up to 300% more CO32− than control tests without impurity addition. The results also suggested that there is some preference to incorporate not only CO32−, but also Cl− and SO42− into the crystal lattice over OH− and Al(OH)4− [80]. Comprehensive studies on anion impurities and their effects on DSP were conducted over two decades [13, 52, 54, 61, 71, 81]. The composition of sodalite formed after predesilication and digestion in liquors containing added sodium salts such as NaCl, Na2CO3, or Na2SO4 salts was investigated, which shows that the magnitude of anion incorporation into sodalite under pseudo-Bayer conditions follows the trend: OH− < Al(OH)4− < Cl− < CO32− < SO42 − . The incorporation of chloride (Cl−) and sulphate (SO42−) was also observed to follow a Langmuir-type isotherm with increasing anion concentration, which was later supported by Whittington et al., Smith et al., and Vogrin et al. [13, 81, 82]. These studies were beneficial as they modelled quantitative anion incorporation into DSP, although they did not relate to other fundamental factors such as reaction kinetics. In addition to being incorporated into the DSP structure, anions can promote or suppress the phase transformation

9

sequence. For example, Seimiya et al. discovered DSP in bauxite residue transformed from zeolite A to sodalite after desilication with the addition of NaCl, Na2CO3, and Na2SO4 [80]. Breuer et al. demonstrated that the addition of sodium salt such as sodium oxalate, carbonate, and sulphate promoted formation of cancrinite over sodalite at 200 °C [7]. However, other studies have found that specific salts might inhibit the transformation of zeolite A to other phases [62]. In summary, it is critical to understand the behaviour of each anion influencing secondary nucleation rates, crystal growth, and agglomeration processes under Bayer DSP precipitation conditions.

Conclusions and Outlook This review summarizes both chemical and physical properties of DSPs and reviewed key issues relating to current challenges faced by the alumina industries for silica contamination in the Bayer liquor, thermodynamics and solubility of DSPs, phase transformation, and crystallization enlargement of DSPs during Bayer process that may affect the refinery aluminium grade or subsequent separation of DSP with other minerals in bauxite residue. The concentration of NaOH and NaAl(OH)4 is a primary factor that affects the concentration of silicate in Bayer liquor. The solubility of silicate in aluminate-free solutions is higher than in aluminate-bearing solutions under Bayer liquor conditions, and the relationship between silicate concentration and caustic concentration is positively correlated while the equilibrium silicate concentration decreases with decreasing aluminate concentration. The temperature has minimal impact on solubility within a range of 30–75 °C but has been shown to intensify at Bayer digestion temperatures. The crystalline phase also affects silica levels in the Bayer liquor, with the early amorphous phase formed during the Bayer process contributing to higher silica concentration. For mineral phase transformation during the Bayer process, sodalite and cancrinite are common types of DSPs (digestion-soluble products) formed during the process from kaolinite. These DSPs are known to form because of solution-mediated silica-rich phase transformations, and the crystallization pathway occurs via the Ostwald successive transformation step rule. Anion impurities present in Bayer liquors can impact the phase transformation rates and trace amounts of impurities can cause significant effects on the crystal formation process. Future studies need to be done to clarify the chemical composition and mineral structures of DSPs formed during Bayer process in the presence of anion impurities.

10 Acknowledgements The authors gratefully acknowledge the financial support from Rio Tinto and the Alumina Workshop Scholarship for this project. We acknowledge the facilities and the scientific and technical assistance of the Australian Microscopy and Microanalysis Research Facility at the Centre for Microscopy and Microanalysis, The University of Queensland.

References 1. Smith, P., The processing of high silica bauxites—review of existing and potential processes. Hydrometallurgy, 2009. 98(1): p. 162–176. 2. Misra, C. and C. Misra, Industrial alumina chemicals. Vol. 17. 1986: American Chemical Society Washington, DC. 3. Peng, H. and J. Vaughan. In-Situ XRD Investigation of Bauxite Dehydroxylation. in TMS Annual Meeting & Exhibition. 2018. Springer. 4. Faulstich, F.R.L., et al., Raman spectroscopic analysis of real samples: Brazilian bauxite mineralogy. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2011. 80(1): p. 102–105. 5. Laws, M.P., Reactions of Kaolinite in the Predesilication Processing of Bauxite Ores. 2005: University of Melbourne, School of Chemistry. 6. Lowe, J.L. and C.U.o.T.D.o.A. Chemistry, DSP in the Bayer Process: A Fundamental Study of Its Precipitation and Role in Impurity Removal. 2007: Curtin University of Technology. 7. Breuer, R., L. Barsotti, and A. Kelly, Behavior of silica in sodium aluminate solutions. Extractive Metallurgy of Aluminum, 1963. 1: p. 133–157. 8. Oku, T. and K. Yamada, The dissolution rate of quartz and the rate of desilication in the Bayer liquor. Essential Readings in Light Metals: Alumina and Bauxite, Volume 1, 1971: p. 247–254. 9. Gilkes, R., et al. Caustic insoluble aluminium containing nanominerals in bauxite from South Western Australia. in Proceedings of the 9th International Alumina Quality Workshop. 2012. 10. Jamialahmadi, M. and H. Müller-Steinhagen, Thermodynamic relationships for the solubility of silica in Bayer process liquor. Aluminium (Forschung), 1992. 68(3): p. 230–234. 11. Peng, H., et al., The effect of leaching temperature on kaolinite and meta-kaolin dissolution and zeolite re-precipitation. Minerals Engineering, 2021. 170: p. 107071. 12. Peng, H., J. Vaughan, and J. Vogrin, The effect of thermal activation of kaolinite on its dissolution and re-precipitation as zeolites in alkaline aluminate solution. Applied Clay Science, 2018. 157: p. 189–197. 13. Whittington, B., B. Fletcher, and C. Talbot, The effect of reaction conditions on the composition of desilication product (DSP) formed under simulated Bayer conditions. Hydrometallurgy, 1998. 49(1): p. 1–22. 14. Barnes, M.C., J. Addai-Mensah, and A.R. Gerson, The solubility of sodalite and cancrinite in synthetic spent Bayer liquor. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 1999. 157(1): p. 101–116. 15. Müller-Steinhagen, H., Determining silica solubility in Bayer process liquor. JOM, 1998. 50(11): p. 44–49. 16. Komada, N., et al., Thermodynamic properties of sodalite at temperatures from 15 K to 1000 K. The Journal of Chemical Thermodynamics, 1995. 27(10): p. 1119–1132. 17. Mostafa, A.G., et al., Prediction of heat capacities of solid inorganic salts from group contributions. Industrial & Engineering Chemistry Research, 1996. 35(1): p. 343–348.

H. Peng et al. 18. Mostafa, A.G., J.M. Eakman, and S.L. Yarbro, Prediction of standard heats and Gibbs free energies of formation of solid inorganic salts from group contributions. Industrial & Engineering Chemistry Research, 1995. 34(12): p. 4577–4582. 19. Park, H. and P. Englezos, Thermodynamic modeling of sodium aluminosilicate formation in aqueous alkaline solutions. Industrial & Engineering Chemistry Research, 1999. 38(12): p. 4959–4965. 20. Moloy, E.C., Q. Liu, and A. Navrotsky, Formation and hydration enthalpies of the hydrosodalite family of materials. Microporous and Mesoporous Materials, 2006. 88(1): p. 283–292. 21. Zeng, L. and Z. Li, Solubility and Modeling of Sodium Aluminosilicate in NaOH–NaAl (OH)4 Solutions and Its Application to Desilication. Industrial & Engineering Chemistry Research, 2012. 51(46): p. 15193–15206. 22. Kurdakova, S., et al., Thermodynamic properties of synthetic calcium-free carbonate cancrinite. Physics and Chemistry of Minerals, 2014. 41(1): p. 75–83. 23. Liu, Q., H. Xu, and A. Navrotsky, Nitrate cancrinite: Synthesis, characterization, and determination of the enthalpy of formation. Microporous and Mesoporous Materials, 2005. 87(2): p. 146–152. 24. Ogorodova, L., et al., Cancrinite and cancrisilite in the Khibina-Lovozero alkaline complex: Thermochemical and thermal data. Geochemistry International, 2009. 47(3): p. 260–267. 25. LaMacchia, R., Toward a better understanding of desilication product (DSP) precipitation kinetics, in Ninth International Alumina Quality Workshop. 2012, AQW Incorporated: Perth, Western Australia. p. 214–218. 26. Gasteiger, H.A., W.J. Frederick, and R.C. Streisel, Solubility of aluminosilicates in alkaline solutions and a thermodynamic equilibrium model. Industrial & Engineering Chemistry Research, 1992. 31(4): p. 1183–1190. 27. Staker, W., Internal Communication. 2017. 28. Königsberger, E., P.M. May, and G. Hefter, A comprehensive physicochemical model of synthetic Bayer liquors, in Seventh International Alumina Quality Workshop. 2005, AQW Incorporated: Perth, Western Australia. p. 74–77. 29. Adamson, A.N., E.J. Bloore, and A.R. Carr, Basic principles of Bayer process design, in Extractive Metallurgy of Aluminum, G. Gerard and P.T. Stroup, Editors. 1963, Interscience: NY. p. 23–58. 30. Smirnov, M.N., The metastable solubility of silica in aluminate solutions. The Soviet Journal of Non-Ferrous Metals, 1964. 37(1): p. 16–22. 31. Leiteizen, M.G., Kinetics of converting bauxite silica to sodium aluminosilicate. The Soviet Journal of Non-Ferrous Metals, 1972. 13(5): p. 37–40. 32. Cresswell, P., Factors affecting desilication of bayer process liquors. Proceedings of Australian Chemical Engineering Annual Conference, Chemeca 84, 1984. 12: p. 285–292. 33. Hewett, K.J., A.J. White, and G.I.D. Roach, Silica solubility in plant liquors, in AJW23/MDR (Confidential Report). 1987. 34. Smith, P.G., B.L. Fletcher, and C.C. M., Silica equilibrium in caustic aluminate solutions, in CSIRO Minerals Products Communication MPC/P-019. 1992. 35. Sizgek, G.D. and D.D. Aguila, Silica equilibrium equation for a full range of QAL plant liquors, in Queensland Alumina Limited. 1994. 36. Sipos, P., The structure of Al(III) in strongly alkaline aluminate solutions - A review. Journal of Molecular Liquids, 2009. 146(1): p. 1–14. 37. Anderson, G.M. and D.A. Crerar, Thermodynamics in geochemistry: the equilibrium model. 1993: Oxford University Press, USA. 38. Gout, R., et al., Raman spectroscopic study of aluminum silicate complexes at 20°C in basic solutions. Journal of Solution Chemistry, 2000. 29(12): p. 1173–1186.

Chemical Thermodynamics and Reaction Kinetics of Bayer Process Desilication 39. Peng, H., J. Vaughan, and M. Zieba, The thermodynamic approach to predicting silicate solubility. 2015. 40. Peng, H. and J. Vaughan, Aluminate effect on desilication product phase transformation. Journal of Crystal Growth, 2018. 492: p. 84–91. 41. Pokrovski, G., et al., Structure and stability of aluminium-silica complexes in neutral to basic solutions: Experimental study and molecular orbital calculations. Mineralogical Magazine A, 1998. 62: p. 1194–1195. 42. Zeng, L. and Z. Li, Dissolution Behavior of Al, Si and Fe of Diaspore Concentrate in NaOH−NaAl (OH) 4 Solutions at Elevated Temperature. Industrial & Engineering Chemistry Research, 2013. 43. Gout, R., et al., Raman spectroscopic study of aluminum silicate complexes at 20 C in basic solutions. Journal of Solution Chemistry, 2000. 29(12): p. 1173–1186. 44. Agnew, S.F. and C.T. Johnston, Aluminum solubility in complex electrolytes, in WM2013: Waste Management Conference: International collaboration and continuous improvement. 2013: United States. 45. Subotic, B. and J. Bronic, Theoretical and practical aspects of zeolite crystal growth. Handbook of Zeolite Science and Technology, Marcel Dekker Inc., New York–Basel, 2003: p. 129. 46. Duecker, H.C., A. Weiss, and C.R. Guerra, Synthetic crystalline zeolite. 1971, Google Patents. 47. Ruan, S., et al., Desilication of hematite, goethite and iron powder seeded low alumina to caustic liquors. Hydrometallurgy, 2017. 169: p. 297–305. 48. Ruan, S., The Mechanism and Kinetics of Sodium Aluminosilicate Crystallisation in Synthetic Bayer Spent Liquor. 2015, University of South Australia. 49. Teas, E.B. and J.J. Kotte. The effect of impurities on process efficiency and methods for impurity control and removal. in Bauxite Symposium IV, Kingston. 1980. 50. Smith, P., et al. Understanding growth of DSP in the presence of inorganic ions. in Proceedings of the 6th International Alumina Quality Workshop. 2002. Brisbane Convention & Exhibition Centre Queensland, Australia. 51. Vogrin, J., et al., Influence of chloride on sodium aluminosilicate solubility in Bayer liquor. Microporous and Mesoporous Materials, 2020. 299: p. 110086. 52. Vogrin, J., et al., The anion effect on sodium aluminosilicates formed under Bayer process digestion conditions. Hydrometallurgy, 2020. 192: p. 105236. 53. Xu, B. and P. Smith, The effect of iron sources on caustic and alumina recovery from synthetic bayer DSP (sodalite). Hydrometallurgy, 2012. 129: p. 26–29. 54. Lowe, J., et al., Incorporation of impurity anions into DSP: insights into structure and stability from computer modelling. Molecular Simulation, 2006. 32(01): p. 35–44. 55. Lowe, J., et al. Morphology and crystallinity: Insights into the mechanism of growth of DSP. in 7th International alumina quality workshop. 2005. 56. Whittington, B. and T. Fallows, Formation of lime-containing desilication product (DSP) in the Bayer process: factors influencing the laboratory modelling of DSP formation. Hydrometallurgy, 1997. 45(3): p. 289–303. 57. Barnes, M.C., J. Addai-Mensah, and A.R. Gerson, The mechanism of the sodalite-to-cancrinite phase transformation in synthetic spent Bayer liquor. Microporous and Mesoporous Materials, 1999. 31(3): p. 287–302. 58. Barnes, M.C., J. Addai-Mensah, and A.R. Gerson, The kinetics of desilication of synthetic spent Bayer liquor and sodalite crystal growth. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 1999. 147(3): p. 283–295.

11

59. Subotić, B., et al., Transformation of zeolite A into hydroxysodalite: I. An approach to the mechanism of transformation and its experimental evaluation. Journal of Crystal Growth, 1980. 50(2): p. 498–508. 60. Peng, H., D. Seneviratne, and J. Vaughan, Role of the amorphous phase during sodium aluminosilicate precipitation. Industrial & Engineering Chemistry Research 2018. 57(5): p. 1408–1416. 61. Peng, H., M. Ding, and J. Vaughan, The Anion Effect on Zeolite Linde Type A to Sodalite Phase Transformation. Industrial & Engineering Chemistry Research, 2018. 57(31): p. 10292–10302. 62. Reyes, C.A.R., C. Williams, and O.M.C. Alarcón, Nucleation and growth process of sodalite and cancrinite from kaolinite-rich clay under low-temperature hydrothermal conditions. Materials Research, 2013. 16(2): p. 424–438. 63. Shi, L., et al., Desilication of low alumina to caustic liquor seeded with sodalite or cancrinite. Hydrometallurgy, 2017. 170: p. 5–15. 64. Radomirovic, T., et al., Crystallization of sodalite particles under Bayer-type conditions. Hydrometallurgy, 2013. 137: p. 84–91. 65. Vogrin, J., et al., Synthesis of zeolites using kaolin in concentrated sodium hydroxide-aluminate solutions. Applied Clay Science, 2023. 244: p. 107106. 66. Baur, W.H. and R.X. Fischer, LTN-type zeolite framework as an interpenetrating net of KFI-and SOD-type parts homeomorphic to cuprite, Cu2O. Acta Crystallographica Section B: Structural Science, 2007. 63(2): p. 229–234. 67. Abdullahi, T., Z. Harun, and M.H.D. Othman, A review on sustainable synthesis of zeolite from kaolinite resources via hydrothermal process. Advanced Powder Technology, 2017. 28 (8): p. 1827–1840. 68. Johnson, E. and S.E. Arshad, Hydrothermally synthesized zeolites based on kaolinite: a review. Applied Clay Science, 2014. 97: p. 215–221. 69. Zheng, K., et al., The influence of sodium carbonate on sodium aluminosilicate crystallisation and solubility in sodium aluminate solutions. Journal of Crystal Growth, 1997. 171(1): p. 197–208. 70. Bosnar, S., et al., Influence of anions on the kinetics of zeolite A crystallization:: a population balance analysis. Journal of crystal growth, 2004. 267(1–2): p. 270–282. 71. Wang, S., et al., Revealing the effect of anions on the formation and transformation of zeolite LTA in caustic solutions: an in-situ synchrotron PXRD study. Crystal Growth & Design, 2023. 23(5). 72. Riley, G., et al., Plant impurity balances and inclusion in DSP, in Fifth International Alumina Quality Workshop. 1999, AQW Incorporated: Bunbury, Western Australia. p. 404–414. 73. Lowe, J.L., DSP in the Bayer process: A fundamental study of its precipitation and role in impurity removal, in Department of Applied Chemistry. 2007, Curtin University of Technology. 74. Flint, E.P., L. Shartsis, and L.S. Wells, Method of reducing the concentration of silica in sodium aluminate solutions (US2519362 A). 1950. 75. Feher, I., et al., Reducing or compensating for sodium hydroxide loss produced during alumina manufacture. Fr Demande, 1973. 2: p. 166–188. 76. Teas, E.B. and J.J. Kotte, The effect of impurities on process efficiency and methods for impurity control and removal, in Proceedings of Bauxite Symposium, No. IV. 1980, Journal of the Geographic Society of Jamaica. p. 100–129. 77. Whittington, B.I., B.L. Fletcher, and C. Talbot, The effect of reaction conditions on the composition of desilication product (DSP) formed under simulated bayer conditions. Hydrometallurgy, 1998. 49(1–2): p. 1–22. 78. Seimiya, S. and M. Shietoshi, Studies on sodalite compounds in the Bayer process (2nd report). Journal of Japan Institute of Light Metals (Keikinzoku), 1962. 12(5): p. 286–291.

12 79. Seimiya, S. and S. Mori, Study on sodalite compounds in the bayer process (3rd Report): Dissolution phenomenon of sodalite compounds in sodium aluminate solution. Journal of Japan Institute of Light Metals, 1962. 12(6): p. 351–354. 80. Seimiya, S., Some properties of sodalite in red mud, in Extractive Metallurgy or Aluminum, G. Gerard and P.T. Stroup, Editors. 1963, Interscience: NY. p. 115–132.

H. Peng et al. 81. Smith, P.G., et al., The composition of DSP formed under predesilication and high temperature Bayer digestion conditions. Light Metals, 2001. 1: p. 5–11. 82. Vogrin, J., et al., The influence of sodium sulphate on sodium aluminosilicate solubility in Bayer liquor aiding the desilication process. Hydrometallurgy, 2023. 219: p. 106079.

Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites Michael Coley, Anthony Greenaway, Alicia Buckley, Khadeen Henry, Jheanell James, and Jason Brown

Abstract

Introduction

Low-temperature alumina plants in Jamaica traditionally process hematite-rich ores with manageable concentrations of mineral impurities. As these ores become scarce, the new bauxite mines have lower available alumina, higher goethite, and higher concentrations of phosphorus, chromium, and manganese contaminants. Since the minerals associated with some of the minor impurities have not been identified, this paper explores whether the contaminants accumulate in the coarse or fine bauxite fractions. Samples of hematite-rich bauxite from the traditional mining areas and also goethitic ores from the new mines were wet-sieved into three size fractions (+200, −200/+325, and −325 mesh). Each fraction was investigated using X-ray fluorescence (XRF) and X-ray diffraction (XRD) spectroscopies. Following digestion, their caustic soluble concentrations of phosphorus, chromium, and zinc were also determined. The traditional bauxites had no correlation between size fraction and the alumina, iron, calcium, or phosphorus concentrations in the ore, however larger particles had slightly more manganese impurities. In contrast, larger particles from the new mines had higher concentrations of iron and chromium impurities while smaller particles contained marginally more alumina. During digestion, the new bauxites gave higher caustic soluble P, Zn, and Cr and will be more challenging and costly to process into alumina that meet the required specifications. Keywords




Jamaican bauxite Hematite Caustic soluble phosphorus




Goethite




Size fraction

M. Coley (&)
A. Greenaway
A. Buckley
K. Henry
J. James
J. Brown Department of Chemistry, The University of the West Indies, Mona, Kingston, 7, Jamaica e-mail: [email protected]




The Bayer process uses high-pressure digestion conditions to extract alumina minerals from bauxite [1]. During the process, a variety of impurities often dissolve and eventually impact the liquor characteristics, alumina yield, and quality, often making the alumina unsuitable for smelter operations [2]. Traditionally mined Jamaican bauxites are gibbsitic, red in colour, and contain hematite as the main iron mineral. These ores, generally found at low altitudes, contain low concentrations of phosphorus and metal impurities. As supplies of these ores decline, new mining areas with goethitic bauxite having lower available alumina and higher concentrations of chromium, zinc, and phosphorus are being explored [3]. These yellow bauxites are mainly found in hillside locations that are dominated by pocket deposits with highly variable mineral compositions [4, 5]. The local Bayer plants have well-established methods for processing hematite-rich ores containing 0.1–0.4% P2O5. However, as the use of goethitic bauxites becomes imminent, plants require new strategies to manage bauxite blends with higher concentrations of phosphorus (1–2.5% P2O5), zinc (0.02–0.06% ZnO), and chromium (0.06–0.20% Cr2O3). High P2O5 concentrations in bauxite have significant implications for soluble phosphorus control and lime requirements; liquor viscosity and alumina yield and purity may also be impacted [3]. Poorly controlled soluble phosphorus may reduce mud settling rates, lower filtration efficiency, cause scale formation, and increase fines production during precipitation [6]. The main phosphorus mineral in goethitic bauxite is crandallite (CaAl3(PO4)2(OH)5
H2O), however apatite (Ca5(PO4, CO3)3 (F, OH, Cl), wavellite (Al3(PO4)2(OH)3(H2O)5, and variscite (AlPO4
2H2O) may also be present. These minerals dissolve differently in caustic, for example, apatites are sparingly soluble while crandallite dissolves quite readily. Hence, unless the phosphorus mineralogy is known, it is difficult to establish the lime requirements needed for phosphorus control

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_2

13

14

M. Coley et al.

during predesilication and digestion [7, 8]. Managing the metal impurities (Cr2O3, ZnO, and MnO) is also challenging as their low concentrations in the ore make it difficult to identify their source minerals using traditional X-ray diffraction (XRD) techniques. It is also difficult to predict their caustic soluble concentrations and to design protocols for their control [7, 9]. Jamaican bauxites are mainly fine particles but these may aggregate to form larger units. The presence of sand, limestone, and a variety of other minerals often results in crushed bauxites having a fairly wide range of particle size. This paper compares the characteristics of traditionally mined, red, hematite-rich Jamaican bauxite with samples of yellow, goethitic ores that are slated for mining in the near future. These new bauxites have high phosphate and metal impurity concentrations. Crushed exploration samples from both mines were wet-sieved and fractioned into three typical categories: >75 lm (+200 mesh), 45–75 lm (−200/+325 mesh), and 60% of bulk particles in the +200 fraction. The remaining two bauxites (DCS10B and DCS14B) were dominated by fines (*50% −325 fraction). For the new mining area, only two bauxites had >50% of particles in the +200 fraction. On average, 45.8% of particles from the new mines were in the −325 fraction in comparison to only 32% of particles from the traditional mines. For both bauxite mines, less than 10% of particles were of intermediate size (−200/+325 fraction). Data in Table 3 shows that the % available alumina was slightly higher for samples from the traditional mines. Across both mining areas, the larger sized fractions of each bauxite sample (+200 and −200/+325 fractions) had similar % available alumina; these were consistently higher than the alumina content of the −325 bauxite fractions.

16

M. Coley et al.

Table 3 Size fraction and % available alumina of bauxite samples from the traditional and new mining areas Traditional hematite-rich mining area

New goethitic mining area

Goethitic bauxites

Sample fractions

% mass

% available alumina

Goethitic bauxites

Sample fractions

% mass

% available alumina

DCS11A

+200

68.7

47.9

DCN30A

+200

38.6

40.3

−200/+325 DCS10B

8.6

47.0

−200/+325

11.8

40.5

−325

22.7

44.7

−325

49.5

37.6

+200

39.8

42.6

+200

47.1

9.3

41.3

−200/+325 DCS21B

−325

44.9

44.2

+200

38.7

43.6

51.0

40.2 46.7

9.7

46.0

−200/+325

6.4

41.2

22.7

43.1

−325

55.0

39.2

+200

34.4

+200

61.1

40.1

−200/+325

11.5

41.1

−325

27.4

37.4

+200

39.9

43.0

DCN10A

DCN16A

−200/+325 DCN21

6.0

40.9 41

−325

59.5

37.7

+200

56.1

40.4

9.6

43.4

−200/+325

11.9

40.2

−325

50.5

41.5

−325

32.0

36.9

+200

73.1

38.6

+200

58.6

7.5

39.1

−200/+325 Average size fractions

47.1

67.6

−200/+325 DCS16A

47

8.0

−325

−325

DCS14B

−200/+325

+200 −200/+325

DCS08B

DCN47A

−325

19.4

37.5

+200

58.4

43.2

9.4

43.0

32.3

40.7

−200/+325 −325

XRD Analysis Figure 1 shows the overlaid diffractograms of the unsieved DCN21A bauxite sample and its +325 and −325 fractions. DCN21A is from the new mining area. Scans of the sieved and unsieved samples are similar which suggests that their mineralogies are also similar. The broad peak at values >15° 2h concealed the boehmite peaks. This broad peak was characteristic of all the scans recorded on the instrument during the time and resulted from interferences due to the sample holder. Fractions of the traditionally mined bauxites were not examined via XRD.

XRF Analysis Traditional Bauxite Mines The XRF data for bauxites from the traditional mines are in Table 4 and those for the new mining area are in Table 5. As with previous tables, samples are organized based on the XRF concentrations of P2O5 in the original unsieved samples. Data for the traditionally mined ores show that the concentrations

DCN29

−200/+325 Average size fractions

7.8

39.1 41

−325

33.7

37.8

+200

45.6

41.9

−200/+325 −325

8.6

41.8

45.8

38.9

of Al2O3, SiO2, CaO, P2O5, and ZnO are roughly similar within the three size fractions for each bauxite (Table 4). Figure 2 shows that the MnO concentrations increase with particle size for all but one sample (DCS11A). Significant differences in MnO concentrations are observed for the fractions of samples DCS21B and DCS14 and it would be worthwhile to probe these fractions to explore whether there is improved opportunity to identify the Mn minerals in each. Similar differentiation was not observed for any other element within the traditional bauxite mines.

The New Bauxite Mines The % Fe2O3 and % Cr2O3 in goethitic bauxites from the new mining area occur in higher concentrations in the +200 bauxite fractions (Table 5). The % Fe2O3 were lowest in the −325 bauxite fractions except maybe for sample DCN21A. The fine fractions tended to have higher % P2O5, however the trends were not always consistent (Fig. 3). The % P2O5 in fractions of DCN29 from the new mining area (Fig. 3) are well differentiated and are potentially useful for identifying the phosphorus minerals in each. Studies of their caustic soluble concentrations could also be pursued. No defined

Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites

17

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

4

20

10

30

40

50

60

70

2-Theta - Scale

DCN21A

DCN21A +325 fraction

Dunsieved

DCN21A

D -325 fraction

Fig. 1 Overlaid X-ray diffractograms of DCN21A, a bauxite from the new mining area: scans are of unsieved DCN21A and the +325 and −325 fractions Table 4 XRF data for fractions of bauxite samples from the traditional mining area Sample

% Al2O3

% Fe2O3

% SiO2

% CaO

% P2O5

% Cr2O3

% ZnO

% MnO

DCS 11A +200

47.2

18.2

0.8

0.03

0.11

0.164

0.014

0.16

DCS 11A +325

47.3

17.7

0.8

0.04

0.09

0.137

0.014

0.16

DCS 11A −325

47.9

18.2

0.82

0.04

0.1

0.139

0.016

0.16

DCS 10B +200

47.3

19.5

2.49

0.27

0.15

0.147

0.023

0.34

DCS 10B +325

47.0

18.4

2.38

0.21

0.14

0.110

0.023

0.21

DCS 10B −325

47.1

18.4

2.4

0.19

0.15

0.105

0.024

0.17

DCS 21B +200

46.6

19.6

0.58

0.04

0.15

0.113

0.028

1.11

DCS 21B +325

46.2

19.1

0.57

0.04

0.15

0.108

0.021

0.76

DCS 21B −325

46.7

19.6

0.61

0.04

0.15

0.110

0.022

0.71

DCS 8B +200

47.1

18.3

2.34

0.08

0.16

0.079

0.027

0.53

DCS 8B +325

47.1

18.5

2.34

0.06

0.16

0.078

0.024

0.37

DCS 8B −325

47.7

19.2

2.44

0.06

0.17

0.082

0.025

0.34

DCS 14 +200

47.9

16.0

2.6

0.11

0.18

0.121

0.102

2.53

DCS 14 +325

47.8

16.4

2.59

0.10

0.19

0.120

0.071

1.64

DCS 14 −325

47.9

16.6

2.61

0.10

0.19

0.121

0.045

0.90

DCS 16A +200

46.0

18.0

3.48

0.15

0.77

0.121

0.089

0.38

DCS 16A −325

46.5

18.5

3.44

0.19

0.81

0.114

0.093

0.36

trends were observed between the % Al2O3, SiO2, CaO, MnO, and ZnO in relation to the particle size of samples from the new mining area. An assessment of the XRF data for bauxite samples from both mining areas shows that some metal species concentrate in the coarse or fine bauxite particles but this is not widespread.

It is probable that the characteristics of bauxite fractions of each sample are mainly determined by small constituent bauxite particles that aggregate together. The data so far are not conclusive however they suggest that fractioning bauxite samples followed by use of conventional XRF, XRD or other traditional techniques will not concentrate mineral species

18

M. Coley et al.

3.0

% MnO

2.5 2.0 1.5 1.0 0.5 0.0

11A 11A 11A - 10 B 10B 10B - 21B 21B 21B - 8B 8B 8B - 14 14 14 - 16A 16A +200 +325 325 +200 +325 325 +200 +325 325 +200 +325 325 +200 +325 325 +200 325

Sample Names - Traditional Mines Fig. 2 % MnO in fractions of bauxite samples from the traditional hematite-rich mining area. (DCS precedes the name of each sample in the chart)

Table 5 XRF data for fractions of bauxite samples from the new mining area Sample

% Al2O3

% Fe2O3

% SiO2

% CaO

DCN 30A +200

46.2

22.9

0.34

0.09

DCN 30A +325

48.0

17.4

0.32

0.09

DCN 30A −325

48.4

16.1

0.32

DCN 10A (+200)

46.0

18.6

0.45

DCN10A (−325)

46.1

16.8

DCN 16A +200

42.9

25.8

DCN 16A −325

47.9

DCN 21 +200

46.0

% P2O5

% Cr2O3

% ZnO

% MnO

0.85

0.316

0.080

2.19

0.8

0.169

0.100

2.35

0.10

0.81

0.141

0.105

2.32

1.02

3.08

0.241

0.106

1.17

0.47

1.16

3.15

0.139

0.090

0.75

1.17

0.24

1.62

0.711

0.041

0.06

16.4

1.4

0.40

1.98

0.198

0.033

0.04

14.8

1.83

3.21

5.12

0.169

0.097

0.29

DCN 21 +325

46.6

14.7

1.89

3.03

5.28

0.128

0.098

0.34

DCN 21 −325

46.3

14.2

1.98

2.95

4.85

0.113

0.100

0.33

DCN 29 +200

44.8

21.8

0.36

1.87

6.03

0.391

0.053

0.24

DCN 29 +325

47.0

16.0

0.36

2.58

7.63

0.208

0.053

0.22

DCN 29 −325

47.0

12.2

0.36

3.65

10.20

0.135

0.054

0.16

12 10

% P2O5

8 6 4

2 0

30A 30A 30A 10A 10A 16A 16A 21 21 21 29 29 29 (+200) (+325) (-325) (+200) (-325) (+200) (-325) (+200) (+325) (-325) (+200) (+325) (-325) Samples Names - New goethite-rich mine Fig. 3 % P2O5 in fractions of bauxite samples from the new mining area (DCN precedes the name of each sample in the chart)

Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites

enough to provide improved mineralogy information. The approach seems promising for samples with unusually high impurity concentrations such as DCN29 however.

Elemental Correlations The mineral impurities in bauxites are typically at low concentrations and are difficult to identify using traditional XRD. Simple plots of the elemental concentrations of species in the ore may give clues about their mineral associations [15]. This approach is useful where distinct minerals of the low-concentration species do not exist or where the mineral concentrations are extremely low.

3.0

% MnO

2.5 y = 19.34x - 0.71 R² = 0.34

2.0 % MnO

% MnO

The plots below are for fractions of the bauxite samples from each mining area. Figures 4a, c, and e are for currently mined bauxites while Figs. 4b, d, and f are for the new ores. A plot of % MnO versus % ZnO for bauxite fractions from the traditional mine (Fig. 4a) shows a strong correlation with only one bauxite (DCS16A) deviating significantly from the line. Feret and See identified lithiophorite ((Al,Li)(Mn4+,Mn3+) O2(OH)2) as the main manganese mineral in Jamaican bauxite. They proposed that Zn2+ substitutes for Li+ and results in a strong correlation between the Zn and Mn [16]. They proposed that zinchophorite, a zinc-enriched lithiophorite, is the zinc-contaminating mineral in Jamaican bauxite.

% MnO - DCS 16A

y = 26.05x - 0.16 R² = 0.88

2.0

19

1.0

1.5 1.0 0.5

0.0 0.00

0.03

a

0.06

0.09

0.0

0.12

y = 1.09x + 0.11 R² = 0.14

0.6 0.4 0.2

0.15

y = 1.93x + 1.07 R² = 0.777

8 6 4 2

0.0

0 0.00

0.10

0.20

0.30

d

% CaO

0.20

0.0

1.0

2.0

3.0

4.0

% CaO

0.8

0.15

y = 0.037x - 0.41 R² = 0.824

0.6

% Cr2O3

% Cr2O3

0.10

% ZnO

10 % P 2O 5

% P2O5

0.8

0.10

0.4

y = -0.0038x + 0.19 R² = 0.031

0.05

0.2 0.0

0.00

e

0.05

12

1.0

c

0.00

b

% ZnO

0.0

10.0

20.0 % Fe2O3

30.0

0.0

f

10.0

20.0

30.0

% Fe2O3

Fig. 4 Correlation of selected elemental oxides in bauxites from the traditional (a, c and e) and new goethitic mining areas (b, d and f)

20

M. Coley et al.

Although the number of samples in this study is small, the strong correlation observed between % MnO and % ZnO in the hematitic bauxites suggests occurrence of a single mineral that contains both metals. A plot of % MnO versus % ZnO for the goethitic bauxites suggests that the mineral associations of zinc and manganese are much more complex. Zinc concentrations in these ores are also consistently higher and more than one mineral sources might be present. The higher zinc concentrations and possibly the presence of uncharacteristic zinc-bearing contaminants might have implications for the management of soluble Zn in the new bauxite. Figure 4c shows a very weak correlation between % P2O5 and % CaO. This contrasts with the fairly strong relationship observed in Fig. 4d which shows the plot for the new bauxite mine. Previous work has identified crandallite as the main P mineral in the new mine [6, 11]. Bauxites from both mines will likely have very different caustic soluble P2O5 concentrations. Figure 4e plots % Cr2O3 against % Fe2O3 for the currently mined ore while Fig. 4f gives a similar plot for the new mine. Clear differences in the mineral associations are again observed. This suggests need for careful study of caustic soluble chromium in the new bauxites. Based on the strong correlation in Fig. 4f, the chromium impurity is probably associated with the iron mineral in the new mine.

samples and has an extensive database on the minerals found in bauxites across the world. For a given bauxite, the programme apportions the metal species to their natural host minerals such that they are distributed to match the XRD profile while achieving a balance with the XRF concentrations [14]. Only one bauxite sample was studied via XDB in this preliminary work. Fractions of the unsieved and sieved portions of sample DCN21A (+200, −200/+325, and −325) were examined. The XDB output (Table 6) predicts the distribution of the metal oxides among the various bauxite minerals in the sieved and unsieved fractions of DCN21A. The programme identified gibbsite as the main alumina mineral in all the samples. All samples had additional alumina in boehmite, crandallite, and aluminous goethite; however, nordstrandite was only detected in the −325 fraction. Based on the XDB analysis, crandallite was the only P mineral detected in the unsieved DCN21A. Analysis of the +200 mesh DCN21A fraction identified both crandallite and hydroxyapatite however. While the predictive ability of XDB relies heavily on the quality of the XRD scans and the accuracy of the XRF data, the results obtained for DCN21A demonstrate the potential of using XDB to study fractions of bauxite samples, especially those with unusually high concentrations of mineral impurities. Further work is required to determine whether these approaches can eventually identify the presence of low-concentration mineral species, especially those associated with Cr or Zn.

XDB Analysis The XDB software was applied to fractions of the bauxite samples to predict their mineral constituents by relating their measured XRF and LOM data with the XRD scan obtained in each case. The programme was specially developed for bauxite

Caustic Soluble Concentrations The concentrations of caustic soluble P2O5, Cr2O3, and ZnO were studied for three bauxite samples from the new mining

Table 6 XDB predictions of the mineral species in the unsieved and sieved fractions of sample DCN21A (new mining area) Crandalite

Alumino-goethite

Gibbsite

Boehmite

Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325)

Fe2O3 (%) TiO2 (%) CaO (%)

2.44

2.32

2.35

2.18

P2O5 (%)

6.16

5.86

5.95

5.53

6.64

6.32

6.41

5.95

15

15.26

14.7

14.3

1.69

1.72

1.66

1.61

SiO2 (%) Al2O3 (%)

Anatase

Quartz

38.78

33.47 34.78 33.8 Hydroxyapatite

not observed

4.14

4.43 4.51

Norstrandite

Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325) Unsieved (+200) (+325) (-325)

Fe2O3 (%) TiO2 (%)

1.98

1.91

2.02

1.99

CaO (%)

0.67

P2O5 (%) SiO2 (%) Al2O3 (%)

0.51 1.89

1.98 0.86

Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites

21

Table 7 % caustic soluble P2O5, Cr2O3, and ZnO in different size fractions of bauxite samples from the new mining area Sample

% P2O5 (XRF)

% caustic soluble P2O5

% Cr2O3 (XRF)

% caustic soluble Cr2O3

% ZnO (XRF)

% caustic soluble ZnO

DCN21 (+200)

5.12

55.0

0.169

14.8

0.097

12.1

DCN21 (+325)

5.28

52.5

0.128

17.2

0.098

10.6

DCN21 (−325)

4.85

55.0

0.113

15.5

0.100

11.0

DCN16A(+200)

1.62

88.0

0.711

10.0

0.041

1.2

DCN16A (−325)

1.98

80.5

0.198

21.8

0.033

1.7

DCN30A (+200)

0.85

99.5

0.316

25.0

0.080

6.2

DCN30A (+325)

0.80

100.5

0.169

35.0

0.100

6.35

DCN30A (−325)

0.81

101

0.141

38.5

0.105

5.05

area (Table 7). DCN30A is a typical low P bauxite but its low % CaO suggests presence of a non-crandallite P mineral, possibly variscite (Table 5). DCN21A has excess % CaO beyond that due to crandallite alone; however, for sample DCN16A, the %CaO:%P2O5 ratio suggests that crandallite is likely the only P mineral that is present (Table 5) [11]. The caustic soluble concentrations of bauxite fractions from the currently mined areas were not studied as their soluble concentrations in plant liquors are typically low and are well known. Strategies to manage their build-up in the traditional low-temperature process are well established. On the other hand, the solubility characteristics of P2O5, Cr2O3, and ZnO from the new bauxite mines are unclear and the extent to which they dissolve from each bauxite fraction has not been reported. Based on data from the three bauxites, there are no major differences in the extraction of phosphorus from the fractions of each ore (Table 7). There were major differences however between each sample. Several factors could be responsible for the differences, for example, the nature of the P mineral (e.g., in DCN30A) or the presence of excess CaO in the bauxite (e.g., in DCN21A), among others [11]. On average, about 70% of the P2O5 that enters with the bauxite is expected to end up in the liquor [11]. The results in this study, however, show a much wider range and illustrate the difficulties that will be encountered in managing process impurities from these ores. As seen for soluble phosphorus (Table 7), there is no significant differentiation in the soluble Zn concentrations among the different bauxite fractions. Cr2O3 was more extractable from the finer fractions of DCN16A and DCN30A, however there is insufficient data from which to draw conclusions.

Management of Caustic Soluble P2O5, ZnO, and Cr2O3 from the New Mining Area Of the three impurities studied in this work, the management of P2O5 will likely require the most significant effort and possibly will incur the greatest cost. The solubilities of the P2O5 minerals in this work (52.5–100%) are much higher than those observed for the Zn (1.2–12.1%) and Cr impurities (10.0–38.5%) (see Table 7). Bauxite blending has long been used as an option for managing high concentrations of bauxite impurities, hence this section explores the use of blending to manage the high % P2O5 in the new bauxite. Samples of currently mined bauxites were digested by themselves and in blends with the goethitic ores. The mg/L caustic soluble P2O5 was measured in each case and is reported in Table 8. When bauxite from the traditional mines (0.22% P2O5) was digested under low-temperature conditions, the average P2O5 concentration was 169 mg/L. In contrast, separate digestions of unsieved ores from the new mines containing 1.0%, 2.0%, and 3.0% P2O5 gave soluble P2O5 concentrations of 981 g/L (75% caustic soluble P2O5), 2287 mg/L (85.5% caustic soluble P2O5), and 3019 mg/L P2O5 (74.9% caustic soluble P2O5), respectively (Table 8). Separate blends comprising 25, 50, and 75% ore from the new mines (1, 2, and 3% P2O5) were combined with the matching proportions of the traditionally processed ore (i.e., 75, 50, and 25% of traditional ore; 0.22% P2O5). The soluble phosphorus concentrations (mg/L P2O5) of digest liquors from each goethitic bauxite sample and its blends with the low P ore (from current mines) were plotted against the % P2O5 in the bauxite. The plots gave linear relationships with R2 > 0.995 (Table 8). As the total % P2O5

22

M. Coley et al.

Table 8 Caustic soluble P2O5 based on digests of a traditional hematite-rich bauxite, digests of a goethitic bauxite, and digests of blends from both ores Blends

1% P2O5 (new bauxite) with 0.2% P2O5 (traditional bauxite)

2% P2O5 new bauxite) with 0.2% P2O5 (traditional bauxite)

Total P2O5 (%)

mg/L P2O5

% caustic soluble P2O5

Total P2O5 (%)

mg/L P2O5

100% traditional bxt (TB)

0.22

165

57.5

0.22

75% TB, 25% NB

0.42

387

70.9

50% TB, 50% NB

0.63

615

25% TB, 75% NB

0.83

100% new bauxite (NB)

1.03

Equation and R2

y = 1005.9x − 49.4; R2 = 0.999

3% P2O5 new bauxite) with 0.2% P2O5 (traditional bauxite)

% caustic soluble P2O5

Total P2O5 (%)

mg/L P2O5

186

64.8

0.22

157

54.9

0.68

592

67

0.94

836

68.4

75.1

1.13

1150

78.3

1.66

1527

70.8

914

84.7

1.59

1697

82.1

2.37

2245

72.9

981

75.2

2.04

2227

85.5

3.09

3019

74.9

y = 1139.8x − 120; R2 = 0.997

in the blends increased, higher soluble P2O5 concentrations were observed. This means that blends of the new bauxites and the traditional ores can be successfully processed and the resulting soluble P2O5 concentrations can be predicted (Table 8). Lime requirements for phosphorus control can therefore be estimated. Accuracy of the estimates will be impacted by the phosphorus mineralogy of the ores and the calcite concentrations. When goethitic bauxite with 1.0% P2O5 was blended 50/50 with traditional bauxite containing 0.22% P2O5, the resulting soluble P2O5 in the digestion liquor was 615 mg/L. This is almost four times higher than the 165 mg/L P2O5 that the plant would be accustomed to if the traditional ore (0.22 % P2O5) was processed. Much higher soluble P2O5 concentrations occur with bauxite blends that contain higher proportions of the new bauxite. Blending bauxite from the new mines with low P2O5 ores is the best option currently available for managing the high phosphorus concentrations in the new bauxite mines. Even with blending, higher soluble P2O5 concentrations will occur. This will require more CaO to precipitate the impurity and will result in a higher mud load. New processes for managing soluble P2O5 are therefore required if these high phosphorus bauxites are to be economically processed to yield high-quality smelter-grade alumina [17]. The processing of goethitic bauxites is usually associated with slow mud settling and lower alumina yields. Increased soda and alumina losses, higher energy costs, and lower quality alumina also occur [3, 18]. The processing of ores from the new bauxite mines will likely experience some of these challenges; however, in addition, the management of P2O5, ZnO, and Cr2O3 impurities in the bauxite will further complicate alumina production from these replacement ores [19, 20].

% caustic soluble P2O5

y = 994.7x − 90.4; R2 = 0.999

Conclusions On average, bauxite from the traditional mining areas had larger +200 fractions while the new bauxites had larger −325 fractions. For the traditional mines, concentrations of Al2O3, SiO2, CaO, P2O5, and ZnO were similar for all three fractions of each bauxite; however, the % MnO values were higher in the +200 size fractions. In the new mining areas, there were higher % Fe2O3 and Cr2O3 in the +200 fractions but Al2O3, SiO2, CaO, MnO, and ZnO were similar for the three fractions of each bauxite sample. Nevertheless, bauxites with high mineral concentrations (MnO, Cr2O3, or P2O5) tended to show enrichment in one of the fractions. Study of these fractions using XDB or other sophisticated techniques may show enhanced potential to identify minerals that were not observed during analysis of the bulk, unsieved bauxites. The total impurity concentrations (% P2O5, % Cr2O3, % ZnO, and % MnO) in the new bauxites are sometimes more than five times higher than in the traditionally processed ores. Processing these ores will therefore require strategies to manage these impurities. There are options to manage soluble zinc impurities but these are traditionally associated with high-temperature bauxite plants. Pilot-scale recirculating experiments are probably needed to assess their relevance for use in low-temperature processing of the new bauxite feed. Managing soluble Cr2O3 may also be challenging and more data are therefore needed to better understand the behaviour of these impurities. Whether the new bauxites are processed by themselves or in blends with traditional hematitic ores, much higher caustic soluble impurities will result. Managing these will likely

Challenges with Characterizing and Processing Goethite-Rich Jamaican Bauxites

impact production costs and also alumina yield and purity. If the process slurries settle slowly, this will further complicate the processing operations.

References 1. Ostap, S., Effect of Bauxite Mineralogy on its Processing. In Bauxite Proceedings of 1984 Bauxite Symposium. The Society of Mining Engineers of the American Institute of Mining, Metallurgical and Petroleum Engineers Inc.: Los Angeles CA. pp 651–671. 2. Teas, E.B. and J.J. Kotte. The effect of impurities on process efficiency and methods for impurity control and removal. In Bauxite Symposium IV, Kingston. 1980. 3. Lawson D., Rijkeboer A., Andermann L.J., Mooney A. Impact of Jamaican bauxite mineralogy on plant operations. Light Metals 2008, 113–118. 4. Libby, S.C., Grubbs, D.K., Rodenburg, J.K., Wefers, K.A. (1980); The Geology, Mineralogy and Clarification Properties of Red and Yellow Jamaican Bauxites. Jamaica. Jour. Geol. Soc. Jamaica, 176–186. 5. Grubbs, D.K. (1982) The Iron and Phosphate Mineralogy of the Bauxites in Harmons Valley and the Mocho Mountains, Jamaica. Jour. Geol. Soc. Jamaica, 55–61 6. Young N, Coley M, Greenaway A (2019) Mineralogical investigations of Jamaican hematite-rich and goethite-rich bauxites using XRD and solid state 27Al and 31P MAS NMR spectroscopy, Journal of Geochemical Exploration, Vol. 200, 54–76 7. Authier-Martin, M., Forté, G., Ostap, S., See, J. (2001) The Mineralogy of Bauxite for Producing Smelter-Grade Alumina. JOM Journal of Minerals, Metals and Materials Society, 53, 36–40 8. Whittington B (1996) The Chemistry of CaO and Ca(OH)2 Relating to the Bayer Process. Hydrometallurgy, 43, 13–35. 9. Vind, J.; Vassiliadou, V.; Panias, D. Distribution of trace elements through the Bayer process and its by-products. In Proceedings of the 35th International ICSOBA Conference (Travaux46), Hamburg, Germany, 2–5 October 2017; pp. 255–267.

23

10. Kirwan L., Lawson D., Rijkeboar A., Hodnett K., Mooney A., Walker R. (2009) Characterisation of iron mineralogy in Jamaican Bauxite and associated aspects of alumina and soda losses. Light Metals, 133–138 11. Henry, K., Coley, M., Greenaway, A., (2018) The dissolution of Phosphorus from Jamaican bauxites under low temperature Bayer conditions, Hydrometallurgy, 179, 132–140. 12. Sajó, I.E. X-Ray diffraction quantitative phase analysis of Bayer process solids. In Proceedings of the10th International Congress of ICSOBA, Bhubaneshwar, India, 28–30 November 2008; pp. 71– 76. 13. Sajó, I.E. XDB Powder Diffraction Phase Analytical System, version3.107; Software, Copyright Sajó, I.E. 1987–2016; Sajó, I. E.: Budapest, Hungary, 2005. 14. Feret, F., Martin, M., Sajo, I. (1997) Quantitative Phase Analysis of Bidi-koum Bauxites (Guinea). Clays Clay Miner. 45, 418–427. 15. Greenaway, A.M., Bucknor, A.N., Henry K. and Coley, M.D. (2013) Identifying possible associations of minor and trace elements with major elements in Jamaican hillside and catchment bauxites using multivariate analysis tools; Minerals Engineering 53, 124–135 16. Feret, F.; See, J. (2006) Occurrence and Craracterization of Zn and Mn in Bauxite. Light Metals 41–45. 17. Lindsay, S. J. (2005) SGA requirements in coming years. Light Metals 2005, 117–122. 18. Powell KA, Kirwin LJ, Hodnett, K, Lawson D, Rijkeboer A. Characterisation of Alumina and soda losses associated with the processing of geothitic rich Jamaican bauxite. Light Metals 2009, 151–156. 19. Lawson, D., Rijkeboer, A., Dajkovich, D., Jackson, M., Lawrence, H. Approaches to the Processing of Jamaican Bauxite with High Goethite Content. In: Grandfield, J. (eds) Light Metals 2014, 11– 14. 20. Coley, M.D., Greenaway, A. M., Henry-Herah, K. E. Impacts of Mineralogy on Soluble Phosphorus Concentrations during Low Temperature Temperature Processing of Jamaican Bauxite. In: Tomsett, A (ed) Light Metals 2020, 3–11.

Sugar-Derived Causticization Additives for the Bayer Process Amit Desai, Jun Su An, and LoongYi Tan

Abstract

Introduction

Bauxite digestion in the Bayer process results in a buildup of sodium carbonate in the caustic liquor, which reduces the efficiency of alumina production. This effect is mitigated by lime causticisation, during which carbonate is removed, preferably as calcium carbonate. The extent and efficiency of causticisation in Bayer liquors are limited by the formation of tricalcium aluminate (TCA) and alumino-carbonates, which also reduce overall plant productivity due to the consequent loss of alumina from the liquor. We demonstrate that TCA formation can be minimized or eliminated by using sugar-derived molecules, specifically gluconic acid, a mixture of gluconic and glucaric acids, and sodium gluconate. This reduction in TCA formation improves the efficiency of causticisation by three to four times as determined by the rate constants for carbonate consumption or by the relative increase of C/S ratio. This in turn reduces alumina and caustic losses in the causticiser, with the potential to improve the plant liquor C/S and hence the overall productivity of the Bayer circuit. Keywords







Bayer process Causticization Organic acids and salts Tricalcium aluminate inhibition

A. Desai (&)
J. S. An
L. Tan Solugen Inc., 14549 Minetta Street, Houston, TX 77035, USA e-mail: [email protected] J. S. An e-mail: [email protected]




Bayer process liquors are inevitably contaminated with carbonate, which arises from base-catalyzed oxidation of organic compounds and dissolution of inorganic carbonate from the bauxite, and absorption of CO2 from the air [1, 2]. The carbonate content of Bayer liquor is conventionally expressed as the caustic to soda ratio, C/S, where C = [NaOH] in units of g/L of Na2CO3 and S = C + [ Na2CO3] in the same units. C/S = 1 for pure NaOH. Lime causticisation is used to limit the carbonate content of the liquor and thus maintain its productivity. Improvements in the causticization process therefore lead to better liquor productivity, and hence higher alumina production, as well as lower lime consumption. The C/S of the main liquor stream in Bayer refineries is generally in the range of 0.80– 0.95, depending on carbonate inputs and the efficiency of the causticisation process. The desired solid product of causticisation (calcium carbonate) is rarely formed directly in Bayer liquors, but mostly via alumino-carbonate intermediates, with tricalcium aluminate (TCA) as the undesired byproduct. TCA formation not only reduces causticisation and increases lime usage but also results in alumina loss that directly reduces the overall alumina yield of the Bayer plant. The relative proportions of calcium carbonate and TCA depend on a range of equilibrium and kinetic factors [3, 4]. Approaches to limiting TCA formation include causticising at high temperature (e.g., at 145 °C) in a single stage [5] or multiple stages [6], or at normal temperatures ( Co > Ni

D2EHPA

[19]

D2EHPA 10% Fe pH 2 55% ALIQUAT 336 5% pH 1 24%

UO2(SO4)34−, NO3–, Cl–, HSO4–, Fe (III)

Alamine 336

[20]

25 °C. The liquors obtained were analyzed using inductively coupled plasma optical emission spectrometry (ICP-OES 700 series. The precipitates were oven-dried at a temperature of 50 °C for a duration of 24 h. then were calcined in a muffle furnace at 1200 °C for 1 h. The calcined sample was then digested in a microwave oven with 6.5 mL of phosphoric acid and 3.5 mL of sulphuric acid. The resulting solution was then analyzed using the ICP-OES 700 Series. Multi-element ICP standards with concentrations of 100 mg L−1 (Al, Ca, Ti, Fe, K, Mg, Na, Ga, Sc, Zr, V, and Li) and mono-elemental ICP standard solutions with concentrations of 1000 mg L−1 for silicon, 10,000 mg L−1 for gallium, and 1000 mg L−1 for scandium were used for the preparation of the solutions. For the sample preparation phase, a 65% HNO3 solution with a 3% concentration was used for dilutions. The curve was constructed using concentrations of 1, 2, 4, 6, 8, and 10 ppm. The preferred wavelengths for the elements analyzed by ICP are as follows: Al (396 nm), Ca (396 nm), Fe (238 nm), K (766 nm), Mg (280 nm), Na (589 nm), Si (251 nm), Ti (336 nm), V (292 nm), Ga (294.3 nm), Sc (361.3 nm), and Zr (343.82 nm). Scanning electron microscopy coupled with energy dispersive spectroscopy (SEM–EDS) was carried out on the calcined sample using the Phenom equipment and model Pro-X, with backscattered electrons and a voltage of 10 kV. By exciting the sample with an electron beam, it is possible to create a spectrum where the peak positions identify the element, while the signal intensity corresponds to its concentration.

Results and Discussion Bauxite Sample Characterization The values found in the XRF characterization of the bauxite sample show the following compositions for the oxides: aluminum 63.79%, silicon 1.87%, iron 4.27%, calcium, potassium, magnesium, and manganese 0.01%, and sodium 0.05%.

Acid Leaching Solid: Liquid Variation and Temperature Variation For the solid:liquid ratio parameter, increasing the ratio above 1:10 results in a decrease in the viscosity of the paste (sample + acid), making the leaching stage more difficult. The 1:10 ratio was used to recover iron, titanium, aluminum, scandium, and gallium from the red mud. The tailings were leached with a sulphuric solution at 80 °C for 2 h [23]. The leaching test results show the partial or total extraction of the metals in the bauxite sample. They maintained a concentration of 0.5 M for sulfuric acid with an S/L ratio of 1:10 at a temperature of 90 °C; 6% of aluminum and 18% of iron were leached. At an S/L ratio of 1:5, 9% aluminum and 15% iron were extracted. It was observed that temperature was the parameter that influenced the increase in the extraction rate of metals, regardless of the S/L ratio, since at 25 °C and 50 °C, the leaching percentage remained at 5% only for iron, while aluminum was not leached. The leaching of lithium and aluminum from a bauxite, which also originates from China, showed that increasing the temperature from 80 to 100 °C resulted in a change in the leaching percentages of Al and Li from 75.78% and 73.68% to 88.64% and 96.35%, respectively [24]. Increasing the temperature causes the molecules to have more energy, significantly improving the leaching rate. Based on the results observed, a temperature of 90 °C and a solid: liquid ratio of 1:10 were selected for the subsequent tests. Variation in Concentration (M) Extraction efficiency is generally achieved with high acid concentrations. Iron is often co-extracted with the elements of interest in acid-leaching processes, with an increase in concentration [25]. Sulphuric acid is one of the best leaching agents. However, some disadvantages exist, such as silica gel formation [26]. At a pH lower than 7, soluble silica is found as orthosilicic acid H4SiO4. These monomers bind together and form polysilicic acid, forming a jelly [24].

Development of a Hydrometallurgical Process to Obtain High-Purity Alumina Using Bauxite

93

Fig. 1 Effect of sulfuric acid concentration (90 °C, 1 h and 1/10 S/L) and Effect of hydrochloric acid concentration (90 °C, 1 h and 1/10 S/L)

The results obtained in the tests with sulfuric acid (Fig. 1) from 0.5 to 3 M were 12% to 97% extraction for iron and 10% to 68% for aluminum, respectively. However, when the acid concentration was increased to 4.5 and 6 M, it was impossible to obtain a solution due to the formation of silica gel. Hydrochloric acid can leach more iron (III), while sulfuric acid removes more representative aluminum concentrations because ferrous sulfate, the intermediate product in the reaction, is more stable. For this reason, aluminum is leached in more representative concentrations. Hydrochloric acid is responsible for producing concentrations of FeCl3, a stable product in water [27]. At a concentration of 0.5 M with hydrochloric acid (Fig. 1), around 15% of the aluminum and iron were leached out, while at a concentration of 1.5 M, 85% of the iron was removed and 25% of the aluminum. By increasing the acid concentration to 6 M, the iron extraction rate, which was 96%, increased to 97% at 4.5 M and remained constant throughout the rest of the tests. However, for aluminum, a significant increase in the extraction rate was observed, from 38% at 3 M to 53% at 6 M (Fig. 1). These results showed that hydrochloric acid is an exciting leaching agent for

removing iron. It can be concluded from these tests that concentration is a relevant factor in leaching processes.

Time Variation Sulphuric acid was selected at a concentration of 3 M (Fig. 2) for the time variation test. It was possible to assess that the leaching rate for iron in 1 h was 96%, and when the time was increased to 4 h, the leaching rate was 98.3%. For aluminum, the leaching rate increased from 68% in 1 h to 99.2% in 4 h, with only 3% of silicon and 0.5% of titanium leached. H2SO4 became a viable alternative for the following stages of the project due to the high aluminum recovery rate. Hydrochloric acid (Fig. 2) proved to be a potential leaching agent for iron. A working concentration of 3 M was selected. In 1 h, the iron leaching rate was 96%; in 4 h, it was 98%, but the same effect was not observed for aluminum, whose maximum removal was 38%. The silicon content reached 6% in 4 h. For the following stages of the work, solvent extraction, precipitation, and calcination, the liquor obtained from leaching with 3 M sulfuric acid, 4 h, 90 °C, and an S/L ratio of 1/10 was selected as the optimum condition for the leaching stage.

Fig. 2 Effect of time with sulfuric acid (3 M, 90 °C and 1/10 S/L) and effect of time with hydrochloric acid (3 M, 90 °C and 1/10 S/L)

94

B. da Rocha Pereira et al.

Solvent Extraction

100

100

D2EHPA Alamine 336

80 60 40 20 0

-1

Studies on the Variation of Extractant Concentration (v/v) The extractant concentration was varied to check the extraction percentage for iron and aluminum, and the pH varied from −0.5 to 1.5. The volume/volume variations studied were 10%, 15%, 20%, and 25%. The percentage of extraction for both extractants can be observed in Fig. 4 for iron as the volume/volume ratios were varied. The extraction of iron using the Aliquot 336 extractant (quaternary ammonium salt), when there was an increase in the concentration of the extractant from 5 v % to 15 v %, the extraction efficiency went from 5.2% to 4.2% [32]. Higher extractant concentrations are favorable for metals such as calcium and sodium, and it can be related that the extraction behavior of iron with Alamine 336 is due to the difference in the concentration of elements in the solution, in which aluminum has 34 g/L, while for iron it has 2.6 g/L [33]. However, it is co-extracted when the iron is present in the same solution. The concentrations of metals in the solution and the selectivity of the extractants have a significant impact on the purification process, particularly with the use

Percentage of extraction Al(%)

Percentage of extraction Fe(%)

Effect of pH and Extractants During the solvent extraction process, pH influences the release or exchange of hydrogen ions in the extractant composition, thus facilitating the extraction of metals [12]. Therefore, the increase in pH provides different extraction orders for different metal cations [10]. In the review by Shakibania et al. [21], D2EHPA showed the ability to extract various metals, such as vanadium, nickel, and aluminum. This extractant showed selectivity for iron in a solution with high concentrations of vanadium. The extraction of iron with the 10% v/v D2EHPA extractant was 55% at pH 2 [20]. The authors studied the separation of vanadium and iron in a solution containing both elements. They used organophosphate extractants such as D2EHPA and Cyanex 272. It was found that increasing the pH range from 0.5 to 2.0 resulted in 100% extraction of iron using both extractants [28]. When testing the effect of pH, using D2EHPA, around 80% of the iron was extracted at pH 1, while at lower pH values, the iron removal rate remained around 40% [29]. Alamine 336 showed considerable performance in acidic pH solutions, extracting 90% of the iron and around 15% of the aluminum as the pH increased from −0.5 to 1.5. In a liquor produced through acid leaching of bauxite residue, the solvent extraction technique was used to separate the impurities and recover the scandium present in this liquor, iron being one of the impurities. Evaluating the effect of pH from 0.5 to 2.0, it was reported that only 20% of the iron was removed using the extractant Alamine 336 [30]. Amine-based extractants have a two-step reaction mechanism, the first in the form of protonation and the second in the form of ion exchange [31].

Figure 3 shows the tests using the D2EHPA extractant. At pH −0.5, 30% of the iron was extracted, while at pH 1.5, 62% was extracted. Aluminum was co-extracted from 10% at pH = −0.5 to 52% at pH = 1.5. Fe extraction remained at around 90% at all the pHs studied for Alamine 336, while aluminum removal was 15% at pH −0.5 and 0 and 25% at pHs 0.5, 1.0, and 1.5. In addition, the pH that showed the best results is close to the pH of the leaching solution (pH −0.5). Therefore, adding impurities to adjust the pH, such as sodium hydroxide, is unnecessary. This test has therefore shown an exciting alternative. However, the percentage of extracted aluminum still needs to be improved.

0

1

2

D2EHPA Alamine 336

80 60 40 20 0

-1

0

pH Fig. 3 Effect of pH on extraction by D2EHPA and Alamine 336 (15 min, 10% v/v, 1:1 A/O, 25 °C)

1 pH

2

Development of a Hydrometallurgical Process to Obtain High-Purity Alumina Using Bauxite

95

Fig. 4 The impact of increased concentrations of extractants Alamine 336 and D2EHPA on the efficiency of iron extraction in the bauxite sample as a function of pH (25 °C, 15 min, 1:1 A/O)

of the extractant Alamine. At concentrations of 10% v/v, about 5% is extracted. When the value rises to 20%, the rate drops to 4% [31]. Therefore, increasing the v/v ratio can be beneficial for some extractants, such as D2EHPA 20 and 25% (Fig. 4), while the Alamine 336 (Fig. 4) extractant did not show the same working dynamics. For the D2EHPA extractant, it was noted that, for iron, the 10% v/v concentration and the pH −0.5–1.5 variation resulted in an extraction rate of 30% to 50%, as shown in Fig. 4. When the concentration was increased to 15% v/v, the extraction rate remained at 70% at all pH values. At 20% v/v, the extraction efficiency increased linearly from 40% (pH −0.5) to 90% (pH 1.5). At a v/v ratio of 25%, the extraction rate was 80% at pH values −0.5 to 1 and reached 100% at pH 1.5. It can be seen that aluminum was co-extracted by increasing the ratio from 10 to 25%, reaching almost 40% at pH 1.5 in the 25% v/v ratio. Therefore, the 25% v/v concentration was selected for the temperature variation tests. Concerning the Alamine 336 extractant, iron extraction (Fig. 4) became lower as the volume/volume ratio increased. Iron extraction at a volume/volume ratio of 10% was 90% at pH 1.5. At ratios of 15% and 20% v/v, it was 64% and fell to

34% when working with 25% v/v. Concerning aluminum extraction, in the 10% v/v ratio, the percentage was only 15% at pH −0.5, but when the other ratios were analyzed, 8%, 10%, and 13% were obtained for the 25%, 20%, and 15% volume ratios, respectively. When the pH value was raised to 1.5, the extraction percentages were 43% for the 25% v/v ratio, 40% for 20% v/v, and 30% for 15% v/v. This extractant became a potential for further studies, as the 10% v/v ratio extracted around 90% of the iron and 15% of the aluminum.

Studies on Temperature Variation (°C) The temperature tests were conducted at the 25% v/v concentration as defined in the previous tests. Increasing the temperature favored the extraction of iron by D2EHPA, as shown in Fig. 5. However, as the pH increased to 0.5, the iron extraction efficiency dropped from 100 to 90% and remained stagnant until pH 1.5. Increasing the temperature from 25 to 50 °C decreased co-extraction for aluminum, except for pH 1.5. The Alamine 336 extractant was not favored by increasing the temperature to 50 °C. Extraction

B. da Rocha Pereira et al.

100

Alamine 336 10% 50°C D2EHPA 25%-50°C

Percentage of extraction Al (%)

Percentage of extraction Fe(%)

96

100

75

50

25

0 -1

0

1

Alamine 336 10% 50°C D2EHPA 25%-50°C

80 60 40 20 0 -1

2

0

1

2

pH

pH Fig. 5 Effect of temperature by D2EHPA and Alamine 336 (50 °C–15 min)

rates remained lower than 20% for iron and 10% for aluminum at all pH values. D2EHPA was used in a synthetic solution of nickel, calcium, iron, copper, cobalt, zinc, and manganese. When evaluating the effect of the temperature of 25 and 50 °C, the rate of iron extraction increased [34]. A sulfuric leach liquor for separating uranium and iron using Alamine 336 dissolved in kerosene, in a study, a temperature range of 25–50 °C was applied. Thermodynamic studies showed that the extraction of iron is an endothermic reaction. When comparing the results for Alamine 336, it was noted that increasing the temperature did not increase the Fe [35]. From the tests carried out, it was found that the parameters that resulted in the removal of significant percentages of iron from the liquor produced from the leaching were the extractant D2EHPA pH -0.5, v/v ratio 25%, and temperature of 50 °C.

Precipitation and Calcination to Obtain Alumina For the precipitation tests, with the use of ammonium hydroxide from the solvent extraction liquor with D2EHPA, Table 3 Percentage of aluminum and iron precipitation with ammonium hydroxide at 25 °C for 1 and 2 h and pH variation from 3.0 to 4.5 for the D2EHPA (pH −0.5, v/v 25%, test liquor at 50 °C)

the aluminum in the solution was easily precipitated, as shown in Table 3. The pH 3 test for 2 h was selected for the calcination, as it precipitated 100% of the aluminum and only 50% of the iron in the solution, keeping the other 50%. After the production of alumina, it was sent for analysis via SEM–EDS to verify the obtained structure and to determine which elements could be quantified through EDS. It was observed that only aluminum and iron were present. The structure of the alumina produced showed agglomerates of particles with different dimensions and surface morphology similar to pores (Fig. 6) [35]. The X-ray fluorescence shows only the presence of aluminum and oxygen. The purity of the alumina obtained at the end of the process was 99.99%, with 5 ppm of iron, 3 ppm of silicon, and 4 ppm of potassium as impurities. The efficiency of the process for obtaining high-purity alumina is 87.4% of the liquor obtained from the leaching stage followed by the purification stage, solvent extraction followed by precipitation, and calcination. The Bayer process produces 98.7% alumina from bauxite, generating highly alkaline waste (red mud). In order to produce HPA, it is necessary to incorporate other purification methods, such as washing the alumina with concentrated hydrochloric acid. The process developed

Tests

Al 25 °C (%)

Fe 25 °C (%)

1 h pH 4.5

100

85

1 h pH 4.0

85

80

1 h pH 3.0

80

100

2 h pH 4.5

100

100

2 h pH 4.0

95

65

2 h pH 3.0

100

50

Development of a Hydrometallurgical Process to Obtain High-Purity Alumina Using Bauxite

97

was selected because the concentrations recovered were 100% for iron and aluminum, while the other impurities remained in the cake. The solvent extraction tests were essential for the liquor purification process. The D2EHPA extractant removed 18% of the aluminum and 98.7% of the iron. Ammonium hydroxide was the precipitating agent that managed to precipitate 100% of the aluminum available in the solution and was also used because it is eliminated in calcination, generating no impurities for the process. With the studies, it was possible to develop a hydrometallurgical route that produced alumina with a purity of around 99.99% from a bauxite sample.

References

Fig. 6 SEM of alumina obtained (Mag. 9200X)

during this research results in obtaining high-purity alumina through a hydrometallurgical route, with the advantage of reducing the amount of generated waste. This approach represents a viable alternative that can be applied to the waste generated in the bauxite mining process, which is classified as a raw material at risk of supply according to the guidelines of the European Union. The experiments were conducted on a laboratory scale, and so far, large-scale tests have not been carried out to evaluate the efficiency of the reactors and other equipment to be employed. For this reason, the route becomes interesting because it does not generate the red mud residue. The only remaining waste generated in the proposed trajectory involves silicon and titanium, which can undergo leaching processes to ensure the complete utilization of all the material. In addition it has only a few process steps, unlike the Bayer process, which excessively consumes sodium hydroxide, due to the reactive silica present in bauxite. It is not necessary to use other acid purification steps to ensure the purity of the obtained alumina.

Conclusions Acid leaching is an exciting process for recovering metals. The parameter that most influenced the leaching was the concentration of the acids. Increasing the concentration did not always result in higher extraction percentages due to the properties of the acids. The sulfuric acid leaching condition

1. Ruys A. Bauxite: The principal aluminum ore. In: Alumina Ceramics. Elsevier; 2019. p. 39–47. 2. Santos RM dos, Magalhães R da S, Sobrinho N de O, Gomes ÉR. Incorporação de lama vermelha na indústria cerâmica: uma revisão. Res Soc Dev. 2021;10(10):e321101018949. 3. Ruys A. Refining of alumina: The Bayer process. In: Alumina Ceramics. 2019. p. 49–70. 4. Smith P, Power G. High Purity Alumina–Current and Future Production. Miner Process Extr Metall Rev. 2021;13:747–56. 5. Ruys A. Introduction to alumina ceramics 1.1. Alumina Ce. 2019. 37 p. 6. Siscomex. Dados compilados sobre Drawback suspensão e insenção. 2022. p. 49. 7. Abdulvaliyev R, Dyussenova S, Manapova A, Ata A, Beisenbiyeva U. Modification of the phase composition of low-grade gibbsite-kaolinite bauxites. Complex Use Miner Resour. 2021;317:94–102. 8. Tabereaux AT, Peterson RD. Aluminum Production. Vol. 3, Treatise on Process Metallurgy, Volume 3: Industrial Processes. U. S.A: Elsevier Ltd.; 2014. 839–917 p. 9. European Union. Study on the Critical Raw Materials for the EU. 2023. 160 p. 10. Santanilla AJ, Aliprandini P, Benvenuti J, Alberto J, Tenorio S, Crocce D, et al. Structure investigation for nickel and cobalt complexes formed during solvent extraction with the extractants Cyanex 272, Versatic 10 and their mixtures. Miner Eng. 2021;160:106691. 11. Dissanayake DMSN, Mantilaka MMMGPG, Silva RT De, Silva KMN De, Pitawala HMTGA. Laterite and its potential as an alternative-bauxite. Clean Mater. 2021;1(August):100016. 12. Zhao A. Study on Extraction and Separation Performance of Iron and Aluminum from Acid Leaching Solution of High-Iron Bauxite. 2022;63(1):37–45. 13. Zhuk AZ, Vlaskin MS. Synthesis of high-purity aluminum oxide from hydrothermally produced boehmite by high temperature vacuum treatment. Mater Today Proc. 2017;4(11):11580–7. 14. Valeev D, Pankratov D, Shoppert A, Sokolov A, Kasikov A, Mikhailova A, et al. Mechanism and kinetics of iron extraction from high silica boehmite − kaolinite bauxite by hydrochloric acid leaching. Trans Nonferrous Met Soc China. 2021;31(10):3128–49. 15. Kim H, Moon G, Choi I, Lee J, Kumar R. Hydrometallurgical process development for the extraction , separation and recovery of vanadium from spent desulfurization catalyst bio-leach liquors. J Clean Prod. 2018;187(March):10.

98 16. Kyriakogona K, Giannopoulou I, Panias D. Extraction of aluminium from Kaolin: A comparative study of hydrometallurgical processes. Proc World Congr Mech Chem Mater Eng. 2017;133:2–7. 17. Valeev D, Pankratov D, Shoppert A, Sokolov A, Kasikov A, Mikhailova A, et al. Mechanism and kinetics of iron extraction from high silica boehmite-kaolinite bauxite by hydrochloric acid leaching. Trans Nonferrous Met Soc China. 2021;31:3128–49. 18. Zhang K, Qiu L, Tao J, Zhong X, Lin Z. Recovery of gallium from leach solutions of zinc refinery residues by stepwise solvent extraction with N235 and Cyanex 272. Hydrometallurgy. 2021;205(August):105722. 19. Zhao Y, Zheng Y, He H, Sun Z, Li A. Effective aluminum extraction using pressure leaching of bauxite reaction residue from coagulant industry and leaching kinetics study. 2021;9(November 2020). 20. Aliprandini P. O uso da extração por solventes para tratamento de licor de lixiviação de minério limonítico de níquel. Universidade de São Paulo; 2017. 21. Shakibania S, Mahmoudi A, Mokmeli M. Separation of vanadium and iron from the steelmaking slag convertor using Aliquat 336 and D2EHPA : Effect of the aqueous species and the extractant type. Miner Eng. 2022;181(March):107521. 22. Botelho ABJ, Crocce DE, Tenório JAS. Selective separation of Sc (III) and Zr (IV) from the leaching of bauxite residue using trialkylphosphine acids , tertiary amine , tri-butyl phosphate and their mixtures. Sep Purif Technol. 2021;279(September):13. 23. Wu Y, Li W, Vovers J, Lu HT, Stevens GW, Mumford KA. Investigation of green solvents for the extraction of phenol and natural alkaloids : Solvent and extractant selection. Chem Eng J Adv. 2022;442(February):11. 24. Zhang J, Wang Q, Liu X, Zhou G, Xu H, Zhu Y. Provenance and ore-forming process of Permian lithium-rich bauxite in central Yunnan , SW China. Ore Geol Rev. 2022;145(October 2021):104862. 25. Pepper RA, Perenlei G, Martens WN, Couperthwaite SJ. High purity alumina synthesised from iron rich clay through a novel and

B. da Rocha Pereira et al.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

selective hybrid ammonium alum process. Hydrometallurgy. 2021;204(August):105728. Agrawal S, Dhawan N. Hydrometallurgy Microwave acid baking of red mud for extraction of titanium and scandium values. Hydrometallurgy. 2021;204(July):105704. Anawati J, Azimi G. Recovery of scandium from Canadian bauxite residue utilizing acid baking followed by water leaching. Waste Manag. 2019;95:549–59. Alkan G, Yagmurlu B, Gronen L, Dittrich C, Ma Y, Stopic S, et al. Selective silica gel free scandium extraction from Iron-depleted red mud slags by dry digestion. Hydrometallurgy. 2019;185 (March):266–72. Zhou L, Gou M, Luo S. Hydration kinetics of a calcination activated bauxite tailings-lime-gypsum ternary system. J Build Eng. 2021;38(November 2020):102189. Tavakoli MR, Dreisinger DB. Separation of vanadium from iron by solvent extraction using acidic and neutral organophosporus extractants. Hydrometallurgy. 2014;141:17–23. Jantunen N, Kauppinen T, Salminen J, Virolainen S, Lassi U. Separation of zinc and iron from secondary manganese sulfate leachate by solvent extraction. Miner Eng. 2021;173 (May):107200. Botelho Junior AB, Espinosa DCR, Tenório JAS. Characterization of Bauxite Residue from a Press Filter System: Comparative Study and Challenges for Scandium Extraction. Mining, Metall Explor. 2021;38(1):161–76. Kumar JR, Shin SM, Yoon HS, Nam CW, Chung KW, Lee J, et al. Separation and Recovery of Vanadium from Synthetic Leach Liquor Solutions Containing Iron, Calcium, Sodium, Aluminum, and Manganese by the Solvent Extraction Technique. Sep Sci Technol. 2014;6395:11. Wang Y, He Y, Yin S, Long H, Li S. Research on extraction of zinc from spent pickling solution using Aliquat 336. Hydrometallurgy. 2020;193(may):7. Tsakiridis PE. Solvent extraction of aluminium in the presence of cobalt , nickel and magnesium from sulphate solutions by Cyanex 272. Hydrometallurgy. 2005;80:90–7.

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud He Xin, Lv Guo-zhi, Zhang Ting-an, Wang Song, and Wang Long

Abstract

The Bayer process is the primary technology utilized for alumina production but it is limited by its stringent requirements on the quality of the bauxite raw material. Due to the absence of a reasonable and effective large-scale utilization method, a substantial amount of high-alkali red mud generated from the process is kept in storage yards, occupying significant land space and posing grave threats to both the environment and human health. Based on prior research, this paper comprehensively summarizes and scrutinizes the current state of alumina production as well as the management of red mud. Commencing with the phase structure of alumina production, this paper delves into the development of the Bayer process and the Calcification-carbonization method. The aforementioned approaches offer pioneering concepts for promoting sustainable development within the alumina industry. Keywords

Bayer process Bayer process








Calcification-carbonization method Red mud Hydrometallurgy




Introduction Bayer red mud is an insoluble alkaline residue that results from the manufacture of alumina during the Bayer alkaline leaching process. Red mud is a waste material consisting H. Xin
L. Guo-zhi (&)
Z. Ting-an
W. Song
W. Long Northeastern University, Key Laboratory of Ecological Metallurgy of Multimetal Intergrown Ores of Ministry of Education, Special Metallurgy and Process Engineering Institute, Shenyang, 110819, Liaoning, China e-mail: [email protected] Z. Ting-an e-mail: [email protected]

largely of iron oxide and other insoluble substances. Although commonly red due to its high Fe2O3 content, the color can vary from brown to grayish-white depending on specific iron oxide concentrations [1]. The chemical composition of red mud is not uniform and is influenced by the varying chemical makeup of bauxite and different production processes. Typically, red mud generated through the Bayer process contains Al2O3 (8˗30%), Fe2O3 (6–72%), SiO2 (5–30%), Na2O (2–10%), CaO (0–23%), and TiO2 (4– 25%). Additionally, bauxite ores from diverse regions introduce elements such as potassium, magnesium, arsenic, chromium, and a range of rare earth elements. A comprehensive list of the main components of Bayer process red mud, categorized by country of origin, is presented in Table 1. Statistically, for every ton of alumina produced, between 1.5 and 2.5 tons of red mud is generated. Figure 1 illustrates the discharge and comprehensive utilization quantity of red mud in China from 2011 to 2022. By 2022, the global stockpile of red mud will surpass 5 billion tons. Notably, China's stockpile alone accounted for 1.6 billion tons, approximately one-third of the global total. The annual discharge of red mud is increasing by approximately 100 million tons annually [8]. Historically, coastal countries opted for sea dumping to dispose of red mud. However, with the tightening of global environmental regulations in recent years, this method has been explicitly banned. Presently, dam construction serves as the primary storage solution for untreated red mud. Red mud is a problematic waste material that poses a two-fold challenge: it consumes large swathes of land while requiring extensive and costly storage sites. In addition, the highly alkaline red mud waste liquid can pollute groundwater and increase environmental pressure, leading to water and soil alkalization [9]. A red mud storage dam breach occurred on October 4, 2010 at the Ajka alumina plant in Hungary, resulting in the leakage of an estimated 1 million cubic meters of red mud [10]. This event caused extensive

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_12

99

100 Table 1 Main components of Bayer red mud in different regions of the world (wt.%)

H. Xin et al. Country/region

Alumina plant

Al2O3

Fe2O3

Na2O

CaO

China/Guangxi

Pingguo

19.10

9.18

32.20

4.38

14.02

9.39

Australia [2]

Pinjarra

17.10

23.80

36.20

1.60

3.90

3.90

Australia [3]

Worsely

15.60

3.00

56.90

2.20

2.39

4.46

SiO2

TiO2

Brazil/Balkarene

Alunorte

15.1

15.6

45.6

7.5

1.16

4.29

Espana [4]

Alcoa

21.2

4.4

37.5

3.6

5.51

11.45

United States/Arkansas

Alcoa

12.15

4.5

55.6

2*5

–

4.5

Italy [5]

Eurallumina

17.91

9.58

30.45

12.06

7.77

8.61

Jamaica [6]

ALPART

14.2

3.4

50.9

3.18

–

6.87

Turkey [7]

Seydisehir

20.4

15.74

36.9

10.1

–

4.9

our laboratory. It evaluates the treatment costs and economic benefits of the calcification-carbonization method, providing new insights for promoting the sustainable development of the alumina industry.

Alumina Production and Red Mud Generation Bayer Process

Fig. 1 Emissions and comprehensive utilization of red mud in China from 2011 to 2020

damage to surface water and land, leading to the most severe safety accident in the history of the aluminum industry and prompting widespread panic across several European countries. This catastrophe also served as a clarion call for the Chinese aluminum oxide industry, compelling relevant national departments to prioritize the issue of solid waste storage. The “Twelfth Five-Year Plan for the Comprehensive Utilization of Bulk Industrial Solid Waste,” published in 2012, identified the comprehensive utilization of red mud as one of the top ten key projects during the “Twelfth FiveYear Plan” period. This paper provides a systematic study of the current status of alumina production and the reasons for red mud generation from the perspective of phase transformation. Furthermore, it discusses the current state of clean production processes for alumina and the harmless utilization of red mud. Finally, focusing on the calcification-carbonization process, this paper summarizes the process parameters and experimental results obtained by numerous researchers in

In 1889, Bayer discovered that sodium aluminate solution, from which most of the aluminum hydroxide had precipitated, could selectively dissolve the alumina contained in bauxite when bauxite was pressurized and heated in an autoclave. This process is known as the leaching of bauxite, and by alternating leaching and precipitation processes, bauxite can be continuously processed to produce aluminum hydroxide products, forming the Bayer cycle [11]. This process is characterized by its simplicity, low energy consumption, and high product quality. It is economically more efficient for processing high-grade bauxite compared to other methods [12]. The reaction principle is illustrated in Eq. 1 (forward reaction for dissolution and reverse reaction for precipitation). Al2 O3
xH2 O þ 2NaOH þ ð3  xÞH2 O þ aq $ 2NaAlðOHÞ4 þ aq

ð1Þ

The Bayer process, as illustrated in Fig. 2 can be divided into several main steps: ore slurry preparation, high-pressure leaching, dilution of the leached slurry, separation and washing of red mud, decomposition of seed crystals, classification and washing of aluminum hydroxide, calcification of aluminum hydroxide, evaporation of the mother liquor, and causticization [13]. The Bayer process is characterized by its simplicity, a relatively small number of production steps, and its suitability for large-scale equipment and process automation. It offers energy efficiency advantages when processing bauxite with good leaching performance.

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud

101

Fig. 2 Typical flow chart of the Bayer process

However, the Bayer process requires high-grade bauxite, and it is particularly beneficial when processing bauxite with a high aluminum-to-silicon ratio (A/S > 8). In the Bayer process, a portion of the alumina in bauxite reacts with silica and caustic soda to form sodium aluminosilicate hydrate (Na2O
Al2O3
1.7SiO2
n H2O, n  2), which enters the red mud. The reaction is represented by Eq. 2, where the mass ratio of alumina, silica, and sodium oxide is 1:1:0.608. This means that for every 1 kg of silica in bauxite, there is a loss of 1 kg of alumina and 0.608 kg of

sodium oxide. Therefore, when the grade of bauxite decreases, that is, when the aluminum-to-silicon ratio decreases, the economic efficiency of the Bayer process is significantly affected. The influence of the aluminumto-silicon ratio in bauxite and red mud on the alumina dissolution rate can be illustrated by plotting the data obtained from the alumina dissolution calculation formula (Eq. 3). The impact of the aluminum-to-silicon ratios (A/S)O and (A/S)R on the alumina dissolution rate is shown in Fig. 3. Here, (A/S)O represents the original aluminum-to-silicon

102

H. Xin et al.

Fig. 3 Effect of A/S between raw ore and red mud on the actual yield of alumina produced by Bayer process

ratio in bauxite, and (A/S)R represents the aluminumto-silicon ratio in red mud. From the graph, it can be observed that the alumina dissolution rate reaches 80% only when the aluminum-to-silicon ratio in bauxite is greater than 5, considering that the minimum aluminum-to-silicon ratio in red mud is theoretically 1. However, due to practical conditions in alumina production, the actual aluminumto-silicon ratio in red mud is usually greater than 1. If the aluminum-to-silicon ratio in red mud is 1.4, the aluminumto-silicon ratio in the original bauxite needs to be greater than 7 to achieve an 80% alumina dissolution rate. On the other hand, if the aluminum-to-silicon ratio in red mud is reduced to 0.4, an aluminum-to-silicon ratio of 2 in bauxite is sufficient to achieve an 80% dissolution rate. Therefore, there are two ways to improve the alumina dissolution rate: increasing the aluminum-to-silicon ratio in bauxite and reducing the aluminum-to-silicon ratio in red mud. Al2 O3
2SiO2
2H2 O þ mNaOH ¼ NaAlðOHÞ4 þ Na2 O
Al2 O3
1:7SiO2
nH2 O gAl ¼

ðA=SÞO ðA=SÞR  100% ðA=SÞO

ð2Þ ð3Þ

Based on the aforementioned ideas, the current research progress in improving the alumina dissolution rate in the Bayer process mainly focuses on the following two directions:

1. Enhancing the aluminum-to-silicon ratio in bauxite through physicochemical methods to achieve predesilication before dissolution. 2. Exploring ways to alter the structure of red mud to decrease the aluminum-to-silicon ratio and sodium-to-silicon ratio.

Enhancing the Aluminum-to-Silicon Ratio of Bauxite Ore Improving the aluminum-to-silicon ratio is a key objective in bauxite ore processing to enhance the quality and efficiency of alumina extraction. At present, there are some methods to improve the aluminum-silicon ratio of bauxite, such as ore beneficiation methods, chemical desiliconization methods and biological desiliconization methods for the predesiliconization of raw ore to improve the grade.

Ore Beneficiation Methods Ore beneficiation methods primarily exploit the differences in physical properties (particle size, density) and chemical properties (selectivity of beneficiation reagents) between aluminum- and silicon-bearing minerals in bauxite ore. Several methods, including screening, gravity separation, and flotation, are used to achieve the separation of aluminum and silicon minerals.

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud

The application of screening methods depends largely on the microscopic mineralogy of the ore. Rio Tinto discovered that the bauxite ore in the Weipa region of Australia exists in the form of bean-shaped particles, with decreasing silica content as the particle size increases. Therefore, the company's beneficiation plant crushes the mined bauxite ore to below 19 mm and performs wet screening using a 10-mesh sieve, with the tailings returned to the mining area for backfilling to avoid adverse effects on the subsequent processing caused by SiO2. If there is no significant difference in particle size between the target minerals, separation cannot be achieved through screening [14]. Gravity separation exploits the density differences between the target ore minerals for separation. Currently, it is mainly applied to certain local mono-hydrated alumina ores due to their simple relationship between aluminum and silicon minerals, allowing separation through gravity methods. Conversely, trihydrated gibbsite ores have a tighter association between aluminum and silicon minerals, making effective separation unachievable. Flotation relies on the selectivity of beneficiation reagents to achieve desilication in bauxite ore. Since most added beneficiation reagents are high-molecular-weight organic compounds, some may enter the subsequent production process with the concentrate. After a certain enrichment level, they can affect the settling characteristics of red mud, evaporation of the mother liquor, seed decomposition rate, and product granularity [15].

Chemical Desiliconization In the 1940s, the Lauta plant in Germany proposed a chemical desilication method for processing high-silica bauxite ores in Yugoslavia, Austria, and other regions [16]. This desilication method is particularly effective for bauxite ores with fine-grained intergrowth or close association between kaolinite and bauxite, which are difficult to separate. The main process of this method includes preroasting, alkali leaching desilication, and solid‒liquid separation. The addition of thermal activation in the chemical desilication process led to an improved recovery rate of aluminum oxide. This process also effectively removes impurities such as organic matter, sulfur, and carbonates from the ore. However, it should be noted that the calcification process significantly increases the energy consumption of the overall process. Biological Desiliconization The bioleaching method primarily utilizes heterotrophic bacteria to decompose silicates or aluminosilicates in bauxite ores. Through bacterial decomposition, SiO2 is transformed into soluble substances, while Al2O3 remains insoluble [17]. For instance, silicate bacteria can breakdown kaolinite into silica dioxide and aluminum oxide. Compared to other

103

methods, this approach can achieve favorable economic and technical indicators while simultaneously avoiding or minimizing environmental pollution [18]. Bioleaching methods are more suitable for treating gelatinous and ultrafine bauxite ores. However, this method faces challenges such as a slow bacterial leaching rate, long process cycles, stringent environmental requirements, low productivity, and technical issues related to the removal and degradation of heterotrophic bacteria. Therefore, there is still a considerable gap between this method and its widespread industrial application.

Modifying the Balanced Structure of Red Mud Lime Bayer Process The Lime Bayer process is a new aluminum oxide production technology developed by the Zhengzhou Research Institute of Aluminum Corporation of China Limited and the Shenyang Aluminum and Magnesium Design and Research Institute, specifically tailored to the characteristics of China's bauxite resources. The main principle of the lime Bayer process is to add a significant amount of lime during the high-temperature Bayer leaching process, transforming the balanced structure of red mud from sodium aluminosilicate hydrate in the form of sodalite to calcium aluminosilicate hydrate in the form of hydrous garnet. This transformation eliminates alkali from the leaching desilication product, resulting in a substantial reduction in alkali consumption. The Lime Bayer process can utilize the existing Bayer process system with minimal changes to the process equipment. It only requires an increase in lime dosage to treat medium-grade bauxite, producing alumina with lower alkali consumption and energy consumption compared to the conventional Bayer process. Therefore, this process has low investment requirements and is relatively simple to implement. Currently, the Shanxi Branch of Aluminum Corporation of China Limited has completed the construction of a three-phase Lime Bayer process production line with a capacity of 800,000 tons. The experimental results from July 2002, which lasted for 119 h, showed that when using bauxite with an aluminum-to-silicon ratio of 6.5 to 7, the aluminum-to-silicon ratio of the red mud after leaching was 1.7, and the sodium-to-silicon ratio was 0.26, the actual recovery rate of alumina is over 70%. These technical indicators were close to the design specifications. High-Pressure Hydrothermal Method The high-pressure hydrothermal method involves leaching under high temperature (above 280 °C), high alkaline concentration (mother liquor alkali concentration above 400 g/L of Na2O), and high molecular ratio (mother liquor molecular ratio of Na2O/Al2O3 was 30–35). The leaching solution is

104

evaporated to a sodium oxide concentration of approximately 500 g/L, and hydrated sodium aluminate is crystallized out. This is then dissolved to obtain a sodium aluminate solution with a lower molecular ratio, which can be used to produce aluminum hydroxide through seeding and precipitation. Silicon in the red mud exists in the form of NaCa(HSiO4), which can be hydrolyzed to recover Na2O. The equilibrium solid phase of the red mud is CaO
SiO2
H2O, with an A/S ratio of less than 0.5 and a Na2O mass fraction of less than 1%. This method does not cause the loss of Al2O3 and Na2O, but the evaporation process of the leaching solution consumes a considerable amount of energy, with an evaporated water quantity of 18 t/t Al2O3. Meanwhile, the hydrolysis and Na2O recovery process of NaCa(HSiO4) is complex and involves high temperature and alkaline conditions, making the overall process intricate.

The Sub-molten Salt Method The sub-molten salt method is employed for processing low-grade bauxite, utilizing a high-concentration alkali solution (Na2O > 600 g/L), temperatures ranging from 150 to 180 °C, and near atmospheric pressure. This technique yields a leaching solution with a high molecular ratio, along with red mud primarily composed of hydrated sodium silicoaluminate. The leaching solution is diluted, subjected to liquid‒solid separation, and subsequently evaporated for the crystallization and precipitation of solid hydrated sodium aluminate. This solid is then dissolved to acquire a sodium aluminate solution with a lower molecular ratio, serving as a precursor for the production of Al(OH)3 through seeding and precipitation. The separated red mud is treated with hot water washing and lime to remove alkali, resulting in the formation of hydrated garnet as the equilibrium solid phase in the final red mud. Experimental findings indicate that the sub-molten salt method achieves an alumina leaching rate of 96.4% when processing bauxite with an A/S ratio of 8.62. However, the production process of the sub-molten salt method requires a high-alkali concentration of Na2O exceeding 600 g/L, placing high demands on equipment resistance to alkali. Additionally, the leaching solution exhibits high viscosity and surface tension, posing challenges for subsequent liquid‒solid separation. The primary objective of increasing the alumina-to-silica ratio in bauxite is to reduce the amount of silica entering the Bayer process, thereby lowering the consumption of caustic soda and alumina associated with silica. Modifying the structure of the leached residue (red mud) is primarily aimed at altering the bonding of aluminum, silicon, and sodium to achieve a balanced red mud structure with lower alkalinity

H. Xin et al.

and aluminum content, consequently reducing caustic soda and alumina losses.

Harmless Utilization of Red Mud The problem of red mud storage can be addressed by finding a cost-effective method for large-scale resource utilization. The basic principles of comprehensive red mud utilization are to extract valuable metals effectively and utilize red mud as a whole without generating secondary waste [19]. Currently, research on the comprehensive utilization of red mud primarily focuses on three aspects: recovery of valuable elements, preparation of construction materials, and preparation of adsorbent materials. Different processes generate red mud suitable for different comprehensive utilization methods. For example, red mud from the sintering process contains a high content of active components such as 2CaO
SiO2 due to the addition of limestone, making it suitable for cement production. On the other hand, red mud from the Bayer process contains a large amount of adhered and structural alkalis but lacks active components such as 2CaO
SiO2, making it unsuitable for the construction materials industry. Therefore, alternative approaches are needed to achieve comprehensive utilization.

Extraction of Valuable Elements from Red Mud Red mud is a waste residue from the metallurgical industry, and its application in the metallurgical field is an important aspect of comprehensive red mud treatment. The primary application of red mud in the metallurgical industry is the extraction and recovery of valuable components, including iron, aluminum, sodium, and other major valuable elements, as well as trace elements such as titanium, scandium, tantalum, niobium, gallium, vanadium, thorium, and uranium. It is inferred that the recovery of metal elements such as iron, titanium, scandium, gallium, and vanadium from red mud can meet the growing demand for these metals. Given the current situation of gradually depleting mineral resources, the recovery of valuable metals from red mud holds significant value [20]. In recent years, the recovery and extraction of valuable elements from red mud have shown technical feasibility. However, most studies have focused on individual metal elements or elements within the same group, resulting in high costs. Additionally, the remaining residues generated after processing create a new form of solid waste, hindering the achievement of large-scale red mud disposal objectives.

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud

Overall Utilization of Red Mud for Building Materials Both domestic and international practices have demonstrated the significant demand for raw materials in the construction industry. Red mud, with its potential to be used in the production of various construction materials, offers a viable solution for large-scale utilization [21]. However, there are two limiting factors in utilizing red mud for construction material production: high-alkali content and production costs. The high-alkali content in cement can lead to an alkali-aggregate reaction, where alkalis from cement, admixtures, and the environment react with active SiO2 in aggregates (sand and stone), forming alkali silicate gel. This gel can cause concrete to expand and exhibit spider-web-like cracking, leading to structural damage. One possible solution is to use calcium oxide to remove alkalis from red mud, thereby reducing their impact on cement quality. The production costs of using red mud as a cement raw material primarily include red mud treatment and transportation costs. The high moisture content of red mud (50% to 85%) necessitates further dehydration before direct calcification, thereby increasing costs. To mitigate transportation costs, it is advisable to establish cement plants near alumina plants, thereby reducing transportation expenses.

Application of Red Mud in the Environmental Protection Field The application of red mud in the field of environmental protection primarily utilizes its potential to sequester metals and metalloid elements such as arsenic. Both raw and modified red mud can be applied in environmental protection and agriculture. The main characteristics of red mud can be categorized into two types: (1) high alkalinity, which facilitates the hydrolysis or precipitation of metals into hydroxides and carbonates, and (2) a high content of iron, aluminum, and titanium oxides, which provides active surface sites for the adsorption of metals and metalloids. Therefore, red mud holds potential application value for substances and areas contaminated by metals and metalloids [22].

Calcification-Carbonization Process To achieve clean production of aluminum oxide and the harmless utilization of red mud, it is necessary to explore a new method for the low-alkalinity discharge of red mud in the Bayer process for aluminum oxide production. This requires changing the equilibrium phase structure of red mud in the Bayer process, specifically disrupting the balance

105

between sodium, aluminum, and silicon in red mud and seeking a new type of red mud that does not contain sodium and aluminum. Based on this, the Special Metallurgy Team at Northeast University invented a low-cost and large-scale method for the disposal of red mud called the “calcification-carbonization” process. The key steps of the calcification-carbonization process for treating bauxite and red mud include the following: first, calcification is used to transfer all the silicon in bauxite or red mud into hydrated garnet (calcification process), which can remove most of the Na2O from bauxite or red mud; second, CO2 is used to carbonize the calcification-transformed slag (containing hydrated garnet), resulting in carbonized slag mainly composed of calcium silicate, calcium carbonate, and aluminum hydroxide (carbonization transformation); finally, the carbonized slag is subjected to low-temperature aluminum dissolution to obtain a new structure of red mud mainly composed of calcium silicate and calcium carbonate (aluminum dissolution process). The reactions involved in each major process are as follows: (1) Calcification process: Na2 O
Al2 O3
xSiO2
ð6  2xÞH2 O þ 3CaO þ H2 O ! 3CaO
Al2 O3
xSiO2
ð6  2xÞH2 O þ 2NaOH

(2) Carbonization process: 3CaO
Al2 O3
xSiO2
ð6  2xÞH2 O þ ð3  2xÞCO2 ! xCa2 SiO4 þ ð3  2xÞCaCO3 þ 2AlðOHÞ3 þ ð3  2xÞH2 O

(3) Aluminum dissolution process: AlðOHÞ3 þ NaOH ! NaAlðOHÞ4

The processed red mud from the Bayer process primarily consists of calcium silicate (Ca2SiO4) and calcium carbonate (CaCO3) in a new structure. Theoretically, this new structure of red mud does not contain alkaline or aluminum components, making it suitable as a raw material in the cement industry. This solution addresses the fundamental issues of land occupation and environmental pollution caused by red mud, thereby significantly improving the current situation of high solid waste emissions in the aluminum oxide industry. The specific process flow is illustrated in Fig. 4. Drawing upon the fundamental principles underpinning calcification-carbonization treatment technology and processes, subsequent comprehensive and methodical

106

H. Xin et al.

Fig. 4 The flow chart of the calcification-carbonization method [23]

explorations have been undertaken across multiple avenues. These encompass the domains of thermodynamics, kinetics, intrinsic material systems, tangible mineral systems, reactor design, geographically diverse red mud samples, process simulation, scaling strategies, and the sustainable harnessing of calcification-carbonization byproducts.

mud. Zhu Xiaofeng [28] conducted comprehensive studies on the calcification transformation and carbonization decomposition processes of trihydrate bauxite ore and red mud using the calcification-carbonization method. They established the process parameters and successfully conducted scaled-up experiments in a 200 L reactor.

Exploration and Optimization of the Calcification-Carbonization Method

The Reaction Mechanism of the Calcification-Carbonization Method

Pan Lu (2011) employed the calcification-carbonization method to treat low-grade monohydrate bauxite and Bayer red mud, comparing the effects of alkali-carbonation and CO2-carbonation on the extraction efficiency of alumina from low-grade monohydrate bauxite [24]. Guo Fangfang (2013) investigated the influence of various parameters in the calcification and carbonization processes on the treatment effectiveness of Bayer monohydrate bauxite red mud and trihydrate bauxite red mud [25]. Li Qiji (2019) studied the process of alumina production through post-calcium addition and calcification-carbonization, determining the impact of the post-calcium addition process on the rate of alumina dissolution and the sodium content in the red mud [26]. Chen Yang [27] utilized calcium carbide slag as a calcium source in the calcification process, conducting a comparative analysis of the effects of diverse calcium sources, such as calcium oxide and calcium carbide slag, on the calcification-carbonization transformation of Bayer red

The core of the calciumization-carbonization method lies in the conversion of sodium silicate slag into hydrated garnet. To investigate the transformation mechanism, Zheng Chaozhen [29] conducted a study on the synthesis of hydrated garnet using pure substances. This research aimed to understand the generation process and regularities of hydrated garnet and put forth three methods for calculating the silicon saturation coefficient of hydrated garnet. Additionally, an exploration was undertaken to investigate the carbonization stability and decomposition mechanism of hydrated garnet. Liu Guanting [30] conducted a systematic investigation into the reaction mechanisms of carbonization decomposition and low-temperature aluminum dissolution in monohydrate bauxite red mud. The study meticulously analyzed the behavioral patterns of sodium, aluminum, silicon, calcium, and other elements during the carbonization process. Building upon these findings, novel schemes for process optimization were proposed, resulting in an

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud

enhanced utilization rate of aluminum and silicon resources in red mud. Chen Yongchao [31] and Guo Yong [32], focusing on the primary silicon-bearing minerals in bauxite, namely, kaolinite and illite, conducted research on the calciumization transformation mechanisms of these minerals. Li Xiaoqi [33] carried out a comprehensive analysis and study of the thermodynamics, process conditions, and kinetics involved in the calciumization and carbonization processes of Bayer red mud. The study determined the thermodynamic conditions and kinetic control steps for each stage of the investigated processes.

Research and Development of Calcified Carbonization Reactor According to the liquid–solid reaction law of calcification carbonization reaction, the laboratory independently developed a new type of stack tube stirring dissolution reactor and Venturi jet flow carbonization reactor [34] and used the reactor to carry out a 200 L scale expansion experiment. After treatment, the aluminum-silicon ratio of trihydrate ore was 4.24, and the aluminum-silicon ratio of secondary carbonization slag was 0.89. The content of Na2O in the slag was 1.01%. The ratio of Al to Si and the content of Na2O decreased from 1.47 and 10.27 wt% to 0.76 and 0.35 wt%, respectively. The recovery rates of alumina and sodium oxide were 48.3% and 96.6%, respectively.

107

levels. Consequently, this method serves as a viable solution to mitigate the environmental repercussions stemming from red mud soilization. Wang Yanxiu [39] conducted a comparative analysis of the soil properties of Bayer process red mud, calcification-carbonation red mud, natural soil, and nutrient soil. The study revealed that both the original red mud and calcification-carbonation red mud fulfill the criteria for agricultural soil in terms of bulk density, total porosity, water retention, and fertility. According to the international soil texture classification, the former was categorized as sandy clay, while the latter exhibited a superior soil texture classified as sandy loam in comparison to Bayer process red mud. Nevertheless, red mud exhibits deficiencies in total nitrogen content, organic matter concentration, soil macroaggregates, and an imbalance in the proportion of readily available nutrients when compared to total nutrients. To address these shortcomings, Chaoxi [40, 41] employed a one-step KOH hydrothermal transformation of red mud to generate potassium silicate minerals. The resulting transformed slag contained a mass fraction of 10.63% effective K2O, 18.32% mass fraction of SiO2, and various mineral elements such as Ca, Mg, and Fe. It satisfied the requirements outlined in the “Organic‒Inorganic Compound Fertilizer Standard” (GB/T 18,877–2020), specifically meeting the Type I standard. When applied as a red mud slag-potassium silicate mineral fertilizer, it exhibited pronounced promotion of Chinese cabbage growth, establishing itself as a high-quality silicate mineral fertilizer that enhances nutrient absorption by plants.

Comprehensive Utilization of New Red Mud After the Calcified Carbonization Method

Optimization of Calcified Carbonization Method

To achieve clean alumina production and harmless utilization of red mud, Wang Kun [35, 36] conducted vortex melting reduction to recover iron from iron-containing tailings after calcification-carbonation. However, complete utilization of the remaining slag was still not achieved. Therefore, Yanxiu Wang [37, 38] used calcification-carbonation red mud as a raw material to prepare silicate cement and iron-aluminum cement. After appropriate component allocation, they were calcined at 1450 °C and 1300 °C. The strength of the silicate cement clinker exceeded that of 425-grade cement and reached the standard of 525-grade cement. The addition of calcification-carbonation red mud was 42.37%, which was 20% higher than that in existing research. The strength of the iron-aluminum cement met the requirements of 325-grade cement, with a calcification-carbonation red mud addition ratio of 65%. The calcification-carbonation method effectively reduces the alkalinity and salinity of Bayer process red mud, rendering it comparable to natural soil in terms of salinity

During the industrial promotion of the calcification-carbonation method, several issues have emerged. The calcification transformation requires the addition of nearly 40% lime, which significantly increases the proportion of solid raw materials and reduces material flowability. To address these challenges, Li Qiji made improvements to the existing calcification-carbonation process. By adding lime after leaching, Li Qiji effectively resolved issues such as poor slurry circulation, pipe scaling, and blockages. Additionally, the heat released during flash evaporation cooling and pressure reduction was fully utilized. Under conditions that did not affect the original reaction and production status, lime addition was adjusted to occur after slurry leaching, replacing flash evaporation. The heat from the residual steam after leaching was utilized to facilitate the calcification transformation. Xie Liqun employed continuous calcification-carbonation treatment on monohydrate bauxite red mud to reduce the calcification-carbonation process. The solution obtained after calcification was directly subjected to carbonation

108

without solid‒liquid separation. Both stepwise and continuous processing achieved the desired effect of aluminum extraction and alkali removal. The residual Na2O resulting from continuous processing without solid‒liquid separation after calcification did not affect the final outcome. Furthermore, the low solubility of alumina in the solution obtained after aluminum leaching in the calcification-carbonation method leads to the formation of a large amount of sodium aluminate solution with a high ratio of sodium aluminate to alumina molecules. If this solution were to be reintroduced into the Bayer process system like the calcification solution, it would increase the molecular ratio of sodium aluminate in the Bayer process and increase the energy consumption for evaporation. To address this issue, Lv [23] conducted research on the treatment process of monohydrate bauxite red mud using calcium aluminate as a calcium source in the calcification-carbonization method. They determined the process conditions and parameters while analyzing the reaction mechanism of aluminum precipitation and the cyclic utilization of calcium aluminate.

Conclusions After nearly a decade of relentless effort, the calcification-carbonization method has undergone systematic laboratory research and scaled-up experiments on over 20 representative raw materials both domestically and internationally. These materials include monohydrate bauxite and red mud from Henan and Shanxi Provinces, as well as highiron monohydrate bauxite and red mud from Yunnan Province. Additionally, trihydrate bauxite and red mud from regions such as Indonesia, Malaysia, Australia, and Guinea were also investigated. A series of Chinese and international patents have been obtained, affirming the efficacy of this method. It has the potential to increase China's bauxite resource reserves by 4–5 times, effectively mitigating the ecological threats posed by the traditional Bayer process. Furthermore, this method enables clean production of alumina and the low-cost large-scale utilization of Bayer process red mud, injecting new vitality into the alumina industry.

References 1. Xue S, Liu Z, Fan J, et al. Insights into variations on dissolved organic matter of bauxite residue during soil-formation processes following 2-year column simulation. Environmental Pollution, 2022, 292: 118326. 2. Liu X, Han Y, He F, et al. Characteristic, hazard and iron recovery technology of red mud—A critical review. Journal of Hazardous Materials, 2021, 420: 126542.

H. Xin et al. 3. Snars K, Gilkes R J. Evaluation of bauxite residues (red muds) of different origins for environmental applications. Applied Clay Science, 2009, 46(1): 13–20. 4. Collazo A, Fernández D, Izquierdo M, et al. Evaluation of red mud as surface treatment for carbon steel prior painting. Progress in Organic Coatings, 2005, 52(4): 351–358. 5. Komnitsas K, Bartzas G, Paspaliaris I. Efficiency of limestone and red mud barriers: laboratory column studies. Minerals Engineering, 2004, 17(2): 183–194. 6. Jústiz-Smith N, Buchanan V E, Oliver G. The potential application of red mud in the production of castings. Materials Science and Engineering: A, 2006, 420(1): 250–253. 7. Altundogan H, Altundoğan S, Tumen F, et al. Arsenic Adsorption From Aqueous Solutions by Activated Red Mud. Waste Management, 2002, 22: 357–363. 8. Deng B, Li G, Luo J, et al. Selectively leaching the iron-removed bauxite residues with phosphoric acid for enrichment of rare earth elements. Separation and Purification Technology, 2019, 227: 115714. 9. Huang L, Chunlei L I, Wang H, et al. Research Progress on Comprehensive Utilization of Red Mud. Journal of Physics: Conference Series, 2021, 2009(1): 012021–012026. 10. Novais R M, Carvalheiras J, Seabra M P, et al. Innovative application for bauxite residue: Red mud-based inorganic polymer spheres as pH regulators. Journal of Hazardous Materials, 2018, 358: 69–81. 11. Narayan A, Mac-Quhae C, Rosales J, et al. Does Alumina-Refining Waste Increase the Nutrient Level in Tropical Mesotrophic Floodplain Lakes?. Bulletin of Environmental Contamination Toxicology, 2021, 107: 506–513. 12. Mishra B, Gostu S. Materials sustainability for environment: Red-mud treatment. Frontiers of Chemical Science Engineering, 2017, 11(003): 483–496. 13. Bartlett R W. Alumina production process. Comprehensive Utilization of Minerals, 1989, (5): 5. 14. Peter Smith. The processing of high silica bauxites-Review of existing and potential processes. Hydrometallugy, 2009, 98(1–2): 162–176. 15. Andrew P.I., Anishchenko N.M., Mishchenkova N.P. Mechanism of the anionic flotation of chamosite and gibbsite. Tsvetnye Metally, 1973, 116(6): 16–20. 16. Barnes M C, Jonas A M, Gerson A R. The kinetics of desilication of synthetic spent Bayer liquor seeded with cancrinite and cancrinite/sodalite mixed-phase crystals. Journal of Crystal Growth.1999,200:251–264. 17. Roth L G, Berns D S, Chang-Hwei C, et al. Biobeneficiation of bauxite using bacillus polymyxa: calcium and iron removal. International Journal of Mineral Processing, 1996, 48(1): 34–41. 18. Banashettappa V. Biobeneficiation of bauxite ore using bacillus polymyxa. 2012, 49(1): 51–60. 19. Shaohan Wang, Huixin Jin, Yong Deng, Yuandan Xiao. Comprehensive utilization status of red mud in China: A critical review. Journal of Cleaner Production.2021,289:125136. 20. Xiaolin Pan, Hongfei Wu, Zhongyang Lv, Haiyan Yu, Ganfeng Tu.Recovery of valuable metals from red mud: A comprehensive review.Science of The Total Environment,2023,904:166686. 21. Jizhe Zhang, Zhanyong Yao, Kai Wang, Fei Wang, Hongguang Jiang, Ming Liang, Jincheng Wei, Gordon Airey.Sustainable utilization of bauxite residue (Red Mud) as a road material in pavements: A critical review. Construction and Building Materials.2021,270:121419. 22. Mengfan Wang, Xiaoming Liu.Applications of red mud as an environmental remediation material: A review. Journal of Hazardous Materials.2021,408:124420.

Research of Cleaner Production of Alumina and Harmless Utilization of Red Mud 23. Guo-zhi Lu, Ting-an Zhang, Li-nan Ma, Yan-xiu Wang, Wei-guang Zhang, Zi-mu Zhang, Long Wang.Utilization of Bayer red mud by a calcification–carbonation method using calcium aluminate hydrate as a calcium source. Hydrometallurgy.2019,188:248–255. 24. Pan Lu. Basic Research on Treating Low-grade Bauxite and Red Mud by Lime-carbonation Process [D]. Shengyang: Northeastern University,2011 25. Guo Fangfang. Research on Calcification-Carbonation Process for Red Mud [D]. Shengyang: Northeastern University, 2015. 26. Li JIqi. New Process for Alumina Production with Back Feeding CaO Technology in Calcification Carbonation Method[D]. Shengyang: Northeastern University, 2019. 27. Chen Y, Lv G, Zhang T A, et al. Carbide Slag as a Calcium Source for Bauxite Residue Utilization via Calcification–Carbonization Processing. Russ J Non-Ferr Met, 2022, 63(2): 132–145. 28. Zhu X F, Zhang T A, et al. Kinetics of carbonated decomposition of hydrogarnet with different silica saturation coefficients. International Journal of Minerals,Metallurgy and Materials, 2020, 27(4): 11. 29. Zheng Chaozhen. Research on the Generation Process and Carbonation Properties of Hydrogarnet [D]. Shengyang: Northeastern University, 2015. 30. Liu G T, Liu Y, Lv G Z, et al. Wet grinding of calcified slag to improve alumina extraction from red mud by the calcification-carbonization method. JOM, 2020, 72(2): 970–977 31. Zhang Z M, Lv G Z, Chen Y C, et al. Alumina extraction from kaolinite via calcification carbonation process. Russ J Non-Ferr Met, 2020, 61(3): 248–256. 32. Guo Yong. Research on Illite Transformation under Calcification-Carbonation Process[D]. Shengyang: Northeastern University, 2020. 33. Li Xiaoqi. Research on Calcification-Carbonation Process and Kinetics of Red Mud in Bayer Process [D]. Shengyang: Northeastern University, 2016.

109

34. Li R B, Zhang T A, Liu Y, et al. Characteristics of red mud slurry flow in carbonation reactor. Powder Technology, 2017, 311: 66– 76. 35. Kun Wang, Yan Liu, Ting-an Zhang, Xiao-fei Li, and Xin Chen. Investigationof thesmelting reduction mechanism and of iron extraction from high-iron red mud. MaterialsResearch Express, 2020, 7: 126514. 36. Kun Wang, Yan Liu, Guozhi Lyu, Xiaofei Li, Xin Chen, and Ting’an Zhang. Recoveryofiron from high-iron Bayer red mud by smelting reduction. TMS Annual Meeting, TMS Light Metals 2020, 92–97. 37. Y.X. Wang, T.A. Zhang, Y.H. Zhang, et al. Mineral transformation in treating low-grade bauxite by the calcificationcarbonization process and preparing cement clinker with obtained residue. Minerals Engineering, 2019, 138:139–147. 38. Wang Y, Zhang TA, Zhang Y, et al. Transformation and Characterization of Cement Clinker Prepared from New Structured Red Mud by Sintering. JOM: the journal of the Minerals, Metals & Materials Society, 2019, 71(8). 39. Wang Y, Zhang TA, Lv G, et al. Assessment of Bauxite Residue for Reclamation Purposes After Calcification–Carbonization Treatment. JOM: the journal of the Minerals, Metals & Materials Society, 2019, 71(9). 40. Xi Chao, Ting‒an Zhang, Guozhi Lyu, Zhipeng Liang, Yang Chen. Sustainable application of sodium removal from red mud: Cleaner production of silicon-potassium compound fertilizer. Journal of Cleaner Production, 2022, 352:131601. 41. Xi Chao, Ting‒an Zhang, Guozhi Lyu, Yang Chen, Qiuyue Zhao, Xuewei Yang, Fangqin Cheng. Research on the mechanism of sodium separation in bauxite residue synergy preparation of potassium-containing compound fertilizer raw materials by the hydrothermal method. Journal of Environmental Management, 2022, 317:115359.

Sustainable Valorization of Bauxite Residue (“Red Mud”): Exploring the Potential of H2 Reduction for Multi-metal Recovery Ganesh Pilla, Tobias Hertel, and Yiannis Pontikes

Abstract

This study presents a process for the simultaneous recovery of metals from bauxite residue (BR) at relatively lower temperatures using H2 reduction to achieve zero-waste discharge (minimal waste generation) of BR. The proposed method involves blending BR with NaOH and subjecting it to H2 reduction (using different process parameters), resulting in an intermediate material enriched in magnetite and water-soluble aluminate. Optimal process parameters were determined (temperature of 600 °C, a reduction time of 120 min, 20 wt.% NaOH addition, and a reduction atmosphere of 5 vol.% H2 + 95 vol.% N2 with a 20 L/h flowrate), resulting in satisfactory recoveries of Fe (74.7%, grade = 40%), Al (81.5%), and Na (91.4%). This approach represents a step towards the sustainable valorization of BR, i.e., the efficient recovery of metals (Fe, Al, and Na) along with a non-magnetic fraction (Ca, Si, Ti, Rare earth elements), a precursor for further Sc, Ti recovery, and an additive eventually for building materials. Keywords





Bauxite residue Red mud economy Metal extraction





Sustainability Hydrogen




Circular

Introduction The production of alumina from bauxite results in the significant generation of bauxite residue (BR) or red mud, a substantial solid waste. About 1 to 2 tons of BR are typically produced for each ton of recovered alumina [1]. The G. Pilla (&)
T. Hertel
Y. Pontikes KU Leuven, Department Materials Engineering, Sustainable Resources for Engineered Materials (SREMat), 3000 Leuven, Belgium e-mail: [email protected]

landfilling of substantial quantities of BR (>170 million tons/year) is associated with management risks, and occasionally even accidents, affecting both the environment and human lives. Globally, less than 3% of the BR is used, with the majority accumulated near the alumina refineries. BR is valued as a secondary resource rich in metals like Fe, Al, Na, Si, Ti, and REEs [2]. Yet, applying conventional beneficiation methods (gravity, magnetic, flotation techniques) for the recovery of these metals leads typically to poor recovery as BR is a complex material due to its intricate composition, small particle size, and high pH (>10). Both pyrometallurgical and hydrometallurgical methodologies have been explored to recover valuable metals from BR, as indicated in previous studies [3, 11, 14, 15]. A pilot-scale pyrometallurgical process has been proposed for the production of high-grade pig iron (with a purity of 95%) and mineral wool using Greek BR as a feedstock. This process has the capacity to utilize 1300 tonnes of dry BR per month, yielding 300 tonnes of pig iron and 865 tonnes of mineral wool [3]. However, it should be noted that this process consumes a significant amount of energy, approximately 2000 kWh per ton, and faces challenges associated with the substantial use of flux and carbon dioxide (CO2) emissions. These factors may limit its practicality for large-scale industrial implementation. The establishment of a hydrometallurgical route for the production of high-purity magnetite (98%), alumina (98%), and titania (98%) from BR has been achieved at the laboratory scale [15]. The potential for upscaling this process to an industrial level hinges on the development of a comprehensive approach for adapting the intricate process routes, effective management of liquid byproduct waste, and the implementation of stringent safety protocols for handling hazardous acids and substances such as H2SO4, NH4OH, and C2H2O4. Recent efforts have focused on the exploration of hydrogen reduction at low temperatures as a promising method for recovering metals from BR [4–10]. However, the limited existing literature addressing the recycling of substantial

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_13

110

Sustainable Valorization of Bauxite Residue (“Red Mud”): Exploring the Potential of H2 Reduction for Multi-metal Recovery

waste quantities raises uncertainties about the feasibility of upscaling such initiatives to achieve a sustainable BR process on an industrial scale. Therefore, this study proposed a method for the simultaneous recovery of Fe (in the form of magnetite), Al, and Na along with non-magnetic fraction (Ca, Si, Ti, and REEs) from H2 reduced intermediate products. Unlike the smelting/pyrometallurgical processes, this method involves a combination of low-temperature H2 reduction ( 99.9) sodium hydroxide (NaOH) was used as an additive to BR for H2 reduction experiments. If the NaOH has impurities, it might affect the extraction efficiency of Aluminum. During water leaching-magnetic separation or wet-milling process, these impurities dissolve and impact the metallurgical-grade alumina production.

Experimental Procedure The process begins with a pelletization (Eirich mixer type EL1) step in which 100 g of BR was blended with NaOH solution (with an addition of 10–25 wt.% based on dry

111

weight). The resulting pellets are within the size range of 10–20 mm and were subsequently dried. BR has a very fine particle size distribution and it’s adequate to use this material in the pellet form for efficient utilization instead of fines and without losing the products. In addition, pellets are preferred fee for better metallurgical properties such as high reducibility and better permeability. A rectangular alumina boat-type crucible containing approximately 100 g of pellets was placed in a lab-scale box furnace under an N2 atmosphere. H2 gas (5 vol.% H2 + 95 vol.% N2 with a flow rate of 20 L/h) was purged for a specified reduction time (30– 120 min) and at a designated temperature (ranging from 400 to 700 °C). After the reduction process, the pellets were cooled to room temperature under an N2 atmosphere (flowrate of 10 L/h) to prevent oxidation of the reduced pellets. Following the reduction, wet grinding (Retsch RS200 model) of the pellets was carried out with a solid-to-liquid (S/L) ratio of 0.5. The resulting wet-milled slurry products were then subjected to water leaching and wet-magnetic separation. As compared to dry magnetic separation, wet-magnetic separation is suitable for fine particles in terms of efficient separation of different metals from BR along with enhancement of scalability. This separation process aimed to separate water-soluble aluminate solution (for Al and Na recovery), magnetic fraction (for Fe recovery), and non-magnetic products (consisting of Ca, Si, and Ti). Within the water-leaching setup, a solid magnet block (with a magnetic strength of 0.29 T) was positioned to facilitate the separation of the magnetic product from the non-magnetic portion. Water leaching was conducted at 60 °C, employing an S:L ratio of 1:10 to prevent Si-gel formation and enhance the effectiveness of Fe, Al, and Na recovery. After a 60-min water leaching period, the magnetic fraction was detached from the magnetic block. The remaining solid was separated from the leach liquor through a two-stage filtration process involving filter papers with pore sizes of 12 µm and 0.45 µm, respectively.

Product Characterization The characteristics of the reduced products were assessed using Wavelength Dispersive X-ray Fluorescence (WDXRF) analysis performed with a 4 kW Bruker S9 Tiger instrument, Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) conducted with an S8 Varian 720 ES axial instrument, as well as X-ray Diffraction (XRD) analysis using a Bruker D2 Phaser instrument with Cu-K radiation within a 10–50° 2h range and a 0.08 step size. The ICDD-PDF database was employed for XRD analysis. Additionally, the SEM–EDS (Scanning Electron Microscopy—Energy Dispersive X-ray Spectroscopy) technique

112

G. Pilla et al.

was employed using an XL 30 FEG instrument to examine the microstructural variation. To determine the recovery of Fe, Al, and Na, the formulas outlined in earlier studies [8] were adopted for calculation purposes.

Process Optimization To determine the optimal parameters for simultaneous Fe, Al, and Na recovery and explore the impact of various process variables and their interrelationships, the study employed a design of experiments (DoE) approach. This approach involved utilizing statistical models [13] through the implementation of response surface methodology (RSM) performed using Design Expert JMP Pro software version 17. The specifics regarding DoE-RSM models are discussed elsewhere [13, 14]. The study examines three process factors, namely reduction temperature (x1) ranging from 400 to 700 °C, reduction duration (x2) ranging from 30 to 120 min, and NaOH addition level (x3) ranging from 10 to 25 wt.%. In the laboratory, the prescribed number of experimental runs (Table 1), utilizing the DoE to include all possible combinations, were executed. For each experimental run, the recovery rates of Fe, Al, and Na were computed. Table 1 Experimental matrix for reduction experiments of 5 vol% H2 + 95 vol% N2 with 20 L/h flowrate and recovery results of Fe, Al, and Na at different runs

Experiment

Results and Discussion Raw BR The Greek raw BR consists of a mixed-phase assemblage of the total −Fe content (22.7 wt%), Al2O3 (23.6 wt%), SiO2 (7.5 wt%), CaO (8.3 wt%), TiO2 (5.1 wt%), and Na2O (3.4 wt%) in the form of hematite/goethite, diaspore/boehmite/ gibbsite, quartz, cancrinite, katoite, calcium carbonate, perovskite/anatase. The SEM–EDS analysis, as depicted in Fig. 1, conducted on the unprocessed BR, demonstrates the presence of sub-micron particles showcasing a complex combination of various major elements within the bulk composition of the sample. This intricate elemental association presents a substantial challenge in the context of metal recovery [12, 15].

Influence of H2 Reduction Parameters on Metal Recovery The reduction tests were performed by increasing the temperature from 400 to 700 °C for 120 min with the addition of 20 wt.% NaOH under an atmosphere of 5 vol.% H2 + 95 vol.% N2 with 20 L/h flowrate.

Variables

Response (experimental data on recovery, %)

Reduction temperature (°C) —1

Reduction time (min)—2

NaOH concentration (wt %)—3

Fe

Al

Na

1

600

120

10

68.6

62.6

77.3

2

600

120

15

72.8

71.3

87.7

3

600

120

20

74.7

81.5

91.4

4

600

120

25

57

81.8

92.4

5

600

30

20

69.8

79.4

88.9

6

600

60

20

70.2

80.6

90.7

7

400

120

20

49.1

76.8

87.9

8

500

120

20

76.5

80.2

90.1

9

700

120

20

50.2

84.0

92

10

700

120

10

60.5

62.4

78.8

11

700

30

10

66.6

59.8

75.2

12

700

30

25

40.1

83.6

89.4

13

400

30

10

34.5

59.8

82.8

Sustainable Valorization of Bauxite Residue (“Red Mud”): Exploring the Potential of H2 Reduction for Multi-metal Recovery

113

Fig. 1 SEM–EDS analysis on the Greek raw BR

The recovery rates of Al and Na increased with rising temperature (Fig. 2). The Al recovery rose from 76.8% to 80%, and Na recovery climbed from 88% to 92.5%, attributed to the enhancement of water-leachable aluminates (NaAlO2) phases. During the H2 reduction process of BR, the added NaOH reacts first with Al and Si phases due to high reactivity than Fe. Eventually, the formed Al phase of soluble NaAlO2 is endothermic and highly stable than Fe3O4 and with increasing temperature NaAlO2 phase formation increases and overall, this phenomena is positively effect on Al recovery from BR [4, 6]. Conversely, the rise in temperature from 400 to 500 °C led to increased Fe recovery and TFe grade, followed by a decline at 700 °C (Fig. 2). This was due to enhanced reaction rates between hematite and magnetite. However, higher temperatures caused magnetite to convert into FeO with weaker magnetic properties, resulting in Fe losses during magnetic separation [9, 12]. Fig. 2 Effect of Temperature (400–700 °C) on recovery of Fe, Al, and Na from H2 reduced products (reduction condition: reduced pellet samples at 400– 700 °C for 120 min with 20 wt.% NaOH under 5 vol.% H2 + 95 vol.% N2 [flowrate = 20 L/h])

Experiments were conducted under a reduction temperature of 600 °C for 30–120 min, employing a gas mixture of 5 vol.% H2 + 95 vol.% N2 (flowrate: 20 L/h), and 20 wt.% NaOH. As depicted in Fig. 3, following water leaching and wet-magnetic separation, prolonged reduction times corresponded to slightly increased recovery rates for Al (79.4 to 81.5%) and Na (88.9 to 91.4%). Similarly, Fe recovery rates are enhanced from 69.8 to 74.7%, accompanied by an upsurge in the TFe grade of the magnetic concentrate (38.3 to 40%). Notably, a shorter reduction (30 min) exhibited the presence of unreduced hematite leading to Fe recovery/TFe grade (due to the low magnetic susceptibility of hematite). Another set of experiments was conducted at 600 °C for 120 min using a 5 vol.% H2 + 95 vol.% N2 gas mixture and 20 L/h flowrate, with NaOH concentration varying from 10 to 25 wt.% (Fig. 4). Increasing NaOH addition to BR for H2 reduction resulted in notable improvements in Al and Na

114

G. Pilla et al.

Fig. 3 Effect of Time (30– 120 min) on recovery of Fe, Al, and Na from H2 reduced products (reduction condition: reduced pellet samples at 600 °C for 30– 120 min with 20 wt.% NaOH under 5 vol.% H2 + 95 vol.% N2 [flowrate = 20 L/h])

Fig. 4 Effect of NaOH (10–25 wt.%) on recovery of Fe, Al, and Na from H2 reduced products (reduction condition: reduced pellet samples at 600 °C for 120 min under 5 vol.% H2 + 95 vol.% N2 [flowrate = 20 L/h])

recovery rates (Al recovery rose from 62.6% to 81.8%, and Na recovery increased from 77.3% to 92.4%). This higher NaOH content facilitated the formation of sodium aluminate phases, enhancing Al and Na recovery, which aligns with findings from prior studies [16]. Conversely, Fe recovery exhibited different trends (Fig. 4). It increased from 68.6% to 74.7% (with an increase in TFe from 36.2% to 40%) from 10 to 20 wt.% NaOH but dropped to 57% (with a TFe Grade of 37.9%) at 25 wt.% NaOH due to an unfavorable water-leachable sodium iron oxide (NaFeO2) phase formation [5, 6]. The development of a dense NaFeO2 layer results in a slower reaction, hindering the conversion of hematite into magnetite. This layer forms a dense film that obstructs H2 diffusion, thereby impeding the reduction of internal hematite [17, 18]. This phenomenon could potentially influence Fe conversion, subsequently affecting Fe-recovery outcomes.

Optimization and Validation of Process Through a comprehensive factorial design involving a series of experimental runs (Table 1) and employing quadratic models, the process parameters for simultaneous Fe, Al, and Na extraction were optimized. The resultant optimized conditions (Fig. 5) encompass a reduction temperature of 600 °C), 20 wt.% NaOH, and a reduction time of 120 min. According to projections, the anticipated recoveries are 82.8% for Al, 70.5% for Fe, and 92.6% for Na. The analysis of variance (ANOVA) by using quadratic models employed in evaluating input process parameters (temperature, time, and NaOH) effect on Fe, Al, and Na recovery. From ANOVA analysis, a “Probability > F (with noise)” value below 0.05 resembles the statistical significance of the model terms. Under this condition (P < 0.05),

Sustainable Valorization of Bauxite Residue (“Red Mud”): Exploring the Potential of H2 Reduction for Multi-metal Recovery

115

The proposed process (Fig. 6) involves a two-stage method for recovering various products from BR. It includes H2 reduction followed by water leaching and magnetic separation to obtain iron, sodium aluminate, and a non-magnetic fraction enriched in Ca, Si, and Ti as major products. The non-magnetic fraction holds the potential for recovering Ti, Si, REEs, and materials suitable for the construction industry. Moreover, alumina can be further extracted via precipitation for sale at higher market values, while the magnetic fraction finds applications as a pigment, and additive in iron-making, and sintering processes. Considering material balance (Fig. 6), processing 1 ton of BR yields 477 kg of magnetic fraction, 75 kg of non-magnetic fraction, and 477 kg of sodium aluminate/alumina after a 2-stage process of H2 reduction (600 °C) followed by a combined water leaching-magnetic separation.

Conclusions Fig. 5 Overlay plots from JMP design expert software indicating the optimal region for the simultaneous extraction of Fe, Al, and Na

the quadratic terms x21 (temperature) and x23 (NaOH addition) are more significant model terms for Fe recovery than the time factor. The model terms of linear x3 and quadratic x23 (NaOH) are more significant than time and temperature for the Al and Na recovery. Overall, temperature and NaOH addition parameters significantly influence the enhancement of simultaneous recovery of Fe, Al, and Na over time. According to the data presented in Table 2, the actual recovery rates for Fe, Al, and Na are 74.7%, 81.5%, and 91.4%, respectively. The deviations in Fe, Al, and Na recoveries are 4.2%, 1.4%, and 1.2%, respectively. A comparison between the projected and actual outcomes reveals that the optimal process parameters obtained using the software employed are remarkably accurate, with all values falling within an acceptable deviation threshold (270 °C), grain size growth is apparent, as

The effect of heat treatment at different temperatures on the microhardness of LPBF A20X is shown in Fig. 4. The hardness of as-printed A20X lies within 114–120 HV, as highlighted in Fig. 4a. In the low-temperature heat treatment regime (180–370 °C), increasing the temperature produces significant softening compared to the as-printed condition. For example, after 6 h of heat treatment at 220 °C, the hardness of the alloy drops by up to 19% (to 97 ± 2 HV) compared to as-printed A20X. At temperatures  220 °C, increasing the duration of heat treatment further reduces the

Fig. 2 Microstructures of LPBF A20X in the a as-printed state and various heat-treated states including b 180 °C for 6 h, c 220 °C for 6 h, d 270 °C for 6 h, e 320 °C for 6 h, and f 370 ° C for 6 h. Images are taken in the plane perpendicular to build direction (XY plane)

Heat Treatment of A20X Alloy Manufactured Using Laser Powder Bed Fusion

379

Fig. 3 Effect of heat treatment on the microstructure of LPBF A20X after various heat treatments: a 450 °C for 6 h, b 480 °C for 6 h, c 505 °C for 6 h, and d 530 °C for 6 h. Images are taken in the plane perpendicular to build direction (XY plane)

Fig. 4 Effect of heat treatment on the hardness of LPBF A20X at temperatures: a below 400 °C and b above 400 °C

material hardness. Above 220 °C, significant softening occurs at shorter dwell times ( |t| is 0.9879, much larger than 0.05, it can be considered that the results measured by these

two methods are equal, indicating that the real-time online measurement of the zone anode current can be realized by folding the fiber optic ring of FOCS, without affecting the anode change operation.

592 Table 2 Paired analysis of measurement values for the individual anode

Y. Meng et al. Item

Value

Item

Value

Current_N

7.99472

T-Ratio

0.057436

Current_T

7.99328

Degree Freedom

31

Mean difference

0.00144

Probability > |t|

0.9546

Samples number

32

Probability > t

0.4773

Correlation

0.99745

Probability < t

0.5227

Fig. 6 Zone measurement of anode current by both the traditional method and new method, respectively: a traditional method; b new method

Table 3 Measurement results of series zone anode current

Cell number

Measurement value(A) Traditional method

5001

5038

6001

Table 4 Paired analysis of zone anode current value

Region number

New method

1

64675

64057

2

62176

60320

3

63766

63590

4

61813

61842

1

63995

64466

2

62602

62427

3

59349

60616

4

67888

69302

1

67815

67191

2

60377

60396

3

61068

60786

4

65540

66024

Item

Value

Item

Value

Current_N

63418.1

T-Ratio

−0.01551

Current_T

63422

Degree Freedom

11

Mean difference

−3.9167

Probability > |t|

0.9879

Samples number

12

Probability > t

0.5060

Correlation

0.95449

Probability < t

0.54940

Accurate Measurement of Anode Current in Aluminum Electrolysis: From Ideal to Reality

593

Continuous Measurement and Analysis of Zone Current So far, a vast number of papers discuss about individual anode current measurement. However, little is reported on the measurement and analysis of zone anode current. Therefore, the new method was used to continuously measure, over a long time, the current in the 4th region of cell 5001, near the duct-end. Figure 7 shows the current curve measured over 5 h from approximately 9:20 to 15:00, and Fig. 8 shows the measurement distribution. Overall, the current in the region follows a normal distribution, with a mean value of 61387A and a standard deviation of 768A. The corresponding coefficient of variation is only 1.25%, indicating that the change in current is not large. However, after a time 14:20 (Fig. 7), the current decreases rapidly from its maximum value, which might be caused by some abnormality in this region. The cell voltage during this time was obtained with a sampling frequency of 0.1 Hz. The zone current value was measured at the same frequency. A zone pseudo-resistance was computed according to Eq. (3) [4]. Figure 9 shows the changes of zone current, cell voltage, and zone pseudo-resistance. R ¼ ðV  Vext Þ=I

ð3Þ

where, R is zone pseudo-resistance, X; V is cell voltage, V; Vext is extrapolation voltage, taken 1.65 V; I is zone anode current, A. Based on Fig. 9a, it can be seen that, although the zone pseudo-resistance is calculated by Eq. (3) and determined by both the cell voltage and the zone anode current, its variation

Fig. 7 Continuous measurement results of current in the 4th region of cell 5001

Fig. 8 Distribution analysis of continuous measurement data

characteristic tends to be more similar to the changes in the cell voltage. Only during the time of approximately 13:30 * 14:40, when the cell voltage is relatively small and the zone current changes obviously, does the variation trend deviate significantly from the voltage. However, when zooming on Fig. 9a, some differences are still present, as shown in Fig. 9b and c. The current control of aluminum reduction cell is based on the pseudo-resistance variation of the entire cell. Since the line current is constant, it is essentially based on the variation of cell voltage. According to Fig. 9a, it can be considered that if the control of the aluminum reduction cell is carried out by the individual region feeding, on the basis of the zone pseudo-resistance, the overall control is basically consistent with the control based on the cell voltage. But according to Fig. 9b and c, if the alumina in acertain region is controlled based on the zone pseudo-resistance, it will be more suitable for the specific requirement in this region, and would enable a higher degree of control. When the zone anode current of the aluminum reduction cell can be accurately measured online, the change of zone pseudo-resistance can computed by the cell voltage and zone anode current. This can be used to control alumina feeding in each region to replace the actual feeding strategy which is feeding all regions with the equivalent alumina supply based on the overall cell pseudo-resistance. As a result, the large reduction of low voltage anode effect and their related PFC emissions should be realized. Obviously, compared to the voltage variation of the entire cell, the zone anode current is also more helpful in diagnosing local faults such as the anode damage, the anode spikes, and the feeder failure. Due to limitations in accuracy, traditional methods such as voltage drop and Hall effect sensor, will add errors when

594

Y. Meng et al.

Fig. 9 Changes of zone current, cell voltage, and zone pseudo-resistance in some period of time: a about 9:20 am *15:00 pm (*340 min); b about 14:01 pm * about 14:04 pm (*3 min); c about 14:48 pm * about 14:51 pm (*3 min)

individual anode measurements are used to describe current in a given zone. Consequently, the changes of zone anode current shown in Fig. 7 might be difficult to measure. Even if the cost of FOCSs drops significantly, to the point where each anode can be measured individually, it is necessary to use a single fiber optic to measure the zone current due to the measurement accuracy. The method to measure zone anode current by folding the optical fiber ring has opened new possibilities for the development of aluminum reduction cell control technology. Based on the support of this technology, it is conceivable that every four anodes could be equipped with an independent feeder for better process control.

Conclusion FOCSs have a series of advantages such as high measurement accuracy, wide measurement range, resistance to background magnetic field interference, and flexibility for deformation. The method of folding fiber optic was proposed to use the FOCS for online accurate measurement of anode current in industrial aluminum reduction cells. (1) The paired analysis of measurement results of the individual anode and zone anode shows that the measurement results obtained by the method of folding

Accurate Measurement of Anode Current in Aluminum Electrolysis: From Ideal to Reality

optical fiber rings are equal to those obtained by the traditional method. The zone anode current can be measured by the FOCSs by using the method of folding the fiber optic. (2) The change of zone pseudo-resistance tends to be similar to the change of cell voltage. Zone pseudo-resistance is more sensitive to reflect the change in this region than cell voltage, so the alumina concentration in a certain region can be controlled individually by using the zone pseudo-resistance. (3) The cost of current measurement and control on the aluminum reduction cell by using high accurate sensors will be reduced significantly by using the FOCSs with the same number of feeding zone. Some relevant engineering application research has already been carried out.

Acknowledgements This work was supported by the National Natural Science Foundation of China (52374349).

References 1. Gariépy R, Couturier A, Martin O, et al. (2014) Preparation and Start-up of Arvida Smelter, AP60 Technological Center. Paper presented at the 143rd TMS Annual Meeting, San Diego, CA, 16– 20 February 2014 2. Zhou D, Yang X, Liu M, et al. (2015) Chinalco 600KA High Capacity Low Energy Consumption Reduction Cell Devolopment. Paper presented at the 144rth TMS Annual Meeting, Orlando, Florida, 15–19 March 2015 3. Lavoie P, Taylor MP., Metson JB (2016) A Review of Alumina Feeding and Dissolution Factors in Aluminum Reduction Cells. Metall Mater Trans B 47:2690–2696 4. Potocnik V, Reverdy M (2021) History of Computer Control of Aluminum Reduction Cells. Paper presented at the 150h TMS Annual Meeting, Orlando, Florida, 14–18 March 2021 5. Boulanger J, Gosselin A, Gaboury S, et al. (2022) Imaging Alumina Distribution Using Low-Voltage Anode Effect Detections in Anodic Current. Paper presented at the 151st TMS Annual Meeting, Anaheim, California, 27 February-3 March 2022

595

6. Tessier J, Cantin K, Magnusson DT (2018) Investigation of Alumina Concentration Gradients Within Hall-Héroult Electrolytic Bath. Paper presented at the 147th TMS Annual Meeting, Baltimore, Maryland, 15–18 April 2018 7. Moxnes B, Solheim A, Liane M, et al. (2009) Improved Cell Operation by Redistribution of the Alumina Feeding. Paper presented at the 138th TMS Annual Meeting, San Francisco, California, 15–19 February 2009 8. Yang S, Zhang H, Zou Z, et al. (2019) Reducing PFCs with local anode effect detection and independently controlled feeders in aluminum reduction cells. JOM 72(1):229–238 9. Wang Y, Tie J, Sun S, et al. (2016) Testing and Characterization of Anode Current in Aluminum Reduction Cells. Metall Mater Trans B 47: 1986–1998 10. Potocnik V, Arkhipov A, Ahli N, et al. (2017) Measurement of DC Busbar Currents in Aluminium Smelters. Travaux 46, Paper presented at the proceedings of 35th international ICSOBA conference, Hamburg, Germany, 2–5 October 2017 11. Bohnert K, Gabus P, Nehring J, et al. (2007) Fiber-Optic Current Sensor for Electrowinning of Metals. J Lightwave Techno 25 (11):3602–3609 12. Fan HT, Tie J, Zeng QY, et al. (2019) Measurement and analysis of anode current in 400kA aluminum reduction pots. Light metals 4:26–30 13. Zeng Q, Li C, Meng Y, et al. (2020) Analysis of interelectrode short-circuit current in industrial copper electrorefining cells. Measurement164:108015 14. Wong CJ, Shi J, Bao J, et al. (2023) A Smart Individual Anode Current Measurement System and Its Applications. Paper presented at the 152nd TMS Annual Meeting, San Diego, California, 19–23 March 2023 15. Tie J, Xiao H, Zhao RT, et al. (2023) Zone anode current measurement system and electrolytic cell measurement system based on individual fiber ring. Chinese Patent 2023109487869. 31 July 2023 16. Tie J, Zhao R, Zhang Z et al. (2020) System and method for measuring anode current of aluminum electrolytic cell. US. Patent 20200032408A1, 2020 17. Xiao H, Tie J, Lei J, et al. (2023) A system, method and electronic equipment for measuring zone anode current in an aluminum reduction cell. Chinese Patent 2023106872489. 9 June 2023 18. Tie J, Zhao RT, Zhang ZF, et al. (2017). Accurate measurement and its application of anodic current in aluminium electrolysis. Metall Indus Automation 41(6):49–54 19. Montgomery DC, Runger GC, Hubele NF (2011) Engineering Statistics. John Wiley & Sons, Inc.

Correlation Between Corrosion Rate and Electrochemical Parameters of Anode Process on the Metallic Electrode in Molten Oxyfluorides Andrey Yasinskiy, Thomas Jamieson, Kamaljeet Singh, Guðmundur Gunnarsson, Jon Magnússon, Dominic Feldhaus, Roman Düssel, Isabella Gallino, and Bernd Friedrich

Abstract

Using metallic anodes enables oxygen evolution during the aluminum reduction in electrolysis cells with vertical electrodes. Although they are considered non-consumable compared to carbon ones, they tend to corrode in molten fluoride media, and the corrosion products end up in aluminum as impurities. The big challenge for the industry is finding the best conditions (anode and bath composition, temperature, and current density) that provide the lowest corrosion rate. The most reliable way to characterize the corrosion rate is to perform an electrolysis test and determine the mass of anode components in the produced aluminum. This method is highly time- and resource-demanding. This paper A. Yasinskiy (&)
B. Friedrich IME Process Metallurgy and Metal Recycling, RWTH Aachen University, 52056 Aachen, Germany e-mail: [email protected] A. Yasinskiy Laboratory of Physics and Chemistry of Metallurgical Processes and Materials, Siberian Federal University, 660025 Krasnoyarsk, Russia Key Laboratory for Ecological Metallurgy of Multimetallic Minerals (Ministry of Education), Northeastern University, Shenyang, 110819, China T. Jamieson
I. Gallino Chair of Metallic Materials, Saarland University, 66123 Saarbrücken, Germany

investigates the connection between the electrochemical parameters of the anode process, which can be measured in minutes, and the results of long-term electrolysis tests. Several parameters, such as exchange current density on clean and oxidized surfaces, were studied in the low-temperature KF-NaF-AlF3-Al2O3 melt at 800 °C. A linear dependency between exchange currents and corrosion rate was found. Keywords











Introduction Metallic anodes hold significant potential for replacing consumable carbon anodes in molten salt electrolysis technologies, particularly in processes like aluminum electrowinning, where oxygen evolution is desirable [1, 2]. The conventional use of prebaked carbon anodes in the aluminum industry simplifies the metal reduction process and reduces costs. However, it comes with a notable drawback—greenhouse gas emissions resulting from several reactions [3]:

K. Singh Department of Engineering, Reykjavik University, 102 Reykjavik, Iceland Department of Materials Science and Engineering, NTNU, 7491 Trondheim, Norway





Exchange current Voltammetry Corrosion rate Inert anode Molten fluoride Aluminum reduction Wear rate

Al2 O3ðdisÞ þ 3=2CðsÞ ¼ 2AlðlÞ þ 3=2CO2ðgÞ

ð1Þ

Al2 O3ðdisÞ þ 3CðsÞ ¼ 2AlðlÞ þ 3COðgÞ

ð2Þ

Al2 O3ðdisÞ þ 2AlF3ðdisÞ þ 3CðsÞ ¼ 4AlðlÞ þ 3COF2ðgÞ ð3Þ

G. Gunnarsson IceTec, Arleynir 8, 112 Reykjavik, Iceland J. Magnússon Arctus Aluminum Limited, Arleynir 8, 112 Reykjavik, Iceland

AlF3ðdisÞ þ 3=4CðsÞ ¼ AlðlÞ þ 3=4CF4ðgÞ

ð4Þ

AlF3ðdisÞ þ CðsÞ ¼ AlðlÞ þ 1=2C2 F6ðgÞ

ð5Þ

D. Feldhaus
R. Düssel TRIMET Aluminium SE, 45356 Essen, Germany © The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_76

596

Correlation Between Corrosion Rate and Electrochemical Parameters …

This result in approximately 1.5 tons of CO2 emitted per ton of aluminum due to reactions (1) and (2), with an additional 0.2 tons of CO2-e per ton of aluminum attributed to perfluorocarbon gas emissions in reactions (3), (4), and (5) when the alumina is depleted from the electrolyte [1]. Numerous experiments [4–6] have shown that replacing carbon anodes with alternative materials such as argon plasma [7], ceramics [8], cermets [9, 10], or metal alloys [11] can facilitate the oxygen-evolving reaction: Al2 O3ðdisÞ ¼ 2AlðlÞ þ 3=2O2ðgÞ

ð6Þ

Metallic anodes, in particular, have demonstrated promise due to their high electrical conductivity, resistance to thermal shock, robustness, ease of fabrication, and electrical connectivity [5]. Alloys like Cu-Ni-Fe [12–16] and Cu-Al [17–19] have been extensively studied in low-temperature electrolytes ranging from 700 to 950 °C [16]. However, the presence of iron (Fe) in the alloy can lead to a preferential attack by the electrolyte, resulting in the formation of FeF2 [20]. Recent research suggests that nickel-rich Ni-Cu-Fe-based alloys may perform better due to the formation of protective oxide scales (Cu2O/CuO, NiO, and NiFe2O4). The performance of metallic anodes is highly influenced by electrolyte temperature. Lower-temperature electrolytes offer advantages such as slower dissolution of the anode oxide scale and reduced risk of anode material damage. KF– NaF–AlF3-based electrolytes, for instance, present a compelling low-temperature alternative to the commonly used NaF-AlF3 electrolytes. They offer high alumina solubility at lower current densities and operate at temperatures around 800 °C, providing improved electrical conductivity compared to KF-AlF3 [21]. The performance of metallic anodes in the electrolysis is characterized by two key parameters: (i) Wear Rate: This parameter, determined through long-term electrolysis tests, calculates the mass of impurities originating from the anode in the bath and the produced metal over time and the surface area exposed to the electrolyte. It assesses anode lifespan and, crucially, the achievable aluminum purity. (ii) Voltage Stability: This parameter reflects anode passivation behavior. In some cases, uncontrollable voltage spikes occur due to insulating compounds in the anode scale, leading to excessive energy consumption. Anode composition significantly influences these parameters. The most reliable method to evaluate anode performance in these aspects is to conduct long-term electrolysis tests in at least medium-scale (20–100 A) laboratory cells. However, this approach, which involves testing a wide range of compositions and conditions to gain a better

597

understanding of the corrosion mechanism, study dependencies, and find the optimal anode, is both time- and resource-intensive. Alternatively, there is ongoing research aimed at electrochemical characterization of anode performance using time- and resource-efficient techniques [4, 19]. These methods hold promise in bridging the gap between real-world anode performance and kinetic parameters of the anode process for oxygen-evolving electrode. As of now, there is no direct link between the two. The research presented here takes a step towards closing this gap by proposing a reliable technique for studying corrosion resistance in a simple and time-efficient manner.

Experimental Methods All experiments were conducted in a low-temperature NaF– KF–AlF3–Al2O3(sat) melt with a similar composition at 800 ± 2 °C. The bath was prepared using dried NaF and KF salts, each with a purity of 99.5% supplied by Sigma-Aldrich. Industrial-grade AlF3 and smelter-grade alumina, serving as a source of O2− ions, were provided by Trimet GmbH. Electrochemical measurements were carried out in a three-electrode cell, consisting of a graphite container containing approximately 300 g of electrolyte, placed within a gas-tight steel reactor. The reactor had a continuous Ag supply and a water-cooled lid. Cylindrical samples of Cu-Ni-Fe alloy, both in as-cast and homogenized states and machined to a diameter of 2 mm, were immersed 10 mm deep in the electrolyte and served as the working electrode. The quasi-reference electrode (qRE) was a Ø1 mm tungsten rod, also immersed 10 mm deep in the electrolyte. The counter electrode was a Ø4 mm glassy carbon rod, immersed 20 mm deep. Temperature measurements were conducted using an S-type thermocouple immersed 10 mm deep. Electrochemical kinetic parameters, potentially correlated with different corrosion mechanisms, were recorded using an Iviumstat.h potentiostat (Ivium Technologies, The Netherlands) controlled by Iviumsoft, following the steps summarized in Table 1. Five different compositions of Cu-Ni-Fe alloys were tested (specific compositions are subject to a non-disclosure agreement and may be disclosed in future publications). One of these compositions was tested in both the homogenized and as-cast states, while the others were solely in the homogenized state. Each composition was tested three times to confirm reproducibility. The exchange current density was determined by extrapolating the linear portions of the Tafel plot to the reversible potential. Stationary potentials and relaxation times were graphically determined from the potential-time

598 Table 1 Electrochemical measurement strategy

A. Yasinskiy et al. #

Procedure

What is measured

What is characterized

1

Weighting before the test

Initial mass

–

2

Equilibration time (3 min)

Open circuit potential

–

3

LSV* after immersion

Exchange current and reversible potential on clean anode

The resistance towards metal oxidation/dissolution

4

Galvanostatic polarization (1 h, i = 0.1 A
cm–2)

Stationary potential

The oxide layer growth and passivation behavior

5

LSV after polarization

Same as in #3 on oxidized anode

The resistance towards oxide dissolution

6

OCP** (relaxation)

Relaxation time and relaxation rate

The stability of oxide layer

7

LSV after relaxation

Same as in #3 on quasi-equilibrated anode

The resistance towards corrosion

8

Weighting after the test

Final mass, mass gain

The oxide layer growth mechanism

Linear sweep voltammetry (potential range: Eocp ± 400 mV, sweep rate 50 mV
s–1) **Open circuit potentiometry (duration 0.5 h)

*

curve recorded during the chronopotentiometry and open-circuit potentiometry. The relaxation rate was calculated as ΔE/Δs (dE/ds). Before measuring the mass of the anode after each test, any visible residues from the bath were mechanically removed from the oxide scale. Electrolysis tests were conducted in a two-electrode cell with vertical plate-shaped electrodes, maintaining an average current of 32 A for 12 h (unless interrupted for technical reasons). The cell was placed in an electric furnace with a constant temperature. Alumina was continuously fed into the cell. After each test, samples of the bath and produced metal were collected. The metal samples were analyzed using Spark OES at DTE Iceland and ICP OES at Trimet GmbH. The bath samples were analyzed with ICP OES at Trimet GmbH. The wear rate, expressed in g
(m2
h)−1, was calculated using the following equation: P P mMeðAlÞ þ mMeðbathÞ ð7Þ WR ¼ s
A where mMeðAlÞ is the mass of anode constituents in aluminum, mMeðbathÞ is the mass of anode constituents in the bath, s is the test duration, A is the total surface area of the anode exposed to electrolyte, Me = Ni, Fe, Cu. Each of the anode compositions was tested various number of times from single measurement to six tests. Totally, 14 electrolysis tests were done.

Results and Discussion The kinetic parameters of the anode process were measured and plotted against the wear rate of an anode with the same (or a closely similar) elemental composition to determine if

there is a statistical correlation between them. The correlation between the wear rate and the first set of electrochemical parameters is illustrated in Fig. 1. The average wear rates of the tested anodes ranged from 1.9 to 6.4 g
(m2
h)−1, depending on the anode composition, homogenization treatment, and operational conditions, such as fluctuations in alumina feeding. The average exchange current densities on a clean (non-oxidized) anode surface fell in the range of 0.4–2.5 mA
cm−2, which is of a similar magnitude as values reported for different metals in molten salts [22]. Transitioning from a clean metal electrode to an oxidized metal electrode led to a notable increase in exchange current, ranging from 15 to 80 mA
cm−2, in agreement with reported data for oxidized electrodes [23]. After galvanostatic oxidation treatment, the reversible potential of the anode shifted positively by approximately 1 V. Allowing the anodes to stay unpolarized in the electrolyte resulted in a slight decrease in exchange currents. During the relaxation process, the oxide scale underwent partial dissolution and recombination due to chemical reactions with the electrolyte penetrating the scale, eventually reaching a state referred to here as quasi-equilibrium. Additionally, relaxation led to a significant shift in the reversible potential towards negative values. The high correlation coefficients obtained for two of the first set of parameters (exchange current densities on a clean anode, Fig. 1a, R2 = 0.75, and on an electrochemically oxidized anode, Fig. 1b, R2 = 0.96) with the anode wear rate suggest that exchange current is linearly connected with the corrosion behavoir of gas-evolving anodes. This is more evident in the region of moderate to high wear rates: if we exclude the data point with the lowest wear rate, these correlation coefficients become 0.98 and 0.99, respectively.

Correlation Between Corrosion Rate and Electrochemical Parameters …

599

Fig. 1 Correlation between average wear rate of anode and average potentiodynamic electrochemical parameters recorded during linear sweep voltammetry: a exchange current density on clean anode, b same on oxidized anode after 1 h galvanostatic polarization at 0.1 A.cm–2,

c same on anode in a quasi-equilibrium state after 30-min relaxation, d– f are corresponding reversible potentials (at iox = ired) recorded with respect to W quasi-reference electrode

The correlation between the wear rate and reversible potential is much weaker. Although these parameters are not ideal for characterizing and predicting corrosion rates, one observation can be made. For a clean anode, an increase in the wear rate leads to a decrease in the reversible potential, whereas after galvanostatic oxidation, the dependency reverses, remaining consistent even after the relaxation time. This represents the shift between two different types of electrodes (from Me-Men+ to Me/MeO-Men+). The aforementioned galvanostatic oxidation and relaxation processes yield another set of kinetic parameters, which can also be plotted against the wear rate. The correlation between the wear rate and the second set of electrochemical parameters is presented in Fig. 2. The stationary potentials ranged from 1.5 to 1.8 V vs. W qRE (without compensation for Ohmic voltage drop), while the onset potential for oxygen evolution fell between 1.0 and 1.1 V. Relaxation times spanned from 120 to 480 s, and the relaxation rates varied between 1.9 and 6.1 V
s–1. There is a strong correlation between the stationary potential during galvanostatic low-current oxidation (Fig. 2a) and the wear rate of the anode, while the other two parameters

depicted in Fig. 2b and c did not exhibit any significant correlation. Electrochemical parameters, along with their corresponding empirical equations and correlation coefficients, are summarized in Table 2. Among all the measured parameters, three demonstrated significantly high correlations with the anode wear rate: the exchange current density on the clean anode surface, the same on the oxidized surface treated with low-current galvanostatic polarization, and the stationary potential during polarization. These parameters appear to represent two competing mechanisms of anode corrosion: (a) Metal fluorination/dissolution, and (b) Oxide dissolution. Both mechanisms result in the transfer of anode material into the bath, leading to the reduction of impurity metals at the cathode. The exchange current density on clean metal characterizes the bond energy between metal atoms in the lattice and describes the kinetics of metal oxidation and reduction when

600

A. Yasinskiy et al.

Fig. 2 Correlation between average wear rate of anode and average galvanostatic electrochemical parameters recorded during chronopotentiometric studies: a stationary (steady-state) potential (vs.

Table 2 Comparison of the correlations between anode wear rate and electrochemical parameters

Parameter

W quasi-reference electrode) achieved during galvanostatic polarization of anode for 1 h at 0.1 A
cm–2, b relaxation time after 1 h polarization measured during open circuit potentiometry, c relaxation rate dE/ds

Empirical equation

Correlation coefficient

Comment

Exchange current density on clean anode (i0cl)

1000i0 = 0.3WR + 0.40

0.750

Strong linear correlation

Exchange current density on oxidized anode (i0ox)

100i0 = 1.6WR − 2.13

0.961

Strong linear correlation

Exchange current density on relaxed anode (i0re)

100i0 = 1.19WR − 0.26

0.551

Weak correlation

Er = 0.057WR − 0.50

0.514

Weak correlation

Reversible potential on oxidized anode (Erox)

Er = -0.071WR + 0.94

0.375

No correlation

Reversible potential on relaxed anode (Erre)

Er = −0.059WR + 0.58

0.713

Linear correlation

Stationary potential (Est)

Est = −0.0487WR + 1.82

0.820

Strong linear correlation

sr = 4.5WR + 242.9

0.003

No correlation

dE/ds = 0.14WR + 3.62

0.024

No correlation

100△mel = 0.11WR + 2.47

0.051

No correlation

Reversible potential on clean anode (Ercl)

Relaxation time (sr) Relaxation rate (dE/ds) Mass gain, electrochemical (△mel)

it is in contact with the electrolyte and not fully protected by the oxide scale. At the same time, the oxide scale is continually dissolved in the electrolyte, and the dissolution products end up in the cathode metal as impurities. The exchange current density on the metal protected by the oxide scale may characterize the kinetics of oxide scale dissolution. The stationary potential during galvanostatic polarization correlates well with the wear rate. Anodes with low wear rates statistically exhibit higher stationary potentials, and vice versa. One possible explanation is that during low-current galvanostatic oxidation treatment, the electrolyte constantly penetrates the forming oxide scale and comes into contact with the clean anode surface. This leads to the appearance of partial currents involving metal dissolution,

oxidation, and fluoridation, in addition to the partial current of the primary reaction, which is the oxidation of oxygen ions on the surface of the oxide scale. The reversible potential for the primary reaction is more positive compared to metal oxidation, so the metal acts as a depolarizing agent, reducing the apparent potential for the entire process. Higher fractions of partial currents for metal oxidation in the total current lead to higher wear rates. Additionally, low-current galvanostatic oxidation may reveal the tendency of the anode composition towards passivation, as observed for Ni-rich compositions and discussed in future publications. Mass gain recorded after electrochemical measurements did not show any correlation with the wear rate due to the complexity of scale growth and dissolution processes.

Correlation Between Corrosion Rate and Electrochemical Parameters …

Although the net change in mass is often used as a parameter characterizing the corrosion rate, it was not proven to be reliable evidence in the presented study. Surprisingly, relaxation time and rate, which were expected to correlate with the wear rate due to their connection to the properties and structure of the oxide scale, electrolyte penetration, and oxide dissolution behavior, did not exhibit such a correlation. Similarly, the exchange currents recorded after the relaxation did not demonstrate a strong correlation with the wear rate. For further studies, steps #6 and #7 from Table 1 can be excluded from the strategy, as they do not provide valuable information for predicting anode wear rates or enhancing our understanding of the process.

Conclusion Estimating the wear rate of oxygen-evolving anodes is achievable through quick and straightforward electrochemical techniques. Notably, there exists a strong correlation between exchange current densities on both the metal and metal oxide surfaces and the 12-h wear rate of the anode. Furthermore, the stationary potential achieved during low-current galvanostatic polarization can serve as a valuable indicator of long-term anode performance, signaling whether the anode is prone to uncontrollable passivation during electrolysis. This proposed technique offers an efficient means of initially characterizing numerous anode compositions and conditions, mitigating the need for prolonged periods and significant material resources typically associated with long-term electrolysis tests. Acknowledgements This work was financed by the Ministry of Economy, Industry, Climate Protection, and Energy of the State of North Rhine-Westphalia within the project ‘CO2-free Aluminium Production’ with the Grant EFO0113D.

601

References 1. K. Du, E. Gao, C. Zhang et al. Nat Commun 14, 253 (2023). 2. A. Suzdaltsev and Y. Zaikov. J. Electrochem. Soc. 170 056506 (2023). 3. A. A. Polyakov et al. J. Electrochem. Soc. 169 053502 (2022). 4. S. K. Padamata et al. J. Electrochem. Soc. 170 073501 (2023). 5. A. S. Yasinskiy, S. K. Padamata, P. V. Polyakov, and A. V. Shabanov, Non-ferrous Metals., 48, 15 (2020). 6. Y. He, K. Zhou, Y. Zhang, H. Xiong, and L. Zhang, J. Mater. Chem. A, 9, 25272 (2021). 7. S. Feng, et al. Chemical Engineering Journal, 472, 145010 (2023). 8. Z. Zhang, J. Xu. Int J Appl Ceram Technol., 20: 329–340 (2023). 9. S. Yaroshevskyi, et al., Open Ceramics, 16, 100458 (2023). 10. Y.-Q. Tao et al. Materials, 15, 5377 (2022). 11. W. Wei et. al. Journal of Materials Science & Technology, 107, 216-226 (2022) 12. S. Jucken, B. Tougas, B. Davis, D. Guay, and L. Roué, Metall. Mater. Trans. B, 50, 3103 (2019). 13. E. Gavrilova, G. Goupil, B. Davis, D. Guay, and L. Roué, Corros. Sci., 101, 105 (2015). 14. V. Jalilvand, et al. Surface and Coatings Technology, 441, 128576 (2022). 15. I. Gallino, M. E. Kassner, and R. Busch, Corros. Sci., 63, 293 (2012). 16. G. Gunnarsson, G. Óskarsdóttir, S. Frostason, and J. H. Magnusson, ed. C. Chesonis “Light Metals.” The Minerals, Metals & Materials Series. (Springer, Cham, Berlin) (2019). 17. S. K. Padamata, A. Yasinskiy, A. Shabanov, T. Bermeshev, Y. Yang, Z. Wang, D. Cao, and P. Polyakov, Trans. Nonferrous Met. Soc. China, 32, 354 (2022). 18. K.-C. Zhou et al. Zhongguo Youse Jinshu Xuebao/Chinese Journal of Nonferrous Metals, 31 (11), 3010–3023 (2021). 19. S. K. Padamata, A. Yasinskiy, and P. Polyakov, Trans. Nonferrous Met. Soc. China, 30, 1419 (2020). 20. T. R. Beck, C. M. MacRae, and N. C. Wilson, Metall. Mater. Trans. B, 42, 807 (2011). 21. A. P. Apisarov et al., Russ. J. Electrochem., 46, 633 (2010). 22. H. Andrews, S. Phongikaroon, J. Electrochem. Soc. 165 E412 (2018). 23. G. Saevarsdottir et al., J. Electrochem. Soc. 170 072508 (2023).

Direct Production of Al-Mn Alloys During the Electrodeposition of Aluminum in a Laboratory Cell Omar Awayssa, Geir Martin Haarberg, Gudrun Saevarsdottir, and Rauan Meirbekova

Abstract

Introduction

This study reports the direct production of aluminum– manganese alloys during the electrodeposition of aluminum in cryolitic melts. For the purpose of measuring current efficiency, experiments were conducted in a laboratory cell. The temperature was changed between 960 and 980 °C at a cryolite ratio (CR) of 2.2 and a cathodic current density (CCD) of 0.9 A/cm2. Up to 3.0 weight percent of manganese was present. Mn2O3 was used as a method of manganese addition. To track the dissolution of manganese during electrolysis, bath samples were routinely taken, and ICP-MS was examined. Al-Mn alloy electrodeposition current efficiency was estimated to be in the region of above 90%. Estimates of aluminum's current efficiency were made. The metal deposits’' hardened surfaces were essentially flat, but some were deformed. Keywords

Aluminum–manganese alloys Aluminium electrodeposition




Current efficiency

O. Awayssa (&) Department of Chemical and Petroleum Engineering, UAEU, 15551 Al Ain, United Arab Emirates e-mail: [email protected] G. M. Haarberg
G. Saevarsdottir Department of Materials Science and Engineering, NTNU, Sem Sælands Vei 12, NO-7491 Trondheim, Norway G. Saevarsdottir School of Science and Engineering, Reykjavik University, Menntavegi 1, 101, Reykjavik, Iceland R. Meirbekova Tæknisetur, Árleynir 8, 112 Reykjavík, Iceland




According to the overall electrochemical reaction described by [1], liquid aluminum (Al) is produced in the Hall–Héroult process by electrolytic reduction of alumina (Al2O3) dissolved in an electrolyte including cryolite (Na3AlF6) at 960– 970 °C: 2Al2 O3 ðdissolved Þ þ 3CðsÞ ¼ 4AlðlÞ þ 3CO2 ðgÞ

ð1Þ

The 3xxx family of aluminium alloys’’ main alloying component is manganese (Mn). The alloy has a greater corrosion resistance and is significantly stronger than commercially available pure aluminium when a small amount of Mn (up to 1.5 wt%) is added to Al. Due to advancements in its mechanical characteristics, the alloy has become suitable for widespread usage in applications requiring good workability and moderate strength [2, 3]. Manganese dissolves in molten aluminium extremely slowly, and this is greatly influenced by the particle size of the manganese that has been introduced [4]. When powdered manganese is introduced to molten aluminium, some of it may float to the surface and create a hard crust, which indicates that part of it may have undergone oxidation [4]. The possibility to produce aluminum–manganese alloys right in cryolitic melts has been reported [5] such that by incorporating MnO, MnO2, or their mixes with alumina in a cryolite-based electrolyte, it has been claimed that the produced alloys contained up to 10 weight percent Mn. As opposed to Al12Mn, which is more likely to develop at a manganese level of 7.7 at. % at T *511 °C, an Al6Mn phase is more likely to form with a manganese value of 14.3 at. % at *T 658 °C [6]. When Mn2O3 is utilized as a precursor for manganese, the reduction process will proceed as follows: Mn3 þ þ 3e ¼ Mn

ð2Þ

Current efficiency measures may be used to assess the utilization of supplied electrical current used to deposit

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_77

602

Direct Production of Al-Mn Alloys During the Electrodeposition of Aluminum in a Laboratory Cell

aluminium which may be estimated by metal weight gain relating the actual produced aluminum to the aluminum that would theoretically be produced based on Faraday’s law. Then CE% may be written as CE% ¼

Wactual Wactual  100  100 ¼ Wtheoretical MIt=zF

ð3Þ

where Wactual is the actual mass of metal produced whereas Wtheoretical is the theoretical mass of metal produced according to Faraday's law. M is the molar mass of aluminum, I is the applied current intensity in A, z is the number of electrons transferred, and F is the Faraday constant of 96,487 C/mol. In practical terms, it is impossible to acquire the amount of aluminium predicted by Faraday's law. In the electrolyte, aluminum always dissolves to some extent. The metal is thus transferred to a location near the cathode outside the diffusion layer where it is oxidized by CO2. During the back reaction, alumina is produced and carbon monoxide is released, as per the following: 2Alðdiss:Þ þ 3CO2 ðgordiss:Þ ¼ Al2 O3 ðdiss:Þ þ 3COðgÞ ð4Þ At the cathode, dissolved impurity species that are more noble than aluminium will undergo reduction [8]. The decrease in current efficiency for aluminum reduction is caused by the current utilized to co-deposit such contaminants as Mn. The alloy's average current efficiency may be estimated using CEalloy % ¼

W alloy  100 W alloy:theoretical

ð5Þ

where Walloy is the total mass of metal produced experimentally whereas Walloy. theoretical is the theoretical mass of alloy produced. The theoretical mass of the produced alloy is given by Faraday’s law as W alloy:theoretical ¼

M alloy It zalloy F

ð6Þ

where Malloy is the average molecular mass of the alloy and zalloy is the average charge transferred for the deposition of the alloy. The two quantities may be estimated for the Al-Mn alloy according to the so-called electrochemical equivalent given by h ih i M Al M Mn zAl : zMn    W equiv: ¼  ð7Þ xAl MzMnMn þ xMn MzAlAl Thus CE % for the alloy can be given by

CE%alloy ¼

603

W alloy  100 W equiv: FIt

ð8Þ

where M Al, M Mn, z Al, z Mn,x Al, and x Mn are the molecular masses of Al and Mn, their charges, and their mass fractions, respectively. This investigation presents findings from a study into the electrochemical deposition of an aluminum-manganese alloy during aluminium reduction in fluoride-based melts in a laboratory cell using industry standards. It is examined how the presence of Mn affects the current efficiency relative to Al, the current efficiency for the alloy, and the texture of the deposit's surface once it has hardened.

Experimental Experiments were carried out in a laboratory cell originally designed by Solli et al. [7] for current efficiency measurements during electrodeposition. The laboratory cell is schematically illustrated in Fig. 1. A graphite crucible with cylindrical sintered alumina side lining of about 10 cm height containing anode, cathode, and electrolyte was used. The anode is cylindrical with a central vertical hole passing through it with an inward inclination angle of 10° as well as horizontal holes penetrating the anode. This design provides good convection within the bath so that gas bubbles from the anode pass through the central vertical hole in the bottom allowing electrolyte to flow up and through the horizontal holes on the sides causing the electrolyte to circulate in a loop. By that, the gas bubbles would have less effect on the diffusion layer and thus the current efficiency would not be significantly affected by increased convection. The liquid aluminium metal product wets a steel plate resting on the bottom of the graphite crucible and acts as a cathode which ensures an almost flat deposit surface and as a result an even current distribution. A steel pin of 21.0 mm height is placed in a 4.0-mm-deep hole at the center of the bottom of the graphite crucible to make contact with the steel cathode plate. The latter is placed on top of a layer of alumina powder after cementing the bottom of the crucible with a layer of cast alumina cement of 7.0 mm thickness. These two layers should prevent loss of the deposit and minimize chances of aluminium carbide (Al4C3) formation. The electrolyte constituents as shown in Table 1 were transferred into the crucible after being dried at 200 °C for 24 h. The cell was then placed in a Pythagoras tube inside a vertical furnace. Two copper lids with greased rubber O-rings were used to seal up the two ends of the furnace making it gas tight. The anode was placed in the bath and held by a steel current collector. The furnace was continuously flushed with argon gas during the experiment in order to prevent air burning of

604

O. Awayssa et al.

Fig. 1 The design of the CE laboratory cell used in this work

Anode Conductor

Alumina Side lining Thermocouple Graphite Anode

Graphite Crucible Electrolyte Steel Plate Alumina

Table 1 Electrolyte components

Cement

Chemicals

Pre-treatment

Quality/supplier

AlF3

Sublimed at 1090 °C for 24h

Industrial grade, Alcoa-Norway

NaF

Dried at 200 °C for 24 h

99.5%, Merck-Germany

CaF2

Dried at 200 °C for 24 h

Precipitated pure, Merck-Germany

Al2O3

Dried at 200 °C for 24 h

Anhydrous (c-alumina), Merck-Germany

Mn2O3

Dried at 200 °C for 24 h

325 Mesh powder, 98%, Alfa Aesar-Germany

cell components. The temperature was recorded during electrolysis using a thermocouple made of Pt/Pt10Rh placed inside a lateral slot of the crucible. A DC power supply was used to supply the current. The operating temperature was varied from 965 to 980 °C with a fixed electrolysis duration of 4 h. The corresponding superheat was varied from 13.0 to 28.0 °C, being calculated from an equation in [10]. The cathodic current density (CCD) was kept at 0.9 A/cm2 for all runs. A cryolite ratio (CR) of 2.2 was used for all runs. The standard electrolyte was 12.0 wt. % AlF3, 5.0 wt. % CaF2, 4.0 wt.% Al2O3, and balance of NaF-AlF3-based cryolite. Manganese (III) oxide was initially admixed with the bath constituents prior to electrolysis. Three concentrations were considered based on Mn content which were 1 wt. % Mn, 2 wt. % Mn, and 3 wt. % Mn. The bath was sampled regularly at constant intervals using quartz tubes while keeping the same position of the sampling in the bath for all runs. The collected metal samples were subjected to mechanical and chemical post-treatments, the latter by aluminium chloride hexahydrate solution for 30–40 min. Bath samples were crushed

into fine powder and dissolved in a mixture of strong acids including HCl, HNO3, and HF. The solutions were digested and agitated to ensure a complete dissolution. ICP-MS was conducted for samples afterwards to determine the Mn content in the bath.

Results and Discussion Manganese Addition Mn2O3 was admixed into the bath before melting. Three concentrations were considered: 1.0 wt. % Mn, 2.0 wt. % Mn, and 3.0 wt.% Mn. Temperatures were 965 °C, 970 °C, 975 °C, and 980 °C.

Bath Analysis Baths for experiments conducted at 965 °C and an initial content of Mn added 1 wt.% and 3 wt. % were analysed for Mn content. As seen in Fig. 2, around 45 % of Mn dissolved depleted during the first half of the experiment (120 min) at 965 °C whereas 21% depleted at 980 °C.

Direct Production of Al-Mn Alloys During the Electrodeposition of Aluminum in a Laboratory Cell

Fig. 2 Decay of Mn in the bath at 1 wt. % content at 965 °C and 980 °C

The electrodeposition in a laboratory cell may be considered a semi-batch process where the rate of depletion of Mn in the bath at any time may be expressed in the form of   A ð9Þ c ¼ co exp  kt V where co is the initial concentration of the impurity species prior to addition, A is the active surface area of the cathode, V is the electrolyte volume, k is the mass transfer coefficient, and t is the time at which the sample has been taken out of the cell.

Electrodeposits Analysis According to the results, regardless of the operating temperature, a rise in the starting concentration of Mn caused an increase in the metal's Mn content to be seen. The ultimate amounts of Mn in the metal at starting bath concentrations of 1.0 wt% and 2.0 wt% Mn were approximately 8.0 wt% and 13.0 wt%, respectively, independent of the operating temperature, suggesting that the latter may have had less of an impact on the solubility of manganese in the bath.

Cell Performance Current Efficiency for Aluminium When examining the impact of Mn in the deposit on CE, blank tests would be a useful point of comparison. Based on the net weight of aluminium present in the solidified deposit, current efficiency for aluminium is calculated. That indicates that the weight of the co-deposited manganese will be subtracted from the overall weight of the deposit after cleaning. An overview of the actual current efficiency for all

605

Fig. 3 Summary of actual CE % for Al at different temperatures and Mn contents initially added to the bath

temperatures at various initial Mn contents supplied to the bath is shown in Figure 3. The highest current efficiency of Al was attained at 965 °C and 3.0 wt% Mn originally added to the bath, whereas the lowest was at the same Mn concentration at 980 °C.

Current Efficiency for Al-Mn Alloys Equations (5–8) were used to estimate the average current efficiencies of Al–Mn alloys. According to each element's proportion in the alloy, the average current efficiency serves as a representation of the current efficiency of the alloy as a whole. Figure 4 provides the actual current efficiency for aluminium and the average current efficiency of aluminum– manganese alloys. The values obtained for Al-Mn current efficiency were comparable to those for blank tests under the same scenarios, which may imply the viability of the approach suggested.

Electrodeposits Shape The solidified deposits’ surfaces of all blank tests were flat as depicted in Fig. 5. The solidified deposits’ surfaces were flat when experiments were carried out under the additions of 1 wt. % Mn and 2 wt. % Mn at 965 °C, 970 °C, 975 °C, and 980 °C. However, the deposits’ surfaces of runs at 3 wt. % Mn at 970 °C and 980 °C were deformed as seen in Fig. 6. It is worth mentioning that the shape of the electrodeposit is related to the current distribution during the electrolysis. Flat surfaces assure even current distribution, and thus reliable current efficiency measurements. Similar behavior has been reported in similar experiments but with different alloying element additions [8–10].

606

Fig. 4 Actual CE% for Al and average CE % for Al-Mn at initial added Mn

Fig. 5 Blank tests using NaF-AlF3 cryolite with no alumina feeding at CR = 2.2, CCD = 0.9 A/cm2, t = 4 h, (left) T = 965 °C,(right) T = 980 °C

Fig. 6 Deposits using NaF-AlF3 cryolite with no alumina feeding at 3 wt. % Mn, CR = 2.2, CCD = 0.9 A/cm2, t = 4 h, (left) T = 970 °C, (right), T = 980 °C

O. Awayssa et al.

Direct Production of Al-Mn Alloys During the Electrodeposition of Aluminum in a Laboratory Cell

607

Conclusions

References

In a laboratory cell designed for studies of the aluminum current efficiency, the co-deposition of manganese to yield Al-Mn alloys was investigated. ICP-MS analysis suggests that regardless of the operating temperature, increasing the initial concentration of manganese precursor supplied to the bath brings about an increase in the amount of Mn in the metal. The average current efficiencies of Al-Mn alloys have a difference of up to 9% in comparison to those estimated for the actual deposition of Al which implies that this path is quite efficient to produce such alloys. All of the blank tests' hardened deposits had flat surfaces, which indicated even current distribution. For the majority of runs at various Mn contents and operating temperatures, the co-deposition of Mn had little impact on the surface morphology of the solidified deposits. High Mn2O3 concentration correlates to 3.0 wt% Mn that was added to the bath at 970 °C and 980 °C, resulting in the deformation of the deposits' surfaces.

1. Grjotheim K, Kvande H (1993) Introduction to Aluminium Electrolysis : understanding the Hall-Héroult process, 2nd ed., Düsseldorf: Aluminium-Verlag. 2. Davis J (2001) Alloying: Understanding the Basics, ASM International.351–416. 3. Davis J (1998) Aluminum and Aluminum Alloys, Metals Handbook Desk Edition, 2nd ed., ASM International. 417–505. 4. Kline J, Yeh W, Preston U (1975) Method of adding manganese to aluminium. USA Patent 3,865,583. 5. King W ( 1976) Aluminium-Manganese Alloy. USA Patent 3,951,764. 6. Solli P, Eggen T, Rolseth S, Skybakmoen E (1996) Design and performance of a laboratory cell for determination of current efficiency in the electrowinning of aluminum. Journal of Applied Electrochemistry.1019–1025. 7. Solheim A, Rolseth S, Skybakmoen E, Støen L, Sterten Å, Støre T (1996) Liquidus Temperatures for Primary Crystallization of Cryolite in Molten Salt Systems of Interest for Aluminum Electrolysis. Metall. Mater. Trans. B. vol. 27B,739-744. 8. Awayssa O, Meirbekova R, Saevarsdottir G, Audunsson G, Haarberg G M (2020) Current efficiency for direct production of an aluminum–titanium alloy by electrolysis in a laboratory cell. In: Tomsett A. (eds) Light Metals. The Minerals, Metals & Materials Series. Springer, Cham. 9. Awayssa O, Saevarsdottir G, Meirbekova R, Haarberg G M (2021) Electrochemical Production of Al-Si Alloys in Cryolitic Melts in a Laboratory Cell. Journal of The Electrochemical Society, 168 046506 10. Awayssa O, Saevarsdottir G, Meirbekova R, Haarberg G M (2021) Electrodeposition of aluminium-titanium alloys from molten fluoride-oxide electrolytes. Journal of The Electrochemistry Communications, 123 106919.

Acknowledgements Financial support from Start-up Fund number 12N132-UAEU and NTNU is greatly acknowledged.

Electrowinning of Al-Sc Master Alloys in the LiF-AlF3-Sc2O3 Melts Andrey Yasinskiy, Ilya Moiseenko, Dmitriy Varyukhin, Anastasia Saparova, Aleksandr Samoilo, Pavel Yuryev, Youjian Yang, Zhongning Shi, Zhaowen Wang, Peter Polyakov, and Bernd Friedrich

Abstract

Introduction

Aluminum–scandium master alloys are highly demanded products used to create multi-functional aluminum alloys and composites. A high cost of Al-Sc master alloys stops the automotive industry from using them widely. This research investigates the possibility to produce Al-Sc master alloys via electrolysis of the LiF-AlF3-Sc2O3 melt. The kinetic parameters of the aluminum and scandium electrowinning were studied by means of voltammetry, stationary polarization, and electrolysis tests. The apparent limiting current density for co-deposition of Aluminum and Scandium on tungsten cathode was in the range from 1.28 to 1.97 A.cm–2 in the temperature range from 860 to 940 °C. Based on the electrochemical measurements, the parameters for galvanostatic electrolysis were selected and electrolysis tests were carried out to obtain Al-Sc master alloys. The microstructure of the obtained Al-Sc master alloys was studied. It was possible to obtain Al-Sc alloys with the concentration of Sc 0.68 wt.%. Keywords














Electrochemical co-deposition Electrowinning Al-Sc Master alloy Molten salt Electrolysis

A. Yasinskiy (&)
I. Moiseenko
D. Varyukhin
A. Saparova
A. Samoilo
P. Yuryev
P. Polyakov Laboratory of Physics and Chemistry of Metallurgical Processes and Materials, Siberian Federal University, 660025 Krasnoyarsk, Russia e-mail: [email protected] A. Yasinskiy
B. Friedrich IME Process Metallurgy and Metal Recycling, RWTH Aachen University, 52056 Aachen, Germany A. Yasinskiy
Y. Yang
Z. Shi
Z. Wang Key Laboratory for Ecological Metallurgy of Multimetallic Minerals (Ministry of Education), Northeastern University, Shenyang, 110819, China

The advancement of technology is closely intertwined with the development of innovative materials and manufacturing techniques. Among these, considerable attention is directed toward alloys and composite materials based on aluminum, which hold great promise in various applications [1–3]. Specifically, aluminum–scandium master alloys have garnered significant interest due to their role in producing multi-functional aluminum alloys and composites [4, 5]. Al-Sc alloys find extensive applications in the manufacturing of sports equipment, including baseball and softball bats, bicycle frames, lacrosse sticks, and tent poles. Their exceptional strength properties make them the material of choice for these items. In cutting-edge sectors, such as aerospace, both NASA and Airbus are actively assessing Al-Sc alloys for potential use in aircraft components [6, 7]. However, their widespread adoption, particularly in the automotive industry, is hindered by the high production costs associated with existing methods. Currently, Al–Sc master alloys are primarily manufactured through an aluminothermic reduction process, involving scandium salts beneath a layer of salt flux at approximately 900 °C [4]. This process requires costly scandium fluoride and pure aluminum as raw materials, encompassing expenses for their production and transportation within the overall production cost of Al-Sc master alloys. Additionally, a notable drawback of this method is the rapid accumulation of oxides within the salt flux, necessitating periodic replacement or regeneration. An alternative and more cost-effective approach to producing Al-Sc master alloys involves utilizing relatively inexpensive Sc2O3 within alumina reduction cells according to the equations: Al2 O3 ðdisÞ þ 3=2CðsÞ ¼ 2AlðlÞ þ 3=2CO2 ðgÞ

ð1Þ

Sc2 O3 ðdisÞ þ 3=2CðsÞ ¼ 2ScðdisÞ þ 3=2CO2 ðgÞ

ð2Þ

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_78

608

Electrowinning of Al-Sc Master Alloys in the LiF-AlF3-Sc2O3 Melts

3Al2 O3 ðdisÞ þ Sc2 O3 ðdisÞ þ 4C ¼ 2Al3 ScðlÞ þ 4CO2 ðgÞ

ð3Þ

This method eliminates several steps such as cooling, transportation, and subsequent heating of the resulting pure aluminum and obviates the need for an additional reactor dedicated to Al-Sc master alloy production [8, 9]. However, this technology is characterized by high operating temperatures (950–960 °C), limited scandium extraction, the presence of hazardous gaseous byproducts, and overall low energy efficiency. Electrowinning of elemental scandium and the Al3Sc intermetallic compound occurs at potentials roughly 0.44 V and 0.31 V more negative than aluminum electrowinning [4]. Several alternative methods have been proposed to address these challenges, including the electrolytic production of Al-Sc alloys in various molten salt compositions at lower temperatures [10–14]. Among these approaches, electrolysis in KF-AlF3- and NaF-AlF3-based melts with Al2O3 and Sc2O3 additives at temperatures ranging from 700 to 900 °C, using a liquid aluminum cathode and carbon anode, has shown promise [11, 15–17]. Recent pilot-scale implementations of these methods have further supported their feasibility [18]. The occurrence of potassium electrowinning, facilitated by depolarization in the presence of scandium, may be attributed to the potential formation of intermetallic compounds. This situation complicates the precise estimation of kinetic parameters, given that scandium electrowinning coincides with aluminum electrowinning, and there's a likelihood of forming intermetallic compounds involving scandium, aluminum, and potassium. Under conditions of stationary polarization, the addition of Sc2O3 oxide to the KF-AlF3 melt leads to a notable increase in the limiting current density for both aluminum and scandium electrowinning [4]. This increase is approximately from 0.45 to 0.81 A cm−2. Simultaneously, potassium electrowinning becomes evident at a potential of approximately −0.95 V and remains relatively unchanged. The limiting current density for aluminum electrowinning from the KF-AlF3 melt ([KF]/[AlF3] = 1.5 mol/mol) saturated with Al2O3 is around 0.70 A cm−2 [19]. Another possible electrolyte potentially suitable for electrowinning of Al-Sc master alloys is LiF–AlF3-based. Although it may help overcome some drawbacks related to potassium co-deposition due to wider electrochemical window, it was not studied extensively. Additionally, the electrical conductivity of LiF-based melts is higher compared to NaF- and KF-based ones. Electrolytes containing portions of LiF in 2.2NaF/AlF3-3 wt.% Al2O3-5 wt.% LiF-2 wt.% Sc2O3 were used to prepare alloys with 0.14–0.41 wt.% Sc at current efficiency of 88.7% and 79.27% for cathodic current density of 0.5 and 1 A
cm−2, respectively [20]. Guo et al. [21] used the LiF-ScF3-ScCl3–Sc2O3 at 850 °C with Al

609

cathode to successfully produce Al-Sc alloys with maximum of 3.437% of Sc content in the alloy. The studies of Al reduction process in LiF-based melts [22–25] showed that the mechanism of aluminum reduction in LiF-based melts is complicated and deserves further study to find the optimal process parameters. In light of these considerations, it becomes imperative to delve into the kinetics of electrode processes within LiF-AlF3-based melts. The present study aims to contribute to this understanding by investigating the kinetic characteristics of aluminum and scandium electrowinning from LiF-AlF3 melts containing Sc2O3 additives during the synthesis of Al-Sc master alloys.

Experimental Methods Preparation of Electrolytes The molten electrolytes were meticulously prepared using high-purity, dried fluorides. To achieve the desired composition, a precise quantity of Sc2O3 was introduced into the melt using a steel tube. The total mass of the melt in the cell was approximately 500 g.

Electrochemical Measurements The electrochemical measurements were conducted within a three-electrode graphite cell, in an argon atmosphere, across a temperature range of 860–940 °C. A finely grained, graphite rod served as the auxiliary electrode. Tungsten rods with a 2 mm diameter were employed as working electrodes. The potential of the working electrode was measured relative to the potential of the tungsten quasi-reference electrode. The stationary polarization characteristics were determined by recording the steady-state electrode potential as a function of the supplied cathode current. This was achieved using PGSTAT AutoLab 302N equipment in conjunction with NOVA 2.1.6 Software, both from The MetrOhm (the Netherlands). Additionally, voltammograms were generated by varying the potential sweep rate at 900 °C. To ensure precise measurements, the “I-Interrupt” procedure was employed to identify and offset any ohmic voltage drop in the measurement circuit. The melt's temperature was monitored using a K-type thermocouple, with fluctuations limited to ± 2 °C. A series of short-term (2–4 h) electrolysis tests were performed in a galvanostatic mode within a two-electrode graphite cell with varying contents of aluminum and scandium oxides, at current densities of 0.12 A
cm−2, 0.6 A
cm−2, and 1.15 A
cm−2. The tests were conducted at temperatures exceeding the liquidus temperature, specifically at 860 °C and 910 °C.

610

Analysis

A. Yasinskiy et al.

scanning electron microscope EVO50 Carl Zeiss with the Energy350 energy-dispersive spectrometer.

The electrolyte analysis was conducted using X-ray phase analysis. X-ray diffraction patterns were recorded on a Shimadzu XRD7000 X-ray diffractometer, Japan, with CuKa radiation and a monochromator. The scanning range was set from 5 to 70° on the 2h scale, with a step size of 0.03°, and a scanning speed of 1.5 degrees per minute. The prepared finely dispersed powder sample was manually pressed into a standard quartz glass cuvette. X-ray powder diffraction analysis (XRD) was performed using the Information Retrieval System for XRD (IRS XRD) program. This analysis utilized the PDF2 database of XRD standards for phase identification and quantitative XRD (QXRD) through the multi-reflex method known as “corundum numbers”. The corundum numbers of phases (Imax/Icor) were used to normalize the intensity (I) of XRD standard lines to a common scale (corundum intensity) and are available in the PDF2 database. The analysis of the cathodic metal was carried out using the laboratory spectrometer FOUNDRY-MASTER LAB. The detection and measurement of the intensity of emission lines in the plasma's spectral composition were performed using CCD detectors. The analysis of the emission spectrum enabled the determination of the elemental composition of the sample under examination. The assessment of whether the metal met specified requirements, its grade, and deviations from the required standards was determined by comparing the obtained data with the reference database integrated into the instrument's software. The elemental composition of the inclusions in the structure of the metal sample was determined using a

Cyclic voltammetry (CV) allows the study of the kinetics of the electrode process over a wide range of potentials to establish the potential difference between the onsets of Al and Sc. Furthermore, by analyzing the current–potential curve, one can draw important conclusions about the mechanism of reduction and oxidation, which can be used to enhance the electrowinning process. To determine the electrochemical window and study the kinetic parameters of both Al and Sc reduction, such a study was performed at a temperature of 900 °C. Figure 1 represents the results of cyclic voltammetry conducted in a 64LiF-36AlF3 melt with an initial addition of 2 wt.% Sc2O3. Using potential sweep rates ranging from 0.1 to 0.8 V
s−1, several electrochemical processes were identified. Aluminum electrowinning initiates at potentials more negative than approximately -0.06 V relative to the potential of the quasi-reference electrode (“Red 1” wave in Fig. 1b), while the “Red 2” deposition peak emerges in the potential range of −0.25 to −0.35 V. At potentials more negative than the “Red 2” peak, a negative current associated with aluminum electrowinning persists. The cathodic wave “Red 3”, appearing at around −0.70 V, is likely associated with Sc reduction on the Al surface. The potential window between Al and Sc varies between 0.42 and 0.48 V, which agrees with previously reported values [4]. The potential difference between onsets for Sc and Li is more challenging to recognize. Still, a rough estimation gives

Fig. 1 Results of CV performed in 64LiF-36AlF3 (2 wt.% Sc2O3) with W working electrode at 900 °C: a dependency between peak potentials and ln(m) for three reduction and three oxidation peaks (waves),

b cyclic voltammograms recorded at four different scan rates, c dependency between peak current densities and square root of scan rate for abovementioned six peaks (waves)

Results and Discussion

Electrowinning of Al-Sc Master Alloys in the LiF-AlF3-Sc2O3 Melts

the value around 0.1 V, which places the undesirable co-deposition of Li within the bounds of possibility. The current peaks “Red 1” and “Red 2” are likely both associated with Al reduction. The potential difference is between 0.13 and 0.20 V. One possible explanation is that the first wave represents the reduction of Al on W with the formation of an intermetallic compound, while the second peak manifests the reduction of Al on the Al surface when there is no uncovered W substrate left. On the other hand, the corresponding oxidation peaks (“Ox 1” and “Ox 2”) are very pronounced, suggesting that the reason may be different, namely, a two-step reduction process with a two-electron transfer on the first step and a one-electron transfer on the second step. However, the potential difference for this case is too high, making this phenomenon deserving of further study. The oxidation peak “Ox 3” is likely related to the oxidation of Sc. Analyzing cyclic voltammograms in terms of plotting peak potentials against ln(m) (Fig. 1a) and peak currents against the square roots of sweep rates (Fig. 1c) allow us to delve deeper into the mechanism of the electrode process. A linear correlation between ip and c1/2, and Ep remaining unchanged at different sweep rates, indicates a diffusioncontrolled process. Both conditions are fulfilled only for the oxidation of Sc and the “Ox 2” process for Al. Without further studies, it is impossible to arrive at a definitive conclusion regarding the mechanism. From a technical standpoint, calculating diffusion coefficients for these processes based on the CV data is not feasible. Still, one can calculate apparent diffusion coefficients D* using the Randles–Ševčík equation, which is normally used to calculate

611

the true diffusion coefficients of electroactive species when the product is soluble:


v
D ip ¼ 0; 4463
ðz
FÞ
C
R
T 3 2

12

ð4Þ

where z is the number of electrons participating in the reaction; F is the Faraday constant, equal to 96,485 C/mol; C is the ion concentration, in mol
cm–3; v is the sweep rate in V
s−1; D is the diffusion coefficient in cm2
s−1; R is the universal gas constant, equal to 8.314 J
(K
mol)−1; and T is the absolute temperature in Kelvin (K). All kinetic parameters collected from cyclic voltammetry are listed in Table 1 with apparent diffusion coefficients calculated for three different electroactive species. The apparent diffusion coefficients for aluminum ions spanned from 0.6
10–7 to 1.2
10–6 cm2
s−1. In contrast, for scandium ions, they were 0.2
10−5 cm2
s−1. In the galvanostatic potentiometry, we gathered data on how the cathodes potential changes over time. These findings allowed us to establish relationships between potential and current density. The results obtained in chronopotentiometry in the similar electrolyte at 860–940 °C are presented in Table 2. Notably, high limiting current densities were observed. In the temperature range of 860 to 940 °C, the limiting current density ranged from 1.28 to 1.97 A
cm−2. The mass transfer coefficients (kc) were in the range from 0.49
10–4 cm
s−1 to 0.76
10–4 cm
s−1, which is in better agreement with theoretical values for various molten salt systems compared to the diffusion coefficients obtained in cyclic voltammetry (CV).

Table 1 Kinetic parameters for reduction and oxidation of Al and Sc on W electrode Scan rate, V
s−1

ip, A
cm−2

D*
105 cm2
s−1

Ep, V w.r.t. qRE

Reduction peaks m

m1/2

0.1

0.32

0.2

0.45

0.4

0.63

−0.92

0.8

0.89

−0.22

ln(m)

Red1

Red2

Red3

Red1

Red2

Red3

Red1

Red2

Red3

−2.30

−0.32

−0.49

−0.25

−0.14

−0.27

−0.67

0.009

0.162

0.339

− 1.61

−0.37

−0.63

−0.32

−0.14

−0.31

−0.68

0.006

0.134

0.278

−0.46

−0.79

−0.39

−0.14

−0.34

−0.7

0.004

0.105

0.206

−0.68

−1.04

−0.46

−0.15

−0.34

−0.7

0.005

0.091

0.144

Average

0.006

0.123

0.242

Standard deviation

0.002

0.027

0.074

Oxidation peaks m

m1/2

ln(m)

Ox 1

Ox 2

Ox 3

Ox 1

Ox 2

Ox 3

Ox 1

Ox 2

Ox 3

0.1

0.32

−2.30

1.05

0.70

0.20

0.05

−0.12

−0.62

0.093

0.330

0.217

0.2

0.45

−1.61

1.19

0.86

0.24

0.05

−0.12

−0.61

0.060

0.249

0.156

0.4

0.63

−0.92

1.61

1.19

0.37

0.05

−0.12

−0.61

0.055

0.238

0.186

0.8

0.89

−0.22

2.31

1.47

0.43

0.07

−0.12

−0.6

0.056

0.182

0.125

Average

0.066

0.250

0.171

Standard deviation

0.016

0.053

0.034

612 Table 2 Kinetic parameters of the Al-Sc co-deposition obtained from stationary galvanostatic polarization

A. Yasinskiy et al. T, K

ilim A
cm−2

ires A
cm−2

kc
104, cm
s−1

860

1.28

0.29

0.49

900

1.52

0.43

0.58

940

1.97

0.46

0.76

In the stationary polarization studies, high residual currents (ires) were recorded before reaching the potential of diffusion-controlled co-deposition of Al and Sc. This current increased with an increase in temperature and was in the range from 0.29 to 0.46 A
cm−2. Electrolysis experiments demonstrated the promise of lithium cryolite as an electrolyte for simultaneous aluminum and scandium electrodeposition on the cathode. After a 2-h electrolysis session at a current density of 1.15 A/cm2, we produced an alloy containing approximately 0.7 wt.% scandium. A microstructural analysis of the resultant sample suggests the possible formation of aluminum–scandium intermetallic compounds. The microstructure of the examined sample reveals three zones: an extensive zone of the metallic base, a transitional zone with sizes ranging from 180 to 280 µm, and a metallic “droplet” zone with aluminum as the main component. The structure of the “droplet” consists of an aluminum base with inclusions of large plate-like intermetallics. The length of the large plates ranges from 93 to 432 µm, while the smaller ones fall within the 41–62 µm interval. Cross sections of the “droplet” reveal clusters of irregularly shaped smaller inclusions, measuring 5–10 µm.

Conclusion This study explores the kinetics of aluminum and scandium electrowinning from LiF-AlF3 melts with Sc2O3 additives, shedding light on the synthesis of Al-Sc master alloys. Cyclic voltammetry was employed to delve into the kinetics of the electrode process. Aluminum electrowinning, scandium electrowinning, and the potential formation of intermetallic compounds were analyzed. Results indicate that scandium electrowinning coincides with aluminum electrowinning, complicating precise kinetic parameter estimation. Stationary polarization studies revealed high limiting current densities in the temperature range of 860 to 940 °C, demonstrating the feasibility of the proposed method. Electrolysis experiments using lithium cryolite as an electrolyte produced an alloy containing approximately 0.7 wt.% scandium, with possible intermetallic compound formation. This study contributes insights into the kinetics of aluminum and scandium electrowinning, laying the groundwork for more efficient and cost-effective production methods for Al-Sc master alloys in the future. Going

forward, the research will be focused on unraveling the kinetics of simultaneous metal deposition, optimizing electrolysis parameters such as temperature, electrolyte composition, and current densities, and exploring the feasibility of creating other rare-earth metal–aluminum alloys. Acknowledgements The work was performed by the Laboratory of Low-Carbon Metallurgy and Energy under the state assignment of the FSAEI HE “Siberian Federal University”, a participating organization of the REC “Yenisei Siberia”, as part of the national project “Science and Universities”, project number FSRZ-2024-0004. The results presented in the paper were obtained in 2021.

References 1. A. Azarniya, A. Karimi Taheri, and K. Karimi Taheri, Journal of Alloys and Compounds, 781, 945 (2019). 2. K. Y. Chervyakova, N. A. Belov, M. E. Samoshina, and A. A. Yakovlev, Russian Journal of Non-Ferrous Metals, 59(2), 200 (2018). 3. A. A. Filatov, P. S. Pershin, A. V. Suzdaltsev, A. Yu. Nikolaev, and Yu. P. Zaikov, Journal of The Electrochemical Society, 165 (2), E28 (2018). 4. A. Y. Nikolaev, A. V. Suzdaltsev, Y. P. Zaikov, Journal of The Electrochemical Society, 166 (8) D252–D257 (2019). 5. N. J. Ricketts, Light Metals, 2019, 1395 (2019). 6. R. Jones, et al. Additive Manufacturing Letters 2:100026. (2022). 7. A. R. Rhamdani, et al. Mineral Processing and Extractive Metallurgy Review (2023) https://doi.org/10.1080/08827508. 2023.2243009 8. C. Guan, J. Xue, J. Zhu, and Q. Liu, 3rd International Symposium on High-Temperature Metallurgical Processing, 243 (2012). 9. Y. Shtefanyuk, V. Mann, V. Pingin, D. Vinogradov, Y. Zaikov, O. Tkacheva, A. Nikolaev, and A. Suzdaltsev, Light Metals, 2015, 589 (2015). 10. Sh. Yang, B. Gao, Zh. Wang, Zh. Shi, Y. Ban, H. Kan, X. Cao, and Zh. Qiu, Light Metals, 2007, 54 (2007). 11. Q. Liu, J. Xue, J. Zhu, Y. Qian, and L. Feng, ECS Transactions, 50 (11), 483 (2012). 12. Q. Liu, J. Xue, J. Zhu, and Ch. Guan, Light Metals, 2012, 685 (2012). 13. M. Harata, K. Yasuda, H. Yakushiji, and T. H. Okabe, Journal of Alloys and Compounds, 474, 124 (2009). 14. Y. Castrillejo, A. Vega, M. Vega, P. Hernandez, J. A. Rodriguez, and E. Barrado, Electrochimica Acta, 118, 58 (2014). 15. A. V. Suzdaltsev, A. Yu. Nikolaev, and Yu. P. Zaikov, Tsvetnye Metally, 2018(1), 69 (2018). 16. A. V. Suzdaltsev, A. A. Filatov, A. Yu. Nikolaev, A. A. Pankratov, N. G. Molchanova, and Yu. P. Zaikov, Russian Metallurgy (Metally), 2018(2), 133 (2018). 17. Zh. Tian K.Yang, X. Hu, Y. Lai, and J. Li, 10th International Symposium on High-Temperature Metallurgical Processing, 321 (2019).

Electrowinning of Al-Sc Master Alloys in the LiF-AlF3-Sc2O3 Melts 18. Zh. Tian, Ya. Lai, Sh. Yang, J. Dang, X. Hu, K. Zhang, and J. Li, Pat. CN104746106A (2015). 19. A. Yu. Nikolaev, A. V. Suzdaltsev, P. V. Polyakov, and Yu. P. Zaikov, Journal of The Electrochemical Society, 164(8), H5315 (2017). 20. Z. Wang, C. Guan, Q. Liu, and J. Xue. In 6th International Symposium on High-Temperature Metallurgical Processing, 215– 22 (2015). Hoboken, NJ: John Wiley & Sons

613 21. R. Guo, X. J. Zhai, and T. A. Zhang. Preparation of Al-Sc Alloy by LiF-ScF3-ScCl3 molten salt electrolysis. Materials Science Forum 675– 677:1125–28 (2011). 22. A. Yasinskiy, et al. Metals, 11(9), 1431 (2021). 23. S.K. Padamata, Yasinskiy, A.S., Polyakov, P.V. Journal of the Electrochemical Society, 168(1), 013505 (2021). 24. A. Yasinskiy, et al. Metals, 11(7), 1053 (2021). 25. A. Yasinskiy, et al. Minerals, Metals and Materials Series, 6, pp. 519–524 (2021).

Factors Affecting the Performance of Oxygen-Evolving Ni–Fe–Cu Anodes in Low-Temperature Molten Fluoride Electrolyte for Aluminium Electrowinning Gudrun Saevarsdottir, Geir Martin Haarberg, and Sai Krishna Padamata

Abstract

There is a strong push towards developing an alternative low-carbon process for the primary production of aluminium. Electrolysis with inert, oxygen-evolving anodes is an attractive alternative, and alloys of Ni–Fe– Cu are promising. This paper describes results from electrochemical studies to examine the performance of such anodes for different sodium–potassium cryolite electrolyte compositions from CR = (NaF + KF)/AlF3 = 1.3 to 1.4, and KR = KF/(KF + NaF) = 0.3–0.4. Voltammetry curves reveal how surface treatment in the form of anode pre-oxidation reduces passivation current and improves anode stability. Also, the effect of the polarization history of the anode on the voltammetry characteristics and the overall long-term stability of the anode is addressed, but leaving the anodes unpolarized in the electrolyte damages the dense, protective oxide layer on the anode surface, although pre-oxidation offers some protection. Keywords







Aluminium electrolysis Inert anodes gases Oxygen-evolving electrodes




Greenhouse

Introduction The currently used aluminium electrolysis process (i.e., the Hall–Héroult process) is energy-intensive and responsible for 1.175 billion tonnes of CO2 emission in 2021 [1], G. Saevarsdottir (&)
S. K. Padamata Department of Engineering, Reykjavik University, 102 Reykjavik, Iceland e-mail: [email protected] G. Saevarsdottir
G. M. Haarberg Department of Materials Science and Engineering, NTNU, 7491 Trondheim, Norway

including indirect emissions from energy production. This process uses consumable carbon anode and the liquid aluminium product serves as cathode. The carbon anodes require regular replacement as they are consumed. The overall reaction governing the aluminium electrowinning in the Hall–Héroult process is as follows: 1:5CðsÞ þ Al2 O3ðsolÞ ¼ 1:5CO2ðgÞ þ 2AlðlÞ E0 ¼ 1:19V and E0DH ðCE ¼ 96%Þ ¼ 2:037 V at 960  C

ð1Þ CO2 is the primary anode product, while CO is coevolved at the anode, making up around 10% of the anode gas [2]. The E0 is the standard reversible potential for reaction (1), at a process temperature of T = 960 °C (E0 = ‒DG0/nF), DG0 is the standard Gibbs energy of reaction, n is the number of electrons transferred in the reaction, and F is Faraday constant. E0DH is the necessary voltage to cover the reaction enthalpy for the reaction (E0DH = DH0(CE)/nF [3]), DH0 is the standard enthalpy of the cell reaction for reactants entering at ambient temperature and products leaving at electrolyte temperature, and CE is the current efficiency (REF). About 1.5 t CO2 per t of Al is emitted during the process [4], while a global average of 0.6 t CO2(eq) per t of Al is released in the form of perfluorocarbon gas emissions [5]. As a part of the global de-carbonization push, there is an urgency to find an alternative process to produce aluminium where the emission of greenhouse gases approach zero as much as possible. Replacing consumable carbon anodes with oxygen-evolving inert anodes could substantially reduce emissions and the quest for an inert anode production process is as old as the Hall–Héroult process. The success has been limited due to the cryolite-based electrolyte, which typically would be Na3AlF6 + * 12% AlF3, at a process temperature of 960 °C, it proved practically impossible to identify a material which would survive in the corrosive environment. Conventional electrolyte is characterized by the so-called Cryolite ratio, CR = (NaF)/AlF3. Typical industrial electrolyte is close to CR * 2.1, but addition of

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_79

614

Factors Affecting the Performance of Oxygen-Evolving Ni–Fe–Cu Anodes …

AlF3 to reduce the CR even further lowers the liquidus. If some of the Na in the electrolyte is replaced by K, an acceptable alumina concentration can be maintained, even at low CR and temperatures [6]. In this instance, the electrolyte is described by two parameters, CR = (NaF + KF)/AlF3 and the potassium ratio, KR = KF/(KF + NaF). By using electrolytes of CR = 1.3–1.5 and KR = 0.2–1, the cell can be operated below 800 °C, which is a much easier environment for anode material candidates. As a consequence, significant progress has been made in finding a suitable anode material [7]. A few inert anode processes are currently being developed and scaled up, but none of them has reached industrial scale to date. When aluminium is electrodeposited using oxygen-evolving anodes, the following reaction takes place: Al2 O3ðsolÞ ¼ 1:5O2ðgÞ þ 2AlðlÞ ; 0

E ¼ 2:20V and

0 EDH

ðCE ¼ 95%Þ ¼ 2:924V at 960  C ð2Þ

As the chemical energy in carbon is no longer present in the process, more electric energy is required to cover the theoretical minimum. The potentials in Eq. (2) refer to standard conditions at 960 °C to facilitate comparison to Eq. (1). However, due to modifications of the electrolyte composition, the inert anode cells are expected to operate at around 800 ˚C, which facilitates operation of inert anodes without anode failure due to chemical attack. At 800 °C, the reversible potential for the overall cell reaction would be around 2.31 V, which is 1.15 V higher than in the Hall–Héroult process, and the increase in rea*98ction enthalpy of the oxygen-evolving reaction is equivalent to a voltage increase of 0.9 V. To compensate for the increased voltage requirement, anode–cathode distance (ACD) should be reduced to minimize the overall cell voltage of the inert anode cell. This is hard to accomplish if the electrodes are horizontal, as in the Hall–Héroult process, but operating with vertical electrodes enables a more compact design. The vertical electrode cell (VEC) (Fig. 1) [8] consists of vertical inert anodes and wettable cathodes in parallel to each other. Wettable cathodes are essential to reduce the ACD in the VEC, so that the metal produced is captured, and for the operating cathode surface to be molten metal, which has excellent kinetic properties [9]. It should be noted that if the ACD is too small, two main issues could occur: (a) back-reaction, caused by the interaction of Al with oxygen, (b) contamination of Al at the cathode by possible anode corrosion products, (c) uneven saturation of the alumina raw material, and (d) insufficient ohmic heat generation in the electrolyte to cover the reaction enthalpy as well as heat losses. Thus, it is important to determine an optimal ACD value.

615

Fig. 1 Concept of VEC with inert anodes and wettable cathodes (used with permission of Arctus Aluminium, 2023)

Up to date, three material types have been investigated: metals or alloys, ceramics, and cermets [7, 8], An appropriate anode material should have good thermo-mechanical properties, high fluoridation resistance (formation of metallic fluoride), the ability to form a stable dense oxide layer that has low dissolution in the molten fluoride electrolyte, and a low wear rate of less than 1 cm/year when the anode is operated at 0.8 A/cm2 [10]. Much of published research on inert anodes has focused on metallic alloy anodes as they have high electrical conductivity, are resistant to thermal shocks, and are easy to fabricate [11]. Metallic anodes such as Ni–Fe, Ni–Fe-Al, Cu-Al, and Ni–Fe–Cu have been examined in various electrolyte compositions and varying conditions [9]. Of these alloys, the Ni–Fe–Cu system is considered to have superior anode material properties, as the alloy has the ability to form NiFe2O4 spinel oxide, which is stable in fluoride melts [12]. Cu concentration plays a key role in the anode performance of the oxide layer [13]. For the wetted cathodes, the consensus material seems to be TiB2, which has excellent stability in molten fluorides and is perfectly wetted by liquid aluminium in those electrolytes [9]. Figure 1 shows a VEC concept that is in used by Arctus Aluminium (Iceland). Here, the metallic Ni–Fe–Cu anode and TiB2 cathodes are vertically placed, and the oxygen evolution takes place at the anode and the cathode is completely wetted by the aluminium reduced on its surface. The electrochemical measurements discussed in this paper are carried out using a typical three-electrode cell, and the reference electrode is Al/AlF3. The electrochemical methods used in this paper are cyclic voltammetry and chronopotentiometry, performed in a setup illustrated in Fig. 2. Our previous paper clearly discusses the cell setup and other details involving the experimental methods [14].

616

G. Saevarsdottir et al.

Fig. 3 Open-circuit potential (OCP) on the anodes (pre-oxidized and untreated) taken immediately after their immersion in electrolyte and after 30 min of immersion anodes in KF-NaF-AlF3-Al2O3 (CR = 1.4, KR = 0.35) at 825 ˚C

Fig. 2 The cell setup used for electrochemical measurements [14]

Factors Influencing the Performance of the Anodes Effect of Pre-oxidation In order to enhance the anode performance, it is proposed to pre-oxidize the anodes before use. The anodes are oxidized in air at 800 °C for 8 h to form a protective oxide layer on the anode surface. Figure 3 shows the Open-circuit potential (OCP) versus time of pre-oxidized and untreated anodes made of Ni42Fe38Cu20 alloy. In the case of pre-oxidized anode, when the anode was initially immersed in the electrolyte, a slight decrease in OCP from an initial value of OCP = 1.34 V can be seen. This could be due to electrochemical reactions re-ordering the pre-existing oxide layer at the OCP, making it more robust to chemical attack. Subsequently the pre-oxidized anode was left unpolarized for 30 min in the electrolyte, and OCP was again recorded. The OCP remained slightly higher than the initial OCP value, or around OCP = 1.37 V, indicating that the oxide layer formed during the oxidation process was still present and not completely dissolved, though a higher potential is required to suppress dissolution reactions. In the case of the untreated anode, of the same composition, the OCP was significantly higher than for the pre-oxidized anode or around 1.47 V, and was constantly fluctuating when recorded immediately after its exposure to the corrosive electrolyte. This indicates that due to the lack of the protective oxide layer, a higher voltage is required to

protect the alloy from corrosion. A clear increase in the OCP to 1.58 V after 30 min of exposure to electrolyte can be seen. This could be due to the formation of metallic fluorides on the surface of the untreated anode, which further weakens its resistance to corrosion. The structure of the oxide layer of pre-oxidized anodes before use and after 30 min of submersion in the electrolyte is shown in Fig. 4. Figure 4 shows how 30 min of electrolyte exposure affects the oxide layer of a pre-oxidized anode, as tested for OCPs in Fig. 3. It is apparent that after the 30 unpolarized minutes in the electrolyte, the oxide layer is somewhat thinner, but visibly more porous, and electrolyte components had penetrated the oxide layer. Figure 5 shows linear sweep voltammetry (LSV) curves taken after the immersion of pre-oxidized and untreated anodes of this same composition in the electrolyte for 10 min. In the case of untreated anode, a clear passivation peak can be seen starting below the reversible potential for oxygen evolution, and a significant background current is present as oxygen evolution takes over. Please note that the values in the figure are not infrared (IR) compensated, thus the shift in the observed peak potential, as the peak current densities are quite high. This increase is associated with the active dissolution and reoxidation happening on the anode surface, which slows down as a thin protective layer is formed. In the case of pre-oxidized anodes, the passivation peaks are much smaller, around a third of those for the untreated anode, but still significant. That indicates that there are reactions associated with a re-ordering of the oxide layer, but much less prominent than for the untreated anode. From thermodynamics, some possible reactions contributing to the electrochemical reactions happening on the anode surface at potentials below oxygen evolution are shown in Fig. 6.

Factors Affecting the Performance of Oxygen-Evolving Ni–Fe–Cu Anodes …

617

Fig 4 SEM images and elemental mappings for the oxide layer of a freshly oxidized anode a and an identical anode after being suspended in the electrolyte for 30 min b without any polarization

Fig. 5 Linear sweep voltammetry curves at different scan rates taken on Ni–Fe–Cu anodes in KF-NaF-AlF3-Al2O3 in anodes in KF-NaF-AlF3Al2O3 (CR = 1.4, KR = 0.3) at 825 ˚C (note: no IR compensation)

618

Fig. 6 Standard potentials of half-reactions vs. Al at 1098 K, calculated using HSC Chemistry HSC 10 software, licensed to Reykjavik University

Effect of Cu Concentration As earlier stated, the Ni–Fe–Cu anodes have the ability to form NiFe2O4 oxide scale, which has low solubility in cryolite melts [14]. Cu concentration in the Ni–Fe–Cu alloy plays a key role in the formation of the nickel ferrite oxide layer. CuO2 scale is readily formed on the alloy, which is the outmost layer. Meanwhile, NiO and Fe2O3 are forming on the surface of the alloy, which will interact with each other and form the NiFe2O4 scale. CuO2 then dissolves, and Nickel spinel becomes the outermost layer. Comparing Fe– Ni-Cu anodes with 5% and 20% Cu content shows how the Cu content and being subject to a period of electrolysis in the cell contributes to anode stability. Four types of anodes were used in the test, 48Ni-47Fe-5Cu and 42Ni-38Fe-20Cu, untreated or/and after 2 h of polarization at 0.5 A/m2. Figure 7 shows LSV curves with a sweep rate of 0.05 V.s‒1 in an electrolyte composition of CR = 1.3, KR = 0.7 at T = 770 °C. For the 20% Cu anodes, there is a clear passivation peak at around 2.2 V with a fresh anode and a bit higher voltage after anodization, indicating that the thicker oxide layer contributes to the overpotential. Comparison to Fig. 5 indicates that this peak may be linked to activity related to CuO or CuO2. The 5% anodes display this same peak, but also a more significant peak at lower potentials, that are likely attributable activities involving iron or nickel oxide species. The overall passivation activity contributes more to the current density for the 5% Cu anode than for the 20% Cu anode, indicating that the higher Cu content improves the stability of the anode.

G. Saevarsdottir et al.

Figure 8 shows steady-state polarization curves for those four cases. The curves indicate that the anodic overpotentials for oxygen evolution are higher for the 5% Cu anode than for the 20% Cu anodes, even though the background passivation current at potentials below the oxygen evolution is higher for the lower Cu content anode. Embedded in Fig. 8a, b is the galvanostatic polarization voltage for the two anodes during the 2-h period at 0.5 A cm−2. The potential for the 5% Cu anode increases steadily, indicating increased resistance due to a growing oxide layer, while the voltage for the 20% Cu anode is relatively stable, which corresponds to less growth of the oxide layer. It is observed on the steady-state polarization curves in Fig. 8 that the 2-h polarization increases the overpotentials for both anode compositions, likely due to the contribution from the oxide layer formed. The lower background passivation current in the steady-state polarization curves for both anode compositions after polarization indicates that oxide layer formed during the polarization has a protective function, although not comparable to pre-oxidation treatment. All of this is an indication of the importance of Cu content in the Ni–Fe–Cu system when applied as oxygen-evolving anodes for aluminium electrolysis.

Operating Temperature Figure 9 shows 0.01 V.s‒1 sweep rate IR compensated LSV curves for anodes with 20% Cu content, that were not pre-oxidized, taken at three different electrolyte temperatures. The anodes were lifted from the electrolyte between measurements while the electrolyte temperature stabilized at a new setpoint. The sweep rate is low but none the less, several peaks associated with dissolution and passivation can be seen on anodes when the electrolyte temperature is at 840 ˚C, and significant peaks at both 1.7 V and 2.2 B V were observed for the 5% Cu anode at 770 °C in Fig. 7, though in this case for a 20% Cu anode, which at 770 °C had no peak at 1.7 °C. Equivalent peaks are hardly visible at 760 ˚C, though one may with some effort be detected at 2.2 V, such as was visible for the 20% alloy in Fig. 7, admittedly at a higher sweep rate. It can also be seen that for a given anode current density, the overpotential for oxygen evolution is much lower at higher temperatures compared to the one at 760 ˚C. These curves, on anodes that were not pre-oxidized for protection, show the effect of electrolyte temperature on the anode stability, and how a decrease in process temperature

Factors Affecting the Performance of Oxygen-Evolving Ni–Fe–Cu Anodes …

619

Fig. 7 Anodic polarization LSV curves with a scan rate of 0.05 V.s‒1 for 5% and 20% Cu anodes before and after 2 h of anodic polarization at 0.5 A/m2

Fig. 8 Steady-state polarization curves for 5% and 20% Cu anodes before and after 2 h of anodic polarization 0.5 A/m2. The curves for the 2-h galvanostatic polarization at 0.5 A cm−2 are embedded in the figures

contributes to longer anode life. An anode composition which is stable at 760 °C may not be so at 840 °C.

Conclusion The paper covers some findings on factors affecting the performance of Ni–Fe–Cu alloys serving as oxygen-evolving inert anodes for aluminium deposition. It is demonstrated that pre-oxidizing the anodes in air for 8 h at 800 °C forms an

oxide layer, which helps protect the anodes from active dissolution and passivation when immersed in the electrolyte. It is also shown that an anode composition with 20% Cu outperforms anodes with 5% Cu composition in terms of stability during electrolysis, due to the stability and limited growth of the oxide layer. Anodes that have not been pre-oxidized will form an oxide layer during electrolysis which provides a form of protection from electrolyte attack. Alloys that show good stability at a relatively low electrolyte temperature may not be sufficiently stable at a higher temperature.

620

Fig. 9 Linear sweep voltammetry curves (0.01 V.s‒1) taken on Ni–Fe– Cu anodes in KF-NaF-AlF3-Al2O3 (CR = 1.4, KR = 0.3) at 760, 800, and 840 ˚C Acknowledgements The authors would like to thank Jon Hjaltalin Magnusson (Arctus Aluminium), Gudmundur Gunnarsson, Gudbjorg Oskarsdottir, and Rauan Meirbekova (IceTec), who have supported this work, for example, by supplying the anode alloy and cathode material. Special thanks to the Alcoa Foundation, which funded this work (2019:223136) as well as Rannis-The Icelandic Centre for Research (2019:207242-051) which contributed to the funding.

References 1. “IAI international Aluminium Institute, statistics, greenhouse gas emissions from the aluminium sector, accessed August 30th 2023.” [Online]. Available: 1.https://international-aluminium.org/ statistics/greenhouse-gas-emissions-aluminium-sector/ (accessed on 30th August 2023). 2. T. A. Aarhaug, A. Ferber, O. Kjos, and H. Gaertner, “Online Monitoring of Aluminium Primary Production Gas Composition by Use of Fourier-Transform Infrared Spectrometry,” in Light Metals 2014, 2014, pp. 647–652. https://doi.org/10.1002/ 9781118888438.ch109.

G. Saevarsdottir et al. 3. H. Kvande and W. Haupin, “Inert anodes for AI smelters: Energy balances and environmental impact,” JOM, vol. 53, no. 5, pp. 29– 33, May 2001, https://doi.org/10.1007/s11837-001-0205-6. 4. G. Saevarsdottir, H. Kvande, and B. J. Welch, “Aluminum Production in the Times of Climate Change: The Global Challenge to Reduce the Carbon Footprint and Prevent Carbon Leakage,” JOM, vol. 72, no. 1, pp. 296–308, Jan. 2020, https://doi.org/10. 1007/s11837-019-03918-6. 5. G. Saevarsdottir, S. K. Padamata, B. N. Velasquez, and H. Kvande, “THE WAY TOWARDS ZERO CARBON EMISSIONS IN ALUMINUM ELECTROLYSIS,” Light Metals, 2023. 6. A. Redkin et al., “Recent Developments in Low-Temperature Electrolysis of Aluminum,” 2013. 7. S. K. Padamata, K. Singh, G. M. Haarberg, and G. Saevarsdottir, “Review—Primary Production of Aluminium with Oxygen Evolving Anodes,” Journal of The Electrochemical Society, vol. 170, no. 7, p. 073501, Jul. 2023, https://doi.org/10.1149/1945-7111/ace332. 8. Y. He, K. Zhou, Y. Zhang, H. Xiong, and L. Zhang, “Recent progress of inert anodes for carbon-free aluminium electrolysis: a review and outlook,” J. Mater. Chem. A, vol. 9, no. 45, pp. 25272– 25285, 2021, https://doi.org/10.1039/D1TA07198J. 9. S. K. Padamata, K. Singh, G. M. Haarberg, and G. Saevarsdottir, “Wettable TiB2 Cathode for Aluminum Electrolysis: A Review,” Journal of Sustainable Metallurgy, vol. 8, no. 2, pp. 613–624, Jun. 2022, https://doi.org/10.1007/s40831-022-00526-8. 10. R. P. Pawlek, “Inert Anodes: An Update,” in Light Metals 2014, J. Grandfield, Ed., Cham: Springer International Publishing, 2016, pp. 1309–1313. https://doi.org/10.1007/978-3-319-48144-9_219. 11. A. S. Yasinskiy, S. K. Padamata, P. V. Polyakov, and A. V. Shabanov, “An update on inert anodes for aluminium electrolysis,” Non-Ferrous Metals, 2020, vol. 1, pp. 15–23. 12. P. Meyer et al., “Comparative study on the chemical stability of Fe3O4 and NiFe2O4 in molten salts,” Materials Science and Engineering: B, vol. 228, pp. 117–122, Feb. 2018. https://doi.org/ 10.1016/j.mseb.2017.11.025. 13. E. Gavrilova, G. Goupil, B. Davis, D. Guay, and L. Roué, “On the key role of Cu on the oxidation behavior of Cu–Ni–Fe based anodes for Al electrolysis,” Corrosion Science, vol. 101, pp. 105– 113, Dec. 2015, https://doi.org/10.1016/j.corsci.2015.09.006. 14. G. Saevarsdottir, G. M. Haarberg, M. Bourmaud, K. Singh, and S. K. Padamata, Anodic Behaviour of Ni42Fe38Cu20 Electrode in Molten Fluoride Salts. Journal of The Electrochemical Society, vol. 170, no. 7, p. 072508, Jul. 2023, https://doi.org/10.1149/19457111/ace5e3.

Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts: The Riddle of Dissolving Alumina Rafts Solved Jonathan Alarie, László I. Kiss, Lukas Dion, Martin Truchon, Sébastien Guérard, and Jean-François Bilodeau

Abstract

With increasing concerns about environmental performance and workplace security, the production of primary aluminum encounters many challenges. Among them is the introduction, dissolution, and distribution of large quantities of alumina into a small volume of electrolyte with a large horizontal surface using only a limited number of discrete injection points. Typically, one kilogram of alumina is injected at each location every minute or so, which agglomerates and creates rafts limiting the dissolution rate. More importantly, to prevent instability, the rafts must disappear from the rafts injection point before the next addition, either as a result of dissolution, transport, or disintegration. The work presented uses a dimensional analysis to quantify the dissolution and disintegration rate of alumina raft. A theoretical approach is proposed to understand the key behaviors observed from extensive experimental work on the dissolution of different macroscopic forms of alumina. J. Alarie (B) · L. I. Kiss · L. Dion · M. Truchon GRIPS, University of Quebec at Chicoutimi, 555 Boul Universite, Saguenay, QC G7H 2B1, Canada e-mail: [email protected] L. I. Kiss e-mail: [email protected] L. Dion e-mail: [email protected] M. Truchon e-mail: [email protected] REGAL, Aluminium Research Centre, 2325 Rue de l’Université, Québec, QC G1V 0A6, Canada S. Guérard · J.-F. Bilodeau Arvida Research and Development Centre, Rio Tinto, 1955 Mellon Boulevard, Jonquière, QC G7S 4K8, Canada e-mail: [email protected] J.-F. Bilodeau e-mail: [email protected]

Finally, the potential application of the model proposed is also presented. Keywords

Aluminum electrolysis • Alumina dissolution • Cryolite properties

Introduction The production of primary aluminum is done through the electrolysis of dissolved aluminum oxide in a cryolite melt. The dissolution of alumina is of primary concern to keep the process stable. When the content of dissolved alumina in the melt is extended beyond the operational range, instabilities appear in the thermal and electrical equilibrium. The most common of these instabilities is the anode effect, which leads to unnecessary greenhouse gas emissions, increased power consumption, and possible damage to the cell, due to localized overheating of the cell lining. On the other hand, higher concentration promotes the formation of muck under the liquid metal pad. This muck also increases the instabilities of the cell and locally modifies the electrical resistance of the cathode, causing uneven consumption of the latter. Therefore, it is essential to create a steady flux of alumina that dissolves into the electrolyte, which is challenging due to the nature of the feeding strategies. The cells are fed through periodic 1000 g additions of powdered materials from 3 to 5 points over a surface of more than 70 m2 . Such cold alumina poured on the cryolite surface rapidly agglomerates to form solid floating rafts. These rafts limit the solid-to-liquid contact necessary for an efficient dissolution of the alumina. Following recent progress related to the effect of the liquid composition on the diffusivity of alumina in cryolite melt [1] and on the raft formation dynamics [2], this work builds on to describe the mechanisms of the raft sintering, disintegration and dissolution. A model is then proposed to predict the alumina dissolution. The differences between the model and experimental

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_80

621

622

J. Alarie et al.

Table 1 Conditions range used in the parametric study Parameter (◦ C)

Superheat AlF3 (w%) CaF2 (w%) Al2 O3 (w%) Drying temperature (◦ C)

Minimum

Maximum

5 9 4 3 100

15 14 7 6 800

Fig. 1 Sketch of the discretized raft

results are discussed highlighting the benefits and limitations of this approach. Finally, a potential industrial application of this model is discussed.

Experimental Method The effect of the bath conditions on alumina dissolution was estimated using a parametric study of injections in a laboratory experimental setup. The parameters tested are presented in Table 1 and were distributed in eight conditions. These tests were conducted in a round crucible of 185 mm high and 203 mm in diameter placed in a radiative furnace. The sample was produced by the injection of 40 g of alumina, previously dried at temperatures reported in Table 1 for 1 hour. The alumina was poured into a curved funnel, with its extremity placed at 25 mm from the surface of the melt. The apparent mass of the sample was recorded with a gravimetric apparatus previously developed (see Alarie et al. [3]), with the help of a sample holder. The holder consisted of Nichrome wire with a stainless steel ring of 12 mm in height and 12 cm in diameter to limit the horizontal spreading of the powder and hold the powder by its center to the bath-air interface.

Raft Dissolution Model To simulate the dissolution of an alumina raft it is necessary to understand the different phenomena in action. For simplicity, the raft is considered to be cylindrical and at rest in the instants following the injection process. The dynamics of the injection have been described by Alarie et al. [1] and allow the estimation of the alumina that goes through the interfaces, as an initial alumina cloud, and the remaining fraction that forms the alumina raft. At the moment of its formation, this raft is considered completely dry. Thus, the wetted particles are associated with the alumina cloud and are not considered at this point. Following injection, the process becomes static and the relative movement between the alumina particles is null. In this phase, the active phenomena affecting the alumina particles are the infiltration of the liquid, the dissolution, and the phase change of the alumina as well as the dry and liquid phase sintering of the particles.

The rate of these phenomena is driven by heat and mass transfer dynamics. An accurate knowledge of the properties of the materials is then important to obtain a model with adequate precision. For this purpose, the properties of the bath were calculated according to the method presented by Alarie et al. [1]. The properties range of the bath used for calculations are presented in Table 1. Note here that the alumina is considered as spheroid particles of 80 µm in diameter (ds = 8 · 10−5 m) with low deformation, usually similar to those of sand particles, which have a shape factor (c f ) of 1.4 (see the VDI heat atlas [4]). The thermal properties of alumina come from an analysis presented by Alarie et al. [2]. The simulation considered is one-dimensional and the raft simulated is similar to the ones created in the laboratory setup. Thus, they are the results of an injection of 40 g, forming a raft that is 12 cm in diameter and 3 mm in height. This heap of powder has been discretized in 51 layers as presented in Fig. 1. For every time step, the simulation proceeds with a specific sequence of phenomena solved explicitly. First, the raft is a dry heap of powder filled with air with fluoride gas. Then, the infiltration rises in the raft, allowing more particles to be in contact with the liquid. The liquid contact also enables the liquid phase mass transfer, that is to say, the sintering of alumina, the cause of its phase change, and the dissolution. During this time, dry phase sintering of the grain modifies the porosity of the particles not in contact with the liquid. Finally, the dissolution of the alumina is calculated. This dissolution also includes the mechanisms needed for the alumina to reach the lower part of the raft and be caught by the convective flow under the raft surface. More precisely, the simulation goes through the following phases at each time step: 1. The infiltration of the raft. 1.1 Porosity of the raft. 1.2 Height of the liquid. 2. The heat transfer mechanisms. 2.1 Effective thermal conductivity of each layer. 2.2 Convection heat transfer at the bottom of the raft.

Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts …

2.3 Thermal conduction. 2.4 Radiative heat loss. 2.5 Enthalpies influences.

623

liquid goes up a layer, the bath in each layer also rises to the next layer, bringing a hotter bath in each layer. 

3. The sintering of the alumina.

H=

3.1 Dry sintering of the powder. 3.2 Liquid phase sintering (alumina phase change).

re f f =

4. The dissolution of the raft. 4.1 Internal dissolution of grains. 4.2 Effective dissolution of the grain in the layer of the raft. 4.3 Convective mass transfer at the bottom of the raft. 4.4 Diffusion of alumina in the raft. The calculation of each of these phases is presented thereafter.

The Infiltration of the Raft Porosity of the Raft The initial porosity of the raft is one of the primary parameters that influence the behavior in this porous media. It was determined with the model presented by Yu et al. [5]. This model is presented in Eq. 1, resulting from experimental measurements to obtain a model for a poured packing of the spheres from an initial parameter (φ0 = 0.567), and two constants (a = 0.247 and b = 0.749). This model is presented in Eq. 1. φ = φ0 + (1 − φ0 )e−ads

b

(1)

It is also important to consider the internal porosity of the alumina particles. This parameter was determined, using scanning electron microscopy, to be near 0.15 V%. Details on the method are presented by Alarie et al. [2]. Finally, the porosity of the wetted layer is updated at each time step from the sintering and dissolution of the alumina.

Height of the Liquid The infiltration behavior is modeled with the help of the Washburn equation, as presented by Kirdponpattara et al. [6] and reported in Eq. 2. It uses the effective particle radius (re f f ) in combination with the surface tension of the liquid (γ ), the angle of contact between the liquid and the solid in air (θ ) as well as the liquid viscosity (μ) and the time (t) to find the height of infiltration (H ). The effective particle radius is calculated from Eq. 3, presented by Hapgood et al. [7]. The use of the effective radius of the particle is to model the real porosity connection to the equivalent vertical tube, necessary for the use of the Washburn equation [6]. In the model, the bulk bath properties are used to obtain the height of the rising front of the liquid in the powder. Thereafter, each time the

re f f γ cosθ t 2μ

c f ds  3 1−

(2) (3)

The Heat Transfer Mechanisms Convection Heat Transfer at the Bottom of the Raft From the lower surface, a convective heat transfer from the bulk bath is considered. Since the crucible in the experimental setup is heated from the side and because the sample represents less than 0.5 w%, the bulk temperature is considered constant during the test. Note that this case is also similar to that of the heating produced in industrial cells. The raft is considered as a flat plate with parallel flow and the heat transfer is calculated according to Eqs. 4–6 [8]. These equations use a velocity of 35 mm/s, as measured in our experimental setup to calculate the Reynolds number. Then, the heat transfer coefficient (h) is found from the Nusselt number using the diameter of the raft as its characteristic length (L c ) and the bulk bath thermal conductivity (λ). Knowing the temperature difference between the bottom layer of the raft and the bulk temperature (T ) and the surface of contact between the raft and the liquid ( A) finally allows finding the convective heat flux ( Q˙ c ) N u = 0.664Re0.5 Pr 1/3 h Lc Nu = λ Q˙ c = h AT

(4) (5) (6)

Thermal Conduction The heat transfer in this simulation involves finding an effective thermal conductivity for each discretized layer of the raft. This effective conductivity has been found with the help of the Zehner-Bauer-Schlünder model presented in the VDI heat atlas [4]. This is a unit cell model that estimates the effective conductivity of a powder bed knowing its porosity and the thermal conductivity of the components as the primary parameters. They also consider the contribution of the radiative heat transfer and the shape of the particle to the effective thermal conductivity. This results in a thermal conductivity that changes over time and space in the raft. Moreover, the dry and wet thermal conductivities are nearly one order of magnitude apart, with a value of around 0.15 W/mK and 0.95 W/mK, respectively. In addition, an equivalent temperature must be found to estimate the heat flux between each of the layers. This is achieved with the help of the thermal capacity to weigh the

624

J. Alarie et al.

temperature of the bath and the particles at this average temperature. These hypotheses allow computing the heat transfer through the body of the raft.

Radiative Heat Loss For the top of the raft, only a radiative heat transfer is considered. This approximation was used following previous work, which highlighted that with high temperature in the experimental setup, the conventional heat transfer becomes negligible. It is also a limiting case because the convection on top of the sample can help to raise its temperature more rapidly in the first seconds, which is expected to help the overall dissolution rate of the raft. The absence of this heat input can only lower the dissolution rate in the model. The radiative heat loss ( Q˙ rad ) was calculated with the help of Eq. 7, using the emissivity of the grains ( p ), the Stephan-Boltzmann constant (α) and the temperature of the grains (T p ) and the environment (T∞ ). 4 ) Q˙ rad =  p α A(1 − φ)(T p4 − T∞

(7)

Reaction Enthalpies Finally, the enthalpies of the different reactions, taken from Dassylva-Raymond [9], are considered as follows. First, the heat flux from the adjacent layers is computed, along with the heat capacity of the liquid bath from its liquidus to its actual temperature is calculated. Then, three cases of figures can happen: 1. The total heat flux is positive and can melt solidified bath, if any. In this case, the melting of the bath absorbs energy until all of the solid bath is melted. Then the bath temperature rises. 2. The total heat flux is negative, but there is sufficient heat capacity to absorb it. Then the bath temperature only decreases. 3. The total heat flux is negative and brings the bath temperature to its liquidus. Then, the bath temperature remains at its liquidus temperature due to the energy released by the bath that solidifies. The first two cases are pretty straightforward and raise or lower the temperature of the layer of the raft according to simple thermal balance. The third case is of more concern. It is known that the solid bath that usually forms is mainly consisting of pure cryolite. Therefore, the solidification of the bath inside the raft changes the additive mass fraction in the remaining liquid. This behavior is also considered in the simulation and lowers the liquidus temperature of the melt for the successive time step. Consequently, only a small amount of solid bath is formed in the raft. Moreover, no layer completely solidifies its liquid content at any time of the simulation. This is due to the infiltration of the bath, the convective heat transfer

under the raft, and the effective conductivity of the dry layer above the liquid. These three phenomena cause the temperature of the alumina grains to rise quickly before contact with the liquid.

The Sintering of the Alumina It is clear here that the initial properties of the alumina are of first importance in the raft sintering kinetic. Properties such as α-alumina content, the BET surface, the MOI, the LOI, and the presence of adsorbed fluoride in the alumina particles are known to influence the dissolution process of alumina, but, moreover, its sintering behavior. Note that the relation is not linear, nor direct, for the influence of these parameters on the dissolution, because they impact the sintering process. The exact contribution of each of these parameters to the alumina sintering is exposed thereafter.

Dry Sintering of the Powder The dry sintering kinetics is usually modeled with the Avrami equation, reported in Eq. 8 as presented by Macêdo et al. [10]. It relies on a kinetic constant (k) and an exponent (n) to fit experimental data to the Avrami equation. This kinetic depends on the gas in contact with the alumina and the volatile molecules created during the sintering. To find the experimental parameters (k and n), the exact composition of the gas in the pores of the raft and their effect on sintering should be the subject-specific investigations. Without these informations, the data reported by Bagwell and Messing [11] are here used. Their work estimated the influence of water vapor in the sintering of alumina. It appears that this speed is similar to that of Shacklee, reported in Bagwell and Messing [11], which concerns fluoride gas influence. For this reason, an exponent of n = 2.1 was selected to use in the Avrami equation with Al F3 vapor. Along with the value reported by Shacklee in Bagwell and Messing [11] (x = 1 and t = 30 min), the constant k of the Avrami equation was approximated to be 0.03. With these parameters, the dry sintering speed has been estimated in an alumina raft. In the dry sintering process, the gas composition in contact with the grains is of first importance. Two specific gases, in addition to the air, can be identified; water vapor and fluoride gas. Bagwell and Messing [11] demonstrated the importance of the water vapor in the particles on the kinetics of dry sintering. The effect of the water vapor, which comes from the MOI content, has a proportional effect on the sintering rate just a little lower than that of the fluoride vapor. For these simulations, the MOI content effect has been associated with the effect of the fluoride vapor and a unique value has been applied to the dry sintering process. A better description of the composition of the gas in the porosities of the raft should be the subject of future experimental tests. The LOI param-

Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts …

eter is related to the moisture content in the alumina crystal. It allows the release of water vapor all along the liquid phase sintering. This creates bubbles inside the body of the raft, which increases in volume as the vapor is released and the temperature of the gas increases. Since the mass fraction of AlF3 in the liquid bath is nearly constant (around 35 w% of the total mass of the bath ± 4), we can also assume that the vapor composition is nearly constant and that the dry sintering of alumina particles is the same for all conditions. n (8) x = 1 − e(−kt )

Liquid Phase Sintering (Alumina Phase Change) The liquid phase sintering is the less known of the mass transfer phenomena that occurs in a raft. It relies on the alumina diffusivity in the liquid to establish a dynamic equilibrium between the dissolution of the alumina and its precipitation in α-alumina, which corresponds to the phase change of alumina. In fact, the γ -alumina dissolution, catalyzed by the presence of fluoride [12], quickly saturates the melt inside the alumina particle, while the α-alumina content of the particle remains in the solid form. Therefore, due to the local pressure gradient created by the curvature of the surfaces, the dissolved alumina precipitates on these α-alumina seeds, forming alumina platelets [13]. Bagwell and Messing [11] also described the influence of the number of α-alumina seeds in the alumina crystal. The αalumina seed content in the particle, which has been related to the weight fraction in this simulation, acts as precipitation sites for the dissolved alumina. Bagwell and Messing [11] also suggested that the sintering kinetics is favored proportional to the number of seeds until a given value. In the case of the alumina raft dissolution, this also means that alumina with a higher value of α-alumina content undergoes faster sintering than that with less α-alumina phase until this threshold. Above this value, it is important to understand that a smaller fraction of the raft undergoes sintering since the alumina is already in more stable form. Therefore, the increase in porosity created by the dissolution-precipitation phenomenon inherent to the sintering does not occur, thus limiting the effective diffusion of the alumina and then the overall dissolution process. Note that the shape of the internal porosities of the grain, usually related to the BET surface of the grain, also creates a suitable site for the precipitation of the dissolved alumina. A change in the BET surface of the alumina modifies the sintering behavior of the alumina. Macêdo et al. [10] showed that the rate of sintering of the alumina is proportional to its surface area. This is expected since a larger area of the surface increases the availability of precipitation sites and the availability of the alumina to dissolve. The transport of the dissolved alumina is then strongly inherent to the diffusivity of the alumina in the liquid. This dissolution precipitation happens very fast in the first

625

seconds of the bath presence and slowly decays over time. The equilibrium is reached when the alumina platelets have a diameter of about 10 µm, as reported in the literature [13–16]. Once the equilibrium is established, the alumina platelets begin to slowly dissolve and shrink, leading to the efficient part of the raft dissolution. The Lifshitz-Slyozov-Wanger (LSW) theory (from Kang [17]) was used to simulate the liquid phase sintering. According to this theory, the grain growth of the alumina can be estimated using the gas constant (R), the diffusivity of the alumina in the liquid (D), the molar ratio (C∞ ) of alumina dissolved around the particle and its molar volume (V¯ ). Equation 9 then allows finding the particle radius (r ) at a given absolute temperature (T ) after a certain time (t). Once the sintering phase of the simulation is completed, the porosity of the raft is updated accordingly.  r=

3

8 Dγ C∞ V¯ t + r03 9 RT

(9)

Here, it is important to understand that the particles involved are the α-alumina seed and not the alumina grains. As mentioned earlier, the dissolved alumina precipitates on the α-alumina present in the grains to form the platelets. Note that for simplicity, the platelets are considered as a sphere for this calculation instead of the octagonal prism that is presented in the literature. The number of seeds and their size are of primary concern to adequately simulate the sintering behavior. Moreover, the size at which they reach equilibrium drives the alumina surface available for dissolution in further steps. Unfortunately, this information is not available in the literature, and usage of a specific hypothesis was required to proceed with this step of the work. The mass fraction of α-alumina in an alumina particle, obtained by chemical analysis, was used to find its volume in the grain. Thus, with the assumption that the platelets are spherical particles, the volume of α-alumina was divided into several seeds of a given diameter to match the appropriate value. The diameter of the seeds was fitted to the size observed using scanning electron microscopy, obtained in experimental tests on this specific subject.

The Dissolution of the Raft Convection plays an important role in the heat and mass transfer affecting the raft dissolution. This convection is driven by the bath properties (density, viscosity, alumina solubility, and alumina diffusivity) and can be calculated from the composition of the melt [1]. With these properties, it is easy to find the Grashof, Reynolds, and Schmidt numbers useful in the calculation of the Sherwood number with reported semi-empirical relations. These relations, reported in Eqs. 10 and 11, relate

626

J. Alarie et al.

the dissolution flux of an alumina surface (φ) to the Sherwood number by the use of the contact area (A), the bath density (ρl ), the alumina mass fraction gradient (C), the characteristic length of the raft (L c ) and the diffusivity of alumina in the melt. Once the Sherwood number is known, the alumina flux through a given surface can be calculated, using the convective mass transfer coefficient and the alumina solubilities. The detail on the diffusivity of the alumina in cryolite melts can be found in Alarie et al. [1]. hm L c D φ = h m ρl AC

Sh =

(10) (11)

Internal Dissolution of Grains At the beginning of the dissolution-precipitation behavior caused by the liquid phase sintering, a part of the dissolved alumina is allowed to diffuse from the inside of the alumina particle to the bath trapped inside the raft. The amount of diffusing alumina increases as the equilibrium of precipitation is achieved. This diffusion is also limited by the presence of the solid alumina in the diffusion path, leading to a lower effective diffusion from the alumina particle to the bath inside the raft. This apparent dissolution of the particles then increases the local alumina content in the bath, which also be transported to the surface of the raft. To find the dissolution rate of the seeds, a Sherwood number of 3 is used, which corresponds to the limiting case where the Grashof number tends to zero. Effective Dissolution of the Grain in the Layer of the Raft In the effective dissolution of an alumina grain, it is necessary to consider the flow of the bath around it. Since there is infiltration, which creates a liquid flow in the porous media, a higher Sherwood number is expected. Moreover, the bubbles created by the desorption of the gas trapped inside the grains create a dynamic flow in the raft. The volume of the bubble is crucial for the dissolution process. This is because the growth of the bubble push saturated bath, situated in the raft porosities, outside of the raft, thus momentarily increasing the effective diffusion of the dissolved alumina. On the contrary, the release of the bubble from the raft body creates a void that the bath in the surrounding pores fills, bringing fresh bath in contact with the alumina. This decreases the effective diffusion of the alumina for a very short time but enhances the overall dissolution of the alumina. Consequently, the LOI content of the alumina is favorable to the dissolution process, but this effect is also dependent on the viscosity of the gas and liquid inside the raft. The model presented here is not yet mature enough to take this dynamic into account. Therefore, it is approximated by an additional increase in the Sherwood number, which was fixed at 5.

Convective Mass Transfer at the Bottom of the Raft The last layer of mass transfer represents the effective dissolution of the alumina raft through its surface in contact with the bulk electrolyte. Here again, this dissolution is inherent to the effective diffusion of the alumina inside the porosities of the raft. However, this part of the dissolution is subject to intensive agitation of the liquid, which leads to a larger convective mass transfer, strongly influenced by the bath properties. The convective dissolution of the raft is then modeled by Eq. 12, which represents the dissolution of a flat plate parallel to the flow. (12) Sh = 0.664Re0.5 Sc1/3 Diffusion of Alumina in the Raft The diffusivity of alumina in cryolite is influenced by the presence of solid particles along the path. To take this into account, the model from Weissberg [18], presented in Eq. 13 from Kim et al. [19] was used. It uses the diffusivity of alumina in cryolite (D) to obtain the effective diffusivity (De f f ) from the porosity () of the porous media. Two levels of porosity need to be considered, that of the alumina particles and that of the raft. Accordingly, the effective diffusivity in these two cases is different and expressed as in Eq. 13. De f f =

D 1 − 0.5ln

(13)

Results The experimental measurements made on powder poured on the bath surface still imply more phenomena than those that were described in the previous section. Therefore, considering the dissolution behavior in a single injection is unlikely to give a proper estimation of the accuracy of the proposed method. Nevertheless, Fig. 2 presents a typical result from a single test to understand the overall raft behavior, along with a simulation run with the same conditions. It can be seen that the mass of the raft increases rapidly at the beginning of the test. Therefore, a large amount of gas is generated and escapes the raft, causing an abrupt change in the apparent weight of the raft, which is also lower due to its gas content. Also, the mass of the raft seems to attain its maximum near 35 s, when half of the raft is infiltrated and the dissolution of alumina is not quite set. After this maximum, the mass of the raft decreases almost linearly until the complete dissolution. Note here that, since the dissolution of the raft was recorded on video, the opening of the furnace induced cooling at the surface of the bath, limiting the dissolution of the top layer of the raft. For comparison, Fig. 2 also presents the apparent weight calculated with the model developed in this work, as well as the real alumina mass in the raft. Note that the bath

Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts …

627

Fig. 2 Typical gravimetric curves obtained in the parametric study compared to the output of the model

temperature is colder than the liquidus of the bulk bath for less than 20 s. This is due to the rising of the hot bath during the infiltration of the raft and because the heat transfer from the bulk bath is more intensive in the alumina particles. Therefore, a very low amount of cryolite is allowed to solidify in the lower layers of the raft, and then no solid barrier is created to stop the alumina diffusion. From Fig. 2, it can be seen that the dissolution of alumina begins in the first ten seconds of contact with the melt. However, at the beginning, only the first layers can dissolve. Then, as the liquid rises in the raft, the dissolution of the subsequent layers begins and increases the alumina mass fraction in the raft. This dissolved alumina can then reach the lower surface of the raft to be dispersed in the bulk electrolyte. Then, when enough layers dissolve, the alumina dissolution rates increase and become quite linear after 60 s, until 200 s. Afterward, the lower layers begin to disappear more rapidly, thus limiting the number of layers that effectively dissolve, and the dissolution rates decrease accordingly.

Discussion The models and hypotheses employed to build the simulations presented here have their own limitations and the coupling between them can be enhanced by appropriate models and experiments. Nevertheless, this first attempt to simulate the whole behavior of an alumina raft dissolution gives results in agreement with the experimental tests and illustrates the high potential of the described model. The simulations presented can benefit from the addition of specific values, namely for the BET surface, MOI, and LOI. The use of adequate values should increase the reliability of the model for the reasons presented here. The following considerations should be taken into account to enhance the model’s reliability.

The use of spheroid particles of alumina with constant diameter for the calculation is surely one of the most influential parameters in error mitigation. The real alumina raft has a much broader range of particle size and shape. This greatly influences the porosity of the raft, which controls the infiltration velocity and the liquid-to-solid ratio. These parameters then influence the effective thermal conductivity of the raft, the sintering of the alumina, and the effective diffusion of alumina. For more precision, a size distribution function should be used in the equations of the model. The one-dimensional simulation presented here is indeed simpler than reality and thus participates in the error of the model. In fact, a radial infiltration of the raft can be observed in the dissolution test performed. This is inherent to the size distribution of the particles as well as the angle of repose of the heap of powder. A three-dimensional approach should lead to results closer to the real behavior. However, the onedimensional simulation has the advantage of requiring less computational resources with a computed versus simulated time ratio of 2:1. The optimization of the program used to compute the model should allow for real-time simulation, which is beneficial for considering the potential coupling of the method with other mathematical models. The Lifshitz-Slyozov-Wanger theory assumes a dispersed system with a solid fraction that tends to zero. This is not the case in an alumina raft and the use of the LSW theory is here only a rough approximation of the sintering behavior of the alumina particles. Moreover, models of sintering with infiltration already exist and should be applied here (see German [20]). The use of the bath properties of the first layer in the Washburn equation gives only an estimation of the infiltrating front. In fact, the properties of the bath, at the infiltration front should be used instead. However, this only changes slightly the position of the front, since two things arise from this situation.

628

J. Alarie et al.

Fig. 3 Dissolution rate of an alumina raft at a different mass fraction of AlF3 additives

First, the decrease in temperature increases the viscosity and slows down the infiltration. Second, the solidification of cryolite, if the temperature drops under the liquidus, decreases the cryolite ratio (CR), which in turn decreases the viscosity. More work is needed here to quantify the exact influence that those two behaviors will have on the infiltration speed. However, calculations made in this work are fairly close to the observed value in experimental tests. The main difference between the simulation and the experimental results comes from the presence of gas inside the raft. The model presented here is not yet capable of simulating the desorption of the gas and its influence on the amount of liquid inside the raft. Moreover, the gas desorption creates voids in the porosities of the raft. As the gas amount in these voids increases, their volume expands, pushing the cold bath saturated with alumina outside of the raft. When these internal bubbles are released, the voids created suck more hot bath with low alumina content from the bulk liquid. Then, the formation of bubbles inside the raft limits the dissolution, while the bubble forms but increases the amount of fresh bath inside the raft.

Industrial Concerns The model presented here can greatly help the industry to see the effect of their operating conditions on the dissolution of a batch of alumina. For example, Fig. 3 shows the impact of the mass fraction of aluminum fluoride on the dissolution rate of a raft in a standard industrial bath. Figure 3 shows that an increase in aluminum fluoride has a small impact on the dissolution rate of a raft. This can be explained by the fact that the dissolution of a raft is limited by the diffusivity of the alumina between the layers of the

raft more than by the concentration gradient. Therefore, since the AlF3 content of the bath increases the diffusivity of the alumina (see Alarie et al. [1]), the raft will dissolve at a fairly similar rate. However, the increase of the AlF3 allows reducing the temperature of the cell, by about 15 ◦ C in the case presented here. This will lead to a favorable effect such as the limitation of the sintering of the alumina. Then, the raft created will be more fragile and will disintegrate more easily. Unfortunately, the actual model cannot take this parameter into account, but a more sophisticated model will be built, from the one presented here, to address these concerns.

Conclusion The dissolution of alumina rafts has been modeled by the coupling of different models already existing in the literature. This model takes into account the infiltration of the liquid, as well as the heat and mass transfer inside the raft. Results show that the dissolution of alumina occurs in the first seconds of contact with the liquid and that temperature does not limit the dissolution at this moment. This model was compared with experimental results and the difference between the two sets of results is discussed. In the end, some improvements have been suggested, but the model proves to predict the alumina raft dissolution rate with impressive precision for its simplicity. The results obtained from the model will therefore be an essential tool for anyone who wants to improve the dissolution of alumina in industrial cells. Acknowledgements The authors want to thank Rio Tinto, the Natural Sciences and Engineering Research Council of Canada, and the Fonds de recherche Nature et Technologies of Quebec, by the intermediary of doctoral scholarship, and the Aluminum Research Centre REGAL, for their technical and financial support for this project.

Dimensional Analysis Applied to the Dissolution and Disintegration of Alumina Rafts …

References 1. Alarie, J. et al. Determination of the mass transfer coefficient for the dissolution of alumina samples immersed in a cryolitic bath. Acta Materialia (Submitted) (2023). 2. Alarie, J., Kiss, L., Dion, L., Guérard, S. & Bilodeau, J.-F. Detailed analysis of alumina raft formation kinetics upon injection in cryolite. In Preparation (2023). 3. Alarie, J. et al. Validation of the gravimetric method to properly follow alumina dissolution in cryolitic bath, 680–687 (Springer, 2020). 4. Chemieingenieurwesen, V. D.-G. V. u. VDI heat atlas 2nd ed. edn (Springer, Berlin ;, 2010). URL http://public.ebookcentral.proquest. com/choice/publicfullrecord.aspx?p=3065605. 5. Yu, A. B., Bridgwater, J. & Burbidge, A. On the modelling of the packing of fine particles. Powder Technology 92 (3), 185–194 (1997). 6. Kirdponpattara, S., Phisalaphong, M. & Newby, B.-m. Z. Applicability of washburn capillary rise for determining contact angles of powders/porous materials. Journal of Colloid And Interface Science 397, 169–176 (2013). 7. Hapgood, K. P., Litster, J. D., Biggs, S. R. & Howes, T. Drop penetration into porous powder beds. Journal of colloid and interface science 253 (2), 353–66 (2002). 8. Cengel, Y. A. & Ghajar, A. J. Heat and mass transfer : fundamentals & applications Fifth edn (McGraw-Hill Education, New York, N.Y., 2015). 9. Dassylva-Raymond, V. Analyse et modélisation du comportement des agrégats d’alumine dans le procédé Hall-Héroult. Thesis (2015). URL https://constellation.uqac.ca/4121/.

629

10. Macˇedo, M. I. F., Bertran, C. A. & Osawa, C. C. Kinetics of the γ → α-alumina phase transformation by quantitative x-ray diffraction. Journal of Materials Science 42 (8), 2830–2836 (2007). 11. Bagwell, R. B. & Messing, G. L. Effect of seeding and water vapor on the nucleation and growth of α-al2o3 from γ -al2o3. Journal of the American Ceramic Society 82 (4), 825–832 (1999). 12. Shaklee II, C. A. Matrix-mediated, chemical vapor transport synthesis of alpha alumina platelet-shaped particles (The Pennsylvania State University, 1994). 13. Johnson, A. R. Alumina crusting and dissolution in molten electrolyte. Journal of Metals 34 (3), 63–68 (1982). URL : //WOS:A1982NK02400009. 14. Hill, R. F., Danzer, R. & Paine, R. T. Synthesis of aluminum oxide platelets. Journal of the American Ceramic Society 84 (3), 514–520 (2001). 15. Townsend, D. & Boxall, L. Crusting behavior of smelter aluminas, 613–621 (1984). URL https://onlinelibrary.wiley.com/doi/abs/10. 1002/9781118647851.ch91. 16. Østbø, P. N. Evolution of alpha phase alumina in agglomerates upon addition to cryolitic melts. Ph.D. thesis, Trondheim (2002). 17. Kang, S.-J. L. Sintering: densification, grain growth and microstructure (Elsevier, 2004). 18. Weissberg, H. L. Effective diffusion coefficient in porous media. Journal of Applied Physics 34 (9), 2636–2639 (1963). 19. Kim, J.-H., Ochoa, J. A. & Whitaker, S. Diffusion in anisotropic porous media. Transport in Porous Media 2 (4), 327–356 (1987). 20. German, R. M. Sintering theory and practice (1996).

Fundamental Loss of Current Efficiency During Aluminium Electrolysis and Its Correlation with Sodium Content Dissolved in the Aluminium Lukas Dion and Paul Desclaux

Abstract

Primary aluminium producers are inclined to maximize the cells’ current efficiency in order to enhance metal production and reduce production cost. For timely decisions, sodium content of the cathodic aluminium has been used as a performance indicator related to individual cell performances. This paper pinpoints the straightforward quantitative theoretical relation which exists between sodium content and current efficiency. This relation is based on the fundamental thermodynamics of the changing bath composition in the boundary layers and on mass transfers at the anodic and cathodic interfaces. Few simplifying hypotheses are used to predict the cell’s optimal current efficiency under a specific set of operating conditions. The proposed calculation methodology is described and a critical discussion is performed to highlight the impact of different factors on the current efficiency along with future considerations necessary to improve the current efficiency estimations. Keywords











Current efficiency Sodium content Aluminium dissolved Diffusion coefficients Transfer equilibrium

Introduction Primary aluminium production using the Hall–Héroult process has been ongoing for more than a century. Nonetheless, even now, there are very little practical ways to actually L. Dion (&) GRIPS, Université du Québec À Chicoutimi, 555 Boulevard de L’université, Chicoutimi, QC G7H 2B1, Canada e-mail: [email protected] P. Desclaux 1685 Castner, G7S 3A5, Jonquière, QC, Canada

improve the amount of aluminium produced by an aluminium smelter. The simplest solution is to increase the number of cells available to produce the aluminium, with the considerable offside of requiring capital, in the range of hundreds of millions of dollars, and the physical space for such additions. A more practical solution is to increase the cell amperage, which, by definition, should increase the amount of aluminium produced proportionally. However, this solution brings additional heat input for the electrolysis cell, which needs to be extracted from the cell to avoid disruption of the cell stability. Therefore, many aluminium producers, especially in regions with low energy cost, are consistently working in this direction, but significant changes require technological change applied over multiple years. The other available solution is to improve the current efficiency of the cells. This indicator, representative of the ratio of metal produced in comparison to the theoretical amount, is sensitive to a large number of factors such as the cell stability, thermodynamic conditions, magnetohydrodynamics, and chemistry. It is therefore unsurprising that the current efficiency of electrolysis cells has been known to be different from cell to cell, and even variable over time in specific cells based on the operating conditions. The typical current efficiencies for recent cell technologies presented in the literature fluctuate between 90 and 95%. This paper presents a mathematical relation to estimate the theoretical maximum current efficiency of electrolysis cells. These set of equations are dependent on the fundamental thermodynamics resulting from the change in bath composition occurring in the boundary layers at the anodic and cathodic interface.

Previous Research on Current Efficiency Due to the fluctuating nature of this indicator, and its importance for aluminium productions, the relationship between current efficiency and other parameters has been

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_81

630

Fundamental Loss of Current Efficiency During Aluminium …

631

studied quite heavily in the past. The current efficiency, defined by Eq. (1), is a ratio between the mass of metal produced to its theoretical amount. CE% ¼

Amount produced P ¼ MIt Theoretical amount vF

ð1Þ

where P is the amount of metal produced in [kg], M is the molar mass of the aluminium [kg/mol], I is the cell current [A], t is the time of production, v is the valence of the metal (v = 3 for Al), and F is Faraday’s constant of 96,485 [C/mol]. While the theoretical amount of metal is only dependent on time and the current intensity, the practical amount produced has a number of possible variations and all analyses generally investigate the factors associated with the loss of current efficiency. These factors can be classified into different categories.

Loss of Current Efficiency Caused by the Re-Oxidation of Aluminium and Aluminium Ions As the aluminium is produced in the electrolysis cells, it accumulates at the bottom of the cell until it is tapped and transported to the cast house. However, the aluminium can dissolve in the electrolysis bath under specific conditions, form a metal fog, and come in contact with the carbon dioxide produced resulting in the occurrence of a back reaction (Eq. 2), causing a loss of metal in the cell. 2Al þ 3CO2 ¼ Al2 O3 þ 3CO

who may contribute significantly to the loss of current efficiency [6–10].

Loss of Current Efficiency Caused by Electronic Transport and Short Circuits During electrolysis, a few hundred thousand amperes are generally passing through the cell. However, if the current has a resistance path of lower value outside the electrolyte, this portion does not contribute to metal generation and contribute to the loss of current efficiency. An important phenomenon reported to affect the loss of current efficiency is the electronic transport within the electrolyte. Some authors [11–13] investigated this phenomenon to quantify its importance but different results exist with regard to its magnitude. In addition to possible electronic conduction, short circuits may occur when a segment of the carbon anode doesn’t consume as rapidly as the rest of the anode, if an anode is set unproperly, or if deformation of metal pad is of significant importance. Under such circumstances, the anode may be in direct contact, temporarily or permanently, with the aluminium; therefore, offering a direct path without the need to cross the electrolyte. While this phenomenon is known to occur periodically, it is rarely investigated under industrial conditions due to the detrimental nature of this occurrence on the process. Hyland [14] reproduced numerically this phenomenon and the results indicated that a significant amount (10–80%) of the anode current is shorted for the duration of the events.

ð2Þ

The occurrence of the back reaction is considered a major contributor to the loss of current efficiency and was studied by numerous investigators [1–5] when attempting to understand the current efficiency’s behavior.

Loss of Current Efficiency Caused by Electrolysis of Other Metals The electrolyte contains a significant number of constituents from the complex chemistry necessary for sustaining and optimising the process. Among these, the most common impurities are iron oxides, calcium, silicon and phosphorus, which are consequences from the raw materials used for the aluminium production. During the electrolysis, these impurities are also reduced electrochemically, therefore wasting portion of the current which did not produce aluminium while also contaminating the metal pad in the cell. There are also consistent reports in the industry of cyclic patterns of oxidoreduction associated with the phosphorus

Correlations Between the Current Efficiency and Process Indicators As mentioned, the loss of current efficiency may be caused by a multitude of factors. While attempts have been made to develop comprehensive model of these phenomenon in a coupled or uncoupled way, primary aluminium producers are satisfied with accurate correlations between the current efficiency performances of a cell, or smelter, and other key process indicators (KPIs) that are easier to measure as part of the operational routine. A recent and extremely comprehensive statistical model was developed by Côté et al. [15] using process data collected from 30 different potlines over a decade of operations. The strategy was to develop a predictive tool to estimate the current efficiency of potlines in real time based on a set of specific KPIs which would allow to act rapidly once a drift in behavior is detected. The model identified seven different predictors: Anode–cathode distance (ACD), solubility of metals in the bath, iron content in bath, cell instability

632

multiplied by current, metal height, pot age, and average duration of the control alumina feed tracking. While some indicators were clearly defined as a cause of the loss of current efficiency, other indicators were simply a direct, or parallel consequence of the loss of current efficiency. Nonetheless, both type of indicators is extremely relevant to anticipate the current efficiency of potlines over an extended period of time. Therefore, this strategy is extremely beneficial for aluminium producers to facilitate long-term decisions or to detect drift in the process that can affect the performances of the overall cells. However, there are numerous reports in the literature that present the sodium content dissolved in the metal as a relevant indicator to estimate the current efficiency of individual cells. The articles of interest on this topic provided two main explanations to corroborate this correlation. The first explanation considers the migration of sodium fluoride anions towards the cathodic boundary layer to support the transport of electrical charges [16–22]. Due to the equilibrium between the fluoride complexes in the cathodic boundary layer of the bath, the increase of the concentration of sodium fluoride is the result of a lower transfer rate at the interface which correlates with a slower aluminium dissolution rate. The second explanation pushes the concepts further and assumes the possible formation of a thin cryolite layer at the bath–metal interface [9, 23, 24]. It was hypothesized that conditions in a stable boundary layer may provide sufficient changes in the liquidus temperature of the bath to provide solidification of a bath layer at the bath–metal interface (BMI) which would corroborate the significant migration of sodium fluoride described in the previous paragraph.

Theoretical Interaction Between Current Efficiency and Sodium Dissolved in Metal Main Hypothesis This demonstration is a first step towards a comprehensive and unified model for the theoretical estimation of the loss of current efficiency. Therefore, the validity of this model is constrained by the five main hypotheses at the base of this theoretical work. (1) The loss of current efficiency is considered to be exclusively attributed to the oxidation of the aluminium dissolved in the electrolyte at the anode. All other mechanisms (undissolved Al reactions, parasite reactions, short circuit, and electronic currents) that contribute to the loss of current efficiency are not considered.

L. Dion and P. Desclaux

(2) The electrolysis cell is considered exclusively in steady-state conditions. It implies that the perturbations and wave at the bath–metal interface (BMI) are periodic and with a similar amplitude. It only considers that the BMI is free and that the anodic and cathodic boundary layers are stable over time. (3) The main medium considered in this work is the electrolytic bath. Its temperature is considered stable and homogenous in all its volume, including within the boundary layers of the bath. (4) The composition of the electrolytic bath is constant over time. However, the chemical concentration of the constituents within the boundary layer is allowed to change as a function of its distance with the boundary. (5) The migration of NaF ions is the only contributor that allows the passage of electrical charges through the electrolytic bath.

General Equation of Current Efficiency From the definition of the current efficiency described in Eq. 1, we can redefine the considered variables as: g¼

m mo  mp mp ¼ ¼1 m0 mo mo

ð3Þ

where η is the current efficiency, mo is the theoretical amount of aluminium produced, and m is the metal produced by the cell. This last variable can be further refined as the difference between the theoretical amount produced and an amount of metal loss (mp). From Eq. 1, we know that over an equivalent period of time, a specific amount of aluminium can be considered as a specific amount of current. Thus it is possible to rewrite Eq. 3 as: g¼1

Ip Io

ð4Þ

where Io is the total current of the cell and Ip is the current corresponding to the amount of aluminium (mp) loss by oxidation at the anode. As this reaction occurs under stable conditions, following our second hypothesis, we can assume the concentration state of dissolved aluminium in the electrolysis cell corresponds to Fig. 1. Here the dissolved aluminium concentration at the interface with the aluminium is equivalent to the saturation of the

Fundamental Loss of Current Efficiency During Aluminium …

633

where 3 is the electronegativity of the aluminium and F represents Faraday’s constant. Then, by inserting back this term into Eq. 4, knowing that the total current of the cell (I0) is the product of the anodic current density (ia) by its surface (Sa). We are able to obtain Eq. 8 describing how to theoretically calculate the current efficiency based on the mass transfer at the interface of our system. g¼1 Fig. 1 Concentration of aluminium dissolved in the bath as a function of the position in the system considered

aluminium in the bath (Cs) and the aluminium concentration at the carbon is equivalent to zero (C0) due to the backreactions (Eq. 2) that occur with the CO2 generated from the electrolysis. The considered system is in a steady state of transfer, thus the average concentration of aluminium in the bath (Cav) reaches a value that is directly dependent on the equilibrium between the rate of kinetic transfer of the dissolved aluminium within both opposite boundary layers. This equilibrium is described in Eq. 5 and the isolated variable Cav can be obtained from Eq. 6. Sc
kc
½C s  C av  ¼ Sa
ka
C av Cav ¼

Sc
k c
C s Sa
k a þ Sc
k c

ð5Þ ð6Þ

where kc and ka are the transfer coefficient of dissolved aluminium at the cathode and at the anode, respectively, while SC and Sa correspond to the respective cathode and anode surface areas available for mass exchange. Following hypothesis #2, #3, and #4, all variables from Eq. 6 can be considered constant through time; Sc and Sa are dependent on the geometry of the cell. The saturation concentration (Cs) is dependent on the bath chemistry and temperature while kc and ka are dependent on the hydrodynamic conditions in the respective boundary layers. By knowing that both sides of Eq. 6 are at equilibrium, and more precisely considering that the right side of the equation directly represents the continuous flow of aluminium that is being loss through the backreaction. It is possible to define the metal loss term (Ip) from Eq. 4 by isolating the left part of Eq. 5 and merging it with the results from Eq. 6 as presented in the next equation. I p ¼ 3
F
Sa
k a
C s


Sc
k c Sa
k a þ Sc
k c

ð7Þ

3
F
ka
C s 1
a ia 1 þ SSa
k
k c

ð8Þ

c

As a result, from the previous equation, only three variables still appear as undefined: Cs, ka, and kc. The identification transfer coefficients (kc and ka) require a deeper analysis and will be developed in the following section. However the saturation coefficient of the aluminium in the cryolite has been the subject of numerous studies and can be identified through experimental work [25–28] or thermochemical calculations [29] based on the chemistry conditions and bath temperature.

Identification of the Transfer Coefficients In order to obtain the transfer coefficients of specific species at the interface of interest, the correspondence with the diffusion coefficients of the respective species will be used as described by Levich [30]. k1 ¼ k2


23 D1 D2

ð9Þ

Accordingly, the relationship between the transfer coefficient of two respective species in the same medium is dependent on the ratio of their diffusion coefficient. Using Eq. 9, the goal is to determine k1, the transfer coefficient of the aluminium at the anode and cathode interface. In both cases, it will be required to identify the diffusion coefficient of the dissolved aluminium (DAl) in the electrolyte. Odegard [31] performed linear sweep voltammetry and chronopotentiometric measurements in laboratory conditions using alumina saturated cryolite to obtain a diffusion coefficient of 10.6 ± 1.1  10–5 cm2 s−1. Dewing and Yoshida [13], using a laboratory setup to perform current efficiency measurement, reported a value of 2.5  10–4 cm2 s−1. Similarly, Xiang et al. [32] reported current efficiency measurements in a small laboratory setup and identified a diffusivity of 9  10–5 cm2 s−1 for the aluminium dissolved in the bath. For the purpose of this work, the results from Dewing and Yoshida will be considered.

634

L. Dion and P. Desclaux

kAl2 O3

ia ¼ 6
F
DAl2 O3

ð10Þ

5E-09

Diffusion coefficient (m 2/s)

To consider the reaction at the anode interface, the secondary species considered will be the alumina. In the boundary layer region close to the anode, the alumina concentration will drop significantly as it is contributing to the electrolysis reaction. The lower limit concentration of the alumina at the anode interface is such that it maintains stable electrolysis conditions without leading to anode effect. Therefore, it is possible to calculate the transfer coefficient of the alumina as a quotient of the alumina concentration difference between the bulk and the anode interface divided by the current intensity, as reported in Eq. 10. To simplify, the transfer coefficient can be calculated as the required amount of alumina moving through the boundary layer to sustain the electrolysis process under steady-state conditions

4.5E-09 4E-09 3.5E-09 3E-09 2.5E-09

950

960

970

980

990

1000

1010

1020

Bath Temperature (°C) BR = 1,05

BR = 1,1

BR = 1,2

BR = 1,25

BR = 1,15

Fig. 2 Diffusion coefficient of alumina as a function of bath temperature and bath ratio (calculated using the Wilke–Chang equation)

where kAl2O3 is the transfer coefficient of the alumina, ia is the anodic current density, and DAl2O3 is the difference of concentration between the bulk and the anodic interface within the boundary layer. Based on results from prior work performed under industrial conditions by the authors, a transfer coefficient for the alumina is proposed as 5.75  10− 5 m/s. The diffusion coefficient of the alumina can be determined from previous literature performed under experimental conditions. While this was the results of numerous investigations, a recent study by Alarie et al. [33] considered this molten salt system to estimate the diffusion coefficient of alumina under different sets of operating conditions. The results from Alarie et al. reported diffusion coefficients for the alumina in the range of 2.51 to 3.29  10–9 m2/s for cryolitic ratio between 2.1 and 2.4 and bath temperature between 950 and 990 °C. A similar strategy was used to estimate the alumina diffusion coefficient in this work. The resulting diffusion coefficient for different bath temperature or bath ratio is presented in Fig. 2. Inserting the respective value of the transfer coefficient of the alumina, along with the diffusion coefficients of alumina and aluminium into Eq. 9, the transfer coefficient of the aluminium dissolved in molten cryolite can be estimated to be between 2.22 and 2.66  10–4 m2/s. The remaining variable to identify is the transfer coefficient of the aluminium on the cathodic side of the system. The identification of this variable is a little more strenuous and requires the consideration of the equilibrium at the BMI directed by the following equation: Al þ 3NaF $ 3Na þ AlF 3

ð11Þ

where the liquid aluminium reacts with the sodium fluoride at the BMI to generate aluminium fluoride and dissolves the sodium. The equilibrium constant (K) described in Eq. 12 has been investigated and reported in previous literature as a function of the bath temperature. KðTÞ ¼

a3Na
aAlF 3 a3NaF

ð12Þ

This equilibrium is ongoing at the BMI. As such, the concentration of NaF and AlF3 will be significantly different than the concentration in the bulk and therefore the activities of those respective species cannot be as easily defined as if it was in the bulk of the bath. To circumvent this challenge, it is necessary to know the sodium concentration of the metal pad in electrolysis cells. Special consideration needs to be taken to avoid sodium volatilisations during the sampling, but accurate techniques were developed for this specific task, such as the Heraeus Electro-Nite sampler QS3012ACC used in the work of Chollier et al.[34]. Once the concentration of sodium is known in the aluminium, it is possible to evaluate the sodium activity in the metal by converting the concentration using the activity coefficient of sodium [35] (cNa) and the known concentration ([Na]ppm) as shown in Eq. 13. The molar masses (MAl and MNa) of the aluminium and sodium, respectively, need to be taken into consideration to obtain the proper units. aNa ¼ cNa


M Al
½Nappm
106 M Na

ð13Þ

Fundamental Loss of Current Efficiency During Aluminium …

635

With a known concentration, it is possible to directly calculate the sodium activity in the metal, which, in turn, can be inserted back into Eq. 12 to obtain the ratio of activities between AlF3 and NaF. aAlF 3 K ðT Þ ¼ 3 a3NaF aNa

ð14Þ

Using the activities of NaF and AlF3 from Dewing [36], corrected for the operating temperature, the ratio at the aluminium interface is obtained from the equilibrium provided by Eq. 14. The bath ratio in the bulk is already known as it is considered constant and in control for the cell. Consequently, the difference in the sodium fluoride concentration between the bulk and the BMI (DNaF) can be calculated. Following hypothesis #5, the transport of the electric charges is supported exclusively by the transport of NaF ions. Consequently, it is possible to determine the transfer coefficient of NaF ions in this boundary layer as cathodic (Sc) and anodic (Sa) surfaces are known. kNaF ¼

Sa ia ic ¼ Sc F
DNaF F
DNaF

ð15Þ

Once kNaF is identified, we can apply Levich equation (Eq. 9) to estimate the value of the transfer coefficient of the dissolved aluminium on the cathode side. This is possible due to the availability of the diffusion coefficient of the sodium fluoride in the literature. Finally, by combining Eq. 9 with Eq. 10 properly adapted for the respective cathodic or anodic constituents, we can simplify the system to obtain the following equation to estimate the current efficiency of an electrolysis cell (η). Here DNaF is the only unknown variable that needs to be determined from measurements of the sodium in the metal:

g¼1


3
F
C s
kAl2 O3
1

1þ




DAl DAl2 O3

23

ia

F
DNaF
kAl2 O3 ia

ð16Þ

2


ðDDAlNaFO Þ3 2 3

which with further mathematical simplification leads to the following equation: 2

g¼1

3
C S
D3Al 2

2

6
DAl2 O3
D3Al2 O3 þ DNaF
D3NaF

ð17Þ

Results and Discussion Following Eq. 17, it is possible to theoretically calculate the respective current efficiency of an electrolysis cell if we know its “real” sodium content, its bath temperature, and the bath chemistry. As such, it is possible to estimate the equilibrium conditions between the sodium dissolved in the metal and the current efficiency of the cell for known cell conditions. The results from Fig. 3 illustrates the equilibrium behavior between the sodium dissolved in the aluminium and the current efficiency for different bath temperature. The general behavior resulting from the proposed equations is similar to previous publications [17] [37] investigating the behavior of individual cells. The logarithmic behavior of this relationship may explain why this indicator is not highly effective, by itself, to rapidly detect drift in complete potlines. In the upper band of the current efficiency, the behavior is almost linear and important change in the sodium concentration may occur (±20 ppm) for only a small variation of the current efficiency (± 1%). By taking into additional consideration that the sodium concentrations from this figure are “true” concentration, the uncertainty of industrial comparison is even increased due to the volatilization of sodium during standard aluminium samples taken for impurity composition. On the other hand, cells with a lower current efficiency will rapidly drop in current efficiency with little effect on the sodium content. Whereas a drop of only 10 ppm of sodium may be associated with a loss of more than 4% of current efficiency. Additionally, it is possible to observe that the bath temperature will change the absolute equilibrium value of the sodium by approximately 3 ppm (low RF) to 6 ppm (high RF) for every variation of 10 °C for the bath temperature. However, it has a negligible impact on the measured slope of the correlation between sodium and the current efficiency. When investigating the effect of the bath ratio on the equilibrium between current efficiency and sodium dissolved in the metal, it is possible to observe that the behavior is quite different than with the bath temperature, as depicted in Fig. 4. It is possible to observe that the bath ratio changes considerably the possible range of sodium content dissolved in the metal over a similar span of current efficiency. Therefore, cells with a bath ratio of 1.3 would operate with sodium variations expected in a narrower range than cell operating with a bath ratio closer to 1.0, thus affecting the easiness of a smelter to use this indicator to distinguish individual cell behavior.

636

L. Dion and P. Desclaux

Fig. 3 Equilibrium between current efficiency and sodium content in the aluminium as a function of the bath temperature (bath ratio = 1, 1 and current density = 0.8 A/cm2)

Fig. 4 Equilibrium between current efficiency and sodium content in the aluminium as a function of the bath ratio (bath temperature = 970 °C and current density = 0.8 A/cm2)

The remaining element to investigate is the impact of the current density. As the current density increases, the mass transfer of ions in the boundary layer needs to increase as well to sustain the process without perturbations, which leads to a variation of the equilibrium conditions. However, the results from Fig. 5 demonstrate that these variations are negligible in comparison to the effect caused by differences

in bath temperature or bath ratio, whereas a difference of 0.2 A/cm2 leads to a fluctuation of barely 2 or 3 ppm in the sodium content dissolved in the aluminium.

Recommendation for the Industry Aiming for Maximum CE

Fig. 5 Equilibrium between current efficiency and sodium content in the aluminium as a function of the current density and bath temperature (bath ratio = 1.1)

As demonstrated in this work, a theoretical correlation exists between the sodium content dissolved in the aluminium and the current efficiency of the cells. This correlation is based on different equilibrium in three separate regions (bath– metal interface, cathodic boundary layer, and anodic boundary layer) where different transfer processes will occur. Ultimately, the current efficiency of electrolysis cells can be improved if we mitigate the conditions in favor of the mass transfer of the elements. By considering the asymptotic part of the curves depicted in Fig. 3, 4, and 5, it is possible to validate that the results are in agreement with previous industrial observations. Thus, it is possible to improve the theoretical maximum current efficiency value of a cell by (a) reducing the bath ratio of the cell, (b) reducing the bath temperature, and c) reducing the current density.

Fundamental Loss of Current Efficiency During Aluminium …

Finally, the theoretical model described in this paper may face limitation with lower performing cells. As the concept is based on mass transfer equilibriums, there is a theoretical limit reached for all cell conditions where the mass transfer kinetic is unable to sustain the additional transfer of aluminium dissolved required to explain further losses in current efficiency. This theoretical lower limit is generally at a current efficiency between 88 and 94%. While it is possible for cells to operate with current efficiency underneath this threshold, additional mechanism is required to theoretically explain these occurrences.

Cluster Identifications of Highand Low-Performing Cells Following the observations presented in Fig. 3, 4, and 5, we can observe that estimating the current efficiency, under industrial conditions, based on the sodium content in the metal is theoretically possible, but remains a challenging task. Nonetheless, the authors propose two solutions in order to use this concept, and therefore the sodium as a potential indicator of cell performances. A. Qualitative solution: As indicated clearly in Fig. 3, the behavior in the upper and lower range of current efficiency is nearly asymptotic. As such, each individual cell should have a near-constant chemistry (above 10% in excess AlF3) and temperature target, while operating with a constant current density. It is assumed that the standard deviation from multiple sequential sodium measurements should be combined with the absolute value to categorize the performances of the cell as a highly or poorly performing pots. Therefore, highly performing pots should have an average concentration of sodium above the median of the potline combined with an important standard deviation. On the contrary, poorly performing pots should have an average concentration of sodium below the median average combined with a narrow standard deviation between sequential measurements (e.g. 1 week with one measurement per day). B. Quantitative solution: The equation system presented in this article could easily be implemented in the control software of electrolysis cells or in the historical data archives. As such, it could be possible to estimate a precise current efficiency value for every corresponding “true” sodium concentration measurements as long the cell chemistry, temperature, and current density are

637

known. This current efficiency estimations could be used as a key performance indicator to categorize cells into differently performing clusters and adjust other parameters accordingly to optimize process. However, the main issue is that the “true” sodium concentration is rarely known. Therefore, a corresponding equivalence should be established as an additional input to the equation system to get a proper estimate of the “true” sodium concentration. This equivalence can be determined (a) empirically from trial and error using individual and potline cell data, (b) experimentally by performing comparative measurements between the Heraeus Electro-Nite sampler and typical measurements, (c) by investigating the theoretical volatilization behavior of sodium during metal sampling, or (d) by a combination of these three methods. Such work is left for future research at the current time.

Conclusion This paper demonstrated a theoretical set of equations to correlate the sodium content dissolved in the aluminium with the current efficiency of an electrolysis cell. These equations are based on the equilibrium at the bath–metal interface and the transfer kinetics occurring at the anodic and cathodic boundary layer. Based on the measured content of sodium in the aluminium, this system allows the calculation of the aluminium dissolution rate in the electrolytic bath which culminates as aluminium which re-oxidizes and causes a loss of current efficiency. Using the equations, the equilibrium system was explored and identified that the bath ratio is the factor which affects the most the correlation between sodium and current efficiency. It also has the most important impact on the theoretical maximum value of the current efficiency that can be obtained. Bath temperature was highlighted as the second most important contributor to this equilibrium and current density, by itself, has a negligible impact on the correlation between current density and sodium content. Ultimately, these equations can provide significant insights for the development of strategies or tools to follow-up individual cell performances. Using these tools, clusters of “low”- and “high”-performing cells can be identified and thorough analysis can be performed to optimise the performances of individual electrolysis cells, leading to significant improvement in the overall smelter production and efficiency.

638

References 1. Lillebuen, B., et al., Current efficiency and back reaction in aluminium electrolysis. Electrochimica Acta, 1980. 25(2): pp. 131–137. 2. El-Demerdash, M.F., S.M. El-Raghy, and F.M. El-Daw, Three dimensional model for current efficiency based on the rate of aluminium transfer to anode, in TMS - Light metals, C. Eckert, Editor. 1999. p. p. 353–358. 3. Haarberg, G.M., Formation of Metal Fog and Dissolved Metals During Electrodeposition from Molten Salts. ECS Proceedings Volumes, 2002. 2002–19(1): p. 789. 4. Liu, Z., et al., Current efficiency predictive model and its calibration and validation, in TMS - Light metals, C. Suarez, Editor. 2012. p. 935–938. 5. Dassylva-Raymond, V., Prédiction de l'efficacité de courant du procédé Hall-Héroult, in Applied science. 2009, Université du Québec à Chicoutimi: Chicoutimi. 6. Keller, R. Mass transfer in advanced aluminum electrolysis. 1984. 7. Keller, R., Laboratory study on the reaction of impurities in aluminum electrolysis. Light Metals, 1982: p. 295–298. 8. Sterten, A., P.A. Solli, and A. Solheim, in Al-Symposium. 1995: Donovaly, Slovakia. p. 209–219. 9. Sterten, A., P.A. Solli, and E. SkyBakmoen, Influence of aluminium impurities on current efficiency in aluminium electrolysis cells. Journal of Applied Electrochemistry, 1998. 28: p. 781– 789. 10. Chollier-Brym, M.-J., et al., Factors Influencing the Distribution of Impurities in the Metal of Hall–Héroult Pots and Their Impacts on Current Efficiency. JOM, 2019. 71(3): p. 1169–1174. 11. Haarberg, G.M., et al., The role of dissolved metal during electrodeposition of aluminium from cryolite-alumina melts, in TMS - Light metals, W. Schneider, Editor. 2002. p. p. 1083–1089. 12. Morris, D.R., A Mathematical Model of the Alumina Reduction Cell. ECS Proceedings Volumes, 1976. 1976–6(1): p. 469. 13. Dewing, E.W. and K. Yoshida, Electronic conductivity in cryolite-alumina melts? Canadian Metallurgical Quarterly, 1976. 15(4): p. 299–303. 14. Hyland, W., The Current Efficiency of a Shorted Anode in a Prebake Cell. Light Metals, 1984: p. 711–720. 15. Côté, P., et al., Predicting Instability and Current Efficiency of Industrial Cells, in Light Metals 2017. 2017. p. 623–629. 16. Tabereaux, A., in The international Harald A. Oye Symposium. 1995: Trondheim, Norway. p. 115–127. 17. Tabereaux, A.T., The role of Sodium in Aluminium Electrolysis : A possible indicator of cell performance, in TMS - Light Metals, W. R. Hale, Editor. 1996. p. 319–326. 18. Thonstad, J., et al., The content of sodium in aluminium in laboratory and in industrial cells. TMS - Light Metals, 2001.

L. Dion and P. Desclaux 19. Haupin, W.E., Understanding boundary layers, in TMS - light metals, R. Huglen, Editor. 1997. p. 319–323. 20. Danielik, V., P. Fellner, and J. Thonstad, Content of sodium and lithium in aluminium during electrolysis of cryolite-based melts. Journal of Applied Electrochemistry, 1998. 28: p. 1265-1268. 21. Fellner, P., et al., The content of sodium in aluminium during electrolysis of the molten systems Na3AlF6–NaCl–Al2O3 and NaF–NaCl. Electrochimica Acta, 2004. 49(9–10): p. 1505–1511. 22. Keller, R., J.W. Burgman, and P.j. Sides, in Light Metals. 1988. p. 629–631. 23. Solheim, A., Crystallization of cryolithe and alumina at metal-bath interface in aluminium reduction cells, in TMS - Light Metals, W. Schneider, Editor. 2002. p. 225–230. 24. Kent, J.H., Journal of Metals (JOM), 1970. 22(11): p. 30–36. 25. Thonstad, J., The solubility of aluminium in NaF-AlF3-Al2O3 melts. Canadian Journal of Chemistry, 1965. 43(12). 26. Yoshida, K. and E.W. Dewing, The apparent solubility of aluminium in cryolite melts. Metallurgical Transactions, 1972. 2 (July): p. 1817–1821. 27. Danielik, V., et al., Solubility of Aluminum in Cryolite-Based Melts. Metallurgical and Materials Transactions B, 2009. 41(2): p. 430–436. 28. Arthur, A.M., The Solubility of Aluminum in Cryolite-Alumina Melts and the Mechanism of Metal Loss. Metallurgical transactions, 1974. 5: p. 1225–1230. 29. Bale, C., et al., FactSage 6.1. 1976–2010. 30. Levich, V.G.e., Physicochemical Hydrodynamics, ed. I. Prentice-Hall. 1962, United States of America. 31. Odegard, R., On the Electrochemistry of Dissolved Aluminium in Cryolitic Melts. Electrochimica Acta, 1988. 33(4): p. 527–535. 32. Nai Xiang, F., K. Grjotheim, and H. Kvandet, Current Efficiency Measurements in Laboratory Aluminium Cells—VIII. Current, Temperature and Cathode Alloy Composition (Al–Cu), Al-Diffusivity. Canadian Metallurgical Quarterly, 1986. 25(4): p. 287–291. 33. Alarie, J., et al., Determination of the alumina diffusivity and dissolution rate for alumina samples immersed in a cryolitic bath. Materiala, 2023. 34. Chollier-Brym, M.-J., et al. Factors Affecting Current Efficiency of Hall-Héroult Process Based on the Variation of Sodium Content in Pot Metal. in ICSOBA. 2016. Quebec City. 35. Yao, P.C. and D.J. Fray, Sodium activity determinations in molten 99.5% aluminium using solid electrolytes. Journal of Applied Electrochemistry, 1985. 15: pp. 379–386. 36. Dewing, E.W., Thermodynamics of the System NaF-AIF3: Part VI. Revision. Metallurgical Transactions B, 1990. 21B(April): p. 285– 294. 37. Dion, L., et al., Sodium content in aluminum and current efficiency —Correlation through multivariate analysis, in Light Metals, M. Cootsey, Editor. 2013: San Antonio, Tx, USA.

Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis David J. Jarvis, Rosanna E. van den Blik-Jarvis, Rosie F. L. Mellor, and Alf Bjørseth

Abstract

The global aluminium industry is trying to reduce its carbon footprint, in line with climate change policy, and great strides are being made in the use of hydropower, solar, and geothermal electricity, as opposed to fossil-fuel powered electricity. Two major smelting challenges remain however: finding inert anodes and wettable cathodes. VSCA, a European materials company, has invented and developed two promising solutions: (1) A nickel-based anode alloy that has been successfully tested for over a week at 960 °C in a Hall-Héroult cell with limited corrosion or passivation; (2) A wettable TiB2 cathode that has survived over 3 months of continuous exposure to liquid aluminium, again at 960 °C, without any signs of degradation or dissolution. Combining an inert anode with a wettable cathode is considered to be the ‘holy grail’ of aluminium electrolysis. This paper shows that green aluminium with near-zero CO2 is within grasp and could be realized with further testing and industrial scale-up. Keywords





Aluminium Electrolysis cathodes CO2 reduction








Inert anodes Wettable Sustainability

Introduction The Hall-Héroult electrolysis process, invented in 1886 [1, 2], is currently the only commercial method for producing primary aluminium. The process requires large amounts of

D. J. Jarvis (&)
R. E. van den Blik-Jarvis
R. F. L. Mellor
A. Bjørseth VSCA, Gaustadsalléen 21, 0349 Oslo, Norway e-mail: [email protected]

electrical energy, typically 14 kWh per kilogram of aluminium [3]. In addition, the process consumes carbonaceous anodes as a necessary part of the electrochemical reaction, as shown in Eq. (1): 2Al2 O3ðdissolvedÞ þ 3CðsÞ ! 4AlðlÞ þ 3CO2ðgÞ

ð1Þ

When combining the electrical- and anode-related emissions, every kilogram of aluminium leads to 10 kg of CO2 emissions, on average around the world [3]. This can be reduced significantly by choosing to use renewable energy sources as the electrical input. Indeed, great strides are now being made in the use of hydropower, solar and geothermal energy, instead of fossil-fuel-powered electricity. Numerous large smelters around the world are thus located next to water reservoirs, solar parks and geothermal power plants. While this is very positive for reducing CO2 emissions, it is still only part of the challenge. The other part of the environmental challenge relates to the two electrode materials inside the cell, namely the anode and the cathode. As mentioned already, carbon is the current material of choice for anodes and it has been since the 1880s. Somewhat ironically, Charles Hall emphasized the use of non-carbonaceous anodes in some of his original patent applications in 1886 (see Fig. 1) [1]. Consumable carbon was mentioned more as a temporary solution until better, more inert anodes could be found that had improved oxidation and fluoridation resistance. Sadly, these inert anodes have not been forthcoming and carbon-free electrolysis has not been realized in the aluminium sector, despite considerable R&D and investments over many decades [4–9]. Another important area of materials technology inside an aluminium smelter is the cathode. Again, carbon is currently the material of choice for the cathode. The downside of this is that liquid aluminium has poor wettability with carbon and this leads to magneto-hydrodynamic movement and sloshing of the metal pad, with the serious risk of short circuits between the anode and the metal. Increasing the anode-

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_82

639

640

Fig. 1 Excerpt from claim 1 of Charles Hall’s original patent application from 1886 [1]

cathode distance (ACD) to 5–6 cm not only gives extra safety margin but also increases cell resistance and, consequently, requires more electricity. This adds more cost and generally creates more CO2 emissions. If a wettable cathode surface could be implemented in a cell, then the ACD could potentially be reduced down to 2 cm, which saves on electricity—not to mention increasing the overall lifetime of the cell itself. Energy savings of 10–20% are achievable using wettable and/or drainable cathodes made of TiB2, for example [10–13]. This paper will concentrate on these two material challenges: the quest for an inert anode and the need for a wettable cathode. Finding viable industrial solutions to both would be ground-breaking, especially when coupled with renewable electricity sources. The authors believe that green aluminium with near-zero CO2 is within grasp and must be a priority for the entire aluminium industry in all countries.

Experimental Studies This experimental study focuses on two novel Hall-Héroult cell electrodes, namely a non-consumable anode and a wettable cathode. The research has been carried out by VSCA, a privately-owned materials and manufacturing company (N.B. VSCA is an acronym for Very Special Ceramics and Alloys). This section is subdivided into processing of anode materials, processing of cathode materials and material characterization.

Anode Design and Processing Geochemical inspiration was drawn from the natural oxide and oxyfluoride minerals that co-exist with molten cryolite in Nature, for example at the Ivigtut cryolite mine in Greenland (N.B. the mined cryolite from Ivigtut was used by both Hall and Héroult in the 1880s) and the Pitinga cryolite mine in Brazil [14]. These large natural cryolite deposits contain about 20 associated pegmatitic minerals that were

D. J. Jarvis et al.

thermodynamically stable in hot molten magmatic cryolite for many millennia during the Earth’s formation. Most of these minerals contain ‘high field strength elements’ (HFSEs) such as niobium, tantalum, yttrium, iron, manganese, tin, zirconium, thorium, uranium, and rare earths. Therefore, it was decided to focus on some of these HFSE alloying additions in our alloy design, with the ultimate intention of forming in situ, cryolite-resistant, mineral coatings when used as a smelter anode [15]. If protective mineral coatings with low band gaps (0–1 eV) could be obtained, then the problem of low electrical conductivity and anode passivation would naturally be overcome. In this vein, a series of castable septenary Ni-based superalloys have been developed by VSCA, comprising Ni-Nb–Ta-Fe–Mn-Cr-Y [15]. The masteralloy constituents were melted at 1400 °C in a vacuum induction melter (VIM) using a yttria-coated refractory crucible. Once molten, the alloy was then cast under inert argon gas into a water-cooled copper mould to produce a 2 kg alloy plate 10  100  200 mm in size. The plate was later cut with a diamond saw into anode test samples for cell testing. Each sample was pre-oxidized in air at 1000 °C for 6 h to generate a non-spalling, oxide layer. Initially, the samples were tested ‘statically’ (that is to say, without any electricity) by simply immersing the alloy in a molten cryolite-alumina bath (Na3AlF6-Al2O3) for 6 weeks at 960 °C in an air furnace. Subsequently, electrolytic testing was performed in a lab-based Hall-Héroult cell—with the help of a leading aluminium company. The samples were typically run for 3– 4 days at 960 °C in a cryolite-alumina bath of realistic smelter composition and with different current densities. These cell parameters are deemed to mimic closely the real-world conditions of modern Hall-Héroult aluminium electrolysis cells. The best-performing samples were tested repeatedly, giving a cumulative electrolysis time of over a week. Cell voltage, current, and temperature were monitored and recorded throughout.

Cathode Design and Processing With the aid of in-house thermochemical modelling, a series of TiB2 cathode samples were designed and produced by self-propagating synthesis (SHS) of pressed pellets of different shapes and sizes [16]. The synthesis was based on Ti64, amorphous boron and TiB2 diluent powders, all in the typical size range of 1–50 microns. Ignition of the green pellets was carried out under inert argon gas to avoid any oxidation during high-temperature reaction. Disc-shaped TiB2 samples (7 mm thick, 35 mm diameter) were initially made for ‘static’ testing in molten aluminium, without any electricity. These disc-shaped samples were placed at the

Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis

bottom of an alumina crucible containing molten aluminium at 960 °C and left for 3 months of continuous melt exposure inside an air furnace. Meanwhile, elongated TiB2 samples (10  10  60 mm) were made specifically for a lab-based Hall-Héroult cell with a vertical electrode configuration, again with the help of a leading aluminium company. This electrolytic testing was performed at 960 °C for several days using a cryolite-alumina bath of realistic composition and with realistic current densities, closely matching the environment of a real-world aluminium smelter. Cell voltage, current and temperature were monitored and recorded throughout.

Characterization Techniques Anode and cathode samples were characterized, before and after testing, using a range of advanced techniques. The metallography of samples typically involved diamond-saw cutting, diamond grinding, diamond polishing, and chemical etching using Kalling’s or Weck’s reagents, for anode and cathode etching, respectively. Optical and electron microscopy were carried out using an Olympus GX53 inverted metallurgical microscope and an FEI Nova NanoSEM 450 with EDX spectroscopy, respectively. Hardness testing was performed using an EmcoTest Durascan G5 with a Vickers indenter. Measurements of electrical conductivity at room temperature were conducted using a Foerster Sigmatest 2.070 device and a pure copper reference sample.

Results and Discussion Anode Results Vacuum induction melting and tilt casting were used to melt the masteralloy constituents into a homogeneous, septenary alloy plate (shown in Fig. 2a). The alloy plate had a fully uniform composition and comprised Ni-Nb–Ta-Fe–

641

Mn-Cr-Y [15]. The as-cast microstructure contained at least three distinct crystalline phases, including solid-solution dendrites, intermetallic phases, as well as some interdendritic eutectic (see Fig. 2b). Hardness testing in different locations gave an average of approximately 500 HV with a 5 kg load (see Fig. 2c). The absence of cracks in the corners of the Vickers indents suggests that the alloy has reasonable ductility, in addition to a very high strength of 1500 MPa. The electrical conductivity of the alloy is of the order 2% of the international annealed copper standard (100% IACS), which is typical for Ni-based superalloys like IN625 or IN718. This electrical conductivity is an order of magnitude higher than conventional graphitic carbon which is only 0.2% of IACS. Samples, cut from the alloy plate, were heated up to 900 °C in air using flame torches and rapidly plunged into cold water, as part of a quench test. The samples did not crack or break and, therefore, it is concluded that the thermal shock resistance of the alloy is excellent. Further cut samples were placed in crucibles of molten cryolite-alumina at 960 ° C for 6 weeks in order to test fluoridation resistance. As part of this immersion testing, it was discovered that the samples were inherently resistant to cryolite attack and only a thin outer layer of metal fluorides was present. This ‘static’ testing was of course without electricity and there were no evolving oxygen bubbles, as would be the case in a real anode cell test. In order to explore the oxidation resistance of the alloy, cut samples were then oxidized in an air furnace at 1000 °C for 6 h. The oxidized sample shown in Fig. 3a was then cross-sectioned to visualize the microstructure and measure the oxide layer. As can be seen in Figs. 3a and b, the oxide layer has a dark-grey-greenish hue and is less than 100 microns in thickness. The oxide layer is remarkably adherent to the underlying metal substrate and there was no evidence of spalling or powdering after oxidation, which suggests a favourable Pilling-Bedworth ratio for the protectiveness of the oxide. Due to the multi-component nature of this

Fig. 2 a Anode alloy blocks after VIM-casting; b an optical micrograph of the as-cast alloy microstructure, comprising three crystalline phases, etched with Kalling’s reagent, and c a typical Vickers hardness indent with a 5 kg load

642

septenary alloy, there is a complex layering in the top 20 microns of the surface oxide. The scanning electron microscope (SEM) and energy dispersive X-ray spectroscopy (EDX) images in Figs. 3c and d show that niobium, iron and manganese oxide are the most dominant oxides near the outer surface. Because of the sluggishness of tantalum diffusion, this element seems to be less surface-active, while yttrium distribution was difficult to detect at all using EDS. Following these separate campaigns of (i) oxidation testing and (ii) fluoridation testing, it was then decided to test the samples in a Hall-Héroult cell, whereby the alloy is subjected to high-temperature oxidation and fluoridation at the same time. Cut samples (10  10  100 mm) were first pre-oxidized and then fixed vertically in a cell with a machined thread. The sample was then lowered into the molten cryolite-alumina bath, of typical smelter composition, and tested at  960 °C for 3 days. Some samples were cumulatively run for over a week, with intermittent stops to check the sample’s appearance. Typical electrochemical cell data are plotted in Fig. 4. Generally speaking, the samples survived well and a stable baseline voltage of about 4 V was observed. When non-carbonaceous anodes were used in this manner, the overall cell reaction (2) becomes

Fig. 3 a Anode sample after pre-oxidation; b optical micrograph of the outer oxide layer after pre-oxidation; c SEM image of the outer oxide layer in cross-section, d SEM–EDS layered image showing the elemental distribution in the outer oxide layer

D. J. Jarvis et al.

2Al2 O3ðdissolvedÞ ! 4AlðlÞ þ 3O2ðgÞ ð2Þ An interesting feature of the left-hand graph (Fig. 4a) is the appearance of voltage fluctuations. Upon further inspection of the time-resolved data, it was seen that the voltage increased to the 8 V cutoff value every 45 s, thereafter dropping back to the 4 V baseline. This could be related to the build-up of small oxygen bubbles at the anode surface, which grow and detach themselves in a highly periodic manner. A thin frothy film of evolving oxygen bubbles on a roughened surface is likely to create this kind of periodic resistance. This effect has been noted before in previous inert anode studies [17]. It appears that oxygen is a much more ‘sticky’ gas, compared to the CO2 gas emitted from carbon anodes in regular smelters, and this presents an interesting physical challenge to overcome. As can be seen in Fig. 5a, the 100 mm-long anode sample was immersed in the cryolite bath at a depth of three-quarters its height, as indicated by the black line. The top part of the sample was therefore exposed to hot oxygen gas throughout testing. Despite the high-temperature and oxygen-rich environment, the sample did not spall. Once retracted from the molten bath at the end of a test run, most

Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis

643

Fig. 4 Anode test data showing a voltage and b temperature. The anode gave a typical cell voltage of 3–4 V at temperatures of 960 °C for 3 days. Samples were run numerous times, with week-long cumulative test durations

Fig. 5 The anode sample after several days of cell testing (a), covered in bath crust as well as a Ni-Nb–Ta-rich oxyfluoride layer. The individual elemental distributions at the anode’s outer surface are shown in the EDS maps in (b)

of the cryolite would drip off under gravity, while a small amount would stick to the sample as a solidified crust. The bottom 5 mm of the anode sample was cut and cross-sectioned and further analyzed in the SEM (see Fig. 5b). The sample showed limited corrosion and produced an outer crust of several 100 microns thickness, comprising various semiconducting phases including doped Ni(II)O, doped Ni(III)F3, and various Nb–Ta oxides and oxyfluorides, such as columbite, microlite, and pyrochlore— all known to co-exist with cryolite in Nature [14] and to exhibit a certain degree of multi-valent electronic conductivity [18]. Owing to the heavily-doped multi-valent nature of these oxyfluorides, almost all of the phases in the outer crust were semiconducting. This could also be observed qualitatively from the electronic charging pattern in the scanning electron microscope. As a result, the semiconducting outer crust did not passivate the sample electrically and electrons were able to flow throughout cell testing. Furthermore, this outer crust did not appear to dissolve in the cryolite bath, which, in turn,

offers two major benefits: (i) good protection of the anode alloy underneath the crust and (ii) minimal contamination of the aluminium pad.

Cathode Results The first batch of cathode samples, synthesized by SHS under argon gas, had a microstructure of fine TiB2 particles with some open porosity (20 vol %) that could be easily wetted by liquid aluminium [16]. A test disc (7 mm thick, 35 mm diameter) was submerged in molten aluminium at 960 °C for 3 months of uninterrupted ‘static’ testing (see Fig. 6a) and showed no visible change of shape or size. The sample sank to the bottom of the crucible, due to its higher density than the melt, and it was well wetted by aluminium on every surface. The majority of the open porosity was filled by melt infiltration of aluminium, to produce essentially a ceramic–metal composite or ‘cermet’ (see Fig. 6b). The solidified aluminium melt, visible in Fig. 6a, was

644

D. J. Jarvis et al.

Fig. 6 a Photograph of the cross-sectioned TiB2 cathode sample after 3 months of continuous exposure to liquid aluminium at 960 °C, and b an optical micrograph of the same sample, showing excellent wetting with aluminium, as well as facile melt infiltration into the interior pores

chemically tested for impurities and there was no trace of titanium, vanadium or boron in the melt, even after 3 months of submersion. Remarkably, the TiB2 sample also survived the large thermal shock of being at room temperature and then suddenly submerged in liquid aluminium at 960 °C, within the space of 1–2 s. The high conductivity of TiB2, its low coefficient of thermal expansion and the sponge-like microstructure of the material are clearly beneficial and offer outstanding thermal shock resistance. The hardness of the TiB2 sample was tested after 3 months in aluminium and it was found to have a Vickers hardness of approximately 450 HV (see Fig. 7a), which is similar to hard alloys, like Stellite®. This bodes well for future application in a real smelter, where erosion and wear also become important. The electrical conductivity of the TiB2 sample was demonstrated qualitatively (see Fig. 7b) and measured quantitatively using an eddy current method. The electrical conductivity was found to be 8% of the IACS. Considering that graphitic carbon is only 0.2% of IACS, it is evident that this cathode material is 40 times more conductive than conventional carbon cathodes currently used in smelters. The main reason for the superior electrical conductivity relates to the facile infiltration of liquid aluminium into the

Fig. 7 a Vickers hardness test of the TiB2 cathode sample after 3 months of continuous exposure to liquid aluminium, and b a photograph showing the high electrical conductivity of the TiB2 cathode

residual open porosity, leading to a continuous, highly conductive pathway of aluminium through the TiB2 cathode. Having passed all the ‘static’ test criteria, it was then decided to test the TiB2 samples in a vertical configuration in a lab-based Hall-Héroult electrolytic cell, again at 960 °C and with realistic bath composition and current densities. Figure 8a shows a photograph of the TiB2 sample prior to testing, with dimensions of 10  10  60 mm. After 24 h of testing, the sample was retracted from the molten cryolite, inspected and cut (see Fig. 8b). A small piece of TiB2 was cut from the bottom (see Fig. 8c) and analyzed in an optical microscope (see Fig. 8d). The sample had a continuous film of wetted, frozen-on aluminium (light phase, 100 microns thick) all around the cathode, which again highlights good wettability. The sample retained its square cross-section of 10  10 mm, proving that the cathode’s dimensional stability was good. Typical electrochemical data plots for the cathode testing are shown in Fig. 9. As can be seen, the voltage was low (approximately 2 V) and stable throughout (see Fig. 9a). As with the aforementioned anode testing, some cathode samples were cumulatively run for over a week, with intermittent stops to check the sample’s appearance. It is also worth adding that the SHS manufacturing process for making TiB2 cathodes is extremely versatile and is

Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis

645

Fig. 8 a photograph of the TiB2 cathode sample prior to cell testing, b after 24 h of cell testing and sectioning, c a cross-section piece of the TiB2 sample cut from the bottom, and d an optical micrograph of the TiB2 sample showing wetted aluminium at its surface (light phase)

Fig. 9 Cathode test data showing a voltage and b temperature versus time. The cathode gave a typical cell voltage of 2 V at temperatures of 960 °C for 1 full day of testing. Samples were run numerous times, with week-long cumulative test durations

solid-solution of (Ti,V)B2 which further increases chemical stability of the cermet. In summary, the TiB2 cathode material, developed by VSCA, exhibits a useful balance of properties: excellent wettability with liquid aluminium, good chemical stability, good dimensional stability, good microstructural stability, good thermo-mechanical stability, and superior electrical conductivity. The manufacturing of cathode tiles is considered safe, environmentally-friendly, energy-efficient, and industrially-scalable to large volumes [16].

Outlook Fig. 10 Various shapes can be produced in TiB2 including a tiles, b balls, and c) letters

capable of making a variety of shapes. Trials have successfully produced tiles, spherical balls and even letters (as shown in Fig. 10a–c). TiB2 tiles can also be integrated with a graphite sub-structure for easier installation in a smelter. Another interesting feature of the process is the fact that old ‘out-of-spec’ aerospace Ti64 powders from 3D printers can be recycled and readily used to make TiB2. The small content of vanadium from the Ti64 alloy forms a

Further laboratory testing is underway for both anode and cathode optimization. Larger samples are currently being produced for (i) 10–20 kg anode plates made from VSCA’s proprietary Ni-based superalloys, and (ii) wettable tiles with 5  100  100 mm dimension made from VSCA’s proprietary TiB2 cermets. These materials will then be further trialed in industrial Hall-Héroult cells for longer durations, i.e. several months, as well as in different horizontal, vertical, and drained electrode configurations. A high priority will be to test the anode and cathode materials in unison, in

646

D. J. Jarvis et al.

order to maximize the combined benefits of electrical and CO2 reduction. Scale-up and automation planning for anode and cathode production have already started at VSCA, to ensure that material supply can match customer demand in the coming years. Life-cycle analyses (LCAs) are also being performed to demonstrate and document the sustainability benefits of this anode-cathode technology, and these will be reported on in future papers.

Combining an inert anode with a wettable drained cathode is considered to be the ‘holy grail’ of aluminium electrolysis. This paper shows that green aluminium with near-zero CO2 is within grasp and could be realized with further testing and industrial scale-up in the coming years. The impact that this would have on global CO2 reduction is tremendous and warrants a special international effort to achieve it before the end of this decade—as proposed by the First Movers Coalition on Aluminium [19].

Conclusion

Acknowledgements This work has been privately funded by the inventors, founders and investors of VSCA (David Jarvis, Rosanna van den Blik-Jarvis, Alf Bjørseth, and André Heinz). The authors gratefully acknowledge Dr. Patricia Almeida Carvalho from SINTEF, Oslo, Norway, for her help with SEM-EDX microscopy.

This paper provides evidence, and increasing optimism, that novel high-temperature electrode materials can be developed to withstand the extremely harsh conditions of an aluminium smelter. • A new nickel-based superalloy anode is presented, for the first time, that survives molten cryolite and oxygen gas at 960 °C for over a week of testing and with limited corrosion. The anode does not passivate; conducts electricity well throughout; survives thermal and mechanical shocks; has inherent creep resistance; and can be cast, welded and machined into shapes using regular manufacturing technology. Whether the anode can be called ‘inert’ is subject to debate; the authors prefer the more modest term ‘very slowly consuming anode’ (VSCA), since no materials are truly inert at temperatures of  1000 °C in the presence of oxygen and fluorine. • A wettable cathode based on TiB2 is also presented for the first time. After long-duration testing in molten aluminium at 960 °C, this cathode material has demonstrated good survivability in both static aluminium and cryolitic environments. These TiB2 samples do not show any measurable corrosion; do not degrade or dissolve in liquid aluminium; do not visibly change shape or crack; show excellent wettability with molten aluminium; possess very high electrical conductivity (40  higher than carbon); have good thermal shock resistance and hardness; and can be produced in a cost-affordable manner using environmentally-friendly techniques, like SHS. One important point to emphasize is the fact that all electrode testing has deliberately taken place at 960 °C (i.e. smelter temperatures) using representative bath chemistries and electrochemical conditions. If successful with further testing, it is conceivable that smelter companies could adopt this new materials technology, without having to change too many process features. A retrofittable solution is obviously faster and more preferable than designing and building completely new smelter configurations at huge cost and capital expenditure.

References 1. Hall CM (1886) US. Patent Serial No. 207,601. Filed 9 July 1886 2. Héroult P (1886) French Patent No. 175,711. Filed 23 April 1886 3. International Aluminium Institute (2021) Primary Aluminium Smelting Energy Intensity. https://international-aluminium.org/ statistics/primary-aluminium-smelting-energy-intensity/ 4. Saevarsdottir G, Padamata SK, Velasquez BN, Kvande H (2023) The way towards zero carbon emissions in aluminum electrolysis. Light Metals 2023:637–645. Springer, Cham. https://doi.org/ 10.1007/978-3-031-22532-1_86 5. He Y, Zhou K, Zhang Y, Xiong H, Zhang L (2021) Recent progress of inert anodes for carbon-free aluminium electrolysis: a review and outlook. J. Mater. Chem. A 9:25272–25285. https:// doi.org/10.1039/d1ta07198j 6. Padamata SK, Yasinskiy AS, Polyakov PV (2018) Progress of inert anodes in aluminium industry: review. Journal of Siberian Federal University Chemistry 11(1):18-30. https://doi.org/10. 17516/1998-2836-0055 7. Sadoway DR (2001) Inert anodes for the Hall-Héroult cell: the ultimate materials challenge. JOM 53: 34-35. https://doi.org/10. 1007/s11837-001-0206-5 8. Pawlek RP (2014) Inert anodes: an update. Light Metals 2014:1309–1313, Springer, Cham. https://doi.org/10.1007/978-3319-48144-9_219 9. Chapman V, Welch BJ, Skyllas-Kazacos M (2011) Anodic behaviour of oxidised Ni-Fe alloys in cryolite alumina melts. Electrochimica Acta 56(3):1227-1238. https://doi.org/10.1016/j. electacta.2010.10.095 10. Ransley CE (1954) British Patent No. GB802905A. Filed 14th January 1954 11. Ransley CE (1962) Refractory carbides and borides for aluminum reduction cells. JOM 14:129-135. https://doi.org/10.1007/ BF03378134 12. Li J, Lü X, Lai Y, Li Q, Liu Y (2008) Research progress in TiB2 wettable cathode for aluminum reduction. JOM 60:32-37. https:// doi.org/10.1007/s11837-008-0104-1 13. Padamata SK, Singh K, Haarberg GM, Saevarsdottir G (2022) Wettable TiB2 cathode for aluminum electrolysis: a review. J. Sustain. Metall. 8:613-624. https://doi.org/10.1007/s40831022-00526-8 14. Bastos Neto AC, Pereira VP, Ronchi LH, de Lima EF, Frantz JC (2009) The world class Sn, Nb, Ta, F (Y, REE, Li) deposit and the massive cryolite associated with the albite-enriched facies of the

Novel Developments for Inert Anodes and Wettable Cathodes in Aluminium Electrolysis Madeira A-type granite, Pitinga Mining District, Amazonas State, Brazil. The Canadian Mineralogist 47:1329-1357. https://doi.org/ 10.3749/canmin.47.6.1329 15. Jarvis DJ, Jarvis RE (2019) EPO Patent No. EP3839084. Filed 20th December 2019 16. Jarvis DJ, Jarvis RE (2023) EPO Patent No. EP23153611. Filed 27th January 2023

647

17. Cassayre L, Utigard TA, Bouvet S (2002) Visualizing gas evolution on graphite and oxygen-evolving anodes. JOM 54:41-45. https://doi.org/10.1007/BF02701696 18. Verwey EJW (1939) Electronic conduction of magnetite (Fe3O4) and its transition point at low temperatures. Nature 144:327-328. https://doi.org/10.1038/144327b0 19. First Movers Coalition on Aluminium (2023) In: www.weforum. org/first-movers-coalition/sectors

Online Monitoring of Metal Oxides in Molten Fluoride Electrolytes Wojciech Gebarowski, Samuel Senanu, Arne Petter Ratvik, Ole Kjos, Henrik Gudbrandsen, and Egil Skybakmoen

Abstract

Reliable monitoring of oxide concentrations in electrowinning of metals in fluoride-based electrolytes is very important for avoiding unwanted reactions, notably perfluorocarbons emissions which are very potent greenhouse gases. A probe, consisting of a working electrode and a counter electrode, has been developed for online monitoring of metal oxide concentrations in molten fluoride electrolytes. During measurements, the immersed probe in the electrolyte is subject to fast anodic polarisation sweeps while the electric current response is recorded. During anodic polarisation, oxide species in the electrolyte are rapidly depleted at the probe, resulting in a response called anode effect on the probe tip, triggering a rapid voltage increase and current drop. The rate of oxide depletion and the consequent anode effect is proportional to the dissolved oxide concentration. This allows to correlate voltage-current–time curves to the oxide concentration. Measurements in cryolite melts with different oxide concentrations showed different voltages at which anode effect occurs. Keywords

Perfluorocarbons Dissolution




Molten fluoride electrolyte

of greenhouse gasses to the atmosphere. Greenhouse gas emissions from the production of aluminium arise from the various reactions occurring. The production of primary aluminium is based on molten salt electrolysis using consumable carbon anodes. These consumable carbon anodes partake in the aluminium producing reactions to produce CO2 as shown in Eq. 1: 1 3 3 Al2 O3 ðdissÞ þ C ¼ AlðlÞ þ CO2 2 4 4

Equation 1 suggests that 1.22 tonnes of CO2 is produced for every tonne of aluminium produced, however, the actual value is higher due to factors such as air burn, loss in current efficiency, Boudouard reaction, dusting, etc. The average industrial value is ca. 1.5 tonne CO2 per tonne Al [1, 2]. In addition, the production of primary aluminium generates a significant amount of perfluorocarbon (PFC) emissions, mainly CF4 and C2F6, as illustrated by Eqs. 2 and 3. These gases are very potent greenhouse gases with large detrimental effect on our atmosphere. The average PFC emissions for the aluminium industry in 2022 is estimated to be 0.7 t CO2 eq./t Al [1].




Introduction The global annual production of aluminium in 2022 was 68.5 million tonnes [1]. Production of aluminium is an energy-intensive process responsible for significant emissions W. Gebarowski
S. Senanu (&)
A. P. Ratvik
O. Kjos
H. Gudbrandsen
E. Skybakmoen SINTEF Industry, P.O. Box 4760 Torgarden, NO-7465 Trondheim, Norway e-mail: [email protected]

ð1Þ

3 3 AlF 3 ðdissÞ þ CðsÞ ¼ CF 4 ðgÞ þ AlðlÞ 4 4

ð2Þ

1 AlF 3 ðdissÞ þ C ðsÞ ¼ C 2 F 6 ðgÞ þ AlðlÞ 2

ð3Þ

The emission of PFCs is mainly related to problems with controlling the alumina concentration in the electrolyte, leading to a phenomenon called anode effect [3]. An anode effect is an unwanted condition at the electrode induced by insufficient amount of dissolved alumina, in the electrolyte. It causes an increase of cell voltage, inactivate some areas of the electrodes, and triggers undesirable anodic reactions as given by Eqs. 2 and 3 resulting in the emission of the PFC

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_83

648

Online Monitoring of Metal Oxides in Molten Fluoride Electrolytes

gases. Since the anode effects are related to process control issues, i.e. alumina feeding and distribution, they can in theory be avoided entirely with proper feed control, thereby eliminating PFC emissions from aluminium production. Dissolution of alumina in the molten cryolite-based electrolyte is one of the most complex aspects of the electrolysis process as many parameters are involved. Firstly, during electrolysis, oxide concentration in the bath is usually kept as low as possible, just to provide an appropriate concentration to maintain a stable process [2, 3]. This ensures a high dissolution rate for each addition of alumina, usually around 1–2 kg alumina batches [2, 3], minimising sludge formation. On the other hand, it creates a delicate balance between appropriate process conditions and the unwanted anode effects. Operating at the low oxide concentrations causes several challenges, especially the oxide distribution within the cell since feeding is limited to a few point feeders. This issue becomes even more challenging for larger cells with larger anodes as these are more prone to so-called partial anode effects [4]. Oxide distribution is directly connected to electrolyte stirring conditions (magnetohydrodynamic (MHD) effects) and feeding arrangement in the cell [3]. Also, the microstructure, particle size distribution, impurity content, etc., have a large impact on dissolution kinetics. Moreover, since quality of raw materials is rarely fixed in the long term, the industry needs to adapt to alumina quality variations and optimise feeding strategy accordingly. Therefore, knowledge about dissolution kinetics of oxide materials is important for making good predictions of the performance of the electrolysis process, both to control and to tune the production appropriately. Several researchers have over the years tried different analytical techniques for continuous monitoring of the oxide concentration [5–10]. The most employed techniques by researchers are based on electrochemical response measurements. The measurements involve applying a rapid voltage sweep to an electrode assembly immersed in a fluoride electrolyte by means of a voltage-controlled source of electric power. By monitoring the current passing through the electrode assembly during each sweep, the voltage and current at which anode effect is induced is recorded. The voltage at which the anode effect occurs is observed to vary with the concentration of oxides in the electrolyte. Haverkamp et. al. [7], observed that the potential at which anode effect occurs increases with increasing oxide concentration. It is thus possible to correlate the oxide concentration with the applied potential at anode effect within a defined range of parameters. This makes the voltammetry technique suitable for determining oxide concentration in fluoride melts when appropriate calibrations are done. The present paper describes the development of a sensor for in situ monitoring of oxide concentration in molten fluoride salts. Besides analytical methods for oxide

649

concentration analysis such as inert gas fusion technique and X-ray diffraction (XRD), that requires sampling and sample preparations, there is no existing, commercially available method for in situ measurements of the oxide concentration in molten fluoride salts. The presented sensor is based on the theory described in the patent by Haverkamp et al. [7].

Experimental The presented results were carried out using setup shown in Fig. 1. Electrolyte with volume approx. 450 cm3 was melted in graphite crucible placed in vertical tube furnace with resistive heating element. Three different electrolyte types were used: conventional industrial cryolite bath (NaF-AlF3CaF2-Al2O3), low-temperature cryolite bath (NaF-AlF3KF-Al2O3), and rare earth metal fluoride (NdF3-PrF3LiF-Nd2O3-Pr2O3). The first bath type was prepared from frozen industrial electrolyte with CR = NaF/AlF3 = 2.1, 5 wt% CaF2, and ca. 1 wt% Al2O3. The low-temperature cryolite melt had a CR = (NaF + KF)/AlF3 = 1.3, KR (potassium fluoride ratio) = KF/(NaF + KF) = 0.8 and prepared using NaF (Merck, 99.9 wt%), KF (Merck, 99.9 wt%), and AlF3 (in-house sublimated, > 99%). Rare earth metal electrolyte was prepared from 88.58 wt% of NdF3 (in-house, 98%) and 11.42 wt% LiF (Sigma-Aldrich, 99.9%). Temperature of bath was 980, 808, and 1048 °C for conventional, low-temperature, and rare earth metals bath, respectively, and it was monitored during experiment using shielded thermocouple immersed in the bath. Aluminium oxide (Merck, 99.9 wt%) was added to the conventional and low-temperature bath using a stainless-steel feeding tube in batches corresponding to 0.5 or 2.5 wt% of the total bath mass. For rare earth metal bath, a mixture of 77.5 wt% of Nd2O3 (Alfa Aesar, 99.9%) and 22.5 wt% of Pr2O3 (Alfa Aesar, 99.9%) was used. During the test, the bath was agitated using a boron nitride stirring element. The default stirring speed was 120 rpm, if not specified differently. The measuring probe was immersed in the bath and kept at the same position during the whole test. The immersed part of probe is depicted in Fig. 2 b and consists of two electrodes: a working electrode in the form of a small area conical tip and a larger area counter electrode. The probe contains a built-in thermocouple to monitor probe temperature, this could be useful in short-term field measurements as it will help know if the probe has been in the melt long enough for representative data. The probe was connected to the in-house built measuring unit (shown in Fig. 1b), responsible for generating a driving polarisation signal, and for recording the probe voltage (potential difference between probe electrodes) and probe current. The measuring unit can generate fast polarisation signals for the probe with a sweep rate up to 50 kV/s and maximum current

650

W. Gebarowski et al.

Fig. 1 Setup for monitoring oxide concentration. a Experimental setup; (1) thermocouple; (2) oxide feeding tube; (3) graphite crucible; (4) electrolyte (molten salt); (5) stirrer; (6) measuring probe. b In-house measuring unit for generating and recording electrochemical responses

and voltage up to 20 A and 15 V correspondingly. Polarisation signal can be voltage- or current-controlled and can have user-programmed shape: linear sweeps or square wave pulses. In this work, the focus is on the use of linear voltage sweeps which is the signal type originally proposed by Haverkamp et al. [7], however, utilizing other types of signal can enhance the probe response performance. The sample rate of signal recording was 100 kHz for all presented results. The relationship between the probe response signal and the oxide concentration in the bath is a non-linear curve which must be experimentally calibrated. A polynomial fitting is utilised to calculate a transfer function from the non-linear curve to translate the probe response data to oxide concentration values. The probe response and the corresponding transfer function are not universal and must thus be determined for different electrolyte systems. To approximate the oxide concentration using the probe response, ex-situ oxide concentration analysis of bath samples collected during the experiments was performed by analysing the oxygen

Fig. 2 a Polarisation curves of probe at different Al2O3 concentrations in an electrolyte using voltage sweep rate 5000 V/s. b Schematic of probe.

content using the inert gas fusion technique with the LECO TC-436 DR (Leco Corp., USA) instrument.

Results and Discussion Figure 2a shows typical polarisation curves of the probe produced by applying fast voltage sweeps in the anodic direction for fluoride electrolytes containing 2 and 7 wt% of Al2O3. In the first stage of both cases, the current increases quite linearly with applied voltage until a rapid drop is observed which manifests anode effect occurrence at the probe tip. During anodic polarisation, oxygen from oxyfluoride ions (formed from alumina and the fluoride species) is removed forming CO/CO2 gas at the anode as shown by the schematic reaction given by Eq. 4: 2 4 ðAl2 O2 F 2 4 ; Al2 OF 6 ; Al2 OF 8 Þ þ C   ! AlF 4 þ ðCO; CO2 Þ þ e

ð4Þ

Online Monitoring of Metal Oxides in Molten Fluoride Electrolytes

Oxyfluoride species can form different complexes with different Al, O, and F ratios depending on oxide concentrations and acidity of electrolyte. There are more possible oxyfluoride complexes that exist in electrolyte, however, these in Eq. 4 are generally accepted in literature as the most stable and common ones [11–13]. Once all the oxyfluoride ions in the probe vicinity are depleted, a passive layer resulting from the decomposition of the fluoride species in the probe vicinity to form PFCs as shown by Eqs. 2 and 3 above occurs. The passive layer causes a passivation of the probe tip which results in blockage or hindrance of current flow described as anode effect [3]. There is a substantial difference between polarisation curves recorded at different alumina content as presented in Fig. 2a. In the case of the electrolyte containing 2 wt% of alumina, the probe passivation occurred at 2.9 A and 3.9 V, while, for 7 wt%, these values were much larger: 15.4 A and 10.4 V, respectively. They are called the critical current density and the critical voltage. Intuitively, the lower concentration of dissolved oxide in electrolyte, the quicker the oxide depletion occurs during voltage sweeps. This large difference in polarisation curves of the probe depending on alumina concentration allows to correlate probe response with alumina concentration and it is the basis of the described method. It is worth noting that the values of critical voltage and current in Fig. 2 are not specific for any electrolyte concentration. Besides the strong dependence on oxide concentration, their response also depends on the electrical resistance between the electrodes in the probe. In addition, since depletion and passivation processes are dependent on ion transport which in turn is a time-dependent process, the sweep rate of polarisation hugely affects the shape of curve. The effect of resistance can be minimised by using a fixed electrode arrangement in the probe. Moreover, by using the electric charge passed through the probe until passivation is reached instead of the critical voltage or the critical current

651

will eliminate the effect of interpolar resistance. Since transport of oxide species in the vicinity of the probe tip is mainly driven by diffusion, the large dependence of charge required to induce anode effect and polarisation sweep rate is expected for different oxide contents. The graph in Fig. 3a presents the current response of the probe during linear polarisation at different sweep rates, from 100 to 10,000 V/s. Shaded area under the curves indicates the charge which was passed through the probe until the passivation occurred. These plots clearly illustrate how significantly the charge-to-passivation value changes with voltage sweep rate applied to the probe. The faster sweep rate results in a decrease in charge required to reach passivation of the probe. For the sweep rate of 100 V/s, the probe was passivated after passing 46.6 mC, while for 10,000 V/s after only 3.1 mC. Time to passivation was also decreased substantially, from 37 ms to almost 1 ms. Smaller charge required to passivate the probe is beneficial due to reduced time for single measurement. In addition, it translates into less tip consumption and longer lifetime of the probe. Stirring of the bath has a strong influence on passivation of the probe, however, this effect is only visible at low alumina concentrations. The graph in Fig. 4 shows the effect of electrolyte stirring on charge-to-passivation of the probe. At 2 wt% of alumina in the electrolyte, faster stirring significantly increases the value of charge making it more difficult to trigger passivation of the probe. Interestingly, at high oxide concentration, this effect is minimal suggesting that at such conditions the diffusion is more important than the convection for ion transport. This observation may indicate a great impact of even moderate stirring on decreasing anode effects occurrence since industrial process operates at low alumina concentrations. The presented probe has been tested in different molten salt systems including rare earth fluoride melts (e.g.,

Fig. 3 a Current response curves of probe polarised at different voltage sweep rates. Numbers next to each curve show charge (integrated shaded area under curve). b Dependence of polarisation sweep rate on charge-to-passivation.

652

W. Gebarowski et al.

Fig. 4 Effect of sweep rate, oxide concentration and stirring speed on mean charge-to-passivation of the probe tip

neodymium, praseodymium, dysprosium fluoride melts) and their corresponding oxides giving satisfactory results. The probe response varies for different electrolyte systems due to unique electrolyte properties such as diffusion coefficients of ion species, electrical conductivity, electrochemical reaction rates, and interfacial conditions (wetting of graphite electrode by electrolyte and gaseous products). Nevertheless, the character of response is always similar which allows to accommodate the probe for use in many different conditions. Additionally, optimisation of the polarisation signal can significantly improve the response signal resolution and stability. Graphs in Fig. 5 show the probe and temperature response during two additions of alumina to the low melting point electrolyte. Each addition contained 0.5 wt% of the total electrolyte mass. Non-filled circle points show probe data rescaled to Al2O3 concentration values based on a calibration curve. Solid square points show results of ex situ oxygen analysis (recalculated to Al2O3 content) based on bath samples taken 5 min before and 5 min after each oxide addition. There is a rapid response of the probe after alumina addition: fast increase followed by plateau region indicating very fast kinetic of alumina dissolution under presented conditions. Most of the added alumina portion was completely dissolved within approx. 2 min after addition. Alumina additions caused temperature decrease by approx. 3 °C which is insignificant drop with minimal impact on the dissolution rate during test. The graphs presented in Fig. 6 show the probe measurement and temperature change during additions of Nd2O3-Pr6O11 to the NdF3-LiF electrolyte. After each oxide addition, there is a rapid increase in probe response

Fig. 5 Probe and temperature response during additions of alumina in low-temperature fluoride melt (NaF-AlF3-KF)

indicating an initial fast dissolution rate for approximately 4 min followed by a slower increase in oxide concentration and finally a plateau which indicates complete dissolution of the added oxide. Similar to the previous example the temperature drops by approx. 4 °C after oxide additions which is not expected to affect dissolution rate significantly. The third example shows kinetic of dissolution of alumina in industrial grade electrolyte (Fig. 7). Unlike the two previous examples, oxide addition was much larger, ca. 2.5 wt% of the total electrolyte mass was added in one batch. In this case, complete dissolution of the whole oxide addition took approximately 35 min which is significantly longer than in the previous cases. The longer period for the dissolution is probably due to a low dissolution rate resulting from the large bath temperature drop after the additions (ca. 8 °C drop in bath temperature). Also, the slow dissolution rate leads to sludge formation which sediments at the bottom of crucible. The stirring speed at 120 rpm used during the experiment was not sufficient to prevent the sludge formation. Alumina trapped in the sludge is more compacted hence transport of electrolyte to and from the alumina grains is substantially moderated. Such situation is very common during real operation of industrial cell where alumina is trapped in sludge or rafts caused by overfeeding or not optimal alumina

Online Monitoring of Metal Oxides in Molten Fluoride Electrolytes

Fig. 6 Probe and temperature response during additions of an oxide mixture consisting of 77.5 wt% Nd2O3 and 22.5 wt% Pr2O3 in a neodymium and lithium fluoride melt (88.58 wt% NdF3 and 11.42 wt% LiF)

distribution in cell. The presented example shows the importance of optimal oxide feeding to prevent the sludge formation for stable industrial cell operation.

Conclusion and Outlook The development of a probe for in situ oxide concentration measurements in molten salts has been developed to give highly reliable results. The performance of the probe on laboratory scale has been verified based on three different electrolyte systems. The results show good probe response, signal-to-noise ratio and stability over time, revealing its potential as a useful tool for investigation of dissolution kinetics of oxides in various molten salt electrolytes. The technique and probe presented in this paper can be utilised for comparing the dissolution rates or kinetics of different oxide raw materials in different electrolyte systems. It can also be used for online monitoring of the oxide concentration during the electrolysis process. The described measuring system was designed for laboratory use but can be adapted to perform measurements in industrial environment with minor changes, for instance, to investigate distribution of oxide concentration in large cells.

653

Fig. 7 Probe and temperature response during one addition of alumina in conventional industrial aluminium fluoride melt

This can be a valuable tool to design better cells with optimal feeding points, hence, improving process operation in addition to reducing PFC emissions. Furthermore, industrial campaigns have been planned to test the probe in industrial settings. This will facilitate adaption of the probe for industrial applications. Acknowledgements The authors gratefully acknowledge financial support from the European Union Horizon-CL5-2021-D2-01 Programme under Grant Agreement No. 101069492, the Swiss State Secretariate for Education, Research, and Innovation (SERI) under contract number 22.00043, the European Union's Horizon 2020 and Innovation Programme under Grant Agreement No. 776559 and SINTEF Internal funding.

References 1. International Aluminium Institute (2023) Primary Aluminium Production. https://international-aluminium.org/statistics/primaryaluminium-production/. Accessed 2 August 2023 2. Grjotheim K, Kvande H (1993) Introduction to Aluminium Electrolysis: Understanding the Hall-Heroult Process, 2nd ed. Aluminium-Verlag, Dusseldorf 3. Thonstad J, Fellner P, Haarberg GM, Hives J, Kvande H, Sterten A (2001) Aluminium Electrolysis: Fundamentals of the Hall-Héroult Process, 3rd ed. Aluminium-Verlag, Düsseldorf.

654 4. Åsheim H (2017) PFC Evolution in the Aluminium Production Process. Ph.D. thesis, Norwegian University of Science and Technology, Trondheim 5. Marinha D, et.al. (2023) Following Alumina Dissolution Kinetics with Electrochemical and Video Analysis Tools. Light Metals 2023: 77–86. https://doi.org/10.1007/978-3-031-22532-1_10 6. Nikolaev A, et. al. (2020) Electrochemical Sensor for Monitoring the Alumina Dissolution and Concentration in a Cryolite-Alumina Melt. J. Electrochem. Soc 167(12): 126511. https://doi.org/10. 1149/1945-7111/abb176 7. Haverkamp RG, Welch BJ (2000) Measurement of Alumina in Reduction Pots, in: U. Patent (Ed.) 6010611, Auckland Uniservices Ltd, New Zealand 8. Crottaz O, et. al. (2001) Development of Techniques for Measuring the Composition of Low Temperature Electrolytes. Light Metals 2001: 1195–1199

W. Gebarowski et al. 9. Haverkamp RG, et. al. (2001) Voltammetry and electrode reactions in AlF3-rich electrolyte. Light Metals 2001: 481-486. 10. Haverkamp RG, Welch BJ, McMullen A (2001) Real Time Alumina Measurement in Industrial Cells. Light Metals 2001: 1193–1194 11. Kvande H, Rorvik H (1986) Light Metals 1986: 451 12. Picard GS, et. al. (1996) Structures of oxyfluoroaluminates in molten cryolite-alumina mixtures investigated by DFT-based calculations. Journal of Molecular Structure-theochem 368: 67–80 13. Robert E, et. al. (1997) Structure and Thermodynamics of Alkali Fluoride−Aluminum Fluoride−Alumina Melts. Vapor Pressure, Solubility, and Raman Spectroscopic Studies. Journal of Physical Chemistry B 101. https://doi.org/10.1021/jp9634520

Smelting 4.0: Digital Strategy for Aluminum Production Ved Prakash Rai and Datta Raju D

Abstract

Hall-Héroult (in smelting plant) has been the key process for aluminum production since 1886, and it has gone through several improvements in the past. However, challenges like low efficiency, premature cell failure, and sustainability-related issues persist even today and the industry is working to overcome them. Industries have significantly benefited by adopting the fourth industrial revolution (I4.0) concepts such as digital twin, remote monitoring, predictive maintenance, data & analytics, etc. They have helped industries to achieve their objectives. The same concepts will also help aluminum industry to address their challenges. This paper describes specific challenges related to aluminum production and defines a roadmap for adopting I4.0 concepts to help solve problems. This paper introduces a structured approach for the digital transformation of complex smelting plants and recommends a roadmap for this industry. Keywords

Aluminum




Industry 4.0




Digital transformation

Introduction

siloed data, and lack of real-time visibility into operations has led industries to find a new way to improve business KPIs (key performance indicators). Figure 1 explains the value chain of aluminum production more focusing on smelting process [1]. Charles Martin Hall and Paul Héroult invented the first commercially feasible method to extract aluminum from alumina in 1886, now the process is called as the Hall-Héroult smelting process [2]. In the Hall-Héroult process, alumina dissolves in a molten cryolite bath which lowers the electrolysis temperature. Electrolysis is performed at 950 0C [3] which is much lower than the melting temperature of alumina. During this electrolysis, at the anode, carbon combines with oxygen to form carbon dioxide, while, at the cathode, liquid aluminum deposits which is tapped at regular intervals. In the Hall-Héroult process, for operations and process control the major technical KPIs are average amperage, specific energy consumption, current efficiency, hot metal production per day, net carbon consumption, AlF3 consumption, number of pot x days in operation (operating rate), age of pot stopped, etc. Similarly, for maintenance, the major KPIs are equipment availability percentage, maintenance costs, mean time to repair (MTTR), mean time between failures (MTBF), etc. For warehouses, the KPIs are inventory accuracy, stockout rate, carrying cost of inventory, etc. (see Fig. 1).

Aluminum Production Process Aluminum production process majorly consists of five steps, namely, bauxite mining, refining by Bayer process, aluminum smelting by Hall-Héroult process, casting, and at the end fabricating according to requirements of the market. Challenges across the value chain like manual processes, V. P. Rai (&)
D. R. D IoT DE Advisory, Tata Consultancy Services, Bangalore, 560100, India e-mail: [email protected]

Industrial Revolutions An industrial revolution is a wave of major technological innovations linked to each other and together bringing about a fundamental change in human society [4]. The intense use of new evolving technologies has now marked the beginning of the fourth industrial revolution. The fourth revolution is preceded by the other three (Fig. 2), the first Industrial Revolution (IR) started during

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_84

655

656

V. P. Rai and D. R. D

Fig. 1 Value chain of aluminum production more focusing on smelting process

1784. It introduced coal as a major source of energy, boosted the use of the steam engine, and helped in transforming heavy manual work into machine-based activities [5]. After the IR 1.0, the second industrial revolution started during 1870 and improved the use of electricity. Mass production of goods started by deploying production lines. It helped in increasing the production rate [5, 6].

The 3rd IR started during 1969 and is known for automated production, by using electronics and information technology organizations were able to automate their production process. All three industrial revolutions that humankind has experienced until now have helped to improve working processes in almost every industry. These technologies have changed human behavior and working style [5].

Smelting 4.0: Digital Strategy for Aluminum Production

657

Fig. 2 Depicts industrial revolutions with timeline

Now the world has entered the era of the 4th industrial revolution. Industry analysts are predicting that this will bring the changes that the industry has never experienced in past decades. Industry 4.0 is characterized by several complementing technologies. It has nine major pillars that are cloud computing, autonomous robots, simulation, horizontal and vertical system integration, industrial Internet of things, cyber security, additive manufacturing, augmented reality, big data, and analytics [5].

Data Management Organizations rely on logbooks to keep a record of shift data which are generated during operational and qualitative parameter measurements like temperature, voltage, bath level, etc., for process control. Data from the lab is manually entered in Excel and then it is analyzed. Data generated by measuring equipment are recorded manually on paper and then it is communicated to the different concerned departments for analysis.

Challenges of Aluminum Smelting Industry

Remote Monitoring of Pots Visibility of cells at remote locations is not available. Performance monitoring of critical parameter is difficult.

Aluminum smelting has several challenges for the last several decades. Challenges faced by smelting process are briefly explained below.

Smelting Operations High Specific Energy Consumption High specific-energy consumption for production of aluminum (*35% of total production cost) and associated effect on sustainability [8, 9]. To produce one tonne of aluminum the average process energy intensity stood at 13,600 kWh [7, 9]. It is estimated that 3.5% of the world's total energy goes to aluminum smelters [10]. Premature Cell/Pot Failure A significant portion of ongoing costs in aluminum smelters includes electrolytic cells reconstruction. The cells undergo different wear mechanisms and failure modes determining their finite lifespan. Smelters must carefully manage all aspects determining cell life to maximize it while minimizing reconstruction costs. Pot design, materials and supplier selection, construction quality, and electrolysis operations all affect the cell life. Replacement of one cell cost around $ 21,179 [11].

Smelting Process Control Batching of Pots for Tapping Quality of the produced metal is deteriorated by the impurities like iron, silicon, manganese, etc. During tapping, pots with high purities should not be tapped with pots of high impurities. For a Potroom manager, it is a daunting task to do batching for metal tapping considering the impurities of the cells as well as operational constraints associated with crane movement. Identification of Sources of Impurities Silicon, iron, vanadium, manganese, etc., are the typical impurities in produced metal. Different raw materials are used in the Hall-Héroult cell and impurities increase during the electrolysis process. The sources of impurities can be anything like carbon anode, soda ash, alumina, etc. Once metal purity deteriorates, it is important to know the exact sources of impurities to improve the quality of metal produced. Managers need to collect data from the different sources and they need to do mass balancing each time manually. Delays in decision-making affect the quality of produced metal.

658

Calculation for Charging Raw Material Calculation of AlF3 and soda ash are taken based on only some parameters and quantity are either decided on the experience of an operator or on some Excel-based calculation. Monitoring of Process Parameters The absence of customized dashboard makes monitoring of the process difficult. Multiple Excel-based reports are generated by operations, quality, maintenance, etc. Costly and Time-Consuming Experiments Experiment for process optimization and parameter settings is a challenge because of its cost and required time to conclude the experimental setting.

Asset Life Cycle Management Maintenance Process An unplanned breakdown of equipment causes production loss as well as increase in maintenance costs. Maintenance activities are performed after an asset having a breakdown. This causes production losses and overtime for employees. Limited visibility of maintenance, repair, and operations equipment. It increases the risk of excess inventory or stockout.

Warehouse Inventory Visibility Lack of real-time visibility of raw material inventory across different warehouses. Data related to inventory are scattered across siloed systems. Forecasting Accuracy The problem of overstocking and stocking out of raw materials such as aluminum fluoride, soda ash, and alumina because of inaccurate forecasting caused by frequent parameter changes, e.g., current fluctuation, blackout [12], high sodium in alumina, etc.

Environment, Health, and Safety Environment Monitoring Globally, the aluminum industry emits approximately 1.1 billion tonnes of CO2-equivalents annually (approximately 2% of the world’s total emissions) [13]. These numbers are growing because an increasing share of aluminum production is derived from electricity from fossil fuels. Global warming has affected the entire world. This has resulted in

V. P. Rai and D. R. D

more frequent and intense extreme weather events that have caused increasingly dangerous impacts on nature and people worldwide. Since the aluminum industry is considered one of the major greenhouse gas emitters, the pressure on the industry is mounting even more heavily than in previous decades to take measures to reduce their greenhouse gas emissions. CO2 emissions represent the single biggest sustainability challenge, with significant contributions of CF4 and C2F6 to the atmosphere.

Personal Protective Equipment (PPE) Adherence Aluminum industry is considered as one of the most hazardous industries in the world. According to data by the National Library of Medicine, in theUSA in aluminum smelting industry average duration of man-hours between accidents was 32,516 [14]. Accident in Potroom can be controlled by monitoring personal protective equipment (PPE) by engineering controls, administrative or right work practices.

Workforce Management Workforce Planning Manually Aluminum smelting is a labor-intensive industry and to produce metal thousands of workforces are involved on the shop floor. Effective distribution of work is quite important to ensure high productivity of the employees. This planning is done manually either on paper or on Excel-based system. Managers need to manually track issues related to attendance, absences, holiday, overtime, shift change, and non-compliance, workforce distribution for tapping, anode setting, bath temperature measurement, voltage checking, etc. Training newly hired employees is challenging because of the complexity and danger associated with the system.

Information Technology Inadequate IT-OT Convergence The OT (operation technology) devices which are present in most of the aluminum plants are programmable logic controllers (PLCs), supervisory control and data acquisition systems (SCADA), cell control systems (CCS), etc. Companies are unable to leverage data that are being generated by several OT devices. The lack of IT-OT integration is a challenge for the industry. A lack of convergence between IT (information technology) and OT had led to the slow and inadequate use of the Data. Cyber Security Aluminum smelters use several OT devices, for example, cell controller which controls the cells, it helps in taking

Smelting 4.0: Digital Strategy for Aluminum Production

decisions regarding feeding of alumina, terminations of anode effect, and anode movement. Similarly to the rectifier substation, cell process control depends on their SCADA (supervisory control and data acquisition) system. Many of the OT devices are connected to the internet. Not only the operations but also several administrative activities depend on software. Vulnerability and risk to the IT-OT system may create a severe risk to the entire plant. In the past, a major aluminum producer had a bitter experience of cyber-attack [15].

Discussion To improve the smelting process and its key KPIs (key performance indicator) through digitalization (Industry 4.0), entire ecosystems need to be connected digitally. Typically, smelting plants have asset like pot tending machines, electrolytic cells, forklift, etc. They need to be connected to manufacturing IT systems and connected among each other. Also, the plant operations should be connected to asset and ecosystem. Below Fig. 3 explains the point of view of a connected smelting plant operates. Smart smelters will be built on a network of self-aware and intelligent systems powered by digital capabilities such as smart machines, integrated operations, intelligent workforce, smart warehouse, and control tower. This is at a lower level, supported by the functional use cases. Horizontal and vertical integration will provide alignment across the entire value chain, from the shop floor to enterprise level. Engineering integration will integrate the process from initiation to the final implementation stage. Advanced analytics will make operations more predictable using data, which will help in decision-making. The industrial cloud will accelerate digital transformation through on-demand access to

Fig. 3 Connected view of a potroom

659

computing resources, such as data storage, applications, and networking. Robust cyber security will protect servers, computers, applications, and data from malicious attacks (Fig. 3).

Typically, a Connected Smelter Will Have Capabilities Like Smart Machines, Integrated Operations, Intelligent Workforce, Smart Warehouse, and Smelting Control Tower Smart Machines Smart machines solution enables real-time tracking, improvement of asset performance by IoT (internet of things) enabling system. Smart, smart machines will help to get more efficient use of equipment. It will Improve asset reliability. For example, real-time data accumulation from pots, cranes, dry scrubber system, etc., to get real-time status and historical insights of the asset through solutions like digital twin for smelting units, remote monitoring of pots, condition-based maintenance, predictive maintenance, and prescriptive maintenance, etc. It would help predict failures and improve efficiency. KPIs impact: Overall labor effectiveness (OLE), equipment availability (EA), maintenance costs, etc. Integrated Operations Integrated operations enables a system to get real-time information from its operational activities so that they can share data seamlessly. As the systems will relate to each other more data will be generated in both structured and unstructured forms and the advanced analytics will help organizations to get several previously undiscovered results. Over a time, system will move from just real-time to

660

predictive operation and eventually it will achieve an autonomous operating environment where it will take decision without much human intervention which will be directly aligned with business goals. For example, real-time data visibility across production, maintenance, quality control, warehouse, etc., through digital solutions like AI (artificial intelligence)-based voltage optimization, digital CAPA (corrective and preventive action) and RCA (root cause analysis), personal-based dashboard, etc. KPIs impact: Specific energy consumption, current efficiency, hot metal production per day, net carbon consumption, operating rate, age of pots stopped, AlF3 consumption, etc.

Intelligent Workforce Intelligent workforce system enables integration of workforce with the secured network in the working environment. An integrated system will help workforce to interact seamlessly, improve health, safety, productivity, training, etc., with the help of immersive and wearable technologies. For example, PPE (personal protective equipment), tracking of employees working in Potroom, workforce distribution for activities like pot tapping and anode setting, visibility of actual number of workforces presents at the shop floor, etc., KPIs impact: Total recordable injury frequency rate (TRIFR), lost time injury frequency rate (LTIFR), absence cost, training completion rate, overall labor effectiveness (OLE), etc. Smart Warehouse Smart warehouse enables integration and coordination of various components and entities within a warehouse using advanced technologies and data-sharing mechanism, thereby

Fig. 4 End-to-end integrated operations

V. P. Rai and D. R. D

enhancing visibility, efficiency, and responsiveness across the entire network. For example, monitoring quantity of cathode block, busbar, soda ash, aluminum fluoride etc. with the help of digital solutions like digital inventory management system, etc. KPIs impact: Inventory accuracy, stockout rate, etc.

Control Tower Control tower augments decision-making by providing real-time visibility and insights across the value chain, in a single location through E2E (End to End) traceability of the complete value chain. For example, close looping of workorder for maintenance of different voltage drops present in the cells using unified platform for work order management. KPIs impact: Specific energy consumptions, number of pot outs, hot metal production per day, etc.

Recommended Use Cases Description The functional use cases discussed below will support digital capabilities discussed in Sect. 3.1. The below picture (Fig. 4) explains the use cases that formulates/covers each capability explained above.

Use Cases for Smart Machines A1. Digital twin for smelting units This creates a virtual representation of the actual cell. It will mirror activities inside the Hall-Héroult cell by modeling chemical reactions, anode adjustment, alumina feeding, metal tapping, etc. KPIs impacted: Specific energy consumption, operating rate, current efficiency, hot metal production per day, age of

Smelting 4.0: Digital Strategy for Aluminum Production

pots stopped, net carbon consumption, AlF3 consumption, etc. A2. Remote monitoring of cells This will track the status of pots by monitoring their critical parameters like voltage, noise, anode effect, line amperage, etc., in remote locations. This will provide information of parameters in near real time and notify the line in charge about abnormalities happening in the pots and potline. KPIs impacted: Specific energy consumptions (kWh/t), number of pot outs (Pots/Month) A3. Condition-based maintenance (CBM) CBM enables real-time visibility of critical equipment (overhead crane, pot tending machines, dry scrubber system, etc.) and provides alerts whenever the parameters (vibration, temperature, etc.) of the machine cross the threshold limits that monitor the actual condition of an asset and decides maintenance activities need to be done. KPIs impacted: Equipment availability, maintenance costs. A4. Predictive maintenance Machine-specific prediction model for equipment like overhead cranes, pot tending machines, dry scrubber systems, etc., which will do multi-parameter tracking resulting in early detection of failures (derailment of crane, excessive wear of end truck wheels, electrification, fault in the power circuit, electrical contact) which are frequently occurring and causing costly breakdown work. KPIs impacted: Equipment availability, maintenance costs.

Use Cases for Integrated Operations B1. Should be voltage calculator An integrated AI based system will calculate optimum average voltage considering parameters like different voltage drops (lining drop, busbar drop, joint drop, etc.) [16], age of the cell, average noise, amperage of the line, etc. A systematic and controlled reduction of cell voltages will reduce power demand. It will optimize the required voltage per pot of the plant by identifying pots running with overvoltage. KPIs impact: Specific energy consumption B2. Integrated raw material charging calculator. An integrated solution will consider parameters like noise of the cell, amperage of the line, temperature of the individual cells, number of anode effects, bath ratio, etc., and will calculate the actual quantity of added material (AlF3, soda ash, etc.) required for an individual cell. KPIs impact: AlF3 consumption B3. Paperless operation A digital solution for real-time data entry and data acquisition of technical parameters like bath level, bath

661

temperature, metal pad, AlF3 additions, joints drop, lining drop, amount of metal tapped, etc., for all the cells in the line triggering automated workflows for correction. KPIs impact: overall labor effectiveness (OLE) B4. Digital Corrective and preventive action (CAPA) and root cause analysis (RCA), This enables analysis of real-time as well as historical data to understand root cause of problems like anode effect, high noise, pot leakages, premature cell failure, high silicon and iron in the cell, etc. Also, preventive actions are provided and tracked to avoid repetition of the issues. KPIs impact: Specific energy consumptions, operating rate, hot metal production per day, current efficiency, age of pots stopped, net carbon consumption, and AlF3 consumption. B5. Digital simulation of smelter. A digital simulation of cell will run different ``What if'' scenarios like changes in the metal pad, lining material, side lining thickness, cathode material, voltage, etc., with the help of simulated model and technical data for enablement of parameter optimization or best scenarios to get maximum efficiency from the cell. KPIs impact: Specific energy consumption B6. AI-based batching of cells. An AI-based tapping planner will group cells according to their purity (% Si, % Fe, % Mn, etc., content), operational constraint like tapping time, and movement of overhead crane. KPIs impact: P1020 production (Si 0.1 % max, Fe 0.2 % max, Al 99.7 % min ) B7. End-to-end raw material quality monitoring. It enables traceability of all the input raw materials (alumina, soda ash, aluminum fluoride, carbon, etc.) added in the process at each stage which allows complete visibility across the quality and quantity of material used in the production process. It will enable end-to-end mass balancing of input material to analyze sources of Si, Fe, Mn, etc., content in the metal. KPIs impact: P1020 Production B8. Laboratory information management system (LIMS). With availability of one integrated system covering alumina plant, smelter operation and process control, fabrication, and procurement can provide all the data at one place. A fully integrated system can capture data and provide visibility of parameters where inter-department and inter-plant communication happens for the same measured sample in the central laboratory. KPIs impacted: OLE. B9. E-Permit to work. An electronic permit to work is a digital work authorization solution that can synchronize different departments

662

V. P. Rai and D. R. D

Fig. 5 Logical implementation roadmap

and unscheduled work like removal of failed cell, repair of crane, etc., and systematize work related to contractors. KPIs impact: OLE

Use Cases for Integrated Workforce C1. PPE tracking. Real-time tracking of PPE adherence like use of safety helmet, industrial shoes, safety goggles, ear plugs, etc., in potlines using advanced camera leveraging digital image processing techniques. KPIs impact: Total recordable injury frequency rate (TRIFR), lost time injury frequency rate (LTIFR) C2. Workforce management system (WFMS). Digital workforce planning enables worker's allocation for the activity as per skill, prediction of labor requirement according to activities of the Potroom, scheduling of activities, time tracking, absence management, analytics, and easy accessibility of data related to workforce. It will help in improving workforce visibility across different functions. KPIs impact: OLE C3. Virtual training. Virtual on-the-job training will simulate real-world experience for critical processes like crane operation, bath temperature measurement, anode setting, metal tapping, metal sampling by leveraging augmented reality and virtual reality. KPIs impact: OLE, total cost of employee training (TCET), training completion rate Use Cases for Smart Warehouse D1. Inventory management system. Digital inventory management platform imports data from different applications and prepares dashboards for goods, e.g., cathode block, busbar, anode rod, aluminum fluoride, soda ash, etc., by considering the demand, supply,

lead times, transit times, etc., of the material. It helps in automatically managing safety stock and obsolescence, and it reduces the risk of stock out. KPIs impact: Inventory accuracy, stockout rate

Use Cases for Control Tower E1. Personal-based dashboard. Personal-based dashboards will give a summary of shift parameters (anode effects, noisy pots, temperature, voltage, etc.) based on real-time data coming from the cell controllers and historical data in the database. It will provide visibility of operations. Based on the summarized reports, personal like shift in charge, line manager, and plant head will take necessary actions to control abnormalities like repeated anode effect, over voltage, high bath level, over tapping, etc. KPIs impact: Specific energy consumptions (kWh/t), operating rate, hot metal production per day (t), P1020 production, net carbon consumption (kg/t), AlF3 consumption (kg/t), OLE E2. Potroom environment monitoring system. Digital environment monitoring system will be used for real-time monitoring of harmful gases like perfluorocarbons, hydrogen fluorides, and carbon monoxide in ambient air in the pot room. KPIs impact: AlF3 consumption (kg/t)

Logical Implementation Roadmap (Short Term, Mid-Term, Long Term) Figure 5 below gives roadmaps for the implementation of use cases. Timeline indicators are for design, development, testing, and rollout. Prioritization of the use cases is done based on complexity, dependency, and time required for implementation.

Smelting 4.0: Digital Strategy for Aluminum Production

Conclusion The benefits of Industry Revolution 4.0 span across safety, production, quality, maintenance, etc. As a result of the use cases that come under the Industry 4.0 umbrella will transform aluminum smelting. Staying competitive in a digitally transforming world is about continuous improvement with an Industry 4.0 focus. By leveraging all the use cases suggested a smelter can leverage the benefit of the new industrial revolution.

References 1. Value chain: https://www.ega.ae/en/about-us/our-value-chain. Accessed 6 September 2023 2. Reverdy, M., Potocnik, V. (2020). History of Inventions and Innovations for Aluminum Production. In: TMS 2020 149th Annual Meeting & Exhibition Supplemental Proceedings. The Minerals, Metals & Materials Series. Springer, Cham. https://doi. org/10.1007/978-3-030-36296-6_175 3. Rai, V.P., Upadhyay, V (2019). Failure Analysis of Low Amperage (70KA) Hall-Heroult Electrolytic Cell. In: APM 2019, IISc Bangalore - Conference on the Advances in Process Metallurgy. 4. The third industrial revolution, Impact of science on society: https://unesdoc.unesco.org/ark:/48223/pf0000075479/PDF/ 075479engo.pdf.multi.Accessed 6 September 2023. 5. Schwab, K. The Fourth Industrial Revolution: What It Means and How to Respond. 2016. Available online: https://www.weforum. org/agenda/2016/01/the-fourth-industrial-revolution-what-itmeans-and-how-to-respond/#. Accessed 6 September 2023. 6. Mokyr, J., The Second Industrial Revolution, 1870–1914: https:// faculty.wcas.northwestern.edu/jmokyr/castronovo.pdf. Accessed 6 September 2023.

663 7. Rai, V.P., Upadhyay, V. (2020). Redesigning of Current Carrying Conductor—The Energy Reduction Initiative in Low Amperage Hall-Héroult Cell. In: Tomsett, A. (eds) Light Metals 2020. The Minerals, Metals & Materials Series. Springer, Cham. https://doi. org/10.1007/978-3-030-36408-3_176 8. Rai, V.P., Upadhyay, V. (2021). Stepped Collector Bar—Continuous Developments in Low Amperage Hall-Héroult Cell to Reduce Voltage Drop. In: Perander, L. (eds) Light Metals 2021. The Minerals, Metals & Materials Series. Springer, Cham. https:// doi.org/10.1007/978-3-030-65396-5_86. 9. Rai, V.P., Upadhyay, V. Initiatives to reduce specific energy consumption in low amperage hall-Heroult electrolysis process. In ICNFM 2023, International conference on Nonferrous Minerals & Metals. 10. Energy area: https://jnarddc.gov.in/en/services/Energy.aspx. Accessed 6 September 2023 11. Rai, V.P., Upadhyay, V. (2020). Restart of Shutdown Pots: Troubles, Solutions and Comparison with Normal Pots to Improve Results. In: Tomsett, A. (eds) Light Metals 2020. The Minerals, Metals & Materials Series. Springer, Cham. https://doi.org/ 10.1007/978-3-030-36408-3_80. 12. Rai, VP, Upadhyay, V (2020) Managing potlines containing low amperage Hall-Héroult cell through operational and technical excellence during blackout and load reductions. Materials Today: Proceedings, 27 (part 3): 2941–2946. https://doi.org/10.1016/j. matpr.2020.03.783. 13. Aluminum for Climate:https://www3.weforum.org/docs/WEF_ Aluminium_for_Climate_2020.pdf. Accessed 6 September 2023 14. Das, BC, Chaudhury, S (1995), Accidents in the aluminum smelting industry, Industrial Health; 33(4):191–8. https://doi.org/ 10.2486/indhealth.33.191. PMID: 8557540. 15. Cyber-attack on Hydro: https://www.hydro.com/en/media/on-theagenda/cyber-attack/. Accessed 6 September 2023. 16. Rai, V.P., Upadhyay, V. (2021). R&D Projects for Improving Aluminium Smelting Technology: An Energy Reduction Initiative. In: Perander, L. (eds) Light Metals 2021. The Minerals, Metals & Materials Series. Springer, Cham. https://doi.org/10.1007/978-3030-65396-5_49.

Study of the Degradation of Ordinary Refractory Bricks in an Aluminum Reduction Cell Mohamed Hassen Ben Salem, Gervais Soucy, Daniel Marceau, Antoine Godefroy, and Sébastien Charest

Abstract

Keywords

On the lining of the aluminum reduction cells, layers of refractories are used to ensure proper thermal equilibrium as well as the protection of the insulating bricks underneath from high temperatures and chemical attacks. These materials, which largely affect the cell's life, are subject to corrosion by the electrolyte bath. In this paper, the breakdowns and/or the degradation of the ordinary refractory bricks (ORBs) located on the sidewalls and below the cathode blocks are investigated during industrial autopsies performed after cell failure or scheduled shutdowns. Subsequent chemical characterization on the ORBs was carried out using Scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM–EDX), X-ray fluorescence (XRF), X-ray powder diffraction (XRD), and Differential scanning calorimetry (DSC) analysis. The XRD results indicate that nepheline, cryolite, and silicon are present in both sidewall and bottom ORBs with higher presence of sodium fluoride phase in bottom ORBs. Comprehensive literature review is also conducted based on the degradation of ORBs and compared to the results obtained during industrial autopsies. Subsequently, ORBs samples will be exposed to electrolytic bath contamination at different temperatures and durations at the laboratory scale. The aim is to examine the contamination history endured by ORBs in an aluminum reduction cell.

Ordinary refractory bricks Autopsies

M. H. Ben Salem
G. Soucy (&) Université de Sherbrooke, 2500, Boul. de l’Université, Sherbrooke, QC J1K 2R1, Canada e-mail: [email protected] D. Marceau University Research Centre on Aluminium (CURAL) – Aluminium Research Centre (REGAL), Université du Québec à Chicoutimi, 555, Boul. de l’Université, Chicoutimi, QC G7H 2B1, Canada A. Godefroy
S. Charest Aluminerie Alouette Inc., 400, chemin de la Pointe-Noire, P.O. 1650, Sept-Îles, QC G4R 5M9, Canada




Corrosion




Cell life




Introduction The Hall-Héroult process, a cornerstone of aluminum production, presents a complex interplay of factors that influence the performance and longevity of reduction cells. This process is characterized by its thermal, chemical, and mechanical aggressiveness, as it operates at high temperatures and involves the exposure of materials to corrosive molten aluminum compounds. The choice of materials within the cell is critical, as they must withstand these harsh conditions. Refractories and insulating bricks are particularly essential components, forming a protective coating around the carbon cathode to shield against chemical degradation by the electrolyte. This is where ORBs play a critical role. They must endure not only the high temperatures but also the chemical reactivity of the bath which generate intense heat and create a chemically challenging environment. The corrosion of ORBs weakens the protective lining around the cathode and the sidewalls, reducing its effectiveness in shielding against chemical degradation. Moreover, the corrosion induces structural changes in the refractory bricks, which can lead to important changes in their thermo-physical and mechanical properties resulting in severe impact on the heat balance and in more severe cases to failure in the lining and or in the cathode block. Therefore, a comprehensive understanding of the chemical reactions taking place within the ORBs, particularly with the aggressiveness of the electrolyte bath, is essential to extend the lifespan of the aluminium reduction cells, a crucial step in both minimizing environmental impact and enhancing the economic sustainability of aluminum production.

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_85

664

Study of the Degradation of Ordinary Refractory Bricks in an Aluminum Reduction Cell

665

Previous Work

Compositional Analysis of Reacted ORBs

Researchers in the field of aluminum production have undertaken a series of studies to investigate refractory materials in the context of aluminum reduction cells. In 1997, Kvam and Øye [1] delved into the homogeneity and degradation of silicon carbide (SiC) side linings, providing insights into the properties and performance of this material as used in aluminum reduction cells. Rutlin and Grande [2] examined fluoride attack on alumino-silicate refractories, shedding light on the mechanisms of fluoride corrosion and its impact on refractory materials. In 2001, Pelletier et al. [3] introduced a unified approach to allow a comprehensive understanding of the corrosion mechanisms in the potlining refractories. In 2003, Pelletier and Allaire [5] extended their exploration of potlining refractory corrosion, focusing on the influence of cathode materials and the interactions with refractories. Siljan et al. [4] presented a state-of-the-art review of alumino-silicate refractories, covering advancements in refractory materials and their performance in the demanding environment of aluminum production. Building upon the insights gained from previous studies, the objective of the work is to take a step further by characterizing the ORBs obtained from industrial autopsies. This comprehensive characterization will provide essential data to unlock the history of degradation within aluminum reduction cells. Furthermore, the proposed methodology includes subjecting these ORBs to laboratory corrosion tests using the electrolyte bath as used in the aluminum industry. This approach aims to bridge the gap between industrial observations and controlled laboratory experiments, offering a holistic perspective on the factors influencing the degradation of ORBs in aluminum reduction cells.

The virgin alumina-silicate ORBs had an initial composition of 72.5 wt% SiO2, 19.5 wt% Al2O3, 5 wt% Fe2O3, and 3 wt % K2O. To investigate ORBs degradation, collected various samples from both the build-up and reacted zones were collected manually at three different cathode lining cross sections. The phase and elemental composition of the reacted ORB material were analyzed using X-ray diffraction (XRD) and X-ray Fluorescence (XRF). In XRD analysis, a step size of 0.05° and a counting time of 1 min were employed within the 2-Theta range of 10–70°, utilizing an X’Pert Pro MPD X-ray diffractometer equipped with a Pixel detector. Elemental mapping of the transformed refractory was conducted using XRF with a Axios Advanced Pan Analytical spectrometer featuring wavelength dispersive analysis (WD-XRF). The XRF results revealed the qualitative presence of elements including F, Na, Al, Si, K, Fe, and Ca within the refractory material.

Autopsies of Ordinary Refractory Bricks Samples were obtained from seven separate autopsies conducted in an industry. These samples were carefully collected from various locations along the sidewalls and beneath the cathodes of different decommissioned pots. Particular attention was paid to preserving the orientation and precise positioning of each sample. This approach was taken to establish a degradation map based on the age and operational history of each cell. The autopsies and laboratory assessments of spent pot linings from these decommissioned cells, all lined with High Quartz Bricks (ORBs HQB: SiO2 content >75 wt% and Al2O3 content 4 mm) after mechanical separation mUBC: mass of melted UBC sample.

Re-Melting of Dross with Salt Flux Re-melting of dross samples was performed, in one out of three experiments, using a typical Cl-based salt flux (NaCl: KCl 1:1 ratio and 5% wt. cryolite) applied in aluminium recycling operations. Salt flux treatment for metal separation is highly applied in recycling plants, either directly in rotary drum furnaces or as second step to retrieve aluminium trapped in the dross from reverberatory furnaces [9, 16]. A mass ratio of 1:2 dross/salt was employed. The mixture was placed in graphite crucibles in the electric resistance furnace under atmospheric conditions at 800 °C. After *25 min the melt was cast. The metallic aluminium was retrieved by mechanical treatment in jaw crusher and

mi + mMD  100 mUBC

ð4Þ

mi: mass of initial cast metal mMD: mass of metal recovered from re-melted dross mUBC: mass of melted UBC sample. In case of UBC addition in AA3104 hot heel, the UBC share in metal recovery was calculated using a mass balance approach. More specifically, the amount of totally produced aluminium metal was considered equal to the amount of aluminium recovered from the AA3104 plus the amount of aluminium recovered from the added UBCs. The share of AA3104 recovery was first calculated based on the data of the reference melting test and assumed constant in all subsequent UBC melting trials. This hypothesis enabled the calculation of the worst-case scenario in terms of UBC share in metal recovery. The share of UBC recovery in the hot heel experiments (MRHH) was calculated using Eq. (5). MRHH ð%Þ ¼

ðmi
100Þ  ðR3104
m3104 Þ mUBC

ð5Þ

mi: mass of initial cast metal R3104: (%) metal recovery from reference AA3104, calculated by the reference tests. m3104: mass of AA3104 hot heel used in the test mUBC: mass of melted UBC sample. For the mechanical separation of dross and re-melting of dross with salt flux, in hot heel tests, the additional retrieved metal from the dross, mx>4 mm and mMD, respectively, was added to the mass of initial cast metal, mi in Eq. (5), in accordance with the principle described in Eqs. (3) and (4).

Results and Discussion Analysis of UBCs from the Market The analysis of the body part of commercial beverage cans from the market showed that the majority of the external can coating was polyester-based, while the internal one was epoxy-based. Typical FTIR spectra of both lacquer types found in the body part of a beverage can are depicted in Fig. 2. This result agrees with FTIR results derived from

Recovery Considerations in the Pyrometallurgical Recycling of Used Beverage Cans

933

Fig. 2 Typical FTIR spectra of polyester (external) and epoxy-based (internal) lacquers found in the body part of a beverage can from the market

Table 1 Carbon content (% wt.) results on UBCs directly from the market and recycling plant

Sample

Average value

Minimum value

Maximum value

UBCs from the market

2.3

1.8

3.0

UBCs from the plant

2.6

1.9

3.5

spectroscopic study on UBC sample collected from the recycling plant bale. Considering carbon content determination, the results are summarized in the following table (Table 1). The average value of carbon content in both cases is similar, which is consistent with the literature data from other studies [13, 17].

Delacquering Considerations Study of Lacquer Decomposition Using Thermal Analysis The TGA results from the thermal decomposition of polyester- and epoxy-based lacquers under four different atmospheric conditions are presented in Fig. 3a and b. Additionally, Table 2 summarizes the results of the main decomposition phenomena of both lacquers, i.e. the initial (Ti), peak (Tp), and final (Tf) temperatures—as derived from the differential thermogravimetry (DTG) curves—and corresponding weight losses. As can be seen, the main decomposition phenomena end up to *480 °C, while the presence of oxygen promotes combustion process in temperatures higher than 550 °C. In more detail, under inert atmosphere (nitrogen) the coating degradation follows one decomposition step, during which volatilization phenomena and cracking of aliphatics prevail leaving pigments, fillers, and carbon residue (char) on aluminium surface [11, 13]. Presence of oxygen leads additionally to combustion reactions, resulting in char removal. The combustion temperature depends on oxygen concentration; the higher the oxygen content, the lower the combustion temperature. In order for the char to be decomposed within the examined temperature range, the presence of oxygen is mandatory. Especially, in

lower oxygen contents, i.e. 4 and 8% v/v O2, complete decomposition shifts to higher temperatures, as the required temperature for char combustion exceeds *580 °C in both lacquers. Nevertheless, when considering real industrial conditions, oxygen requirements should fulfill a twofold objective: facilitate the effective delacquering process including char decomposition, while simultaneously prevent the evolution of aluminium oxidation, a phenomenon known to progress above 550 °C. Typical isothermal curves of both lacquers (the average weight loss from both lacquers is shown) in synthetic air atmosphere are presented in Fig. 4; similar curves were recorded in case of the other examined conditions (not shown here). The effect of atmosphere on delacquering degree is comparatively depicted in Fig. 5, in which the aggregated results are derived from isothermal measurements of both lacquers (average delacquering degree value) after 35 min thermal treatment inside the TGA furnace. Delacquering in nitrogen atmosphere leaves a char residue of *8% wt. at 550 °C. Following the obtained curves, it can be seen that delacquering degree in 8% v/v O2 lies somewhat between synthetic air and nitrogen atmosphere. On closer observation, it is revealed that delacquering follows more similar behavior to inert atmosphere up to *350 °C, while it resembles that of synthetic air from 375 to 550 °C. Interestingly enough, delacquering degree in 4% v/v O2 presents the lowest values compared to all conditions—even compared to nitrogen atmosphere—up to *475 °C. It seems that the presence of oxygen above a critical concentration, i.e.  8% v/v, advances degradation phenomena at temperatures lower than 350 °C and higher than 450 °C compared to nitrogen and 4% v/v O2 conditions. In the intermediate region, nitrogen yields higher decomposition—

934

Th. Tzevelekou et al.

Fig. 3 TGA and DTG curves of a polyester– and b epoxy-based lacquers in various atmospheres: synthetic air, 8% O2 v/v, 4% O2 v/v, and nitrogen

even compared to synthetic air; however, a considerable amount of carbon residue is then left, which can be removed only in the presence of oxygen. In total, those described phenomena might be attributable to competitive degradation and combustion phenomena occurring concurrently in the presence of oxygen. Dedicated study on polymer thermal

degradation kinetics and experimental work is required [18, 19] to determine the effect of atmosphere on involved reactions and rate determining step, which was outside the scope of the present work. The latter is directed in demonstrating the effect of atmosphere on overall delacquering degree at conditions of industrial interest. Indicatively, for

Recovery Considerations in the Pyrometallurgical Recycling of Used Beverage Cans Table 2 Results from the main decomposition steps of polyesterand epoxy-based lacquers

935

Polyester

Step

Ti (°C)

Tp (°C)

Tf (°C)

Dm (%)

Synthetic air

1

372

411

429

83

2

525

549

567

15

8% O2 v/v

1

359

396

431

85

2

544

578

616

15

1

383

423

442

87

2

584

647

697

13

1

378

410

431

92

1

330

370

378

35

2

411

420

443

28

3

527

556

587

32

4% O2 v/v Nitrogen Epoxy Synthetic air

8% O2 v/v 4% O2 v/v Nitrogen

1

403

425

446

80

2

561

592

637

19

1

416

452

468

82

2

590

651

694

17

1

402

430

446

93

35 min processing time a scrap temperature  400 °C could promote high delacquering degrees >70% wt. Almost complete delacquering is achieved at *500 °C for synthetic air and 8% v/v O2, while a temperature of *550 °C is necessary in case of 4% v/v O2. Overall, the results of the isothermal curves may provide a relation between the delacquering degree and temperature to better feed our simulation process model. The studied oxygen range is industrially oriented; even if a particular setpoint may be defined for oxygen regulation at a delacquering furnace, atmosphere variations may occur subject to operational activities. Fig. 4 Typical isothermal TGA curves of polyester- and epoxy-based lacquers (average weight loss) under synthetic air atmosphere

UBC Delacquering in Lab-Scale Furnace Images of ASR UBCs and UBCs after delacquering at each temperature are shown in Fig. 6, along with the actual achieved delacquering degree. According to our previous work [15], comparison between lacquer TGA isothermal measurements and the results of UBC delacquering by the thermal treatment in electric resistance furnace for the same nominal conditions illustrated very good agreement confirming that TGA is a useful and reliable tool to study delacquering kinetics and design of industrial trials. UBCs of 17.5% wt. delacquering degree (calculated by Eq. 1) had a similar appearance to ASR samples. At 50% wt.

936

Th. Tzevelekou et al.

Fig. 5 Aggregated results from isothermal measurements of both lacquers (average delacquering degree) after 35 min thermal treatment inside the TGA furnace

delacquering degree UBCs were black/brown indicating char formation along with unreacted coating on their surface. At 75% wt. UBCs had black and shiny metallic surfaces, while fully delacquered UBCs exhibited shiny metallic surface in their majority.

Aluminium Recovery Considerations Metal Recovery in Direct UBC Melting Figure 7a shows the results of metal recovery (calculated by Eq. 2) in direct UBC melting relying only on initial cast metal weight. Relatively low metal recovery values are revealed, not higher than 60% wt. even in the case of completely delacquered UBCs. The lowest metal recovery (*50% wt.) was obtained for samples of 50% wt.

Fig. 6 Thermal treatment of chopped UBCs at four different temperatures (T = 270, 350, 400, and 500 °C) at a laboratory electric resistance furnace (holding time: 60 min) and achieved delacquering

delacquering degree, while no and partial (17.5% wt.) decoating also led to low values together with high weighing scattering. These results reflect the fact that during laboratory experiments, very fast metal solidification was observed, which hindered efficient metal and dross separation. The most interesting finding though is that half-delacquered UBCs boasted the worst metal recovery, even compared to the ASR ones. It is believed that the remaining char residue on the can sheets is responsible for this result. Presence of this char content along with unreacted coatings (Fig. 6) produced a mixture of honey-like viscosity, which strongly hindered metal coalescence and resulted in trapping a large amount of aluminium metal in the dross [20]. When moving to higher delacquering degrees, i.e. 75 and 95% wt., the recovery increased.

degree. The ASR sample is also depicted. The shiny metallic surface has been revealed in the case of the fully delacquered UBCs (T = 500 ° C)

Recovery Considerations in the Pyrometallurgical Recycling of Used Beverage Cans

937

Fig. 7 Results of metal recovery determination in direct UBC melting method after a initial cast metal weighing (metal recovery results from the three specimens and their average value are included in the graph) and b mechanical (average of two tests) and salt flux separation methods; initial cast metal weighing results are also shown for comparison reasons

Figure 7b presents the metal recovery obtained after mechanical separation (Eq. 3) and salt flux separation (Eq. 4) of the collected dross samples, in comparison to the one calculated based solely on the weight of initially cast metal. As expected, the attained values are much higher than the ones obtained considering only the weight of the initial cast products. As a general outcome, low delacquering degrees up to 50% wt. are accompanied by low metal recovery in both methods. Regardless of the delacquering degree, it is observed that higher metal recoveries >80% wt. were achieved using mechanical separation. It is possible

that the complex structural morphology of the solidified metal-dross mixture (Fig. 1) did not allow efficient salt penetration, wetting, and metal separation during dross remelting, while the presence of non-metallic contaminants in the mechanically recovered metal cannot be excluded.

Metal Recovery in UBC Melting Using AA3104 Alloy Hot Heel Figure 8a shows the results of metal recovery in UBC melting using AA3104 alloy hot heel method relying only on initial cast metal weight. The respective recovery

938

Th. Tzevelekou et al.

Fig. 8 Results of metal recovery determination in UBC melting using AA3104 alloy hot heel method after a initial cast metal weighing (metal recovery results from the three specimens and their average value are included in the graph) and b mechanical (average of two tests) and salt flux separation methods; initial cast metal weighing results and AA3104 recoveries from the three examined separation procedures are also shown for comparison reasons

obtained by the reference trials of the AA3104, i.e. * 90% wt., is also included. It is noted that, as explained in the methodology section, the calculated recoveries in the graphs refer only to UBC share, with the assumption of unvarying AA3104 contribution in recovery (Eq. 5). A descending metal recovery trend is revealed up to 50% wt. delacquering degree, while recovery increases for delacquering degrees 75% wt., reaching a maximum in the case of delacquered UBCs. The increased scattering observed for 17.5 and 50% wt. delacquering degree underlines again the pre-described experimental difficultly in initial separation between metal and dross during casting. In addition, errors inducing discrepancy in the results can also be introduced by the experimental procedure in lab-scale (e.g. manual stirring and skimming). Nevertheless, it is also clear that the ASR UBCs boast a higher metal recovery than the ones obtained by partial or half-delacquered ones. It could be possible that in case of high unreacted coating content (ASR or lower delacquering degree), there is an instant temperature increase due to the direct firing (spontaneous combustion) of the samples in the furnace aiding metal liquation and separation

from dross. This could explain why this recovery minimum is more obvious in hot heel tests, where UBC immersion constrained oxygen contact and direct combustion, compared to the direct UBC melting method. The results exemplify that combination of adhering carbon residue (char) on aluminium sheets along with unreacted coatings has a detrimental effect on recovery, and industrial recycling practices should focus on methodologies for their removal, prior to UBC re-melting [11]. Figure 8b illustrates the metal recovery obtained after mechanical and salt flux separation of the collected dross samples, compared to the one calculated based only on the weight of the initially cast metal. The metal recovery of reference AA3104 was *97 and 95% wt. for salt flux and mechanical separation methods, respectively. High delacquering degrees,  75% wt., result in very high, almost complete metal recoveries. The results indicate that salt flux treatment of the dross obtained by this experimental procedure is more efficient yielding higher recoveries than mechanical separation, more profound in lower delacquering degrees. This is most possibly the positive effect of AA3104

Recovery Considerations in the Pyrometallurgical Recycling of Used Beverage Cans

dross fraction in the cast dross sample, exhibiting more homogeneous structure, thus allowing for more effective flux wetting and metal coalescence during dross re-melting with flux [11, 21]. It is worth mentioning though, that in case of completely delacquered UBCs, mechanical separation also provides maximum recovery values. Finally, a recovery trough (minimum in the graph) close to * 50 % wt. delacquering is again observed. In total also considering the recovered aluminum from the dross, the achieved recoveries by hot heel method at higher delacquering degrees are significantly higher than the ones obtained in the corresponding direct UBC melting trials, demonstrating the beneficial effect of the use of hot heel in constraining oxidation losses. Furthermore, its positive contribution to facilitate can sheet liquation and melt homogeneity is also demonstrated. Hot heel method also resembles more the actual industrial recycling practice of UBC bale bath immersion. Finally, since the lab-scale experiments were performed in a furnace with no special provision for the atmosphere, the results showed that once coated aluminium is delacquered, re-melting by hot heel immersion—even in non-inert atmosphere—yields high recoveries. Lastly, it should be underlined that the results obtained in the current study in terms of quantified delacquering degreerecovery correlation should not be generalized but correspond to the applied lab-scale experimental conditions. Nevertheless, the demonstrated trends may be used as a guiding tool towards the selection of suitable process parameters in favor of optimal recovery.

Conclusions and Future Work UBCs possess both high value and complexity as a secondary aluminum source, emphasizing the need for industrial practices geared towards efficient recycling. The purpose of this work was to elucidate the influence of the extent of UBC thermal pre-treatment on aluminum recovery. Considering TGA findings, employing two distinct melting methods, and utilizing three assessment procedures for aluminum recovery, a comprehensive understanding emerged. The outcomes highlighted the profound effects of delacquering degree on melting behavior and subsequent dross formation. Efficient delacquering contributes to improved metal recovery and minimization of dross formation under controlled process conditions, such as furnace atmosphere and temperature. Parameters of thermal pre-treatment should be carefully selected not to result in conditions resembling intermediate delacquering degrees. We confirmed and thoroughly demonstrated over the whole delacquering range that the remaining char residue along with unreacted coatings

939

hinders metal separation at subsequent re-melting resulting in reduced recoveries. Oxygen presence in de-lacquering atmosphere could be beneficial to allow for char combustion; nevertheless, its optimum concentration should be adjusted in relation to the applied specific process parameters, i.e. temperature profile, as illustrated by the obtained results at 4, 8, and 20.9% v/v O2. Aluminium recycling by hot heel immersion of delacquered UBCs yields almost complete metal recovery. Integration of the results in the developed process model will be examined targeting to improve UBC recycling. This research work yields insights relevant to optimization of recycling efficiency, supporting the preservation of valuable material resources and sustainable development of the aluminum industry. Acknowledgements Special acknowledgements go to Dimitris Voulgaris, Head of the Surface Science and Coatings Department of ELKEME SA for his cooperation and support in this research work.

References 1. Communication document from the Commission, “The European Green Deal”, COM/2019/640 final. https://eur-lex.europa.eu/legalcontent/EN/ALL/?uri=COM:2019:640:FIN. 2. Proposal for a regulation of the European Parliament and of the Council establishing a framework for ensuring a secure and sustainable supply of critical raw materials and amending Regulations (EU) 168/2013, (EU) 2018/858, 2018/1724 and (EU) 2019/102, COM(2023) 160 final, 2023/0079(COD). https:// eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX% 3A52023PC0160. 3. Proposal for a regulation of the European Parliament and of the Council establishing a framework for ensuring a secure and sustainable supply of critical raw materials and amending Regulations (EU) 168/2013, (EU) 2018/858, 2018/1724 and (EU) 2019/1020 – Mandate for negotiations with the European Parliament. 4. Aluminium beverage can recycling rates 2020, European Aluminium, 6/12/2022. https://european-aluminium.eu/wp-content/ uploads/2022/12/14-12-2022_European-Aluminium_Reportalumnium-beverage-can-recycling-rates-2020.pdf. 5. https://www.recyclingtoday.com/news/ubc-recycling-rate-goalscmi/. 6. https://www.spglobal.com/commodityinsights/en/market-insights/ latest-news/metals/013122-aluminum-association-targeting-moreus-states-to-raise-ubc-recycling-conference. 7. A Circularity Case for Aluminium Compared with Glass and Plastic (PET), International Aluminium Institute, March 2022. https://international-aluminium.org/resource/aluminium-beveragecan-study/. 8. Opportunities for Aluminium in A Post-Covid Economy, International Aluminium Institute, March 2022. https://internationalaluminium.org/resource/opportunities-for-aluminium-in-a-postcovid-economy/?_thumbnail_id=6715. 9. Raabe D, Ponge D, Uggowitzer PJ, Roscher M, Paolantonio M, Liu C, Antrekowitsch H, Kozeschnik E, Seidmann D, Gault B, De Geuser F, Deschamps A, Hutchinson C, Liu C, Li Z, Prangnell P, Robson J, Shanthraj P, Vakili S, Sinclair C, Bourgeois L,

940

10.

11.

12. 13.

14.

15.

Th. Tzevelekou et al. Pogatscher S (2022) Making sustainable aluminum by recycling scrap: The science of “dirty” alloys. Prog. Mater. Sci. 128:100947. Stotz PM, Niero M, Bey N, Paraskevas D (2017) Environmental screening of novel technologies to increase material circularity: A case study on aluminium cans. Resour. Conserv. Recycl. 127:96–106. Steglich J, Dittrich R, Rosefort M, Friedrich B (2016) Pre-Treatment of Beverage Can Scrap to Increase Recycling Efficiency. JMSE-A 3(3-4):57–65. Kvithyld A, Meskers CEM, Gaal S, Reuter M, Engh TA (2008) Recycling light metals: Optimal thermal de-coating. JOM 60:47–51. Steglich J, Dittrich R, Rombach G, Rosefort M, Friedrich B, Pichat A (2017) Dross Formation Mechanisms of Thermally Pre-Treated Used Beverage Can Scrap Bales with Different Density. In: Ratvik, A. (eds) Light Metals 2017. The Minerals, Metals & Materials Series. Springer, Cham. Vallejo-Olivares A, Høgåsen S, Kvithyld A, Tranell G (2022) Thermal De-coating Pre-treatment for Loose or Compacted Aluminum Scrap and Consequences for Salt-Flux Recycling. J. Sustain. Metall. 8:1485–1497. Chamakos, N. et al. (2023) Towards the Efficient Recycling of Used Beverage Cans: Numerical Study and Experimental Validation. In: Broek, S. (eds) Light Metals 2023. TMS 2023. The Minerals, Metals & Materials Series. Springer, Cham.

16. Niedermair, F., Wimroither, G. (2011) Latest Trends in Post Consumer and Light Gauge Scrap Processing to Include Problematic Processing Material such as UBC, Edge Trimmings and Loose Swarf. In: Lindsay, S.J. (eds) Light Metals 2011. Springer, Cham. 17. Pichat, A., Vassel, A., Menet, P.Y., Jouët-Pastre, L. (2020) Constellium R&D approach in recycling, from lab to industrial scale. In: Tomsett, A. (eds) Light Metals 2020. The Minerals, Metals & Materials Series. Springer, Cham. 18. Kvithyld, A. et al. (2002) Decoating of aluminium scrap in various atmospheres. In: Schneider, WA. (eds) Light Metals 2002. TMS 2002. The Minerals, Metals & Materials Series. 19. Vyazovkin S, Burnham AK, Favergeon L, Koga N, Moukhina E, Pérez-Maqueda LA, Sbirrazzuoli N (2020) ICTAC Kinetics Committee recommendations for analysis of multi-step kinetics Thermochim. Acta 689: 178597. 20. Gökelma, M., Diaz, F., Öner, I.E., Friedrich, B., Tranell, G. (2020) An assessment of recyclability of used aluminium coffee capsules. In: Tomsett, A. (eds) Light Metals 2020. The Minerals, Metals & Materials Series. Springer, Cham. 21. Capuzzi S, Kvithyld A, Timelli G, Nordmark A, Gumbmann E, Engh TA (2018) Coalescence of clean, coated, and decoated aluminum for various salts, and salt–scrap ratios. J. Sustain. Metall. 4(3):343–358.

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy Shahid Akhtar, Massoud Hassanabadi, and Ragnhild E. Aune

Abstract

Keywords

Removing transition metal impurities is vital in producing high-conductivity aluminum for wire and cable applications. These impurities, primarily titanium (Ti) and vanadium (V), can significantly reduce electrical conductivity. To address this challenge, laboratory-scale experiments were performed, focusing on the performance of aluminum-boron master alloys, specifically Al-3% B (AlB2 phase) and Al-8% B (AlB12 phase), in eliminating Ti and V from electrolysis metal ingots. In the present study, three critical parameters were addressed, i.e., holding time, stirring, and fluxing. All three parameters yielded positive outcomes, enhancing the electrical conductivity of the cast material. Cast house trials were conducted using Al-5% B (AlB2 phase) and Al-8% B (AlB12 phase). Furthermore, the melt quality was assessed through the Porous Disc Filtration Apparatus (PoDFA) method. These trials were conducted in the aluminum wire rod cast house, where the 1370 alloy wire rod production occurs. The results obtained from the cast house trials provided valuable insights. Specifically, using the Al-5% boron master alloy with the AlB2 phase demonstrated its capability to yield a cleaner melt, as evidenced by a reduced total inclusion count.

Electrical Conductor (EC) Grade Aluminum Alloy Al-B master alloys Melt quality Transition metal impurities Porous disc filtration apparatus (PoDFA) Continuous casting line (CLL)

S. Akhtar (&) Hydro Aluminium AS, Commercial Technology, Sunndalsøra, Norway e-mail: [email protected] M. Hassanabadi Hydro Aluminium R&D Center, Finspång, Sweden R. E. Aune Department of Materials Science and Engineering, Norwegian University of Science and Technology (NTNU), Trondheim, Norway
















Introduction The vast availability of crude bauxite in the Earth’s crust and the versatile properties of aluminum (Al) and its alloys have made aluminum a vital material in today’s modern world. The constant demand for aluminum in various branches of the economy has resulted in its consumption steadily increasing [1, 2]. As a result, aluminum has become the most widely used material for electricity transmission and distribution systems today [2]. While pure aluminum has served as a conductive material in various segments of the electrical industry for many years, aluminum alloys exhibit excellent conductivity while combining structural resilience. Moreover, aluminum is significantly lighter than copper (Cu), making it easier to handle due to its lower density, which is roughly one-third that of copper [3]. Additionally, the price of aluminum is relatively stable compared to copper, which experienced significant price fluctuations worldwide during the 1960s and 1970s [4]. Consequently, aluminum found its way into applications where copper had traditionally been the norm. To produce high-conductivity aluminum for wire and cable applications, it is essential to remove transition metal impurities. Elements such as titanium (Ti), vanadium (V), chromium (Cr), zirconium (Zr), and others, which are present as smelter impurities, have a detrimental effect on the electrical conductivity of aluminum and its alloys [5, 6]. These solute impurities are typically eliminated from the molten aluminum by adding boron (B) aluminum master alloys. Boron aluminum treatment is environmentally safer

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_118

941

942

and a cost-effective alternative to methods like adding borax (a powdery white substance also known as sodium tetraborate (Na2B4O7)) to electrolytic cells or employing potassium borofluoride (KBF4) treatment in the mixing or holding furnaces. Importantly, boron aluminum treatment reduces the risk of boron contamination in the electrolytic cells and eliminates refractory attacks by KBF4 [7]. Following the addition of boron aluminum master alloys, which often have an AlB2 or AlB12 microstructure, solute impurities such as Ti and V react with the aluminum borides. These particles precipitate and settle at the bottom of the melt, forming a boride sludge. Boron becomes trapped within this sludge, located at the bottom of the furnace, and clean, high-conductivity aluminum can then be further cast [8]. An alternative but less common method involves adding boron in rod or cast form in the launder. In this case, the borides that form are trapped in Ceramic Foam Filters (CFFs). Regardless of the method used, it is essential to note that boron recovery should not be expected in the final alloy. Over the years, numerous studies [9–15] have explored the mechanisms behind the removal of transition elements from molten aluminum through boron treatment. However, to the knowledge of the present authors, no information is available regarding the effect of boron treatment on the quality of the aluminum melt. In the present study, laboratory-scale trials involving the boron treatment of smelter-grade aluminum using boronaluminum master alloys, specifically those with an AlB12 and/or AlB2 phase, have been conducted. These experiments were also replicated under actual working conditions on the Continuous Casting Line (CCL) at the Hydro Aluminium Karmøy plant in Norway, where 8% AlB12 and 5% AlB2 waffles were employed. Melt quality assessments were carried out through the Porous Disc Filtration Apparatus (PoDFA) method, and the collected data was carefully analyzed.

S. Akhtar et al. Table 1 Chemical composition of the REFINAL 325XF flux MgCl2

KCl

F

Application rate

Density

35–45%

55–65%

0.5–1%

0.2–0.8 kg/t

2.17 g/cm3

Next, 2.5 kg of smelter-grade aluminum was loaded into Salamander® Super crucibles with dimensions of 184 mm in height, 156 mm in top diameter, and 108 mm in bottom diameter. The desired heat size was carefully measured and added to the crucible along with the boron-aluminum master alloys and flux. To achieve the casting temperature of aluminum, the crucible was placed inside a Nabertherm resistance furnace and heated to 998 ± 25 °K. The laboratory experiments’ primary objective was to investigate the effect of boron-aluminum master alloys, the flux, and the stirring duration on the removal efficiency of the transition metal elements Ti, V, Zr, and Cr. To assess this, electrical conductivity measurements were taken on disc samples before and after stirring the refined molten aluminum. Figure 1 presents the horizontal spectrograph disc mold used for sampling, yielding samples with a diameter of 57 mm and a thickness of 9.8 mm. The mold was constructed from steel and coated with boron nitride to ensure proper sampling. Initial samples were collected both before and immediately after stirring and dross removal from the surface of the molten aluminum. The stirring process was carried out using a boron nitride-coated steel impeller. The impeller, mounted in an electric-driven agitator equipped with a speed regulator (see Fig. 2), was operated at 120 RPM. Through the stirring

Experimental Procedure Laboratory Experiments The present section outlines the experimental procedure for optimizing boron treatment in the production of the 1370 electrically conductive grade aluminum alloy. The materials used in the experiments include smelter-grade aluminum provided by the Hydro Aluminium Karmøy plant, REFINAL 352XF alkali metal flux from MQP Ltd, UK, and boron-bearing substances Al-3% B (AlB2) and Al-8% B (AlB12) supplied by KBM. Table 1 presents details of the chemical composition, density, and application rate of the REFINAL 325XF flux.

Fig. 1 Horizontal spectrograph disc mold used for sampling

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy

943

Fig. 2 The experimental setup used for stirring the molten aluminum consisting of an electric agitator, an impeller, and a stand that allowed for adjusting the impeller distance to 3 cm from the bottom of the crucible

process, the surface of the furnace was covered by a thermal insulating refractory plate, effectively preventing excessive heat loss from the molten aluminum. To evaluate the effect of the holding time on the removal efficiency of the transition metal elements, samples were taken at intervals of 15, 30, 45, and 60 min following the stirring process. Additionally, a sample was secured from the bottom of the crucible after casting the remaining molten aluminum into a copper mold. Table 2 summarizes the boron-aluminum master alloys used, their respective % added, stirring durations in minutes for each specific case, and the corresponding % of fluxing additions. Before performing the electrical conductivity measurement, the collected samples underwent a preparation step, which involved the removal of the sprue and scalping 3 mm off the as-cast disc face. The electrical conductivity was

measured at room temperature and a frequency of 60 kHz using an Auto-Sigma 3000 Electrical Conductivity Meter (GE Inspection Technologies, New York, USA) with an accuracy of ± 0.1% IACS at 10% IACS and ± −0.5% IACS at 100% IACS. Elemental analysis of the samples was also performed using Optical Emission Spectroscopy (OES).

Industrial Trials Moving from laboratory-scale experiments to industrial trials, PoDFA samples were collected at the Hydro Karmøy CCL after boron treatment with Al-5% B (AlB2) and Al-8% B (AlB12) waffles. Boron waffles (10 kg each) were added together with cold metal in the industrial 32 ton melting furnaces in Karmøy CCL producing rods for electrical

Table 2 Experimental detail of the laboratory boron treatment of the smelter-grade aluminum Master alloy type and addition to aluminum [%] Al-3%B, 0.20%

Stirring time [min]

Flux [%]

Master alloy type and addition to aluminum [%]

Stirring time [min]

Flux [%]

2

No flux

Al-3%B, 0.20%

2

0.15%

10 Al-8%B, 0.10%

2

10 Al-8%B, 0.10%

10 Al-3%B, 0.10% Al-8%B, 0.05%

2 10

2 10

Al-3%B, 0.10% Al-8%B, 0.05%

2 10

944

S. Akhtar et al.

applications. Several charges were made by using Al-B 8% (AlB12) and Al-B 5% (AlB2) waffles. The addition of Al-B master alloy is calculated according to the formula below B¼

M ðCr þ Ti þ V þ ZrÞ P 2

highlight the critical factors influencing electrical conductivity and melt quality, bridging the gap between controlled lab conditions and real-world industrial applications.

Lab Experiments: A Closer Look at Electrical Conductivity

In the present section, the results of the obtained laboratory experiments are presented and discussed in view of their implications in the context of industrial trials. The findings

The laboratory experiments were designed to clarify the effects of the various parameters studied on the electrical conductivity of aluminum alloys, including the type of boron-aluminum master alloys, stirring, holding times, and fluxing used. Figure 3 presents the electrical conductivity results from the laboratory experiments conducted without fluxing and Fig. 4 with fluxing. Regarding the experiments without fluxing, it can be seen from Fig. 3 that adding boron-aluminum master alloys led to a significant improvement in electrical conductivity. This enhancement was observed across the different types of boron-aluminum master alloys, stirring conditions, holding times after stirring, and the presence of fluxing agents. In the case of the experiments with fluxing, it can be seen from Fig. 4 that comparable or higher electrical conductivities than in the case of experiments performed without fluxing were obtained even before stirring. Additionally, stirring itself led to a further increase in the electrical conductivity. Overall, combining stirring and fluxing of the molten

Fig. 3 Electrical conductivity results without fluxing. The following labels are used for clarity: (i) Ref: reference sample (smelter-grade aluminum). (ii) BS: sample before stirring. (iii) AS: sample after

stirring. (iv) Numbers 15, 30, 45, and 60: time intervals for sampling after stirring. (v) BC: electrical conductivity of the final sample collected from the bottom of the crucible

B = Weight in Kg of Al-xB. M = Weight of metal to be treated. P = % of B in Al-xB. The description of PoDFA method and its assessment criteria is described elsewhere [16, 17]. PoDFA samples after boron treatment from different charges was taken before and after Ceramic Foam Filter (CFF). Metallographic analysis of the solidified residue was carried out for the quantitative measurement of total inclusion content (mm2/kg), and types of inclusions were also identified.

Results and Discussion

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy

945

Fig. 4 Electrical conductivity results from fluxing. The following labels are used for clarity: (i) Ref: reference sample (smelter-grade aluminum). (ii) BS: sample before stirring. (iii) AS: sample after

stirring. (iv) Numbers 15, 30, 45, and 60: time intervals for sampling after stirring. (v) BC: electrical conductivity of the final sample collected from the bottom of the crucible

aluminum consistently resulted in higher electrical conductivity, typically in the range of 0.5–1% International Annealed Copper Standard (IACS). The stirring process proved, in other words, to play a crucial role in distributing the boron-aluminum master alloys evenly throughout the charge, making the melt more homogenous. Given the higher affinity of transition metals for boron over aluminum, the formation of diborides, such as TiB2 and VB2, was facilitated as the diffusion distance between the transition metal elements and the boron-aluminum master alloys was reduced. This also enhanced the kinetics of the reaction responsible for forming the transition metal diborides. The diborides were captured in the presence of fluxes, and the extensive shear within the melt accelerated the transfer of flux particles to the dross. To gain a deeper understanding of the transition metal element removal efficiency, the chemical composition of the melt was analyzed using Optical Emission Spectroscopy (OES). The results are shown in Fig. 5a–d. As can be seen from Fig. 5, the stirring time did not yield any definitive conclusions. However, the graphs indicated that both the Ti and V concentrations decreased with holding time, particularly up to 30 min after stirring. Furthermore, trials with added flux exhibited lower Ti and V concentrations than non-fluxed trials, suggesting that the flux addition,

even without stirring, contributed to removing transition metal elements. However, a slight increase in Ti and V concentrations was observed in the trials involving Al-3% B and Al-8% B with 2 min of stirring and fluxing. This increase could be attributed to the short stirring process, leading to the re-entrainment of trapped Ti and V in the dross or sediment at the bottom of the crucible. The variability in the removal efficiency of transition metal elements between aluminum-boron master alloy variants, Al-3% B and Al-8% B, was not consistently apparent and depended on the specific experimental conditions. Therefore, it was concluded that mixing molten aluminum, adding fluxing agents, and, to some extent, controlling holding times were the key parameters for achieving higher removal efficiency of transition metal elements, consequently leading to enhanced electrical conductivity.

Industrial Trials: Scaling Up the Optimization Efforts The industrial trials sought to translate the insights gained from laboratory experiments into real-world applications. Figure 6 provides a glimpse into these trials by showcasing a

946

S. Akhtar et al.

Fig. 5 Concentration of Ti and V in the experimental samples with and without fluxing

Fig. 6 Cross-section of a fractured wire rod sample

Fig. 7 Longitudinal section of a fractured wire rod sample

cross-section of a wire rod sample that fractured during the drawing operation. The figure illustrates the presence of boride particles, which have a detrimental impact on material strength and load-bearing capacity.

To gain deeper insights, a longitudinal section from the fractured sample was prepared and examined under an optical microscope, as shown in Fig. 7.

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy

947

Fig. 8 a SEM image and corresponding b EDS spectra of a fractured wire rod sample

Fig. 9 PoDFA results using 8% AlB12 waffles where the numbers 1, 2, and 3 correspond to different charge numbers, while A and B denote samples taken before and after passing through the CFF

For a more detailed analysis, SEM coupled with EDS was used to identify the types of particles present. The results of this SEM/EDS analysis are presented in Fig. 8a and b. Figure 9 offers insights from the Porous Disc Filtration Apparatus (PoDFA) trials utilizing 8% AlB12 waffles.

Furthermore, Fig. 10 visually represents the PoDFA filter residue, which contained various particles. To gain a deeper understanding of the analysis, the SEM images and corresponding EDS spectra of PoDFA sample residues (charge with 8% AlB12 phase) are shown in Fig. 11.

948

S. Akhtar et al.

In other words, the presence of boride particles in the fractured sample underscores the importance of effective boron treatment in preventing material failure. The trials using 8% AlB12 and 5% AlB2 waffles revealed stark differences in melt quality. While the former exhibited high inclusion content and oxide numbers, indicative of a poor-quality melt, the latter demonstrated significantly lower inclusion content and oxide numbers, signifying a high-quality melt. Finally, Fig. 13 offers a closer look at a PoDFA micrograph, highlighting boride inclusions (see arrow). These results, derived from the industrial trials, provide invaluable insights into the characterization of materials involved in the boron treatment process.

Fig. 10 Micrograph of the PoDFA filter residue (charge with 8% AlB12 phase)

The figure highlights the presence of borides of transition elements, further emphasizing their role in material quality. The results from the PoDFA trials involving 5% AlB2 waffles are presented in Fig. 12. Here, too, the analysis underscores the presence of borides of transition elements, reinforcing their association with the boron treatment process of the melt.

Bridging Lab and Industrial Insights The fracture analysis of the drawn sample (Fig. 6) provides compelling evidence of the critical role played by effective boron treatment in preventing material failure. This serves as a clear reminder of the critical significance of optimizing boron treatment processes in industrial applications. The stark differences in melt quality observed between trials with 8% AlB12 and 5% AlB2 waffles further underscore the significance of fine-tuning these processes. The

Fig. 11 SEM image and corresponding EDS spectra of PoDFA sample residue (charge with 8% AlB12 phase)

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy

949

Fig. 12 PoDFA results using 5% AlB2 waffles where the numbers 1, 2, and 3 correspond to different charge numbers, while A and B denote samples taken before and after passing through the CFF

Furthermore, the in-depth SEM/EDS analysis (Figs. 11 and 12) provided valuable insights into the nature of the trapped particles. Predominantly, these particles were identified as borides of transition elements, highlighting their intimate connection with the boron treatment process of the melt.

Advancing Aluminum Manufacturing

Fig. 13 Micrograph of PoDFA residue (charge with 5% AlB2 phase)

former exhibited a wide range of inclusion content and oxide numbers, both indicative of poor melt quality. However, the latter consistently demonstrated significantly lower inclusion content and oxide numbers, indicating a high-quality melt.

The combined analysis of laboratory experiments and industrial trials offers a holistic understanding of the crucial factors influencing electrical conductivity and inclusion content in producing 1370 electrically conductive grade aluminum alloy. By optimizing the boron treatment processes, the quality, reliability, and efficiency of aluminum manufacturing were enhanced. The present findings provide a foundation for further research and development in pursuing high-performance aluminum alloys with enhanced electrical conductivity. It also bridges the gap between controlled lab conditions and real-world industrial applications.

950

Conclusions The findings derived from the present study provide essential insights into the refinement of molten aluminum for enhanced electrical conductivity and reduced transition metal impurities. The following key conclusions have been drawn from the investigations: 1. The combination of stirring, fluxing, and holding of molten aluminum after refining resulted in achieving a higher electrical conductivity ranging from 0.5% to 1% IACS. This underscores the effectiveness of these refining processes in improving the electrical properties of the aluminum alloy. 2. While the positive impact of stirring on removing transition metal impurities in molten aluminum is evident, the influence of stirring time remains to be determined. The distinctions of how stirring duration affects the purification process require more in-depth exploration for a comprehensive understanding. 3. Using a short stirring time (2 min) introduces a potential challenge. This brief stirring period might provoke re-entrainment of the trapped Ti and V, especially when flux is introduced during the refining process. This highlights the importance of carefully optimizing the stirring parameters to prevent the unintended reintroduction of impurities. 4. The lab-scale trials did not conclusively demonstrate the superiority of the master alloy with AlB2 phase (Al-3% B) over the one with AlB12 phase (Al-8% B) in refining smelter-grade aluminum from transition metal impurities. However, the industrial-scale work presented a different scenario. It showcased a cleaner melt concerning inclusion content when the aluminum-boron master alloys containing AlB2 phases were employed. This underscores the subtle variations in refining efficiency between laboratory-scale and industrial-scale settings. In essence, our study navigates through the intricacies of aluminum refining processes, shedding light on the crucial role of stirring, fluxing, and holding in enhancing electrical conductivity. The disparities observed between laboratory and industrial results emphasize the importance of considering scale-specific dynamics in pursuing optimal aluminum refining practices.

Future Work The present study opens avenues for future research to enhance refining processes in the production of smelter-grade aluminum. Investigating optimal stirring

S. Akhtar et al.

durations, refining fluxing methodologies, and conducting a comprehensive analysis of various master alloy variants will be essential. Scaling up successful industrial trial outcomes, employing advanced characterization techniques, and exploring diverse material properties are crucial for further exploration. Additionally, considering the broader context of aluminum manufacturing and its applications will contribute to refining processes, ensuring improved efficiency and product quality in the industry.

References 1. Shen, W., Hu, A., Liu, S., Hu, H., Al-Mn Alloys for Electrical Applications: A Review, Journal of Alloys and Metallurgical Systems, Vol. 2 (2023) 100008. https://doi.org/10.1016/j.jalmes. 2023.100008. 2. Davis, J.R., Alloying: Understanding the Basics, ASM International (2001) pp. 351–461. 3. Edward, C., Aluminum and its Alloys Used in Electrical Engineering (2012). https://electrical-engineering-portal.com/ aluminum-and-its-alloys-used-in-electrical-engineering (Assessed 16th Aug 2023). 4. Judianto, H., Aluminum Alloys-Applications (2023). https://www. academia.edu/9645530/Aluminumalloys_applications_ 130422231757_phpapp02. (Assessed 6th July 2023). 5. Dube, G., Removal of Impurities From Molten Aluminum, EU Patent (1983), 0112024. 6. Cooper, P.S., Kearns, M.A., Removal of Transition Metal Impurities in Aluminum Melts by Boron Additives, Aluminum Alloys: Their Physical and Mechanical Properties (1996) pp. 1–3, 217, 141–146. 7. Azo Materials, Boron Masters Alloys for High Conductivity Aluminum (2018). https://www.azom.com/article.aspx?ArticleID= 3445. (Assessed 17th June 2023). 8. Khaliq, A., Thermodynamics and Kinetics of Transition Metal Boride Formation in Molten Aluminum, PhD Thesis, Swinburne University of Technology, Melbourne (2013). 9. Karabay, S., Uzman, I., Inoculation of Transition Elements by Addition of AlB2 and lB12 to Detrimental Effect on the Conductivity of 99.6% Aluminum in CCL for Manufacturing of Conductor, Journal of Materials Processing Technology, Vol. 160, No. 2 (2005) pp. 174–182. https://doi.org/10.1016/j.jmatprotec. 2004.06.015. 10. Khaliq, A., Rafiq, M.A., Ali, H.T., Ahmed, F., Mahmood, S., Ranjha, S.A., Analysis of Melt Quality Related Failure of Electrical Conductor Aluminum Wire, Journal of Mining and Metallurgy, Section B Metallurgy, Vol. 53, No. 1 (2017) pp. 75– 81. https://doi.org/10.2298/JMMB151006030K. 11. Khaliq, A., Rhamdhani, M.A., Brooks, G.A., Grandfield, J., Removal of Vanadium from Molten Aluminum—Part I. Analysis of VB2 Formation, Metallurgical and Materials Transaction B, Vol. 45, No. 2 (2013) pp. 752–768. https://link.springer.com/ article/10.1007/s11663-013-9974. 12. Khaliq, A., Rhamdhani, M.A., Brooks, G.A., Grandfield, J., Removal of Vanadium from Molten Aluminum - Part II. Kinetic Analysis and Mechanism of VB2 Formation, Metallurgical and Materials Transaction B, Vol. 45, No. 2 (2014) pp. 769–783. https://doi.org/10.1007/s11663-013-9975-9. 13. Khaliq, A., Rhamdhani, M.A., Brooks, G.A., Grandfield, J., Removal of Vanadium from Molten Aluminum - Part III. Analysis of Industrial Boron Treatment Practice, Metallurgical and

Optimization of Boron Treatment for Production of 1370 Electrically Conductive Grade Aluminum Alloy Materials Transaction B, Vol. 45, No. 2 (2014) pp. 784–794. https://doi.org/10.1007/s11663-013-0017-4 14. Li, X., Cui, X., Liu, H., Zhu, Z., Liu, J., Zhang, X., Cui, H., Li, H., Pan,Y., Feng. R., Man, Q., Study on the Improvement and Mechanism of AA6101 Electrical Conductivity by Trace TM (Zr, V, Ti) Elements-Assisted Boron Treatment, Journal of Alloys and Compounds (2023) 939. https://doi.org/10.1016/j.jallcom.2023. 168728. 15. Cui, X., Wu, Y., Zhang, G., Liu, Y., Liu, X., Study on the Improvement of Electrical Conductivity and Mechanical Properties

951

of Low Alloying Electrical Aluminum Alloys, Composites Part B, Vol. 110 (2017) 381e387. https://doi.org/10.1016/j.compositesb. 2016.11.042. 16. Liu, L., Samuel, H., Assessment of Melt Cleanliness in A356.2 Aluminum Casting Alloy Using the Porous Disc Filtration Apparatus Technique—Part II Inclusion Analysis, Journal of Materials Science, Vol 32 (1997) pp. 5927–5944. 17. Stanica, C., Moldovan, P., Aluminium Melt Cleanliness Performance Evaluation Using PoDFA, UPB Scientific Bulletin, Series B: Chemistry and Materials Science, Vol. 71, No. 4 (2009).

LAlum—Standardization of Launder Systems for Aluminum Casting Michel J. Quintiano and José G. Hernandez

Abstract

This article aims to present the importance of standardizing the profile and sizes of launder systems used in non-ferrous metal foundries. A global standardization of these aspects would bring significant benefits, such as reducing the lead time for replacement of refractory materials by suppliers and substantial financial gains by minimizing the need for manufacturing custom molds for each project or client. Moreover, standardization would enable foundries to conduct comparative studies among the different materials available in the market since the shape of the launder systems would no longer be an influential factor, focusing instead on the compositions of the refractories used in their construction. Another crucial point is that by standardizing the profile of refractory launder systems, this standardization would extend to the entire channel system, including space for thermal insulation, the profile of metal plates, and lid systems. Keywords

Launder system transport




Aluminum casting




Aluminum

Introduction The aluminum market is a cornerstone of the global economy, with a broad spectrum of applications in multiple industrial sectors. After over two decades of experience in this field, it becomes evident that the vast range of profiles of

M. J. Quintiano (&)
J. G. Hernandez Alum Foundry Supplies, Salto, Brasil e-mail: [email protected] J. G. Hernandez e-mail: [email protected]

refractory launders used for the transportation of liquid metal is both impressive and problematic. The multitude of profiles, far from being an advantage, often emerges as a hurdle, since a single profile could suffice for different aluminum flow rates if properly sized. Currently, the sector faces numerous disadvantages due to the lack of standardization of these profiles. It’s not uncommon for clients to use 4–5 different types of profiles for a single launder line, especially when purchasing modules from different global suppliers. This scenario leads to a chaotic environment for foundries and suppliers, particularly when replacing components. The negative impact of this lack of uniformity is evident in several areas: manufacturing lead times extend beyond 30 days, inventories become inconsistent due to a variety of demands, and the budgeting process becomes complex and lengthy, often requiring technical visits for each situation. The standardization of refractory launder profiles would, therefore, emerge as an innovative solution with multifaceted benefits for the aluminum industry. This would allow greater flexibility in operations, simplify the selection and procurement process of components, and eliminate the uncertainties often associated with customized budgets. Moreover, standardization would help optimize energy usage and enhance operational safety. In summary, the adoption of standardized refractory launder profiles can be considered a crucial strategy to bolster the competitiveness of the aluminum market on a global scale, offering benefits ranging from energy savings to the optimization of manufacturing processes.

The Intricate Balance of Thermal Losses In the sophisticated process of pouring aluminum through refractory launders, we encounter a complex thermal dynamic influenced by two critical interfaces. The first, located at the point of contact between the liquid aluminum and the refractory profile of the launder, triggers thermal

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_119

952

LAlum—Standardization of Launder Systems for Aluminum Casting

953

phenomena of undeniable importance. The second, which occurs on the surface exposed to the environment at the top of the launder, represents one of the areas of greatest thermal loss throughout the launder system [1]. This interaction adds an additional level of complexity, making thermal management in refractory launders a fascinating challenge. It is vital to understand that the interaction between the liquid aluminum and the ambient air at the top of the launder stands as the main contributor to thermal dissipation. However, this observation should not overshadow the fact that there is significant thermal loss through direct contact with the refractory launder profile. In light of these insights, there arises an urgent need to explore approaches that mitigate these various forms of loss. However, this exploration must be conducted with astute consideration for operational safety, designing solutions that not only maximize thermal efficiency but also preserve the integrity of the launder structure, taking into account future eventualities of repairs and interventions. Thus, the harmonization between the intricate balance of thermal losses and the imperativeness of operational safety emerges as a central element in the pursuit of optimizing the pouring process and the development of refractory launders.

advantages and disadvantages. Understanding these characteristics is essential (Table 1) in the quest for a standardized profile capable of meeting the needs of different foundries, promoting efficiency and competitiveness throughout the aluminum industry. Each launder profile has peculiarities that can directly influence the pouring process, making a detailed analysis important to arrive at a standardized model of straightforward interpretation and universal application. In a detailed study of existing trough profiles (Fig. 1), we have identified various positive and negative aspects in different profiles and dimensions. Dimension A, often highlighted for its spaciousness, offers significant ease in terms of cleaning and maintenance. This wide accessibility is advantageous, but when excessively open, it can cause significant aluminum temperature losses, which are less desirable compared to other models. On the other hand, its more compact version minimizes thermal loss but restricts access for interventions and repairs, making troubleshooting more challenging. Meanwhile, Dimension B, with moderate depth in various models, consistently favors maintenance. However, when its depth is exaggerated, it becomes a point of concern as it can complicate maintenance and potentially compromise operator safety. Dimension C has varied benefits. In its larger size, it assists in the extraction of the piece from the mold and facilitates cleaning during operation, but when excessively wide, it can cause thickening in Dimension D, increasing energy demand in the region. In more compact scenarios, Dimension C is necessary to ensure efficient demolding in conventional molds; however, its absence may require more elaborate molds and significantly complicate the quality during the refractory manufacturing process. When we analyze Dimension D, we find an ideal balance in models similar to Dimension P, where energy consumption is optimized. Smaller dimensions, on the other hand, require less energy to stabilize temperature, favoring efficiency.

Current Varieties of Launder Profiles In the context of the aluminum industry, various studies have been conducted aiming at optimizing thermal yield and efficiency in the casting process. Some research has already addressed the pursuit of standardized launder system profiles [2] with the goal of improving the performance of transporting molten aluminum. However, it’s important to emphasize that these attempts often focused on creating exclusive standards for specific flow rates of the foundries themselves, limiting the flow and applicability to the market. Next, we will discuss the most commonly found profiles in the launder system market, presenting their potential

Table 1 Profile description

Dimension

Description

A

Open channel area where the aluminum will be in contact with the environment

B

Height of the wetted profile

C

Angle of the internal channel wall

D

Thickness of the wall at the channel’s lower base

E

Internal channel radius

F

Flat bottom of the inner part of the channel

G

External channel wall

P

Thickness of the wall at the channel’s top part

S

Safety height of the metal level

Note All drawings follow the same wetted area of 155 cm2. The dimension P (30 mm), dimension S (20 mm), and the bottom thickness (35 mm) are consistent across all examples

954 Fig. 1 Profiles found in current days

M. J. Quintiano and J. G. Hernandez

LAlum—Standardization of Launder Systems for Aluminum Casting

955

Dimension E, depending on its size, has the potential to simplify maintenance, dissipate stresses, and even promote a homogeneous metal flow. However, insufficient or absent Dimension E can lead to problems such as cracks and maintenance difficulties. Finally, Dimension G, in some iterations, is straight, which can be problematic, especially with rigid thermal insulation systems. In other variations, its angles are beneficial, but steep inclinations can direct unwanted stresses to the center of the refractory.

In order to keep this metric intuitive, we adopted a standard increment of 25 mm (equivalent to 1 inch). This means that the scale of the profile will consistently grow in this relation, meeting the vast majority of current demands and flows. In particular circumstances, we will be equipped to develop profiles with more subtle increments of 12.5 mm (equivalent to 1/2 inch), aiming to address requirements so specific that a full increment wouldn’t capture. However, it’s prudent to associate a value premium for such specifications, since they will be tied to smaller production batches, which can influence production costs. Initially, the prospect of using a single value as the basis for sizing the profile seemed quite attractive, leading us to investigate the principles of the golden ratio. This exploration brought us to the concept of the golden rectangle (Fig. 2), which, at first analysis, presented itself as the ideal option for our profile, ensuring a proportion between the depth of the trough and the minimal exposure of the aluminum to the external environment. However, as we began to draft the initial designs, it became clear that this approach might be the right direction for the standardized profile. However, critical concerns related to operational safety emerged, forcing us to abandon this concept. This was because, during maintenance or repair processes of the refractory modules, operators faced the challenge of dealing with a narrow top opening, making access hazardous and impractical, with a risk of inappropriate exposure to deep areas of the profile. Furthermore, another essential aspect to consider is the final height of the refractory module, which would become excessively elevated. This would render the project implementation unfeasible, as it would require the repositioning of furnace heights relative to pouring points, thus compromising the project’s economic viability for most companies.

Results and Discussion In our research, we came across studies that addressed various profile shapes [3], aiming to guide towards an ideal profile resembling the proposed one. However, these studies, for the most part, limited themselves to comparative effects between the different profiles, without progressing towards the creation of a trough standard that’s easily interpreted and universally applicable for foundries and refractory manufacturers. Given this, the standardization proposal emerges as an opportunity to overcome this limitation and to combine efforts from various sectors of the aluminum industry to develop a refractory trough profile that is versatile and widely accepted. The joint participation of refractory manufacturers and foundries is essential for the standardization to achieve effectiveness and become a comprehensive and accessible solution for the entire industry. When considering the general requirements for transporting liquid aluminum, such as flow rate, temperature, and properties of the refractory materials, the proposal will aim to create a profile that meets the needs of multiple foundries, offering benefits like greater energy efficiency, inventory reduction, and increased operational safety. Collaboration between the different market players is the way forward towards a standardization that benefits the entire aluminum production chain, promoting the sustainable and competitive development of this vital sector of the global economy.

Definition of the Profile Creation Criterion To ensure the acceptability and replication of the profile in the sector, it is essential that the principles guiding its conception are clear and straightforward. In this vein, our initial approach determined that all internal profiles would be molded from a single dimensional reference. This single metric would encompass all the ideal proportions for a standardized profile, adapting to the flow rate required for each specific application. Fig. 2 Golden rectangle

956

M. J. Quintiano and J. G. Hernandez

Considering profiles with a Dimension A of 250 mm, the names can become somewhat lengthy. Therefore, for simplicity, we have chosen the unit “inch” for profile designations, as shown in Table 2. As shown in Table 2, the first part of the name, such as “Profile 3,” indicates the size of the profile. The section referring to the “length” will detail the extent of the profile, a topic we will discuss later. Lastly, the name of the material used in the profile’s manufacture is determined by each refractory manufacturer, as specified in their technical sheets.

Creation of the Standardized Profile

Fig. 3 Hexagon and connected hexagons

After ruling out the possibility of the golden rectangle, we began to seek another geometric shape that could encapsulate our fundamental principle of proportion based on a single dimension to define the internal profile. After careful consideration, we turned our attention to the hexagon (Fig. 3), since, when modifying one of its dimensions, it also adjusts proportionally. However, our enthusiasm for the hexagon was quickly dampened, as the depth of the resulting profile was excessively shallow compared to the required opening. Continuing our analyses, we found an ingenious and elegant solution: the use of two connected hexagons (Fig. 3) to maintain the necessary proportions. This clever connection allowed us to achieve the ideal ratio we were seeking between the profile’s depth and opening, thus satisfactorily addressing the design challenge.

Once the basic framework that will guide us through all proportions of the profile is established, we will address each aspect of the standardized profile’s development, taking into account the pros and cons of the profiles previously studied. Our aim is to find the perfect balance between energy savings, ease of maintenance, safety, and ease of installation.

Internal Profile Regarding the internal profile, it’s essential for it to have an exit angle to ensure smooth demolding during the manufacturing process. This angle also facilitates the cleaning of the module during process stops. It’s important to note that the larger Dimension C is, the more impact it will have on Dimension D; therefore, finding a suitable balance is crucial. After an in-depth market analysis, we identified a 5-degree angle on each face, totaling a 10-degree internal angle (Fig. 4), as highly efficient for the demolding process Table 2 Nomenclature Profile Name

Dimension A (mm)

Profile 2″  length – material

50

* Profile 2.1/2″  length – material

62,5

Nomenclature Definition of the Profiles

Profile 3″  length – material

75

* Profile 3.1/2″  length – material

87,5

Just as crucial as establishing a standardized profile is adopting a universal nomenclature. This measure will provide a unified language, simplifying communication within the sector. Given that Dimension A will be central in our approach, it makes perfect sense for it to serve as the basis for our nomenclature. As previously suggested, Dimension A will increase in standard increments of 25 mm, with exceptions in increments of 12.5 mm.

Profile 4″  length – material

100

Profile 5″  length – material

125

Profile 6″  length – material

150

Profile 7″  length – material

175

Profile 8″  length – material

200

Profile 9″  length – material

225

Profile 10″  length – material

250

…

…

LAlum—Standardization of Launder Systems for Aluminum Casting

957

This radius is adjusted according to the main dimension provided, hence being the result of the tangency between the two internal faces and the two baseline lines of the lower hexagon.

External Walls

Fig. 4 Side faces and internal profile radius

without compromising the module’s cleanliness. With the side faces defined, our attention turned to the bottom of the internal profile, where it’s vital to minimize turbulence generation and ensure ease of cleaning after operation. Over time, we observed that any corner presents a significant challenge, as it becomes a tension accumulation point in the refractory and certainly a spot prone to metal buildup, complicating maintenance and cleaning. Based on these factors, we have decided that the profile should have a single radius (Fig. 4) so that the profile has a smooth, uniform line, minimizing turbulence and allowing the operator to clean with a single tool movement. Table 3 Dimension P e H

Profile name

Determining the external wall thickness of a profile is a critical issue in its creation, as this measure must be capable of supporting and directing large volumes of metal without collapsing. At times, the fear of collapses leads refractory manufacturers to excessively increase this dimension, resulting in the system’s energy inefficiency and elevated costs, due to the significant increase in the volume of material used. However, adopting a single measurement for this thickness, as with the internal profile, would not be a prudent approach, since different profiles will be selected to transport varied volumes and loads of metal. It is sensible, therefore, to seek a balance point or range between energy efficiency and safety, adjusting the external wall thickness according to the needs of each profile. After field analyses, we observed that for every 25 mm increase in Dimension A, Dimension P should be incremented by 4 mm in relation to the previous profile. Additionally, the lower base of the trough should also be sized in relation to Dimension P of the selected profile, with an increment of 5 mm in Dimension H regarding these measures; see Table 3. The standardized profile will start at the 50 mm (2 inch) profile; with this in mind, Dimension P for this profile will be 18 mm. Subsequent profiles will have an increase of 4 mm in the wall as described earlier.

Dimension P (mm)

Dimension H (mm)

Profile 2″

18

18 + 5 = 23

Profile 2.1/2″ *

18 + 2 = 20

20 + 5 = 25

Profile 3″

18 + 4 = 22

22 + 5 = 27

Profile 3.1/2″ *

22 + 2 = 24

24 + 5 = 29

Profile 4″

22 + 4 = 26

26 + 5 = 31

Profile 5″

26 + 4 = 30

30 + 5 = 35

Profile 6″

30 + 4 = 34

34 + 5 = 39

Profile 7″

34 + 4 = 38

38 + 5 = 43

Profile 8″

38 + 4 = 42

42 + 5 = 47

Profile 9″

42 + 4 = 46

46 + 5 = 51

Profile 10″

46 + 4 = 50

50 + 5 = 55

…

…

…

958

M. J. Quintiano and J. G. Hernandez

Fig. 5 External wall and bottom

Fig. 6 External wall with chamfer and radius

Considering the profiles available in the market, it became evident that Dimension G, when built straight or with a steep incline, can cause issues in assembling the refractory module. Based on this finding, Dimension I (Fig. 5) was introduced, which will always be 5 mm, regardless of the profile size chosen. This dimension aims to facilitate assembly when rigid thermal insulators are used, providing a sliding fit of the module with the thermal insulation. Dimension I also plays a pivotal role in minimizing the impact caused by Dimension D, even though the latter is in a controlled situation, since Dimension C is a conservative measure. Despite the introduction of Dimension I, we still faced the challenge of sharp edges which can harm or even damage the thermal insulation, depending on the pressure and force exerted during the assembly process. To prevent this and facilitate guiding the module into the thermal insulation system, Dimensions J and K (Fig. 6) were created. These dimensions minimize the refractory’s sharp edges and ensure that the better-finished parts of the refractory module come into contact with the thermal insulation. The chamfer in Dimension J and the radius in Dimension K also play a crucial role in dissipating the tension of the refractory module. To ensure the integrity of the outer wall and minimize stresses on the refractory module, it is essential to incorporate radius that eliminates sharp edges and points that could initiate cracks. Dimension L (Fig. 6) plays a critical role, especially since this region is subject to friction with

cleaning tools. The radius in this area effectively distributes the stress or force applied. Regarding dimension M, it further contributes to the dissipation of stress in the refractory. However, to achieve this radius, it’s necessary to re-evaluate the model in some aspects, which may result in additional manufacturing costs. It’s important to note that Dimension M is not mandatory; its inclusion is at the discretion of the refractory manufacturer.

Safety Criteria Considering that operational safety is one of the fundamental principles in creating the standardized profile, it is crucial to correctly define this in order to ensure proper leveling of the troughs and the ideal flow rate. Many studies recommend a safety margin of 30%, a suggestion that we endorse as highly beneficial. After detailed analyses, it became evident that the safety margin should be applied to Dimension N. In this way, we consider 30% of the effective channel height as Dimension S (Fig. 7), which will ensure a safe area for operation in profiles of any size. This approach takes into account variations in profile dimensions and ensures operational safety in all situations. Having the safety area well-defined, we find Dimension B (Fig. 7), which represents the wetted area to be considered in flow calculations. This measurement is essential to determine the capacity of the standardized profile to efficiently transport metal during the process.

LAlum—Standardization of Launder Systems for Aluminum Casting

959

In Fig. 9, we can see the 2″, 2.1/2″, and 3″ profiles with their final measurements derived from Dimensions A and P. It’s worth noting that other profile sizes can be created based on the rules stipulated above. After creating the standardized profiles, we can highlight in Table 4 the wetted area of some of the most commonly used profiles for flow rate calculations.

Standard Lengths

Fig. 7 Wetted and safety area

Standardized Profile Having evaluated and considered all points, we have arrived at the standardized profile (Fig. 8) and completed its self-dimensional concept. The smallest system presented in this standardization proposal will be the 2″ profile. Through it, we can develop all the other profiles by only changing Dimensions A of the hexagon and P of the wall thickness.

Fig. 8 Standardized profile for calculations

Now that we have a standardized profile, it’s possible to introduce the concept of standardized lengths. These lengths are suggested to be kept in stock by refractory manufacturers (Table 5). The concept involves starting with a base length of 500 mm and adding increments of 250 mm. Fractional dimensions can be considered based on the needs of the specific project; however, we recommend these increments to be at least 10 mm, as adjustments can be made at the module junctions. Avoid fractional lengths, such as 1123 mm or 1127 mm, always aiming for whole dimensions with 10 mm increments, like 1120 mm or 1130 mm. This aids in efficient stock management and minimizes costs across the supply chain.

Conclusion This article introduced a systematic methodology for standardizing profiles used in the metal transport industry. We highlighted the imperative need for standardization and articulated a robust model, backed by stringent criteria, to address this issue. The insertion of a deliberate safety margin in Dimension N not only enhances operational safety but also facilitates a more precise identification of the area of Dimension B, which, in turn, optimizes transport efficiency. By utilizing a base profile, we achieved significant versatility, satisfying a broad range of industrial applications, which we believe can cover up to 95% of cases. Additionally, the standardized length system simplifies implementation, reducing both the cost and time required for production. It’s important to note that standardization is not merely a technical exercise. It plays a critical role in the evolution of other standardized products and systems, including cleaning tools, gate systems, and covers. Thus, standardization offers a myriad of advantages, both technical and economic. It simplifies not only the production and inventory management process but also promotes enhanced operational efficiency.

960

M. J. Quintiano and J. G. Hernandez

Fig. 9 Standardized profiles of 2″, 2.1/2″, and 3″

Our preliminary field tests have already indicated significant gains, including a temperature reduction of about 30 degrees in the furnaces after the installation of the standardized profile. It’s worth noting that these data are still in

the collection and analysis phase, and we plan to detail them in future publications. In summary, this article not only offers a viable solution to a pressing issue but also serves as a solid foundation for

LAlum—Standardization of Launder Systems for Aluminum Casting Table 4 Wetted area for flow calculation

Table 5 Nomenclature with length and material

961

Profile name

Wetted area (70% N) cm2

Total area (N) cm2

Profile 2″

19,18

28,82

Profile 3″

29,38

45,04

Profile 4″

42,45

64,86

Profile 5″

57,91

88,28

Profile 6″

75,78

115,31

Profile 7″

117,53

180,16

Profile 8″

169,80

259,44

Profile 9″

231,66

353,12

Profile 10″

301,25

461,22

…

…

…

Profile name

Length (mm)

Straight trough profile 2″  500 mm—Castceram

®

HR

500

Straight trough profile 2″  750 mm—Castceram

®

HR

750

Straight trough profile 2″  1000 mm—Castceram

®

HR

1000

Straight trough profile 2″  1250 mm—Castceram

®

HR

1250

Straight trough profile 2″  1500 mm—Castceram

®

HR

1500

Straight trough profile 2″  1750 mm—Castceram

®

HR

1750

Straight trough profile 2″  2000 mm—Castceram

®

HR

2000

…

future studies in the area. It is hoped that this work will act as a catalyst for the continued optimization of metal transport systems, thus contributing to ever-higher standards of safety and efficacy in related industrial sectors. Acknowledgements We would like to express our deepest gratitude to the Alum Foundry Supplies team. Their invaluable support and the opportunity provided to us were essential for the completion of this work. Special thanks go out to the pillars of our lives: our mothers, fathers, wives, and children. The constant support and love from each of you were the foundation that sustained us throughout this journey. Every word of this work was driven by your faith and encouragement in us. From the bottom of our hearts, thank you very much!

…

References 1. J. Klesch, M. Eng EIT, and G. Product Manager, “Decreasing Thermal Gradients via VDC Tabletop Refractory Design,” Nashville, Oct. 2017. [Online]. Available: www.aluminum-us.com. 2. M. Canullo, F. Daroqui, J. Ottaviani, M. Martín, and R. Acuna, “Launder System for Aluminium Casting,” Materials Science Forum - MATER SCI FORUM, vol. 630, pp. 119–125, Oct. 2009, https://doi.org/10.4028/www.scientific.net/MSF.630.119. 3. J. O. Marthinusen and S. F. Ray, “insural insulating materials, launder design and the use of tempcal thermal modelling,” Second International Workshop on Aluminium Melt Process Technologies in Dublin, Oct. 1998.

Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration Are Bergin, Robert Fritzsch, Shahid Akhtar, Lars Arnberg, and Ragnhild E. Aune

Abstract

Introduction

Filtration is a widely used method for removing non-metallic inclusions/bifilms from molten aluminium, where Ceramic Foam Filters (CFFs) dominate due to their cost-effectiveness and easy use. While many CFFs are phosphate-bonded and have been considered inert when in contact with molten aluminium, recent studies have shown that this is not the case, making it necessary to study even non-phosphate-bonded CFFs. Through aluminium melt pilot-scale filtration trials using a 6082 alloy and commercial CFFs based on sintered alumina (Al2O3) and silica (SiO2), the present study reveals that even these filters can react with the molten aluminium. Microscopic characterization of unused/used CFFs confirmed two notable observations: (i) indications of silicon depletion in the used filters and (ii) minimal reactivity. Based on the microscopic evaluation and LiMCA data, the silicon depletion and the conditions leading to silicon depletion are discussed in view of its effect on filtration efficiency. Keywords











Ceramic Foam Filters (CFFs) Filter degradation Filtration conditions Filtration efficiency LiMCA

A. Bergin (&)
S. Akhtar Hydro Aluminium AS, Commercial Technology, Romsdalsvegen 1, 6600 Sunndalsøra, Norway e-mail: [email protected] R. Fritzsch
L. Arnberg
R. E. Aune Department of Materials Science and Engineering, Norwegian University of Science and Technology (NTNU), Alfred Getz Vei 2, 7034 Trondheim, Norway

During the aluminium production process, minimizing the presence of non-metallic inclusions is crucial to avoid defects in the final product which will compromise mechanical strength, among others. Filtration is employed to remove these inclusions, and ceramic foam filters (CFFs) are a commonly used solution for this purpose. The traditional CFFs are mainly composed of ceramics like alumina (Al2O3), silicon carbide (SiC), and silica (SiO2), and were believed to be inert in contact with molten aluminium dating back to their introduction in the 1970s [1]. However, both industrial experience and recent studies [2–4] have challenged this assumption. Traditionally, most CFFs have been produced using a phosphate binder as described by an early patent [1], which can react with magnesium in the aluminium melt, even at low concentrations, potentially affecting the filter integrity and HSE1 considerations in the aluminium casthouse as well as leading to potential changes in the melt composition [4]. Moreover, under certain process conditions, it is thermodynamically possible that CFFs without phosphate binder and comprising silicon carbide, silica, or alumina-silica can react with the molten aluminium. For example, it has been reported that for an aluminium alloy containing less than *10 wt.-% silicon at 700 °C, silicon carbide can react with the molten aluminium, forming aluminium carbide and dissolving silicon into the melt [5]. Similarly, both silica- and alumina-silica refractories are susceptible to reduction by aluminium and magnesium, causing silicon to dissolve in the aluminium melt [6]. Nevertheless, there are currently no reports of molten aluminium reacting with CFFs without phosphate binders. Trials to investigate this were conducted by Voigt et al. [7] using a short-term filtration setup simulating foundry conditions. These filtration trials tested the chemical stability of CFFs with sintered surface compositions of SiO2 and 1

Health, Safety, and Environment.

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_120

962

Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration

963

3Al2O3•2SiO2 (mullite), among others, using about 2.3 kg of an AlSi7Mg alloy. The obtained results revealed no significant change in the melt chemistry, suggesting that the CFFs did not react with the aluminum melt. Bao et al. [8] and Syvertsen et al. [9] reported pilot-scale filtration tests with Al2O3- and SiC-based CFFs, where the latter filters comprised both silica and silicon carbide in significant amounts. Bao et al. [8] analyzed the melt composition (through disk samples for spectrographic analysis) and the number of aluminium carbide particles before and after filtration without observing any indications of CFFs reacting with the aluminium melt. However, both [8, 9] studies noted that the filter removal efficiency decreased over time for SiC filters in contrast to Al2O3 filters, suggesting the thermodynamically unstable nature of SiC in aluminium melt as the explanation, though without definitive evidence. The current study was conducted as an investigation into the stability of commercially produced CFFs made from sintered alumina and silica for aluminium production. Filters of the same type were tested twice under identical conditions at a pilot-scale filtration setup.

Experimental Procedures and Materials The current study was conducted at Hydro Aluminium Metal’s pilot-scale R&D facility in Sunndal, Norway, as part of a series of trials, some of which were previously presented elsewhere [10]. Although there is significant overlap in experimental details between the current study and the previous paper [10], a thorough description of the experimental procedures is provided here as well. The R&D facility features a loop-shaped arrangement of launder sections, as shown in Fig. 1, enabling continuous circulation of the melt without the need for casting or drainage. The melt circulation is achieved using a mechanical metal pump without a filter to prevent inclusion removal with the pump. The furnace was filled with approximately 8 tons of an aluminium 6082 alloy, with compositions given in Table 1. The melts were prepared with non-fluxed pot room metal and the necessary alloying elements to produce the specified alloy. As part of a larger test campaign, the current study comprised two tests, each conducted with different melts on different days, and with slightly varying conditions. Table 2 summarizes the main difference between the two tests. A new melt was prepared every day, and two consecutive trials were performed daily. The two trials included in the current study were conducted on separate days, i.e. the first trial of one day (Trial A) and the second trial of another day (Trial B) as seen in the table. Between the first and the second trials each day, there was a delay of approximately 2

Fig. 1 A schematic of the filtration test loop with the measurementand addition positions, DFF® filter box equipped with a ceramic adapter, and the mechanical metal pump. This configuration is adapted from reference [10]

h. This delay may have affected the magnesium concentration of the melt due to oxidation. Total trial duration was nearly equal for both tests, with only a minor 6 min difference. Temperature measurements at priming (the step of enabling the metal to flow through the filter) were performed with a laser and a K-type thermocouple for the melt and filters, respectively, and while the melt temperatures were similar, there was a significant difference for the filter temperatures at 105 °C higher for Trial A. The number and type of non-metallic inclusions is important for the filtration process and the filter removal efficiency, and therefore compacted aluminum saw chips were added to the melt during both trials at a fixed spot, as an attempt to have comparable inclusion concentrations entering the filter. The location for saw chip additions is shown in Fig. 1. The additions of 4 kg of saw chips were performed every 10 min starting approximately 5 min after the trial began and were consistent for both trials. Additionally, for Trial B, grain refiner AlTi5B1 from Aleastur, Spain, was continuously added after 43 min into the trial until its completion, at the same location as the saw chips. The grain refiner was added at a rate of approximately 1.5 kg/ton of aluminium, considering the calculated melt flow rate.

964

A. Bergin et al.

Table 1 The chemical composition of the aluminium melts for the two trials (determined using spark optical emission spectroscopy (SOES) with an ARL 4460 instrument from Thermo Fischer Scientific, USA) presented in weight percent (wt.-%) and of the main alloying elements only. Sampling was conducted early in the morning Elements

Si

Fe

Mn

Mg

Al

Melt 1

1.02

0.21

0.66

0.91

97.18

Melt 2

1.07

0.20

0.65

0.90

97.17

Table 2 Experimental details and conditions for the two performed trials Trial

Trial point in time

A

Early morning

B

Late day

Trial duration [minutes]

Melt

Melt temperature at priming [°C]

Filter temperature at priming [°C]

Addition chips

Addition grain refiner

*75

Melt 1

745

605

4 kg every 10 min

N/a

*82

Melt 2

738

500

4 kg every 10 min

*1.5 kg AlTi5B1/ton Al

The melt flow rate for Trial B was estimated at 6.6 tons/h using a method verified experimentally at the Hydro R&D facility. The method includes a refractory insert dam, with a drilled opening of known diameter, and lasers for continuous measurement of the pressure drop over the dam. Unfortunately, this measurement was not performed for Trial A, but no indications of a significant difference in melt flow rate was observed between the two trials. The filtration tests were performed using a DFF® (Drain Free Filtration) filter box with a solid ceramic filter adapter, as shown in Fig. 1. The filter box has been described in detail elsewhere [11–13], and in short it deviates from a regular filter box by performing the priming from underneath using underpressure, and that there is no need to drain the melt after its use. The solid ceramic adapter accommodated four equal CFFs in each corner, reducing the total filtration area compared to a standard-sized filter of, for instance, 23-inch (584  584 mm2). The filters used in both trials (in total 8 filters—two trials and four filters per trial) were trapeziums with a size of 178  178 mm2 at the inlet and 150  150 mm2 at the outlet, and a thickness of 50 mm. These commercial grade/ppi 30 filters were composed of alumina and silica sintered together in a weight ratio of approximately 5/1, with small amounts of magnesium and no binder added, as communicated by the filter producer. The priming of the filters was considered successful based on observations (through inspection windows at the DFF® filter box, the aluminium melt waves from each corner and where they met could be seen) and that the weight of the spent filters containing aluminium was similar. Two liquid metal cleanliness analyzers (LiMCA II (ABB Ldt., Canada)) were used to continuously measure the melt inclusion content, one positioned before and one after the filter box, as indicated in Fig. 1. The average removal efficiency, based on the LiMCA N20 values (number of inclusions >20 lm), was calculated as the average of the

calculated removal efficiency at every measuring point. The melt flow rate was considered for each data point, ensuring that more or less the same volume of melt was measured both before and after filtration (the molten metal used approximately 4 min from Position 1 to 2; see Fig. 1). For further analysis, both an unused filter and spent filters were sectioned and visually inspected. Samples measuring 25  30 mm2 were retrieved from the top section of a spent filter from each trial and an unused filter. Light optical microscopy (LOM) (Axio Vert.A1, Zeiss, Germany) connected to the ZEN core software (Zeiss) and a scanning electron microscope (SEM) (Ultra 55 LE, Zeiss, Germany) equipped with an XFlash 4010 energy-dispersive X-ray microanalyzer (EDS) detector (Bruker AXS, Germany) were used for analysis. The samples were embedded in epoxy and polished using standard metallographic procedures before analysis.

Results and Discussion In Fig. 2, the SEM analysis and EDS silicon mapping of three distinct filter struts (CFFs are a web of solid so-called filter struts) from a new, unused filter are presented. Interestingly, the silicon distribution within the filter was found to be non-uniform. The silicon EDS signal, which correlates to a higher silicon content with a stronger signal, was more pronounced in specific areas of the filter. This observation is especially clear for the higher magnification micrographs, represented by the two lower EDS micrographs in Fig. 2. However, despite these variations, there was no visible systematic pattern in view of the distribution of high and low silicon content. Additionally, upon closer examination of both the SEM– and EDS micrographs at higher magnifications, the regions with lower silicon content appeared to have a different structure from that of the areas with higher

Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration

965

Fig. 2 SEM micrographs and EDS silicon mappings of filter struts retrieved from a new filter which was embedded in epoxy. The analysis was conducted using an accelerating voltage of 10 kV

silicon content, indicating more porosity. Furthermore, the areas with different silicon content appeared to be organized and well-defined from each other. Despite the heterogeneous nature of the silicon content, silicon was observed to be present throughout the entire filter. The examination of the spent filters using LOM revealed the presence of partly discolored filters, as shown in Fig. 3. This discoloration is similar to the visual appearance previously observed for reacted phosphate-bonded CFFs [4]. In Fig. 3, it can be observed that certain areas still exhibit a color similar to that of the new filters, appearing bright white. These areas are believed to be the regions that have

not reacted with the molten metal. Equally, other regions displayed a darker color, indicating that these regions have likely undergone a reaction with the molten metal. Figure 4 provides a more detailed view of a similar filter strut. The leftmost micrograph is from LOM, just like in the previous figure, while the two subsequent micrographs show the same filter strut in SEM and EDS, respectively. The figure reveals a clear correlation between the silicon concentration of the filter and its dark color. Specifically, the EDS analysis barely detected any silicon in the area where the filter appeared dark in color. This correlation is particularly evident at the interface between the light and dark

966

A. Bergin et al.

Fig. 3 LOM micrographs presenting two examples of partly reacted filter struts obtained from a spent filter. The imaging was performed using polarized light and an  5 objective lens

Fig. 4 Analysis of a partly reacted filter strut retrieved from a spent filter from Trial A, employing LOM, SEM, and EDS silicon mapping. The LOM micrograph was acquired using polarized light and an  5

objective lens, while the SEM and EDS analyses were conducted with a 10 kV accelerating voltage

regions, as shown in the two bottom micrographs of Fig. 4, where the side that appeared dark in LOM appeared to be almost black in the silicon elemental analysis from EDS (but still with some Si present). The behavior of other elements (oxygen, magnesium, aluminium, phosphorus, calcium) was also examined. Although there were indications of a slightly higher aluminium concentration in the dark regions, neither of the elements showed any significant difference between the bright and dark regions. When comparing the micrographs of the new filter in Fig. 2 and the spent filter in Fig. 4, with regards to silicon, it becomes apparent that the intensity of the silicon signal in

the dark region of the spent filter was significantly weaker, in comparison to the low-intensity silicon signal observed in the new filter. These observations strongly suggest that the dark color of the filter, both visually and in LOM, is a result of the filters reacting with the aluminium melt, leading to silicon depletion from the filter. Further, supporting evidence for the hypothesis of silicon depletion from the filter during filtration can be observed from the SEM– and EDS analysis of a random filter strut, as shown in Fig. 5. The dark areas in the micrographs, which indicate low silicon content, can be seen to penetrate inwards into the filter strut in an irregular shape, similar to the

Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration

967

Fig. 5 SEM micrograph and EDS silicon mapping of a random filter strut obtained from a spent filter, utilizing an accelerating voltage of 10 kV

cross-section of localized corrosion in metals. When comparing the current micrographs of the spent filter with the corresponding micrographs of the new filter in Fig. 2, these irregularly shaped regions depleted of silicon are notably different from the regular and well-defined low silicon regions found in the new filter. Unfortunately, the current study did not investigate potential changes in the melt composition resulting from silicon depletion. An interesting aspect of the silicon depletion phenomenon was the observed variation in the reacted regions, both within the same filter and between filters, where the distribution of reacted and unreacted regions appeared to be random. To quantify the fraction of reacted area, 20 filter struts per filter were manually examined using LOM and the average reacted area calculated from these measurements. In the case of the filter from Trial A, the analysis revealed a significant variation in the reacted area between filter struts. Some struts showed minimal reaction, with only 15% of their area affected, while others were almost completely reacted. Moreover, when comparing the filters from Trial A and Trial B, it was established that approximately 58% of the total filter area had reacted in Trial A, while the filter from Trial B showed minimal change, with only about 3% of its total area reacted. It is worth noting that, unfortunately, the detailed analysis was not extended to the three remaining filters per trial, apart from them being cut in two for visual inspection. However, there were no evident differences observed between each filter per trial, except for the already mentioned variation in reacted area. The significant difference in the extent of reacted area between Trial A and Trial B raises two important questions: Firstly, how will this silicon depletion affect the filtration performance and filtration efficiency? And secondly, what could be the underlying cause for the considerable variation in filter reaction between seemingly similar trials?

Addressing these questions is crucial to gain a deeper understanding of the impact of silicon depletion on the filtration process and to potentially optimize the performance of CFFs. In terms of the first question, the LiMCA data were analyzed, and the results, along with the calculated removal efficiencies, are presented in Fig. 6, with additional details summarized in Table 3. Each trial is represented by a graph showing the LiMCA N20 values before and after the filter box, as well as the calculated removal efficiencies (since no degasser was used during the trials, the N20 is not affected by microbubbles). In Trial B, a noticeable change in the inclusion count and removal efficiency can be observed towards the end of the trial (approximately after 60 min) due to the addition of a grain refiner rod. However, this particular section is not included in Table 3, as it is not directly relevant for evaluating the correlation between silicon depletion and the filtration performance. Comparing the incoming inclusion count for the two trials, there was a slight difference. In Trial A, the inclusion count remained stable (or with a slight linear decrease) throughout the entire casting duration, whereas in Trial B there was an initial decrease in the first *10 min before the count stabilized, similar to Trial A. From Table 3, it is evident that the average incoming inclusion count was relatively similar for both trials, particularly during the stable regions (the entire trial duration for Trial A). As a result, the filter removal efficiencies between the two trials should be comparable. For Trial A, the removal efficiency showed a general stable trend with a slight linear increase. In contrast, Trial B exhibited relatively stable removal efficiency throughout the trial (again, excluding the development after grain refiner additions). However, Trial A also showed relatively high numbers and sizes of drops in the removal efficiency, as

968

A. Bergin et al.

Fig. 6 Graphs showing the LiMCA N20 values for both trials, displaying the inclusion counts before and after the filter box, along with the calculated removal efficiencies. It is important to note that Trial B involved the addition of grain refiner rod towards the end of the trial, which accounts for the observed changes in LiMCA-counts and removal efficiency

indicated by the standard deviations in Table 3 in addition to the graph. Upon comparing the overall filtration performance, the average removal efficiencies for the two trials were significantly different, with 63.1% for Trial A and 89.9% for Trial B. As the filters from Trial A suffered from silicon depletion, these LiMCA-results suggest that the silicon depletion negatively affects the filtration performance. However, further data would be necessary to draw definitive conclusions and fully understand the implications of silicon depletion on the overall filtration process.

The reason behind the observed difference, where the same filter exhibited significant silicon depletion or almost none despite identical conditions, remains uncertain. While the current results are derived from pilot-scale trials, achieving precise control over all process parameters can be challenging compared to laboratory-scale experiments. Nevertheless, the pilot-scale setup has demonstrated reproducibility across several trials conducted over the years [8, 9, 12, 14, 15]. The only major differences in process conditions between the current two trials was that Trial B involved the addition of a grain refiner towards the end of the trial, and that the filter preheating temperature at priming was 105 °C higher for Trial A. However, it is unlikely that these factors had any significant effect on the observed differences of the filters. The grain refiner addition occurred late in the trial, reducing its potential impact and it is unlikely that the grain refiner could have had any repairing effect. A temperature of 605 °C should not have an effect on the sintered ceramic material, and also the temperature would quickly equilibrate with the aluminium temperature upon priming—equally for both trials. On the other hand, variations in flow patterns and the presence of aluminium-oxide layers within the filters could offer reasonable explanations. In a CFF during use, the flow velocity and the impingement angle of the aluminium melt are expected to vary, making certain filter areas more susceptible to silicon depletion due to higher flows and direct melt impingement. All CFFs are different due to the nature of their production, and specific filters or areas within a single filter may be more “reactive” in this regard. Regarding the potential presence of aluminium-oxide layers, it is commonly assumed that there are no oxide layers present at the CFF surface during filtration. However, the filters are initially filled with air before priming, including the microporosity at the CFF surface, which could facilitate the formation of aluminum-oxide due to the low required oxygen partial pressure for formation (1.01325
10–44 Pa at 700 °C [16]). In other words, the presence of aluminiumoxide layers is not unlikely. The potential variation in this oxide layer presence from test to test, influenced by flow patterns and other factors, might contribute to the observed differences in silicon depletion. If this explanation proves to hold, it implies that the presence or absence of an

Table 3 Summarized details from the LiMCA-graphs in Fig. 6, including the average LiMCA-count before the filter box and the calculated average removal efficiencies along with their respective standard deviations. All data is based on measurements taken before the addition of grain refiner, which was exclusively applied in Trial B Trial

Average LiMCA N20 before CFF [K/kg]

Average LiMCA N20 before CFF—at a stable incoming amount [K/kg]

Average removal efficiency [%]

A

3.6 ± 0.8

3.6 ± 0.8

63.1 ± 12.5

B

5.2 ± 1.6

4.6 ± 0.8

89.9 ± 5.1

Silicon Depletion in Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration

aluminium-oxide layer inside a CFF during use can significantly impact the actual filtration performance of the used filter.

4.

Conclusions The present study investigated the stability of alumina and silica sintered CFFs without binder through two similar filtration trials using a pilot-scale setup with a 6082 alloy. It was discovered that the filters reacted with the aluminium melt, resulting in silicon depletion from the filters. Notably, there was a significant difference in the extent of reacted area between the two filtration trials, with approximately 58 and 3% reacted, respectively. This difference in behavior potentially had an impact on the filtration performance. A discussion was presented to understand the reasons behind the varying behavior of the filters in the two trials, but a definitive conclusion could not be reached. Further research in this area is essential, not only to comprehend the silicon depletion process but also to gain a better understanding of the filtration mechanism of CFFs. Investigating the actual conditions within a CFF during filtration could shed light on how these filters function, leading to potential improvements in filtration efficiency and performance. This line of inquiry presents an interesting opportunity for future studies to enhance the knowledge of CFF behavior and its implications for industrial applications.

5.

6. 7.

8.

9.

10.

11.

12. Acknowledgements The authors express their sincere gratitude to the Research Council of Norway (NFR project nr: 284090) for providing part of the financial support that made the present research possible. Additionally, the authors would like to extend their appreciation to Norsk Hydro ASA for their valuable assistance and collaboration throughout the study. Special thanks are also due to the helpful personnel at the Hydro R&D facility in Sunndal and the Department of Materials Science and Engineering at the Norwegian University of Science and Technology (NTNU) for their contributions and support during the course of this work.

References 1. J. W. Brockmeyer (1982) Ceramic Foam Filter, US. Patent 4, 343,704. 2. L.S. Aubrey, R. Olson and D.D. Smith (2009), Development of a Phosphate-Free Reticulated Foam Filter Material for Aluminium Cast Houses, Materials Science Forum 630, pp. 137–146. https:// doi.org/10.4028/www.scientific.net/MSF.630.137. 3. C.K.W. Solem, R. Fritzsch and R.E. Aune (2018), Preliminary Experimental Study of the Thermal Stability and Chemical

13.

14.

15.

16.

969

Reactivity of the Phosphate-Based Binder Used in Al2O3-Based Ceramic Foam Filters (CFFs), Extraction 2018. https://doi.org/ 10.1007/978-3-319-95022-8_94. A. Bergin, C. Voigt, R. Fritzsch, S. Akhtar, L. Arnberg, C.G. Aneziris and R.E. Aune (2021), Experimental Study on the Chemical Stability of Phosphate-Bonded Al2O3-Based Ceramic Foam Filters (CFFs), Metallurgical and Materials Transactions B 52, pp. 2008–2025. https://doi.org/10.1007/s11663-021-02144-3. D.J. Lloyd (1989), The Solidification Microstructure of Particulate Reinforced Aluminium/SiC Composites, Composites Science and Technology 35, pp. 159–179. https://doi.org/10.1016/0266-3538 (89)90093-6. A. Yurkov (2015) Refractories for Aluminium: Electrolysis and the Cast House, Springer International Publishing, Switzerland. C. Voigt, B. Fankhänel, E. Jäckel, C.G. Aneziris, M. Stelter and J. Hubálková (2015), Effect of the Filter Surface Chemistry on the Filtration of Aluminum, Metallurgical and Materials Transactions B 46, pp. 1066–1072. https://doi.org/10.1007/s11663-0140232-7. S. Bao, M. Syvertsen, A. Nordmark, A. Kvithyld, T. Engh and M. Tangstad (2022), Plant Scale Investigation of Liquid Aluminium Filtration by Al2O3 and SiC Ceramic Foam Filters, Light Metals 2013. https://doi.org/10.1007/978-3-319-65136-1_166. M. Syvertsen and S. Bao (2015), Performance Evaluation of Two Different Industrial Foam Filters with LiMCA II Data, Metallurgical and Materials Transactions B 46, pp. 1058–1065. https://doi. org/10.1007/s11663-014-0251-4. A. Bergin, C. Voigt, R. Fritzsch, S. Akhtar, L. Arnberg, C.G. Aneziris and R.E. Aune (2022), Performance of Regular and Modified Ceramic Foam Filters (CFFs) during Aluminium Melt Filtration in a Pilot Scale Setup, Light Metals 2022, pp. 640–647. https://doi.org/10.1007/978-3-030-92529-1_84. U. Tundal, I. Steen, T. Haugen and J.O. Fagerlie (2016), Apparatus and Method for the Removal of Unwanted Inclusions from Metal Melts, Patent no. WO 2016/126165 A1. U. Tundal, I. Steen, Å. Strømsvåg, T. Haugen, J.O. Fagerlie and A. Håkonsen (2019), Drain Free Filtration (DFF) – A New CFF Technology, Light Metals 2019, pp. 1097–1104. https://doi.org/ 10.1007/978-3-030-05864-7_134. S. Bao, M. Syvertsen, F. Syvertsen, B.E. Gihleengen, U. Tundal and T. Pettersen (2019), Laboratory Scale Study of Reverse Priming in Aluminium Filtration, Light Metals 2019, pp. 1105– 1111. https://doi.org/10.1007/978-3-030-05864-7_135. A. Kvithyld, M. Syvertsen, S. Bao, U.A. Eriksen, I. Johansen, E. Gundersen, S. Akhtar, T. Haugen and B.E. Gihleengen (2019), Aluminium Filtration by Bonded Particle Filters, Light Metals 2019, pp. 1081–1088. https://doi.org/10.1007/978-3-030-058647_132. M. Syvertsen, I. Johansen, A. Kvithyld, S. Bao, U.A. Eriksen, B.E. Gihleengen, S. Akhtar, A. Bergin and A. Johansson (2020), Evaluation of CFF and BPF in Pilot Scale Filtration Tests, Light Metals 2020, pp. 963–971. https://doi.org/10.1007/978-3-03036408-3_130. S. Bao, M. Syvertsen, A. Kvithyld and T.A. Engh (2014), Wetting Behavior of Aluminium and Filtration with Al2O3 and SiC Ceramic Foam Filters, Transactions of Nonferrous Metals Society of China 24, pp. 3922 – 3928. https://doi.org/10.1016/S1003-6326 (14)63552-4.

A PoDFA Benchmarking Study Between Manual and AI-supervised Machine Learning Methods to Evaluate Inclusions in Wrought and Foundry Aluminum Alloys Pascal Gauthier, Vincent Bilodeau, and John Sosa

Abstract

Keywords

The PoDFA inclusion measurement is achieved by identifying the inclusions and their concentration in the melt for each type with a trained operator. The standard technique is realized by using a square grid with an optical microscope to count the total area with each detected square. This manual and non-efficient method requires a lot of time and effort and can generate important variations in PoDFA results for reproducibility and repeatability. In the past, there were many unsuccessful attempts to automatically detect, count, and classify all inclusion types due to the complexity of the application. Disc sampling, image artifacts, polishing defects, and metallurgical constituents are some examples that can interfere with the inclusion detection and the measurement methodology. Commercial image analysis systems with threshold options and Boolean logical operations are not sufficient to automate the solution. The implementation of artificial intelligence technologies such as supervised machine learning algorithms are necessary to automate this complex method. The benchmarking study was achieved between the standard PoDFA methodology compared to the artificial intelligent way. Results show that the new technique exhibits a good correlation and a high potential for industrial use.

Cleanliness PoDFA Method Inclusions Assessment Automated Artificial intelligence AI Machine learning Wrought Foundry Aluminum Alloys

P. Gauthier (&)
V. Bilodeau Rio Tinto Aluminum, Arvida Research and Development Center, Saguenay, QC G7S 4K8, Canada e-mail: [email protected] J. Sosa MIPAR Image Analysis, 8050 N High St, Ste 170, Columbus, OH 43235, USA















Introduction Inclusions and oxide films have a critical impact on the mechanical and corrosion properties for the final downstream aluminum product [1, 2]. These non-desired particles need to be minimized or to be suppressed during the casting process. Today, there are several methods available for process evaluation and the determination of inclusion levels from one step to another. For example, the K-Mold, the Prefill, the LIMCA, and the PoDFA techniques are available [3–7]. Each technique has pros and cons but the PoDFA is still the industry reference and a good tool for identifying the major types of inclusions. It can effectively assess the effects of various operating practices and process treatments on metal cleanliness [6]. The PoDFA analysis method exists for many decades to evaluate the molten aluminum cleanliness for the Aluminum industry. The term PoDFA is the abbreviation of “Porous Disc Filtration Apparatus”. The technique consists of filtrating a molten metal sample through a porous refractory disc (Fig. 1). The inclusions and other foreign particles present in the metal are collected on the surface of the refractory disc surface. Observation by optical microscopy of the metallographic cross-section of the disc allows the identification of different types of inclusions, expressed as total concentrate of inclusions per kilogram of metal filtered (mm2/kg), giving an indication of the metal cleanliness. The GRID method shall be used when you need accuracy and when you evaluate clean metal (for inclusion

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_121

970

A PoDFA Benchmarking Study Between Manual and AI-supervised Machine …

Fig. 1 Metallographic mounting of PoDFA samples

concentrate 1.00

2

On the Importance of Measurement and Process Uncertainty in Certifying the Quality of Aluminium-Based Products

hardness, either Vickers (HV) or Brinell (HB). In other words, the customer gives us their requirements for the mechanical properties in the form of a semi-open interval, which is limited by the minimum required value for the selected property. In the case of semi-finished products made of aluminum alloys that are heat treated, the customer prescribes the type of heat treatment, after which it is necessary to measure the mechanical properties. Even in this case, the customer requirements do not prescribe or do not mention the measurement uncertainty. In most cases, this does not require the use of a larger or an exact number of rods. However, there are also customers who explicitly request that the measurement of mechanical properties be performed on a prescribed number of rods and that all the measured values fall within their requirements. The combined measurement uncertainty of our measurements, as a combination of all the components of uncertainty, including sampling uncertainty, is also in this case the one that determines the width of the guard band at the limits of the (semi-open) interval of the customer requirements, or, in the case of the closed interval, at both limits. For this reason, it is crucial to ensure the credible certification of the mechanical properties. An assessment of the mechanical properties is performed on the basis of destructive methods. So, in this case, as well as determining the chemical composition, it is necessary to define and validate the methodology and know the degree of uncertainty for representative sampling. The uncertainty of the sampling of the test rods for tensile testing is determined in a very similar way, where the pieces are sampled according to the established procedure for N (i.e., 25) tubes at the same location of the product (rod, profile or tube) and then the heat treatment is carried out very carefully (if prescribed) and machining of the test specimens to the prescribed geometry as well as the tensile test itself. Also in that case, the uncertainty of the preparation of the test rods for tensile testing is determined by sampling with special care, and the heat treatment (if prescribed), machining, and tensile testing are performed according to the established procedure.

1005

Customer requirements for the macro- and microstructure of semi-finished products can vary, but mostly relate to the width of the peripheral zone, the grain size and grain fraction, and the size of the inclusions. Without giving details of individual requirements, we can say that in quantitative macro- and microstructural investigations, the reliability of the results depends largely on the representativeness of the examined part of the sample. The uncertainty of the sampling or the degree of the representativeness of the examined part of the sample affects the reliability of the measured values much more than the measurement uncertainty of the measuring instrument, i.e., the microscope. Therefore, the greatest possible attention should be paid to determining this uncertainty, as it directly affects the reliability of the metallographic measurements recorded on the certificate. The uncertainty of the sampling is determined by sampling according to the established procedure, and all the other operations (machining, grinding, polishing, etching, and metallographic inspection) are performed as carefully and repeatably as possible. In this case, too, it is necessary to define the width of the guard band and move away in relation to the limits of the customer requirements for its value. However, this, as we have already said, depends primarily on the uncertainty of the sampling or the degree of the representativeness of the examined part of the sample. Another special feature of metallographic investigations is determining the size and proportion of the non-metallic (oxide and non-oxide) inclusions. Inclusions are irregular in shape and often have very complex morphologies. In some places, they appear in the microstructure as individual particles, and in others as aggregates and agglomerates (clusters of a larger number of particles). They are often present in the form of cracked or broken oxide films, and it is in this form that they are the most difficult to quantify. It is very important that we clearly define the rules for determining the representative sample and measuring the size and proportion of the inclusions with the customer, and then refer to them when printing the certificate.

Uncertainty of the Processing Paths The Role of Measurement Uncertainty in Quantitative Metallographic Investigations Metallographic (macro- and microstructural) investigations can be descriptive and quantitative. In descriptive metallographic investigations, the microstructure of the representative surface of the sample is examined and only the characteristics present are identified. In quantitative metallographic investigations, however, we measure these characteristics, which means that we use the microscope as a measuring instrument.

The process path is the sequence of all processes and sub-processes that must take place in order for a certain product to be successfully realized. The process path begins in the sales department, which receives the customer’s inquiry (and, among other things, also the quality requirements of the ordered product) and ends with the shipment of the product (including a quality certificate that proves the conformity of the actual product with the requirements). It follows from this that the measurement uncertainty when determining any product property consists not only of

1006

the uncertainty of the measurement system (the laboratory equipment with which we perform a certain measurement and the operators who perform this measurement) but also of the multitude of individual measurement uncertainties, the sources of which are the measurements of various parameters or sizes in virtually all parts of the process path used to produce the product. For this reason, the measurement uncertainty of the process path by which a certain product is produced is the broadest form of measurement uncertainty, which encompasses the uncertainty of all the measurements made. Uncertainty is related to repeatability, which is another name for stability. If the measurement system is stable, the measurements are repeatable and, consequently, have the lowest measurement uncertainty. The same applies to process routes. To the extent that a certain process (or, in a narrower sense, technological) path is stable, it is also more repeatable and, as a result, also has a lower measurement uncertainty. It should be noted that the reproducibility of the measurement of the selected property during the final quality control of the product (e.g., tensile strength) is also a measure of the uncertainty of the process path by which this product was produced. In manufacturing, where there is an ever-present antagonism between production technology and quality control, this fact is often forgotten and the uncertainty is attributed exclusively (or at least largely) to the measurement system and perhaps, to a certain extent, to the uncertainty of the sampling of the test subjects. But in reality, this story is significantly more complex and must be treated as such. The main source of most uncertainty (instability or insufficient repeatability) in product-quality certification is the production process or technology. Here, the unwritten rule applies that the uncertainty of the production process or technology is several times greater than the uncertainty of sampling, which is also up to three times greater than the uncertainty of the measurement system. If we take the previously mentioned tensile strength as an example, the measurement uncertainty of the measuring system for determining the tensile strength is usually (on the reference material) ±1%, the sampling uncertainty is mostly within the limits of ±2–3%, while the process path uncertainty can exceed ±10%. An analysis of all three types of measurement uncertainty (measurement system, sampling, and process path) can therefore help us improve processes and reduce risks, which will be presented in more detail below.

V. M. Kevorkijan et al.

Quality Assured Tools Designed to Measure and Process Uncertainty Guard Bands as a Tool to Ensure the Consistency of Comparative Measurements Guard bands, schematically presented in Fig. 2, are a tool which ensures that our measurements are comparable to the measurements that will be performed by the customer (at their premises or at an external laboratory). In fact, none of the standards currently in force explicitly require the introduction of guard bands. But that does not mean that we do not need them in practice. It turns out that guard bands are simply necessary if we want to ensure that we and the customer both measure the same result or are comparable within some predetermined probability of this event. We are therefore interested in the conditions relevant to the event, in order for us to measure in the same way as the customer with a probability higher than some predetermined value.

Fig. 2 Guard bands at the limits of the customer’s demand interval: a indicates the lower and b the upper limit of the customer’s demand interval, u is the sum of the measurement uncertainty of the measurement system, um, and the uncertainty of the process path, up. The value of u determines the width of the guard bands

On the Importance of Measurement and Process Uncertainty in Certifying the Quality of Aluminium-Based Products

At first glance, guard bands appear unfavorable for the manufacturer, because they “narrow” the intervals of the customer requirements. With the “narrowing of intervals”, the customer requirements, which we pass on to production, become more stringent, which is usually resisted by the technology. However, it should be borne in mind that the “narrowing of the intervals” of the customer requirements is a necessary consequence of eliminating the inconsistency of comparative measurements, which would occur at the limits of the intervals of the customer requirements if we did not introduce these guard bands. The desired probability of the event that both the customer and producer obtain the same result is the one that resolves the trade-off between production technology and the certification process. A production process that cannot be sufficiently authentically certified is risky and, in principle, unacceptable. It is important to note that customers do not prescribe guard bands. The customers only require the conformity of the comparative measurements over the entire interval of the given requirements. However, that can only be ensured in practice by prescribing stricter conditions to the production process or a “narrower” interval of the requirements, which is narrowed by a well-defined multiple of the combined measurement uncertainty.

1007

The Multicolor Quantification of Testing Results and the Compliancy of the Processing Paths The quality “traffic light” is a tool that measures the values of selected product properties or assigns the appropriate color to the selected parameter of the process path: green, amber, or red, depending on its position within the interval of the customer requirements or standard. In a machine analysis of test results, the tool is based on data on the measurement uncertainty of measurement systems collected in the e-register. It is similar for the monitoring of process paths, where the starting point for checking compliance is the e-register of the measurement uncertainty of process parameters. With its proactive analysis of test results, the quality traffic light can also diagnose the possible causes of low-quality products and help propose measures to eliminate them. The design of the quality traffic light and the way it works in a digitized environment are presented in Fig. 3. If the measured values of the product’s properties and the process parameters meet the customer requirements, it is colored green in the information system. This means that both reliably meet the customer requirements, as they are far from the required limit values by more than the width of the guard bands.

Fig. 3 Operation of a proactive, quality-indicating “traffic light” in a digitized manufacturing environment

1008

If the measured values deviate from the customer requirements by more than the width of the guard band, the quality traffic light turns this value red and classifies the product as a reject. Between red and green there is an intermediate amber color. This is reserved for products whose properties and parameters of the process path lie within the guard bands. This group of products also does not meet the customer requirements, but in these cases the deviations are minimal. Measurement values colored green can be exported to a quality certificate, as they comply with the customer requirements and can successfully pass the comparative measurement with the customer. Results marked by the scoreboard as red are rejects. These are products that cannot be repaired to meet the customer requirements. Although they represent a direct cost of non-quality, they can also be very useful, since an analysis of the possible causes of the occurrence of these errors, carried out by artificial intelligence, is an excellent starting point for improving the production process. The third group includes measurements that the quality scoreboard marks as amber. With on-the-spot corrections, if the lack of quality is recognized early enough, some of these products can still be repaired and delivered to the customer. When corrections are no longer possible, such products can be offered to customers with less stringent requirements, for whom the achieved level of product quality is completely satisfactory. Even in this case, the tool diagnoses the possible causes of poor quality and suggests measures to eliminate only these. Another function made possible by this tool is determining the compliance of the actually used production process with those requested by the customer. For example, the customer can request the use of specific production equipment and/or specific technological operations, which must also be documented on the quality certificate. The tool enables a real-time determination of the required compliance and printing of the determined status on the certificate.

V. M. Kevorkijan et al.

By consistently using such a proactive tool, it is possible to reduce costs, raise quality, and implement continuous improvements in the production process. The tool becomes especially effective in an increasingly or already fully digitalized business system or in the e-confirmation of the feasibility of customer requirements, e-production, e-quality control, and e-certification.

Conclusion In this paper, we examined the importance of measurement and process uncertainty in conformity assessment and closing the quality-assurance loop in the aluminum industry. With the help of continuously updated data on measurement and process uncertainty, we built guard bands and a proactive quality traffic light: two simple, cost-effective, and user-friendly quality-assurance tools. The consistent use of guard bands and quality traffic lights reduced non-quality costs by 70–80% and improved the stability of the production processes. It also reduced the risks of incorrect decisions and supplier risks. Tools are the key to the digital transformation of the aluminum industry, as they are a prerequisite for the autonomous communication of digitally controlled processes (from opening to closing the quality-assurance loop).

Reference 1. Thomas Belliveau, Denis Choquette, Olivier Gabis, Michael Ruschak, John Sieber, Albert Wills, and Rebecca Wyss, “Using Guard Bands to Accommodate Uncertainty in the Spark AES Analysis of Aluminum or Aluminum Alloys When Determining Compliance with Specified Composition Limits”, In Light Metals 2017, Ed. Arne P. Ratvik, TMS, pp. 275–284.

Characterization of Aluminum Dross Compositions Using Rietveld XRD Technique, Standardless XRF Method and Carbon Analyzer Hussain Al Halwachi

Abstract

Aluminum dross is a by-product generated during molten aluminum treatment and casting process. Knowing the composition of the dross will provide better environmental solutions through recycling and reusing. In this research, Rietveld XRD method is used to identify the crystalline phases present in dross. The XRD analyses were used to develop template files to ease the phase identification process. The Alpha Alumina phase is carried out by Rietveld XRD and separate XRD application since it is a crucial phase in aluminum smelting process. Omnian XRF standardless application was able to obtain the trace elements available in dross. Unifying sample preparation for XRD and XRF provided faster data reporting. In addition, Carbon content in dross is provided by a modified calibration of Leco Carbon analyzer. This methodology identified the crystalline phase of dross, the available trace elements, and the carbon content successfully with high accuracy and the fastest possible reporting time. Keywords

Dross




XRD




XRF

Introduction Aluminum dross is generated during melt transfer, metal treatment, and casting process due to the reaction between the hot metal with the atmosphere [1, 2]. Hot aluminum metal has high oxygen affinity to form aluminum oxide (Al2O3). Due to this exothermic reaction, there will always be a film of dross formed on top of hot metal during the H. Al Halwachi (&) Alba Laboratory, Aluminium Bahrain (Alba), P.O. Box 570 Manama, Kingdom of Bahrain e-mail: [email protected]

different processes of molten metal [3, 4]. Up to 10% of the annual aluminum production is lost in the form of dross [5], which is a huge volume. There are three types of dross, white, black, and saltcake [3]. The dross generated in aluminum smelting process is classified as white dross. The white dross is considered as a hazardous waste. Toxic gasses are generated when dross contacts humidity or water such as methane and hydrogen sulfide [2]. Such gases have a bad environmental impact. In addition, dross itself is considered as a by-product waste from the aluminum smelting process which some countries started looking for an environmental solution for it, while others dump it in landfills. Knowing the chemical composition of white dross will guide for better environmental solutions through reusing and recycling. It is not easy to identify the composition of aluminum dross due to the heterogeneity and complexity of the material. The dross composition is affected by several factors such as atmosphere temperature, furnace operations, and type of alloy [2]. Due to the heterogenous nature of dross material and complex composition, it is not easy to carry out the qualitative analysis with routine techniques [3]. Therefore, the crystalline phases are carried out by X-ray diffraction method (XRD), due to its ability to identify phases present in mixtures [3]. There are many approaches of XRD for phase identification, but Rietveld method was considered as a suitable solution for such material. Since the composition of white dross is affected by several factors, the crystalline phases will not be the same for each collected sample, like the case of an electrolyte bath, which contains known phases such as cryolite and chiolite. This means that peak identification will be one of the challenges in dross characterization. In this study, template files were generated to ease the phase identification for the routine daily activities. In addition, not all elements available in dross can be identified by XRD phases, because some of the elements are below the detection limits of XRD. Some of the trace elements are present in dross, but they don’t form a crystalline phase. The qualitative and quantitative analyses of these elements are obtained

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_127

1009

1010

in this study by standardless X-ray fluorescence (XRF), which is a modern approach for providing the elements in unknown materials. The data gathered from both XRD and XRF provided a clear picture about the total composition of dross material. Since Alpha Alumina is one of the crucial phases for reduction cell operation, it was carried out also by XRD Alpha application for Primary Alumina. Due to the global environmental focus on carbon emissions especially from the industrial sector, carbon content is considered as essential information for any further treatment or recycling for dross.

Analytical Method and Procedure A total of 24 dross samples were collected from three cast house plants, TAC station treatment and metal recovery plant at Aluminium Bahrain smelter (Alba) in order to collect all types of dross generated during transferring metal and producing final products (Fig. 1). The collected samples were left to cool and then transferred to the Laboratory for sample preparation steps. Small pieces of aluminum were removed, and dross material was crushed manually using a pestle and mortar due to the hardness of the material. The prepared powder was pressed in a steel ring.

Standardless XRF Application The analysis of trace elements in such a complex material like dross is usually carried out by ICP or atomic absorption, both techniques are very accurate but involve chemical usage and a lot of preparation steps. In this study, XRF Standardless technique is used to obtain the trace elements. Omnian software was used to provide the trace elements in dross (Fig. 2). The spectrum of each sample was evaluated to ensure that no element was missed. What made the Fig. 1 To the left: dross material collected at cast house, to the right: dross sample preparation

H. Al Halwachi

process faster was that I measured the same pressed powder ring of XRD machine, so one sample was prepared for XRD and XRF measurement. The trace elements also are required to specify the chemistry of the phases before starting the process of phase identification.

Crystalline Phases Identification Peaks scanning from (10 to 80 2h) was carried out using a Cubic XRD machine manufactured by Malvern PANalytical, using copper X-ray tube. The scans were treated by High score software, which is linked with ICDD reference database for crystalline patterns and provides qualitative analysis by comparing the scanned samples with the reference pattern using Rietveld approach (Fig. 3). The challenge here is to find which phases are available in samples and how to identify them using the reference database. Due to the sensitivity of this step, the phase identification was cross checked and confirmed by Malvern PANalytical Laboratory -Netherlands. To ease the daily work for dross samples, templates are generated depending on the nature and location of collected dross samples in the plant. Alpha Alumina was measured by XRD alpha Alumina application for primary Alumina, although the nature of dross and primary Alumina are different, good matching was found between Alpha Alumina by Rietveld XRD method and XRD Alpha Alumina application for primary Alumina.

Carbon Content in Dross Carbon content has become crucial in any substant nowadays due to the worldwide focus on carbon footprint and greenhouse gas emissions. Knowing the carbon content in dross will assist in taking the right decision for recycling or reusing in the aluminum smelting process. Carbon content in dross

Characterization of Aluminum Dross Compositions Using Rietveld XRD Technique, Standardless XRF Method and Carbon Analyzer

1011

Fig. 2 Spectrum of XRF peaks obtained by Omnian standardless software

Fig. 3 Spectrum of dross sample number 8 obtained by Rietveld XRD method

was measured using Leco CS844, which is used to analyze carbon in several materials such as cast-iron samples and electrolyte bath. For analyzing electrolyte baths, the machine is modified by placing a halogen trap, which is the same requirement for dross material since halogens might present in dross and might cause damage to Leco’s furnace. Since carbon was expected to be low in dross, the existing carbon calibration for electrolyte bath was modified by adding more standards in the lower range (Fig. 4) to make sure that the machine would be able to provide accurate analysis.

Results The trace elements for samples 1, 2 and 3 are listed in Table 1. Quantitative analysis for dross sample number 8 is shown in Table 2. Table 3 contains the carbon content of all 24 dross samples with an average of 0.29%. Alpha Alumina comparison between Rietveld XRD analysis and Alpha Alumina for Primary Alumina application is summarized in Table 4.

1012

H. Al Halwachi

Fig. 4 Linear correlation of carbon calibration gave 0.000103 RMS error and R2 of 0.999187

Table 1 Trace elements by XRF standardless for samples 1, 2 and 3 Sampl 1 Sampl 2 Sampl 3 Element Value (%) Element Value (%) Element Value (%) O 35.3 O 26.4 O 35 F 0.93 F 15.8 Al 33 Na 0.979 Na 8.55 F 19 Mg 3.441 Mg 0.187 Na 11 Al 56.5 Al 47.2 Ca 2 Si 0.643 Si 0.012 Mg 0.2 P 0.008 P 0.002 Fe 0.06 S 0.025 S 0.052 S 0.05 Cl 0.025 Cl 0.013 Si 0.03 K 1.046 K 0.019 K 0.02 Ca 0.341 Ca 1.692 Sr 0.01 Ti 0.213 Fe 0.048 Ni 0.01 V 0.122 Ni 0.008 CI 0.01 Mn 0.027 Zn 0.017 Zn 0.003 Fe 0.135 Ga 0.003 Ga 0.003 Ni 0.004 Sr 0.017 P 0.002 Zn 0.007 Ga 0.003 Sr 0.264 Ba 0.023

Table 2 Quantitative analysis for dross number 8

Phases

%

Spinel

43.8

Wurtzite

17.6

Alpha Alumina

18.4

Beta Alumina

2.4

Diaoyudaoite

3.2

Harmunite

6.5

Sodium Aluminum Oxide

8

Villiaumite

0.2

Kahlenbergite

0.3

Silicon Strontium

1.4

Characterization of Aluminum Dross Compositions Using Rietveld XRD Technique, Standardless XRF Method and Carbon Analyzer

1013

Table 3 Carbon content in dross samples ID Dross 1 Dross 2 Dross 3 Dross 4 Dross 5 Dross 6 Dross 7 Dross 8 Dross 9 Dross 10 Dross 11 Dross 12

C% 0.33 0.06 0.12 0.52 0.36 0.16 0.21 0.30 0.32 0.52 0.16 0.19

ID Dross 13 Dross 14 Dross 15 Dross 16 Dross 17 Dross 18 Dross 19 Dross 20 Dross 21 Dross 22 Dross 23 Dross 24

C% 0.29 0.17 0.04 0.07 0.08 0.12 0.12 0.08 0.11 0.05 0.18 0.06

Table 4 Alpha Alumina comparison between Rietveld XRD analyses and Alpha Alumina application for primary Alumina Sample Alpha phase (%) Rietveld XRD (%) Dross-8 17.5 18.4 Dross- 11 38.1 39.0 Dross- 16 16.0 16.6 Dross-17 23.0 23.7 Dross-18 18.8 19.1 Dross-19 8.0 7.3

Discussion In this study, I was exploring the components of dross, for the first time in the history of Aluminium Bahrain smelter (Alba), so a lot of challenges were faced during sampling, sample preparation, and analyses. The crystalline phase identification was the most difficult part since the content of dross is complex and carrying out this practice might involve a big percentage of errors. One of the most important crystalline phases found in dross by Rietveld XRD method is the corundum (Alpha Alumina). The amount of corundum in primary Alumina used in aluminum electrolysis process is usually between 2 and 10%, a controlled corundom content will ensure a smooth electrolysis process and decomposition of Alumina. In contradiction, dross material contained very high Alpha Alumina varying from 10 up to 40%, and due to the importance of this phase, dross samples were tested also by XRD application specified for testing Alpha Alumina in Primary Alumina, and the results found close to Rietveld XRD method, although both applications for different materials. Such a high percentage of Alpha Alumina is really

Diff -0.9 -0.9 -0.6 -0.7 -0.3 0.7

a concern if a decision is taken to recycle dross in reduction cells. The trace elements in dross are usually obtained by ICP or atomic absorption technique, both techniques involve chemical usage and long procedures of sample preparation. In this study, I used a standardless XRF application to obtain all the available trace elements with no additional sample preparation steps, the same pressed powder sample is used for XRD Rietveld method, Alpha Alumina test, and XRF standardless application. This shortens the reporting time and eliminates chemical usage, in addition, XRF standardless method can obtain a qualitative and quantitative analysis of all the trace elements that exist in the sample, while other technical approaches such as atomic absorption technique, will measure the element that the user specified to the instrument such as aluminum, Silicon, iron, and others, by preparing standards and blank along with unknown samples. Atomic absorption will not report tungsten values as an example if the user did not search for it and prepared standards to measure it, which contradicts with XRF standardless approach. During the initial work of dross analysis at Alba, I started measuring the phases by XRD, and then the trace elements by XRF. When I proceeded further with the

1014

H. Al Halwachi

Fig. 5 Sequence of tests for obtaining dross analyses

project, I found that it is better to measure trace elements firstly to know which elements exist, then measure Alpha Alumina, then the information of both techniques will support me in identifying the right phases by XRD (Fig. 5). The trace elements will help evaluate the decision taken in identifying the phases, as an example it is impossible to find Spinel (Mg–Al phases) if we did not find the Mg peak by XRF. Carbon content in dross was obtained using Leco CS844 machine, the average carbon content for all samples was 0.29% and the maximum value obtained was 0.52% which is on the lower side compared with carbon dust in an electrolyte bath. From the data gathered, carbon contents in bath vary between (0.55–2%). Carbon content in dross is a piece of information required in case the material will be reused in the smelting process.

Conclusion Dross phase identification was successfully obtained by Rietveld XRD method for all samples collected during dross generation at Alba plants. The phases identification of all samples involved in this study were used to generate phases temples to guide the worker during phase identification and ease the process. Trace elements were accurately obtained by XRF standardless method. Alpha Alumina was also measured by XRD application for primary Alumina due to the importance of this phase in aluminum smelting process. Alpha Alumina results were compared with Rietveld XRD method and good matching was found. The results of trace

elements and Alpha Alumina are required prior phase identification process for right phase selection and error elimination. Carbon content in dross was obtained by a carbon analyzer and found in low concentration. The developed procedures and methodology in this study provided accurate analyses and a full overview of dross contents. Acknowledgements I would like to thank Sheida Makvandi from Malvern PANalytical for all the analytical support given.

References 1. Zhang S, Zhu W, Li Q, Zhang W, Yi X (2019) Recycling of Secondary Aluminum Dross to Fabricate Porous c-Al2O3 Assisted by Corn Straw as Biotemplate. Journal of Materials Science and Chemical Engineering 7:87–102. https://doi.org/10.4236/msce. 2019.712010. 2. Gomez A, Lima N, Teno J (2008) Quantitative analysis of aluminum dross by the Rietveld method. J-stage 49 (4):728–732. https://doi.org/10.2320/matertrans.MRA2007129. 3. Solem C.K.W, Deledda S, Tranell G, Aune R (2023) Sampling Procedure, Characterization, and Quantitative Analyses of Industrial Aluminum White Dross. Journal of Sustain Metallurgy 9:95–106. https://doi.org/10.1007/s40831-022-00624-7. 4. Solem C.K.W., Ekstrøm K.E., Tranell G., Aune R.E. (2020). Evaluation of the Effect of CO2 Cover Gas on the Rate of Oxidation of an AlMgSi Alloy. In: Tomsett, A. (eds) Light Metals 2020. The Minerals, Metals & Materials Series. Springer, New York, p 1141– 1147. 5. Drouet M.G, Handfield M, Meunier J, Laflamme C, (1994) Dross treatment in a rotary arc furnace with graphite electrodes. JOM 46:26–27. https://doi.org/10.1007/BF03220691.

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys Bin Zhang and Gary P. Grealy

Abstract

Iterations of thermomechanical modeling on DC casting of high-strength crack-prone alloy billets have been conducted to understand billet start-up phase thermal stress development and to optimize tooling design. Transient thermal stress development during cast start at the billet butt, billet surface, and inside the billet are investigated; Connection of cracking with stress development at billet butt is presented; Effects of starting head, wiper placement, and wiper placement location on the thermal stress development are also examined. The billet surface rebound temperatures and cast-in TC temperatures from the modeling agree well with those measured in the laboratory and in the field; the casting campaigns on the AirSlip® mold package exhibit great billet quality and high pit recovery. Keywords













Thermomechanical Modeling AirSlip® Billet DC casting Thermal stress High strength Aluminum alloy Crack-prone Cold cracking Hot tears





Introduction With the development of aircraft/aerospace and defense industry, in recent years, more cast houses are requesting the capability of casting high strength aluminum alloys (2xxx, 7xxx and Al–Li alloys). High strength aluminum alloys are becoming candidates in other vehicle industries such as

railway and automotive to reduce weight. High strength aluminum alloys have long freezing ranges and low thermal conductivities, causing them greater thermal stress formation during casting; their as-cast microstructures consist of supersaturated primary aluminum dendrites and more eutectic intermetallic constituents, resulting in a low ductility and fracture toughness. These characteristics make high strength aluminum alloys very crack-prone when compared with their common alloy counterparts, i.e., hot cracking during the last period of solidification and cold cracking after solidification during DC casting and subsequent removal from the casting pit. AirSlip® casting technology [1, 2] was engineered almost four decades ago and has been widely accepted within the industry for ease of operation, great billet quality, high productivity and pit recovery. For some cast houses, it is still challenging to cast high strength alloys of higher alloy contents. A new iteration of AirSlip® mold package has recently been developed with the specific objective of simplifying the casting of high-strength, crack-prone alloys while maintaining excellent cast product quality along with high pit recovery. Modeling of thermal stress development during DC casting has been conducted to understand or predict the formation of hot tears [3–9], cold cracking [10–14], and optimize tooling [15, 16]. It is believed, when accurate boundary conditions, thermophysical and thermomechanical properties [17–21] are available, high fidelity modeling results on thermal stress could be achieved. In this work, modeling is conducted to understand the thermal stress development of industrial billet DC casting process and assist in the development of the new AirSlip® package for casting high strength, crack-prone aluminum alloys.

B. Zhang (&) Research and Development, Wagstaff Inc., Spokane, WA 99216, USA e-mail: [email protected] G. P. Grealy Technical Service, Wagstaff Inc., Spokane, WA 99216, USA e-mail: [email protected] © The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_128

1015

1016

B. Zhang and G. P. Grealy

Model Description A 3-D transient fluid flow, heat transfer, fluid–structure interaction, and thermal stress evolution model of DC casting was developed using the commercial CFD package, Flow-3D®. This package solves equations including momentum, energy conservation, continuity, motion, and stress. For the fluid-structure interaction (FSI), it solves stresses within solid components, while for the thermal stress evolution (TSE), it solves them within solidified fluid regions. In the TSE model, a temperature dependent elastic– plastic constitution relationship is employed for the solidified aluminum alloy. The Huber-von Mises yield condition of no hardening and the Prandtl-Reuss equations of perfect plasticity with an isotropic linear response for solidified aluminum from the yield point is modeled. The descriptions of the mathematical equations can be found elsewhere [22]. The model includes a quarter section of a 14″ AirSlip® mold and a starting head. The 3-D modeling is conducted instead of 2-D axial symmetry modeling to capture the features on the starting heads. The 3-D solid model is presented in Fig. 1 for differences between the starting heads. SH-A is starting head for common alloy casting and SH-B is starting head for high strength alloy casting with the improved design. A total of *0.52 million cells are constructed in the model and the grid sizes are *4.5 mm.

Initial and Boundary Conditions The mold filling was not modeled, instead, AA7068 aluminum alloy was programmed inside the mold with an adiabatic baffle arranged on top of the cross feeder to keep the incoming metal at a constant temperature of 667 °C; In doing so, it is assumed that the extra volume of liquid metal on top of the cross feeder has no influence on the thermal stress field. The initial temperatures for the mold/starting head were assumed to be at an ambient temperature of 27 °C. Cooling water temperature was assumed at 27 °C. A temperature dependent heat transfer coefficient was applied for metal with mold during AirSlip® casting process, a heat transfer coefficient of 200 W/m2/K was used prior to

solidification, and it linearly decreases from 200 W/m2/K at liquidus temperature to 100 W/m2/K at dendrite coherency point temperature (assumed at fraction solid of 0.5), then to 20 W/m2/ K at solidus temperature, finally to 5 W/m2/K at room temperature. The heat transfer coefficients between cooling water and billet were constants, 40 KW/m2/K for the water impingement region (from inverse heat transfer analysis) and 15 KW/m2/K for the water free falling region. Wiper itself was not physically modeled and below the wiper or below the water free falling region, a constant convection heat transfer coefficient of 30 W/m2/K and environment temperature of 27 °C were assumed at the billet surface. Wiper distance (WD) in the model is defined as the distance below the mold from mold tang to the wiper blade top surface including the wiper blade deflection during casting.

Thermophysical and Thermal-mechanical Properties Temperature dependent density, specific heat, and thermal conductivity are used for 7068 alloy. AA7068 aluminum alloy thermophysical properties was from JMatPro® V8.01, the liquidus and solidus temperatures for AA7068 alloy are 630.3 °C and 473 °C, respectively, and the latent heat release follows the nonlinear function of temperature within the solidus and liquidus temperatures. Measured temperature dependent elastic modulus/yield strengths for 7055 alloy [20] was approximated for the AA7068 alloy; at temperatures above solidus temperature, thermal strain is assumed to be zero. Temperature dependent A36 steel properties were used for the starting heads.

Probe Locations in the Model Five locations from billet in the model were probed at cast length of 382 mm for temperature, stress and plastic strain tensor components: (1) butt center, (2) butt at flat island edge for SH-A starting head or butt at clinch bottom for SH-B starting head, (3) butt under the rim (*10 mm from billet surface), (4) billet center where maximum stress was exhibited and (5) billet surface where maximum tensile

Fig. 1 3-D model for the starting heads: a common alloy stating head (SH-A) and b high strength alloy stating head (SH-B)

(a) SH-A

(b) SH-B

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys

1017

Fig. 2 Probe locations on billets at cast length of *382 mm from the two starting heads

(a) SH-A stress was exhibited. The five locations along with the defined terms for result comparison are presented in Fig. 2. Casting starts after initial mold fill and subsequent holding for a certain time with a start casting speed of V0 mm/min for a cast length of *152 mm, then ramped up to the run cast speed of V1 mm/min within a cast length of *203 mm. To simulate casting of a length of *510 mm, the model takes approximately 120 hrs. using Flow-3D® version 22.2 running on two 18-core each Intel®Xeon™ Gold 6254 CPUs (3.10 GHz).

Result and Discussion Common Alloy Starting Head (SH-A) versus High Strength Alloy Starting Head (SH-B) The starting head fulfills multiple functions, it helps with the initial billet butt formation and helps pull the billet out of the mold at the cast start. Starting head design for billet casting could be very different, because of the difference in materials, shape, and geometry features of the starting head, the heat transfer and solidification of the billet butt will vary, therefore, the thermal stress development on the billet butt will be not the same. The modeling of SH-A and SH-B in the same mold and under the same casting conditions were conducted and contours of temperature, and stress/plastic strain tensor components are presented in Fig. 3 at 382 mm cast length. WD = d0 mm is the same for billets cast with either starting heads. The corresponding values at five locations are labeled in the contours. Temperature contour for the billet butt is presented in Fig. 3a1, a2, the SH-B tends to show a colder butt center and billet center, warmer butt clinch bottom, and under the butt rim than the SH-A. For the water impingement zone, the

(b) SH-B billet surface temperatures in °C are the same for both starting heads and no significant difference is seen on the stress and plastic strain tensor components, so this location will not be discussed further. Butt Center: the circumference stress (rhh) and radial stress (rrr) change from tensile for SH-A to compressive for SH-B; the axial stress (rzz) are all small, compressive for SH-A and tensile for SH-B. Tensile maximum principal stress (r1) is significantly reduced for SH-B. The circum

 ference plastic strain ephh , radial plastic strain eprr , and
 axial plastic strain epzz are significantly lower for SH-B than for SH-A. The higher tensile stress/plastic strain for SH-A is directly the result of the constraining effect at the central button when contraction occurs. The change from tensile to compressive (rhh, rrr), significantly less r1 and tensile
 plastic strain ephh ; eprr for SH-B are very beneficial to reducing butt center cracking. Billet Center: tensile rhh, rrr, rzz, and r1 tend to be less for SH-B than for SH-A. Tensile ephh and eprr , compressive epzz are also one times lower for SH-B than for SH-A. Therefore, SH-B is a better choice for reducing billet center cracking. Butt Clinch Bottom: rhh and rzz change from tensile for SH-A to compressive for SH-B, tensile rrr and r1 are less for SH-B; ephh , eprr and epzz are all lower for SH-B than for SH-A. Again, the tensile stress/plastic strain for SH-A is the result of constraining effect at the flat island when contraction occurs. The stress change from tensile to compressive or reduced tensile stress and plastic strain indicates the chance of billet butt cracking at the clinch bottom region for SH-B will be less than butt cracking at the flat island edge for SH-A. Rim Bottom: compressive rhh is more, compressive rrr and rzz are less, and tensile r1 is more for SH-B than for SH-A; plastic strains are significantly less for SH-B than for SH-A. Due to the significantly reduced plastic strain, billet butt cracking under the rim bottom will also tend to improve for SH-B.

1018

B. Zhang and G. P. Grealy

64.2

246.1

18.8

248.5

87.4

203.1

-30.0

216.6

-45.3

91.2

251.0

219.6

45.4

53.0

-10.6

(f1)

-5.0

(d2) σzz

-3413

-11.6

-3568

p

(g1) ε

5034

p

-1211 1367

(f2) ε

θθ

8589

22.3

(h1) εpzz

-2774 -4904

-8015

3020

1521

rr

1407 12918

-1516

747

-1885

91.2

46.1

39.4

(c2) σrr

εpθθ

3096

1.34

44.3

77.8

-3842

2047

(e1) σ1

64.3

-20.0

769 7254

1.5

48.9

-1886

88.4

10.3

17.6

-1206

3092

(e2) σ1

18.9

(b2) σθθ

7.8

1.5

(d1) σzz

-36.3

(a2) T

64.2

-7.1

89.5 -4.0

200.2

90.1

-47.2

10.0

(c1) σrr

64.3

246.1

47.8 -13.4

(b1) σθθ

(a1) T

200.9

88.4 -10.9

-15.0

240.9

17.6

5195

p

(g2) ε

rr

-11609

-10229

(h2) εpzz

Fig. 3 Contours at cast length of 382 mm, wiper distance (WD) = d0 mm. SH-B: a1–d1 and e1–h1; SH-A: a2–d2 and e2–h2 (Units: temperature in °C, stress in MPa, labelled plastic strain in µe)

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys

If tensile stress (rhh, rrr) and tensile plastic strain (ephh , ephh ) are large enough at the last stage of solidification, hot tear will initiate at the butt center or at the flat island edge/clinch bottom; Billet butt is also a place where other surface defects (cold folds, oxide patch/films, etc.) might end up with when the process is not in control. These defects will decrease the alloy’s capability to resist butt cracking. As is presented, butt center from SH-A is of greater tensile r1/rhh/rrr and ephh /eprr , cold cracking is likely to occur here, plus the presence of above-mentioned butt surface/near butt surface defects. In the billet center from SH-A above the button, greater principal stress and tensile plastic strain are also exhibited, this will accelerate the crack propagation immediately along the axial upward once the crack at billet butt is formed. For SH-B, because of the change from tensile stress to compressive, significantly reduced tensile stresses at butt center, at the clinch bottom, and at the billet center above, plus less tensile plastic strain there, the occurrence of cracking (either hot cracking or cold cracking) will be much reduced. Thus, the developed SH-B is better at mitigating butt cracking. Indeed, this has been validated from high pit recovery casting campaigns.

With Wiper versus W/O Wiper Wiper technologies have been used in the DC casting industry for high-strength crack-prone alloys. The purpose is to let the surface of the cast product to reheat from the hot center, in such a way, the thermal gradient across the billet from edge to center will be decreased; thus, the thermal stress is reduced, and cracking avoided. One other benefit with wiper application is to get rid of water under the butt and eliminate the significant thermal stress formation when film boing changes to nucleate boiling under the butt [13]. The model for SH-B without wiper activated was conducted and contours of temperature, stress and plastic strain tensor components are presented in Fig. 4a–h at 382 mm cast length. When compared with that for WD = d0 mm presented in Fig. 3a1–h1, for billet butt temperatures, the SH-B with wiper shows apparently hotter billet butt than the SH-B w/o wiper. The average radial temperature gradient at *5 cm above the starting head is *4  lower for billets cast with a wiper (*0.16 °C/mm) than that w/o a wiper (*0.69 °C/mm). It is noted, although the wiper is *4.4 cm below the sump bottom with WD = d0 mm, the sump depth tends to increase *0.8 mm, and mushy zone width increases *3.7 mm when compared with those w/o a wiper. For billet surface in the water impingement zone, the difference between with a wiper and w/o a wiper is insignificant on temperature, stress/plastic strain tensor components and it will not be discussed further.

1019

Butt Center: compressive rhh and rrr are greater with a wiper than w/o a wiper; tensile rzz and r1 are all small and they tend to be more for with a wiper than w/o a wiper. Compressive ephh /eprr and tensile epzz tend to be small with a wiper. It is noted there are no plastic strains for w/o a wiper. For the butt center, the largest benefit of using a wiper seems to be having greater compressive rhh/rrr. Billet Center: with a wiper, tensile rhh, rrr and r1are about *10.0 MPa less than w/o a wiper; and tensile rzz tends to be more than w/o a wiper. Tensile ephh /eprr and compressive epzz tend to be more for with a wiper than w/o a wiper. Butt Clinch Bottom: with a wiper, both compressive rhh and tensile rrr/r1 tends to be *0–3.0 MPa less, compressive rzz tends to be more for with a wiper than for w/o a wiper; tensile ephh tends to be more, compressive eprr /epzz tends to be less. With a wiper at this distance, it seems the benefit for with a wiper at clinch bottom is marginal. Rim Bottom: with a wiper, rhh changes from tensile to compressive, compressive rrr tends to be less, compressive rzz is more, and tensile r1 is *12.2 MPa less than that for w/o a wiper, respectively. Both tensile ephh and compressive eprr and epzz are more with a wiper. Under the rim bottom, the greatest benefits are the change of rhh from tensile to compressive and decreased tensile r1 for wiper use; they are beneficial to avoid the butt rim cracking. If plastic strain is also more at the last stage of solidification, it might assist in early tear formation for with a wiper. However, it was found the circumferential cold fold formed at the very cast start usually helps to stop such tears from propagating upward. The application of a wiper tends to produce surface layer of greater compressive stress (rhh/rrr) and less tensile r1 without inducing great plastic strains at butt center and under the rim bottom; The other gain with a wiper use is the reduction of thermal stress in the billet center above the butt center. These are beneficial to keeping the crack precursors (cold folds/tears/folded oxide film), if any, to be in closure.

Wiper Distance (WD) Wiper distance below the mold is an important practice parameter, too much distance away from the mold may not yield efficient thermal stress reduction, too close a distance from the mold may extend sump depth and mushy zone length and result in more and larger defects such as porosity and increased macrosegregation in the billet center. In practice, a wiper is usually placed close to the bottom of the sump. To understand the influential magnitude from a wiper distance on the thermal stress reduction and metallurgical impact, three wiper distances are investigated. The contours of temperature, stress, and strain tensor components are

1020

B. Zhang and G. P. Grealy

245.7

19.1

64.6 177.6

98.1 17.5

50.0

82.0

141.5

-41.8

(e) σ1

0

(f)

1.8

-1251 -889

481 -2743

1985

-6.6

(d) σzz

-1890 408

5358

1.83

- 35.8

(c) σrr

3141

20.0

47.2 -8.8

12.6

(b) σθθ

98.6

13.3

98.6 -11.8

-30.7

(a) T

64.6

17.8

-2614

-3696

εpθθ

0

(g)

εprr

0

1710

(h)

εpzz

Fig. 4 Contours at 382 mm cast length for billet from SH-B without wiper: a–h

presented in Fig. 5 for the SH-B with a wiper at WD = d025 mm and WD = d0-50 mm. In comparison with a wiper at WD = d0 mm (Fig. 3a1–h1), as the wiper distance (WD) reduces from d0 mm to 25 mm closer and to 50 mm closer, the radial temperature gradients get smaller and they are *0.64 °C/cm, *0.42 °C/cm and *0.22 °C/cm, respectively, much lower than that of *5.0 °C/cm for when no wiper is used. The axial temperature gradient for the butt remains relatively constant for all wiper distances including w/o wiper, and the average is *4.8 °C/cm. Butt Center: when wiper is 25 mm closer to the mold than d0 mm, the compressive rhh and rrr both decrease *10 MPa, but it seems when a wiper is more than 25 mm closer to the mold, there is no further reduction on rhh/rrr. Wiper distance does not seem to have a significant influence on tensile rzz and r1. The differences on ephh , eprr , and epzz are all very small from wiper distance change. More reduction on compressive stress from a closer wiper distance to the mold is not beneficial to keeping crack precursors to be in closure on the butt center. Billet Center: as a wiper is 25 mm or 50 mm closer to the mold than d0 mm, the tensile rhh/rrr/rzz/r1 all decrease continuously. Taking maximum principals stress r1 as an

example, it is equivalent to a reduction of *10.3%, *17.7%, and *24.2% for wiper distance (WD) at d0 mm, d0-25 mm, and d0-50 mm when compared with that for w/o a wiper. Both tensile ephh /eprr and compressive epzz decrease as a wiper is closer to the mold. As a wiper is closer to the mold, more reductions on tensile stresses and plastic strains are beneficial for eliminating butt cracking and crack propagation upward. Butt Clinch Bottom: as a wiper is 25 or 50 mm closer to the mold than d0 mm, rhh/rrr/rzz/r1 are all less than *30 MPa, compressive rhh/rzz decreases, tensile rrr/r1 changes a few MPa. As a wiper distance (WD) is 50 mm closer to the mold than d0 mm, the tensile ephh increases, compressive eprr decreases and tensile epzz decreases. Rim Bottom: as a wiper is 25 mm or 50 mm closer to the mold than d0 mm, rhh/rrr/rzz/r1 are also less than *30 MPa, and compressive rhh/rrr/rzz decreases *5.0– 9.0 MPa, tensile r1 decreases *2.0/ *1.0 MPa. As a wiper is closer to the mold, the tensile ephh , compressive eprr and compressive epzz all decrease continuously. The reduction on compressive stress on the rim bottom may not be better but the reduction on plastic strains is beneficial as a wiper distance is closer to the mold. Billet Surface: when a wiper distance is 25 mm or 50 mm closer to the mold than d0 mm, tensile rhh/rrr/rzz/r1 tends

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys

246.5

19.2

64.3 277.6

235.8

243.5

-3.2

3.4

-22.6

-34.7

(e1) σ1

(f1) ε

7.0

1.7

(e2) σ1

0

0

(f2)

0

(h1) ε

rr

zz

-1447

-730

367 -1944

εpθθ

-325

p

363

2734

-1463 -2540

(g1) ε

θθ

4101

13.2

721

-1764

74.7

(d2) σ zz

p

3210

1.7

-1233

-1701

0

p

66.0

(c2) σ rr

-2756

2025

-2.3

-37.7

12.1

742 5296

39.7

-9.0

-1882

81.2

1.6

(d1) σ zz

-5.4

3115

6.0

1.6

72.6

(b2) σ θθ

64.3

8.8

(c1) σ rr

74.7

(a2) T

-3.9

18.4

314.5

277.1

- 36.1

21.8

66.0

268.2

7.5

-34.5

(b1) σ θθ

273.1

40.8

79.6 -3.5

-29.3

(a1) T

247.6

17.4

81.2 -0.1

232.6

1021

-2157

-3078

0

p

(g2) ε

344

0

p

rr

(h2) ε

zz

Fig. 5 Contours at 382 mm cast length for billet from SH-B, wiper distance (WD) = d0-25 mm: a1–d1 and e1–h1 and wiper distance (WD) = d050 mm: a2–d2 and e2–h2

1022

B. Zhang and G. P. Grealy

Billet Temperature Validation and Casting Commissioning

150

300

125

250 200 150 Under Bu Rim

100

Bu Center

50 0

Billet Center

During casting, the billet surface rebound temperature histories were recorded using a traveling surface contact TC as the billet emerged below the wiper; two TCs (one at the billet center and the other at the quarter diameter) were cast in the billet to measure the billet internal temperature until the cast-in TC at the billet center passed through the bottom of the wiper. The cast and modeling results are presented in Fig. 7a, b. The 1st rebound temperature history (RT1) starts when the billet butt is *25 mm in length below the wiper, the rebound temperature from the model (TC1) matches surprisingly well on both trending and magnitude with that measured. The 2nd rebound temperature history (RT2) starts at *609 mm cast length, the measured surface temperature just below the wiper is *145 °C and the peak rebound temperature is *219 °C; Model rebound temperature tends to be *6 °C hotter. The 3rd rebound temperatures (RT2) are repeated measurements, one at right below the wiper and the other at *40 mm below; the measured surface temperature below the wiper is *150 °C and the rebound temperature at *40 mm below the wiper is *214 °C; the rebound temperature from the model (TC3) agrees very well with that measured (RT3). The cast-in TCs start to record the metal temperature at *550 mm cast length. In general, modeling results at both locations match well with the measurements. However, billet center temperature from the model is lower at early solidification, this is associated with the modeling method that the continuous casting process is realized by using upward moving components/boundaries, more than real convection is probably induced in the center of the thimble; When the metal temperatures are below 280 °C (*1100 mm), both probes in the model show *29 °C lower temperatures at greater than *1200 mm cast length, this might be due to the assumption of fixed

Billet CTR3, d0-50 mm Slurry Zone / Mushy Zone Depth, cm

350

σ1, Max Principal Stress, MPa

Billet Bu Temperatue, °C

To validate the model, three strands casting of 7055 alloy billets were conducted in the Wagstaff® research laboratory

through the mold package with SH-B and a wiper distance at do mm. The target chemistries for the 7055 alloy are presented in Table 1 along with that for 7068. The two chemistries are like each other, and the casting practices developed were the same.

Billet CTR3, d0-25 mm

100

Billet CTR3, d0 mm Billet CTR3, No Wiper

75 50 25 0 -25

0

W/O Wiper 1

d02mm

d0-25mm 3

d0-50mm 4

(a) temperature at billet butt

5

0

100

200 300 Cast Length, mm

400

(b) σ1 at billet center

500

600

10

33

9

32

8

31

7

30 29

6 Mushy Zone, fs = 0.5-1.0

5

28

Slurry Zone, fs = 0.0-0.5

4

Sump Depth, cm

to increase; the tensile ephh tends to increase, compressive eprr tends to decrease and compressive epzz tends to increase. Wiper influence is not significant on billet surface in water impingement region. Below the wiper, the billet external is reheated from the hotter internal, Newtonian cooling condition is met and the temperature gradient across the billet is small, the closer the wiper is to the mold, the hotter the billet will be below the wiper, this is summarized in Fig. 6a for billet butt temperature dependency on the wiper distance. As a wiper is closer to the mold, both the compressive stress (rhh, rrr and r1) and plastic strain (ephh , eprr and epzz ) on the bottom butt surface tend to reduce, this is a result from the reduced temperature gradient across the billet associated with the closer wiper distance. The reduced compressive stress may not be beneficial to keep crack precursors in closure, but the reduced plastic strain is. The greatest benefit from closer wiper placement is the reduction of tensile stress (rhh, rrr, rzz, and r1) at the billet center as summarized in Fig. 6b. However, as a wiper is closer to the mold, the surface secondary water cooling is reduced and that may have influences on the sump depth and slurry zone/mushy zone lengths. As presented in Fig. 6c, at wiper distances of d0 mm and d0-25 mm, although the sump bottoms are at *44 and *14 mm above the wipers, they still show slightly extended sump depth and mushy zone length. At a wiper distance 50 mm closer to the mold, with the sump bottom at *21.2 mm below the wiper, the sump depth is *5 mm deeper and the mushy zone length is *15 mm longer; this could be a problem where more and larger porosity might form in the center of the billet. Thus, the determination of WD = d0 mm is reasonable, as it reduces billet center stress, maintains the compressive stress on the butt surface, and keeps the mushy zone depth and sump depth within limits.

27

Sump Depth

26

3 0

1 W/O Wiper

d20 mm

3 d0-25mm

d04-50mm

5

(c) sump/slurry/mushy zone

Fig. 6 Effect of wiper distance on: a billet butt temperatures, b billet center stress evolution, and c sump depth, slurry zone and mushy zone depths

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys

1023

Table 1 Alloy chemistries

700

400 TC3

350

TC2

TC1

RT3

RT2

RT1

600 500

250

Temperature, °C

Temperature, °C

300

200 150 100

400 300

50

100

0

0

0

500 308

1000 679

1500 1050

2000 1421

Billet CTR

200

Billet QTR ctr qtr 0 0

Cast Length, mm

500 308

1000 679

1500 1050

2000 1421

Cast Length, mm

(a) surface rebound temperature

(b) cast-in internal billet temperature

Fig. 7 Comparison between the model (solid lines) and the cast (broken lines): a billet surface rebound temperatures, and b cast-in TC temperatures inside the billet

thermal environment boundary in the model below the wiper, but the real thermal environment around the billet inside the pit varies with time as longer billets are withdrawn. The large diameter AirSlip® casting mold package has been successfully commissioned over a range of diameters and high-strength crack-prone alloys. The pit recovery was 100%, and both the surface quality and internal metallurgical qualities were excellent. 6  strand and 16  strand customer commissioning casts were successfully completed and are presented in Fig. 8a, b, respectively.

Summary Transient thermomechanical modeling on billet DC casting of high-strength crack-prone alloy has been conducted to understand thermal stress development and assist tooling design. The results show the following: • When the SH-A starting head is used, significant higher tensile stress (rhh/rrr/r1) and plastic strain magnitude (tensile ephh /eprr and compressive epzz ) are present at billet center (butt center and subsequent billet center above the butt).

Fig. 8 Commissioning of high-strength crack-prone alloys through AirSlip® mold package

(a) 7055

(b) 7068

1024

• When the SH-B starting head is used, the tensile thermal stress changes to compressive (rhh/rrr), r1 and plastic strain magnitude (ephh /eprr /epzz ) are significantly reduced at the butt/billet centers. These are very beneficial to keeping the crack precursors (cold folds/tears/folded oxide film) at the butt to be in closure. • The application of a wiper significantly reduced the radial temperature gradient (*0.64 °C/cm with a wiper distance at d0 mm vs. *5.0 °C/cm w/o a wiper) and the billet center thermal stress (rhh/rrr/r1) is much decreased. • The closer the wiper is to the mold, the greater is the reduction of thermal stress (rhh/rrr/rzz) and plastic strain magnitude (ephh /eprr /epzz ) at billet center. For maximum principal stress r1, the reductions are *10.3%, *17.7%, and *24.2%, respectively, for wiper distances at d0 mm, d0-25 mm and d0-50 mm, however, the sump depth and mushy zone tend to be longer. The surface rebound temperatures below the wiper from the model and castings agree very well. In general, the cast-in TC temperatures and probed temperatures in the model are also in agreement. The developed large diameter AirSlip® mold package has been successfully commissioned over a range of diameters and high-strength crack-prone alloys, great pit recovery, surface quality, and internal metallurgical qualities are achieved. Acknowledgements The authors are very grateful to Wagstaff Inc. for granting permission to publish this paper and co-workers for assistance to develop high-strength crack-prone alloy AirSlip® casting technology in large diameters.

References 1. Faunce JP, Wagstaff FE, Shaw H (1984) New casting method for improved billet quality. In: McGeer JP (ed) Light Metals 1984. The Minerals, Metals &Materials Society, Pittsburgh; Springer, New York, p 1145–1158. 2. Wagstaff FE, Wagstaff WG, Collins RJ (1986) Direct chill metal casting apparatus and technique, US patent 4598763, 8 July 1986. 3. Moriceau J (1975) Thermal stresses in continuous dc casting of al alloys, discussion of hot tearing mechanisms. In: Rentsch R (ed) Light Metals 1975. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 9571–585. 4. Katgerman L (1982) A mathematical model for hot cracking of aluminum alloys during dc casting. JOM 34 (2): 46–49. 5. Drezet JM, Rappaz M (2001) Prediction of hot tearing in dc-cast aluminum billets. In: Anjier JL (ed) Light Metals 2001. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 887–894.

B. Zhang and G. P. Grealy 6. Benum S, Mortensen D, Fare H, Overlie HG, Reiso O (2002) On the mechanism of surface cracking in dc cast 7xxx and 6xxx extrusion ingot alloys. In: Schneither W (ed) Light Metals 2002. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 967–974. 7. Sengupta J, Cockcroft SL, Maijer DM, Larouche A (2005) Quantification of temperature, stress, and strain fields during the start-up phase of direct chill casting process by using a 3D fully coupled thermal and stress model for AA5182 ingots. Mater. Sci. Eng. A 397 (1–2): 157–177. 8. Jamaly N, Philion AB, Cockcroft SL and Drezet, JM (2012) Hot tearing susceptibility in dc-cast aluminum alloys. In: TMS (ed) Supplemental Proceedings: Volume 2: Materials Properties, Characterization, and Modeling, The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, pp 259–266. 9. Jamaly N, Philion AB, Drezet JM (2013) Stress-strain predictions of semisolid Al-Mg-Mn alloys during direct chill casting: effects of microstructure and process variables. Metall. and Mater. Trans. 44B (10): 1287–1295. 10. Du J, Kang BSJ, Chang KM, Harris J (1998) Computational modeling of dc casting of aluminum alloy using finite element method. In: Welch B (ed) Light Metals 1998. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 1025– 1030. 11. Suyitno, Kool WH, Katgerman L (2004) Finite element method simulation of mushy zone behavior during direct-chill casting of an Al-4.5 Pct Cu alloy. Metall. and Mater. Trans. 35A (9): 2917– 2926. 12. Eskin DG, Lalpoor M, Katgerman L (2011) Cold cracking during direct-chill casting. In: Lindsay SL (ed) Light Metals 2011. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, pp 669–674. 13. Zhang B, Shaber C (2011) Aluminum ingot thermal stress development modeling of the Wagstaff epsilon rolling ingot dc casting system during the start-up phase. In: Prasad A, Taylor JA, Grandfield JF (eds) Aluminum Cast House Technology 2011. Trans Tech Ltd, Switzerland, p 196–207. 14. Wang YB, Krane MJ, Trumble KP (2018) Transient thermal stress development in direct chill cast ingots with application of a wiper. Int. J. of Cast Metals Research 31 (4): 193–208. 15. Schneither W, Jensen EK (1990) Investigating about starting cracks in dc casting of 6063 type billets. Part I: experimental results. In: Bickert CM (ed), Light Metals 1990, The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, pp 743–748. 16. Jensen EK, Schneither W (1990) Investigating about starting cracks in dc casting of 6063 type billets. Part II: modeling results. In: Bickert CM (ed) Light Metals 1990. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 749–755. 17. Wan J, Lu HM, Chang KM (1998) As-cast mechanical properties of high strength aluminum alloy. In: Welch B (ed) Light Metals 1998. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 1065–1070. 18. Philion AB, Thompson S, Cockcroft SL and Wells M (2008) Tensile properties of as-cast alloys AA3104, AA6111 and CA31218 at above solidus temperatures. Mater. Sci. Eng. A497 (1–2): 388–394. 19. Alankar A, Wells M (2010) Constitutive behavior of as-cast aluminum alloys AA3104, AA5182 and AA6111 at below solidus temperatures. Mater. Sci. Eng. A527 (29–30): 7812–7820.

Thermomechanical Modeling on AirSlip® Billet DC Casting of High-Strength Crack-Prone Aluminum Alloys 20. Wagstaff Inc (2012) Characterization on as-cast tensile properties of 7055 aluminum alloy at below solidus temperatures. Internal report, 10 December 2012. 21. Wang YB, Krane MJ, Trumble KP (2017) Quantifying as-cast and homogenized AA7050 mechanical properties trough compression

1025

testing. In: Ratvik AP (ed) Light Metals 2017. The Minerals, Metals & Materials Society, Pittsburgh; Springer, New York, p 399–407. 22. Flow Science Inc. (2022) User Manual: Theory, Rev 2, 2022.

A Passive Approach to Butt Swell Management S. R. Wagstaff, R. B. Wagstaff, B. Opdendries, A. Anestis, S. Pinis, G. Pashos, E. Xenos, and A. Mavroudis

Abstract

The thickening of the base of rolling slab ingots is colloquially known as butt swell. Because traditional DC casting systems are vertical and not truly continuous, this remains one of the remaining sectors of recovery loss in the casthouse and at the scalper. Butt swell is currently managed by using dynamic molds which adjust the mold bore as a function of casting recipe to deliver flat ingot profiles. In this paper, we analyze the metallurgical origin of butt swell in the DC casting process. Based on this analysis, a novel bottom block design is proposed which reduces the extent of the butt swell in the start-up regime of casting. Experimental results are reported in comparison to a standard reference case demonstrating a reduction in butt swell. Subsequent rolling trials also indicate additional scrap reduction in the flat rolling process. Keywords

DC casting




Butt swell




Casting




Bottom block

Introduction Direct Chill (DC) molds used to produce aluminum rolling slab ingots typically have mold bores which are designed to deliver flat ingots for a given alloy cast at a specified speed. Hakonsen [1] presented a well-known model to predict the curved profile of the rolling face for a variety of commercial S. R. Wagstaff (&)
R. B. Wagstaff
B. Opdendries Oculatus Consulting, Spokane Valley, WA 99016, USA e-mail: [email protected] A. Anestis
S. Pinis
E. Xenos
A. Mavroudis Elval, 61st Km Athens-Lamia National Road, 32011 Oinofyta, Viotia, Greece G. Pashos Elkeme, 61st Km Athens-Lamia National Road, 32011 Oinofyta, Viotia, Greece

alloys at a desired casting speed. The vertical DC process is semi-continuous and is limited in length by the depth of the casting pit where it takes place. As a result, the process must be restarted many times each day. The start-up procedure is a careful balance of thermal distortion and heat input and extraction. Most start-up practices involve a careful ramp-up of casting speed from an initial lower velocity to a steady-state velocity to better control these forces. Figure 1 contains mold bore openings for a desired 620 mm  1940 mm AA3104 ingot cast at two different casting speeds indicative of start-up (35 mm/min) and steady-state (60 mm/min) conditions. Under higher casting speeds, the mold bore must be wider than a bore designed to cast at slower speeds. Thus, during start-up when casting at a speed slower than the designed steady-state speed, a “swell” or thickening of the ingot butt occurs. Conversely, if casting at a speed higher than the designed mold bore, an excessive pull-in will occur, resulting in an ingot with a bow-tie cross-section. Butt swell must be removed at the scalper along with the surface shell zone in order to deliver a flat ingot with acceptable geometry and surface quality to the hot mill. The sum of the butt swell and shell zone depth dictates the total required material removal depth at the scalper. For under-powered scalpers or those with shorter cutting knives, butt swell can mean lost productivity as multiple passes must be taken per side to remove the unwanted material. As mold technology and process control have improved over the past three decades, casting speeds have steadily increased industry-wide, leading to larger mold bores and, consequently, an increase in butt swell. This has coincidentally occurred with a reduction in shell zone as well, which has reduced the need for excessive scalping depths, thus, leaving butt swell to be one of the last remaining recovery difficulties at the scalper. The pull-in of rolling faces occurs mechanistically via a rotation mechanism. Figure 2 is a schematic representation of this process. In this representation, the sump can be represented by a right triangle with the vertical leg

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_129

1026

A Passive Approach to Butt Swell Management 345 340

Half Thickness (mm)

Fig. 1 Calculated mold bore openings (MBO) for a 620 mm  1940 mm AA3104 ingot cast using two different casting speeds indicative of start-up (35 mm/min) and steady-state (60 mm/min) conditions

1027

335 330 325

35 mm/min 60 mm/min

320 315 -1000.00

-500.00

corresponding to the solid cooled wall adjacent to the mold and water curtain. The hypotenuse represents the liquid– solid interface or sump. The third leg (not pictured) represents the half thickness of the ingot. As solidification and cooling occur, contraction along the sump (hypotenuse) causes the overall length to decrease. As the outer edge of the ingot can be seen as rigid, this causes the apex of the triangle to pivot inwards. In a more robust treatment, the solidifying rolling face should be seen as a prism. Thus, the contraction is actually two dimensional in nature along the slant face and is closer represented by the thermal distortion of a constrained beam [2] than by a triangle. This simplification allows us to better understand the importance of the sump depth on the degree of pull-in. Deeper sumps will inherently increase the length of our imaginary hypotenuse. Since thermal contraction is represented as a function of

Fig. 2 Schematic representation of pull-in via an apex twist mechanism driven by the solidification and contraction of metal along the liquid–solid interface

0.00

Distance from Center (mm)

500.00

1000.00

original length, longer sump interfaces will lead to more contraction and thus, an increased degree of pull-in. Weaver [3] may have been the first to identify this relationship between sump depth and mold pull-in. In his treatment, which was improved upon by Hakonsen, the sump depth was determined to be a function of TV2 where T is the ingot thickness and V is the casting velocity. An alloydependent constant was used in front of this parameter to account for thermal property variations (conductivity and contraction) between alloys. Weaver demonstrated a defined linear relationship between TV2 and both sump depth and pull-in. Figure 3 is a plot of data extracted from [3] where the TV2 term has been eliminated and instead, pull-in is seen to be linearly correlated with the depth of the sump. During the start-up regime, sump depth is limited by slower casting speeds and shallow bottom blocks. The consequence of this is reduced pull-in and butt swell as has previously been described. Weaver concluded that the only way to overcome this thermal restriction is through the development of a flexible mold that dynamically adjusts bore profiles as a function of position within the cast. Lawrence [4], Marchand [5], Hakonsen [6], and Cordill [7] have all demonstrated flat ingots using such a system. The small recovery and production loss associated with butt swell makes the financial justification of these systems difficult to justify. As a result, they have seen limited adoption in the industry. The shallow sump during start-up appears to be the limiting factor to reduce butt swell. To provide a passive approach to manage butt swell, a proposal was made to manufacture a deeper bottom block to use during casting start-up. This deeper block would allow for sump depths greater than those possible in traditional systems. By providing a deeper sump on cast start, the mold pull-in would be increased and thus, butt swell reduced without the need for dynamic mold bore profiles.

1028 60

Mold Pull-In (mm)

Fig. 3 Data extracted from [3] representing the linear relationship between sump depth and mold pull-in

S. R. Wagstaff et al.

50 40 30 20 10 0

0

100

200

300

400

500

Sump Depth (mm)

600

700

800

900

Experimental Approach To test the hypothesis, a deeper dished bottom block was designed and manufactured to complement a 620 mm  1940 mm Wagstaff LHC mold for AA3104. Figure 4 is a CAD rendering of the designed block. The depth in the center of the block is 11″ (274 mm) and tapers down to 12 inches (304.8 mm) on either end to allow for water drainage. Recognizing the thermal stresses generated in a typical block, the bottom plate was manufactured from stainless steel to provide extra rigidity to the system. The difference in thermal conductivity between steel and aluminum was discussed as a possibility to increase effective sump depth but given the volume of metal, the effect was deemed to be minimal. The peripheral ring was milled from a homogenized ingot of AA6061, and extra “buttresses” were added again to protect against excess thermal distortion. This block was mounted onto a typical starting head base with four additional bottom blocks of a traditional shape. Given the height difference between the blocks, spacers were used under the four traditional blocks to make the lips of all five blocks co-planar. A longer spout and pin were manufactured to avoid excess dross generation caused by the cascading of the metal. Due to the volume differences between the blocks, special precautions had to be taken to ensure that start-down could be safely initiated at the same time. A temporary dam was put in place between the deeper block and the standard blocks such that the filling of the deeper block was initiated before the other ingots. Once the metal level system determined adequate metal had been poured into the block, the dam opened, and the other ingots were filled. The casting automation system regulated flow independently into each of the ingots, so the metal level was identical for all 5 ingots before start-down was initiated according to the recipe. Water flow rates and speed ramps were kept identical for all 5 ingots with no special precautions taken for the deep block.

Fig. 4 Longitudinal (a) and Transverse (b) cross-sectional views of the deep block. (c) is an Isometric view of the same block

Once casting was complete, ingots were removed from the pit and butt swell profiles were measured using a laser plane. The standard ingot butts were cut according to practice (butt swell is known to be excessive) and all 5 ingots were sent to the scalper for laser profile measurement. Profiles were verified at the scalper to ensure measurement accuracy.

A Passive Approach to Butt Swell Management

Fig. 5 Image of ingot cast with deep bottom block exiting the pit

Results Casting of the deeper block system was completed without incident. Figure 5 is an image of the butt of the cast ingot as it was being withdrawn from the pit. As can be seen, the butt is crack free and there are no cold-shuts within the block portion. There is very little curl at the short face, and as a result, there is no observed notch 50–75 mm above the lip as is typical. Figure 6 is a measurement of the butt swell profile for the ingot cast with the deep bottom block and the adjacent ingot cast with the standard block. While not eliminated, the butt swell was reduced by approximately 5 mm at the lip and reached a maximum differential of around 8 mm 30 cm from the lip. At approximately 100 cm from the lip, the two profiles converge together and remain nearly co-linear through the rest of the measured section. It should be noted that these data points represent only one side of the measured ingot, and thus, the total butt swell over the entire butt would be double that reported.

Numerical Model To get a better understanding of results, a simple numerical model of the sump evolution was developed using ANSYS Fluent. The liquidus and solidus were assumed to be equivalent in this model to aid in determining the sump interface

1029

and to help with numerical stability. Boundary conditions along the sides of the evolving ingot were modelled using values from [8]. A constant heat transfer resistance between the ingot and the two different blocks was assumed to be 0.00033 m2K/W. The simulation also ignored filling dynamics and time and instead initiates once the mold and bottom block are full of metal at 697 °C. The sump shape and position are then mapped as the casting speed ramps up from 35 mm/min to a steady-state value of 60 mm/min. Figure 7 are snapshots of the sump interface viewed from the short/end face along the longitudinal plane. Here we can see that the deep bottom block initially begins with a much squarer profile and later evolves into a shape consistent with parabolic growth. The deep block also has a squarer profile until somewhere between 500 and 800 s when the shape becomes similar to the standard case. Figure 8 is a plot of the sump depth for the same two cases, where the maximum depth is reported as a function of time. Here we can see that the deep block initially begins with a sump depth approximately 2.4x greater than the standard case, a condition which is maintained through most of the start-up phase. The deep block has a local maximum depth (viewed as a minimum) at approximately 800 s of cast duration after which the sump depth increases and appears to converge to the steady-state depth exhibited by the standard case. This same overshoot is observed in Fig. 7 when comparing the lines for 800 and 1100 s. Between these two cases, the sump depth decreases (becomes shallower) before reaching the final shape. The standard case has no local maximum depth and instead, the depth increases gradually to the steady-state value of approximately 720 mm.

Discussion From Fig. 6 it is clear that the deeper bottom block did, in fact, exhibit lower butt swell than the standard case. This varied between 5 and 7 mm per side, for a total of 10– 14 mm less swell total across the entire mold bore. If we take the calculated sump depths from Fig. 8 for the standard and deep block cases, 140 mm and 360 mm respectively, we can apply these depths to Weaver’s data from Fig. 3. The predicted difference in pull between these two depths is approximately 15 mm, which corresponds with the measured data. The numerically predicted depth did not account for any loss in contact between the block and the ingot due to curl, which may have led to a slight under-estimation of sump depth and accounted for the slight difference between the predicted and actual pull-in. To completely remove butt swell from the ingot, the sump would need to initially represent the steady-state value. This would be between 700 and 720 mm using data from Figs. 3 and 8 respectively. This block would be twice as deep as the current deep block and

1030

S. R. Wagstaff et al.

25

Swell (mm)

20 Standard Flat Boom Deep Boom Block

15 10 5 0 -5

0

50

100

150

200

Distance from Lip (cm) Fig. 6 Butt swell profile along ingot centerline for the deep bottom block and standard flat bottom block

Standard Flat Botom

-0.1

-0.1

-0.2

-0.2

-0.3

-0.3

-0.4 -0.5

200s

-0.6

500s

-0.7

800s 1100s

-0.8 -0.9

0

0.1

0.2

0.3

Deep Botom Block

0

Sump Depth (m)

Sump Depth (m)

0

-0.4 -0.5

200s

-0.6

500s

-0.7

800s 1100s

-0.8

0.4

Distance from Center (m)

-0.9

0

0.1

0.2

0.3

0.4

Distance from Center (m)

Fig. 7 Sump profiles for standard flat bottom and deep bottom blocks as generated by the computational model at 200, 500, 800, and 1100 s

thus, would represent considerable metal volume lost at the hot mill unless proper scalping could be achieved. The sump overshoot predicted in the numerical prediction of the deep block was not something observed appreciably in practice. This would have exhibited itself as a local narrowing of the ingot profile as discussed in the introduction. The

numerical model did not factor in connections to the starting head base and thus, may have over-predicted the temperature of the steel base plate, causing an excess in sump depth. If one assumes that the motivation for butt swell reduction is a bottleneck at the scalper, we can see that the implementation of the deeper block will save approximately

A Passive Approach to Butt Swell Management

0 -0.1

Sump Depth (m)

Fig. 8 Maximum sump depth from numerical model for standard flat bottom and deep bottom block ingots as a function of time

1031

-0.2

Standard Flat Boom Deep Boom Block

-0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9

0

500

1000

1500

2000

2500

Time (s) one scalper pass per side. If a center is forced to cut ingot butts in order to meet a certain thickness threshold, we can see from Fig. 6 that implementation of a deeper block will eliminate the need for removal of 350 mm from the ingot butt. This can lead to shorter ingot requirements and thus, higher annual throughput. The goal of this project was to reduce the butt swell in rolling slab ingots. However, through the course of this investigation, other observations were made that are worthy of mention: (1) Reduced Alligatoring. It is perhaps obvious that the deeper block simply makes the butt heavier, thus, leading to increased scrap loss even with the reduced swell. This is acknowledged especially for casthouses and rolling mills which cut ingot butts as standard practice. However, the square profile caused by a sawn ingot or a flat block is known to produce alligatoring on the rolling mill as the deformation within the mill is non-uniform. As the surfaces deform preferentially, once the ingot has been reduced in thickness sufficiently, the surfaces become significantly longer than the center, causing the ingot to resemble the open jaws of an alligator or crocodile. This section must be removed to prevent additional cracking further downstream in the rolling process. It was recognized in the 1940s by Russian metallurgist D.A. Petrov [9] that a gradually tapered shape which allowed for a “nose” along the ingot center would reduce or eliminate alligatoring. As the surfaces in contact with the rolls will deform preferentially, having an ingot center longer than the faces meant that at a point

later in the rolling process, the rolling faces and center became co-planar. If any portions of the plate needed to be sheared off, it was significantly less than that required with a square-nosed ingot. It was recognized early in the project that the ingot butt formed by the deep block met the condition for “Petrov’s Lock” and, thus, the ingots cast as a part of this trial were rolled without having butts sawn as standard practice. It was observed that no alligatoring occurred even when identical rolling practices were used. As the un-scalped surface of the ingot was pressed to the end of the transfer bar, the un-scalped surface was removed as a part of a minimal shearing exercise. Further work is needed to optimize the trade-offs between the rolling mill and casthouse, but the initial results were quite promising. (2) Reduced Butt Curl. When using a standard flat block, care must be taken to avoid butt curl, especially in ingots with high width to thickness ratios. The mechanisms behind curl have been treated elsewhere and will not be discussed here. It is sufficient to note that curl occurs because of the initial planar nature of the solidifying ingot. The young ingot is thin and susceptible to thermal distortion. Thus, when water impingement occurs and layers further up are colder than layers below, the ingot butt acts as a bimetallic strip, causing the ingot to deform into an inverted arch. A great deal of research has been spent on various butt curl reduction mechanisms primarily focused on reducing the severity of the quench. During the present trial (W/T ratio = 3.12), it was observed that minimal curl occurred. Absent of any cracks to relieve stress, it was hypothesized that the

1032

S. R. Wagstaff et al.

deeper block allowed for less thermal gradient and a stronger butt due to the increased sectional modulus. Additional trials are required on different alloys more susceptible to curl to validate the results and optimize casting practices.

Conclusion In this test, a passive means to control butt swell was investigated. A hypothesis was formulated that a deeper initial sump would increase the pull-in at the early states of casting. A deeper block was fabricated and used on an industrial casting center alongside traditional flat bottom blocks. Results indicated that butt swell was reduced by 25% and was consistent in thickness to a standard ingot with 350 mm of butt removed. This indicates that the deeper block was indeed capable of reducing butt swell, though not completely eliminating it in the current embodiment. Further investigation is required to optimize the shape of the block with respect to “Petrov’s Lock” and the reduction in alligatoring at the hot mill. Such a process would mean significant scrap reductions. Additional work should also be performed to optimize the casting practices of these types of blocks. With an increased resistance to curl, wider ingots may be possible and more robust start-up practices may increase worker safety. The authors are grateful to ELVAL-HALCOR for the use of their casting pit, and unwavering support of furthering knowledge.

References 1. A. Hakonsen, (1997) A Model to Predict the Steady State Pull-In During DC-Casting of Aluminum Sheet Ingots In: Huglen, R.(ed) Light Metals 1997, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 675–682. 2. H. G. Fjaer and A. Hakonsen, (1997) The Mechanism of Pull-In During DC-Casting of Aluminum Sheet Ingots In: Huglen, R.(ed) Light Metals 1997, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 683–690. 3. C.H. Weaver, (1976) An Empirical Model to Explain Cross-Section Changes of D.C. Sheet Ingot During Casting In: Light Metals 1976, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 441–456. 4. R. Lawrence, (1976) Cross-Section Shape Control of D.C. Sheet Ingot Using a Flexible Mold In: Light Metals 1976, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 441–456. 5. P. Marchand et al., (1998) An Adjustable Mould for Sheet Ingot Production In: Welch, B.(ed) Light Metals 1998, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 1197–1201. 6. A. Hakonsen et al., (2012) Square rolling slabs from the start of casting-the elimination of butt swell In: Suarez, C. (ed) Light Metals 2012, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 1113–1116. 7. C. Cordill, (2020) Mold Shape Control for Direct Chill Ingot Casting In: Tomsett, A. (ed) Light Metals 2020, The Minerals, Metals, &Materials Society, Pittsburgh; Springer, New York, p. 887–891. 8. S.R. Wagstaff (2015) Experimental observations and analysis of macrosecregation in rolling slab ingots. Massachussetts Institute of Technology M.S. Thesis. 9. N.I. Svede-Shvets (1955) Transverse rolling of flat DC ingots with “Petrov’s Lock”, Aluminievye splavy, Moscow: Oborongiz, p. 294– 302.

Characterization of Cr-Bearing Intermetallics Causing Pinhole Formation in Twin Roll Cast 8079 Aluminum Alloy Thin Foils Yusuf Özçetin, Ali Ulus, Onur Birbaşar, and Feyza Denizli

Abstract

8079 alloy, produced by the twin roll casting method, is widely used in the production of thin foils due to its hardness and elongation properties. However, some quality problems must be overcome in cold rolled foils that are thinned to a thickness of around 7 microns. At the beginning of the problems that need to be solved, pinhole formations whose number is above the acceptance criteria come. In this study, pinholes caused by intermetallic or intermetallic-like structures containing Cr were investigated. However, in order to solve the problem, it should be known which metallurgical or mechanical reasons cause pinhole formation. A number of characterization methods were applied to foil samples. It is aimed to make inferences about which production step or which operation in the casting system is the source of the problem. Keywords

Aluminum Pinhole




Characterization




Solidification




8079




Introduction Aluminum sheets with a thickness of less than 200l are considered foils. These aluminum foils are used in a wide range of applications, from heat exchangers to the packaging Y. Özçetin (&)
A. Ulus
O. Birbaşar
F. Denizli Istanbul, Turkey e-mail: [email protected] A. Ulus e-mail: [email protected] O. Birbaşar e-mail: [email protected] F. Denizli e-mail: [email protected]

industry. Especially foils with a thickness of less than 50l generally find a place in food packaging applications[1]. The main reasons for using aluminum as a packaging material are; Its thermal conductivity and impermeability to oxygen. Additionally, aluminum can be laminated and printing can be easily applied. The major alloying elements of 8xxx series aluminum alloys are iron and silicon. 8079 alloy, a member of this alloy family, is frequently used in flexible packaging applications. It has much higher hardness compared to the low mechanical properties provided by the 1XXX series. Foils up to 6.35 microns thick can be produced from this alloy. However, producing such thin foil is a job that requires precision and the correct optimization of all production processes. Quality criteria in the final product must be met [2]. The number of pinholes is one of the main quality acceptance conditions. Micro-sized holes on the foil are called pinholes. As a result of these holes exceeding a certain size and number, the produced foil is not suitable for use in food packaging applications[3, 4]. In the twin roll casting (TRC), the molten metal passes through a bunch of filtering operations(Ceramic Foam Filtration, Deep Bed Filtration, additional other filterings) and solidifies between and by contact with two rotating rolls to produce an aluminum strip [5]. Due to high solidification rates, the microstructure is quite heterogeneous. While a very fine microstructure forms on the plate surfaces in contact with the roller, the microstructure becomes coarser towards the center of the plate. This situation is directly related to the SDAS (Secondary Dendritic Arm Spacing)Solidization Rate theory. This heterogeneous structure created during casting is eliminated by homogenization and annealing processes [6]. In the liquid metal preparation phase before casting, it is important to properly add all melt components. In addition, it must be ensured that the melted liquid metal is effectively filtered during the filtration stages. Cleaning of the casting sprues is also an issue that needs to be taken into

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_130

1033

1034

Y. Özçetin et al.

Fig. 1 Schematic representation of twin roll casting system

consideration. Ceramic-based particles that may mix with the liquid metal may cause pinhole formation in the final foil. In addition, since coarse intermetallics will exhibit a hard mechanical behavior like ceramics, they can penetrate aluminum foil, which is a relatively soft matrix. Therefore, heat treatments after casting are of great importance. The schematic representation of the casting system is given in Fig. 1. When the general pinhole problems in foils are examined; These problems can be divided into indogenic and exogenic. Indogenic problems; Residual gases in the hot metal

Fig. 2 Darkroom view of pinhole formations

(H2, Cl), incompletely melted elemental additives (Na, Fe, Si, Mn, Cu), inclusions (SiO2, Al2O3, MgO, TiB2), and graphite used as a separating agent in the casting roll can be shown as exogenic problems in rolling. These are situations that occur due to contamination of the aluminum material during processing. Exogenic problems are: Al (aluminum smut), Al2O3, SiO2, cellulose, and Fe–Cr (rolling roll). Hole defects in the aluminum foil can be seen as shown in Fig. 2 on the LED illuminated table in the dark room [7, 8]. In this study, it was aimed to characterize intermetallic or intermetallic-like structures that are similar to Al–Fe-Si

Characterization of Cr-Bearing Intermetallics Causing Pinhole …

1035

Fig. 3 Process flow-chart of packaging foil

intermetallics but also contain Chromium. It was questioned whether the problem was caused by casting or rolling. Then, according to the findings, the change in the number of pinholes after the operational actions taken is given.

Materials and Methods 8079 aluminum alloy is cast by twin roll casting machines in industrial scale. After casting operations, thermo-mechanical processing is conducted to produce thin foils about 7 microns of thickness. (Fig. 3) Samples taken from the cast coils and foil which showed quality problems are characterized by SEM (ZEISS EVO 15) and EDS analysis in order to identify the inclusions and root cause of the pinholes found. Before and after results of the Struck brand in-line Pinhole Detector device for the problematic coils are shared.

Experimental Studies The alloy in accordance with the composition of the specified 8079 alloy in EN standards was prepared in the melting furnace. Then, after the necessary fluxing and mixing processes, the coil is transferred to the casting runner. After transfer, grain refiner is added and it is passed through CFF and PDBF filtering operations, respectively. Finally, the liquid metal ready to solidify passes through the rollers and solidifies in the form of a plate. The cast coil-shaped aluminum sheet is subjected to rolling and annealing processes as specified in the production route. Finally, the pinhole results of the coil were first examined in the pinhole detector

results on the line. Figure 4 shows the detector results of a coil that does not meet the pinhole acceptance criteria. According to PHD results; Regions dominated by yellow, orange and red colors indicate high density pinhole areas. According to the acceptance criteria, only 10% of the yellow area of the coil length is acceptable, but in the results of the coil shown, this percentage is much higher. The detector counted approximately 650,000 pinholes in different size classes in the lower layer of the material. Samples were taken from areas where the amount of pinholes was dense and examined with SEM. In images taken in BSD mode; The discontinuity inside the hole can be clearly seen. It has been clearly seen that a structure with a size of approximately 100 square micrometers caused the hole. In addition, this structure of approximately similar size and content was found in the analyzes taken from other hole samples. When the EDS analysis results are examined, the basic elements that make up the structure; It appear to be Al, Fe, and Si. This is the inherent and expected intermetallic structure of the alloy, but in addition to these three elements, there is also a very low amount of Cr. In the classical analysis of aluminum foils, it is normal to see Al-Fe-Si intermetallics distributed in very small sizes in the matrix, as in Fig. 5. The coarse intermetallic structure coming from the casting becomes very small after homogenization annealing and as seen, it spreads into the matrix in the form of small intermetallics. In fact, the hardening mechanism of the foil is these intermetallics dispersed in the matrix. When the Al-Fe-Si-Cr structure that caused the error was examined, it was seen that this structure did not shrink after annealing and remained very large compared to other normal intermetallics. Here, it automatically comes to mind that the

1036

Y. Özçetin et al.

Fig. 4 PHD detector image of a problematic foil coil

Fig. 5 SEM and EDS analysis of pinhole causer structure

Cr-containing intermetallic structure is insensitive to homogenization annealing. But the question to be considered here is this. Did Cr mix with the liquid metal during casting operations and did it form intermetallics due to metallurgical reasons? Or is Cr mechanically embedded into the solid material and coated with intermetallics? The process carried out in order to be more detailed and to increase the accuracy of analysis in this regard is as follows: aluminum foil samples containing defects were separated from the foil itself with the help of a conductive carbon

tape and transferred to the carbon tape. Thus, the metallic foil was removed from the analysis and only the defect structure that caused pinhole formation was examined in Fig. 6. When the pinhole causer structure transferred to carbon tape was subjected to SEM–EDS analysis, it was seen that; Elemental peaks taken in aluminum foil were taken in almost the same proportions. This means that the existence of a metallurgical intermetallic can be mentioned. The idea that aluminum pollutes the analysis has been moved away.

Characterization of Cr-Bearing Intermetallics Causing Pinhole …

1037

Fig. 6 SEM and EDS analysis of pinhole causer structure on carbon band

Of course, it is not known exactly how Cr is located in the structure. For this, more advanced characterization methods are needed. But what is thought at this stage is that Cr somehow mixes with the liquid metal during the casting operation and forms an intermetallic compound. Considering that Cr comes from liquid metal, some changes were made in the casting operation. First of all, all possible Cr-containing media in contact with liquid metal were cut off from contact with liquid metal or inert coatings were applied on them. Thus, the transfer of Cr to the liquid

Fig. 7 PHD detector image of a non-problematic foil coil

metal was prevented in all casting operations. Additionally, improvements have been made to the filtration system. The cell size of the CFF filter has been increased from 60 to 80 mesh. In addition, the placement of gravel (ceramic filtering particle) and alumina balls in the DBF filter has been made more effective in filtering. After the revisions, the cast coils were rolled to a foil thickness of 7 microns. In macro darkroom observations, the presence of dense pinholes was not observed. A small number of pinhole formations were observed at the level

1038

considered normal, independent of any error factor. Additionally, when the PHD analysis of one of the produced coils was examined, it was seen that; Only low intensity pinhole areas shown in blue were observed. In addition, the number of pinholes in a single foil layer was read as approximately 4500. A dramatic decrease compared to the number of pinholes in defective coils was observed. (Fig. 7).

Conclusion • This study attempted to identify pinhole causer structures at 8079 aluminum alloy foils. • It has been observed that the particle causing the pinhole is an Al-Fe-Si-Cr based structure. • According to EDS analysis, this structure is thought to be a Cr-containing intermetallic structure formed due to metallurgical reasons. • It has emerged that the liquid metal must be kept away from any media containing Cr. • Improvements made in filtering systems directly reduce the number of pinholes in the final material. • For further research; advanced characterization techniques like XPS, FIB-SEM, and TEM should be applied in terms of a deep understanding of lattice orientation and crystal structure of existing Cr.

Y. Özçetin et al.

References 1. R. E. Sanders, (2012) Continuous Casting for Aluminum Sheet: a Product Perspective, JOM, Vol. 64, No. 2, 2012, ©2012 TMS. [2] R. E. Sanders, (2012) Continuous Casting for Aluminum Sheet: a Product Perspective, JOM, Vol. 64, No. 2, 2012, ©2012 TMS. 2. Aluminum Foil Manual, The Aluminum Association, 2004, pp. 4–8. 3. W. Chen, P. Zhao, Y. Zhou, Y. Pan, Effects of Homogenization Conditions on the Microstrucures of Twin-Roll Cast Foil Sock of AA8021 Aluminum Alloy, Materials Science Forum, ISSN: 1662– 9752, Vol. 877, pp. 296–302. 4. Ściężor W. et. al., (2016) Study of the Structural Properties and Segregation of Alloying Elements in Strips from the TRC Line, Key Engineering Materials, Vol. 682, pp. 38-45. 5. Mutharasan R., Apelian D., Romanowski C., (1987) A Laboratory Investigation of Aluminum Filtration Through Deep-Bed and Ceramic Open-Pore Filters, Physical & Mechanical Metallurgy, Vol. 33 (12) pp. 12-18. 6. R. E. Akdogan, H. M. Akdoğan, O. Birbaşar, M. Günyüz, Influence of Strip Thickness on As-Cast Material Properties of Twin-Roll Cast Aluminium Alloys, The Minerals, Metals & Materials Society 2019, C. Chesonis (ed.). Light Metals 2019, The Minerals, Metals & Materials Series, pp 1137–1141. 7. O. Keles, M. Dundar, (2007) Aluminum foil: Its typical quality problems and their causes, J. Mater. Process. Technology. 186 (2007) 125–137 8. F. Denizli. (2022) Quality Defects Metallurgical Root Cause Analysis for Aluminum Thin Foil Production, Light Metals 2022, TMS.

Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll Casting with a Steel/Copper Roll Pair Seval Aksoy Aydın, Ece Harputlu, Hikmet Kayaçetin, Cemil Işıksaçan, and Erdem Atar

Abstract

Introduction

Utilizing copper rolls in twin-roll casting (TRC) due to high process productivities provided by high thermal conductivity of copper when used as a shell material is a well-known application to fulfill the increasing material demand seasonally and EN-AW 3105 is one of those aluminum alloys, which is produced by using a steel/ copper roll pair. However, as-cast material microstructure after solidification is distinctive on the material surface which is in contact with the copper roll when compared to the other surface in contact with the steel roll. These microstructural discrepancies observed at as-cast thickness are inherited in the final product after rolling and annealing processes to a certain extent. In this study, corrosion properties of an EN-AW 3105 aluminum strip produced via TRC were scrutinized and it was concluded that the material surface, which solidified in contact with the copper roll performed better corrosion properties with respect to other surface. Keywords

Copper roll




Corrosion




Twin-roll casting

S. A. Aydın
E. Harputlu
C. Işıksaçan (&)
E. Atar Department of Materials Science and Engineering, Gebze Technical University, Gebze, Kocaeli, Turkey e-mail: [email protected]

The demand for lightweight, corrosion-resistant materials in various industrial applications has led to continuous advancements in the field of aluminum alloys, renowned for their excellent combination of strength-to-weight ratio and corrosion resistance. Among these alloys, 3105 aluminum alloy stands out for its versatility and suitability for an array of applications, ranging from the automotive industry to packaging and construction materials. The corrosion resistance of aluminum alloys is very crucial, especially in applications exposed to harsh and corrosive conditions, such as marine environments and architectural structures. Corrosion can lead to significant material degradation, structural failure, and economic losses. Therefore, a comprehensive investigation into the corrosion properties of 3105 aluminum alloy produced through twin-roll continuous casting becomes indispensable for industries relying on these materials. This study focuses on the examination of the corrosion properties of 3105 aluminum alloy produced using a steel/ copper roll pair, aiming to compare the corrosion behavior of both strip surfaces, which solidified under contact with steel and copper rolls, separately. This method, which involves the use of a steel/copper roll pair during the casting process, has gained attraction for its ability to produce aluminum alloys with enhanced casting productivity. Understanding the corrosion behavior of these specially produced materials is vital not only for ensuring their structural integrity but also for optimizing their performance in various environments.

S. A. Aydın e-mail: [email protected] E. Harputlu e-mail: [email protected]

Experimental Studies

E. Atar e-mail: [email protected]

In this study, commercial grade AA 3105 aluminum alloy was cast utilizing an industrial scale twin-roll caster with steel/copper roll pair with a strip gauge of 6 mm. The content of the alloy used in this study is listed in Table 1.

E. Harputlu
H. Kayaçetin
C. Işıksaçan Assan Alüminyum, R&D Center, Dilovası, Kocaeli, Turkey e-mail: [email protected]

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_131

1039

1040 Table 1 Chemical composition of the as-cast sample, wt%

S. A. Aydın et al. Si

Fe

Cu

Mn

Mg

Ti

0.13

0.52

0.07

0.39

0.27

0.01

As-cast strip samples were cold rolled to 0.33 mm and back-annealed in a laboratory scale furnace to obtain H26 temper. Electrochemical and immersion tests were separately applied to both surfaces for the examination of the corrosion behavior of the 3105 aluminum strip. The open circuit potential measurement was conducted in a 1 M NaCl solution according to ASTM G69 test standard. After completing the open circuit potential (OCP) measurements of the surfaces, potentiodynamic polarization tests were conducted. The potentiodynamic polarization tests were performed according to ASTM G102 standard, starting from the equilibrium potential and within a voltage range of -0.4 to + 0.4 V at a scanning rate of 0.5 mV/s in a 1 M NaCl solution. Data obtained from anodic and cathodic polarization measurements were used in this extrapolation method. The immersion corrosion tests, also known as mass loss tests, were conducted according to ASTM G31 standard. In the immersion tests, samples having standard dimensions were immersed in a 0.1 M NaCl solution for 10 days at room temperature. Measurements aiming to determine the volta potential difference between the matrix and intermetallics, which have a significant impact on the corrosion behavior of aluminum alloys, were conducted using a non-contact Park System XE7 scanning Kelvin probe force microscope (SKPFM). A cantilever made of Cr-Au material was used as the probe. SKPFM examinations were conducted at a scanning rate of 0.20 Hz over an area of 400 lm2 on the surfaces. Complementary tests were performed by electrical conductivity measurements, optical microscope, and SEM/EDS analysis.

Results and Discussion In the microstructural examinations carried out on the final aluminum strip with SEM (Fig. 1), it was determined that the sizes of intermetallic particles on the surface in contact with the copper roll were relatively much smaller compared to the intermetallic particle sizes on the surface in contact with the steel roll. During the dimensional measurements of intermetallic particles conducted on copper roll side, it was found that intermetallics generally had sizes in the sub-micron range and exhibited an irregular morphology. On the other hand, much of the intermetallics on steel roll side had sizes ranging from 1 to 2 µm and some particles were observed to have a rectangular shape (Fig. 1d).

The findings obtained from SEM examinations on the surfaces in contact with steel and copper rolls revealed that the intermetallic particles exhibited differences in size and morphology. This indicates that the surfaces of the strip solidified under different conditions due to different heat extraction capacities of the roll materials. Based on the thermal conductivity values of steel and copper rolls, the surface of the strip, which is in contact with the copper roll, solidifies faster than the surface in contact with the steel roll. As a result, intermetallics on the copper roll side are believed to remain smaller in size due to the difficulty of atomic diffusion caused by the oversaturation of the matrix by solute elements and exhibit a more homogeneous distribution in the matrix before they have the chance to grow sufficiently. This situation is believed to lead to different microstructural characteristics on both surfaces of the strip. EDS spot analysis and elemental mapping conducted on the intermetallic particles are presented in Table 2 and Fig. 2, respectively. Results indicate that intermetallics with irregular shapes are enriched in iron (Fe) and silicon (Si), while copper (Cu) and manganese (Mn) show a homogeneous distribution in the structure. It is also observed that irregularly shaped intermetallic particles on the surface in contact with the steel roll are more enriched in iron (at.% 4.50 ± 0.15 Fe) and silicon (at.% 1 ± 0.18 Si) than those found on the surface in contact with the copper roll (at.% 2.80 ± 0.14 Fe and at.% 0.80 ± 0.16 Si). It is determined that the rectangular shaped intermetallic particles on the surface in contact with the steel roll are enriched in iron (at.% 4.60 ± 0.23 Fe), bear no silicon and are assumed to be Al3Fe phase [1]. The results of elemental mapping and spot EDS analyses conducted during SEM examinations reveal that the intermetallics on the surface in contact with the copper roll consist of Al-Si-Fe, while the intermetallics formed on the surface in contact with the steel roll do not have a single chemical composition. Instead, they exhibit differences in chemical composition both within themselves and compared to the intermetallics formed on the copper roll side. These findings mentioned above can be attributed to the different cooling rates/capacities of the steel and copper rolls. Particularly, considering that iron has a very low solubility in the aluminum under equilibrium solidification conditions, it is believed that the faster cooling/solidification rate of the liquid metal on the surface in contact with the copper roll results in iron being trapped within the matrix structure (due to the difficulty of atomic diffusion) before it

Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll …

1041

Fig. 1 SEM images of the intermetallics at surface a–c copper roll side b–d steel roll side

Table 2 Chemical content of the intermetallics

At % Steel roll side

Copper roll side

Irregular particles

Rectangular particles

Irregular particles

Si

1

–

0.8

Fe

4.5

4.6

2.8

has the time to diffuse into the intermetallics, leading to a lower iron content in the intermetallics formed on the surface in contact with the copper roll. On the other hand, since the cooling rate on the steel roll side is relatively lower compared to the copper roll side, iron is able to diffuse into the intermetallics at a much higher rate. This situation has been confirmed by electrical conductivity measurements (Table 3) carried out on both surfaces. According to the results, the electrical conductivity of the surface in contact with the copper roll was measured as

27.75 ± 0.04 MS/m, while the electrical conductivity of the surface in contact with the steel roll was measured as 27.93 ± 0.03 MS/m. Based on this, it is believed that the lower electrical conductivity of the surface on copper roll side is due to the faster cooling, resulting in the matrix being more saturated with solute elements. This observation is in agreement with the previously mentioned spot EDS analyses and validates the measurement results. The potential (V)-time (s) graphs obtained from open circuit potential (OCP) measurements on both surfaces are

1042

S. A. Aydın et al.

Steel roll side

Fig. 2 EDS mapping of the intermetallics

Irregular particles

Rectangular particles

Copper roll side Irregular particles

Si

Fe

Table 3 Electrical conductivities (EC) of the strip EC, MS/m Steel roll side

Copper roll side

27.93

27.75

presented in Fig. 3. The OCP measurement was carried out by immersing the sample surfaces in a 1 M NaCl solution for 3600 s, and the potential change occurring on the surface over time was obtained relative to the reference SCE electrode. Upon examining the curves presented in Fig. 3, it was determined that the surface in contact with the copper roll had an approximate stable potential value of -0.805 V, while the surface in contact with the steel roll had an approximate stable potential value of -0.847 V. The measurement results revealed that the surface in contact with the copper roll exhibited a relatively more positive behavior, approximately 42 mV more positive, compared to the surface in contact with the steel roll. These measurements provide insights into the susceptibility or tendency of a material to corrode due to its activity. In this context, a metal with a more negative potential value, or in other words, a metal with a higher ionization tendency, is more prone to corrosion. Accordingly, the surface in contact with the steel roll having a more negative potential value indicates a higher tendency to corrode compared to the surface in contact with the copper roll.

The potential (V) and current (lA) values obtained during polarization measurements conducted after OCP measurements are graphically presented in Fig. 4. It is determined that the polarization curves of the examined strip surfaces exhibited similar behavior. Additionally, the corrosion potential (Ecorr) which is determined by using the Tafel extrapolation method carried over the polarization curves, corrosion current density (icorr) values which are obtained by dividing the corrosion current (Icorr) by the area of the surface in contact with the solution, along with the corrosion rates calculated according to Eq. 1 are provided in Table 4. Corrosion Rate ¼

K1  icorr  EW q

ð1Þ

q: density (g/cm3) K1 = 3.27  10−3, mm g/lA cm yr (constant) EW: Equivalent weight (the mass of the metal that will corrode, in grams). Based on the results obtained through the Tafel extrapolation method, it has been determined that the corrosion potential (Ecorr), corrosion current density (icorr), and corrosion rate (mm/year) values of the surface in contact with the copper roll are relatively lower compared to the surface in contact with the steel roll. These findings reveal that the strip surface on the steel roll side has a higher corrosion rate (0.022 mm/year), or in other words, lower corrosion resistance, compared to the surface in contact with the copper roll

Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll …

1043

Fig. 3 OCP results

Fig. 4 Polarization curves Table 4 Corrosion parameters and rate Ecorr (V)

icorr (lA/cm2)

Corrosion Rate (mm/year)

Copper roll side

−0.888

1.65

0.018

Steel roll side

−0.899

2.05

0.022

(0.018 mm/year). When both surfaces are compared in terms of corrosion rate, it has been determined that the surface in contact with the steel roll will undergo corrosion approximately 22% faster over the course of 1 year. Following the calculations, it has been determined that the corrosion rate values obtained for these surfaces are in line with the OCP measurements, confirming the consistency of the results.

Regarding SEM examinations conducted on the surfaces exposed to corrosion (Fig. 5), it was observed that intermetallic particles moved away locally from the matrix due to their dissolution in the areas where they reside. This phenomenon generally indicates that the Al-Si-Fe and Al–Fe based intermetallics exhibit a more cathodic behavior compared to the matrix. According to the results of research on the corrosion behavior of Al alloys in the literature, the chemical compositions and dimensions of intermetallic particles have a significant impact on the corrosion behavior of the Al alloy. They have been reported to affect the potential difference (cathodic activity) between the matrix and intermetallic particles and serve as initiation sites for pitting/corrosion attack [2]. Furthermore, it has been noted that the intermetallics with a higher iron content exhibit higher cathodic activity (relatively making the intermetallics more cathodic and the matrix more anodic) and consequently lead to corrosion in the form of pits/grooves at a much faster rate [1, 3]. In this context, as previously detailed, the intermetallic particles on the surface in contact with the steel roll are less in number but larger in size compared to the intermetallic particles on the surface in contact with the copper roll. Additionally, the higher iron content of the intermetallics (Al-Si-Fe and Al–Fe based) on steel side, leads to higher cathodic activity (increasing the micro-galvanic corrosion driving force) compared to the matrix. This is believed to result in a faster and more severe development of corrosion in the form of pits/grooves on the surface of the strip in contact with the steel roll. Cross-sectional optical microscope (OM) studies (Fig. 6) revealed that the corrosion type developing on the surfaces was pitting corrosion. Furthermore, it was observed that the pits/grooves formed on the surface in contact with the steel

1044

Fig. 5 SEM images of both surfaces after corrosion a–c copper roll side b–d steel roll side

Fig. 6 OM images of both surfaces after corrosion a copper roll side b steel roll side

S. A. Aydın et al.

Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll … Table 5 The corrosion rates calculated based on the results of the immersion corrosion tests Mass loss (g)

Corrosion Rate (mm/yr)

Copper roll side

0.0018

0.020

Steel roll side

0.0027

0.030

roll were deeper, larger, and more pronounced compared to those formed on the surface in contact with the copper roll. This indicates that corrosion on this surface occurred at a faster rate and in a more aggressive manner. The results of immersion corrosion tests conducted for a duration of 10 days (240 h) in 0.1 M NaCl solution are provided in Table 5. According to the corrosion rate (mm/yr) values calculated using Eq. 2, it was determined that the corrosion rate of the surface in contact with the steel roll (0.030 mm/year) was higher than that on the surface in contact with the copper roll (0.020 mm/yr). This result indicates that the surface on steel roll side corrodes approximately 50% faster than the surface on the copper roll side. It was also observed that the results of the immersion corrosion tests were consistent with and supportive of the results of the electrochemical corrosion tests applied to these surfaces. Corrosion Rate ¼

WK qAt

W: mass loss (g) K = 8,76  104, mm/yr (constant) q: density (g/cm3) A: surface area (cm2) t: duration of test (h).

Fig. 7 Volta potential maps a copper roll side b steel roll side

ð2Þ

1045

Volta potential maps, which help to gain insight about the driving force for micro-galvanic corrosion, were created using Scanning Kelvin Probe Force Microscopy (AFM/SKPFM) in order to provide a clearer understanding of the corrosion behavior/rates of the surfaces. AFM images of the surfaces, where volta potential measurements were performed, are shown in Fig. 7 and the averages of the measurements are presented in Fig. 8. According to the AFM/SKPFM measurements, the volta potential difference between the intermetallics and the matrix, which is the driving force for micro-galvanic corrosion, was approximately 0.275 V on the surface in contact with the copper roll. In contrast, this value was 0.438 V for the surface in contact with steel roll indicating that the driving force for corrosion was higher on steel roll side. These results revealed that the potential difference between the intermetallics and the matrix on the surface in contact with steel rolls was higher than that on the surface in contact with copper rolls. Therefore, it is observed that the high potential difference between the matrix and intermetallics, which reduces the corrosion resistance of the material or increases the micro-galvanic corrosion driving force, is supportive and explanatory of the electrochemical and immersion corrosion test results provided previously. It has been determined that the volta potential measurement results obtained on 3105 alloy are consistent with the volta potential measurements carried out on AA3003 Al alloy by research groups in the literature [2]. In the mentioned study, it has been noted that due to the more cathodic behavior of iron-rich intermetallic particles compared to the Al matrix, micro-galvanic interactions occur between the intermetallic and the matrix leading to corrosion in the form

1046

S. A. Aydın et al.

Fig. 8 Average volta potential difference values obtained by SKPFM

of pits/cavities. The results obtained from AFM/SKPFM measurements and the findings from corrosion tests have been explained together with the reasons provided below: (i) The higher iron content of intermetallics (Al–Fe-Si and Al–Fe based) due to slower solidification and the lower saturation of the matrix by solute elements (as indicated by the electrical conductivity measurements) result in a higher volta potential difference between the matrix and intermetallics on the surface in contact with the steel roll. This explains why corrosion in the form of pits/cavities starts earlier and develops more rapidly and aggressively on the surface in contact with the steel roll compared to the surface in contact with the copper roll. (ii) The surface in contact with the copper roll solidifies at higher rates compared to the surface in contact with the steel roll leading to smaller intermetallic sizes and lower iron content in the particles. Additionally, the matrix has an enhanced saturation with solute elements. This is the cause of lower volta potential difference between the matrix and intermetallics compared to the surface on steel roll side. This causes corrosion on the surface in contact with the copper roll to proceed slower.

Conclusions In this study, the effects of microstructural differences on the corrosion behavior of a 3105 aluminum alloy due to the application of a steel/copper caster roll pair have been investigated and the results obtained are summarized below:

(i) It was found that the intermetallics on the surface in contact with the copper roll are homogeneously distributed and smaller in size with a single chemical formula (Al–Fe-Si). Whereas, the intermetallics on steel roll side are coarser and have different morphologies (irregular shaped and rectangular shaped) with distinct chemical compositions (Al–Fe-Si and Al–Fe). (ii) The surface in contact with copper roll has a lower electrical conductivity (27.75 ± 0.04 MS/m) indicating a more saturated matrix with solute elements compared to that in contact with the steel roll (27.93 ± 0.03 MS/m). (iii) The surface in contact with the copper roll exhibited a more positive potential value when compared to the surface in contact with the steel roll indicating a more noble behavior. (iv) The corrosion rate of the surface in contact with the copper roll was found to be slower compared to the surface in contact with the steel roll. (v) The surface in contact with the copper roll exhibited less and shallower pitting after corrosion tests. (vi) According to AFM/SKPFM results, it was observed that the potential difference between the intermetallics and the matrix on the surface in contact with the copper roll was lower than that on the surface in contact with the steel roll. Considering all these findings for 3105 aluminum alloy produced by a steel/copper roll pair, it is evident that the surface in contact with the copper roll exhibits higher corrosion resistance compared to the surface in contact with the steel roll due to superior thermal conductivity of copper leading to higher cooling rates on that surface.

Corrosion of EN-AW 3105 Aluminum Strip Produced via Twin-Roll …

References 1. Ambat R., Davenport A.J., Scamans G.M., Afseth A., (2006), Effect of Iron-Containing Intermetallic Particles on the Corrosion Behaviour of Aluminium, Corrosion Science, 48, 3455–3471. 2. Verkens D., Revilla R.I, Huizenga R., Marcoen K., Günyüz M., Işıksaçan C., Terryn H., Graeve I.D., (2021), Effect of Sr Addition

1047 to a Modified AA3003 on Microstructural and Corrosion Properties, Journal of the Electrochemical Society. 3. Zamin M., (1981), The Role of Mn in the Corrosion Behavior of Al-Mn Alloys, National Association of Corrosion Engineers, 37 (11), 627-632.

In Situ Experimental Study of the Nucleation and Growth of Fe-Al Based Intermetallics: An Insight for Designing Next-Generation Recycling Friendly Aluminium Alloys G. Salloum-Abou-Jaoude, K.-H. Cheong, S. Akamatsu, Ph. Jarry, and S. Bottin-Rousseau

Abstract

Introduction

Today’s focus in the aluminium industry is promoting a circular economy, thus meeting the US and the EU's ambition of reducing greenhouse gas emissions. This means replacing primary aluminium with post-consumer scrap, which contains larger quantities of intermetallic forming impurities, particularly Fe and Si, which are detrimental to mechanical properties. We need to develop new generations of impurity-tolerant aluminium alloys enabling reduced carbon footprint through increased recycled content. To meet this objective, it is necessary to deepen our understanding of AlFeSi secondary phase nucleation and growth during aluminium alloy solidification. In this work, model 6xxx recycling-friendly alloys were manufactured and unique in situ directional solidification experiments were performed. For the first time, nucleation and growth of a-AlFeSi and b-AlFeSi intermetallics were observed in real time in thin samples of model 6xxx alloys. We discuss the effect of chemistry modification and solidification parameters on the AlFeSi phase nucleation and growth. Keywords








Solidification Aluminium alloys Fe-intermetallics Recycling




In situ investigation

G. Salloum-Abou-Jaoude (&)
K.-H. Cheong
Ph. Jarry Constellium Technology Center C-TEC, Parc Economique Centr’alp, 725 Rue Aristide Bergès, CS10027 Voreppe, France e-mail: [email protected] S. Akamatsu
S. Bottin-Rousseau Institut Des NanoSciences de Paris, Sorbonne Université, CNRS UMR 7588, Case Courrier 840, 4 Place Jussieu, 75252 Paris Cedex 05, France

Aluminum alloys are highly sought after by automakers thanks to their low density, corrosion resistance, and excellent formability [1]. Such demand for alloys with exceptional forming properties has led to the development of high-end aluminum products using substantial proportions of primary aluminum from smelters. By using primary aluminum, the level of impurities such as iron (Fe) and silicon (Si) in the alloy can be greatly reduced. Such low levels of impurities do not prevent the formation, but reduce the volume fraction of Fe-based intermetallic compounds (Fe-IM), which are detrimental to the mechanical properties of the alloy. Today, the goal of the aluminum industry is to reduce its carbon footprint, and one way to achieve it is by recycling post-consumer scrap. The energy consumption of remanufacturing recycled Al alloys represents only 6.2% of that of producing Al alloys from pure Al [2]. Such a drastic difference in energy consumption leads to the urge to increase the recycling share during alloy production. However, residual elements such as Fe and Si accumulate during recycling and encourage the formation of Fe-IM, which leads to lower formability-related properties unsuitable for premium automobile body sheets. Since the presence of increasing levels of Fe is probably inevitable, several methods can be applied to neutralize its negative effects. Solidification parameters along with chemical additions can modify the Fe-IM nature and morphology [3]. A well-known chemical modification method is the addition of Manganese (Mn), which promotes a–Al(Fe, Mn)Si (Chinese script and/or compact shape) instead of the brittle b–AlFeSi (needle shape) Fe-IM [1]. Such alloying addition has been widely reported to be able to enhance the mechanical properties of Fe-IM containing alloys [1, 4, 5]. In order to develop new recycling friendly alloys, the understanding of the nucleation and growth of secondary Fe-IM in AlFeSi alloys needs to be deepened. With such knowledge, one could design alloys that target the formation

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_132

1048

In Situ Experimental Study of the Nucleation and Growth …

of a-Fe-IM instead of b-Fe-IM during solidification or apply a precise set of parameters to generate an Fe-IM morphology that is less detrimental to the final mechanical properties of the manufactured product… In situ monitoring of direction solidification seemed to be the best way forward to provide direct real time information on the phenomena acting on Fe-IM phase selection, nucleation, and growth. For several decades now, in situ, X-ray radiography of directional solidification experiments has proven to be the method of choice for metallic alloys. It has been successfully applied to Al-based alloys with relatively large (  4 wt%) concentrations of Cu [6, 7] and Ni [8], among other systems. In those alloys, a high contrast between the solid and the liquid is provided by a large difference in the atomic numbers between Al and the redistributed solutes (Beer Lambert law) [9]. Unfortunately, 6xxx alloys mainly contain light elements such as Si and Mg. This entails a faint X-ray absorption contrast, thus minimizing the chances of success for real-time imaging during solidification. In a recent study by Bottin-Rousseau et al. [10], it has been shown that eutectic growth patterns can be observed in real time by regular optical microscopy during the directional solidification of the binary Al-Cu system in thin samples. The (about 10-µm thick) metallic film is prepared by a PVD-like method (plasma sputtering). The reflected-light mode imaging can be made in situ, without any further preparation such as polishing or etching. The apparatus was designed on the model of previous thin-sample directional solidification setups utilized for lowmelting transparent and metallic alloys [11–13]. This setup showed great potential. Given that the apparatus has only been applied to a specific metallic binary alloy with regular eutectic morphology and high secondary phase fraction which are relatively easy to observe, this work aims to evaluate the feasibility of the study on industrially relevant model alloys. These alloys contain at least three alloying elements with a lower fraction of secondary phases. By observing the solidification in real time, one might be able to better understand the conditions for nucleation and growth mechanisms of Fe-IM, thus helping in designing next-generation recycling friendly alloys.

1049

Fig. 1 Flow chart of the processes leading up to directional solidification

Identification of the Composition In this study, the alloy composition was identified by using thermodynamic calculations from FactSage© using FTlite database, which is an Integrated Thermodynamic Databank System. The starting point for a base alloy was the composition range of an industrial 6xxx alloy. The aim was to identify a composition that contains the same intermetallic phases that are present in industrial 6xxx alloys. We can aim for example to have a-Fe-IM as a secondary phase during solidification. We present in Fig. 2 the identification of two ternary model alloys C1 and C3. In this manuscript, we will focus on showing the results from ternary eutectic point sample C3. Figure 2 shows the solidification path of C1 and C3 on a liquidus projection of pseudo-ternary Al-Fe-Si diagram calculated using FactSage. The composition of C3 (Al—12.8 wt% Si—0.7 wt% Fe) is at the ternary eutectic point (3), which is the intersection of three cotectic curves as shown in Fig. 2. This is the ternary eutectic reaction when cooling a liquid with the composition of C3: ð3Þ: at T ¼ 577:24  C Liquid ! Al þ Si þ b-AlFeSi

Materials and Methods This section is dedicated to describing the processes leading to the in situ directional solidification for the observation of nucleation and growth of Fe-IM in Al alloys. Figure 1 shows the preparation and processing route leading to in situ directional solidification experiments of industrially relevant model alloys. The details of the steps will be elaborated in the following subsection.

Several objectives can be attained by using C3 as a reference: (1) Establishing the feasibility of in situ observation of the solidification of a complex system (not binary alloy). (2) Observing the intermetallic phase selection sequence and comparing it with existing thermodynamic databases.

1050

G. Salloum-Abou-Jaoude et al.

Fig. 2 Liquidus projection of pseudo-ternary Al-Fe-Si diagram and solidification path of C1 (red), C3 (blue). 1–2, 2–3: Monovariant lines, eutectic valleys. 3: Ternary eutectic point

(3) Identifying the physical parameters affecting intermetallic phase nucleation and growth in order to better understand and eventually control their size distribution during casting. At a later stage, comparing C1 and C3 could help to identify the morphology of a-AlFeSi that will be obtained during the solidification of C1, given that a-AlFeSi is not supposed to be present during the solidification of C3 samples (according to thermodynamic calculations, Fig. 2).

Thin Sample Manufacturing Via Magnetron Sputtering Thin Al–Fe-Si films were deposited on clean sapphire substrates via magnetron sputtering in a clean room. Pure Al, Fe, and Si sources (99.99 mol%) were used as targets. Figure 3 shows the assembled sample for directional solidification. The nominal total thickness for the Al-Fe-Si was 10 µm. Al was deposited first, to ensure good wetting, followed by Fe and Si. The ratio between the thickness of the layers was determined so that the composition of the film after melting corresponds to the desired value (C1 or C3 per example). After sputtering, another sapphire plate was pressed onto the free surface of the Al-Fe-Si film and both plates were glued together along the edges using ResbondTM 908, a heat

resistant alumina-based adhesive. The assembly was done in a glovebox under a protective Ar atmosphere. The use of sapphire substrates and the gluing were to prevent the de-wetting in the liquid state.

In Situ Bridgman Solidification Experiments The directional-solidification apparatus was designed and built in-house [10] in order to observe solidification in situ and in real time. It was extensively used to observe eutectic growth in thin Al-Al2Cu regular eutectic samples. The same setup was used in this work to observe the nucleation and the growth of Fe-IM in industrially relevant model alloys. The setup is a Bridgman directional solidification furnace placed under a microscope connected to an image acquisition device (Fig. 4).

Results: Effect of Pulling Speed on Intermetallic Morphology In this section, we focus on in situ observations during the directional solidification of a thin Al—12.8 wt% Si—0.7 wt % Fe alloy (C3). This composition corresponds nominally to the invariant ternary eutectic point (Eq. 3). The sample was inserted in the directional solidification module and the Hot/Cold blocks were heated to 630/530 °C

In Situ Experimental Study of the Nucleation and Growth …

1051

Fig. 3 a Photo of an assembled Al-Fe-Si sample, top view. b Schematic illustration of Al-Fe-Si sample, side view

Fig. 5 Growth of an isolated Si crystal (large black crystal) after nucleation in the liquid 5 min after velocity jump from 1 to 5 µm/s. Small dots in the liquid are bubbles Fig. 4 Schematic illustration of the working principles of the in situ directional solidification setup. z: axis of translation of the sample and thermal gradient; x and y axis are in the isothermal plan

by increments, establishing a temperature gradient between the heaters of about 10 K/mm. The aim is to place the solid/liquid interface at the center of the field of view (FoV). After a period of stabilization/homogenization, the thin sample was pulled along the z direction from the hot block to the cold block at various speeds (V = 1, 5, 0.2, 0.1, 10 µm/s) to investigate the effects of V on the microstructure and to familiarize with the optical observation of different intermetallic phases. The first phenomenon we observed is that Si crystals nucleated first (black facetted particle on the top left corner of Fig. 5) during the early stages of solidification at higher temperatures than the rest of the microstructure. In Fig. 5, the nucleation and growth of a new Si crystal occurred in the liquid, ahead of the coupled-growth front we see at the bottom half of the picture, a short time after a

velocity jump from 1 to 5 µm/s. This suggests that the actual composition of the thin sample is slightly off-eutectic (on the Si-rich side). This observation has been confirmed on several occasions throughout the experiments. Three solid phases were clearly identified throughout the directional solidification experiments of sample C3. We expect from our thermodynamic calculations (Eq. 3) to have at T  577 °C simultaneous growth of the ternary eutectic containing 3 phases: • FCC-Al • Diamond-Si • b-AlFeSi Thanks to the optical observation capability of our experimental setup, it was relatively easy to recognize the different phases during directional solidification. Figure 6 presents a comparison of the in situ observed microstructure (a) with conventional metallurgically prepared bulk model

1052

6xxx alloy (b). We can clearly identify the Si phases that appear in black and b-AlFeSi that have faceted plate like morphologies and appear in grey (blue arrows in Fig. 6). According to our thermodynamic predictions, in our C3 sample, we do not expect to have a-AlFeSi, and indeed we confirm that no a-AlFeSi (red arrows in Fig. 6 b) appears to be present in the microstructure. One should also be precise that the phase that appears in white in Fig. 6 a, b is the FCC-Al. This is also consistent with metallographic observations of a typical 6082 industrial alloy [14]. In Fig. 7 we present the microstructure obtained at various solidification velocities, namely V = 5, 0.2, and 0.1 µm/s (in chronological order). As shown in Fig. 6 a, the three solid phases (FCC-Al, Diamond-Si, and b-AlFeSi) are identified using the optical contrast. All three phases are present and grow simultaneously at the solidification front, thus indicating an essentially coupled growth regime. Based on our measurements of the temperature profile along z, the position of the solidification front corresponds to T  580 ± 5 °C. Considering the experimental uncertainties, our observations are consistent with the thermodynamic prediction of  577 °C (see above). In Fig. 7 we can clearly see the effect of pulling speed on the ternary eutectic morphology. With lower pulling speed,

G. Salloum-Abou-Jaoude et al.

the eutectic spacing gets larger and the irregular eutectic morphologies get thicker. One can try to measure the spacing k at a certain velocity between same nature phases. The spacings were taken at an area where eutectic growth is relatively regular and periodic. This was done by counting the number of phases present at an isotherm in the selected area divided by the width of the area. Spacings between b-AlFeSi crystals could not be taken as a reference given their scarcity due to their lower phase fraction in comparison to Si. For the sake of demonstration, we report here the inter Si irregular eutectic spacing kSi at different pulling velocities: • V = 0.1 µm/s: kSi 2 [100, 500 µm] • V = 1 µm/s: kSi  50 ± 20 µm • V = 5 µm/s: kSi  8 ± 1 µm The morphologies of the different phases are clearly identifiable in Fig. 7; black phases are Si, grey phases are b-AlFeSi, and the white is FCC-Al. Most of the Si and b-AlFeSi crystals exhibit partly faceted shapes, more or less tilted with respect to the growth direction. Two kinds of morphologies can be observed: highly branched, seaweed-like for Si crystals (right and left side of Fig. 7) and needle-like morphology for b-AlFeSi and Si crystals (throughout the sample). Facets correspond to deep minima of the surface free energy c, and possibly to strongly non-linear kinetic effects. This means that the surface energy anisotropy is strong. The crenelated shapes observed in particular in Fig. 7 at V = 0.1 and 0.2 µm/s may be linked to the existence of so-called forbidden orientations. Such morphology has already been observed in previous studies carried out using rotating furnace apparatus on In-Bi [15] and Al-Cu systems [16]. The faceted growth also goes along with a very irregular microstructure, with spacings between Si crystals, taken as an empirical reference characteristic, ranging from 100 to 500 µm as mentioned previously.

Conclusions In this work, a novel experimental approach is used aiming at observing the nucleation and growth kinetics of Fe and Si intermetallics in model aluminium alloys. The feasibility of the idea of monitoring in situ and real time the nucleation and the growth of Fe-IM in industrially relevant model Al alloys was confirmed and we conclude that: Fig. 6 a Detail of the three-phase growth pattern of C3 sample, V = 1 µm/. b Optical micrograph of an as-cast 6xxx alloy. Blue arrows point at plate like b-AlFeSi, Red arrows point at Chinese script like a-AlFeSi. The black facetted phases are Si phases

• The apparatus provided satisfactory spatial and temporal resolutions allowing to study early stages of nucleation and the subsequent growth kinetics of different Fe and Si based intermetallics.

In Situ Experimental Study of the Nucleation and Growth …

1053

Fig. 7 In situ panorama of the solidification front and the microstructure of C3 at V = 5, 0.2 and 0.1 µm/s at a fixed temperature gradient of 10 K/mm

• Thermodynamic predictions calculated using FactSage, FTlite database were consistent with our experimental observations regarding the temperature of the solidification front temperature and the presence of FCC-Al and two other phases. • Further comparison of the images with conventional metallography showed contrasts consistent with previous observations of the FCC-Al, Diamond-Si and b-AlFeSi phases. • Reducing the pulling speed during solidification, increased the eutectic spacing and generated thicker eutectic structures more or less tilted with respect to the growth direction. • Two kinds of morphologies were observed: highly branched, seaweed-like for Si crystals and needle-like for b-AlFeSi and Si crystals. Outlook: • Future phase composition quantification and crystallographic characterization are needed to enable us to confirm the nature of the phases observed. This may also help us explain the change in growth morphologies observed during solidification. • This apparatus is currently being used to study the effect of Chemical additions and/or solidification parameters on intermetallic phase selection and morphology modification. Since C1 and C3 are ternary Al-Si-Fe alloys, they can serve as references for studies on the effect of Mn and other elements on Fe-IM phase selection and growth. • The ultimate goal of this work is to be able to use in situ capability to monitor in real time the solidification of high impurity model 6xxx alloys. This will strengthen our fundamental understanding on the governing phenomena controlling secondary phase nature selection, nucleation,

and growth morphology modification, giving insights for designing next-generation recycling friendly aluminium alloys.

References 1. M. H. Khan, A. Das, Z. Li, and H. R. Kotadia, “Effects of Fe, Mn, chemical grain refinement and cooling rate on the evolution of Fe intermetallics in a model 6082 Al-alloy,” Intermetallics, vol. 132, p. 107132, May 2021, doi: https://doi.org/10.1016/j.intermet.2021. 107132. 2. H. Lu, Z. Hou, M. Ma, and G. Lu, “Effect of Fe-Content on the Mechanical Properties of Recycled Al Alloys during Hot Compression,” Metals, vol. 7, no. 7, p. 262, Jul. 2017, doi: https://doi. org/10.3390/met7070262. 3. Z. Que, Y. Wang, and Z. Fan, “Formation of the Fe-Containing Intermetallic Compounds during Solidification of Al-5Mg-2Si-0.7Mn-1.1Fe Alloy,” Metall. Mater. Trans. A, vol. 49, no. 6, pp. 2173–2181, Jun. 2018, doi: https://doi.org/10.1007/ s11661-018-4591-6. 4. W. Zhang, B. Lin, P. Cheng, D. Zhang, and Y. Li, “Effects of Mn content on microstructures and mechanical properties of Al-5.0Cu-0.5Fe alloys prepared by squeeze casting,” Trans. Nonferrous Met. Soc. China, vol. 23, no. 6, pp. 1525–1531, Jun. 2013, doi: https://doi.org/10.1016/S1003-6326(13)62626-6. 5. S. Shabestari, “The effect of iron and manganese on the formation of intermetallic compounds in aluminum–silicon alloys,” Mater. Sci. Eng. A, vol. 383, no. 2, pp. 289–298, Oct. 2004, doi: https:// doi.org/10.1016/S0921-5093(04)00832-9. 6. G. Salloum-Abou-Jaoude, G. Reinhart, C. Hervé, M. Založnik, T. A. Lafford, and H. Nguyen-Thi, “Quantitative analysis by in situ synchrotron X-ray radiography of the evolution of the mushy zone in a fixed temperature gradient,” J. Cryst. Growth, vol. 411, pp. 88–95, Feb. 2015, doi: https://doi.org/10.1016/j.jcrysgro.2014. 10.053. 7. L. Abou Khalil, G. Salloum-Abou-Jaoude, G. Reinhart, C. Pickmann, G. Zimmermann, and H. Nguyen-Thi, “Influence of gravity level on Columnar-to-Equiaxed Transition during directional solidification of Al – 20 wt.% Cu alloys,” Acta Mater., vol.

1054

8.

9.

10.

11.

12.

110, pp. 44–52, May 2016, doi: https://doi.org/10.1016/j.actamat. 2016.03.007. H. N. Thi et al., “Preliminary in situ and real-time study of directional solidification of metallic alloys by x-ray imaging techniques,” J. Phys. Appl. Phys., vol. 36, no. 10A, p. A83, Apr. 2003, doi: https://doi.org/10.1088/0022-3727/36/10A/317. G. Salloum Abou Jaoude, “In situ investigation by X-ray radiography of Microstructure Evolution during Solidification of Binary Alloys,” These de doctorat, Aix-Marseille, 2014. Accessed: Sep. 06, 2023. [Online]. Available: https://www.theses.fr/ 2014AIXM4351 S. Bottin-Rousseau et al., “Locked-lamellar eutectic growth in thin Al-Al2Cu samples: in situ directional solidification and crystal orientation analysis,” J. Cryst. Growth, vol. 570, p. 126203, 2021, doi: https://doi.org/10.1016/j.jcrysgro.2021.126203. S. Akamatsu, S. Bottin-Rousseau, and G. Faivre, “Determination of the Jackson–Hunt constants of the In–In2Bi eutectic alloy based on in situ observation of its solidification dynamics,” Acta Mater., vol. 59, no. 20, pp. 7586–7591, Dec. 2011, doi: https://doi.org/10. 1016/j.actamat.2011.08.036. S. Akamatsu, S. Bottin-Rousseau, M. Perrut, G. Faivre, V. T. Witusiewicz, and L. Sturz, “Real-time study of thin and bulk

G. Salloum-Abou-Jaoude et al.

13.

14.

15.

16.

eutectic growth in succinonitrile–(d)camphor alloys,” J. Cryst. Growth, vol. 299, no. 2, pp. 418–428, Feb. 2007, doi: https://doi. org/10.1016/j.jcrysgro.2006.11.271. V. T. Witusiewicz, U. Hecht, and S. Rex, “In-situ observation of eutectic growth in Al-based alloys by light microscopy,” J. Cryst. Growth, vol. 372, pp. 57–64, Jun. 2013, doi: https://doi.org/10. 1016/j.jcrysgro.2013.02.033. G. Mrówka-Nowotnik, J. Sieniawski, and M. Wierzbińska, “Intermetallic phase particles in 6082 aluminium alloy,” Arch. Mater. Sci. Eng., no. Vol. 28, nr 2, pp. 69–76, 2007. S. Akamatsu, S. Bottin-Rousseau, M. Şerefoğlu, and G. Faivre, “Lamellar eutectic growth with anisotropic interphase boundaries: Experimental study using the rotating directional solidification method,” Acta Mater., vol. 60, no. 6, pp. 3206–3214, Apr. 2012, doi: https://doi.org/10.1016/j.actamat.2012.02.033. M. Medjkoune, “Étude expérimentale des effets d’anisotropie interfaciale en solidification directionnelle d’alliages eutectiques Al-Al2Cu en échantillons minces,” These de doctorat, Sorbonne université, 2022. Accessed: Sep. 06, 2023. [Online]. Available: https://www.theses.fr/2022SORUS016

Measurement of the Heat Transfer in the Primary Cooling Area of a Laboratory Direct Chill Casting Plant for Alloy Design Andreas Weidinger, Sebastian Samberger, Florian Schmid, and Stefan Pogatscher

Abstract

Keywords

The thermal transfer performance in the primary cooling area during direct chill casting plays a crucial role in the microstructure formation process and thus exerts a major impact on the mechanical properties of cast aluminium alloys. Understanding the influence of cooling rate on phase formation is essential for optimizing alloy design. This study presents a comprehensive analysis of heat transfer based on experimental results and its impact on phase formation during the solidification process of aluminium alloys. Measurements were conducted on a laboratory-scale direct chill casting plant (Indutherm VCC3000), with a focus on the primary cooling area. Therein, cooling gradients were determined using five thermocouples placed inside the graphite-mould walls and three implemented in the starter block. Tests were conducted as block casting and no evaluation of the casting speed was possible. Based on the obtained results, the influence of heat transfer on the cooling gradient of the investigated system is discussed. In this regard, the research provides valuable insights for future studies into the correlation between heat transfer, cooling rate, and phase formation during the solidification of aluminium alloys in a laboratory direct chill casting plant. The findings contribute to the optimization of alloy design, facilitating the production of aluminium alloys with tailored microstructures and improved mechanical properties.

Aluminium alloys Production routes Direct chill casting Temperature measurement Heat transfer Thermocouples Lab scale

A. Weidinger (&)
S. Samberger
S. Pogatscher Christian Doppler Laboratory for Advanced Aluminum Alloys, Chair of Nonferrous Metallurgy, Montanuniversitaet Leoben, Franz-Josef-Straße 18, 8700 Leoben, Austria e-mail: [email protected] F. Schmid AMAG Rolling GmbH, Postfach 32, 5282 Ranshofen, Austria

















Introduction State-of-the-art direct chill (DC) casting process is a semicontinuous process widely used in the production of nonferrous metals such as Aluminium, Copper, Lead, and their alloys [1]. In the DC casting process, molten metal is poured into a water-cooled mould, where it is rapidly cooled and solidified. Initially, after the contact with the walls, the material pulls away from the mould when it solidifies due to thermal shrinking. This procedure creates an air gap between the material and mould and reduces the heat transfer towards the mould [2]. At the bottom of the mould, the water leaves the primary cooling zone and enters the secondary cooling of the strand. Here, the cooling medium is in direct contact with the strand surface. This process step removes 80–95% of the heat content from the cast product, as shown by various studies [3, 4]. The primary cooling zone and the formation of a stable shell is one of the limiting factor in terms of casting speed and the formation of surface defects such as cold shots. The secondary cooling zone, on the other hand, is one of the critical factors for the final casting quality [5–7]. Heat transfer rate during the primary cooling phase of the DC casting process is a critical factor in determining the final properties and quality of the ingot. The primary cooling stage in this casting process is characterized by two main heat transfer mechanisms namely convection and conduction. Convection heat transfer occurs at the interface of metal to mould, where the molten metal is cooled by the flowing water. Conduction heat transfer occurs through the mould wall, as the heat from the molten metal is transferred through the walls to the water in the cooling channels [8–10].

© The Minerals, Metals & Materials Society 2024 S. Wagstaff (ed.), Light Metals 2024, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-031-50308-5_133

1055

1056

A. Weidinger et al.

Baserinia et al. [11] showed a simple density-based model and a thermoelastic deformation model to analyse the DC casting process with a main focus on air gap formation by the use of a computational fluid dynamic analysis (CFD). Regarding the fundamentals of these two models, they took the following heat transfer coefficients into account: • such as the heat conduction inside the formed ingot to the air gap (kAlu), • the heat conduction and transfer through the air gap that can form between the ingot shell and the inner mould surface (hGas kGas), • the heat conduction through the mould (kGp, km) • and the heat conduction and transfer from the cooler into the water (hw, cpw). These factors together form the effective heat transfer coefficient for the primary cooling (hpc) among the ingot surface into the primary cooling water. The reciprocals of the conductivities specified lead to the thermal resistances (Rpc) [6, 11, 12]. Taking all these influencing parameters into consideration the heat flux in general for the primary cooling zone (Q_ pc Þ can be therefore calculated according to the following Equitation (1). Q_ pc ¼ hpc
ðT0  TW Þ

ð1Þ

Here, (T0) relates to the ingot surface temperature and (TW) represents the temperature of the cooling water running through the cooler channels which are displayed in Fig. 1 [11–13]. The difference between the ingot surface T0 and the primary water-cooling TW which is based on the individual

Fig. 1 Temperature profile with transitions between the surface of the strand and the cooling water of the cooling unit redone after [14]

resistance is shown as the temperature profile in Fig. 1. Depending on the mould geometry, the heat flux densities in the thermal boundary layer of the water and in the mould wall are expressed differently [14]. The individual heat flows have a significant influence on the solidification process during DC casting, depending on several factors, such as the used alloy, the design of the mould (conical, parallel), and the selected process parameters such as casting temperature and speed. The last two determine the thickness of the solidifying shell and the duration that the molten metal is in contact with the mould. Furthermore, the final ingot geometry affects the rate of heat transfer, as it determines the surface area that is in contact with the mould. The amount of heat dissipated per time correlates with the flow rate and the temperature difference between the inlet and outlet temperature of the cooling water. Each alloy has different properties in the liquid state, such as its thermal conductivity and specific heat, which affect the rate of heat transfer. The fundamental task of the mould and the primary cooling system is to extract enough heat from the incoming melt to form a stable outer shell [12, 15]. To evaluate all parameters appropriately it is of great interest to address the heat transfer in the primary cooling area of a laboratory DC casting plant by measuring the temperature in the intermediate area. This makes the real-time monitoring of the temperatures during the casting process between the inner surface of the mould and the cooling water channels reasonable. First, it can be used to optimize the casting process, by identifying the factors that affect the rate of heat transfer and by determining the optimal casting conditions. Second, it can be used to design new alloys, by understanding the relationship between the alloy composition and the rate of heat transfer.

Measurement of the Heat Transfer in the Primary Cooling Area …

Experimental Methods Based on findings from previous studies [16, 17], the experimental methods are extended in this research to assess the heat transfer within the primary cooling region of a laboratory DC casting plant. Subsurface thermocouples were utilized to record temperature variations at distinct depths beneath the liquid meniscus. These temperature measurements serve as the foundation for calculating the cooling gradients at the interface between the mould and the metal. An advantage of this approach is its capacity to facilitate measurement of temperature gradients within the mould walls without the need for cast-in thermocouples (TC) in the final ingot.

Experimental Set-Up Figure 2a illustrates the used laboratory direct chill casting plant for this study from Indutherm (VCC3000). Figure 2b schematically presents the structure of the primary cooling in its laboratory use. As stated in prior work [16, 17], the current state of the aggregate represents already an adaptation (No. 2 in Ref. [16, 17]). The adapted continuous casting plant is instrumented and evaluated further in this work. The alloy chosen for casting is an industrial grade EN-AW-5083 (0,3% Si, 0,3% Fe, 0,1% Cu, 0,5% Mn, 4,8% Mg) with a selected casting temperature of 720 °C (*100 K above Tliq.). Prior to initiating the casting process, the starter block with mounted TC’s (K 1, 2, 3) had to be inserted from below the mould, ensuring correct placement up to specified markings for all

Fig. 2 Illustration of the Indutherm VCC3000 vertical continuous caster (a), Schematic overlook of the primary and secondary cooling used in the vertical DC casting process (b)

1057

tests. The positioned equipment is secured with the assistance of the pressure rollers (as shown in Fig. 1a). No secondary cooling with water was used during any of these experiments. The test was conducted at a water flow rate for primary cooling of approximately 0.03 kg/s (1.8 L/min). Data acquisition involved the use of a high-speed USB measuring device (USB-2408-2AO) equipped with eight analog input channels connected to the TC’s in the mould wall (TC 1, 2, 3, 4, 5) and TC’s in the casted block (K 1, 2, 3). The software LabView was used to record and store this data. A data acquisition frequency of 50 Hz has shown the lowest noise at highest acquisition rate. After programming the casting parameters, the experimental protocol was as follows: 1. Three times argon purging of the furnace chamber, crucible, and mould. Because the whole melting and casting system cannot be closed due to the experimental set-up, a vacuum atmosphere was not possible. 2. Independent heating of the material in the crucible up to the selected melting temperature. 3. Simultaneous initiation of the temperature recording programs. 4. Holding the melt at the required temperature for 10–15 min. 5. Opening the stopper for a calculated duration to fill the mould completely with melt. 6. Closing the sealing plug after the specified time has elapsed. 7. Shutting off the induction heating, allowing residual material in the crucible and cast ingot to solidify in the mould.

1058

8. Before removing the cast block, stopping the temperature recording and withdrawing the thermo-cables from the casted block. The generated samples have a cross-section of 30  110 mm and a defined length of 14 cm. The experimental set-up involves the strategic placement of five 0.5 mm thermocouples (TC, Type-K, stainless steel sheathed) within the mould walls, enabling the monitoring of the thermal history close to the solid shell. Situated on different height positions, the captured data is related to different heat transfer mechanisms such as the heat transfer in the liquid sump (TC 1, 2), the transition area when the strand starts to separate from the mould (TC 3, 4) and the heat conduction towards the primary cooling and the convective heat transfer (TC 5). Figure 3 schematically shows the positions of the five TC, marked with red dots. To ensure the correct positioning and precision, the TC’s were inserted from the bottom of the mould through the grinded channels and were positioned 2 mm away from the inner mould surface. Additionally, an optical pyrometer positioned 90 mm above the mould bottom was used to measure the exterior mould temperature (not pictured in Fig. 3). The temperature difference in the primary cooling water was continuously monitored via two thermocouples installed in the inlet and outlet of the pipe (not shown in Fig. 3). To complement the measurement of the total heat flux from the middle of the mould in the liquid material towards

A. Weidinger et al.

the primary cooling water, three TC (K 1, 2, 3) marked in Fig. 4a (red dots), were installed in the starting block. These thermocouples were inserted from beneath the starter block into protective Al2O3 tubes. This ensures consistent measuring heights and reusability of the used TC across tests. The selected measuring heights correspond to those in the mould wall (see Fig. 3). Figure 4 presents the measurement set-up. From the prepared measurement block in the mould cavity (a), to the data acquisition by means of thermo-cables and recording by using LabView (b). In addition, the inlet and outlet cooling water temperature were extracted from the data measured by the furnace Indutherm VCC3000.

Results and Discussion The experimental results indicate that the primary cooling rate of the DC casting process varies significantly within the height of the mould. For this study, the focus relied on one pouring temperature of the alloy, which was defined as 100 K above the melting temperature. All the other parameters are set to a fixed value during the process of block casting. Figure 5 shows the time-based thermal profile of the cooling curves on behalf of the Adaption No. 2 [16, 17] for the pouring temperature of 720 ± 7 °C. In all the figures the origin of the time axis (x-axis) marks the start of the casting

Fig. 3 Schematic of the mould with the five arranged TC's positioned in 2 mm distance from the mould inner surface to measure the live temperature during the casting. Measurement of mould temperature at 60 mm, 95 mm, and 120 mm height from the bottom

Measurement of the Heat Transfer in the Primary Cooling Area …

1059

Fig. 4 Measurement set-up of the prepared starting block schematical placed in the mould cavity (a), Whole set-up to measure the temperature of the five TC in the mould wall (TC 1, 2, 3, 4, 5) and the three TC in the casted block (K 1, 2, 3) (b)

(opening stopper) when the liquid metal was released into the mould. When using a TC Type-K in a time-critical process it is of utter importance to take the standard deviation into account, as a drift of the Seebeck-coefficient, caused by atomic structure fluctuation during temperature changes, may cause false temperature impressions. Consequently, it is essential to include a standard deviation to the determined temperature considerations. In accordance with EN 60,854-1:2013 [18], Type-K thermocouples of class 2 show a deviation of ±0.75%. However, as this standard aims for unused thermocouples, this study chose a margin of ±1% by any considering of the data [18, 19]. Figure 5a displays all eight recorded datasets over the cooling process at the chosen casting temperature. For a better overview and representation of the cooling process, Fig. 5b includes TC 3, 4, and K 2 respectively to display the

middle part of the investigated system (compare Fig. 4a). In Fig. 5c, the constant primary cooling water inlet temperature of 18 °C is shown in combination with the varying water outlet temperature. In addition to this the mould outside temperature measured by the machine on a similar height as TC 2 and 3 with an optical pyrometer is plotted. To avoid detrimental effects caused due to a non-steady state in block casting it is crucial to divide the temperature profiles into three areas of interest. Therefore, inflection points can be used to identify different stages of the cooling process and to separate the data into areas that can be analysed separately. Area one starts with the beginning of the block cast and ends after 140 s, when the cooling ingot begins to form a first air gap. In this phase, the highest cooling rate occurs in the middle part of the mould because heat can be dissipated ideally due to an optimal fit to the

Fig. 5 Profile of the cooling curves at a casting temperature of 720 ± 7 °C for the whole set-up of eight TC (a). For better visibility, the cooling curves for TC 3, 4, and K 2 that represents the middle part of the system are depicted individually (b) as also the primary cooling

water inlet and outlet temperature plus the outside temperature of the mould (c). For all the figures, zero point on the time axis marks the start of the casting

1060

A. Weidinger et al.

Table 1 Overview of the cooling gradients calculated according to Eq. (2) in the different temperature ranges based on time segments from the start during the block casting with a fixed pouring temperature of 720 ± 7 °C. Additionally, it is the maximum heat flux of the different positions and the effective primary cooling thermal resistance (Rpc) of the system shown Cooling gradient [K/s] Pos

0–140 s

140–500 s

500–1000 s

Max

TC 1, 2

−1.16 ± 0.7

−0.37 ± 0.2

−0.10 ± 0.0

−2.6

K1

−0.77 ± 0.2

−0.48 ± 0.2

−0.11 ± 0.0

−1.3

TC 3, 4

−1.46 ± 1.0

−0.32 ± 0.1

−0.08 ± 0.0

−2.1

K2

−2.29 ± 2.1

−0.38 ± 0.2

−0.10 ± 0.0

−4.7

TC 5

−0.98 ± 1.1

−0.14 ± 0.0

−0.05 ± 0.0

−5.0

K3

−1.05 ± 0.6

−0.29 ± 0.1

−0.09 ± 0.0

−3.6

mould wall and thus exhibit the best heat transfer and no influence of the radiant heat from above. The second area covers the time between 140 and 500 s after opening the stopper. The third area is from 500 to 1000 s after cast start. In both stages between 140 and 1000 s of solidification, a decreasing cooling rate from top to bottom is evident. This is due to the stabilized temperature profile after a certain time with a maximum at the top of the block. Table 1 lists a summary of the different cooling gradients in the various time ranges. This ensures that each curve is considered at the same point in time regardless of the geometric arrangement as discussed above (0–140 s, 140–500 s, 500–1000 s). The cooling gradient is calculated as the change in temperature divided by the change in time. This can be expressed mathematically as: coolinggradient ¼

DT Dt

ð2Þ

where DT is the change in temperature and Dt is the change in time. By comparing the values of the TC’s inside the melt (K 1, 2, 3) with the implemented TC’s in the mould wall (TC 1, 2, 3, 4, 5) the following assumptions can be drawn. Until 140 s the whole process is very unstable due to the occurrence of a lot of fluctuation inside the mould. This leads to the different cooling gradients and rather big standard deviations over the height. With longer cooling time the cooling gradients inside- and outside of the mould get progressively similar. The middle part including TC 3, 4, and K 2 show the smallest differences over the entire three stages. The maximum cooling rate of −5.0 K/s measured for TC 5 is approximately two times higher than for TC 1, 2 or TC 3, 4 as this position TC 5 is the only one covered fully by the primary water-cooling device. This circumstance reveals the enormous importance of a functioning primary cooling in this area. For successful casting, it is therefore essential to ensure that the first strand shell is formed as quickly as

possible and that the shell separates from the mould by thermal contraction.

Conclusion The experimental results presented in this study shed light on the complex dynamics governing heat transfer during the primary cooling phase of the DC casting process on a laboratory scale. It's noteworthy that other parameters were held constant throughout the quasi block casting process, maintaining consistency and enabling to focus on the heat transfer impact. The results obtained in this study show that the exploration of heat transfer conditions in the primary cooling area of a laboratory DC casting plant is possible. The findings can be summarized as follows: • Measuring the temperature inside and outside the mould walls shows in general a small difference over all three stages of the casting and solidification process. As a result, it is ensured that this measurement set-up is suitable for further studies for live monitoring of heat flow from cast metal to cooling medium. • The experimental methods used in this study are relatively simple and can be easily implemented in the setting. • Combined with a CFD developed model of this system it would be possible to predict the heat transfer process during the DC casting process efficiently. A comparison of the two sets of data can be used to validate the measured data. The influence of the cooling gradient is crucible for the formation of primary phases, segregation, and grain growth. On this behalf, it is under great interest to further optimize the laboratory DC casting process for designing new alloys. This research is the first step to understand how heat transfers during DC casting. By understanding this process and

Measurement of the Heat Transfer in the Primary Cooling Area …

measuring the temperature values, we can improve the quality and efficiency of casting for producing samples with adequate sizes for processing and mechanical testing. Acknowledgements This work was supported financially by the Austrian Federal Ministry of Labour and Economy the Christian Doppler Research Association and the Montanuniversity Leoben. All of them are gratefully acknowledged on this point. Additionally, the authors wish to express their sincere thanks to AMAG rolling GmbH for the great support.

References 1. Ostermann F (2014) Application-technology Aluminium. Springer, Berlin Heidelberg 2. Drezet J-M, Rappaz M, Grün G-U, Gremaud M (2000) Determination of thermophysical properties and boundary conditions of direct chill-cast aluminum alloys using inverse methods. Metall Mater Trans A 31:1627–1634 3. Han Q, Viswanathan S, Spainhower D, Das S (2001) The nature of surface cracking in direct chill cast aluminum alloy ingots. Metall Mater Trans A 32:2908–2910 4. Sengupta J, Thomas B, Wells M (2005) The use of water cooling during the continuous casting of steel and aluminum alloys. Metall Mater Trans A 36:187–204 5. Weckman DC, Niessen P (1984) The Mechanism of Cold Shut Formation on DC Continuously Cast Non-Ferrous Alloy Products: Part I: The Influence of Alloy Freezing Range, Casting Speed, and Liquid Metal Superheat. Int J Mater Res 75:332–340 6. Weckman D, Niessen P (1984) Mathematical models of the DC continuous casting process. Can Metall Q 23:209–216

1061 7. Verwijs J, Weckman D (1988) Influence of mold length and mold heat transfer on horizontal continuous casting of nonferrous alloy rods. Metall Trans B 19:201–212 8. Hadden R, Indyk B (1979) Heat transfer characteristics in closed head horizontal continuous casting. pp. 250–255 9. Herzog U, Rönigk B, Voßkühler H (1968) Probleme der Wärmeabfuhr beim Stranggießen von Kupfer. Int J Mater Res 59:99–104 10. Ohm L, Engler S (1989) Treibende Kräfte der Oberflächenseigerungen beim NE-Strangguß. Metall 43:520–524 11. Baserinia AR, Caron EJ, Wells MA, et al (2013) A numerical study of the direct-chill co-casting of aluminum ingots via FusionTM technology. Metall Mater Trans B 44:1017–1029 12. Caron EJ, Baserinia AR, Ng H, et al (2012) Heat-transfer measurements in the primary cooling phase of the direct-chill casting process. Metall Mater Trans B 43:1202–1213 13. Verein Deutscher Ingenieure VDI-Gesellschaft Verfahrenstechnik und Chemieingenieurwesen (GVC) (2006) Vdi-Wärmeatlas. Springer 14. Schwertfeger K Heat withdrawal in continuous casting of steel. The Making Shaping and Treating of Steel-Casting Volume (2003): 12 15. Mackenzie DS, Totten GE (2003) Handbook of aluminum. Dekker New York 16. Weidinger Andreas (2022) Adaptierung einer induktiv beheizten Stranggussanlage für Al-Legierungen. Master Thesis 162 17. Weidinger A, Samberger S, Schmid F, Stemper L and Pogatscher S (2023) Industry-oriented sample preparation with an in- ductively heated laboratory continuous casting plant for aluminum alloys. EMC Duesseldorf Conf Proceeding 18. Berlin, Beuth Verlag GmbH (2017) DIN EN 60584-1:2014-07, Thermocouples - Part 1: EMF specifications and tolerances (IEC 60584-1:2013) 19. Samberger, S., Stemper, L., Schmid, F., & Pogatscher, S. Effect of Heating/Cooling-rates on the aging behaviour of Al alloys. Beitrag in European Metallurgical. Conf 2023 Düsseld Nordrh-Westfal Dtschl

Influence of Chemistry and Direct Chill (DC) Casting Parameters on the Formation of Altenpohl Zone in 5xxx Alloys Akash Pakanati, Snorre Rist, Thomas Hartmut Ludwig, Eystein Vada, Shiva Talatori, and Jan-Erik Ødegård

Abstract

Introduction

Anodization of 1xxx and 5xxx aluminium alloy sheets improves the corrosion resistance, in addition to providing a bright and shiny appearance. These properties find application in a variety of building exterior and decorative purposes. However, Altenpohl zones are a typical defect observed in some of the 1xxx and 5xxx series aluminium alloys depending on their Fe-to-Si ratio. This defect is a consequence of different iron bearing particles responding differently to etching during the anodizing process. Considerable research has been conducted to study Altenpohl zone formation in 1xxx and 5xxx alloys. This paper aims to improve and extend the knowledge base where the influence of alloy chemistry, casting speed, and casting temperature on the formation and extent of Altenpohl zones in a commercially available 5005 alloy has been studied. Based on the results, recommendations on chemistry and process parameters are provided to control Altenpohl zone formation during DC casting of rolling slabs. Keywords

Altenpohl zone




DC casting




Defects

A. Pakanati (&)
S. Rist
E. Vada
J.-E. Ødegård Hydro Research and Development Center, 6600 Sunndalsøra, Norway e-mail: [email protected] T. H. Ludwig Hydro Aluminium AS, 0283 Oslo, Norway

The earliest observations of Altenpohl Zone can be traced back to Ref [1] in 1950s. In this early work, the phenomena have been linked to macrosegregation of Iron. Considerable attention from several research groups has helped in understanding the Altenpohl zone formation. It is generally agreed that the presence of different iron bearing particles leads to the formation of this Altenpohl zone [2–5]. When sheet ingot slices are subjected to etching, the response to etching is dependent on the type of iron bearing particles. An illustration is provided in Fig. 1. In Fig. 1a, we show a representative cross-section of a sheet ingot. The shell zone is marked, and we can see an outer zone, light in color and an inner zone, dark in color. In this article, we refer to the inner darker zone as Altenpohl zone. The outer zone contains predominantly Alm Fe whereas the inner zone mainly contains Al3 Fe particles [6]. Sheet ingots are subjected to scalping and if the scalping plane (illustrated in Fig. 1) does not lie along the Altenpohl zone, this does not have any effect on the surface/product quality. The resulting rolling surface is illustrated in Fig. 1c. However, if the Altenpohl zone lies within the scalping plane (represented in Fig. 1b), the resulting rolling surface will have alternating light and dark bands as shown in Fig. 1d. This is called structural streaking and results in scrapping of the ingots. Several researchers have investigated this defect, which is usually observed in 1xxx [7–9] and 5xxx [6, 10, 12, 13] alloys. The consensus is that the cooling rate plays a major role in the creation of the iron particles. The formation of Alm Fe is attributed to a high cooling rate [5] (>8–10 K/s) and hence their presence in the outer periphery. Al3 Fe forms at a low cooling rate (