investigation of the materials and paste relationships to ... · distribution la plus homogène de...

INVESTIGATION OF THE MATERIALS AND PASTE RELATIONSHIPS TO IMPROVE FORMING PROCESS

AND ANODE QUALITY

Thèse

Kamran Azari Dorcheh

Doctorat en génie des matériaux et de la métallurgie

Philosophiae doctor (Ph.D.)

Québec, Canada

© Kamran Azari Dorcheh, 2013

iii

Résumé

Des anodes de haute densité et de qualité uniforme sont d'un grand intérêt dans la production d'aluminium

primaire. La variation dans les propriétés des matières premières ainsi que le grand nombre de variables est

reconnue comme étant un grand défi et conduit à des anodes de qualité très variable. Dans ce projet, une

d'essai de comportement au compactage de la pâte d’anode a été développée. Le comportement au

compactage de la pâte d'anode a été utilisé comme indice de qualité intermédiaire pour réduire le nombre de

variables.

Différentes combinaisons de temps et de températures de mélange ont été utilisées pour faire de la pâte

d’anode afin de comprendre les effets des variables de mélange. Les pâtes ont été compactées, cuites et

caractérisés par la suite. Le mélange effectué à 178 C pendant 10 minutes a été celui permettant d’obtenir la

distribution la plus homogène de coke et de brai et la densité maximale des échantillons verts et cuits.

Il est admis que le comportement à la déformation de la matrice liante (particules fines de coke+brai) contribue

fortement au processus de compactage. Les mélanges des matrices liantes et des pâtes d’anodes, avec

différents ratios de brai et de particules de coke, ont été compactés à des taux de déformation différentes. La

compaction de la matrice liante et de la pâte d'anode, avec les compositions classiques utilisées dans ce

projet, n’est pas significativement dépendante de la vitesse de mise en forme.

L'effet de la forme et de la densité des particules sur la densité de la pâte compactée a été étudié. La densité

apparente des particules, la densité apparente (Scott) et la densité apparente vibrée du lit de particules ont été

mesurées pour différentes fractions de cinq sources de coke éponge. Des pâtes d’anode ont été produites à

partir des cinq cokes et ensuite compactées. Il a été suggéré que la densité apparente vibrée n'est pas un

facteur suffisant pour déterminer les propriétés de compaction et la densité de la pâte compactée et aussi que

les facteurs de forme et la densité des particules sont des paramètres importants qui devraient être

considérés.

v

Abstract

Hall-Héroult electrolysis process is the major method for the production of primary aluminium. High density and

consistent quality anodes are of great interest in this process. Variations in raw materials properties are known

as a big challenge that lead to the production of inconsistent and low quality anodes. Anode quality is

associated with a large number of variables that makes it difficult to control and improve the quality of anodes.

In this project, a compaction test was developed and compaction properties of anode paste were used as an

intermediate quality index to reduce the number of variables.

Different combinations of the time and temperature of mixing were used to make the anode paste in order to

understand the effect of mixing variables. The pastes were compacted under a uniaxial load of 60 MPa and

baked. The green and baked samples were characterized through various techniques. The selected mixing

effectiveness resulting in a homogeneous distribution of coke and pitch, the maximum green and baked

density and the minimum air permeability was obtained by mixing at 178 C for 10 minutes.

Deformation behaviour of binder matrix (fine coke+pitch) as a deformable phase was assumed to have a great

contribution to the compaction process. Binder matrix compositions with different pitch to fine coke ratios were

compacted at the strain rates of 2.910-4 s-1 and 2.910-3 s-1. Anode pastes with different formulations were

also compacted at two different strain rates. Compaction of binder matrix and anode paste with the

compositions used in this project was not a time dependent process.

The last part of experiments was conducted to understand the effect of particle shape and particle density on

the packing and compaction properties. Apparent density of particles, vibrated bulk density and Scott density

were measured for different size fractions of 5 sources of sponge coke. Anode pastes were made from the

cokes and compacted. It was suggested that vibrated bulk density was not a sufficient factor to determine the

packing properties and compacted density while on the other hand the shape factors and particle density

should be considered as important parameters in this regard.

vii

Table of contents

Résumé ............................................................................................................................................................... iii

Abstract ............................................................................................................................................................... v

Acknowledgments ............................................................................................................................................ xvii

Preface .............................................................................................................................................................. xix

1 Chapter 1. Introduction ............................................................................................................................... 1

1.1 History................................................................................................................................................ 1

1.2 Aluminium production ........................................................................................................................ 1

1.2.1 Hall-Héroult process.................................................................................................................. 2

1.3 Problems............................................................................................................................................ 6

1.4 Hypothesis ......................................................................................................................................... 6

1.5 Objectives .......................................................................................................................................... 7

2 Chapter 2: Literature review ....................................................................................................................... 9

2.1 Raw materials .................................................................................................................................... 9

2.1.1 Calcined petroleum coke ........................................................................................................... 9

2.1.2 Pitch ........................................................................................................................................ 14

2.1.3 Anode butt ............................................................................................................................... 19

2.1.4 Particle granulometry .............................................................................................................. 21

2.1.5 Rheological properties ............................................................................................................ 27

2.1.6 Particle shape and texture ...................................................................................................... 29

2.2 Anode manufacturing process ......................................................................................................... 31

2.2.1 Blending .................................................................................................................................. 31

2.2.2 Mixing ...................................................................................................................................... 34

2.3 Summary ......................................................................................................................................... 36

3 Chapter 3: Materials and methods............................................................................................................ 37

3.1 Introduction ...................................................................................................................................... 37

3.2 Description of raw materials............................................................................................................. 37

3.2.1 Calcined petroleum coke ......................................................................................................... 37

3.2.2 Coal tar pitch ........................................................................................................................... 38

3.3 General experimental procedures .................................................................................................... 39

3.3.1 Scott density ........................................................................................................................... 39

3.3.2 Vibrated bulk density (VBD) .................................................................................................... 39

viii

3.3.3 Particle apparent density ......................................................................................................... 40

3.3.4 Particle shape factors .............................................................................................................. 41

3.3.5 Mixing ...................................................................................................................................... 41

3.3.6 Compaction test ...................................................................................................................... 42

3.3.7 Green and baked apparent density ......................................................................................... 43

3.3.8 Porosity ................................................................................................................................... 44

3.3.9 Determination of the optimum pitch/coke ratio ........................................................................ 44

3.3.10 Specific surface area (BET) .................................................................................................... 45

3.3.11 Mercury porosimetry ................................................................................................................ 46

4 Chapter 4: Validation of compaction test to evaluate paste quality ........................................................... 49

4.1 Résumé ............................................................................................................................................ 49

4.2 Abstract ............................................................................................................................................ 49

4.3 Introduction ...................................................................................................................................... 50

4.4 Experimental procedures ................................................................................................................. 51

4.5 Results and discussion .................................................................................................................... 52

4.5.1 Vibrated bulk density ............................................................................................................... 52

4.5.2 Compaction test and green density ......................................................................................... 53

4.6 Summary .......................................................................................................................................... 55

5 Chapter 5: Influence of mixing variables on paste characteristics, compaction behaviour and anode

properties .......................................................................................................................................................... 57

5.1 Résumé ............................................................................................................................................ 57

5.2 Abstract ............................................................................................................................................ 57

5.3 Introduction ...................................................................................................................................... 57

5.4 Materials and methods ..................................................................................................................... 58

5.4.1 Sample preparation ................................................................................................................. 58

5.4.2 Characterization of anode paste ............................................................................................. 59

5.4.3 Characterization of green samples .......................................................................................... 59

5.4.4 Characterization of baked samples ......................................................................................... 61

5.5 Results ............................................................................................................................................. 61

5.5.1 Green paste ............................................................................................................................ 61

5.5.2 Green anodes.......................................................................................................................... 63

5.5.3 Baked anodes ......................................................................................................................... 70

5.6 Discussions ...................................................................................................................................... 72

ix

5.6.1 Mixing time .............................................................................................................................. 72

5.6.2 Mixing temperature ................................................................................................................. 73

5.7 Summary ......................................................................................................................................... 74

6 Chapter 6: Influence of formulation on the compaction behaviour of paste .............................................. 75

6.1 Résumé ........................................................................................................................................... 75

6.2 Abstract............................................................................................................................................ 75

6.3 Introduction ...................................................................................................................................... 76

6.4 Materials and methods .................................................................................................................... 77


6.5.1 Binder matrix ........................................................................................................................... 78

6.5.2 Anode paste ............................................................................................................................ 84

6.5.3 Influence of Blaine number vs. pitch ratio ............................................................................... 90

6.6 Summary ......................................................................................................................................... 91

7 Chapter 7: Influence of shape and density of coke particles on the compaction behaviour of paste ........ 93

7.1 Résumé ........................................................................................................................................... 93

7.2 Abstract............................................................................................................................................ 93

7.3 Introduction ...................................................................................................................................... 94

7.4 Materials and methods .................................................................................................................... 94


7.5.1 Coke particle characteristics ................................................................................................... 96

7.5.2 Particle shape and compaction behaviour of the paste ......................................................... 105

7.6 Summary ....................................................................................................................................... 109

8 Chapter 8: General discussion and conclusions ..................................................................................... 111

8.1 Introduction .................................................................................................................................... 111

8.2 Overview ........................................................................................................................................ 111

8.3 Effect of mixing time ...................................................................................................................... 112

8.4 Effect of mixing temperature .......................................................................................................... 113

8.5 Effect of Blaine number and pitch ratio .......................................................................................... 113

8.6 Effect of compaction rate ............................................................................................................... 114

8.7 Effect of shape factors ................................................................................................................... 115

8.8 Future work .................................................................................................................................... 115

Bibliography .................................................................................................................................................... 117

Appendix A: Image processing ........................................................................................................................ 125

x

List of contributions.......................................................................................................................................... 127

xi

List of Tables

Table ‎1.1: Comparison of Söderberg and prebaked anodes [9] ........................................................................... 4

Table ‎2.1: Typical composition of petroleum coke [33] .......................................................................................... 9

Table ‎2.2: Comparison of needle and granular cokes [33, 36, 37] .......................................................................... 11

Table ‎2.3: Pitch sensitivity values [51] ................................................................................................................. 19

Table ‎2.4: Influence of butt quality on the anode properties [70] ......................................................................... 19

Table ‎3.1: Real density and chemical composition of cokes used for anode fabrication ................................... 38

Table ‎3.2: Ball milling parameters for coke D, size distribution and specific surface area of fine cokes ........... 38

Table ‎3.3: Properties of coal tar pitch used as binder in anode samples .......................................................... 38

Table ‎3.4: Mass of coke particles impregnated with resin for apparent density measurement ......................... 40

Table ‎3.5: Particle shape aspects measured by image analyzer ...................................................................... 41

Table ‎3.6: Size distribution of coke particles in the paste samples ................................................................... 42

Table ‎4.1: Particle size distribution of pastes .................................................................................................... 51

Table ‎4.2: Coke source and pitch content of pastes ......................................................................................... 51

Table ‎4.3: Shape factors and VBD of coke particles ......................................................................................... 52

Table ‎5.1: Mixing time and temperature for making anode pastes .................................................................... 59

Table ‎6.1: Composition and compaction parameters for binder matrix samples ............................................... 78

Table ‎6.2: Composition and compaction parameters for paste samples ........................................................... 79

Table ‎6.3: Initial apparent and relative densities of binder matrix with different compositions before compaction

.......................................................................................................................................................................... 82

Table ‎6.4: Specific pitch content for binder matrix compositions ....................................................................... 83

Table ‎6.5: Relative density of binder matrix and paste compositions compacted at 60 MPa at a DR of 10

mm/min ............................................................................................................................................................. 90

xii

List of Figures

Figure ‎1.1: Cumulative primary aluminium production [6] ..................................................................................... 1

Figure ‎1.2: Schematic representation of aluminium electrolysis cell with prebaked anodes [9] ............................ 2

Figure ‎1.3: Anode consumption figures ............................................................................................................... 6

Figure ‎2.1: Microtexture of anisotropic and isotropic cokes (100x) [36] ............................................................... 10

Figure ‎2.2: Relative effects of coke variations, formulation and baking temperature on anode properties [29] ... 13

Figure ‎2.3: Pitch-aggregate system [51] .............................................................................................................. 14

Figure ‎2.4: Binder-aggregate structure [51] ......................................................................................................... 18

Figure ‎2.5: The results of radial scanning of X-ray computed tomography [71]................................................... 21

Figure ‎2.6: Influence of granulometry on anode properties [40] .......................................................................... 22

Figure ‎2.7: Relationships between VBD of particles (a), viscosity of paste (b) and electrical resistivity (c) and

baked apparent density of anode (d) with fineness and amount of fine coke for three formulations of fine,

intermediate and coarse particles [82]. ................................................................................................................ 24

Figure ‎2.8: The correlation of interparticle distance and pitch layer thickness, derived from image analysis, with

electrical resistivity [83] ........................................................................................................................................ 24

Figure ‎2.9: Effect of particle size on specific surface area. A) pitch carbon 850 C, B) pitch carbon 1150 C, C)

petroleum coke [22] ............................................................................................................................................. 26

Figure ‎2.10: Influence of dust content on anode properties. Adapted from [40] .................................................. 27

Figure ‎2.11: Relationships between permeability (a) and tensile strength (b) of anode with fineness and

amount of fine coke for three formulations of fine, intermediate and coarse particles [82]. ................................. 27

Figure ‎2.12: Particle-binder configuration in anode structure [94] ....................................................................... 28

Figure ‎2.13: Paste sensitivity to viscosity controlling parameters [94] ................................................................. 29

Figure ‎2.14: A schematic of anode manufacturing process ............................................................................... 31

Figure ‎2.15: Modified coke blending strategies [77, 114] ....................................................................................... 33

Figure ‎3.1: Scott volumeter for bulk density of coke fractions ........................................................................... 39

Figure ‎3.2: Setup for vibrated bulk density ........................................................................................................ 40

Figure ‎3.3: Schematic illustration of impregnated coke particles and analyzed surfaces .................................. 41

Figure ‎3.4: Mixer used to make paste samples ................................................................................................. 42

Figure ‎3.5: Rigid die and MTS Servohydraulic press for compaction of paste .................................................. 43

Figure ‎3.6: Schematic illustration for the measurement of the sample height before compaction (x). After

putting the punch on the sample, the punch height entered into the die was marked (y). x obtained by z-y

where z is the total mould height. ...................................................................................................................... 43

Figure ‎3.7: Baking cycle for anode samples ...................................................................................................... 45

Figure ‎3.8: Apparent density for green (a) and baked (b) samples made with different pitch contents ............. 45

xiii

Figure ‎4.1: Schematic representation of paste compactability test ................................................................... 50

Figure ‎4.2 : Vibrated bulk density of blended coke fractions ............................................................................. 53

Figure ‎4.3: Examples of repeatability of compaction curves ............................................................................. 54

Figure ‎4.4: Compaction curves with variations in paste formulations and source of coke ................................ 54

Figure ‎5.1: CT scanner used for green samples (Siemens Somatom Sensation 64) ........................................ 60

Figure ‎5.2: Schematic representation for characterization of green and baked samples .................................. 60

Figure ‎5.3: BET surface area of the pastes mixed for 10 minutes at different mixing temperatures ................. 61

Figure ‎5.4: BET surface area and total volume of pores for the pastes mixed at 178 C for different mixing

times .................................................................................................................................................................. 62

Figure ‎5.5: Incremental pore volume vs. pore size for the pastes mixed at 178 C for different mixing times .. 62

Figure ‎5.6: Effect of mixing time at a constant mixing temperature (178 C) on the compaction behaviour and

apparent density of green samples ................................................................................................................... 63

Figure ‎5.7: Effect of mixing temperature at a constant mixing time (10 min) on the compaction behaviour and

apparent density of green samples ................................................................................................................... 63

Figure ‎5.8: Dependence of green apparent density on (a) mixing time at a constant mixing temperature of 178

C; (b) mixing temperature for a constant mixing time of 10 minutes................................................................ 64

Figure ‎5.9: Dependence of open porosity of section 1 on (a) mixing time at a constant mixing temperature of

178 C; (b) mixing temperature for a constant mixing time of 10 minutes ......................................................... 64

Figure ‎5.10: Constructed images form CT numbers and the profile of CT number along the diameter (black

line) of green samples made with different mixing times and temperatures ...................................................... 65

Figure ‎5.11: Variations in Rt and Ra with mixing parameters............................................................................. 65

Figure ‎5.12: CT numbers in the entire and the section 1 of anode samples mixed at 178 C for different mixing

times; (a) standard deviation of CT numbers; (b) average and fluctuations of CT numbers ............................. 66

Figure ‎5.13: Results of CT numbers in the entire and the section 1 of anode samples mixed for 10 minutes at

different mixing temperatures; (a) standard deviation of CT numbers; (b) average and fluctuations of CT

numbers ............................................................................................................................................................ 67

Figure ‎5.14: Variations in the average CT number of a green sample mixed for 10 minutes at 178 C (a) on the

surface of the slices along the height; (b) in the radial direction on the slices as a function of height ............... 67

Figure ‎5.15: Effect of mixing time and mixing temperature on the pore size distribution of the green samples 68

Figure ‎5.16: Examples for microstructure of the green samples; 10 minutes at 158 C (a, b, c); 10 minutes at

178 C (d, e); 6 minutes at 178 C (f). The samples are impregnated with resin and polished. White arrows

indicate open pores filled by resin while the black ones demonstrate the pores filled by pitch (light gray: coke,

medium gray: resin, dark gray/black: pitch) ....................................................................................................... 68

Figure ‎5.17: Variations of binder matrix thickness with (a) mixing time at a constant mixing temperature of 178

C; (b) mixing temperature for a constant mixing time of 10 minutes................................................................ 69

xiv

Figure ‎5.18: Distribution of binder matrix thickness for samples mixed at 178 C for (a) 6 minutes; (b) 10

minutes; (c) 15 minutes; (d) 20 minutes ............................................................................................................ 69

Figure ‎5.19: Distribution of binder matrix thickness for samples mixed for 10 minutes at (a) 158 C; (b) 168 C;

(c) 178 C; (d) 188 C ....................................................................................................................................... 70

Figure ‎5.20: Dependence of baked apparent density and open porosity of baked samples on (a) mixing time at

a constant mixing temperature of 178 C; (b) mixing temperature for a constant mixing time of 10 minutes .... 70

Figure ‎5.21: Effect of mixing time and mixing temperature on the air permeability of the baked sections ........ 71

Figure ‎5.22: Influence of mixing time and mixing temperature on the BET surface area of the baked sections 72

Figure ‎6.1: Compaction curves for the binder matrix samples made from a fine coke with SSA of 2.9 m2/g (BN

2300) and P/FC ratios of 30/100, 34/100 and 38/100 ........................................................................................ 80





Figure ‎6.4: Compaction curves for the binder matrix samples made with a P/FC ratio of 34/100 and fine cokes

with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively) ...................................................... 82

Figure ‎6.5: Compaction curves for the binder matrix samples made with a P/FC ratio of 38/100 and fine cokes

with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively) ...................................................... 83

Figure ‎6.6: Compaction curves for the paste samples made from a fine coke with SSA of 2.9 m2/g (BN 2300)

and P/C ratios of 14/100, 16.2/100, 19/100 and 22/100 .................................................................................... 85





Figure ‎6.9: Compaction curves for the paste samples made with a P/C ratio of 14/100 and fine cokes with SSA

of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively) ..................................................................... 86

Figure ‎6.10: Compaction curves for the paste samples made with a P/C ratio of 16.2/100 and fine cokes with

SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively) ............................................................. 87

Figure ‎6.11: Compaction curves for the paste samples made with a P/C ratio of 19/100 and fine cokes with


Figure ‎6.12: Compaction curves for the paste samples made with a P/C ratio of 22/100 and fine cokes with



and P/C ratios of 16.2/100 and 22/100 .............................................................................................................. 88


and P/C ratios of 16.2/100 and 22/100 .............................................................................................................. 89

Figure ‎6.15: Densification of the paste samples at 10 MPa as a function of time ............................................. 89

xv

Figure ‎7.1: Shape factors for different size fractions of the cokes; (a) aspect ratio; (b) sphericity; (c) roundness;

(d) convexity; (e) compactness ......................................................................................................................... 97

Figure ‎7.2: Loading plot for the PCA model on the particle shape factors ........................................................ 98

Figure ‎7.3: Apparent density for different coke fractions ................................................................................... 98

Figure ‎7.4: Optical microscopic images (25x) for -8+16 mesh size fraction of the 6 cokes used for making the

anode samples; (a) coke A; (b) coke B; (c) coke C; (d) coke D; (e) coke E; (f) shot coke ................................ 99

Figure ‎7.5: The fraction of (a) open and (b) closed pores in various size fractions of the cokes ....................... 99

Figure ‎7.6: Angle of repose for different coke fractions ................................................................................... 100

Figure ‎7.7: Scott density (a); and vibrated bulk density (b) for different size fractions of the cokes ................ 101

Figure ‎7.8: Schematic illustration of roughness and waviness (convexity) of a surface .................................. 101

Figure ‎7.9: Loading plot for the PLS model to predict VBD from particle shape and density .......................... 102

Figure ‎7.10: Observed vs. predicted VBD derived from the PLS model (R2=0.767) ....................................... 102

Figure ‎7.11: Coefficients plot for the variables to predict the VBD .................................................................. 102

Figure ‎7.12: Variable importance plot for the variables to predict the VBD ..................................................... 103

Figure ‎7.13: Hausner ratio for different size fractions of the cokes ................................................................. 104

Figure ‎7.14: Percentage of inter-particle voids after VBD test ........................................................................ 105

Figure ‎7.15: Compaction curves for anode samples made from sponge cokes with different apparent density

and shape factors ............................................................................................................................................ 106

Figure ‎7.16: Compaction curves for anode samples made from a mixture of shot coke and sponge cokes; (a)

coke A + shot coke; (b) coke B + shot coke; (c) coke C + shot coke; (d) coke D + shot coke; (e) coke E + shot

coke ................................................................................................................................................................. 108

xvi

Dedicated to my parents and my wife

xvii

Acknowledgments

I have worked with a great number of people throughout my PhD. It would not have been possible to write this

thesis without the help and support of the kind people around me. It is a good opportunity for me to particularly

mention some of them here.

I would like to express my deep and sincere gratitude to my thesis director, Prof. Houshang Alamdari, for his

confidence in me and giving me the opportunity to be part of his group to conduct this project. I appreciate his

patience, continuous encouragement, immense knowledge, and his availability for discussions. His valuable

suggestions and comments are also appreciated. His guidance helped me all during the time of my research

and writing this thesis. I would also sincerely appreciate my co-director, Prof. Mario Fafard, for his insight,

support, knowledge and valuable comments during these years.

I would like to extend my gratitude to my co-director in Alcoa, Dr. Angelique Adams, for his insight, support,

and valuable discussions throughout this project. I gratefully acknowledge Dr. Donald Ziegler from Alcoa for

his step by step guidance, valuable discussions and comments. I would like to express my gratitude to Gilles

Dufour for his support and insight. I would also thank Pierre Mineau and his colleagues in the laboratory at

Alcoa, Deschambault plant, for collaboration in sample analysis and evaluation. Moreover, I am grateful to

Alcoa for providing me with the materials, experimental setups and equipment in Deschambault plant which

helped me in this research.

I would like to thank Alcoa, Natural Sciences and Engineering Research Council of Canada (NSERC), and

Fonds de Recherche du Québec-Nature et Technologies (FRQ-NT) for the financial support that was provided

by the intermediary of the Aluminium Research Centre – REGAL. Part of the support was provided through the

Industrial Innovation Scholarship program (Bourse de recherche en milieu de pratique).

I wish to thank Gholamreza Aryanpour, Donald Picard, Hicham Chaouki and Guillaume Gauvin for their great

help, guidance and the time they dedicated to this project. Special thanks go to Maude Larouche for

microstructural and image analysis of the samples. I especially appreciate Hugues Ferland, Vicky Dodier,

Daniel Marcotte, Marc Choquette and André Ferland for their availability, collaboration and technical

assistance in the laboratory. Thanks to all of the staff of Mining, Metallurgical and Materials Engineering

Department of Laval University for their help and support.

Special thanks to my friends and colleagues Hamed Heidari, Francois Chevarin, Behzad Majidi, Ramzi Ishak,

Asem Hussein, Geoffroy Rouget and Julien Lauzon-Gauthier for their kind support, help and suggestions in

this project. Many thanks to my friends Mohammad Ghasdi, Majid Heidari, and Milad Mardan, Franck Armel

xviii

Tchitembo, Mousa Javidani and all my friends at the department for making a joyful environment and the great

moments we had together.

I express my sincere gratitude to my parents and brothers for all their love, support and encouragement to aid

me to fulfill my dreams. Words fail me to express my appreciation to my loving, supportive, encouraging, and

patient wife, Neda, for her faithful support during these years.

xix

Preface

Hall-Héroult electrolysis process is the major production route for primary aluminium. Many efforts and

modifications have been made to improve the energy efficiency in this energy intensive process. In spite of the

successful attempts made to reduce energy consumption from 50 kWh/kg Al in 1900 to less than 14 kWh/kg Al

in 2009, this process is still a major consumer of electrical energy. Canada with 2.7 million tons of primary

aluminium in 2012 is the third producer of this metal in the world, so any improvement in energy efficiency of

the process will bring great economic and environmental advantages. Considering the fact that 90% of

aluminium in Canada is produced in Quebec province, improvement of energy efficiency will have a

considerable impact on economy and environment in this province.

Carbon anodes used in the aluminium smelting cells consume 10% of electrolysis cell power. These anodes

are one the critical elements that could be investigated to reduce the energy consumption. Anode quality

aspects such as density, electrical resistivity and reactivity not only affect the energy efficiency but also carbon

consumption, environmental and technological concerns. Consistent and homogeneous anodes are more

important than single high quality anodes. Deficit in anode grade materials, variations in raw materials

properties and thus in process parameters are big challenges to consistent anode fabrication. This thesis

aimed to reduce the impact of materials variations on the anode quality.

This doctoral thesis is presented to the department of mining, materials and metallurgical engineering of Laval

University. The project was part of a Collaborative Research and Development (CRD) program between Alcoa

and Laval University (MACE3 industrial research chair, Aluminium Research Centre-REGAL) and was

supported by the Natural Science and Engineering Research Council of Canada (NSERC) and Fonds de

Recherche du Québec-Nature et Technologies (FRQ-NT). This doctoral project was carried out under

supervision of Professor Houshang Alamdari and co-supervision of Professor Mario Fafard. Dr. Angelique

Adams, ―Director, Global Technology Development at Alcoa‖, was the industry supervisor. Dr. Donald Ziegler,

―Program Manager-Modelling at Alcoa‖, was the chief technical link between the Alcoa program for anode

fabrication and MACE3.

The works carried out as part of the project under the title of: ―Investigation of the materials and paste

relationships to improve forming process and anode quality” have been reported in this thesis. The second

chapter of this thesis reports the previous works reported in anode technology literature focusing on the

influence of materials characteristics and process parameters on the anode quality. It is followed by materials

selection and proposed methodology in chapter 3. Five different industrial sources of sponge calcined

petroleum coke as well as a shot coke and a coal tar pitch were used as starting materials. It was suggested to

use intermediate quality aspects, representing the materials and process parameters, to reduce the number of

xx

variables to be controlled. Compaction behaviour of anode paste was proposed as a quality index for the first

time. A compaction test method was developed and efforts were made to validate the test, reported in chapter

4. Preliminary compaction tests revealed that the test was sensitive to the variations in materials and process

parameters. Therefore, the test was used for the rest of the experiments in the subsequent steps.

In chapter 5 the influence of mixing variables including mixing time and temperature on the paste

characteristics, compaction behaviour, and green and baked anode properties was studied. A complete

discussion was presented to explain how mixing variables affect the coke and pitch distribution in the

compacted samples. The optimum mixing parameters resulting in the best mixing effectiveness and density

were selected and used to consolidate the samples. Influence of pitch content and fineness of fine coke on the

compaction behaviour of binder matrix and paste was investigated and reported in chapter 6. Compaction was

made at different deformation rates to study the effect of compaction rate and the contribution of paste

viscosity to the evolution of density. Chapter 7 reports the importance of different shape factors of coke

particles to the packing and compaction of the paste. It was explained that particle shape and density were

important aspects to determine the packing behaviour and void fraction. General discussion and conclusions

are presented in chapter 8 followed by suggested future works. The list of publications derived from this

doctoral thesis is presented.

1 Chapter 1. Introduction

1.1 History

Metallic aluminium is not found in the nature due to its high chemical reactivity with oxygen. It is found as

silicates and oxides with various levels of hydration and impurity. It has been estimated that 8% of earth crust

is composed of aluminium [1-3] and it is the second metallic element on the earth behind silicon being 27.5% [1].

Aluminium is the most consumed non-ferrous metal in the world. It is used for transportation, packaging,

building, electrical and machinery applications. The wide application of aluminium is due to its good strength-

to-weight ratio, good electrical and thermal conductivity, resistance to corrosion, recyclability, formability,

coating ability and non-toxicity in contact with a wide range of foods. Primary aluminium production was

around 6800 metric tons in 1900 and reached to 4.5 million metric tons in 1960 [4], when it surpassed copper

production with 3.9 million tons [5]. The world primary aluminium production was about 33.9 million tons in 2006

[4], while more than 75% was produced by seven countries [6] (Figure 1.1). It reached 44.9 million tons in 2012

[7], 6% of which was produced in Canada and 4.5% in the USA. With current trend, the primary aluminium

production is predicted to reach 60 million tons in 2020 [8].

Figure 1.1: Cumulative primary aluminium production [6]

1.2 Aluminium production

In 1808, Humphry Davy claimed that alumina is the oxide of an undiscovered metal [9]. In his trial efforts to

obtain this metal, he produced a small amount of Al-Fe alloy [10]. Hans Christian Oersted produced a little

sample of more pure aluminium in 1825 by reducing aluminium chloride with potassium amalgam [9, 10].

2

Friedrich Wohler produced aluminium globules in 1827 by the same method and demonstrated its light weight

and malleability [9]. In 1854, Henri Saint-Claire Deville used cheaper sodium instead of potassium to produce

aluminium and the first commercial aluminium plant started in 1855 by using Deville’s method [9, 10]. In 1886,

after large scale generation of electrical power, Paul L. T. Héroult in France and Charles M. Hall in the USA

independently produced aluminium by electrolysis of dissolved alumina in molten cryolite at 980-1000 C [2, 9,

10]. Two years later, Karl Josef Bayer developed a process to extract alumina from bauxite [9]. These two

methods together resulted in the principal aluminium production process named Hall-Héroult process.

Hall-Héroult process is still the principal process for primary aluminium production, however it is energy

intensive and pollutant. In addition, investment cost is high, because of the large number of production units

required. The supply of required materials including petroleum coke and coal-tar pitch (CTP) is limited and

bauxite resources are not sufficient. These disadvantages motivate the innovation of alternative processes

such as Toth process [1, 11], nitride process [10], Elliot-Mitt process [11], Alcoa chloride process [1, 11, 12], ALCAN

process [11] and carbothermic reduction [9, 11], but none of them has found large scale commercial application.

1.2.1 Hall-Héroult process

As shown in Figure 1.2, a Hall-Héroult cell consists of a brick-lined outer steel shell as a container and support.

Inside the shell, cathode blocks are cemented together by ramming paste. The molten electrolyte is

maintained at high temperature inside the cell. A single Söderberg electrode or a number of prebaked carbon

blocks are used as anode, while the principal formulation and the fundamental reactions occurring on their

surface are the same. Each cell is 10-15 m long, 4-5 m wide and 1-2 m high. A pot line may consist 100-300

cells [11].

Figure 1.2: Schematic representation of aluminium electrolysis cell with prebaked anodes [9]

3

Electrolyte: The electrolyte is a molten bath of cryolite (Na3AlF6) and alumina. Cryolite is a good solvent for

alumina with low melting point, satisfactory viscosity, low vapour pressure, and low reactivity with electrodes

and products. Its density is also lower than that of aluminium which allows natural separation of the product

from the salt. The cryolite ratio (NaF/AlF3) in pure cryolite is 3. Its melting temperature is 1010 C and forms a

eutectic with 11% alumina at 960 C [9, 13]. In industrial cells the cryolite ratio is between 2 and 3 to decrease

the melting temperature to 940-980 C [9, 13].

Cathode: Carbon cathodes are made of anthracite and graphite which are calcined at around 1200 C before

being crushed and sieved. Aggregates are mixed with coal-tar pitch, moulded and baked. Carbon purity is not

as stringent as for anode, because metal contamination from cathode is not significant [9]. Carbon cathode

must have adequate strength, good electrical conductivity and high resistance to wear and sodium penetration.

Anthracite cathodes have higher wear resistance [14] and slower creep with lower amplitude [15] than graphitic

and graphitized petroleum coke cathodes. Instead, dense cathodes with more graphitic order have lower

energy consumption [14] and swelling due to sodium penetration [14, 16]. The swelling may result in early and

non-uniform deterioration of cathode block.

Anodes: Carbon anodes have a specific situation in aluminium smelting. Anode quality affects technological,

economical and environmental aspects of aluminium production. Energy efficiency is related to the nature of

anode materials, as well as the porosity of baked anodes. Around 10% of cell power is consumed to overcome

the electrical resistance of prebaked anode (50-60 m) [9]. Carbon is consumed more than theoretical value

due to a low current efficiency and non-electrolytic consumption. Inhomogeneous anode quality due to the

variation in raw materials and production parameters also affects its performance and the cell stability.

Söderberg (in-situ baking) and prebaked anodes have been industrially used through the history of aluminium

smelting. Söderberg electrodes, used for the first time in Norway in 1923 [11], are composed of a steel shell and

a carbonaceous mass which is baked by the heat being escaped from the electrolysis cell. Carbon-based

materials such as coke are crushed, heat treated and classified. These aggregates are mixed with pitch as

binder, briquetted and loaded into the shell. Temperature increases from top to the bottom of the column and

in-situ baking takes place. Significant amount of hydrocarbons are emitted during baking which is a

disadvantage of this type of electrodes [9, 11].

Prebaked anodes are divided into graphitized and coke types. For manufacturing of the graphitized anodes,

anthracite and petroleum coke are calcined and classified. They are then mixed with coal-tar pitch and

pressed. The pressed green anode is then baked at 1200 C and graphitized. Graphitized anodes are not

produced commercially today. Coke anodes are made of calcined petroleum coke and recycled anode butts as

filler aggregates and coal-tar pitch (binder). The anodes are manufactured by mixing aggregates with coal tar

pitch to form a paste with a doughy consistency. This material is most often vibro-compacted but in some

4

plants pressed during which it is deformed and densified. The green anode is then sintered at 1100-1200 C

for 300-400 hours without graphitization to increase its strength through decomposition and carbonization of

the binder. Higher baking temperatures increase the mechanical properties and thermal conductivity, and

decrease the air and CO2 reactivity [17]. The specific electrical resistance of the coke-type anodes is higher

than that of the graphitized ones, but they have higher compressive strength and lower porosity [11].

Although lower capital, labour and energy are required for Söderberg anodes, an increasing tendency towards

prebaked anodes has been in place since 1970s. One of the main reasons for this is the environmental

problems caused by Söderberg anodes. Other reasons are development of improved forming techniques such

as vibrocompaction, higher electrical conductivity [1, 11], and lower carbon and power consumption of prebaked

anodes [10] which may be due to their higher baking temperature [18]. Table 1.1 compares the properties of

prebaked and self-baking anodes [9].

Table 1.1: Comparison of Söderberg and prebaked anodes [9]

Prebaked anodes Söderberg anodes

Electrical resistivity Density Pitch content, % Anode consumption Power efficiency Anode effect Manufacturing cost Labour Alumina feeding Collection of fumes Environmental problems

Low High 14-18 Low High Rare High High

Central (easy) Easy Low

High Low 25-5 High Low

Frequent Low Low

Sidewise (difficult) Difficult

High

Carbon anodes are consumed during electrolysis. They cause high energy consumption and greenhouse gas

emissions in smelting and anode plants. They disturb cell performance due to an uneven wear and mechanical

failure. Despite numerous investigations that have been done to develop alternatives for carbon anodes, the

replacement does not seem to be realistic in the near future. Graphitized anodes with better electrical

conductivity are very expensive for industrial application. Composite anodes (85% alumina,15% carbon) with

smaller polar distance and lower energy consumption (9 kWh/kg Al) have large weight and volume [9].

Development of inert anodes with higher energy efficiency, longer lifetime and lower CO, CO2 and PFCs

(Perfluorocarbons) emissions has also been studied for a long time. Despite relative success in test

operations, they have not been yet implemented industrially. This is basically due to problems such as high

capital cost [19], failure of ceramic and cermet anodes due to thermal shock [20] and aluminium contamination

due to anode corrosion and wear [20, 21].

Reactions: Although the mechanism of alumina electrolysis in cryolite is not completely understood, it is

generally accepted that cryolite decomposes to sodium, fluorine and tetra/hexa flouroaluminate ions [2, 11].

5

Alumina dissolves in cryolite at 950-970 C and forms aluminium-oxyflouride ions which react with anode to

produce carbon oxides and hexaflouroaluminate ions. Theses ions discharge on the cathode to produce

aluminium and sodium fluoride [2, 11, 12]. Although the formation of CO is thermodynamically favoured at

temperatures as high as 1010 C, the presence of considerable overvoltage (difference between reversible

and polarization potentials) changes the thermodynamic equilibrium and a mixture of CO and CO2 is produced

[10, 12]. The overall reducing reactions can be written as:

Al2O3+3/2 C=2Al+3/2CO2 G=264460+3.75TlogT-92.52T cal (1.1)

Al2O3+3C=2Al+3CO G=325660+3.75TlogT-155.07T cal (1.2)

Boudouard reaction may also take place as a side reaction:

CO2+C2CO G=40800-41.7T cal (1.3)

CO2 reacts with carbon dust as well as the anode to form carbon monoxide. Formation of CO consumes twice

as much carbon as required for CO2 formation. Carbon consumption shows that the primary anode product is

CO2. By increasing the current density up to 1 A/cm2, the proportion of CO2 increases and carbon consumption

decreases [10, 18, 22].

Energy consumption: The electrolysis of alumina is the most energy consuming operation from mine to

metallic aluminium [9]. Theoretically, with current efficiency of 100%, 5.64 kWh is required to produce one

kilogram of aluminium at 977 C with no backward reaction, while it will be 6.34 kWh when heating of

reactants to reaction temperature is taken into account [10]. But more power is consumed because the current

efficiency is less than 100%. The main reasons are: electrical resistivity of anode, variations in the electrical

conductivity of electrolyte due to alumina dissolution, reaction of molten aluminium with cathode [9] and

reduction of CO2 by molten aluminium [10].

Current efficiency is affected by bath temperature, cryolite ratio, anode-cathode distance (ACD), current

density [9] and carbon dust [23]. The average energy consumption is near 15 kWh/kg Al , which has been

decreased 2.5 times over 100 years [11]. The lowest energy consumption achieved today is reported to be less

than 13 kWh/kg Al for modern and high amperage cells with a current efficiency of 95% [24]. Aluminerie

Alouette in Canada has reduced the consumed energy from 14 kWh/kg Al to 12.7 kWh/kg Al by choosing the

best operating parameters, reducing the heat losses in stacks and training the operators [25].

Carbon consumption: Theoretical carbon consumption is 0.33 kg/kg of aluminium [9, 10] while the real anode

consumption is more than 0.4 kg/kg Al for prebaked anodes [24] due to the current efficiency and the excess

consumption (Figure 1.3). Carbon consumption decreases with increasing current density [10, 18] and anode

density. Higher calcination temperatures of coke [26] and higher baking temperatures [10, 17] lead to larger

crystallite size and more homogeneous structure of pitch coke and filler coke which reduce electrical resistivity

6

and anode consumption. There are also references claiming that under-calcination reduces anode

consumption [27, 28]. Excess carbon consumption is determined by the chemical loss due to anode reactivity

with air and CO2, and the physical loss due to dusting, low flexural strength and low thermal shock resistance

(TSR) [9]. Selective burning of binder results in carbon dust formation without contributing to alumina reduction.

Figure 1.3: Anode consumption figures

1.3 Problems

Aluminium industry challenges include high energy consumption, pollutant emissions, high capital investment

and variations of raw materials sources that result in anode quality variation. Although inert anode technology

may overcome most of these issues, it does not seem to be ready for industrial implementation in the near

future. On the contrary, optimization of existing carbon anodes to improve energy efficiency and to reduce

emissions may have immediate application. Consistent high quality anodes are a basic requirement of anode

plants and are of more importance than the quality of individual anodes. Variations in raw materials properties

are considered as one of the most significant challenges in anode manufacturing industry which affect anode

quality and consistency. The variations happen due to market situation and deficit in anode grade raw

materials. For example, there is a regional deficit in coal tar pitch in North America and a recent interest to use

non-traditional cokes in the anode formulation.

When the materials properties are changed, the paste formulation and process parameters including mixing

variables and compaction parameters should be re-adjusted in such a way to compensate for the effects of the

variations and to keep the anode quality consistent. A large number of materials properties, process

parameters including coke characteristics and granulometry; and paste formulation exist which can affect

anode consistency [29-32]. After decades of research, extensive works are still required to know the individual

and simultaneous effects of these factors. Such an enormous number of variables involved in anode

manufacturing process makes it difficult to control and optimize the final anode quality. This, in turn will lead to

low consistency.

1.4 Hypothesis

Reducing the large number of materials properties and manufacturing process parameters into few paste

quality indices helps to work with smaller number of variables and to evaluate the influence of variations on the

Electrolytic cons, 334/CE

Theoretical cons, 334 kg C/t Al

Al2O3+3/2 C=2Al+3/2CO2

Excess cons.

Gross cons.

Net cons. Butts

7

properties in an intermediate step before compacting and baking. This is of great importance since the general

approach in the industry is to characterize green and baked anodes and there is no control over the paste

quality before anode forming. Compaction behaviour of anode paste is a paste quality index to study the

influence of materials variations, paste formulation and mixing parameters. This index can be used to adjust

the compaction process regardless of the factors in previous steps.

1.5 Objectives

The main objective of this project was to reduce the impact of materials variations on anode quality. Studying

the influence of physical properties of raw materials, paste formulation and mixing variables on the quality and

compaction behaviour of the paste helps to achieve the project objectives.

This project had four specific objectives. The first objective was to define a paste quality index to evaluate the

paste quality before compaction. This index should be a representative of all materials and process

parameters existing before compaction of the paste. Moreover, it should be sensitive to the variation of these

parameters. Compaction behaviour of the paste was used as the paste quality index. A test setup was

designed and validated to determine the index.

The second objective was to evaluate the influence of mixing parameters on the paste properties, compaction

behaviour of the paste and homogeneity of the green anodes. In this regard, mixing time and temperature

were used as mixing variables.

The third objective was to study the effect of pitch content and size distribution of fine coke on the compaction

behaviour and density of green anode. Dependence of green density on the compaction rate was also

determined for different paste compositions.

The fourth objective was to study the influence of particle shape, texture and density on the compaction of

anode paste. Similar paste formulations were made using different sources of coke and the compaction

behaviour and final green density were evaluated.

2 Chapter 2: Literature review

Final quality and performance of prebaked anodes are affected by a combination of factors such as raw

materials properties, paste formulation and production parameters. In the first part of this chapter, different

types of raw materials and the effects of their characteristics and proportions on anode quality are reviewed. In

the second part, mixing is considered as a part of the production process, while forming and baking are not the

subjects of this research.

2.1 Raw materials

Calcined petroleum cokes, together with recycled anode butts are used as aggregates to produce green anode

blocks using binder pitch. There are different types of coke and pitch with varying constituents and properties.

In addition to the types and properties of raw materials, influential factors such as granulometry of particles

and proportion of materials in the paste (formulation) determine the final anode density, electrical resistivity,

strength and reactivity. These will be considered in the following subtitles.

2.1.1 Calcined petroleum coke

Calcined petroleum coke is the principal carbonaceous material for making anodes and has the highest mass

fraction in anode paste with at least 60%. Coke is used in the forms of particles with different size fractions as

aggregates, and dust (fine) to produce binder matrix with pitch. Petroleum coke is a mix of coal and heavy

hydrocarbons and is obtained from thermal decomposition and polymerization of residuals of raw petroleum

distillation [33]. Table 2.1 shows its typical composition. Petroleum coke, hereinafter coke, is preferred to

metallurgical coke (coke of coal) for using in aluminium industry due to its lower surface area, volatile content

and ash content which extend the service life [34].

Table 2.1: Typical composition of petroleum coke [33]

Carbon Hydrogen Nitrogen Sulphur Oxygen

88-95% 3-4% 1-2% 0.6-6% 1-7%

There are two types of coke according to the coking technology: fluid coke and delayed coke [33]. More than

90% of coke used in aluminium industry is delayed coke [33]. Ross et al. [35] found that replacing a fraction of

delayed coke with fluid coke has detrimental effects on wetting properties and anode quality. When they

increased the percentage of fluid coke from 0% to 18%, the distortion of stub hole increased, air reactivity

residue decreased and anode consumption increased by 24 kg/ton Al. They found that un-milled fluid coke is

not wetted by pitch and accumulates on the outer edge of agglomerates.

10

There are three types of coke according to their texture: anisotropic, isotropic and sponge coke [33]. The coke

derived from aromatic residuals is anisotropic or needle coke [33, 36]. Properties of this kind of coke change with

orientation. Figure 2.1-a shows the microtexture of anisotropic cokes. It has coarse flaws and lamellar domains

up to 100 m in length [37]. It is relatively soft and is easy to become graphite [33]. It has small volumetric

density [33], small coefficient of thermal expansion (CTE) [33, 36, 37], high porosity [33], high real density [36],

unidirectional and interconnected pores [33] which are wider in the middle and taper to a point toward the ends

[36]. This type of coke affects anode properties such as thermal expansion, reactivity and electrical resistance [33].

When the origin of the coke is asphaltic residuals, it is called isotropic or granular coke [33, 36]. This exhibits

similar properties in all directions. It has fine mosaic texture of about 1 m in diameter [37] and is difficult to

graphitize [33]. The isotropic coke has slitted-closed pores [33, 36], high volumetric density [33], low real density [33,

36], low crystallinity [33] and high coefficient of thermal expansion [33, 36, 37] which reduces the thermal shock

resistance. Isotropic cokes are more reactive because oxidation is kinetically easier on non-graphitic surfaces

[38]. Porosity in isotropic cokes tends to smaller size distributions [39]. They are harder to grind and result in finer

particle size distribution and higher amount of fine coke comparing to anisotropic cokes [40]. Table 2.2

compares the properties of anisotropic and isotropic cokes. Shot coke is an isotropic coke that is coproduced

with isotropic and anisotropic cokes. The origin of shot coke is asphaltene portion of feedstock with high

carbon residue (H/C molar ratio<1.5) [41]. Shot coke consists of small individual spheres, typically smaller than

5 mm in diameter, or clusters of 15 mm. During carbonization process, rapidly carbonizing constituents

segregate from a less reactive matrix. The viscosity and surface tension of the segregated region is higher

than the surroundings and eventually it takes a spherical form to reduce interfacial energy [42].

The coke of paraphinic-naftenicos residuals is sponge coke that is generally the best coke for anode

production [33]. The formation mechanism and coke structure determine the coke properties. Hardness and

density of sponge coke increase with decreasing the asphaltic content [33].

a) Anisotropic, needle coke b) Isotropic, granular coke

Figure 2.1: Microtexture of anisotropic and isotropic cokes (100x) [36]

11

Table 2.2: Comparison of needle and granular cokes [33, 36, 37]

Coke type Origin Texture Graphitization Volumetric density

CTE Reactivity Real density

Pores

Anisotropic (needle)

Aromatic residuals

Coarse flaws

Easy Small Small Lower High Unidirectional and interconnected

Isotropic (granular)

Asphaltic residuals

Mosaic Difficult High High Higher Low Slitted-closed

The demand for anode grade cokes is going to exceed the supply. There is thus a general interest to use a

mix of anisotropic and isotropic cokes to take advantage of useful properties of both of them. Neyrey et al. [36]

studied the effects of coke structure on anode properties and reported that it influences coefficient of thermal

expansion, reactivity and real density without changing other properties. When the percentages of needle and

granular structure changed from 53% to 26% and from 0% to 14%, respectively, the CTE increased from 3.72

to 4.33 (10-6/K) and real density decreased from 2.095 to 2.074 (g/cc) [36]. Neyrey et al. [36] also examined the

effects of non-traditional cokes on anode quality. Their results showed that non-traditional cokes with coke

structures close to traditional cokes had no significant detrimental effect on bench scale anodes. But, they may

negatively affect the density, reactivity and CTE of real scale anodes due to some granular structure they

have. The negative effect on density can be overcome by optimization of anode formulation and the forming

process [36]. Addition of 20% of shot coke to anode formulation did not reveal a catastrophic impact on anode

performance in industrial scale [43]. This is in contrast to the general belief in aluminium industry that shot coke

with high CTE and low open porosity results in higher susceptibility to thermal shock cracking, lower

mechanical strength, and dusting problem. Low open macroporosity reduces the penetration of pitch and thus

bonding of the structure together. When 20% of shot coke was added to the fine fraction, CTE, density and

pitch demand did not change greatly but air reactivity was deteriorated significantly. Addition of the same

portion of shot coke to coarse fraction led to an increase in CTE and density of anode but pitch demand was

reduced. Great improvement in baked apparent density was explained by higher apparent and vibrated bulk

density (VBD) of shot coke. Anode reactivity did not change significantly.

Before being used in anode paste, green coke undergoes calcination for several reasons such as increasing

C/H ratio, grain strength, thermal conductivity and purity, and reducing electrical resistivity, air reactivity and

shrinkage during baking of anode [33]. Calcination temperature has a great impact on coke properties

specifically the reactivity and the crystallite height (Lc). Oxidation rate (R) of carbon is associated with

temperature (T), activation energy (E) and pre-exponential factor (A):

R=A exp (-E/RT) (2.1)

According to TGA experiments of Senior et al. [34], activation energy of green coke for air oxidation is slightly

higher than that of a calcined coke but the pre-exponential factor of green coke is about 10 times greater which

12

results in a higher air reactivity of green coke. Crystallite height (size) and real density of coke increase with

higher calcination temperatures, but the effect is more significant for crystallite height [26]. The probable

reasons are that crystallite height increases in greater percentage than real density as long as no significant

desulphurization occurs and real density slightly decreases when sulphur is volatilized [26]. The results of

Belitskus [29] also showed that graphite-like structure of coke increases with higher calcination temperatures

and this leads to a lower rate of air oxidation of anode.

Crystallite height and real density depend on type or nature of coke and are also indicators of equivalent

treatment temperature or degree of baking. Heat-up time and soaking time at final temperature are considered

to determine the equivalent temperature [44]. According to the laboratory results of Dreyer [44], real density of

baked anodes increases with equivalent temperature leading to a lower CO2 and air oxidation. The increase in

crystallite size and real density of coke is more significant at temperatures above the calcination temperature.

Coke properties affect anode properties and several investigators have studied the impacts of coke

characteristics on anode quality. Belitskus and Danka [26] found out that BET surface area, total pore area

derived from mercury porosimetry, and VBD of coke fractions (except the -200 mesh particles) are related to

the range of particle size. There is however no correlation between VBD and BET surface area. This is more

likely because the surface area is a function of pore size while VBD is affected by total porosity [26]. Optimum

pitch content, green and baked apparent density of anode were related to the VBD of coke and porosities

smaller than 5 m which were not accessible to pitch [26]. Electrical resistivity of anode was well associated

with electrical resistivity of coke and to some extent with VBD [26] which reconfirms the results of Fischer and

Perruchoud [30] who had reported the relationship between specific electrical resistance of anode with electrical

resistivity of coke and final baking temperature. Belitskus [29] studied the effects of several factors on anode

quality. Figure 2.2 reveals that the impacts of coke variations and baking temperature on baked apparent

density, electrical resistivity and anode consumption are greater than those of granulometry (-8+200 mesh)

and binder level (1% below optimum level). He believes that pitch has a limited ability to penetrate pores

smaller than 5 m in diameter and that the pores filled with pitch will contain low-density binder coke after

baking. Therefore, particle size and binder level cannot effectively compensate the coke variations. Jonville et

al. [31] reported the influence of coke source on industrial scale anode quality in two aluminium plants with

similar technologies. They have confirmed the relationship between apparent density and electrical resistivity

of anode with the same properties of coke. Anode thermal conductivity increases with increasing coke bulk

density and baking temperature [32]. The impact of coke bulk density was 0.4 W/mK when it was increased

from 0.78 g/cm3 to 0.88 g/cm3 [32].

13

Figure 2.2: Relative effects of coke variations, formulation and baking temperature on anode properties [29]

Gas permeability and coefficient of thermal expansion of anode are principally affected by the fraction of open

pores larger than 50 m [30] and coke characteristics [30, 37], respectively. Thermal shock resistance of anode

increases with increasing grain stability [37, 45] and decreasing CTE and isotropy of coke [37]. Grain stability

depends on large cracks and flaws [37, 45] as well as calcinations conditions [37] and correlates well with flexural

strength of anode [45, 46]. Results of Perruchoud et al. [45] showed that flexural strength of vibrated anodes

increased from 7 MPa to more than 14 MPa when grain stability of coke increased from 60% to 95%.

Breakage of coke grains results in creation of surfaces without wetting with binder and produces weak points

in the anode. Although Grain stability is more important for pressed anodes, to some extent it determines the

transgranular breakage of vibrated anodes due to thermal shock and mechanical stress.

Contrary to those authors who associated coke properties with laboratory scale anodes, Leach et al. [46]

investigated the relationships between coke intrinsic characteristics and real scale anode properties. They

claimed that some coke properties which are usually considered as effective factors, did not show significant

effects. For example, they found weak relationships between air and CO2 reactivity of coke and the same

properties of anode, while air and CO2 reactivity of anode were increased with increasing V and Na content of

coke, respectively. This is in contrast to the results found by Fischer and Perruchoud [30] that showed the

dependence of anode air reactivity on coke air reactivity and anode baking temperature.

Surface characteristics of filler aggregates affect their wetting behaviour with pitch [35, 47-49]. Adams [50]

conducted a research to identify the influence of surface properties of coke and butt particles on wetting and

pitch penetration. He found that pore size distribution of coke and butt were similar and any difference in pitch

penetration might have been due to differences in wetting behaviour of filler particles. Separate penetration

tests were conducted for coke and butt with pitch content in the range of 19.5 to 22%. For the entire range of

Effe

ct o

n b

aked

ap

par

ent

den

sity

, g/c

m3

-0

.04

-0

.08

-0

.12

Maximum 1250C

950C

1% below optimum

level

All cokes

Commercially calcined cokes

All cokes

Commercially calcined cokes

1% below optimum

level

Minimum 1250C

1150C

1030C

950C

Co

ke

vari

atio

ns

P

arti

cle

siz

e

dis

trib

uti

on

Bin

der

le

vel

B

akin

g te

mp

erat

ure

Co

ke

vari

atio

ns

P

arti

cle

siz

e

dis

trib

uti

on

Bin

der

le

vel

B

akin

g te

mp

erat

ure

+4

+8

+12

+

16

Effe

ct o

n e

lect

rica

l res

isti

vity

,

.m

Co

ke

vari

atio

ns

P

arti

cle

siz

e

dis

trib

uti

on

Bin

der

le

vel

B

akin

g te

mp

erat

ure

Minimum

1150C

1250C

950C

1030C

Effe

ct o

n c

on

sum

pti

on

du

rin

g el

ectr

oly

sis

% o

f th

eore

tica

l

+3

+6

+9

+

12

No consistent

effect found

14

pitch, penetrated coke particles revealed lower mass loss during TGA tests. A thicker pitch film around the

coke particles was also measured by image analysis. This confirms a better wetting and infiltration of pitch into

recycled butt aggregates. Coke particles, on the other hand, were hydrophobic and their surface tension was

lower than that of butts. Butt grains had more oxygen groups than cokes because they had been in contact

with CO2, air and cryolite at a high temperature and thus they are expected to have different surface chemistry

[50].

2.1.2 Pitch

It is hypothesised that pitch has three roles in the anode paste; being adsorbed on the surface to wet the

particles, being penetrated in the coke pores, and acting to bond filler aggregates together to provide sufficient

strength in green blocks (Figure 2.3). Pitch is decomposed during baking and forms strong bonds between

coke grains through carbonization to enhance strength and electrical conductivity.

Figure 2.3: Pitch-aggregate system [51]

Coal-tar pitch (CTP) has been used as a binder in anode paste from the beginning. Tar is produced during the

coking of coal for producing metallurgical coke [33, 52, 53] and its quality and constituents depend on the

operation conditions in the coking ovens. Coal-tar pitch is an aromatic hydrocarbon being formed after

distillation of tar of coal [33, 52, 53]. It contains 5 types of solid particles with different natures:

1- Primary quinoline insolubles (QI): They are carbonaceous spheres with less than 2 m in diameter

and C/H ratio more than 3.5 [52] which are formed by thermal cracking and incomplete combustion of

tar volatiles [33, 52].

2- Secondary quinoline insolubles or mesophase: These particles are formed by polymerization of

aromatics at high temperatures [33, 52]. Formation of mesophase is an irreversible transition from

amorphous to liquid crystals [54]. In this transition the carbon atoms in high molecular weight

hydrocarbons reorder in two-dimensional layers. Mesophase formation is a nucleation and growth

Aggregate dust in cohesive volume

Absorbed binder

(penetration)

Binder

Aggregate grain

Inaccessible porosity

Gas void

(required for outgassing

and cohesive capillary forces)

Boundary layer

(adsorbing pitch)

15

process which begins at about 420 C to reach spheroids or globules with a diameter of 100-150 m [54].

Mesophase particles are larger than primary QI, with more hydrogen [55] and a C/H ratio of less than 3

[52]. They are optically anisotropic and visible in polarized light [54].

3- Carry over: They are dragged particles of coal and coke which come out with tar vapours and are

larger than primary and secondary quinilone insolubles [33, 52].

4- Carbon black: These particles exist and sometimes are intentionally added to pitch to increase the QI

content [52].

5- Ash [52]

Lack of coal tar [56], large amount of Polycyclic Aromatic Hydrocarbons (PAH) in coal-tar pitch [57] and

increasing demand for pitch in aluminium industry resulted in using petroleum tar in combination with coal tar

to produce binder pitch [55]. Petroleum pitches (PP) are more fluid [58] with lower PAH [59] and degree of

aromaticity [55] than coal-tar pitch, but they have higher VOCs [59], lower QI content [58, 59] and lower carbon yield

[58, 59]. QI particles are more reactive than anisotropic carbon materials [58]. Thus, petroleum pitches with low QI

exhibit lower air reactivity and dusting [58]. Perez et al. [58] revealed that anodes made with petroleum pitch have

lower compressive strength and higher electrical resistivity than coal-tar pitch anodes [58]. Fernandez and

Alonso [59] studied the pitch of anthracene oil as another alternative binder for anode production. Anthracene

oil is the heaviest fraction of coal tar distillation with low PAH and reduced volatility and is competing petroleum

pitch. Pitch of anthracene oil has less QI than does CTP but it wets petroleum coke and can be used in high

ratios (50-80%) with CTP, while petroleum pitches cannot be used in ratios higher than 30-40% [59].

Anthracene oil pitches are volatile-free while petroleum pitches have 4-5% of volatiles. This is an advantage of

anthracene pitches when considering the carcinogenic effect of VOCs in working area. Coke of anthracene oil

pitch has graphite-like structure which may optimize electrical conductivity [59].

Pitch is usually characterized by the following physico-chemical properties [48, 49, 54, 55, 59-61]:

Selective fractioning of constituents by using quinoline and toluene as solvents to derive quinoline

insoluble contents with a molecular weight of above 3000 and toluene insoluble (TI) contents with a

molecular weight of above 1000 [55].

Softening point

Coking value

Viscosity

Density / specific gravity

16

Volatile content

Ash content

Sulphur, carbon and hydrogen contents

C/H ratio for resins (QI) and resins (TI-QI)

Although the determination of these parameters gives a relative picture of pitch quality, in some cases, they do

not explain why some pitches behave badly and result in low quality anodes [48, 49, 60, 62]. It may be due to the

fact that the wetting ability of pitch is not considered in these characteristics. Determination of pitch wetting

ability may be helpful to choose the appropriate formulation and granulometry of raw materials as well as the

optimum process parameters to obtain high quality anodes [62]. The presence of secondary QI or mesophase

which can be determined by microscopy techniques [33, 54, 60] and low (C/H) ratio [52, 60] results in lower binder

properties and has a detrimental effect on wetting ability and infiltration of pitch into coke pores [49, 60]. A certain

amount of mesophase larger than 50 m reduces the wetting and penetration of pitch in the coke [55, 60]. The

mechanism of mesophase influence on wetting properties and anode quality is explained in the literature [60]. In

contrast to mesophase, resins affect the anode quality positively [63]. Compared to amorphous materials,

resins have better graphitizability and result in graphitized product with lower electrical resistivity [63]. However,

pitches with higher resin levels require higher temperature for full penetration [49].

In addition to the quality and content of pitch, characteristics of filler coke and anode butts such as surface

properties and particle size distribution are secondary factors affecting pitch penetration. These factors are

discussed in sections 2-1-1 and 2-1-4, respectively. Test methods to determine wetting ability and penetration

time of pitch have been explained by several authors [48, 49, 62].

The content and size of quinoline insoluble or mesophase vary with temperature and have an interrelation with

softening point [53-55]. Heat treating of pitch results in growth of mesophase, increase of QI content [55] and

higher coking values [64]. During vacuum flash distillation the pitch is exposed to a maximum temperature of

about 325 C for a few minutes and a low-mesophase pitch with softening point of 110 C is obtained [53]. On

the contrary, in conventional distillation, heat treating is performed at temperatures above 400 C for a few

hours resulting in formation and growth of mesophase [53]. Vacuum distilled pitch with 0.2% mesophase results

in less pitch accumulation on filler (better wetting), higher baked density, improved air and CO2 reactivity in

comparison with conventional pitch with 2.9% mesophase [53]. High softening point pitches have more and

larger mesophase due to processing at high temperatures [54]. Larger size of mesophase leads to higher

viscosity of pitch [54] and longer time is required for full penetration into the coke [48, 62]. When pitches with

higher softening point are used, higher pitch content and temperature is required to achieve the maximum

17

compaction [65]. According to Perruchoud et al. [55], pitches with 6-8% of QI are the most frequent class with

more than 30%. Two thirds of distillers produce pitches with a softening point of 110-115 C.

Quinoline insoluble content of the pitch affects the pitch requirement [61] and anode properties [52, 61, 66]. Rhedey

[61] compared the quality of anodes derived from pitches with low (5.1%) and medium (11.1%) QI contents and

revealed that a decrease of 1% in QI content reduces the pitch consumption and coke yield by 0.12% and

0.75%, respectively. This is in agreement with Alscher [52] findings. The reduction of coking value is due to

lower C/H ratio in low QI pitches [61]. Compressive strength also decreased by 17 kg/cm2 when medium QI

pitch was replaced with low QI pitch [61]. Anodes made with a mixture of low and medium QI pitches showed a

lower strength and higher anode consumption than the anodes made with a reference pitch (12% QI). Low QI

pitch did not show negative effects on other anode properties [61]. Alsher [52] also reported that pitches with low

QI content have performed as well as pitches with 10-14% QI but with lower electrical resistance. Anodes

made by medium QI content pitches (6-8.5%) produced by vacuum distillation revealed lower electrical

resistivity, CO2 reactivity, and permeability with higher green and baked density compared to high QI pitches

[52]. McHenry et al. [66] obtained a more anisotropic coke structure, higher real density of pitch coke and lower

coking value using low QI pitches. During baking, pitch covers the grains and partially diffuses into the pores.

After volatilization, it converts to anisotropic binder coke and flaw domains form that decrease the mechanical

strength of anode [52]. The presence of 6-8.5% of dispersed primary QI prevents the formation of flaw domains

[52]. More QI form clusters which has no more advantage, while it increases the pitch consumption for sufficient

wetting of solid particles [52]. It can be concluded that low and medium QI content pitches can be used alone or

in combination and heat treatment, that results in mesophase formation and poor anode properties, is not

necessary to increase the QI content.

Pitch content in anode formulation is an effective factor for anode quality. Proper amount of pitch wets the

particle surfaces, fills the free space between the cokes and leaves just enough space for pitch expansion

during baking (Figure 2.4) [51]. This prevents any excessive shrinkage, expansion, crack formation and porosity

in the structure. Sverdlin et al. [65] classified anode pastes into four classes. Weak bonded structures with

negligible amount of pitch have similar properties to coke mixtures and interactions between the structural

elements are attributed to friction forces. For low strength structures with a maximum pitch of 16%, the thin

adsorbed layers of pitch on particle surface act as lubricant. Strong structures contain about 20% of pitch and

physico-chemical forces are attributed to wetting and impregnation of coke particles that increase the strength.

In plastic structures with more than 20% pitch, particles are separated by pitch layers and strength is due to

intermolecular forces in the binder.

18

Figure 2.4: Binder-aggregate structure [51]

Various models have been developed and used to determine the optimum pitch demand [51]. Optimum pitch

content is usually determined based on the maximum baked apparent density and depends on the following

factors [51]:

1- Raw materials properties: porosity and surface properties (wettability) of coke and wetting behaviour

of pitch.

2- Paste formulation: granulometry of aggregates, butt content and fineness and content of dust. Pastes

with finer particles have larger surface area and consequently need more pitch to achieve the

maximum density.

3- Processing parameters: preheating temperature, mixing time and temperature, and forming technique

and conditions. The required amount of pitch depends on the pitch distribution throughout the paste.

Inadequate mixing temperature and time result in the formation of inhomogeneous paste and higher

amount of pitch is needed to improve this situation. Apparent density of pressed anodes is rather

insensitive to paste viscosity, determined by pitch content and temperature, while vibrated anodes

have a sensitive behaviour. At high forming pressures lower amount of pitch is required to obtain the

maximum baked density.

Hulse [51] used production and pilot data. She summarized the influence of individual parameters on pitch

requirement, as shown in Table 2.3.

Maximum green density increases with higher pitch content [51, 65]. Baked apparent density, thermal

conductivity, flexural and compressive strength of anode are increased with increasing pitch content while

electrical resistivity, air permeability, coefficient of thermal expansion, air and CO2 reactivity are decreased [40,

66]. All these characteristics except green density show an optimum and the peak occurs at optimum pitch

content. Ross et al. [35] reported that distortion of stub hole occurs with increasing pitch content over the

optimum amount.

Optimum

Void

Underpitched Overpitched

Coke particle Unbound pitch

(excess)

Surface adhesion pitch

(boundary layer)

19

Table 2.3: Pitch sensitivity values [51]

Property Determined pitch content

Bulk density 0.01 g/cm3 0.125%

Dust (3000 Blaine) content 10% 1%

Dust fineness (at 30% dust) 1000 Blaine 1-1.2%

Mixing temperature 1 C 0.15%

Softening point (Mettler) 10 C 0.55-0.7%

Primary QI content 1% 0.11-0.14%

Mesophase content 1% 0.167%

Butt content 10% 0.6%

2.1.3 Anode butt

The life of an anode in a typical modern cell is about 25 days [67]. After removing, they are cooled, cleaned

from bath contaminants, crushed and recycled. The recycled anode butt is used as aggregate in anode paste

and its physical, chemical and mechanical properties affect the baked anode quality [68-70]. Anode butts are

classified into hard and soft categories. Hard butts are similar to the original anodes while the soft butts are

more severely attacked during electrolysis. Soft butts are more porous and have a lower mechanical strength,

thermal conductivity, ignition temperature and density but higher permeability and reactivity than do hard butts.

Increasing the anode butt content from 5% to 35% resulted in an increase in apparent density and flexural

strength of anode and a decrease in air permeability [70]. The effect of hard butts on anode density and

strength is much more evident than that of the soft butts. Table 2.4 shows the summary of Fischer and

Perruchoud [70] results on anode butts. Recycling of soft anode butts in a closed loop has a detrimental effect

since anode porosity may increase gradually and the air burning of butt grains may become preferential rather

than selective burning of binder matrix [70].

Table 2.4: Influence of butt quality on the anode properties [70]

Anode butts

Anode properties Hard clean Hard dirty Soft clean Soft dirty

++ Very positive

+ Positive

Neutral

- Negative

-- Very negative

--- Catastrophic

Apparent density ++ ++ + +

Mechanical strength ++ ++ + +

Permeability ++ ++

CO2 reactivity residue - - --

Air reactivity residue - -- ---

Ignition temperature - -- ---

Purity - -- - --

The amount and particle size of anode butts is important to anode quality. Belitskus [68] demonstrated an

increase in baked apparent density from 1.46 to 1.50 g/cm3 with increasing the butt content from 0% to 40%.

This increase is in agreement with Fischer and Perruchoud [70] finding. Electrical resistivity and shrinkage

20

during anode baking decreased slightly with increasing the butt content up to 40%. Carbon consumption did

not change significantly but it varies from case to case and depends on butt particle size and contamination

with bath materials [68]. Optimum pitch content was decreased from 18% to 17% with increasing butt content

from 0% to 40% [68]. It can be due to the better wettability or higher density of butts than coke grains. It should

be noted that these results may not be applicable to industrial scale anodes, because the butt particles larger

than 9.5 mm were too large for a bench scale sample of 50150 mm and have been discarded in this work.

Weng and Vera [69] studied the effect of size and amount of butt particles. For laboratory scale anodes, the

baked density increased by 0.01 g/cm3, electrical resistivity decreased by 10 m and air burning decreased

by 5 mg/cm2 h with reducing the maximum butt particle size from 38 mm to 19 mm at a constant butt fraction of

25%. The results for full size baked anode cores did not show the same trend. Anode cores with 25% of 19

mm butt particles had lower density, electrical resistivity and coefficient of thermal expansion, but higher

flexural strength and air burning than those containing 38 mm butt grains. They found that the quality of

anodes containing 19 mm particles was improved with increasing the butt content from 25% to 30% while the

quality of anodes with 38 mm grains was slightly deteriorated. It is stated elsewhere that the worldwide range

of butt content up to 32% has an impact of 0.3 W/m K on anode thermal conductivity [32].

The air and CO2 reactivity of anode butts are influenced by porosity and concentration of inorganic elements.

X-ray computed tomography, chemical and TGA analysis have been used for reactivity measurements on

drilled cores as a function of position in anode butt [71, 72]. The results of X-ray tomography showed that 20 mm

from air-end and electrolyte-end had lower apparent density but higher real density than did the rest of the butt

(Figure 2.5) [71]. The volume percentage of pores for both the air and bath ends was 1.3 to 2.6 times greater

than that of the middle part [72]. Besides, the ends had the most concentration of inorganic elements such as

Fe, F, Ca and Na [71, 72]. Higher porosity and concentration of inorganic elements resulted in higher air and CO2

reactivity and lower half burn-off temperature [72]. This is in agreement with another study [70] where CO2 and air

reactivity increased with increasing the sodium content of butts and amount of soft butts (lower ignition

temperature due to porous nature), respectively. The active carbon sites in micropores are connected to the

outer surface by transitional and macropores which are diffusion paths for air and CO2 [73]. The concentration

profile and reactivity measurements showed that the layer just above the most bottom layer had much

inorganics and higher reactivity than the bottom layer. Suriyapraphadilok et al. [72] explained it by the fact that

the bottom surface has more porosities and the molten impurities move up with CO2 through the pores. The

upper layers have fewer and tighter pores where the inorganic elements accumulate.

Although higher porosity and concentration of inorganic elements at the ends of butt accelerate the burning

rate, the results of TGA analysis have shown that the reaction rate decreases with time and approaches to that

of the middle part [72]. It may be due to the decrease in gasifiable material and coalescence of pores which

21

results in lower reaction surface [71]. It is suggested that butts, specially 10-20 mm of the top, should be

cleaned of reactive parts before recycling to the anode production to avoid the increase in reactivity and anode

consumption [72].

Figure 2.5: The results of radial scanning of X-ray computed tomography [71]

The effects of other carbon material scraps on the anode quality are not the same as those of anode butts.

Anodes made with 100% mixer scraps had the same properties as anodes made with fresh paste [68]. Anodes

with 100% green scrap had higher baked apparent density of 0.02-0.03 g/cm3, lower electrical resistivity and

lower carbon consumption. It may be due to better pitch penetration into coke pores because of multiple

pressing of paste [68]. Using 0-25% potlining scraps, on the other hand, dramatically increased the carbon

consumption from 115% to 125% of the theoretical value and increased the green apparent density due to

high bath content [68].

2.1.4 Particle granulometry

Size distribution of coke and butt aggregates as well as fineness or size distribution of fine coke affect green

and baked density [40], compressive strength [40], pore size distribution [22], electrical resistivity [40, 47], reactivity

and even chemical composition of anodes [74]. Various packing theories have explained the influence of

Air Bath

0 20 40 60 80 100 120 140 160

0 20 40 60 80 100 120 140 160

0 20 40 60 80 100 120 140 160

1.700

1.600

1.500

1.400

1.300

1.200

1.100

2.050

2.025

2.000

1.975

1.950

10.0

20.0

30.0

40.0

50.0

Ap

par

ent

Den

sity

(g/

cm3)

Ab

solu

te D

ensi

ty (

g/cm

3)

Po

rosi

ty (

% V

ol.)

Distance from airburn side (mm)

22

continuous and discontinuous size distributions on packed density of particles [40, 75]. Figure 2.6 reveals the

influence of granulometry on anode properties and shows that there is not a specific size distribution that

maximizes all properties. An optimum ratio of small, intermediate and coarse particles is generally used in

anode formulation to obtain the optimum global quality [76]. One plant may use particles in the size range of

0.6-5 mm, 0.3-0.6 mm and <0.3 mm [77] while another plant may use 2-4 mm, 1-2 mm and 0-1 mm particles [78]

in different percentages. Wider ranges of particle size have higher VBD than narrow ranges of coarse

particles, since the smaller particles fill the intraparticle spaces without increasing the volume [76, 79, 80]. Besides,

the bulk density of closely sized particles increases with decreasing the particle size, because large pores are

annihilated with size reduction [79]. Wharton et al. [22] added recycled material with a wide range of

granulometry to calcined coke and reported a decrease in quantity of smaller pores which resulted in a

decrease from 128% to 120% of theoretical carbon consumption. This can be due to the fact that larger pores

have less surface area for chemical reaction. It should be noted that the granulometry of coke particles affects

the porosity between the aggregates that is not a great percentage of total pore volume. Rorvik et al. [81]

classified the anode pores into four classes and reported that the pores in the binding matrix represent a great

percentage of the total pore volume i.e., close to 70%.

Figure 2.6: Influence of granulometry on anode properties [40]

In addition to the amount and size distribution of aggregates, the granulometry of fine particles (dust) used in

the binder matrix also affects the anode quality. Vidvei et al. [82] used three paste formulations and studied the

effect of fineness and amount of fine coke. The VBD of both dust and mixture of aggregates and dust

increased up to a maximum with smaller sizes of dust. Figure 2.7-a shows that the maximum VBD for the

mixtures was around 1.35 g/cm3 when 60% of dust was below 75 m. The maximum VBD varies also with the

amount of dust. High concentration of dust shifts the maximum VBD towards coarser particles of dust while

low amounts of dust shifts this maximum towards finer particles. Paste viscosity revealed a minimum with

Coarse

Coarse Coarse

Coarse

Intermediate Intermediate

Intermediate Intermediate

Maximum

Fines

Fines Fines

Fines

Maximum

Minimum

Minimum

Porosity

Thermal Shock

Resistance

Resistance to

Compaction

Air

Permeability

23

increasing the fineness of dust. This minimum occurred at the fineness that resulted in maximum VBD. Over

this critical fineness, the viscosity of the paste was increased. Increasing dust content resulted in an increase

in the minimum viscosity (Figure 2.7-b). Electrical resistivity sharply decreased from 94 to 62 m when the

amount of particles smaller than 75 m increased from 10% to 45% (Figure 2.7-c). Green density did not

change with the amount and fineness of dust while baked density was improved (Figure 2.7-d). Tensile

strength, compressive strength and Young's modulus increased with finer dust particles and reached their

maximum value around the size range where maximum VBD occurred (60% of -75m). Gas permeability

decreased from 11 to 4 nPm with increasing the portion of -75 m particles from 20 to 60%. This agrees with

another work [83] which showed a 2 nPm decrease of permeability with smaller fine particles. Rorvik et al. [83]

used image analysis to determine the distribution of coke, pitch and porosity in green and baked anodes. They

revealed that the fraction of particles smaller than 75 m increases with longer milling times and results in

shorter distance between particles, thinner pitch layer and lower electrical resistance. Figure 2.8 demonstrates

that when the percentage of the inter-particle distance of less than 10 m increased from 50% to 80%, the

percentage of the pitch layer thinner than 10 m increased from 12% to 23% and the electrical resistivity

decreased from 92 to 62 m [83].

Fineness of fine coke has an influence on the shrinkage of anodes during baking [84]. For hard particles of coke

that lead to irregular particles after grinding, shrinkage decreases with increasing the fineness (% -63 m) of

fine coke. On the other hand, increasing the fineness of dust with rounded particles that are produced from soft

cokes increases the shrinkage.

24

Figure 2.7: Relationships between VBD of particles (a), viscosity of paste (b) and electrical resistivity (c) and baked apparent density of anode (d) with fineness and amount of fine coke for three formulations of fine, intermediate and

coarse particles [82]. Highest fine content, intermediate fine content, lowest fine content

Figure 2.8: The correlation of interparticle distance and pitch layer thickness, derived from image analysis, with electrical resistivity [83]

a b

c d

25

Improvements in the properties of real scale anodes with using finer particles of dust were reported [85].

Fineness is determined by Blaine number (BN) in the industry that is an indication of specific surface area.

Anode formulation was optimized in an anode plant to enhance anode quality and to reach higher amperages.

Blaine number of dust (<250 m) was increased from 3000 to 4000 during 8 years and baked apparent

density (BAD) was improved by 0.05 g/cm3. The influence of increasing the Blaine number from 4000 to 4400

and using different pitch contents was studied. At high Blaine numbers, underpitching occurred because of the

existence of a larger surface area resulting in lower green apparent density (GAD). About 0.9% more pitch was

required to reach the original value of GAD at BN of 4000. On the other hand, BAD was higher than its original

value and increased to 1.6 g/cm3 with increasing of pitch. Baked loss was less than original value of 5.6% but

increased with increasing of pitch. It was still below the original value at 15.6% of pitch. The more carbon loss

during baking is reduced, the more carbon is available for electrolysis. Contrary to other works [82, 83], some

results indicated that air permeability did not decrease with finer dust [85]. It was the same as the original value

of 0.6 nPm and decreased to 0.4 nPm with increasing the pitch content to 15.6%. The results of electrical

resistivity did not agree with those of Vidvei’s et al [82]. It was 2 m more than the original value of 56 m

but decreased to 54 m with increasing of pitch. Flexural strength did not change with increasing the BN for

400 point at constant pitch of 14.7%, but increased 8% with increasing the pitch to 15.6%.

Perruchoud and Fischer [45] reported that even low quantities of coarse dust result in high permeability and low

burning quality of the anode. However, it is stated in another work [37] that high amount of dust with high

fineness that requires a great amount of pitch, results in uneven distribution of pitch and baking cracks. Smith

et al. [86] investigated the effect of dust granulometry on the properties of bench scale binder matrix samples.

They used dusts with Blaine number of 1000 to 7000 and found that apparent density and mechanical

properties have a maximum for a Blaine number of 3000-4000. Permeability decreased with increasing the

fineness while reactivity revealed an opposite behaviour and increased. They explained that for coarser dusts,

there were more pores larger than 50 m that caused a sharp increase in permeability. On the contrary, the

number of 0.5-15 m pores, which have both required criteria for reactivity (being large enough for gas

diffusion and providing sufficient surface area) decreases [87] that will in turn lead to a lower reactivity. It should

be noted that they used constant pitch content (28%).

The influence of dust granulometry on the required pitch content and the properties of binder matrix (paste

without aggregates) has been investigated [88]. Increasing the fineness from 3000 to 6550 Blaine at a constant

pitch/dust ratio of 28/72 led to deterioration of GAD, BAD, electrical resistivity and strength. This is due to the

fact that finer particles have larger surface area and need more pitch to wet the surface. In fact, binder matrix

is more sensitive to pitch content than anode paste. Air permeability decreased and CO2 reactivity increased

for finer dust. Increasing fineness resulted in forming of a large amount of small pores that was responsible for

26

the changes in the air permeability and electrical resistivity. Higher total volume of pores increased the

reactivity. When binder matrix was made with varying dust granulometries from 1220 to 6550 and optimum

amount of pitch, GAD did not change with finer dust while BAD reached a maximum level at 4430 Blaine and

then decreased. Strength increased slightly and electrical resistivity revealed a minimum level at 4430 Blaine.

There was a tendency towards smaller pore sizes with finer dust and so air permeability decreased. CO2

reactivity revealed a great increase since finer granulometry needs more pitch and selective burning of pitch

coke increases the reactivity.

Specific surface area of particles changes with particle size and determines the pitch requirement [85] but the

relationship between particle size and surface area is not the same for different materials [22, 85]. When Wharton

et al. [22] reduced the particle size of petroleum and pitch coke from 150 to 70 microns, their surface area

increased by 1 m2/g and 10 m2/g, respectively (Figure 2.9). The difference is due to the amount of closed pore

volumes coming to the surface after grinding. Grindability of particles depends on the presence of medium size

internal pores and breakage happens through the pores [45]. A higher internal porosity of a particle, a higher

surface area and a lower internal porosity is expected after grinding.

Figure 2.9: Effect of particle size on specific surface area. A) pitch carbon 850 C, B) pitch carbon 1150 C, C) petroleum coke [22]

Size distribution of fine coke affects the penetration ability and distribution of pitch. According to Couderc et al.

[49] a particle size below 100 m is too small and does not allow sufficient penetration of pitch while a particle

size above 200 m is too large to prevent the pitch infiltration. Cao et al [48] found that coke dusts of -125 m

and -250 m had similar penetration behaviour with the same pitch while penetration into -500 m coke

particles was faster due to the larger inter-particle voids.

0 0.5 1.0 1.5

Particle size (m104)

Surf

ace

area

(m

2 g

-1)

0

5

10

1

5

A

B

C

27

Apart from the particle size of dust, the amount of dust is also important in anode quality. Hulse [40] showed

that green and baked apparent densities increase and air permeability decreases with increasing the amount

of dust with a 3000 Blaine at a constant pitch content (Figure 2.10). Experimental results of Vidvei et al. [82]

showed a slight impact of dust content on anode properties (Figures 2.7-c,d and 2.11). Tkac [89] used a half

fractional factorial design and claimed that dust (-63 m) amount alone does not affect the properties

significantly. According to his results, the interactions of dust content with vibropressure and heating rate affect

the CO2 and air reactivity, respectively.

Figure 2.10: Influence of dust content on anode properties. Adapted from [40]

Figure 2.11: Relationships between permeability (a) and tensile strength (b) of anode with fineness and amount of fine coke for three formulations of fine, intermediate and coarse particles [82]. Highest fine content, intermediate fine content, lowest fine content

2.1.5 Rheological properties

Temperature, pitch content and granulometry of particles, more specifically the fineness of fine coke, influence

the viscosity and rheological properties (flow behaviour) of anode paste. Binder matrix is the deformable phase

in the paste that fills the inter-particle voids. Several authors have investigated the rheological behaviour of

binder matrix. Gildebrandt et al. [90, 91], Kravtsova et al. [92] and Vershinina [93] demonstrated that the viscosity of

binder matrix increases significantly when the amount of fine coke exceeds over 50 wt.% in binder matrix. At a

high concentration of particles the structure changes from a loose configuration with lubricated contacts to a

a b

1.6

1.61

1.62

1.63

1.64

1.65

1.66

12 14 16 18 20

Gre

en a

pp

aren

t d

ensi

ty (g

/cm

3 )

Pitch content (%)

25% dust

35% dust

45% dust

1.5

1.51

1.52

1.53

1.54

1.55

1.56

1.57

12 14 16 18 20

Ba

ked

ap

par

en

t d

en

sity

(g/c

m3 )

Pitch content (%)

25% dust

35% dust

45% dust

0

1

2

3

4

5

6

7

8

12 14 16 18 20

Air

per

mea

bili

ty (n

Pm

)

Pitch content (%)

25% dust

35% dust

45% dust

28

close configuration with direct particle-particle interactions (Figure 2.12) [94]. Thus, friction and interactions

between the particles increase and becomes the dominant behaviour.

Figure 2.12: Particle-binder configuration in anode structure [94]

Viscosity of binder matrix decreases with increasing temperature and particle size of fine coke [90-93].

Experimental data showed that yield point (Py) and plastic viscosity (*) of binder matrix increases with

increasing the degree of dispersity (fineness) of fine coke particles resulting in a higher Newtonian viscosity

() [92]:

⁄ (2.2)

where is the shear rate. This was more evident at temperatures below 200 C. The number of contacts

increases with the increase in the dispersity of particles while the thickness of pitch layer between the particles

decreases. This leads to an increase in the interaction between the particles and the strength of pitch-coke

structure increments [92]. Hulse [94] also explained that particle surface area increases with increasing dust

content and fineness, and more pitch is adsorbed on the surface. As a matter of fact, phase volume and

therefore paste viscosity and elasticity increase. Noting that phase volume is the amount of space taken by

solid particles.

Behaviour of binder matrix changes from Newtonian to viscoelastic when the amount of fine coke exceeds 50

wt.% [95, 96]. For viscoelastic materials, viscosity and elastic behaviour contribute to deformation and

compaction. According to Hulse [94], paste has a granulo-viscoelastic behaviour. Viscosity decreases with

increasing the pitch content and temperature and thus has a larger contribution to compaction. On the other

hand, higher amount of coke and finer particles increase the elastic behaviour. Figure 2.13 schematically

shows how the sensitivity of paste to viscosity controlling parameters (temperature and pitch content) changes

with concentration of particles. At low deformation pressures, viscosity contributes to deformation and thus

Close configuration Loose configuration

Particle Lubricated contact Binder Direct contact

29

paste is sensitive to pitch content and temperature [94]. In other words, at high pressures the elastic behaviour

governs the deformation behaviour and paste is not sensitive to these parameters. It is reported that pressed

anodes in contrast to vibrated anodes, are not sensitive to viscous attributes of paste [94].

Figure 2.13: Paste sensitivity to viscosity controlling parameters [94]

Apart from rheological behaviour, viscosity of binder matrix influences the quality of filler-binder interface.

Pressure is applied during forming to eject the trapped air and to reach the desired density. Sometimes extra

pressure is exerted which results in the formation of a thinner binder layer and filler-filler point contacts. After

removing the load, stress relaxation and expansion occur that will lead to breakage of filler-filler interfaces. The

effect of stress relaxation on filler-binder interfaces depends on the viscosity of binder matrix. For low viscosity,

the filler-binder bridge will only stretch while for high viscosity the stress induced by drawing apart the filler

particles will exceed the strength of interface and completely or partially fissured interfaces will appear [97].

Poor interface has a negative impact on strength [97-99] and, to a lesser extent, the electrical resistivity [98, 99].

Fissured interfaces are associated with large particles, because stress concentrates around the large grains.

2.1.6 Particle shape and texture

Shape, porosity and roughness of coke and butt aggregates influence the flow and packing characteristics of

particles and consequently the anode density. Vibrated bulk density is traditionally used to determine the

required pitch content and sometimes is used as an indicator for internal porosity (within porosity) of particles

with the same size [100]. There is however a paradox between VBD and particle porosity. Experimental results

of Bowers et al. [76] showed that bulk density of natural particles is higher than that of crushed particles of the

same size and source. The crushed particles have more inter-particle porosity and result in lower anode

density. It indicates that size distribution and intra-particle porosity are not the only factors affecting the bulk

density.

Increasing Concentration of Solids

Vis

cosi

ty C

on

tro

llin

g In

flu

ence

e.

g. T

emp

erat

ure

Dep

end

ence

Low Voidage Medium Voidage High Viodage

Combined effect Temperature insensitive region

Liquid phase dominates temperature effects

Interparticle friction effect

Liquid viscosity effect

Boundary layer thickness decreases and particle friction

increases with viscosity of boundary layer

therefore mix becomes temperature dependent

30

Compactness, roughness and elongation of particles are the measures determining particle shape and

consequently flow behaviour and packing density. These shape factors can be measured by fluid imaging and

image analysis [76]. Compactness is an indicator of the distance between pixel groups in the particle image,

roughness is a measure of irregularity of a particle and elongation is similar to ellipse aspect ratio but

considers area and perimeter information [76]. Particles with higher compactness, lower roughness and

elongation will have less inter-particle friction and better flowing and packing behaviour [76, 100]. Vitchus et al.

[100] studied the paradox between bulk density and particle porosity using four types of coke and particle shape

factors. They found that natural particles have lower roughness and elongation compared to roll-crushed and

jaw-crushed grains. Although internal porosity of natural grains was 3-4% more than that of crushed grains, the

inter-particle porosity of the natural grains was 3-8% lower and their VBD was improved by 0.03-0.1 g/cm3 [100].

This is due to the fact that inter-particle free space greatly affects the overall density. Thus, coke particles with

lower intra-particle porosity will not necessarily result in the formation of denser anodes. It should be

emphasized that the Bowers [76] and Vitchus [100] have used a single range of particle size (-30+50 US mesh).

Wider size ranges, larger or smaller fractions which may influence the results have not been considered.

Beliskus [79] expressed the effect of particle shape on the bulk density. He confirmed that spherical particles

pack better than plate-like or needle-like particles and may result in higher VBD even with lower real density

and higher particle porosity. Sverdlin et al. [65] reported that irregular shape and rough surface of particles up to

12 mm prevent the movement in a viscose medium.

The influence of particle shape, roughness and surface area has been extensively studied for various

particulate materials. Particle shape changes with particle size and grinding technique used to make the

powder [79, 101]. Inter-particle friction increases with increasing the angularity, surface area and surface

roughness of particles [75, 101, 102]. When inter-particle friction increases, the resistance to flow and particle

bridging increase and packing properties worsen resulting in lower initial and final packing fraction of a powder

[75, 101, 103, 104]. Although irregular particles give a lower apparent density, they improve the green strength due to

mechanical interlocking [101, 104]. In other words, spherical shape particles will form weak structures. Thus,

irregular but round particles are often used to obtain a satisfactory flow and strength [101]. The ratio of tapped to

apparent density (Hausner ratio) is used as an indication of inter-particle friction. The difference between

tapped and apparent density depends to a great extent on the shape and size of particles. Usually particles

with lower apparent density fraction exhibit a higher increase in density fraction after tapping [105]. Hausner ratio

for spherical particles is slightly above unity. For more irregular particles, vibration or tapping results in a rapid

improvement in packing that is independent of apparent density, and leads to a higher Hausner ratio [101].

31

2.2 Anode manufacturing process

Anode manufacturing process consists of several steps, most importantly, materials handling and selection,

dosing and blending, mixing, forming and baking (Figure 2.14). After crushing and size classification of coke

and butt they will be preheated, weighed and blended with pitch to obtain the desired recipe. The materials are

then mixed to obtain the anode paste. Paste has two phases, a mixture of fine particles of coke and pitch

(binder matrix), and large aggregates. The paste is then densified by vibrocompacting or pressing. Finally, the

green anode is baked (carbonized) at 1100-1200 C to obtain the desired properties such as strength, density

and electrical resistivity. In this thesis only mixing parameters and the influence of paste formulation on the

compaction behaviour and green density were highlighted.

2.2.1 Blending

Anode-grade cokes are getting scarcer in quantity and more inferior in quality. On the other hand, the

specifications of the anode grade cokes are getting more stringent and tighter. Anode plants, therefore, have

to use more than one source of coke. Changing periodically from one source to another results in variations in

process parameters and baked anode properties and also unstable behaviour in the pot. Blending of two or

more cokes is of great interest as a practical solution to achieve the desired anode properties with minimum

impacts of coke variations.

Figure 2.14: A schematic of anode manufacturing process

Coke blending can be performed before the calcination process to achieve a more homogenous product.

However, the problem is to adjust the calcination parameters for a mixture of different cokes. Volatile

32

materials, density, and granulometry are the coke properties determining the calcination parameters [106].

Calcination behaviour of the blended and unblended cokes showed that blends of similar green cokes can be

calcined under similar conditions to the individual cokes and the quality of the resulting material match well

with the expected quality. Calcination of the blends of dissimilar cokes under similar conditions resulted in a

quality widely different from the expected properties [106].

In an anode plant, the anodes were made with a blend of two different sources of coke and anode properties

were studied [107]. The main source was a high bulk density and high sulphur coke with low air and CO2

reactivity while the second source was a low bulk density and low sulphur coke. 70% of high density coke and

30% of low density coke were mixed before coke classification. Plant tests showed that the BAD and GAD of

the test anodes were the same as those of anodes made with a single high density coke. The air permeability

of the test anodes were 0.2 nPm higher than that of the single source anode due to the higher porosity of the

second coke. However, it has not been reflected in BAD. It is expected that anode reactivity declines with the

amount of sulphur in coke [108-111]. Although the sulphur content of the second coke was lower, the air and CO2

reactivity residues were the same as the single source anode. The same reactivity residue can be due to the

fact that the cokes have been blended before classification and both low and high sulphur cokes exist in the

binder matrix which is more sensitive to CO2 reactivity. When cokes with different grindability are blended

before classification, there is a chance that a single coke may accumulate in a special range of particle size.

Besides, the distribution of metallic elements with an influence on the reactivity may also be non-uniform [74].

Another aluminium smelting plant also investigated the mixing of low and high sulfur cokes before

classification [112]. Plant results showed that anodes with the blended coke have a dusting behaviour and a CO2

reactivity residue very close to those of anodes with high sulfur coke. In addition, they could increase the

consistency of anode quality and the stability of plant operation with blending strategy instead of periodic using

of two sources of coke.

A new strategy is used for coke blending which is based on blending after classification while coke ratios are

chosen for each granulometry (Figure 2.15). This way it is possible to use the strong aspects of cokes for the

quality dominant portion to improve the blended quality. Tonti et al. [113] substituted the dust fraction with low

reactivity cokes in different percentages. The air and CO2 reactivity residues increased by 13% and 11%,

respectively, when the dust fraction was completely substituted with low reactivity coke. They claimed that

using low reactivity cokes exclusively for dust fraction was more effective than using it for the whole

aggregates with the same percentage. When the low reactivity cokes were added to the aggregates, the baked

density reduced due to higher macroporosities of these cokes.

33

Figure 2.15: Modified coke blending strategies [77, 114]

This new strategy was used in a plant to improve the anode density and to reduce the selective burning of

binder matrix [77]. They claimed that using high-sulphur low-reactivity cokes exclusively for fines fraction

reduced the excess carbon consumption, while SO2 emission was kept under the environmental limits.

Similarly, Low-porosity high-VBD cokes were used for coarse fraction exclusively to improve the density and to

reduce the optimum pitch level and cost. It was shown that low-VBD coke can be used for fines fraction

without affecting the anode quality, because the fines fraction porosity has negligible effect on the anode

density. They used two high-sulphur cokes with high and low VBD and compared three types of industrial

anodes: 100% high-VBD coke, 68% high-VBD coke and 32% low-VBD coke with conventional blending, 68%

high-VBD coke and 32% low-VBD coke with blending after classification. The results showed that the

reactivity, BAD, permeability and electrical resistivity of the anodes prepared with the new blending strategy

were superior to those of anodes made with the conventional blending method and were closer to those of

anodes made with 100% high-VBD coke. There was no significant difference for flexural strength. The same

strategy was used in a plant in Canada [114]. They used a mix of low (1.8%) and high (3.3%) sulphur cokes for

coarse and intermediate fractions and a high-sulphur coke for fines fraction. Using this method, they could

eliminate the high reactivity cokes from binder matrix and meet the environmental regulations for SO2

emissions. Plant results showed a sharp reduction in the percentage of dusty pots and more stable dusting

conditions. Besides, there were less variations in the sizing that resulted in a decrease of 0.3% in the

pitch/coke ratio because a single coke with a stable hardness was fed into the ball mill.

The blending strategy can be used for different sources of pitch. As discussed in 2.1.2, coal tar pitches are

preferred to petroleum pitches due to higher QI content, better binding properties and coke textures, while

petroleum pitches have lower PAH emissions. Although coal-tar and petroleum pitch components do not have

chemical interactions during blending, interactions exist in the carbonization stage [115]. The presence of CTP in

COKE

A

COKE B

FINE

COKE A INTERM.

COKE A

COARSE

COKE A

FINE

COKE B INTERM.

COKE B

COARSE

COKE B

PROCESS

Z

Y X

Ball Mill

A B

CRS

CRS

INT

INT F

INES

34

the pitch blend improves the wetting ability of pitch, shortens the temperature range of PP weight loss and

increases the microstrength of the pitch coke [115]. The CTP components surround the mesophase and restrict

its growth and coalescence [115]. Söderberg anodes with a mix of CTP and 30-45% PP had the same strength

and resistivity as coal-tar pitch anodes, while bulk density and air reactivity were slightly higher [58].

2.2.2 Mixing

Mixing of blended raw materials is the last step to prepare the anode paste. Sufficient mixing power and time

as well as proper mixing temperature are the most important parameters determining the mixing effectiveness

and anode quality. An efficient mixing results in a homogeneous distribution of coke and pitch [80, 116-118], and

lower porosity in the paste that improves anode characteristics such as density [119] and thermal shock

resistance (TSR) [116].

Electrical resistivity, green and baked apparent density of pressed and vibrated anodes were used as

indicators of the mixing efficiency [119]. Extending mixing time in a sigma blade mixer from 5 to 60 minutes led

to an increase of 0.02 g/cm3 in the GAD and BAD of pressed anodes. Two thirds of this increase occurred

within the first 30 minutes. Electrical resistivity also decreased by 5 m within 30 minutes and then

increased slightly. The amount of penetrated binder in coke pores increases with extending mixing resulting in

enhanced density and electrical conductivity. However, the amount of bridging coke decreases with mixing

time leading to a higher density but poorer electrical contacts between aggregates [119]. In addition, overmixing

results in grain crushing and leads to a different granulometry and newly developed surfaces that require more

pitch for impregnation [89]. Shorter optimum mixing time was reported for vibrated anodes [119]. Green and

baked apparent density increased by 0.015 and 0.02 g/cm3, respectively, and resistivity decreased by 4 m

after 15 minutes of mixing. Overmixing resulted in degradation of the vibrocompacted bench scale anodes. It

may be due to the fact that the amount of coke fine increases with extended mixing time and results in a higher

viscosity of binder which avoids deep penetration of binder into coke pores during vibration. Long mixing time

at high temperature may also have negative effects on the anode quality.

Individual and interaction effects of mixing time, mixer type, heating rate, pitch and fine content on the porosity

were investigated [89]. For samples with 14% pitch, extending mixing time from 5 to 10 minutes resulted in a

larger reduction in porosity, comparing to samples containing 16% pitch. Longer mixing time improves pitch

penetration and uniform coating of coke particles and consequently reduces the formation of voids. Samples

with 16% pitch have more pitch available for bonding and filling the voids. Shorter mixing time is therefore

sufficient for good binder distribution [89]. The continuous pitch film that forms over the coke particles in

extended mixing times results in a better coke bridging after baking [89]. Thickness and continuity of the pitch

film as well as its penetration into pores, determine the cold compression strength of anodes [89]. Stokka [80]

obtained similar results for extended mixing times. When he increased the mixing time in a sigma blade mixer

35

up to 30 min, the total volume of pores of 1-100 m decreased by 30 mm3/g. He also revealed that porosity

was reduced with increasing pitch content, because the void-filling effect of pitch is more than its penetration

into coke pores [80].

Belitskus [119] investigated the effects of mixing and moulding temperatures on anode properties. For pressed

anodes, a super heat of 30 C above the softening point of pitch improved the anode density compared to 10

C of super heat. For vibrated anodes the optimum mixing and moulding temperatures were 40 C above the

pitch softening point. Although the preheating of coke and pitch did not significantly affect the bench scale

anode properties, it was noted that it might reduce the mixing time in the anode plant where the heat transfer

was slower. Stokka [80] claimed that increasing the mixing temperature from 150 to 180 C could reduce the

paste porosity from 40 to 17 mm3/g after 5 minutes of mixing in a sigma blade mixer. In other words, it can

reduce the mixing time to achieve a given porosity volume. However, he stated that long mixing time (more

than 30 min) at high mixing temperature (180 C) might result in a higher level of porosity. Because low

viscosity of pitch at high temperatures enhances the pitch penetration into the coke pores and the remained

pitch is not enough to fill the intra-particle voids [80]. Similar influence of mixing temperature was reported on

the properties of binder matrix electrodes. Mixing temperature below 180 C resulted in inferior properties of

binder matrix samples [88].

Mixing has been combined with a kneading effect in recent decades which applies shear force. The coke

pores are therefore filled with binder matrix and a low porosity paste is produced. The higher the energy of

mixing, the smaller are the particles that penetrate into the coke cavities and, the more uniform is the

dispersion of pitch [116]. Uniform distribution of pitch may reduce its consumption and the volatilisation pressure

during baking, resulting in lower cracking rate [116]. Perrushoud et al. [32] stated that the worldwide range of

mixing power, 5-12 kWh/t, had an impact of 0.3 W/mK on the anode thermal conductivity. Thus, specific mixing

energy is directly related to the TSR of anode [116]. Stokka [80] used an intensive mixer after mixing with one

kneader and two kneaders in series. He measured the total volume of pores of 1-100 m as an indicator of

mixing efficiency and found that intensive mixing reduced the porosity by about 50% and increased the

homogeneity of pore size distribution in the paste. He concluded that mixing efficiency of kneader was not

satisfactory. Tkac [89] also reported that intensive mixing results in a homogeneous mixing and continuous pitch

film over particles in shorter mixing times compared to sigma blade mixer. Intensively mixed samples had

lower total porosity and smaller pores. When the paste was intensively mixed for 10 minutes, the paste was

homogeneous enough and vibration time had little effects on the properties. Longer mixing time and higher

pitch content is required with a sigma blade mixer to achieve a low porosity paste.

The paste produced with an intensive mixer has lower gas content and allows using higher temperatures in

forming step [118]. A 15 C increase in the atmospheric vibrocompacting temperature was possible by using

36

intensive mixing with a specific mixing energy of above 8 kWh/t [118, 120]. Modern plants use a continuous remixer-

cooler after continuous intensive hot-mixer to be able to adjust the mixing and forming temperatures

independently. In the cooler, mixing continues and temperature decreases to the desired temperature with

water injection. A lower reduction in the paste temperature results in a lower PAH emission [118]. The minimum

temperature at the mixer outlet depends on the softening point of pitch [121]. The outlet temperature of the

cooler should be optimum for anode forming and depends on the plasticity of the paste. Paste viscosity is a

function of size distribution of aggregates (coarse/fine ratio), pitch viscosity, pitch/coke ratio, coke porosity,

preheating and mixing temperatures [116, 121]. Paste temperature at the outlet of cooler should be well above the

softening point of pitch for good densification and well above 100 C for surely removing the moisture [121]. The

paste temperature for pressing can be lower than the required temperature for vibrocompaction [121].

2.3 Summary

The literature about anode making is principally focused on the influence of raw materials characteristics and

paste formulation on the quality of green and baked anodes. Many authors have investigated the impact of

production routes for calcined petroleum coke and coal tar pitch, particle size distribution and pitch/coke ratio.

A few publications have considered the effect of variations in raw materials on the final properties of anode but

there is not enough information on the evaluation of variations and making corrective decisions in an

intermediate stage. Most of the authors have studied the impact of anode making procedure and the related

parameters on the final quality and performance in the smelting cells. However, the behaviour of the material

during compaction, and the influence of materials variations and paste composition on the compaction

behaviour have not been taken into account. The influence of mixing parameters on the quality and

homogeneity of the paste has not been effectively considered in these works. This project had a more

comprehensive look on the effects of mixing parameters on the distribution of materials and porosity in the

green samples. In addition, compaction behaviour of paste, and the influence of paste formulation and coke

particle characteristics on this behaviour were investigated.

3 Chapter 3: Materials and methods

3.1 Introduction

Calcined petroleum coke and coal tar pitch were used as starting materials. The cokes were crushed and

classified into varying size fractions. A setup was made for mixing the materials and compacting the resulted

anode paste. Uniaxial compaction test was performed in a rigid mould. Sensitivity of the test to materials

variations and process parameters was validated and the test was used for the objectives of the project. The

influence of mixing variables on the properties, microstructure and compaction behaviour of the paste was

studied and optimum mixing factors were determined. Several pastes were made at the optimum mixing

conditions and the effect of formulation and compaction rate was evaluated. Shape factor and density of coke

particles were considered as variables and their influence on the compaction behaviour was investigated.

General information about the starting materials, experimental procedures and characterization methods is

explained in this chapter and followed for all objectives otherwise stated. Further details for each objective are

provided at the materials and methods section of the following chapters.

3.2 Description of raw materials

3.2.1 Calcined petroleum coke

Six commercially available calcined petroleum cokes including 5 sponge cokes (A-E) and a shot coke were

used for making the pastes and anode samples. Table 3.1 shows the real density (after excluding the pores)

and chemical composition of the calcined cokes. The cokes were crushed by jaw and roll crushers and sieved

into six size fractions using a Sweco vibro-energy separator. The size ranges were -4+8, -8+16, -16+30, -

30+50, -50+100 and -100+200 US mesh. Particles in the range of 8+16 US mesh (-2.38+1.41 mm) were milled

in a ball mill to make fine coke fractions. Total weight of steel balls in the ball mill was 26.8 kg with the

following size distribution:

+25 mm 35%

-25+19 mm 21%

-19+13.3 mm 10%

-13.3+9.5 mm 34%

The milling time and batch weight were controlled to obtain the desired fineness. Fineness was assessed by

Blaine number with the unity of cm2/g that is an indication of specific surface area (SSA) of the powder. It

should be pointed out that BN may differ from the exact value of specific surface area and it is usually

employed as a comparative variable. Size distribution and Blaine number (BN) of the fine fractions were

measured by sieve analysis and laser diffraction particle size analyzer (Malvern Mastersizer 2000),

38

respectively. Specific surface area of the fine cokes was measured by BET method (Micromeritics, TriStar II).

Table 3.2 shows the milling conditions, BN and SSA for coke D that was the principal coke used in this thesis.

Table 3.1: Real density and chemical composition of cokes used for anode fabrication

Coke Real density

(g/cm3) Na [ppm] Si [ppm] %S Ca [ppm] V [ppm] Fe [ppm] Ni [ppm]

A 2.075 <50 10014 1.10.03 1307 1206 704 905

B 2.063 <50 20028 3.20.1 <10 26013 603 1508

C 2.074 <50 507 3.160.09 201 28014 905 1307

D 2.057 1007 12017 2.130.06 1307 36018 46023 25013

E 2.073 <50 21029 1.730.05 24012 905 34017 23012

Shot coke

2.004 14010 7511 3.850.12 372 141471 583 34517

Table 3.2: Ball milling parameters for coke D, size distribution and specific surface area of fine cokes

Ball milling parameters Blaine

number (cm2/g)

Particle size distribution (wt.%)

BET surface area (m2/g)

Batch weight

(kg)

Initial particle

size (mm)

Milling

time (min)

+149

m

-149+74

m

-74+53

m

-53+37

m

-37

m

4 -2.38+1.41 60 2300 22.8 35.5 10.7 10 21 2.9

2 -2.38+1.41 34.5 4000 1.6 26.4 20.2 16 35.8 4.1

2 -2.38+1.41 49 6300 0 10.7 20.6 26.2 42.5 6.1

3.2.2 Coal tar pitch

A single batch of coal tar pitch was used for all the experiments in this thesis. Softening point and quinoline

insoluble content of the pitch were 109 C and 15.5%, respectively. Viscosity was measured by a Brookfield

R/S plus rheometer at different temperatures and is indicated in Table 3.3.

Table 3.3: Properties of coal tar pitch used as binder in anode samples

Mettler softening

point (C)

Quinoline insoluble (%)

Density (g/cm3) Viscosity (Pa.s)

158 C 168 C 178 C 188 C

109 15.5 1.31 4.08 3.15 1.12 0.66

39

3.3 General experimental procedures

3.3.1 Scott density

Bulk density of coke fractions was measured by a Scott volumeter with a 25 cm3 cylindrical cup (Figure 3.1).

The procedure in ASTM B329 [122] standard test method was followed. The empty cup was weighed and a

minimum of 35 cm3 of powder was poured into the top funnel. The powder was allowed to run into the cup until

it was overflowed and a pile of powder was formed on top of the cup. The excess powder was removed

carefully by passing the edge of a spatula parallel and in contact with the top of the cup. The cup was then

weighed and the mass of the powder was divided to the volume of the cup to obtain the bulk density. Each

measurement was repeated three times and the average was calculated.

Figure 3.1: Scott volumeter for bulk density of coke fractions

3.3.2 Vibrated bulk density (VBD)

Vibrated bulk density of coke particles was measured by a vibrating table using ASTM D4292 standard test

method [123]. One hundred grams (0.01) of particles in a given size range were poured in a vibrating funnel

and fed into a graduated cylinder during 70 to 100 s while the cylinder was vibrated at 60 Hz with an amplitude

of 0.2 mm. The volume of vibrated particles was read after 120 s of vibration and VBD was calculated. The

average of three measurements was reported. Figure 3.2 shows the standard setup for this test.

40

Figure 3.2: Setup for vibrated bulk density

3.3.3 Particle apparent density

Apparent density is defined as the ratio of the mass of a particle over its external envelope volume, which

includes the internal pores. A given mass of coke particles was poured in a 1 inch diameter mounting mould

and the pile of the particles was flattened. Table 3.4 indicates the mass of sample for different size fractions.

The particles were impregnated with Epofix polishing resin under vacuum. The mould was filled enough with

resin to have a few millimeters of resin above the particles. The upper surface of the mounted samples was cut

to obtain a flat surface. The height and diameter of the impregnated samples were measured to the nearest

0.01 mm and the sample volume was calculated. The sample was then sectioned vertically as indicated in

Figure 3.3 and the two marked surfaces were polished. Area fraction of the particles on the polished surfaces

was determined by image analysis and averaged. The average area fraction was used as an estimation of

volume fraction and multiplied by the sample volume to obtain the apparent volume of coke particles only. The

mass of particles was divided to particles volume to obtain the apparent density of particles.

Table 3.4: Mass of coke particles impregnated with resin for apparent density measurement

Size fraction Mass (g)

-4+8 US mesh 6.00

-8+16 US mesh 5.50

-16+30 US mesh 4.50

-30+50 US mesh 4.00

41

Figure 3.3: Schematic illustration of impregnated coke particles and analyzed surfaces

3.3.4 Particle shape factors

Coke particles of a single size fraction were spread on a sheet. A Nikon Ephiphot optical microscope equipped

with an image analysis system (Clemex, vision) was used to measure five particles shape factors comprising

aspect ratio, sphericity, roundness, compactness and convexity. Definition of the shape factors are given in

Table 3.5. More spherical particles have lower aspect ratio and higher values for sphericity and roundness.

Higher values of convexity and compactness show lower irregularity (angularity) of a particle. These factors

describe various aspects of a particle which may be related to flow behaviour and packing of the particles.

Table 3.5: Particle shape aspects measured by image analyzer

Shape aspect Definition

Aspect ratio

Ratio of longest dimension to shortest dimension.

Sphericity 4..Area/(Perimeter)2

Roundness

4.Area particle /.(diameter circle)2 Quantifies the roundness of object’s edges.

Convexity Convex perimeter/perimeter Quantifies the jaggedness of object’s edges. Convex perimeter Convexity

Compactness Ratio of area over convex perimeter 4..Area/(Convex perimeter)2

3.3.5 Mixing

A Hobart N50 mixer was installed in an oven to ensure a uniform temperature (Figure 3.4). The classified size

fractions of coke with the proportions presented in Table 3.6 were individually weighed and added into the

mixer bowl to avoid variations in the granulometry of samples. The cokes were preheated at desired mixing

temperature for at least 2 hours. Then solid pitch was added on the coke and preheated for 30 minutes at the

same temperature. The materials were mixed for the desired mixing time and the produced paste was

discharged in a preheated funnel equipped with a slide gate to fill the mould.

Weighing and mounting

with polishing resin

Polishing for

image analysis Cutting planes Flattening and

measurement

42

Figure 3.4: Mixer used to make paste samples

Table 3.6: Size distribution of coke particles in the paste samples

Size range

(US mesh) -4+8 -8+16 -16+30 -30+50 -50+100 -100+200 Fine coke

Wt.% 22.0 10.0 11.5 12.7 8.8 10.8 24.2

3.3.6 Compaction test

Binder matrix: The binder matrix compositions were compacted in a cylindrical steel mould with an inner

diameter of 63 mm (Figure 3.5). Before compaction, a preload of 14.2 kPa was applied by putting the punch on

the paste. The height of the samples (x) was measured (Figure 3.6) and the bulk density of paste was

calculated and used as the initial point of the curve. Compaction tests were conducted at desired constant

displacement rates (DR) and a maximum force of 220 kN. Uniaxial pressure was progressively increased to 70

MPa. The tests were performed at 150 C inside a three zone split-tube furnace mounted on a MTS

Servohydraulic press (Figure 3.5). Force-displacement data were obtained from the machine using a 250 kN

MTS load cell and a 150 mm position transducer (LVDT) of the press. Mass, diameter and height of the

samples were measured and used to plot the compaction curves. Having the real time height of the sample

inside the mould, geometrical density was calculated at each point of the force-displacement curve. The

geometrical density was divided by the real (theoretical) density of the material to obtain the relative density.

Evolution of the relative density was then plotted as a function of applied pressure. Real density of coke and

pitch were used to calculate the real density of pitch/coke compositions. Each sample was repeated twice to

ensure the repeatability of the test. The tests were repeatable, thus the results of one test for each sample

were presented in this thesis.

43

Figure 3.5: Rigid die and MTS Servohydraulic press for compaction of paste

Figure 3.6: Schematic illustration for the measurement of the sample height before compaction (x). After putting the punch on the sample, the punch height entered into the die was marked (y). x obtained by z-y where z is the total mould height.

Anode paste: The paste samples were compacted with a similar procedure to that used for binder matrix

samples. The pastes were compacted in a cylindrical rigid mould with an inner diameter of 68.3 mm at 150 C

to a maximum pressure of 60 MPa at different displacement rates. Two replicates for each sample were made

to verify the test repeatability.

3.3.7 Green and baked apparent density

Height and diameter of entire green compacts were measured at three points with 90 of difference in angle

and averaged. Mass of the sample was divided by the calculated volume to obtain the green apparent density

(GAD). Green compacts were not strong enough and after cutting a green sample into pieces, the edges were

x

y

z

44

not sharp. Thus the calculated volume was not precise. Therefore, to determine the apparent density of a part

of the green sample, the samples were covered with a known volume of paraffin to avoid water penetration

into the open pores. Water displacement method with a precision of 0.1 ml was implemented to obtain the

volume and green apparent density was determined (0.002 g/cm3). Baked samples were solid enough and

precise shape could be obtained after cutting. However, baked apparent density (BAD) was also determined

by water displacement method to keep consistency of the results.

3.3.8 Porosity

A Micromeritics helium pycnometer (AccuPyc II 1340) was used to measure the volume of dry coke particles

as well as green and baked compacted samples. Pycnometer consists of two chambers with known volumes.

The sample with a known mass is in one chamber and the pressure in the other chamber is increased to a

given pressure while the valve between the chambers is closed. Then the valve is opened and gas enters the

sample chamber. Equilibrium pressure is measured and the gas volume in the sample chamber is calculated

by Equation 3.1. Sample volume is obtained by subtracting the calculated volume from the total volume of the

chamber. Helium penetrates into the accessible pores and thus subtracts the volume of open pores. The

measured volume consists of the volume of the material and the volume of closed pores. Pycnometer density

was defined as the ratio of mass of the material to the volume measured by the pycnometer. The fraction of

total, open and closed porosity could be determined using the apparent, real and pycnometer densities, and

Equations 3.2 to 3.4. Volume of closed pores within the coke particles was calculated by Equation 3.5.

(3.1)

(3.2)

(3.3)

(3.4)

(3.5)

3.3.9 Determination of the optimum pitch/coke ratio

Optimum amount of pitch is usually determined based on the maximum baked apparent density [51]. Coke D

with the size distribution of Table 3.6 and fine coke with a BN of 4000 was used and paste samples were made

with various pitch/coke ratios (P/C) in the range of 14/100 to 18/100. The pastes were compacted at 150 C to

a maximum pressure of 60 MPa. Mass, diameter and height of the green samples were measured and green

apparent density was calculated. The samples were cut from the middle in the height and the upper parts were

45

baked in an electric resistance furnace. The samples were packed in -8+14 mesh coke fraction in alumina

tubes to protect against air oxidation. During baking, pitch volatiles are released that may cause cracks in the

sample. The generation rate of the volatiles thus should be lower than the rate they escape from the sample.

Figure 3.7 shows the baking cycle used to avoid cracking. Soaking time was 12 hours at the maximum baking

temperature of 1130 C. Green apparent density increased with increasing P/C ratio while baked apparent

density showed a plateau when P/C exceeded 16/100 (Figure 3.8). Thus, the pitch to coke ratio of 16.2/100

was selected as the optimum pitch ratio for coke D and used for making anode samples using this coke.

Figure 3.7: Baking cycle for anode samples

Figure 3.8: Apparent density for green (a) and baked (b) samples made with different pitch contents

3.3.10 Specific surface area (BET)

Langmuir Theory explains the physical adsorption of gas on a solid surface in monolayer. Brunauer-Emmett-

Teller (BET) is an extension of the Langmuir theory for multilayer molecular adsorption. This theory is

described elsewhere [124] and is the basis for the measurement of the specific surface area of the materials.

The amount of gas adsorbed at a given pressure allows determining the surface area.

46

Coke powder and baked samples were degassed at 250 C under nitrogen flow while for green samples it was

conducted at 50 C. The samples were submerged in liquid nitrogen and specific surface area was measured

by a BET surface analyzer (Micromeritics TriStar II). Coke powder, paste and anode present a low specific

surface area comparing to porous powders. Therefore, low relative pressures of nitrogen that correspond to

low surface area materials were used.

3.3.11 Mercury porosimetry

Equation 3.6 presented by Washburn is the principal equation used in mercury porosimetry. Washburn

equation can be obtained by equilibrium between the force or work required to force mercury into and out of

the capillary pore. Contact angle of mercury and carbon is greater than 90 and mercury does not wet carbon.

Thus, P is positive and a pressure greater than ambient must be applied to the liquid mercury to force the

liquid into the capillary pores. This method requires the evacuation of the sample and subsequent

pressurization to force mercury into the pores. The pressure difference across the mercury interface is the

absolute applied pressure and the equation reduces to Equation 3.7 [125]:

(3.6)

(3.7)

: Surface tension

: Angle of contact

P : Excess pressure above the ambient

r : Capillary radius

Assuming constant values of and at a constant temperature, the product is constant. It means that as

the pressure increases, mercury will intrude progressively into narrower pores. The wide range of pore size

measurable by mercury porosimetry needs a wide range of pressures.

Low pressure measurement consists of cell evacuation, filling with mercury and intrusion. Sample was

weighed to the nearest 0.001 g and put in a 15 cm3 cell (penetrometer). The cover was sealed by grease and

tightened. The penetrometer containing the sample was installed horizontally in the mercury porosimeter

(Micromeritics, AutoPore IV). Air was evacuated from sample cell until a pressure of 0.016 atm was obtained.

The open end of cell stem was submerged under mercury and nitrogen gas above the mercury forced the

liquid into the stem to fill the voids around the sample. Pressure was gradually increased to 1.6 atm to intrude

the mercury into the large pores and voids between the particles. The change in the length of mercury column

in the cell stem was transformed to the intruded volume (pore volume). Then the penetrometer was transferred

to the high pressure chamber.

47

High pressure measurement: Pressure was progressively increased to 4082 atm and the intruded volume was

recorded. At this point, depressurization started and pressure was gradually decreased to 0.9 atm to extrude

the mercury.

Entrance radius of pores is proportional to the applied pressure and the volume of intruded mercury is

recorded. Mercury porosimetry thus provides size distribution of pores and the pore volume for a given size.

4 Chapter 4: Validation of compaction test to

evaluate paste quality

4.1 Résumé

Le comportement au compactage de la pâte d’anode a été utilisé comme indice intermédiaire de suivi de la

qualité afin de réduire le nombre de variables à contrôler dans les étapes précédant de la production d’anode.

Deux sources de coke ont été utilisées. Différentes fractions granulométriques des deux cokes ont été

caractérisées à l'aide d'un système d'analyse d'image pour déterminer leurs facteurs de forme. Ces facteurs

ont été corrélés à la densité apparente vibrée du lit de poudre. Une procédure d'essai de compactage a été

mise en place. Le comportement au compactage des pâtes d’anode a été étudié en fonction des

caractéristiques des particules de coke et de la formulation de la pâte d’anode afin de valider le test de

compaction. Ce test a été appliqué à quatre pâtes d’anode réalisées à partir des deux sources de coke, de

distributions de taille des particules et de teneur en brai (liante) différents. Il a été observé que les essais de

compactage sont reliés aux changements des caractéristiques des matières premières et à leur proportion

dans la formulation d'anode. En conséquence, l'essai de compactage a été utilisé comme un outil pour évaluer

la qualité des pâtes d'anode variant en fonction des matières premières.

4.2 Abstract

Compaction behaviour of paste was used as an intermediate quality index to reduce the number of variables to

be controlled in the subsequent previous production steps. Two different sources of coke were used. Different

size fractions of the cokes were characterized using an image analysis system to determine the shape factors

where the characteristics of each coke were correlated to its vibrated bulk density. A compaction test method

was designed and the influence of coke particle characteristics and paste formulation on the compaction

behaviour of anode pastes was studied to validate the test. The test was applied on four anode pastes made

with different sources of coke, particle size distributions and pitch contents. It was observed that the

compaction test was significantly sensitive to changes in raw materials characteristics and their proportion in

the anode formulation. Consequently, the compaction test was used as a tool for evaluating anode paste

quality in relation with material variations.

50

4.3 Introduction

Intermediate quality indices are required to control each step of the process and to take the corrective actions

within the subsequent steps in order to keep the anode quality consistent. Characterizing the green and baked

anodes is the general approach in the anode manufacturing industry and there is no control on the paste

quality before forming. However, green and baked properties are influenced by a number of variables including

raw materials properties and mixing and compaction parameters. By defining a specific paste quality as an

intermediate process quality index it will be possible to limit the large number of variables and to evaluate the

properties in an intermediate step before compaction and baking. The paste quality index should be

representative of and sensitive to the variations of materials and process parameters existing before

compaction.

Compaction behaviour of paste was defined as the quality index and was studied via a compactability test

method. A sample of paste is taken after mixing and is compacted at desired pressure and temperature.

Variations in force and displacement are acquired and relative density v.s. pressure is plotted as compaction

behaviour of the paste. A schematic of compactability test is shown in Figure 4.1. The compaction curve is a

characteristic (fingerprint) of paste and can be used as a feedback to adjust the formulation and mixing

variables to compensate for the variations in materials, or to adjust the compaction process parameters,

regardless of the factors in previous steps.

Figure 4.1: Schematic representation of paste compactability test

Sensitivity of compactability test to variations in coke characteristics and paste formulation was evaluated to

validate the test. This test method allows studying the influence of materials properties (coke density, shape

factor and granulometry), paste formulation and mixing factors on the compaction behaviour of paste and the

green density. The results of this chapter were presented in TMS 2011 annual meeting and published in Light

Metals 2011, P. 1161-1164.

51

4.4 Experimental procedures

Three formulations from cokes A and B were made with different particle size distributions to reveal the effect

of granulometry on compaction behaviour. Table 4.1 shows the granulometry of the samples. Particle size

distributions for coke A consisted of a reference distribution, which was used in the industry, and two other

distributions; one with 5% more large aggregates and the other with 5% more fine particles than reference. In

addition, one formulation from coke B with reference granulometry was made to reveal the effect of particle

properties from different coke sources. The coke sources and pitch to coke ratios are listed in Table 4.2. Pitch

contents were chosen based on the fine fraction in the formulation. Large amount of fine particles increases

the specific surface area of coke particles and consequently a higher content of pitch is required for wetting

these fine particles.

Table 4.1: Particle size distribution of pastes

Size range (US mesh)

Reference size

5% More fine

5% More large

+6 13.6% 12.8% 14.4%

-6+16 15.6% 14.8% 16.4%

-16+30 16.3% 15.4% 16.9%

-30+50 10.4% 9.8% 11%

-50+100 8.1% 7.5% 8.6%

-100+200 10.5% 9.7% 10.7%

-200 25.5% 30% 22%

Table 4.2: Coke source and pitch content of pastes

Reference granulometry 5% More fine 5% More large

Coke A and B A A

Pitch/coke 13.8/100 15.8/100 13/100

Shape factors and vibrated bulk density of coke particles were measured to study the effects of coke particle

characteristics on their flow and packing behaviour. Vibrated bulk density was measured for four fractions of

cokes A and B as well as the blend of all size fractions. Single coke fractions were impregnated by a polishing

resin under mechanical vacuum and polished for microscopic analysis. Particle shape features including

aspect ratio, compactness, convexity, sphericity, roundness and porosity, as defined in chapter 3, were

measured. A more circular particle, for example, has lower aspect ratio and higher values for compactness

and roundness. Higher values of convexity show smoother surface of the particle.

A rigid cylindrical mould with the internal diameter of 89 mm was used to study the compaction behaviour of

anode pastes. Coke fractions and pitch were preheated at 185 C for 120 and 30 minutes, respectively, and

then they were mixed at 185 C for 10 minutes. Compaction test was carried out at 130 C, using an MTS

52

Servohydraulic press working at a constant displacement rate of 10 mm/min and a maximum force of 220 kN

(35 MPa). The displacement-force data was acquired at a rate of 10 readings per second. The apparent

density of pressed samples was calculated for each displacement value as a function of applied force, as

explained in chapter 3. Relative density vs. applied pressure was presented as compaction curves.

4.5 Results and discussion

4.5.1 Vibrated bulk density

Four shape factors were characterized for single size fractions of cokes A and B to study the relationships

which may exist between particle characteristics and their flowability. Table 4.3 shows the image analysis

results and VBD for single coke fractions. It was observed that, for the same size range, the VBD of particles

increased with lower values of aspect ratio; with lower level of porosity; and with higher values of

compactness, convexity, sphericity and roundness. More spherical particles are expected to display better

packing behaviour and higher bulk density values than plate-like or needle-like particles [75, 76, 79, 100]. Although

coke B has lower real density than coke A, it provides higher VBD for all fractions except for -200 mesh

particles. Coke B particles revealed lower aspect ratio and higher compactness and roundness values than did

coke A. Since these factors affect particle flowability, coke B has better packing behaviour which most likely

compensates the effect of lower real density of coke and results in higher VBD values. For -200 mesh

particles, on the other hand, coke A with better flowability factors and lower porosity resulted in higher VBD.

Table 4.3: Shape factors and VBD of coke particles

Particle characteristics

Size range (US mesh) and source of coke

-6+16 -16+30 -30+50 -200

A B A B A B A B

Aspect ratio 1.86 1.69 1.87 1.74 1.99 1.75 2.02 2

Compactness 0.607 0.657 0.719 0.761 0.707 0.766 0.663 0.638

Convexity 0.802 0.766 0.915 0.957 0.95 0.963 0.93 0.885

Sphericity 0.406 0.4 0.678 0.794 0.707 0.804 0.583 0.507

Roundness 0.41 0.474 0.527 0.569 0.5 0.567 0.437 0.427

%Porosity 28.73 25.98 23.41 19.51 21.34 20.74 3.3 4.64

VBD (g/cm3) 0.841 0.953 0.927 1.02 0.919 1 0.969 0.945

The vibrated bulk density of blended coke fractions used to make the samples is shown in Figure 4.2. Bulk

density of coke A increased with reducing the large fractions (>300 m) and increasing the fine portion (<74

m) in the formulation. These results were expected since the smaller particles fill the inter-particle voids

without increasing the volume [40, 79]. It has also been reported by other authors [82] that higher amounts of fine

particles, results in higher VBD of dust fraction and multi-fraction mixtures, up to a maximum value. In addition,

53

the bulk density of particles increases with decreasing the particle size, because large pores are destroyed

with size reduction [79].

Figure 4.2 : Vibrated bulk density of blended coke fractions

With regard to blends of cokes A and B with reference particle size, it was observed that they exhibit different

bulk densities, as shown in Figure 4.2. This difference is an indication of the influence of physical properties of

coke particles on the flowability and rearrangement during vibration. More spherical and less irregular particles

of coke B reduced the inter-particle friction and bridging of the particles.

4.5.2 Compaction test and green density

Compaction tests were carried out for four formulations to study the influence of size distribution and texture of

particles on the compaction behaviour of paste. The formulations were prepared using two sources of coke

with different particle size distributions and three pitch/coke ratios, as shown in Tables 4.1 and 4.2

respectively. The test was repeated three times for the same formulation to ensure the repeatability. Figure 4.3

shows that the tests were repeatable, thus the results of one test for each sample are shown in Figure 4.4. It is

observed that there is a meaningful difference among these curves and the test is capable of illustrating the

effects of materials variations on the paste compaction behaviour. When using coke A, the final apparent

density decreased by increasing the content of large fractions; while the formulation with higher percentage of

fine particles revealed higher apparent density compared to the reference granulometry. These observations

agree with other studies where it was reported that green apparent density increases when using higher

amounts of fine fraction and using a suitable amount of pitch [40]. Small and fine particles display better filling

ability to the inter-particle voids which will progress to improved density. In addition, by increasing the fine

content, more binder matrix is available to fill the coke pores. However, fine content beyond a specific upper

limit may not have any significant effect on the green density [82] or even may reduce it due to more inter-

particle friction and particle bridging [75]. For the three pastes made with coke A, the trend of VBD of the coke

54

blends was the same as that of green apparent density. Both VBD and green apparent density were increased

with increasing the fine fraction.

Figure 4.3: Examples of repeatability of compaction curves

Figure 4.4: Compaction curves with variations in paste formulations and source of coke

The pastes with the same particle size and pitch content, but different sources of coke, showed quite

separated curves below 20 MPa. This is related to particle characteristics as observed in the case of VBD

results. The interesting result was that coke A with more elongated and rough particles and lower VBD

provided samples with a green apparent density close to that of samples from coke B. This is in accordance

with other studies where it was reported that although aggregate bulk density is an indication of porosity and

packing in a powder bed, there is no consistent correlation between aggregate bulk density and green

apparent density [40]. The results showed that VBD could not be used alone as an indication of compaction

behaviour. In other words, cokes with higher vibrated bulk density will not necessarily result in denser anodes.

55

The influence of real density, porosity and shape of particles on the VBD and GAD was more deeply

investigated and is explained in chapter 7.

4.6 Summary

Paste characterization may be used as an indicator to estimate anode quality. Paste compaction test was

sensitive to variations in raw materials properties. Compaction behaviour, as a paste property, can be used to

study the influence of materials properties, paste formulation and mixing factors on the compaction behaviour

of paste during the forming process. The data and curves obtained from this particular test provide a better

knowledge of the paste behaviour when it is formed by hydraulic pressing and may be used for modelling the

compaction behaviour. Shape factor and texture of the particles influence the bulk density of coke and may be

used to describe the compactability of a particle bed. Vibrated bulk density, however, is not the only factor

which controls the density after compaction.

5 Chapter 5: Influence of mixing variables on paste

characteristics, compaction behaviour and anode

properties

5.1 Résumé

Les effets des paramètres de mélange sur la qualité des pâtes d'anode, des anodes vertes et des anodes

cuites résultantes, ont été étudiés. Quatre durées et quatre températures de mélange ont été appliquées pour

fabriquer des pâtes d'anode à l'échelle laboratoire. La surface spécifique et le volume des pores de la pâte ont

été utilisés comme indicateurs de l'efficacité de mélange. Les pâtes ont été compactées dans un moule

cylindrique avec une contrainte axiale maximale de 60 MPa. Les échantillons verts ont ensuite été cuits à

1130 C pendant 12 heures. Les échantillons verts et cuits fabriqués en modifiant la température et le temps

de mélange ont été caractérisés par des mesures de densité, de porosité et de distribution de taille de pores.

Pour le mélangeur utilisé dans la fabrication des anodes de laboratoire, une durée et une température

optimales de mélange ont été suggérées permettant d’obtenir une meilleure efficacité de mélange, une densité

maximale, et une perméabilité à l'air et une surface spécifique minimales.

5.2 Abstract

The effects of mixing parameters on the quality of anode paste, as well as green and baked anodes were

studied. Four mixing times and four mixing temperatures were applied to make anode pastes at laboratory

scale. Specific surface area and volume of the pores in the paste were used as the indicators for mixing

effectiveness. The pastes were compacted in a cylindrical mould at a maximum pressure of 60 MPa. The

green samples were then baked at 1130 C for 12 hours. Density variations as well as porosity and pore size

distribution were measured for the green and baked samples as a function of mixing temperature and time. For

the setup used in this project, an optimum mixing time and mixing temperature were suggested, resulting in a

better mixing effectiveness, maximum density, minimum air permeability and specific surface area.

5.3 Introduction

In a homogeneous paste, coke aggregates are uniformly distributed in the binder matrix (a mixture of pitch and

fine coke particles). Pitch and binder matrix have also the possibility to penetrate into the coke pores during

mixing. Better mixing can result in formation of a more homogeneous paste and a better penetration of pitch.

This can lead to a compacted paste with lower amount of trapped air between and within the particles. The

anode properties are thus improved by reducing electrical resistivity, air permeability and carbon consumption

58

that can eventually lead to a longer service life and improved energy efficiency in the electrolysis cell. A good

mixing process has also the potential to reduce the pitch consumption leading to lower level of volatiles and

thus reducing the internal pressure and cracking rate during the anode baking cycle [116]. Efficient mixing can

be performed using sufficient mixing power and time and proper mixing temperature.

Although effects of raw materials and anode making parameters on the baked anode quality have been

extensively investigated, little has been published on the influence of mixing variables. Belitskus [119] studied

the effects of preheating temperatures, mixing time and mould temperature on the green and baked apparent

density as well as electrical resistivity of bench scale anodes. He reported an optimum mixing time for a given

raw material and mixer, beyond which mixing could be detrimental to anode properties. The effects of pitch

content and mixing time, temperature and intensity on the volume of intra-particle pores (0.02-100 m) were

explained by Stokka [80]. He found that with increasing mixing temperature and pitch content, a shorter mixing

time was required to obtain a given volume of paste porosity. Clery [126] used paste density as an indication for

optimizing the mixing process to obtain anodes with consistent green apparent density. He described that

using an intensive mixer after the kneader can improve the mixing effectiveness. However, he did not describe

how mixing variables influence paste and green anode properties.

The importance of anodes in the aluminium industry and the state of limited research performed on the mixing

process indicate the necessity of studying the effect of mixing variables which is expressed by homogeneity

and properties of green and baked anode. This part of project aimed to investigate the influence of mixing time

and temperature on the paste properties, its compaction behaviour and green and baked anode features using

several characterization techniques. X-ray computed tomography (CT), mercury porosimetry, air permeability

and BET surface area were carried out to determine the mixing effectiveness. This work can be used to

optimize the commercial anode fabrication parameters. The results of this chapter on paste characterization

were presented in THERMEC 2011 and were published in the journal of Advanced Materials Research, Vol.

409, 2012, P. 17-22. The results on green and baked anodes were published in Powder Technology, Vol. 235,

2013, P. 341-348.

5.4 Materials and methods

5.4.1 Sample preparation

Coke D with the size distribution shown in Table 3.6 was used for making the anode paste. A fine fraction with

a Blaine number of 4000 was used for all samples. The pitch/coke ratio was 14.8/100 for characterization of

pastes samples. At the time of paste characterization the optimum pitch/coke ratio for the particle granulometry in

Table 3.6 had not been determined and a pitch/coke ratio close to that used for industrial particle granulometry

was used. For making green and baked anodes the pitch/coke ratio was 16.2/100, as was determined to be the

59

optimum pitch ratio in chapter 3. Coke fractions and pitch were preheated in an oven for 120 and 30 minutes,

respectively, at the mixing temperatures. The mixing temperatures were selected based on the softening point

of pitch. The preheated materials were mixed as described in chapter 3 using seven combinations of mixing

time and temperature, indicated in Table 5.1.

Table 5.1: Mixing time and temperature for making anode pastes

Sample 1 2 3 4 5 6 7

Time (min) 6 10 15 20 10 10 10

Temperature (C) 178 178 178 178 158 168 188

The pastes were compacted at 150 C and a constant displacement rate of 10 mm/min to a maximum

pressure of 60 MPa. The details for compaction conditions applied to the pastes were explained in chapter 3.

Compaction curves (relative density vs. pressure) were plotted to study the influence of mixing variables on the

compaction behaviour of paste.

5.4.2 Characterization of anode paste

Specific surface area of the paste samples was measured using a BET surface analyzer. In addition, a

mercury porosimeter was used to measure the volume and size distribution of open pores in the paste

samples to study the influence of mixing time and temperature on the penetration of binder matrix into the

porosities and cracks within the coke particles.

5.4.3 Characterization of green samples

Apparent density of the green compacts was calculated using geometrical volume and mass. X-ray

tomography was also used to determine the homogeneity of the distribution of porosity, pitch and coke

particles in the green samples as an indication of mixing effectiveness. X-ray computed tomography is a non-

destructive method to obtain 3D attenuation of X-ray in material. The attenuation depends on atomic number

and bulk density of the material. When attenuation is calibrated, the intensity of X-ray, expressed by CT

numbers in Hounsfield unit (HU), can be associated with density [127] as Adams [128] and Picard [129] did for

apparent density of carbon anodes. One sample for each set of mixing variables was selected and the entire

green sample was scanned with a voxel resolution of 0.15 0.15 0.6 mm3 (pixel length pixel width slice

thickness) using a Siemens Somatom Sensation 64, shown in Figure 5.1. The CT numbers gave a 3D density

map of the sample. MATLAB software was used to construct the images from CT numbers and to analyse the

data. Average and standard deviation of CT numbers were calculated. Maximum height of CT number profile

(Rt) is the difference between the maximum and minimum CT numbers along the diameter of a slice. Ra is the

average of the distance between the successive peaks and valleys of the profile. The Rt and Ra for CT number

60

fluctuations were determined along the diameter of 10 slices from each green sample and the average was

calculated. The 10 slices had similar positions in each entire green anode.

Figure 5.1: CT scanner used for green samples (Siemens Somatom Sensation 64)

Two sections with a thickness of 20 mm were cut from the bench scale green compacts as shown in Figure

5.2. The volume of sections 1 and 2 was measured using water displacement method as explained in chapter

3. The green samples (section 1) were then cut and the volume of few chunks with a weight of 50-62 g was

measured by helium pycnometer. Gas pycnometer results for this material are not sensitive to the sample size

when it lies in cm range. The percentage of open porosity in the total volume of green samples was

determined using the green apparent density of section 1 and the pycnometer density (Equation 3.3).

Figure 5.2: Schematic representation for characterization of green and baked samples

The influence of mixing variables on the volume and size distribution of open pores in green samples was

determined using a mercury porosimeter. Green parts with the weight of 6.5-8.5 g were used for pore structure

analysis. A small part of each green sample with a surface area of 3-3.7 cm2 was impregnated by a polishing

resin under mechanical vacuum and polished. Optical microscopic images were processed with MATLAB and

the average thickness of binder matrix was determined. The image processing procedure is explained in

D=68 mm

GAD

CT scan

1

2

3 Air permeability

He-pycnometer

BET surface area Core sample

D=50 mm

GAD

Baking 1130 C

12 hr

GAD

BAD

Hg-porosimeter

He-pycnometer

Microscopy

61

appendix A. The variations of binder matrix thickness with mixing parameters were studied. The penetration of

binder matrix into the coke porosities was qualitatively compared.

5.4.4 Characterization of baked samples

Section 2 of the samples was baked at a maximum temperature of 1130 C for 12 hours. Water displacement

method was used to determine the baked apparent density (BAD). The baked samples were core drilled to

disks with a diameter of 50 mm and thickness of 20 mm. Air permeability was determined following ISO

standard 15906 [130] using R&D Carbon equipment. The disks were then cut and the volume of few chunks with

a weight of 40-47 g was measured by a helium pycnometer. The percentage of open porosity in baked

samples was determined using the baked apparent density of section 2 and pycnometer density (Equation

3.3). Specific surface area of the baked chunks with a weight of 2.3-2.8 g was measured using BET method.

5.5 Results

5.5.1 Green paste

Figure 5.3 shows the specific surface area of the paste as a function of mixing temperature at a constant

mixing time of 10 minutes. Paste surface area decreased by 36% when mixing temperature was increased

from 158 C to 188 C. The effect of mixing time on the BET surface of paste is shown in Figure 5.4. The

specific surface area reduced by 25% with increasing mixing time from 6 to 15 minutes, at a constant mixing

temperature of 178 C. Further increase in mixing time resulted in a 7% increase in the surface area.

Figure 5.3: BET surface area of the pastes mixed for 10 minutes at different mixing temperatures

62

Figure 5.4: BET surface area and total volume of pores for the pastes mixed at 178 C for different mixing times

Secondary axis of Fig. 5.4 shows the total volume of pores in the paste samples as a function of mixing time,

measured by mercury porosimetry. It shows the same trend as that observed for the BET surface area of the

pastes with increasing mixing time. Pore volume reduced by 25% when mixing time was increased from 6 to

10 minutes. Longer mixing time up to 20 minutes led to 54% increase in the volume of pores. These results

are in agreement with those published by other authors [80, 89]. They have reported that for a given mixing time

which depends on the mixer characteristics and pitch ratio, paste porosity is reduced to a minimum and then is

increased with further increase in mixing time. Figure 5.5 shows the differential volume of pores in the pastes

with different mixing times. When mixing time was increased from 6 to 10 minutes, the size of pores

contributing to total pore volume was reduced from 360 m to less than 200 m. With mixing times longer

than 10 minutes, larger pores (200-380 m) appeared in the paste structure.

Figure 5.5: Incremental pore volume vs. pore size for the pastes mixed at 178 C for different mixing times

178 C, 6 min 178 C, 10 min 178 C, 15 min 178 C, 20 min

63

Compaction tests were carried out for seven pastes to study the influence of mixing time and temperature on

the compaction behaviour of the pastes. It can be observed in Figures 5.6 and 5.7 that there is a meaningful

difference between the compaction curves. Apparently, the test is capable of illustrating the effects of mixing

parameters on the paste compaction behaviour.

Figure 5.6: Effect of mixing time at a constant mixing temperature (178 C) on the compaction behaviour and apparent density of green samples

Figure 5.7: Effect of mixing temperature at a constant mixing time (10 min) on the compaction behaviour and apparent density of green samples

5.5.2 Green anodes

Green apparent density variations were investigated to reveal the influence of mixing variables. Figure 5.8

shows the average of green apparent density for the entire samples as well as for sections 1 and 2, as a

function of mixing time (Fig. 5.8-a) and temperature (Fig. 5.8-b). It can be observed that there is a meaningful

158 C 168 C 178 C 188 C

64

variation in GAD depending on both variables. At a mixing temperature of 178 C, when mixing time was

extended from 6 to 10 minutes, the obtained density increased; while mixing for longer than 10 minutes led to

a slightly lower green apparent density as shown in Fig. 5.8-a. For a mixing time of 10 minutes in Fig. 5.8-b,

when mixing temperature increased from 158 C to 178 C, the average GAD of the entire samples increased

from 1.47 g/cm3 to 1.51 g/cm3. Further increase of mixing temperature to 188 C reduced the green density to

1.49 g/cm3. As expected, Figure 5.8 confirms that a density gradient exists and GAD reduces from top to the

bottom of the samples, produced by a one-side compaction method.

Figure 5.8: Dependence of green apparent density on (a) mixing time at a constant mixing temperature of 178 C; (b) mixing temperature for a constant mixing time of 10 minutes

Volume percentage of open pores in section 1 was calculated using Equation 3.3 and is shown in Figure 5.9.

As expected, the results follow a trend in agreement with GAD variation, i.e., the minimum volume of open

pores was obtained when the anode paste was mixed for 10 minutes at 178 C.

Figure 5.9: Dependence of open porosity of section 1 on (a) mixing time at a constant mixing temperature of 178 C; (b) mixing temperature for a constant mixing time of 10 minutes

Previously highlighted density gradients were investigated using X-ray computed tomography by means of

density profiles in the green samples. Figure 5.10 demonstrates an example of CT profile of 0.6 mm thick

slices located at the middle of the height of three samples that were made with different mixing times and

65

temperatures. It also indicates the fluctuations in the profile along the diameter of the slices. The Rt and Ra

were used as indications for homogeneous distribution of coke, pitch and porosity in the green samples.

Smaller Rt and Ra values are associated with more homogeneous samples. Figure 5.11 reveals the variations

in the Rt and Ra with mixing variables along the diameter (average for 10 slices). They showed a similar trend

to GAD. The minimum fluctuations in density and therefore the best homogeneity was observed for the

samples that were mixed at a minimum temperature of 178 C for 10 minutes.

Slices in the middle of sample height

6 min, 178 C 10 min, 178 C 10 min, 158 C

Rt= 308 HU Rt= 255 HU Rt= 381 HU

Ra= 157 HU Ra= 124 HU Ra= 208 HU

Figure 5.10: Constructed images form CT numbers and the profile of CT number along the diameter (black line) of green samples made with different mixing times and temperatures

Figure 5.11: Variations in Rt and Ra with mixing parameters

66

Standard deviation of CT number was also used to express the degree of homogeneity in the distribution of

porosity, pitch and coke particles throughout the paste. The CT number of a voxel represents the presence

and the amount of pitch, coke and air (pore) in the voxel. Pitch is a carbonaceous material and its CT number

is close to that of coke. In addition, binder matrix that fills the voids between the large particles contains pitch

and fine coke particles. Thus, the major difference between the CT numbers is due to the presence or

dispersion of the pores in the voxels, rather than the non-uniform distribution of pitch or coke particles. In other

words, CT with the voxel resolution of 0.15 0.15 0.6 mm3 can reveal large inhomogeneity in the distribution

of pitch and coke, such as accumulation of pitch and agglomeration of the fine particles, where a voxel

consists of only pitch or coke.

Figure 5.12-a shows standard deviation of CT number as a function of mixing time for both entire green

samples and the section 1 of each sample. CT number decreases when mixing time increases from 6 minutes

to 10 minutes but it does not show a meaningful variation when the time is extended beyond 10 minutes. It

may be an indication that mixing for 10 minutes at 178 C resulted in the homogeneous distribution of porosity,

pitch and coke aggregates. Figure 5.12-b shows that for both entire green samples and the section 1 of each

sample, the average CT number increases by increasing mixing time up to 10 minutes. Again, this average

does not change significantly by further increase of mixing time. This is in agreement with GAD results

presented in Figure 5.8-a since the average CT number is directly related to the GAD.

Figure 5.12: CT numbers in the entire and the section 1 of anode samples mixed at 178 C for different mixing times; (a) standard deviation of CT numbers; (b) average and fluctuations of CT numbers

Standard deviation of CT numbers as a function of mixing temperature is presented in Figure 5.13-a. It is

noted that the standard deviation decreases significantly by increasing the mixing temperature from 158 C to

178 C and it increases by further increase in mixing temperature. The average CT number vs. mixing

temperature for both entire samples and section 1 of each sample (Figure 5.13-b) confirms the GAD results

presented in Figure 5.8-b, assuming once again that GAD is directly related to the average CT number.

67

Figure 5.13: Results of CT numbers in the entire and the section 1 of anode samples mixed for 10 minutes at different mixing temperatures; (a) standard deviation of CT numbers; (b) average and fluctuations of CT numbers

Section 1 of the green samples demonstrated larger average of CT number compared to the entire samples. It

confirms the density gradient in the entire anode and higher density in top layers of the sample, as indicated in

Figure 5.8. Figure 5.14-a shows an example of gradient in apparent density determined by average CT

number along the height of a green sample. Figure 5.14-b demonstrates both vertical and radial gradient in

density. Density decreased from the central axis to the border of the sample which is due to the friction

between the paste and the mould (wall effect). It is more obvious towards the bottom of the sample.

Figure 5.14: Variations in the average CT number of a green sample mixed for 10 minutes at 178 C (a) on the surface of the slices along the height; (b) in the radial direction on the slices as a function of height

Pore size distribution of green samples was determined by mercury porosimetry. The samples consist of large

pores in the range of 300-500 m, medium pores between 5 m and 30 m, and small pores smaller than 5

m. Figure 5.15 shows that with extending mixing time from 6 to 20 minutes at 178 C, the size of largest

medium pore decreased from 24 m to 12 m. The maximum size of medium pores also reduced from 26 m

towards 11 m with increasing mixing temperature from 158 C to 188 C.

68

Figure 5.15: Effect of mixing time and mixing temperature on the pore size distribution of the green samples

Examples for optical microscopic images of green samples (1, 2 and 5) are presented in Figure 5.16. In

sample 5, where mixing was performed at 158 C for 10 minutes, the distribution of pitch and particles was not

uniform (Figure 5.16-a) and some open pores were not filled by binder matrix (Figure 5.16-b). In sample 2,

mixed at 178 C for 10 minutes, the distribution of particles was more uniform than that in sample 5 and did

not indicate agglomeration (Figure 5.16-d). In addition, more pores were filled with binder matrix (Figure 5.16-

e). In sample 1 with a mixing temperature of 178 C and shorter mixing time of 6 minutes, the pores were

partially filled or the pore entrance was just blocked by binder matrix (Figure 5.16-f).

Figure 5.16: Examples for microstructure of the green samples; 10 minutes at 158 C (a, b, c); 10 minutes at 178 C (d,

e); 6 minutes at 178 C (f). The samples are impregnated with resin and polished. White arrows indicate open pores filled by resin while the black ones demonstrate the pores filled by pitch (light gray: coke, medium gray: resin, dark gray/black:

pitch)

a

900 µm

d

900 µm

c

100 µm

b

900 µm

f

100 µm

e

100 µm

Mixing temperature: 178 C

158 C 168 C 178 C 188 C

69

Average thickness of binder matrix is a measure of distance between the aggregates and is shown in Figure

5.17. It is expected that a more effective mixing may result in a lower binder matrix thickness (particle

distance) and therefore a higher GAD. The thickness of binder matrix decreased with increasing mixing time to

10 minutes and then increased above that for 6 minutes. This large increase is not in accordance with GAD of

the samples where after 10 minutes of mixing GAD did not change significantly. Figure 5.18 shows that the

distribution of binder thickness moved towards thicknesses below 100 m with increasing mixing time from 6

to 10 minutes. Mixing over 10 minutes increased the percentage of binder thickness in the range of 100-200

m and 200-400 m and led to higher average thicknesses.

Figure 5.17: Variations of binder matrix thickness with (a) mixing time at a constant mixing temperature of 178 C; (b) mixing temperature for a constant mixing time of 10 minutes

Figure 5.18: Distribution of binder matrix thickness for samples mixed at 178 C for (a) 6 minutes; (b) 10 minutes; (c) 15 minutes; (d) 20 minutes

Binder matrix thickness decreased with increasing mixing temperature from 158°C to 178°C (Figure 5.17-b).

Mixing at 188°C increased the binder thickness that is in agreement with GAD of the samples. Figure 5.19

demonstrates that increasing mixing temperature up to 178°C led to a higher percentage of binder thickness in

the range of <100 m and 100-200 m.

70

Figure 5.19: Distribution of binder matrix thickness for samples mixed for 10 minutes at (a) 158 C; (b) 168 C;

(c) 178 C; (d) 188 C

5.5.3 Baked anodes

Apparent density of the baked sections (section 2) was studied as a function of mixing parameters. As

indicated in Figure 5.20-a, when mixing time was increased from 6 to 10 minutes at a mixing temperature of

178 C, the average BAD was improved from 1.47 g/cm3 to 1.52 g/cm3. Extended mixing time to 15 and 20

minutes decreased the density to 1.485 g/cm3 and 1.489 g/cm3, respectively. For a mixing time of 10 minutes,

when mixing temperature increased from 158 C to 178 C, the average BAD increased from 1.44 g/cm3 to

1.52 g/cm3, as shown in Fig. 5.20-b. This is in agreement with the GAD results in Figure 5.8-b where mixing

temperature of 178 C resulted in the highest green density. Mixing at 188 C slightly reduced the baked

density to 1.485 g/cm3.

Figure 5.20: Dependence of baked apparent density and open porosity of baked samples on (a) mixing time at a constant

mixing temperature of 178 C; (b) mixing temperature for a constant mixing time of 10 minutes

The percentage of open porosity in the total volume of baked sections was calculated using the baked

apparent density and the density determined by He-pycnometer. The results are compared with baked

apparent density in Figure 5.20 and are in a good accordance, as expected. Mixing for 10 minutes at 178 C

71

resulted in the minimum percentage of open porosity. Air permeability of the baked sections indicated in Figure

5.21 validates the results of apparent density and open pore content of the anodes.

Figure 5.21: Effect of mixing time and mixing temperature on the air permeability of the baked sections

In order to validate the effect of mixing variables on the mixing effectiveness, the specific surface area (BET

surface area) of small chunks of the baked sections was measured. Mixing at 178 C for 10 minutes resulted

in the lowest BET surface area (0.176 m2/g), as observed in Figure 5.22-a. Specific surface area then

increased to 0.237 m2/g with longer mixing times as apparent density reduced. Figure 5.22-b shows that BET

surface area reduced from 0.226 m2/g to 0.166 m2/g with increasing mixing temperature from 158 C to 168

C. Higher mixing temperatures up to 188 C resulted in a slight raise in the BET surface area. The trend of

BET surface area as a function of mixing temperature is not in accordance with that of baked apparent density.

In addition, one may expect that the BET value will increase due to the decrease in the density but such a

trend is not clearly observed. This inconsistency is partially due to the considerably large error bars on some

points, especially at low BET values. It could also be due to the effect of pore size distribution. Large pores

have a great contribution to density but a small contribution to specific surface area. The small pores, on the

other hand, have the opposite effect of greatly contributing to BET value. For instance, one may note that in

Fig. 5.15-b, the amount of very large pores decreases as the mixing temperature increases and, beyond 178

C it begins to increase. At 178 C, the sample exhibits a maximum density, thus the minimum BET value is

expected. This is not observed since the small pores compensate for the effect of total pore volume fraction.

72

Figure 5.22: Influence of mixing time and mixing temperature on the BET surface area of the baked sections

5.6 Discussions

5.6.1 Mixing time

Extending mixing time from 6 to 10 minutes resulted in an increase in the green apparent density and a

decrease in the volume of open pores of green anodes (Figure 5.9-a) when mixing was performed at 178 C.

The trend of the average CT number in Figure 5.12-a also confirms the GAD determined by water

displacement. Mixing for longer times of 15 and 20 minutes led to a minor decrease in both GAD and average

CT number. It suggests that mixing time is a variable with considerable influence on the anode quality. For the

mixing process and the mixer used in this study, 10 minutes was the optimum mixing time which may have

resulted in a better distribution of coke particles and pores in the paste, more uniform distribution of pitch over

the aggregates and lower binder matrix thickness. The continuous pitch film that forms over the particles may

have two consequences; a better particle rearrangement and densification during compaction, and improved

compression strength after baking due to a better pitch-coke bridging between the coke particles [89]. The

hypothesis of more uniform distribution is supported when the variations in the CT number are considered

(Figures 5.11-a and 5.12-b). Anodes with mixing time of 10 minutes demonstrated the minimum standard

deviation for CT number and minimum height of CT number profile which is an indication for more

homogeneous paste. It may have also resulted in more penetration of binder matrix into the coke pores which

contributed to enhancing the GAD. Figure 5.15-a reveals that with longer mixing time smaller pores in green

samples were filled by binder matrix and the maximum size of medium pores (30-5 m) was reduced from 24

m to 12 m. Similar effect of mixing time was observed on green paste properties in Figures 5.4 and 5.5.

Baked samples with 10 minutes of mixing also showed the maximum apparent density and therefore the

minimum percentage of open pores and air permeability. A decrease in BET surface area for baked samples

was also observed when mixing time was increased from 6 to 10 minutes. It can be another indication for

penetration of binder matrix into the coke pores with a more effective mixing.

73

Mixing for longer than 10 minutes resulted in a decrease in BAD and an increase in open porosity, air

permeability, binder matrix thickness and BET surface area. Belitskus [119] had reported the degradation of

anode properties with mixing over an optimum time duration. Particle breaking may occur due to over mixing

and, therefore, new surfaces will appear which may have a lower chance to be wetted by pitch. In addition, the

coke granulometry is changed. These uncovered surfaces as well as non-optimum granulometry deteriorate

the paste compaction and lead to lower density and higher porosity and permeability. When particles are

broken, closed pores in coke particles have the chance to become open to the new surfaces. These empty

pores contribute to the increase in open porosity and specific surface area, as demonstrated in Figures 5.20-a

and 5.22-a, respectively. Figure 5.15-a also indicates that the volume of large pores has increased with mixing

over 10 minutes. Longer mixing times may also result in the coke attrition and incorporation of new coke fines

into the pitch [119]. This increases the viscosity and results in lower penetration ability of pitch. High viscosity of

pitch may also decrease the number of binder-coke bridges that form between coke particles. This will in turn

result in paste degradation and lower density.

5.6.2 Mixing temperature

Increasing mixing temperature from 158 C to 178 C resulted in an increase in the green apparent density,

average CT number and baked apparent density of the anodes, indicated in Figures 5.8-b, 5.13-a, and 5.20-b,

respectively. Mixing at 188 C for 10 minutes led to a decrease in GAD, average CT number and BAD.

According to pitch softening point of 109 C, the pitch is not fluid enough at 158 C. The viscous pitch results

in agglomeration of the fine particles without penetration into the open pores. High standard deviation of CT

numbers, Ra value and binder matrix thickness confirm the non-uniform distribution of pitch in anodes mixed at

158 C. Such a low mixing effectiveness and heterogeneity in the paste led to the highest porosity,

permeability and BET surface area, as shown in Figures 5.9-b, 5.21-b and 5.22-b, respectively. Increasing the

mixing temperature to 178 C reduced the viscosity of pitch, as indicated in Table 3.3, which may have

improved its distribution and flowability and increased its penetration and pore filling effect. The results of

microscopic study in Figure 5.16 indicate that after mixing at 178 C more coke pores are filled by pitch

comparing to the mixing temperature of 158 C. Figure 5.15-b exhibits that higher mixing temperature led to

pitch penetration into smaller pores and the maximum size of medium pores reduced to 11 m. This

eventually resulted in the maximum density and consequently the minimum level of porosity and permeability.

Further increase of mixing temperature to 188 C resulted in a decrease in apparent density and deteriorated

anode porosity and permeability. It can be due to the fact that viscosity of pitch decreases at higher

temperatures and improves the pitch penetration into the coke pores. Thus, the amount of remained pitch may

not be enough to fill the voids between the particles [80, 119]. It is shown in Figure 5.15-b that the volume of large

pores increased with increasing the mixing temperature over 178 C. This increase in the volume of voids

74

between the particles can be responsible for the increase observed in BET surface area when the temperature

was increased from 168 C to 188 C (Figure 5.22). It has also been reported that optimum mixing time

decreases at higher mixing temperatures [80]. A mixing time of 10 minutes at 188 C may thus be longer than

the optimum time at this temperature.

5.7 Summary

Mixing the anode constituents is an important step in the anode production so that the mixing variables can

influence the mixing effectiveness and consequently the anode quality. The results in this chapter showed that

there is an optimum mixing time and temperature at which the penetration of binder matrix and distribution of

coke, pitch and porosity can be improved leading to enhanced anode density. Mixing parameters including

time and temperature influenced homogeneity, pore volume, pore size, and therefore the density and

permeability of the anodes. It can be suggested that for the materials and experimental setup used in this

project, mixing at 178 C for 10 minutes revealed the best mixing effectiveness and resulted in the lowest

volume of pores, specific surface area and permeability. This optimum mixing temperature is in accordance

with the temperature used in commercial anode plants that is often 60 to 70 C over the softening point of

pitch.

6 Chapter 6: Influence of formulation on the

compaction behaviour of paste

6.1 Résumé

Les particules de coke sont rigides et supposées être indéformables au cours du processus de compactage de

la pâte d'anode. Ainsi, le comportement de déformation de la matrice liante joue un rôle déterminant dans ce

processus de compactage. La matrice liante est généralement caractérisée par une teneur en brai et une

distribution de taille des fines particules de coke. La compaction dans un moule rigide de différentes

formulations de matrice liante a été étudiée dans ce chapitre. Les matrices liantes ont été compactées à une

contrainte uniaxiale maximale de 70 MPa à 150 C. Deux vitesses de déformation (2.910-4 s-1 et 2.910-3 s-1)

ont permis d'évaluer la contribution du comportement visqueux du matériau à la compaction de la matrice

liante en fonction des deux paramètres étudiés (formulation de matrice liante et taux de déformation). Une

étude similaire de compactage avec des vitesses de déformation de 1.810-4 s-1 et de 1.810-3 s-1 a été

appliqué sur des échantillons de pâte avec différentes teneurs de brai. Cette étude a révélé que la compaction

de la matrice liante et de la pâte d'anode avec des compositions classiques n'est pas significativement

dépendante de la durée de compaction. En conséquence, la contribution de la viscosité du matériau n’est pas

considérée comme étant essentiel dans le processus de compactage du matériau.

6.2 Abstract

Coke particles are rigid and assumed to be non-deformable during compaction. Thus, deformation behaviour

of binder matrix plays a crucial role in the anode paste compaction process. Binder matrix is usually

characterized by pitch content and fineness of the fine coke particles. Compaction of binder matrix with

different compositions in a rigid closed die was studied in this chapter. Binder matrix compositions were

compacted to a maximum uniaxial pressure of 70 MPa at 150 C. Two strain rates of 2.910-4 s-1 and 2.910-3

s-1 enabled us to evaluate the contribution of viscous behaviour of the material to the compaction of binder

matrix as a function of its composition and deformation rate. A similar experimental compaction procedure with

strain rates of 1.810-4 s-1 and 1.810-3 s-1 was applied to paste samples with different pitch contents. This

study revealed that the compaction of binder matrix and anode paste with conventional compositions was not

a significant time dependent process. Therefore, contribution of viscosity may not be significant to the

compaction of the material.

76

6.3 Introduction

Green density of anode depends on the compaction behaviour of the paste and is strongly correlated with final

density, electrical [26] and mechanical [131] properties of the anode. The main target of anode makers is therefore

to obtain high and homogeneous density through the anode in order to decrease its electrical resistivity and to

increase its service life. This part of project was performed to study the compaction behaviour of anode paste and

to reveal the effect of its rheological behaviour on its compaction during forming.

During mixing, fine cokes are embedded into the liquid pitch resulting in a viscous material (binder matrix)

which surrounds the large coke aggregates. This viscous material deforms during compaction and fills the

voids either between the large aggregates or inside them. The coke aggregates are considered as a non-

deformable phase of the paste although being subjected to rearrangement during compaction. Coke

aggregates and binder matrix are therefore the principal constituents of anode paste, which may influence its

compaction properties, and consequently the green density obtained after compaction.

Since the binder matrix surrounds the coarse coke particles and deforms during compaction, its rheological

parameters could be important in determining compaction behaviour of paste. In turn, volume fraction and

granulometry of fine coke could be considered as two major parameters affecting the rheological properties of

the binder matrix and its capability in filling the voids. The research works on prebaked anode paste have been

mostly focused on the effect of fine coke granulometry on the pore filling capability of binder matrix and on the

properties of the final product. Hulse [94] reported that prebaked anode paste had a granulo-viscoelastic

behaviour depended on temperature, pitch content and coke particle characteristics such as size distribution,

shape and roughness. By increasing the pitch content and temperature, the viscosity of paste decreases. This

decreased viscosity will have a larger contribution in the compaction. Higher coke content and coke fineness

on the other hand enhance the elastic behaviour. Figueiredo [85] showed that by using smaller particle size of

the fine coke and optimizing the pitch content, a larger density, lower electrical resistivity and air permeability

could be obtained. Similar improvement was reported on the baked properties of binder matrix only when the

weight fraction of fine coke particles smaller than 96 m was increased from 43% to 86% and an optimum

pitch content was used [88]. It has also been reported that the granulometry and the amount of fine coke

influence the capability of the binder matrix in filling the voids between coke aggregates. Vidvei [82] found out

that decreasing the particle size of the fine coke (60% smaller than 75 m) led to an improvement in anode

density and a decrease in electrical resistivity. However, the behaviour of paste during compaction was not

reported in the literature.

Rheological properties of binder matrix have been extensively investigated for Söderberg anodes where the

anode is not pre-formed but baked in-situ over the electrolysis cell. Gildebrandt et al. [90, 91], Kravtsova et al. [92]

and Vershinina [93] studied the effect of coke content, temperature, and particle size of fine coke on viscosity of

77

binder matrix. Their experimental data showed that the viscosity of binder matrix decreased with increasing

temperature and particle size of fine coke but when the amount of fine coke (<74 m) exceeded over 50 wt.%

of binder matrix, it increased significantly. The possible explanation for the effect of particle size on viscosity

has been given by Hulse [94] and Kravtsova [92] who ascribed this effect to direct particle-particle interaction,

number of contacts and thickness of pitch layer on the particle surface. Sakai [95, 96] found that 50 wt.% of fine

coke in binder matrix was a critical point since its behaviour changed from Newtonian to viscoelastic when

coke amount exceeded that point. Again, the influence of binder matrix formulation on compaction behaviour of

paste and green density of anode has not been considered in these works since they have been mostly carried

out on Söderberg anodes where the compaction of paste was not of great interest. There is thus a lack of

information on the effect of binder matrix parameters on anode paste compaction behaviour.

Anode forming process takes place within a relatively short period of time, typically one minute for vibro-

forming and half a minute for pressing. Considering the viscous properties of binder matrix, at a first glance,

one may expect that the compaction rate plays a significant role in paste compaction behaviour. Strictly

speaking, it is expected that the paste shows a time dependent behaviour and the viscous phase, i.e., binder

matrix, flows under pressure and fills the voids over time leading to a better densification at longer pressing

times. On the other hand, at high coke volume fractions, a solid skeleton may form and the compaction

behaviour may be governed by the strength of the particle/particle contacts. Therefore, this part of project

aimed to reveal whether the binder matrix and paste exhibited a time-dependent behaviour and if this time

dependency affected the compaction behaviour of anode paste. However, in addition to rheological behaviour,

other phenomena such as air entrapment might affect the time dependency and compaction behaviour of the

material that can be studied in future works. In the first step, the compaction characteristics of binder matrix as

a function of weight fraction and particle size of fine coke were studied. Then, the compaction tests were

performed on paste with different pitch contents and fine coke granulometries to better understand the viscous

behaviour and its importance in the compaction of paste with different formulations. The results of this chapter

were published in Powder Technology, Vol. 246, 2013, P. 650-657.


Fine fractions of coke D with the Blaine number of 2300, 4000 and 6300 were used for making the samples.

Specific surface areas of fine cokes were 2.9 m2/g, 4.1 m2/g and 6.1 m2/g, respectively, and size distribution is

shown in Table 3.2. Each fine coke with a given Blaine number was mixed with pitch at different pitch to fine

coke ratios (P/FC), indicated in Table 6.1, to produce a binder matrix. Lower P/FC ratios were used for the

samples with a BN of 2300 since this Blaine number represents lower specific surface area to absorb the pitch.

Fine coke and pitch were preheated and mixed at 178 C for 10 minutes, as explained in chapter 3. Total

mass of fine coke and pitch was 274 g for all formulations.

78

Table 6.1: Composition and compaction parameters for binder matrix samples

Sample

BET surface

area of fine

coke (m2/g)

P/FC ratio

Displacement rate

of the press

(mm/min)

Average deformation

rate (s-1)

Uniaxial

pressure

(MPa)

Binder

matrix

2.9 (BN 2300) 30/100, 34/100, 38/100 1,10 2.910-4, 2.910-3

70 4.1 (BN 4000) 34/100, 38/100, 42/100 1,10 2.910-4, 2.910-3

6.1 (BN 6300) 34/100, 38/100, 42/100 1,10 2.910-4, 2.910-3

Bulk density of the binder matrix compositions was measured and the samples were compacted at 70 MPa, as

described in chapter 3. Compaction tests were performed at two constant displacement rates (DR) of 1 and 10

mm/min. Average deformation rates were calculated using the initial and final heights of the samples (Table

6.1). Compaction curves (relative density vs. pressure) were derived from the tests for further analysis.

For the second part of the tests, the influence of displacement rate on the compaction of anode pastes

(including large aggregates) with different compositions was studied. Fine cokes with a BN of 2300, 4000 and

6300 were used to make the paste samples. Tables 3.6 and 6.2 show the size distribution of coke particles

and pitch to coke ratios (P/C), respectively. The paste samples were made with the same mixing parameters

as used to prepare the binder matrix. They were then compacted at 150 C to a maximum pressure of 60 MPa

at different displacement rates of 0.1, 1 and 10 mm/min (Table 6.2). Total mass of coke and pitch was 488 g

for all compositions.

For further investigation of the contribution of viscosity to the compaction of paste, a creep test was performed

on the paste samples. The paste samples were first compacted with a displacement rate of 10 mm/min until an

axial pressure of 10 MPa was reached (Table 6.2). Then, the pressure was kept constant at 10 MPa for one

hour during which displacement was being recorded. Densification of the pastes at constant pressure was

conducted to study the influence of compaction time.


6.5.1 Binder matrix

Figure 6.1 shows the effect of P/FC ratio and displacement rate on the compaction curves of the binder matrix

samples when a fine coke with a SSA of 2.9 m2/g is used. For each sample, two DRs were used; 1 and 10

mm/min. As expected, it can be seen that P/FC ratio has a significant effect on the compaction behaviour of

the material. The slope of the curves at early stages of compaction increases drastically when P/FC ratio

increases. This is due to the fact that movement and rearrangement of particles happen at the early phase of

compaction [132] since the material is not closely packed after filling the die. Pitch is assumed to act as a

lubricant to help particle rearrangement. Therefore, the samples with higher P/FC ratios reached a higher

79

relative density at lower pressure and revealed a steeper slope. Displacement rate, on the other hand, has a

minor influence on the compaction curves. At low P/FC ratio, i.e., 30/100, displacement rate did not influence

the compaction behaviour and the curves coincided. The effect of DR became more distinctive as the P/FC

ratio increased and resulted in a material with lower viscosity. This is the characteristic of viscoelastic materials

for which the compaction time contributes to densification [133].

Table 6.2: Composition and compaction parameters for paste samples

BET surface

area of fine

coke (m2/g)

P/C ratio Displacement rate of the

press (mm/min) Average deformation rate (s-1)

Uniaxial pressure

(MPa)

2.9 (BN 2300)

14/100 10 1.810-3 60

16.2/100 1, 10 1.810-4, 1.810-3 60

19/100 1, 10 1.810-4, 1.810-3 60

22/100 1, 10 1.810-4, 1.810-3 60

4.1 (BN 4000)

14/100 10 1.810-3 60

16.2/100 0.1, 1, 10 1.810-5, 1.810-4, 1.810-3 60

16.2/100 10 1.810-3 10*

19/100 1, 10 1.810-4, 1.810-3 60

22/100 1, 10 1.810-4, 1.810-3 60

22/100 10 1.810-3 10*

6.1 (BN 6300)

14/100 10 1.810-3 60

16.2/100 0.1, 1, 10 1.810-5, 1.810-4, 1.810-3 60

16.2/100 10 1.810-3 10*

19/100 1, 10 1.810-4, 1.810-3 60

22/100 1, 10 1.810-4, 1.810-3 60

22/100 10 1.810-3 10*

* Creep test on the paste at 150 C and 10 MPa for 1 hour.

For the compositions with the SSA of 4.1 and 6.1 m2/g the effect of P/FC on the compaction curves showed

the same trend (Figures 6.2 and 6.3) with the exception that the strain rate dependency shifted toward higher

P/FC as the Blaine number increased. For both numbers, compositions with a P/FC ratio of 34/100 did not

reveal a viscous behaviour during compaction since the curves of different DRs practically coincided. For the

compositions with the P/FC ratios of 38/100 and 42/100 a non-negligible difference was observed when

different DRs were used and lower DRs resulted in better densification. This indicates that binder matrix

samples with the SSA of 4.1 and 6.1 m2/g start to reveal a viscous behaviour when P/FC ratios exceed 38/100.

80

Figure 6.1: Compaction curves for the binder matrix samples made from a fine coke with SSA of 2.9 m2/g (BN 2300) and P/FC ratios of 30/100, 34/100 and 38/100


81


For a given granulometry for fine coke (BN), the influence of strain rate on compaction behaviour of binder

matrix was less significant at lower P/FC ratios. For a specific size distribution of particles when solid

concentration increases, particle-particle interaction becomes a dominant mechanism that restricts movement

of particles and masks viscous effect of the pitch. In other words, the viscosity of binder matrix increases with

increasing the content of fine coke [90, 91, 94], and subsequently, the contribution of compaction time is

decreased.

According to Figures 6.1 to 6.3, for a given granulometry of fine coke, compositions with higher P/FC ratio

required lower pressure to reach a specific relative density. The reason lies in the fact that changing the

proportion of material constituents, i.e., pitch and coke contents, not only changes the whole material

properties such as viscosity and yielding characteristics [90-93] but also varies some structural aspects. For

example pitch wets the coke particles and acts as a lubricant. A high P/FC ratio, therefore, provides a

continuous layer of pitch on the particle surface which reduces the interparticle friction and facilitates the

rearrangement of the particles that contribute to compaction. In addition, it was observed that as the pitch

content is higher in binder matrix, the compaction curve begins from a higher initial relative density, as

indicated in Table 6.3. This is basically due to the fact that the initial density was recorded after applying a

small pre-load of 14.2 kPa. As stated above, the slope of the compaction curves increases with increasing the

P/FC ratio and is considerably high at very early stages of the compaction process. Consequently, the

samples with higher P/FC ratio are much more sensitive to pre-load than those with lower P/FC ratios are. This

explains the difference between the initial densities of different samples.

82

Table 6.3: Initial apparent and relative densities of binder matrix with different compositions before compaction

BET surface area of

fine coke (m2/g) Pitch/fine coke ratio

Initial apparent density

(g/cm3)

Initial relative

density

2.9 (BN 2300)

30/100 1.0111 0.5562

34/100 1.0615 0.5907

38/100 1.1145 0.6269

4.1 (BN 4000)

34/100 0.9497 0.5285

38/100 0.9914 0.5577

42/100 1.0397 0.5906

6.1 (BN 6300)

34/100 0.9671 0.5382

38/100 0.9864 0.5549

42/100 1.0128 0.5753

In order to evidence the effect of particle size distribution, some compaction curves are presented for the same

P/FC ratios while varying the fine granulometry (Figures 6.4 and 6.5). Again, it can be seen that granulometry

(BN) has a significant effect on the compaction behaviour of the material. Similar to the effect of P/FC ratio, the

slope of the curves at early stages of compaction increases drastically when BN decreases. In addition, it

should be noted that displacement rate has a negligible influence on the compaction curves

.

Figure 6.4: Compaction curves for the binder matrix samples made with a P/FC ratio of 34/100 and fine cokes with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively)

83

Figure 6.5: Compaction curves for the binder matrix samples made with a P/FC ratio of 38/100 and fine cokes with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively)

For a given P/FC ratio, compositions with larger granulometry needed lower pressure to reach a specific

relative density. For large particles, the specific surface area decreases and less pitch is required to wet the

particle surface. Therefore, more pitch is available to fill the voids and to lubricate the particle/particle

interfaces and so facilitating their movement. For example, for the P/FC ratio of 38/100 when the SSA of fine

coke was reduced from 6.1 to 2.9 m2/g, the specific amount of pitch (pitch volume per unit of coke surface)

increased from 0.048 to 0.1 cm3/m2, as shown in Table 6.4. Another parameter which may affect the

compaction behaviour is the strength of the particle bed. Strength of a powder structure, as expressed in

Equation 6.1, depends on the number of contacts between the particles and the strength of a contact [92].

Strength of a contact is proportional to particle radius and has an inverse relationship with the square of inter-

particle distance (Equation 6.2). For a constant volume fraction of pitch, increasing the particle size has

therefore two consequences; to decrease the number of contacts and to thicken the pitch layer over the

particles. Consequently, by increasing the particle size the strength of the structure is reduced and therefore, a

lower pressure is required for the compaction.

Table 6.4: Specific pitch content for binder matrix compositions

BET surface area of

fine coke (m2/g) 2.9 4.1 6.1

P/FC ratio 30/100 34/100 38/100 34/100 38/100 42/100 34/100 38/100 42/100

Specific pitch

content (cm3/m2) 0.080 0.090 0.101 0.064 0.071 0.079 0.043 0.048 0.053

84

(6.1)

⁄ (6.2)

S: strength of the powder structure

m: number of contacts per unit volume of material

Sc: strength of a contact

r: particle radius

h: thickness of pitch layer (particle distance)

The influence of deformation rate on the compaction of binder matrix is negligible and is not significantly

affected by P/FC ratio or particle size distribution. The only exception is the composition with SSA of 2.9 m2/g

and P/FC ratio of 38/100 for which the pitch content is more than the optimum ratio of 35/100 reported by

Smith [88]. This suggests that the densification of the binder matrix in the composition range of this study

depends basically on the applied pressure and the compaction time may not have a significant contribution to

compaction behaviour.

6.5.2 Anode paste

The same experiments were conducted to record the compaction behaviour of anode pastes comprising

binder matrix and large coke aggregates. Three Blaine numbers of 2300, 4000 and 6300 were chosen for the

fine fraction of coke. The total pitch to coke (P/C) ratios of 14/100, 16.2/100, 19/100 and 22/100 were used to

make the samples. The compaction curves using two displacement rates of 1 and 10 mm/min are presented in

Figures 6.6 to 6.8.

For a given pitch quantity, decreasing the displacement rate from 10 mm/min to 1 mm/min did not result in a

better densification. Comparing Figures 6.6 to 6.8 reveals that the Blaine number of fine coke in the range of

these experiments does not change the compaction behaviour of the anode paste. The P/C ratio, however,

showed a considerable effect on densification of anode paste, with higher P/C resulting in a steeper slope of

the curves and higher final density. The possible reasons were explained in section 6.4.1.

85

Figure 6.6: Compaction curves for the paste samples made from a fine coke with SSA of 2.9 m2/g (BN 2300) and P/C ratios of 14/100, 16.2/100, 19/100 and 22/100


86


Some compaction curves are presented in Figures 6.9 to 6.12 to show the influence of particle size distribution

(BN) and SSA of fine coke for the same P/C ratio. It can be seen that the effect of particle size distribution of

fine coke on the compaction behaviour of the material is not evident. Higher BN of fine coke slightly increases

the final density.

Figure 6.9: Compaction curves for the paste samples made with a P/C ratio of 14/100 and fine cokes with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively)

87

Figure 6.10: Compaction curves for the paste samples made with a P/C ratio of 16.2/100 and fine cokes with SSA of 2.9, 4.1 and 6.1 m2/g (BN 2300, 4000 and 6300, respectively)


88


An extremely low displacement rate of 0.1 mm/min was also applied for two samples. As shown in Figures

6.13 and 6.14, this low deformation rate resulted in a slight improvement of compacted density. This

improvement, although negligible, was the motivation to investigate the compaction behaviour and to reveal

whether the anode paste continues to flow under a constant pressure. Densification of the anode pastes was

performed under a constant pressure of 10 MPa (confined creep tests). The results are presented in Figures

6.13 and 6.14 where the vertical lines show the densification under constant pressure during one hour.

Densification at constant pressure is also plotted as a function of time in Figure 6.15. It can be seen that the

densification occurs essentially during the first 2 minutes and then the curves flatten. After 1 hour of creep test

the maximum increase in relative density was 0.019.

Figure 6.13: Compaction curves for the paste samples made from a fine coke with SSA of 4.1 m2/g (BN 4000) and P/C ratios of 16.2/100 and 22/100

89

Figure 6.14: Compaction curves for the paste samples made from a fine coke with SSA of 6.1 m2/g (BN 6300) and P/C ratios of 16.2/100 and 22/100

Figure 6.15: Densification of the paste samples at 10 MPa as a function of time

The results presented in Figures 6.13 and 6.14 suggest that the paste flows under a constant pressure.

However, the deformation derived from this flow is negligible compared to that from instantaneous

deformation. Time dependency of compaction behaviour did not change even with increasing the P/C ratio to

more than its optimum value (16.2/100). In the industry an optimum P/C ratio of 13/100-15/100 is used to

obtain a maximum baked apparent density. Considering the fact that deformation rates in an industrial

compaction process is roughly ten times higher than the maximum rate chosen in this study, deformation rate

dependency could be observed in an industrial case. For example, compaction time may not be enough to

allow particle rearrangement. It is also possible that air entrapment occurs and influences the final density.

90

6.5.3 Influence of Blaine number vs. pitch ratio

It is worthwhile to evaluate which parameter among those varied in this study was more effective on the

compaction of binder matrix and anode paste. The influence of pitch content and Blaine number (SSA) of fine

coke on the final relative density of binder matrix and paste is summarized in Table 6.5. It should be stated that

these relative densities were calculated using the geometrical density of the samples when they were taken

out of the mould and expansion had occurred. When the composition is changed a new material with different

intrinsic characteristics is formed which affects the compaction process. At a constant P/FC ratio, decreasing

the particle size of fine coke resulted in lower relative density of binder matrix since the increased surface area

led to under-pitching. Similar results have been reported in the literature for the green apparent density of

compacted binder matrix with changing the BN from 1220 to 6550 [88]. An increase in the relative density of

binder matrix was observed with changing the P/FC ratio from 34/100 to 38/100. This influence of P/FC ratio

was more evident with increasing the SSA where the largest improvement in the relative density (1.9%) was

observed for the sample with a SSA of 6.1 m2/g (Table 6.5). When the P/FC ratio was increased from 34/100

to 38/100, the binder matrix with a SSA 6.1 m2/g presented the lowest increase in the specific pitch content

(0.005) (Table 6.4). This reveals that binder matrix compositions with higher Blaine number (dry samples) are

more sensitive to pitch content that is in agreement with the literature [51]. In addition, this is the influence of

fineness of fine coke that leads to higher relative density. While using an optimum amount of pitch, increasing

the fineness of fine coke can contribute to improving the relative density of binder matrix.

Table 6.5: Relative density of binder matrix and paste compositions compacted at 60 MPa at a DR of 10 mm/min

Sample P/C ratio

Relative density (%)

SSA 2.9 m2/g

(BN 2300)

SSA 4.1 m2/g

(BN 4000)

SSA 6.1 m2/g

(BN 6300)

Binder matrix 34/100 90.88 88.46 87.35

38/100 92.03 90.03 89.24

Paste

14/100 78.93 78.97 79.48

16.2/100 79.31 79.62 79.85

19/100 81.90 82.12 82.15

22/100 83.61 84.18 84.24

For anode pastes with a constant P/C ratio, using finer particles (BN 6300) did not result in a meaningful

change in relative density (Table 6.5). This is in agreement with a previous work [82] where higher BN resulted

in an insignificant increase in the green apparent density of compacted paste. However, it has been reported

that it may improve the baked apparent density and thus the anode properties [82]. The better anode properties

were explained by improved void filling effect of finer particles. When the BN of fine coke is increased, a

91

considerable improvement in the green density can be achieved if under-pitching is avoided by using higher

pitch content [85]. Table 6.5 clearly shows that for the range of parameters studied in this work, the effect of

pitch content on the relative density of anode samples is more significant than that of the fineness of the fine

fraction.

6.6 Summary

The uniaxial compaction characteristics of binder matrix and anode paste revealed useful information about

their behaviour during densification. It has been shown that the displacement rate in the range studied, which

is lower than that in the industry, has little effect on the achieved density of both binder matrix and anode

paste. Creep (time-dependent component of densification) contributes to less than 2% of the relative density.

Pitch-to-coke ratio was found to be the most important parameter determining the compaction behaviour of the

mixtures. Presence of coarser coke particles in the binder matrix (lower SSA) results in its better densification.

This unexpected behaviour is attributed to the specific pitch content which decreases by increasing the specific

surface area of the coke with higher Blaine numbers. At a constant pitch ratio, the effect of BN on densification

of the anode paste is, however, mitigated.

7 Chapter 7: Influence of shape and density of coke

particles on the compaction behaviour of paste

7.1 Résumé

L'influence des propriétés des particules de coke (facteurs de forme et densité) sur le comportement de la

pâte d'anode vis-à-vis du compactage et de la densité finale compactée a été étudiée. Cinq cokes éponges et

un shot coke provenant de différentes sources industrielles ont été utilisés pour fabriquer des échantillons

d’anode. Les facteurs de forme et la densité apparente des différentes fractions de particules ont été mesurés.

De plus, la fraction volumique de pores ouverts et fermés des particules a été déterminée. La densité

apparente d’un lit de particules a été mesurée avec et sans vibrations. La fraction volumique de pores entre

des particules a été calculé après vibration afin d’étudier les propriétés de compaction. Des pâtes d’anode à

l’échelle laboratoire ont été réalisées et compactées. Les résultats ont révélé que la forme des particules est

un facteur déterminant dans la détermination des propriétés de compaction et de densité apparente

compactée des anodes vertes.

7.2 Abstract

The influence of coke particle properties including shape factors and density on the particle packing,

compaction behaviour of anode paste and the final density of green anode was studied. Five sponge cokes

and one shot coke from different industrial sources were used to make the samples. The shape factors and

apparent density for different size fractions of particles were measured and the volume fraction of open and

closed pores within the particles was determined. The bulk density of the particle bed was measured with and

without applying vibration. Interparticle void fraction was then calculated to determine the packing behaviour of

the particles. The pastes were made and compacted to make laboratory-scale anodes. It has been revealed

that particle shape is an important factor determining the packing properties and compacted density.

94

7.3 Introduction

Coke aggregates rearrange in the binder media during the compaction process to reduce the voids. Density of

green anode depends on the compaction behaviour of the paste and compaction behaviour in turn depends on

the properties and proportion of the materials and process parameters. This part of project aimed to study the

influence of shape factors of coke particles on the compaction behaviour of anode paste.

Size distribution and internal porosity of coke particles are not the only factors affecting the bulk density [76].

Shape and roughness of coke aggregates have an influence on particle flow and packing, and consequently

on the anode density. Coke particles with more irregular shapes and higher roughness have higher inter-

particle friction and inferior flowing and packing properties [76, 100]. Therefore, cokes with lower internal porosity

will not necessarily result in denser anodes, instead, both internal porosity and shape of particles will influence

the anode density. Belitskus [79] confirmed that spherical coke particles pack better than plate-like or needle-

like ones, thus even particles with more porosity may result in a higher vibrated bulk density. Sverdlin et al. [65]

reported that irregular shape and rough surface of coke particles up to 12 mm prevented the movement in a

viscous medium. Edwards et al. [43] demonstrated that anode density increased with addition of 20% of

spherical shot coke to anode paste. Shot coke is an isotropic coke resembling small balls (< 5 mm) and is

produced through a different coking process than the traditional sponge ones [41]. Great improvement in baked

apparent density was explained by higher apparent density and better packing properties of the spherical coke

particles. The influence of particle shape and surface characteristics has been extensively studied for different

particulate materials. In general, inter-particle friction increases with increasing the angularity, surface area

and surface roughness of particles [75, 101, 102]. When the interparticle friction increases, the particle bridging

increases resulting in a higher resistance to rearrangement and a lower packing fraction of a powder [75, 101, 103,

104].

Dependence of anode properties on the paste formulation and anode making parameters has been the

principal subject of anode making literature. Little has been published on the influence of particle

characteristics, more specifically the particle shape. In this work particle density and shape factors were first

determined. They were then associated with the vibrated bulk density of particles, compaction behaviour of

anode paste and density of green anodes.


Five calcined sponge cokes (A-E) and a calcined shot coke were used with a coal tar pitch to make laboratory-

scale anode samples. Table 3.1 shows the chemical composition and real density of the cokes. Table 3.6

presents the size fractions of coke and their proportions used in the anode paste formulation. Five particle

shape factors that may be associated with particle packing were measured for every single fraction of the

95

cokes. The shape factors included aspect ratio, sphericity, roundness, compactness and convexity. The

measurement technique is explained in chapter 3 and definition of the shape factors are given in Table 3.5.

Spherical particles reveal high sphericity and roundness while angular particles have low convexity and

compactness values.

Apparent density of particles was determined by the analysis of optical microscopic images, as described in

chapter 3. Apparent density is the ratio of the mass of a particle over its external volume, which includes the

internal pores. For each fraction of coke the pycnometer density of particles with a batch weight of 40-55 g was

measured by a He-pycnometer. Volume fractions of open and closed pores within the particles were calculated

using Equations 3.3 and 3.4, respectively. Scott density and vibrated bulk density of single fractions of the

coke particles were measured by a Scott volumeter and a vibrating table (chapter 3). Void fraction between the

vibrated particles was calculated according to Equation 7.1.

(7.1)

Angle of repose was also measured for each coke fraction according to ASTM C1444 standard method. The

friction between particles is an indication of particle shape and their surface roughness.

A multivariate analysis was performed using Principal Component Analysis (PCA) method to determine the

correlations between the shape factors of particles. PCA is a latent variable method with a wide range of

applications and serves to reduce several variables into fewer numbers of latent variables or scores capturing

most of the information contained in the original variables [134]. Latent variables are linear combination

(regression) of the input variables as expressed in Equation 7.2.

(7.2)

where X is the matrix of input variables, P is the matrix of regression coefficients or loadings, T is the matrix of

scores and E is the matrix of residuals. The loadings define the relative importance of each variable to the

scores and are selected in such a way to describe the maximum variation in the variables.

Projection to Latent Structures (PLS) was implemented to predict the VBD from particle characteristics. PLS is

another latent variable method, also known as partial least squares, that serves to predict the matrix of output

parameters (Y) from the matrix of input variables (X) [134]. A score matrix for input variables (T) and a score

matrix for output results (U) are defined with PCA. Then PLS tries to maximize the correlations between the T

and U matrices using the appropriate loadings. Apparent density and particle shape factors were used as input

variables in the PLS model to predict the VBD.

Two series of anode paste samples were made. For the first series, the size fractions of each sponge coke

(Table 3.6) were mixed with pitch at a pitch to coke ratio of 16.2/100. For the second series of samples, the -

96

8+16 mesh size fraction of sponge cokes was replaced with the same size range of shot coke. Pitch to coke

ratio and mixing parameters similar to the first series of pastes were applied. The total mass of coke and pitch

was 488 g for all samples. The anode pastes were compacted at 150 C and 60 MPa with a constant

displacement rate of 10 mm/min and compaction curves were plotted.


7.5.1 Coke particle characteristics

Figure 7.1 shows the histograms representing the shape factors for different size fractions of the cokes. As

expected, shot coke particles were considerably rounder and more spherical than sponge cokes. Coke A with

the highest aspect ratio and the lowest sphericity, roundness and compactness revealed the most elongated

and irregular particles among the sponge cokes. On the other hand, coke B had noticeably the most spherical

particles and the highest compactness values just behind shot coke. Roundness and compactness of cokes C

and E were superior to those of coke A but inferior to those of coke B.

PCA with two latent variables showed the correlations between the shape factors. First latent variable

explained 67.5% and the second latent variable explained 29.5% of the variations in the particle shape factors

(cumulative R2=0.97). Figure 7.2 shows the loading plot for the latent variables. This Figure shows that

sphericity and roundness are positively correlated with convexity and compactness of particles, respectively.

Aspect ratio is negatively correlated with roundness and compactness.

Figure 7.3 shows the results for apparent density of the particles. As expected, the pores within the particles

were annihilated during size reduction and thus apparent density increased with reducing the particle size.

Although shot coke has the lowest real density, its apparent density is the highest suggesting very low porosity

of this type of coke as illustrated in Figures 7.4 and 7.5.

97

Figure 7.1: Shape factors for different size fractions of the cokes; (a) aspect ratio; (b) sphericity; (c) roundness; (d) convexity; (e) compactness

98

Figure 7.2: Loading plot for the PCA model on the particle shape factors

Figure 7.3: Apparent density for different coke fractions

99

Figure 7.4: Optical microscopic images (25x) for -8+16 mesh size fraction of the 6 cokes used for making the anode samples; (a) coke A; (b) coke B; (c) coke C; (d) coke D; (e) coke E; (f) shot coke

Figure 7.5: The fraction of (a) open and (b) closed pores in various size fractions of the cokes

Figure 7.6 shows the angle of repose for different size fractions of the cokes. The angle of repose experienced

a drastic increase with reducing the particle size from 16 to 30 mesh size. This is due to the fact that the

number of particles and therefore the number of contacts and inter-particle friction increase for smaller

particles. For particles smaller than 30 mesh size the angle of repose did not show an increasing trend with

size reduction. This is in agreement with convexity of particles in Figure 7.1-d that is an indication of particle

roughness and has a strong correlation with angle of repose.

a b c

d e

f

100

Figure 7.6: Angle of repose for different coke fractions

Figure 7.7 demonstrates the histograms of Scott density and VBD for different size fractions. Scott density of

loose particles showed a maximum with decreasing the particle size (Figure 7.7-.a). This is due to the fact that

the apparent density of particles increases with size reduction, as indicated in Figure 7.3, enhancing the Scott

density. On the other hand, two factors act against the particle density. The first factor is the inter-particle

friction that increases with reducing the particle size, as shown in Figure 7.6, resulting in a higher volume of

voids between the particles [75, 101]. The second factor is the fact that the particles in a narrow single size

fraction cannot fill the inter-particle voids and thus inter-particle porosity increases with size reduction [75]. For

particles with a mesh size of less than 30, there is a slight increase in particle density. Thus, the increase in

inter-particle space becomes the dominant parameter that masks the effect of particle apparent density. So,

the Scott density begins to decrease. However, VBD revealed an increasing trend with reducing the particle

size down to 50 mesh (Figure 7.7-b). This suggests that vibrating the bed of loose particles breaks the weak

particle-particle bridges and enhances the particle movement and packing factor.

PLS was employed to predict the VBD from the shape factors and apparent density of particles. Two latent

variables explained and predicted 76.7% of the variance in VBD (cumulative R2=0.767). The R2 value shows

that part of variance in VBD is left unexplained since some significant variables have not been measured and

considered in the model. Convexity that is used as a variable in the model shows the low frequency and widely

spaced irregularities or waviness of the particle surface (Figure 7.8). But roughness that is the high frequency

and finely spaced surface irregularities is not measured and considered. Roughness determines the interaction

of a particle with its environment as well as the friction coefficient and can help to explain more variations of

VBD.

101

Figure 7.7: Scott density (a); and vibrated bulk density (b) for different size fractions of the cokes

Figure 7.8: Schematic illustration of roughness and waviness (convexity) of a surface

Figure 7.9 shows the loading plot for the latent variables and once again reveals the correlation between the

shape factors observed in Figure 7.2. This Figure also demonstrates the positive correlation between the

particle apparent density and VBD. The predicted VBD shows a good agreement with the observed VBD as

plotted in Figure 7.10.

Apparent density had the largest regression coefficient (0.37) to predict the VBD as illustrated in Figure 7.11.

This revealed that apparent density is an important parameter that worth to be measured in anode making

plants. Sphericity and convexity with 0.28 and 0.25 had the largest coefficients after the apparent density.

Compactness, aspect ratio and roundness had very small coefficients with large confidence intervals that

showed a low repeatability for these variables. This may be due to the fact that multivariate analysis was

performed on a small number of samples (20) that reduced the precision of the model. In addition, it may be

the effect of the errors in the measured shape factors. Figure 7.12 shows the variable importance plot (VIP)

that illustrates how important a variable is to the model. This plot again demonstrates that for this dataset the

most important parameters are apparent density, convexity and sphericity to predict the VBD.

Waviness

Height

Roughness

Width

Roughness

Height

Waviness

Width

102

Figure 7.9: Loading plot for the PLS model to predict VBD from particle shape and density

Figure 7.10: Observed vs. predicted VBD derived from the PLS model (R2=0.767)

Figure 7.11: Coefficients plot for the variables to predict the VBD

103

Figure 7.12: Variable importance plot for the variables to predict the VBD

The ratio of VBD to Scott density is a characteristic of inter-particle friction and is known as Hausner ratio [101,

135]. This ratio indicates the capability of particles rearrangement during vibration of a loose particle bed.

Irregular particles with higher inter-particle friction have a low Scott density but vibration or tapping can make a

significant improvement in the Scott density [101]. Thus, it is expected that particles with irregular shapes and

high inter-particle friction reveal a higher Hausner ratio while more spherical particles have a Hausner ratio

slightly above unity. Figure 7.13 shows that -50+100 and -100+200 fractions which revealed high inter-particle

friction and therefore low Scott density have a higher Hausner ratio than the larger fractions. Hausner ratio

shows the friction and agreed well with the angle of repose of the particles. For a given particle size, cokes

with a lower angle of repose revealed a lower Hausner ratio. Shot coke with considerably more spherical

particles and the lowest angle of repose resulted in the maximum Scott density and VBD of the particle bed but

the lowest Hausner ratio (1.07). In contrast, Hausner ratio did not agree well with particle shape factors. It was

expected that coke A with the most elongated and irregular particles would result in the highest Hausner ratio

but this is not seen in Figure 7.13. This may be due to the amplitude of the vibration procedure applied to the

particles. Coke A particles had the lowest compactness, i.e., the most irregularity among the cokes and

therefore vibration with the amplitude used in this work was not able to liberate the particles and break the

bridges in the particle bed. It can be assumed that friction determines the flow behavior of freely moving

particles but shape factors are more attributed to the particle rearrangement and bridge formation in a bed of

particles in a confined volume. Therefore, the influence of particle shape cannot be clearly seen in Scott

density while it can be better revealed in VBD and compaction results.

104

Figure 7.13: Hausner ratio for different size fractions of the cokes

Void fraction between the particles after VBD test reveals the influence of particle shape and roughness on the

packing behaviour. This parameter is a measure of packing ability of particles and is shown in Figure 7.14. For

all size fractions of coke illustrated in Figure 7.14, cokes B, C and E revealed a better packing ability and thus

a lower inter-particle void fraction than did cokes A and D. This is in agreement with the shape factors in

Figure 7.1. For example, it was expected that coke A with low sphericity, roundness and compactness

(irregular particles) and high aspect ratio would restrict particle movement and rearrangement resulting in a

high volume of voids after vibration.

VBD in Figure 7.7-b was improved for smaller particles which is a simultaneous effect of the changes in

particle apparent density and inter-particle voids. Void fraction for cokes A and D showed an increasing trend

with size reduction. The number of contacts and thus the interaction between the particles increase for smaller

particles. Cokes A and D have irregular particles that amplified the particle interaction and resulted in a higher

volume of inter-particle voids. Therefore, higher apparent density for smaller particles (Figure 7.3) could not

significantly contribute to the VBD (Figure 7.7-b). For cokes C and E that have more regular shape and

spherical particles with a better packing ability, the void fraction did not change significantly with particle size.

Instead, it was even reduced towards -30+50 fraction. Lower void fraction along with higher apparent density

led to a continuous increase in VBD towards -30+50 fraction for cokes C and E (Figure 7.7-b).

105

Figure 7.14: Percentage of inter-particle voids after VBD test

An interesting point was a disagreement between the VBD results and inter-particle voids (Figures 7.7-b and

7.13). VBD results did not follow the void fraction trend and revealed that VBD test was not a reliable method

for estimating the packing ability of coke particles. VBD may be achieved by dense particles while they do not

necessarily result in a packed bed of particles since packing properties basically depend on particle shape

factors. It was shown that different cokes with equal VBD values might present different void fractions. It is

thus suggested to consider complementary parameters such as shape factors and particle porosity along with

the VBD.

7.5.2 Particle shape and compaction behaviour of the paste

Figure 7.15 shows the compaction curves of anodes made from different sources of sponge coke. There was a

significant difference between the curves and the green density of the anodes. Since the samples had similar

formulation and underwent similar compaction procedures, the difference can be due to the different shape

factors and pore fractions within the particles. In this part, the compacted green densities obtained with

different cokes are compared and the particle characteristics, reported earlier in Figures 1-7, are taken into

account to explain the difference. However it is assumed that the effect of particle shape factors on the

compaction process is greater than that of inter-particle friction because pitch acts as a lubricant that reduces

the friction.

106

Figure 7.15: Compaction curves for anode samples made from sponge cokes with different apparent density and shape factors

Particles of coke E were rather more porous than coke C (Figure 7.5) that resulted in slightly lower apparent

density (Figure 7.3). However, coke E revealed a somewhat higher Scott density and VBD than coke C. Coke

E also revealed an improved compaction behaviour and higher green density. This unexpected result may be

attributed to superior shape factors of coke E, i.e., higher sphericity, roundness and compactness that

enhanced the flowing ability and packing properties of the particles under vibration. In addition, it resulted in a

better compaction behaviour of paste in such a way that it compensated the influence of lower density of

particles. The influence of particle shape on compaction properties of other particulate materials has been

previously reported by several authors [103, 132, 136].

Apparent density of coke A was lower than that of coke E for the two larger fractions that account for 32 wt.%

of coke in the paste. For -16+30 mesh size fraction, both cokes A and E revealed similar apparent density but

for the -30+50 mesh size fraction, coke A had a higher density (Figure 7.3). This suggests that coke A had a

higher fraction of large pores than did coke E and therefore coke A underwent a larger increase in the

apparent density with reducing the particle size. Coke A had more elongated and irregular particles than coke

E (Figure 7.1) resulting in a higher void fraction. Although coke A may have a similar or superior particle

density to coke E for more than 60 wt.% of the coke mixture, it revealed a considerably lower Scott density,

VDB (Figure 7.7) and compacted density (Figure 7.15). The enhanced packing properties and compaction

behaviour of coke E can be associated with the superior shape factors for this coke.

Comparison between cokes A and C may lead to a similar conclusion. Coke A had a lower apparent density

than coke C for the two larger fractions but then surpassed the apparent density of coke C for smaller size

107

fractions (Figure 7.3). In addition, coke A had more elongated and irregular particles than coke C (Figure 7.1).

Although coke A had particles with lower density for only -4+8 and -8+16 mesh sizes (32 wt.% of coke

mixture), it resulted in a meaningfully higher void fraction, lower Scott density and VBD for all size fractions,

and a lower compacted density. It can be due to the fact that coke C has more spherical and less irregular

particles that improved the packing of the particles and also flow and compaction properties of the paste.

Coke B has generally denser particles than coke A (except for -16+30 mesh). Coke B particles are also

significantly more spherical and have more regular shapes. Both factors resulted in lower inter-particle voids,

considerably higher Scott density and VBD for coke B particles. However coke B exhibited a similar

compaction curve and green density to coke A while a higher compacted density was expected. This can be

explained when the fraction of open pores within the particles is considered. Coke A has a larger fraction of

open pores (Figure 7.5) that increases the possible penetration of binder into the coke porosity [43]. The

impregnated pores contribute to higher density of particles and consequently that of green compact. Thus,

coke A could compensate the effect of lower density and inferior particle shape factors and achieved a green

density as high as that for coke B.

Coke B had a lower apparent density than coke C for at least 43 wt.% of the coke mixture. On the contrary,

coke B had superior shape factors. The Scott density and VBD of both are similar but coke C exhibited a

higher green density. It shows that in this case the influence of apparent density of particles has been more

important than that of the particle shape.

Coke D had an apparent density equal to or lower than coke B. Sphericity, roundness and compactness of

coke D were also inferior to those of coke B. Superior density and shape factors for coke B contributed to

higher Scott density, VBD and green density, as expected.

There was also a disagreement between the VBD results and green density in some cases. For example, coke

B revealed a meaningful higher Scott density and VBD than did coke A, but a similar green density was

achieved for anode samples made out of these cokes. In another case cokes B and C revealed similar Scott

density and VBD but a higher green density was obtained for coke C. This indicates that VBD test for coke

particles is not always a good indicator to determine the compacted density.

Figure 7.16 shows the compaction curves for anode samples when the -8+16 mesh size fraction of sponge

cokes was replaced with the same size fraction of shot coke. The anode samples containing 10 wt.% of shot

coke revealed an improved compaction behaviour and green density comparing to the samples made with only

sponge coke. Apparent density of shot coke (-8+16 mesh) was at least 9% above that of the sponge cokes

used in this project. In addition, shot coke particles were obviously more spherical than the sponge cokes as

evidenced in Figure 7.4. A combination of both factors resulted in the maximum Scott density and VBD for shot

coke particles. The enhancing effect of particle shape of shot coke on the compacted density of anodes was

108

expected and reported in another work [43]. The influence of particle shape and density was so great that low

real density of shot coke did not deteriorate the green density.

Figure 7.16: Compaction curves for anode samples made from a mixture of shot coke and sponge cokes; (a) coke A + shot coke; (b) coke B + shot coke; (c) coke C + shot coke; (d) coke D + shot coke; (e) coke E + shot coke

109

7.6 Summary

Studying coke particle characteristics including the shape and porosity and their influence on the packing and

compaction behaviour revealed useful information. Multivariate analysis showed that particle porosity,

sphericity and convexity are the most important parameters contributing to the vibrated bulk density of a bed of

coke particles. Although particle density was a determining factor for the compacted density of anode samples,

particle shape was so important that in some cases to some extent compensated the negative effect of particle

porosity. Particle shape determines the rearrangement and packing behaviour of particles during compaction.

This study suggests that particle shape and porosity are the material properties that should be considered

along with the real density and vibrated bulk density to determine the compaction behaviour and final density

of anode.

8 Chapter 8: General discussion and conclusions

8.1 Introduction

This project was a part of a research program for improvement of energy efficiency in Hall-Héroult process

through improvement of anode production and quality. The present thesis was mainly focused on the influence

of raw materials properties on the compaction behaviour of anode paste and on the density and properties of

green compacted anodes. This chapter presents an overview of the work taken place within the framework

of this project and the original aspects of the work are highlighted. In addition, the important conclusions

drawn from the work are outlined. Finally, some suggestions for the refinement of the present work for

future research are provided.

8.2 Overview

The objective of this project was to reduce the impact of raw materials variations on anode quality and

consistency. The problem of anode consistency arises from the large number of material-related factors and

process parameters influencing the anode quality and thus they should be controlled. Materials properties are

not independent from each other and a series of properties may change with changing the source of materials.

Various process parameters should therefore be re-adjusted to compensate the influence of variations and to

keep the anode quality consistent.

This project first aimed to reduce the number of parameters to be controlled. It was proposed to use a paste

quality index that would be sensitive to and representative of the materials properties in order to design the

process parameters according to this index rather than to a large number of material parameters. Compaction

behaviour of paste was introduced as a paste quality index. Once the materials properties are changed, the

compaction curve of the resulted paste will be changed. To keep the quality consistent, the process

parameters in downstream production steps can be adjusted according to the compaction behaviour of paste

regardless of the numerous upstream parameters. In other words, compaction behaviour is a finger print of

paste that reflects the influence of upstream variables.

Paste compaction behaviour is an intermediate quality index that can be used to estimate the quality of green

and baked anodes. After mixing and compaction, the green anodes are cooled and green density is then

measured in the laboratory. This procedure from mixing to green density takes a considerable time up to 6

hours. In contrast, compaction behaviour can be obtained shortly after mixing step, i.e., in half an hour to

provide the first quality factor to the process engineers. As the second advantage, compaction curve

112

demonstrates the whole path for density evolution, which helps researchers to understand and model the

material behaviour during compaction process.

The compaction behaviour of paste was derived from a compaction test method developed in this project. The

paste samples (488 g) were compacted in a cylindrical steel mould with an inner diameter of 68.3 mm.

Compaction was performed at 150 C to a maximum uniaxial pressure of 60 MPa and at desired displacement

rates. The height of the paste in the mould was measured after putting the punch on the paste (14.2 kPa). The

bulk density of paste was calculated and used as the initial point of the curve. Uniaxial pressure was

progressively increased to 60 MPa. Force-displacement data were recorded using the load cell and the

position transducer (LVDT) of the press. Mass, diameter and height of the samples were used to plot the

compaction curves. The height of the samples was recorded continuously during compaction and the

geometrical density of samples was calculated at each point of the force-displacement curve. The geometrical

density was divided by the theoretical density of the material to obtain the relative density. Evolution of the

relative density was then plotted as a function of applied pressure. Anode pastes made from different raw

materials and recipes were compacted to validate the test. The test was used to study the influence of

materials properties, paste formulation, and mixing parameters on green density of anodes.

8.3 Effect of mixing time

The influence of mixing time on the homogeneity of green anodes as well as the green and baked anode

properties was evaluated. The optimum mixing variables for the setup used in this project were then

determined. Homogeneous distribution of coke, pitch and porosity in anode paste is an indication of mixing

efficiency which influences the final anode properties. The homogeneity of the samples was characterised

using X-Ray computed tomography and optical microscopy.

Standard deviation of CT numbers derived from computed tomography as well as the maximum height of the

CT number profiles (Rt, Ra) revealed a decrease with extending mixing time from 6 to 10 minutes at 178 C. In

addition, the mean thickness of binder matrix was reduced from 159 m to 131 m and binder matrix was

penetrated into the intra-particle pores. These results suggest that mixing for 10 minutes was more efficient

than 6 minutes and provided a more uniform distribution of materials in the anode.

The volume of open pores in green and baked anodes was decreased with extending mixing time from 6 to 10

minutes at 178 C and resulted in enhanced anode density, as expected. In addition, size distribution of pores

in green samples reduced towards smaller size ranges and specific surface area of baked samples decreased,

which may be an indication for penetration of binder matrix into the pores. Smaller size and lower volume of

pores led to a reduction in the air permeability from 2.02 npm to 1.41 npm. Enhanced mixing effectiveness

113

may have led to a uniform and continuous pitch film over the coke particles that improved the compaction

behaviour and consequently the green and baked properties.

Mixing for more than 10 minutes, however resulted in inferior anode quality. Longer mixing times of 15 min and

20 min resulted in a lower apparent density, higher specific surface area and air permeability of the samples.

This can be attributed to two phenomena. Particle breaking may have occurred due to over mixing that

produced new surfaces and pores open to the surface. The new surfaces and pores had lower chance to be

wetted and filled by pitch, respectively, and thus contributed to inferior compaction behaviour and higher

volume of pores. Second phenomenon was the coke attrition and incorporation of new coke fines into the pitch

at longer mixing times. This increased the viscosity of pitch and deteriorated the ability of pitch penetration into

the pores.

8.4 Effect of mixing temperature

The influence of mixing temperature on the homogeneity and properties of anodes was more evident than that

of mixing time. The thickness of binder matrix decreased from 200 m to 131 m and CT results showed a

more homogeneous paste with increasing mixing temperature from 158 C to 178 C after 10 minutes of

mixing. Green apparent density also increased by 2.7% and the air permeability was improved. Pitch viscosity

decreases with increasing the temperature and enhances the flow and penetration behaviour of pitch. At 158

C and 168 C the binder was agglomerated and had a limited possibility to fill the pores. The agglomeration

was reflected in Rt and Ra values, pore size distribution and BET surface area.

Higher mixing temperature of 188 C reduced the green and baked apparent density because the pitch was

more fluid at that temperature and could penetrate into the pores more easily. Thus, the amount of pitch was

not enough to fill the voids between the particles and that was why the volume of large pores and specific

surface area increased. In addition, interactions might have been existed between mixing time and

temperature that were not considered. In other words, 10 minutes may be longer than the optimum time

required for mixing at 188 C. The reason for using mixing at 178 °C for 10 minutes as the optimum mixing

variables for making the samples was that they resulted in the best mixing effectiveness and anode properties.

8.5 Effect of Blaine number and pitch ratio

Compaction tests were conducted on binder matrix compositions with different fineness for fine cokes (BN)

and pitch to fine coke ratios. For a given BN of fine coke, the rearrangement of particles was enhanced with

increasing the pitch ratio. Thus, a lower pressure was required to obtain a given relative density and an

increase in the slope of compaction curves was observed at initial stages of compaction. Green samples with

higher apparent density were obtained with increasing pitch ratio in the range of 34/100 to 42/100. The

114

samples with finer particles, i.e., larger surface area were more sensitive to pitch ratio and the maximum

increase in green density (2.2%) occurred for the samples with a BN of 6300 (SSA: 6.1 g/cm3) with increasing

the pitch/fine coke ratio from 34/100 to 38/100.

For binder matrix samples, increasing the fineness of fine coke at a constant amount of pitch resulted in under-

pitching and the final density decreased. This can be explained by the fact that higher specific surface area of

coke will adsorb a higher quantity of pitch. Besides, finer particles increased the interaction between the

particles by increasing the number of contacts per unit volume of material and by reducing the inter-particle

distance. Subsequently, the strength of the structure was increased that required a higher pressure for

compaction.

The influence of fineness of fine coke in anode paste was not significant at constant pitch ratio and slightly

denser samples were obtained with increasing the BN from 2300 to 6300. It is expected that providing the

required amount of pitch for finer particles may avoid under-pitching and will lead to improving the green

density. This is important since there is a trend in the industry to use finer fine cokes.

For a given BN of fine coke in the anode paste, pitch to coke ratio had a significant influence on green density.

Higher pitch content formed a pitch film over the particles to reduce the required pressure for compaction and

to improve the green density. In addition, more pitch was available to fill the inter-particle voids. The increase

in green density was at least 3.4% that happened with increasing the pitch/coke ratio from 14/100 to 19/100 for

a fine coke with a BN of 6300.

8.6 Effect of compaction rate

Dependence of compaction behaviour on compaction time was evaluated for binder matrix and paste.

Deformation rates of 2.910-4 s-1 and 2.910-3 were applied to binder matrix samples. It was observed that at

high coke fractions the particle interaction governs the compaction behaviour and compaction was not time

dependent. The influence of compaction time was observed at pitch to fine coke ratios over 38/100, however

the influence was not significant.

Deformation rates of 1.810-4 and 1.810-3 were applied to anode paste samples. It was expected that paste

might show a time dependent behaviour and binder matrix (viscous phase) might flow under pressure and fill

the voids over time. However, the effect of compaction time for pitch/coke ratios from 16.2/100 to 22/100 and

the mentioned deformation rates were negligible. Deformation rate of 1.810-5 resulted in a slightly higher

green density. Creep tests at 150 C showed that the paste flowed at 10 MPa however the deformation

resulted from this flow was negligible compared to that from instantaneous deformation.

115

It was generally concluded that compaction of anode paste in the range of deformation rates used in this

project was not a time dependent process. This observation would be very useful in determination of

constitutive laws of anode paste which is an essential step for modeling of anode forming process. The share

of time-dependent parameters on densification process can therefore be eliminated with a good degree of

confidence to simplify the model. However, in an industrial case where compaction rate is roughly 10 times

higher than the maximum rate of this project, rate dependency could be observed.

8.7 Effect of shape factors

Particle shape factors and pore fraction within the coke particles have important influence on the bulk density

of a packed bed of particles as well as the green anode density. Five shape factors were measured for a shot

coke and different size fractions of five sponge cokes from different sources. Particle shape showed a great

influence on the interaction between the particles. Higher sphericity and lower irregularity of particles

enhanced the packing properties and along with the low particle porosity contributed to higher vibrated bulk

density. In addition, more spherical and regular shape particles contributed to higher green density for

compacted samples. The effect of particle shape on the green density was more obvious for shot coke with the

most spherical particles. The influence of particle shape factors was so significant that in some cases even

compensated the effect of low density of particles. In other words, low porosity particles may not necessarily

result in a higher green density.

It was revealed that VBD was not a reliable parameter to estimate the anode porosity. Void fraction in a

vibrated bed of particles revealed the packing ability and it did not follow the VBD trend. This is due to the fact

that high VBD may be achieved by low porosity particles while they do not necessarily result in a packed bed

of particles since packing properties depend on particle shape factors. In other words, two different cokes with

equal VBD values may have different packing abilities and void fractions. It is suggested to consider

complementary parameters such as shape factors and particle porosity along with the VBD.

8.8 Future work

Compaction test method developed in this project showed to be a promising method to study the influence of

materials variations and process parameters on the compaction behaviour of anode paste and green density.

It also provides a feedback to tune the subsequent production steps in order to reduce the inconsistency in the

baked anode quality. However, this method may need further refinements to enhance the sensitivity of the test

to the variables. It is suggested to employ this test method for further investigation of the following subjects in

future research.

116

Investigating the effect of intensive mixing on the paste quality and impregnated pores

Energy input is a mixing variable and it is claimed that high energy input, i.e., intensive mixing has a beneficial

effect on the final anode quality. However, the probable influence on the homogeneity of paste and

impregnation of intra-particle pores with binder matrix has not been fully understood. The influence of mixing

intensity on the mixing effectiveness, intra-particle porosity, and green and baked anode properties might be

evaluated.

Studying the interactions between mixing variables

Interactions between mixing time, temperature and intensity can be studied to determine the optimum mixing

parameters. The influence of mixing variables on the mixing effectiveness and anode quality can be evaluated.

In addition, possible interactions between the mixing variables and pitch properties can be assessed.

Studying the governing mechanisms of particle rearrangement during compaction

Rearrangement of particles during compaction influences the final density. Rearrangement itself is a function

of several parameters including rearrangement medium, particle shape and roughness. As a matter of fact,

different mechanisms contribute to rearrangement. In addition, different phenomena may happen during

compaction that will affect the compaction behaviour. For example, particle breakage that is related to particle

grindability may happen. The influence of compaction technology (pressing, vibrocompaction) as well as pitch

ratio and particle characteristics (granulometry, texture, grindability) on the rearrangement can be evaluated.

Developing a test method/model to determine the packing and compaction properties of cokes

This project revealed that vibrated bulk density of coke particles is not a reliable test to estimate the

compacted density. A comprehensive test method or a model can be developed to determine the flow and

packing characteristics of the particles and to estimate the density of compacted paste.

Studying the electrical resistivity and reactivity of pastes with similar compaction curves

This project explained that paste recipe can be modified in such a way to compensate the effect of variations

in raw materials and to obtain a similar compaction curve and green density. However, other anode properties

such as electrical resistivity and reactivity may not be the same. The influence of paste formulation

(granulometry, fineness, pitch content) on the electrical resistivity and reactivity of baked anodes derived from

the pastes with similar compaction curves might be investigated.

117

Bibliography

[1] A. Sverdlin, "Introduction to aluminum", in Handbook of Aluminum: Volume 1: Physical metallurgy and processes, vol. 1, G. E. Totten and D. S. MacKenzie, Eds., first edition, New York, Marcel Dekker, 2003, 1-31.

[2] K. Grjotheim and H. Kvands, Introduction to aluminum electrolysis, second edition, Dusseldorf, Aluminum-Verlag, 1993.

[3] I. J. Polmear, Light alloys; from traditional alloys to nanocrystals, fourth edition, Burlington, Elsevier, 2006.

[4] U.S. Geological Survey, 2012, Aluminum statistics, in Kelly, T.D., and Matos, G.R., comps., Historical statistics for mineral and material commodities in the United States: U.S. Geological Survey Data Series 140, accessed on February 2013. Available: http://pubs.usgs.gov/ds/2005/140/.

[5] U.S. Geological Survey, 2008, Copper statistics, in Kelly, T.D., and Matos, G.R., comps., Historical statistics for mineral and material commodities in the United States: U.S. Geological Survey Data Series 140, accessed on November 2008. Available: http://pubs.usgs.gov/ds/2005/140/.

[6] S. K. Das and W. Yin, "The worldwide aluminum economy: The current state of the industry," JOM, 59 (2007), 57-63.

[7] U.S. Geological Survey, 2013, Mineral Commodity Summaries, accessed on January 2013. Available: http://minerals.usgs.gov/minerals/pubs/commodity/aluminum/mcs-2013-alumi.pdf

[8] S. Das and W. Yin, "Trends in the global aluminum fabrication industry," JOM, 59 (2007), 83-87.

[9] F. Habashi, "Extractive metallurgy of aluminum", in Handbook of Aluminum: Volume 2: Alloy production and materials manufacturing, vol. 2, G. E. Totten and D. S. MacKenzie, Eds., first edition, New York, Marcel Dekker, 2003, 1-45.

[10] K. Grjotheim and C. Krohn, Aluminium electrolysis: The chemistry of the Hall-Heroult process, first edition, Aluminium-Verlag GmbH, 1977.

[11] M. M. Gasik and M. L. Gasik, "Smelting of aluminum", in Handbook of Aluminum: Volume 2: Alloy production and materials manufacturing, vol. 2, G. E. Totten and D. S. MacKenzie, Eds., first edition, New York, Marcel Dekker, 2003, 47-79.

[12] F. Habashi, Handbook of Extractive Metallurgy vol. 2, Wiley-VCH, 1997.

[13] P. A. Foster, "Phase diagram of a portion of system NA3ALF6-ALF3-AL2O3," Journal of the American Ceramic Society, 58 (1975), 288-291.

[14] B. J. Welch, M. M. Hyland, and B. J. James, "Future materials requirements for the high-energy-intensity production of aluminum," JOM, 53 (2001), 13-18.

[15] D. Picard, et al., "Room temperature long-term creep/relaxation behaviour of carbon cathode material," Materials Science and Engineering A, 496 (2008), 366-375.

[16] P. Y. Brisson, et al., "X-ray photoelectron spectroscopy study of sodium reactions in carbon cathode blocks of aluminium oxide reduction cells," Carbon, 44 (2006), 1438-1447.

[17] W. K. Fischer, et al., "Baking parameters and the resulting anode quality", (Paper presented at the TMS Annual Meeting, Denver, CO, United States, 1993), Light Metals 1993, 683-689.

[18] Z. Kuang, et al., "Effect of baking temperature and anode current density on anode carbon consumption," Metallurgical and Materials Transactions B, 27 (1996), 177-183.

http://pubs.usgs.gov/ds/2005/140/

http://pubs.usgs.gov/ds/2005/140/

http://minerals.usgs.gov/minerals/pubs/commodity/aluminum/mcs-2013-alumi.pdf

118

[19] J. Keniry, "The economics of inert anodes and wettable cathodes for aluminum reduction cells," JOM, 53 (2001), 43-47.

[20] D. R. Sadoway, "Inert anodes for the Hall-Heroult cell: The ultimate materials challenge," JOM, 53 (2001), 34-35.

[21] J. Thonstad and E. Olsen, "Cell operation and metal purity challenges for the use of inert anodes," JOM, 53 (2001), 36-38.

[22] R. Farr-Wharton, et al., "Chemical and electrochemical oxidation of heterogeneous carbon anodes," Electrochimica Acta, 25 (1980), 217-221.

[23] T. E. Jentoftsen, et al., "Correlation between anode properties and cell performance", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 2009), Light Metals 2009, 301-304.

[24] Aluminum Industry Technology Roadmap, 2003, The Aluminum Association, accessed on November 2008. Available: http://www1.eere.energy.gov/manufacturing/resources/aluminum/pdfs/al_roadmap.pdf

[25] Aluminum, Aluminum association of canada newsletter, November 2007.

[26] D. Belitskus and D. J. Danka, "A comprehensive determination of effects of calcined petroleum coke properties on aluminum reduction cell anode properties", (Paper presented at the TMS Annual Meeting, 1989), Light Metals 1989, 429-442.

[27] M. J. Chollier-Brym, et al., "Anode reactivity: effect of coke calcination level", (Paper presented at the Light Metals 2009, Warrendale, PA, United States, 2009), Light Metals 2009, 905-908.

[28] B. Samanos and C. Dreyer, "Impact of coke calcination level and anode baking temperature on anode properties", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United states, 2001), Light Metals 2001, 681-688.

[29] D. L. Belitskus, "An evaluation of relative effects of coke, formulation and baking factors on aluminum reduction cell anode performance", (Paper presented at the TMS Annual Meeting, Denver, CO, United States, 1993), Light Metals 1993, 677-681.

[30] W. K. Fischer and R. Perruchoud, "Determining Prebaked Anode Properties for Aluminum Production," Journal of Metals, 39 (1987), 43-45.

[31] C. Jonville, J. C. Thomas, and C. Dreyer, "Influence of coke source on anode performance," JOM, 47 (1995), 23-24.

[32] R. C. Perruchoud, et al., "Anode properties, cover materials and cell operation", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2001), Light Metals 2001, 695-699.

[33] R. J. Tosta M and E. M. Inzunza, "Structural evaluation of coke of petroleum and coal tar pitch for the elaboration of anodes in the industry of the aluminum", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 887-892.

[34] C. L. Senior, et al., "Characterization of reactivity of green and calcined petroleum coke with oxygen for application to combustion systems", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 597-600.

[35] T. Ross, K. Krupinski, and M. McClung, "Plant statistical evaluation of the effect of aggregate composition and pitch level on anode distortion and predicted performance", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 1998), Light Metals 1998, 637-643.

http://www1.eere.energy.gov/manufacturing/resources/aluminum/pdfs/al_roadmap.pdf

119

[36] K. Neyrey, et al., "A tool for predicting anode performance of non-traditional calcined cokes", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 607-612.

[37] M. W. Meier, W. K. Fischer, and R. C. Perruchoud, "Thermal shock of anodes - A solved problem?", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 1994), Light Metals 1994, 685-694.

[38] S. S. Djokić, B. E. Conway, and T. F. Belliveau, "Specificity of anodic processes in cyclic voltammetry to the type of carbon used in electrolysis of cryolite-alumina melts," Journal of Applied Electrochemistry, 24 (1994), 827-834.

[39] R. C. Perruchoud, M. W. Meier, and W. K. Fischer, "Coke characteristics from the refiners to the smelters", (Paper presented at the TMS Annual Meeting, Nashville, TN, United States, 2000), Light Metals 2000, 459-465.

[40] K. L. Hulse, "Impact of raw materials and formulation", in Anode manufacture: raw materials, formulation and processing parameters first edition, Sierre, R&D Carbon Ltd., 2000, 77-158.

[41] A. Guo, et al., "Investigation on shot-coke-forming propensity and controlling of coke morphology during heavy oil coking," Fuel Processing Technology, 104 (2012), 332-342.

[42] S. Eser, et al., "Carbonization of coker feedstocks and their fractions," Carbon, 24 (1986), 77-82.

[43] L. Edwards, et al., "Use of shot coke as an anode raw material", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 2009), Light Metals 2009, 985-90.

[44] C. Dreyer, "Anode reactivity, Influence of the baking process", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1989), Light Metals 1989, 595-602.

[45] R. Perruchoud and W. K. Fischer, "Granular carbon mechanical properties-Their measurment and influence on paste production", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 1992), Light Metals 1992, 695-700.

[46] C. T. Leach, D. G. Brooks, and R. E. Gehlbach, "Correlation of coke properties, anode properties, and carbon consumption", (Paper presented at the TMS Annual Meeting, Orlando, FL, United States, 1997), Light Metals 1997, 481-488.

[47] A. N. Adams and H. H. Schobert, "Characterization of the surface properties of anode raw materials", (Paper presented at the TMS Annual Meeting, Charlotte, NC, United States, 2004), Light Metals 2004, 495-498.

[48] J. Cao, A. Buckley, and A. Tomsett, "Re-examining the pitch/coke wetting and penetration test," JOM, 54 (2002), 30-33.

[49] P. Couderc, P. Hyvernat, and J. L. Lemarchand, "Correlations between ability of pitch to penetrate coke and the physical characteristics of prebaked anodes for the aluminium industry," Fuel, 65 (1986), 281-287.

[50] A. N. Adams, "Characterization of the pitch wetting and penetration behavior of petroleum coke and recycled butts in prebaked carbon anodes," Ph.D. Thesis, The Pennsylvania State University, 2004.

[51] K. L. Hulse, "Binder demand", in Anode manufacture: raw materials, formulation and processing parameters first edition, Sierre, R&D Carbon Ltd., 2000, 159-205.

[52] A. Alscher and R. Wildforster, "Performance of binder pitch with decreased QI-content in anode making - Formation, nature, properties and substitution of quinoline insolubles", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1990), Light Metals 1990, 583-589.

120

[53] R. H. Wombles and J. T. Baron, "Laboratory anode comparison of Chinese modified pitch and vacuum distilled pitch", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 2006), Light Metals 2006, 535-540.

[54] A. S. Tayanchin, et al., "Studying mesophase contents in pitches from different sources", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 623-627.

[55] R. C. Perruchoud, M. W. Meier, and W. Fischer, "Worldwide pitch quality for prebaked anodes", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 509-517.

[56] N. R. Turner, "Recent trends in binder pitches for reduction anodes," JOM, 45 (1993), 39-42.

[57] C. Tremblay, et al., "Estimation of risk of developing baldder-cancer among workers exposed to coal-tar pitch volatiles in the primary aluminum industry," American Journal of Industrial Medicine, 27 (1995), 335-348.

[58] M. Perez, et al., "Formulation, structure and properties of carbon anodes from coal-tar pitch/petroleum pitch blends", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 495-501.

[59] J. J. Fernandez and F. Alonso, "Anthracene oil synthetic pitch: A novel approach to hybrid pitches", (Paper presented at the TMS Annual Meeting, Charlotte, NC, United States, 2004), Light Metals 2004, 499-502.

[60] D. Auguie, et al., "Formation of thin mesophase layers at the interface between filler and binder in prebaked anodes. Effect of mixing on mesophase," Carbon, 19 (1981), 277-284.

[61] P. J. Rhedey, "Laboratory evaluation of a low quinoline insolubles coal-tar pitch as anode binder", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1990), Light Metals 1990, 605-608.

[62] Y. D. Kravtzova, A. S. Tayanchin, and V. K. Frizorger, "Procedure for testing infiltration property of pitches", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 629-633.

[63] M. H. Wagner, et al., "Quality assessment of binder pitches for carbon and graphite electrodes," Fuel, 67 (1988), 792-797.

[64] G. Bhatia, R. K. Aggarwal, and O. P. Bahl, "A relationship for the evaluation of coking values of coal-tar pitches from their physical characteristics," Journal of Materials Science, 22 (1987), 3847-3850.

[65] V. A. Sverdlin, G. F. Vedernikov, and V. K. Fyodorov, "Optimization of technological parameters of aluminum production pot anode block vibration forming", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 1992), Light Metals 1992, 725-730.

[66] E. R. Mchenry, J. T. Baron, and K. C. Krupinski, "Development of anode binder pitch laboratory characterization methods", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 1998), Light Metals 1998, 769-774.

[67] J.-P. Gagne, et al., "Anode butts automated visual inspection system", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 895-898.

[68] D. Belitskus, "Effect of carbon recycle materials on properties of bench scale prebaked anodes for aluminum smelting," Metallurgical and Materials Transactions B, 12 (1981), 135-139.

121

[69] T. L. Weng and V. M. Vera, "Effects of butt particle size and butt fraction on prebake anode properties", (Paper presented at the TMS Annual Meeting, Los Angeles, CA, United States, 1984), Light Metals 1984, 1005-1013.

[70] W. K. Fischer and R. C. Perruchoud, "Interdependence between properties of anode butts and quality of prebaked anodes", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1991), Light Metals 1991, 721-724.

[71] U. Suriyapraphadilok, et al., "Physical, chemical and X-ray computed tomography characterization of anode butt cores", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 617-621.

[72] U. Suriyapraphadilok, et al., "Anode butt cores: Physical characterization and reactivity measurements," JOM, 57 (2005), 35-41.

[73] P. L. Walker, et al., "Catalysis of gasification of coal-derived cokes and chars," Fuel, 62 (1983), 140-149.

[74] R. N. Narvekar, A. Sardesai, and A. B. Prasad, "Importance of granulometry in calcined petroleum coke", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 525-530.

[75] R. M. German, Particle packing characteristics, Princeton, Metal Powder Industries Federation, 1989.

[76] R. Bowers, et al., "New analytical methods to determine calcined coke porosity, shape, and size", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 875-880.

[77] C. Kato, et al., "ALUMAR coke blending facility strategy", (Paper presented at the TMS Annual Meeting, Orlando, FL, United States, 2007), Light Metals 2007, 891-894.

[78] G. Lang, R. Liu, and K. Qian, "Characteristic and development of production technology of carbon anode in China", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 929-934.

[79] D. Belitskus, "Standardization of calcined coke bulk density test", (Paper presented at the AIME Annual Meeting, Dallas, TX, United States, 1982), Light Metals 1982, 673-689.

[80] P. Stokka, "Green paste porosity as an indicator of mixing efficiency", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1997), Light Metals 1997, 565-568.

[81] S. Rorvik, L. P. Lossius, and H. A. Oye, "Classification of pores in prebake anodes using automated optical microscopy", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 531-534.

[82] T. Vidvei, T. Eidet, and M. Sorlie, "Paste granulometry and Soderberg anode properties", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 569-574.

[83] S. Rorvik, A. P. Ratvik, and T. Foosnaes, "Characterization of green anode materials by image analysis", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 2006), Light Metals 2006, 553-558.

[84] J. Chmelar, T. Foosnaes, and H. A. Oye, "Thermal dilation of green anodes during baking", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 2006), Light Metals 2006, 597-602.

[85] F. E. O. Figueiredo, et al., "Finer fines in anode formulation", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 665-668.

122

[86] M. A. Smith, et al., "Evaluation of the effect of dust granulometry on the properties of binder matrix bench scale electrodes", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1990), Light Metals 1990, 651-655.

[87] W. K. Fischer and R. Perruchoud, "Influence of coke calcining parameters on petroleum coke quality", (Paper presented at the AIME Annual Meeting, New York, NY, United States, 1985), Light Metals 1985, 811-826.

[88] M. A. Smith, "An evaluation of the binder matrix in prebaked carbon anodes used for aluminium production," Ph.D. Thesis, University of Auckland, 1991.

[89] M. Tkac, "Porosity development in composite carbon materials during heat treatment," Ph.D. Thesis, Norvegian university of science and technology, Trondheim, 2007.

[90] E. M. Gil’debrandt, V. K. Frizorger, and E. P. Vershinina, "The effect of the granulometric composition and content of a coke charge on the viscosity of pitch-coke compounds," Russian Journal of Non-Ferrous Metals, 50 (2009), 30-32.

[91] E. M. Gildebrandt, et al., "The viscosity of pitches and coke pitch compositions," Russian Journal of Non-Ferrous Metals, 49 (2008), 456-458.

[92] E. Kravtsova, E. Gil’debrandt, and V. Frizorger, "Plastic properties of pitch-coke compositions," Russian Journal of Non-Ferrous Metals, 50 (2009), 114-117.

[93] E. P. Vershinina, E. M. Gil'debrandt, and V. K. Frizorger, "Plastic properties of homogenized coke-pitch compositions," Russian Journal of Non-Ferrous Metals, 52 (2011), 205-208.

[94] K. L. Hulse, "Rheological behavior", in Anode manufacture: raw materials, formulation and processing parameters first edition, Sierre, R&D Carbon Ltd., 2000, 283-334.

[95] M. Sakai, "Dependence of viscoelastic properties of a carbon paste on grain size of coke," Carbon, 17 (1979), 145-148.

[96] M. Sakai, "Viscoelastic properties of a pitch and coke-pitch disperse system," Carbon, 17 (1979), 139-144.

[97] N. Martin, et al., "Binder/filler interfaces in anode carbon: their classification, development, and effects," Carbon, 30 (1992), 487-494.

[98] D. Hays, J. W. Patrick, and A. Walker, "Binder--filler bonding in anode carbon Influence of filler coke surface," Fuel, 62 (1983), 1145-1149.

[99] D. Hays, J. W. Patrick, and A. Walker, "SEM study of binder coke in electrode carbon," Fuel, 62 (1983), 946-952.

[100] B. Vitchus, et al., "Understanding the calcined coke VBD- porosity paradox", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 871-873.

[101] R. M. German, Powder metallurgy science, Princeton, Metal Powder Industries Federation, 1984.

[102] K. Shinohara, M. Oida, and B. Golman, "Effect of particle shape on angle of internal friction by triaxial compression test," Powder Technology, 107 (2000), 131-136.

[103] N. Vandewalle, et al., "The influence of grain shape, friction and cohesion on granular compaction dynamics," European Physical Journal E, 22 (2007), 241-248.

[104] R. M. German, Powder metallurgy and particulate materials processing, Princeton, Metal Powder Industries Federation, 2005.

[105] O. D. Neikov, D. V. Lotsko, and V. G. Gopienko, "Powder characterization and testing", in Handbook of non-ferrous metal powder, first edition, Amsterdam, Elsevier, 2009.

123

[106] R. Narvekar and A. Sardesai, "Performance of blended and unblended green cokes during calcination", (Paper presented at the TMS Annual Meeting, Charlotte, NC, United States, 2004), Light Metals 2004, 483-488.

[107] A. S. Gomes and R. M. Heilgendorff, "Carbon plant performance with blended coke", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 2005), Light Metals 2005, 659-663.

[108] S. M. Hume, et al., "Model for petroleum coke reactivity", (Paper presented at the TMS Annual Meeting, Denver, CO, United States, 1993), Light Metals 1993, 525-531.

[109] S. M. Hume, et al., "Influence of petroleum coke sulphur content on the sodium sensitivity of carbon anodes", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1993), Light Metals 1993, 535-541.

[110] S. M. Hume, R. C. Perruchoud, and W. K. Fischer, "Gas reactivity testing of raw materials and anodes," JOM, 46 (1994), 27-31.

[111] M. Sorlie, Z.-l. Kuang, and J. Thonstad, "Effect of sulphur on anode reactivity and electrolytic consumption", (Paper presented at the TMS Annual Meeting, San Francisco, CA, United States, 1994), Light Metals 1994, 659-665.

[112] L. Edwards, F. Vogt, and J. Wilson, "Coke blending at Anglesey aluminium", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2001), Light Metals 2001, 689-694.

[113] R. T. Tonti, R. D. Zabreznik, and K. Ries, "Anode performance improvement with low reactivity coke additives to the binder matrix", (Paper presented at the TMS Annual Meeting, Warrendale, PA, United States, 1991), Light Metals 1991, 635-639.

[114] M. Gendron, S. Whelan, and K. Cantin, "Coke blending and fines circuit targeting at the Alcoa Deschambault smelter", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2008), Light Metals 2008, 861-864.

[115] M. Perez, et al., "Preparation of binder pitches by blending coal-tar and petroleum pitches", (Paper presented at the TMS Annual Meeting, New Orleans, LA, United States, 2001), Light Metals 2001, 573-579.

[116] S. P. Perez, et al., "Quality improvement for anode paste used in electrolytic production of aluminium," presented at the IAS Annual Meeting, Kowloon, Hong Kong, China, 2005, 523-528.

[117] M. Tkac, T. Foosnaes, and H. A. Oye, "Effect of vacuum vibroforming on porosity development during anode baking", (Paper presented at the TMS Annual Meeting, Orlando, FL, United States, 2007), TMS Light Metals 2007, 885-890.

[118] B. Hohl, A. Pinoncely, and J. C. Thomas, "Anode paste preparation by means of a continuously operated intensive mixing cascade at Aostar Qimingxin Aluminium China", (Paper presented at the TMS Annual Meeting, Charlotte, NC, United States, 2004), Light Metals 2004, 511-515.

[119] D. Belitskus, "Effects of mixing variables and mold temperature on prebaked anode quality", (Paper presented at the TMS Annual Meeting, New York, NY, United States, 1985), Light Metals 1985, 915-924.

[120] B. Hohl and Y. L. Wang, "Experience report - Aostar Aluminium Co Ltd, China anode paste preparation by means of a continuously operated intensive mixing cascade (IMC)", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 2006), Light Metals 2006, 583-587.

[121] A. Kaiser, "New plant for anode paste conditioning and anode pressing with a hydraulic vacuum press", (Paper presented at the TMS Annual Meeting, San Diego, CA, United States, 2003), Light Metals 2003, 561-567.

124

[122] ASTM Standard B329, 2012, "Standard test method for apparent density of metal powders and compounds using the Scott volumeter", ASTM International, West Conshohocken, PA, 2012, www.astm.org.

[123] ASTM Standard D4292, 2010, "Standard test method for determination of vibrated bulk density of calcined petroleum coke", ASTM International, West Conshohocken, PA, 2010, www.astm.org.

[124] S. Lowell, et al., "Surface area from the langmuir and BET theories", in Characterization of porous solids and powders: surface area, pore size and density, first edition, Dordrecht, Springer, 2006, 58-81.

[125] S. Lowell, et al., "Mercury porosimetry: non-wetting liquid penetration", in Characterization of porous solids and powders: surface area, pore size and density, first edition, Dordrecht, Springer, 2006, 157-188.

[126] P. Clery, "Green paste density as an indicator of mixing efficiency", (Paper presented at the TMS Annual Meeting, San Antonio, TX, United States, 1998), Light Metals 1998, 625-626.

[127] J. Hsieh, Computed tomography, principles, design, artifacts and recent advances, second edition, Washington, SPIE-The International Society for Optical Engineering, 2009.

[128] A. N. Adams, et al., "The non-destructive 3-D characterization of pre-baked carbon anodes using X-ray computerized tomography", (Paper presented at the TMS Annual Meeting, Seattle, WA, United states, 2002), Light Metals 2002, 535-539.

[129] D. Picard, et al., "Characterization of a full-scale prebaked carbon anode using X-ray computerized tomography", (Paper presented at the TMS Annual Meeting, San Diego, CA, United states, 2011), Light Metals 2011, 973-978.

[130] ISO Standard 15906, 2007, "Carbonaceous materials for the production of aluminium — Baked anodes — Determination of the air permeability", International Standard, Switzerland, 2007, www.iso.org.

[131] M. Paz, J. R. Boero, and F. Milani, "Coke density and anode quality", (Paper presented at the TMS Annual Meeting, Denver, CO, United States, 1993), Light Metals 1993, 549-553.

[132] D. T. Gethin, et al., "An investigation of powder compaction processes," International Journal of Powder Metallurgy, 30 (1994), 385-398.

[133] R. Lakes, Viscoelastic Materials, Cambridge University Press, 2009.

[134] K. Dunn, (2013, 15/07/2013), Process improvement using data, Available at http://learnche.mcmaster.ca/pid/PID.pdf

[135] R. O. Grey and J. K. Beddow, "On the Hausner ratio and its relationship to some properties of metal powders," Powder Technology, 2 (1969), 323-326.

[136] M. Guden, et al., "Effects of compaction pressure and particle shape on the porosity and compression mechanical properties of sintered Ti6Al4V powder compacts for hard tissue implantation," Journal of Biomedical Materials Research Part B-Applied Biomaterials, 85B (2008), 547-555.

http://www.astm.org/

http://www.astm.org/

http://www.iso.org/

http://learnche.mcmaster.ca/pid/PID.pdf

125

Appendix A: Image processing

In this work, Matlab image processing toolbox was used to study the microstructure of green anode samples

and to measure the thickness of the binder matrix around the coke aggregates. The Image processing

approach can be summarized as follows:

Grayscale micrographs of polished green anodes were taken with the magnification of 25X (Figure

A.1-a).

Grayscale micrographs were converted to black and white (Figure A.1-b)

Two successive steps of dilation and erosion were performed and then holes within the coke regions

were filled. The obtained black and white image was visually compared with the original image and

dilation/erosion ratio was modified to get an image with the closest appearance to the original image.

All the coke particles (white regions in the image) with the equivalent diameter of less than 150 m

were considered as fine coke particles. All the fine particles were then converted to black to consider

them inside a homogeneous binder matrix. By this step a black and white micrograph of green anode

including coke aggregates (in white) and binder matrix (in black) was obtained (Figure A.1-c).

By coding in Matlab, the iso-distance lines (contours) between the aggregates were plotted. The

image was scanned from the bottom to the top at each pixel row and the distances between the coke

regions were recorded. Finally, the mean value of the binder matrix thickness of each micrograph was

calculated by averaging the obtained values of all coke distances (Figure A.1-d).

MATLAB code for converting to black and white, dilation, erosion and filling the intra-particle pores:

files = dir('*.JPG'); for k = 1:numel(files) a = imread(files(k).name); b= im2bw(a); se1= strel('disk', 3); se2= strel('disk',4); c= imdilate(b,se1); d= imfill(c,'holes'); e= imerode(d,se2);

imwrite(e, ['Step one\' files(k).name]); end

126

MATLAB code for eliminating the particles smaller than 150 m: files = dir('*.JPG'); for k = 1:numel(files) a= imread('image.jpg'); b= im2bw(a); coke= bwareaopen(b, 750); print('-dpng','-r200',files(k).name); end

MATLAB code for measuring the thickness of binder matrix: a = imread('image.jpg'); b= im2bw(a); f= bwdist(b,'euclidean'); [x,y]=size(f); for i=1:x g=f(i,:); h=double(g); k=findpeaks(h,'MINPEAKHEIGHT',20); fileID= fopen('half_thickness.txt','a'); fprintf(fileID,'%6.4f\r\n',k); fclose(fileID); end

Figure A.1: Image processing steps to determine the average thickness of binder matrix

a b

c d

3 mm

127

List of contributions

Journal papers:

Kamran Azari, Houshang Alamdari, Gholamreza Aryanpour, Donald Picard, Mario Fafard, Angelique Adams,

―Mixing variables for prebaked anodes used in aluminum production‖, Powder Technology, Vol. 235, February

2013, pp. 341-348. I performed the experimental manipulations, interpretation of data and writing the

manuscript. Assembly of experimental setup, X-ray tomography and data analysis was carried out in

collaboration with Dr. Picard. Prof. Alamdari supervised this work and the manuscript was revised by the co-

authors before submission.

Kamran Azari, Houshang Alamdari, Gholamreza Aryanpour, Donald Ziegler, Donald Picard, Mario Fafard,

―Compaction properties of carbon materials used for prebaked anodes in aluminum production plants‖, Powder

Technology, Vol. 246, September 2013, pp. 650-657. My contribution to this article was the consolidation of

samples, performing all the experiments, interpreting the data and writing the manuscript. This work was

supervised by Prof. Alamdari, Dr. Aryanpour and Prof. Fafard. The manuscript was revised by Dr. Ziegler and

Dr. Picard.

Kamran Azari, Houshang Alamdari, Donald Ziegler, Mario Fafard, ―Influence of coke particle characteristics on

the compaction properties of carbon anodes used in aluminum production‖, submitted to Powder Technology. I

performed all the experimental manipulations, analysis and writing the manuscript under supervision of Prof.

Alamdari. The co-authors revised this work.

Gholamreza Aryanpour, Houshang Alamdari, Kamran Azari, Donald Ziegler, Donald Picard, Mario Fafard,

―Analysis on the die compaction of anode paste material used in aluminum production plants‖, submitted to

Powder Technology. My contribution to this article was performing all the experiments under supervision of

Prof. Alamdari and Dr. Aryanpour and writing parts of the introduction and experimental procedure for the first

draft.

Behzad Majidi, Kamran Azari, Houshang Alamdari, Mario Fafard, Donald Ziegler, ―Simulation of vibrated bulk

density of anode-grade coke particles using discrete element method‖, submitted to Powder Technology. I

performed the morphological studies using image analysis to determine the shape factors of the coke particles.

Conference papers:

Kamran Azari, Hany Ammar, Houshang Alamdari, Donald Picard, Mario Fafard, Donald Ziegler, ―Effects of

physical properties of anode raw materials on the paste compaction behavior‖, TMS Annual Meeting, Light

Metals 2011, San Diego, CA, Minerals, Metals & Materials Soc., 27 Feb.-3 March 2011, pp. 1161-1164. I

128

performed the experiments and interpretation of data under supervision of Prof. Alamdari and Dr. Ammar. The

manuscript was revised by the co-authors.

Kamran Azari, Houshang Alamdari, Hany Ammar, Mario Fafard, Angelique Adams, Donald Ziegler, ―Influence

of mixing parameters on the density and compaction behavior of carbon anodes used in aluminum production‖,

THERMEC 2011, Quebec, Canada, Published in Advanced Materials Research, Vol. 409, pp. 17-22, August

1-5, 2011. I performed the experimental manipulations, analysis and writing the manuscript. Prof. Alamdari

supervised the work and the co-authors revised the manuscript.

François Chevarin, Houshang Alamdari, Julien Lauzon-Gauthier, Kamran Azari, Mario Fafard, Carl Duchesne,

Donald Ziegler, ―Effects of microstructural characteristics on anode reactivity‖, COM 2011, Montreal, Canada,

October 2-5, 2011. My contribution was preparing and assembly of some parts of the experimental setup.

Hicham Chaouki, Stéphane Thibodeau, Houshang Alamdari, Donald Ziegler, Kamran Azari, Mario Fafard,

―Modeling and simulation of green anode forming process‖, COM 2011, Montreal, Canada, October 2-5, 2011.

My contribution to this paper was performing the compaction tests and providing the experimental data.

Behzad Majidi, Kamran Azari, Houshang Alamdari, D.Ziegler, M.Fafard, ―Discrete element method applied to

the vibration process of coke particles‖, TMS Annual Meeting, Light Metals 2012, Orlando, FL, Minerals,

Metals & Materials Soc., March 11-15, 2012, pp. 1273-1277. I performed the microscopic studies to determine

the shape factors of the coke particles.

Kamran Azari, Houshang Alamdari, Asem Hussein, Donald Ziegler, Mario Fafard ―Characterization of packing

ability of coke particles‖. This paper is accepted for publication in the proceedings of TMS annual meeting,

Light Metals 2014. I performed the experimental manipulations, interpretation of data and writing the

manuscript under supervision of Prof. Alamdari. Asem Hussein helped me with determining the apparent

density of coke particles.

Kamran Azari, Behzad Majidi, Houshang Alamdari, Donald Ziegler, Mario Fafard, ―Characterization of

homogeneity of green anodes through X-ray tomography and image analysis‖. This article is accepted for

publication in the proceedings of TMS annual meeting, Light Metals 2014. My contribution to this article was

performing the experimental manipulations, interpretation of data and writing the manuscript under supervision

of Prof. Alamdari. Behzad Majidi helped me with analyzing the microscopical images to determine the binder

matrix thickness.

investigation of the materials and paste relationships to ... · distribution la plus homogène de...

Documents