BOF SLAG HOT-STAGE ENGINEERING TOWARDS IRON … · Besides the creative ideas put forward in every...

Dissertation presented in partial fulfilment of the requirements for the degree of Doctor of Engineering Science

Supervisors: Prof. Bart Blanpain Dr. Muxing Guo Members of the Examination Committee: Prof. Willy Sansen, Chairman Prof. Jozef Vleugels Prof. Patrick Wollants Dr. Shuigen Huang Prof. Bo Björkman (Luleå University of Technology, Sweden) Dr. Ivonne Infante Danzo (ArcelorMittal, Ghent, Belgium)

September 2017

BOF SLAG HOT-STAGE ENGINEERING TOWARDS IRON RECOVERY AND USE

AS BINDERS

Chunwei LIU

© 2017 KU Leuven, Faculty of Engineering Science

Uitgegeven in eigen beheer, Chunwei LIU, Kasteelpark Arenberg 44 bus 2450, B-3001

Heverlee (Belgium)

Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of

openbaar gemaakt worden door middel van druk, fotokopie, microfilm, elektronisch of op

welke andere wijze ook zonder voorafgaandelijke schriftelijke toestemming van de

uitgever.

All rights reserved. No part of the publication may be reproduced in any form by print,

photoprint, microfilm, electronic or any other means without written permission from the

publisher.

i

Acknowledgments

I would like to thank my promoters, Prof. Bart Blanpain and Dr. Muxing Guo, for giving

me the chance to join the HiTemp group. Bart, although you suggested to call directly your

name from the first time we met each other, you are more than a professor to me. Thank

you for your well-structured organization of our group, which has almost the most

researchers in MTM. I will never forget your attitude to life and science, since it has shaped

me in a positive way. Thank Muxing for being my daily supervisor. Without your help, it

would not be possible to finish this thesis. I will never forget your contributions to all of

my articles. I appreciate your time for the revisions you made in manner of words by words.

You have shared not only the scientific knowledge, but also the life experience that I will

keep in mind in the future. Also, I would like to thank Fang, your wife, for her hospitality

and the memorable meals she prepared when I visited you.

Thank you, Prof. Jozef Vleugels, Prof. Patrick Wollants, Prof. Bo Björkman, Dr.

Shuigen Huang and Dr. Ivonne Infante Danzo for your interests to be my jury members.

Your inspiring questions and comments have improved the thesis significantly. Jef has set

an example in terms of your erudition and work efficiency. Gentle Patrick has helped me

to polish the language of my articles. Thank Bo to travel from Luleå to join my preliminary

defense. Big thanks for your time and remarks. Special thanks to Shuigen, who acts like a

daily supervisor to me. Thank you for sharing your rich experience of XRD, EPMA and

other equipments. Your attitude to research will definitely influence me in my future work.

Thank Ivonne for the smooth collaboration in our project and your permission to publish

the data. Prof. Willy Sansen, thank you for your time and energy to chair my defense. It

is my honor to have all of you in my jury committee!

I would thank all the HiTemp members (Pyro and SREMat) for the nice environment to do

research. Besides the creative ideas put forward in every group presentation, I also enjoyed

the HiTemp weekends, social activities and parties. Thank Delia for being my first office

mate and sharing your life experience in Venezuela. Thank you, Amy, Sabrina, Vishal

and Samant to share the office and develop a friendly atmosphere. Special thanks to Amy

and Sabrina to translate the abstract of this thesis to Dutch. I appreciate Vishal for sharing

your life and political opinions, which helps me to define myself. Thank Lieven, Pavel and

Gaurav for the fruitful discussions about the project and for the business travelling

experience in Sweden. Thank Arne to organize the P & O sessions. Thank Tobias and

Remus for the XRD assistance. Of course, I would thank Yiannis, Annelies, Hiroshi,

Acknowledgments

ii

Ghania, Elise, Lesley, Lubica, Lucas, Thomas, Bart, Hoai and other group members,

with whom I exchange the ideas for research and build friendship.

Thank you, the Chinese community in MTM and Leuven. Thank Zhi for your contribution

to my very first publication. It means a lot to me. Thank Xue and Xiebin for being my

mentors in Leuven. The life and work experience you have shared with me helped me a lot.

Thank Lichun & Jiemei for picking me up when I arrived Belgium. Thank Bin & Hao,

Chen, Huayue & Minxian, Liugang & Yanyan and Zhuangzhuang & Xiue for the

delicious food, memorable travels and games. Thank Fei & Gong for organizing the kayaks

in Dinant. Thank Fei (and your families), Ling Zhang, Xuan, Yichuan, Maoxuan, Hao,

Yinan, Jian, Ling Qin and Yong for the fantastic travelling experience. Thank Pengcheng,

Xiaodong & Yaxin, Bolu, Luman, Yannan & Lingling, Jianxun, Yuanyuan & Thomas

and Kai for your company.

Thank Joris, Pieter, Joop, Gert, Paul, Iris, Tom and Rudy for the technical support. I

appreciate the help and kindness from the secretary of MTM. Thank you so much, Kevin,

Huberte, Nadine, Mia and Mieke!

Also, I am grateful to the financial support from China Scholarship Council (CSC) and

Institute for the Promotion of Innovation through Science and Technology (IWT).

Last but not least, I would like to thank my parents. Thank you for always being supportive

to my decision. Your support makes it possible for me to follow my heart. Every time when

I get confused by the life, you open up a window and let the light in. Thank my sister for

taking good care of my parents when I am absent. Thank my fiancé for sharing happy and

sentimental emotions. Your mental company surely contributes to the success of my PhD.

Chunwei

Leuven, 2017

iii

Abstract

Basic oxygen furnace (BOF) slag is a main by-product generated during converter

steelmaking. Valorization of BOF slag contributes to the sustainability of the steel industry

and alleviates the environmental burden significantly. The present thesis is dedicated to

investigate the valorization of BOF slag with respect to recover metal and apply the slag in

constructional productions that create added value. In the first part of the research, the

carbothermic reduction of BOF slag was investigated systematically. Reduction of Fe and

P containing phases was discussed. Effects of Al2O3 and SiO2 additions on the

microstructure and mineralogy associated with the reduction process were also investigated.

Formation and growth mechanisms of the extracted metallic phase were explored. The

second part of this research focuses on optimization of slag solidification to prevent slag

expansion and obtain enhanced cementitious/hydraulic properties. Influence of Al2O3,

basicity (mass ratio of CaO/SiO2), and oxygen partial pressure on the BOF slag was studied

in detail. The effect of Al2O3 addition on the mineralogical modification and crystallization

kinetics of BOF slag was studied by both water quenching and in-situ observations.

Continuous cooling transformation (CCT) and time-temperature-transformation (TTT)

diagrams were constructed. The critical cooling rate to vitrify BOF slag was, for the first

time, determined quantitatively. Crystallization sequence was clarified by integrating the

in-situ observations and post-mortem analysis with thermodynamic calculations. To reveal

the influence of basicity and oxygen partial pressure on the mineralogical and

morphological modification of BOF slag, original and SiO2 modified slags were re-melted

and solidified under argon and/or air atmosphere followed by slow cooling. Experimental

observations were then compared with the results of thermodynamic modelling to achieve

a thorough understanding. The effects of basicity and oxygen partial pressure were then

evaluated with respect to the energy consumption for the slag valorization. Finally, a SiO2

and Al2O3 modified BOF slag was water granulated at a pilot scale. The amorphous and

mineral fractions were measured quantitatively. A mathematical model was developed to

provide an insight into the crystallization behavior during the granulation. Temperature

profiles of the slag particles with different sizes were calculated with the aid of COMSOL

Multiphysics software.

The findings of this thesis suggest an important approach to optimize the

microstructure/mineralogy of the solidified slag through modifying slag chemistry and

solidification conditions. The present study provides a fundamental basis for “Zero waste”

of BOF slag, which will be achieved by innovation of the hot-stage slag engineering with

both metal recovery and slag utilization.

v

Samenvatting

Basic oxygen furnace (BOF) slak is een belangrijk bijproduct afkomstig van het proces in

de staalconvertor. Valorisatie van BOF slak draagt bij tot de duurzaamheid van de

staalindustrie en vermindert de milieubelasting aanzienlijk. Dit proefschrift is gewijd aan

het onderzoek naar de valorisatie van BOF slak door het herwinnen van metaal en het

gebruiken van de slak als constructiemateriaal, waardoor ze toegevoegde waarde krijgt. In

het eerste deel van het onderzoek werd systematisch de carbothermische reductie van BOF

slak onderzocht. De reductie van Fe- en P-houdende fasen werd besproken. De effecten

van Al2O3 en SiO2 toevoegingen op de microstructuur en mineralogie geassocieerd met het

reductieproces werden ook onderzocht, evenals de vorming en groeimechanismen van de

geëxtraheerde metaalfase. Het tweede deel van het onderzoek focuste op de optimalisatie

van de slakstolling om slakuitzetting te vermijden en om verbeterde

cementachtige/hydraulische eigenschappen te verkrijgen. De invloed van Al2O3, de

basiciteit (massaverhouding van CaO/SiO2), en de zuurstofpartieeldruk op de BOF slak

werden in detail bestudeerd. Het effect van Al2O3 toevoegingen op de mineralogische

aanpassingen van de BOF slak en de kristallisatiekinetica werden bestudeerd door middel

van afschrikking in water en in-situ observaties. Continue koelingstransformatie (CCT) en

tijd-temperatuur-transformatie (TTT) diagrammen werden opgesteld. De kritische

koelsnelheid om de BOF slak te verglazen werd, voor de eerste keer, kwantitatief bepaald.

De kristallisatiesequentie werd verklaard door de in-situ vaststellingen en post-mortem

analyses te vergelijken met thermodynamische berekeningen. Om de invloed van de

basiciteit en de zuurstofpartieeldruk op de mineralogische en morfologische wijzigingen

van de BOF slak te verklaren, werden de originele en SiO2 gemodificeerde slakken

opnieuw gesmolten en gestold onder een Ar en/of zuurstof atmosfeer, waarna deze traag

gekoeld werden. De experimentele vaststellingen werden dan vergeleken met de resultaten

van thermodynamische modellering. De effecten van de basiciteit en de

zuurstofpartieeldruk werden dan geëvalueerd rekening houdend met de energieconsumptie

voor de slak valorisatie. Tenslotte werd een BOF slak gemodificeerd met SiO2 en Al2O3

op pilootschaal gegranuleerd. De amorfe en mineraalfracties werden kwantitatief gemeten.

Een mathematisch model werd ontwikkeld om inzicht te verwerven in het

kristallisatiegedrag tijdens de granulatie. Temperatuurprofielen van de slakpartikels met

verschillende groottes werden berekend met behulp van COMSOL Multiphysics software.

De bevindingen van deze thesis zetten een aanzet tot het optimaliseren van de

microstructuur/mineralogie van de gestolde slak door de aanpassingen aan de

samenstelling van de slakken en de stollingscondities. De huidige studie biedt een

fundamentele basis voor “Zero Afval” van BOF slak door innovatie van de hot-stage

slakkentechnologie, rekening houdend met metaalherwinning en slakgebruik.

vii

Symbols and abbreviations

Symbols

∆S Change of entropy [J∙K-1∙mol-1]

Viscosity [Pa∙s]

R Gas constant [J∙mol-1∙K-1]

mi Mass of i [kg]

Q Energy [J]

Rc Critical cooling rate [°C∙s-1]

Phase fraction [%]

T Temperature [°C]

T Temperature [°C]

Tl Liquidus temperature [°C]

Ts Solidus temperature [°C]

Tn Nose temperature [°C]

α Convection coefficient [W∙m-2∙K-1]

∆H Change of enthalpy [J∙mol-1]

A Surface area [m2]

𝐸 Radiation energy per unit area [W∙m-2]

𝜎 Stefan-Boltzmann constant [W∙m-2∙K-4]

V Volume [kg∙m-3]

𝐶𝑃𝑖 Heat capacity of i [J∙g-1∙K-1]

EA Activation energy [J∙mol−1]

𝑋 Crystallized fraction [pct]

𝑘 Crystallization coefficient [s-1]

t Crystallization time [s]

𝜏 Incubation time [s]

n Avrami exponent

Abbreviations

DTA Differential Thermal Analysis

DSC Differential Scanning Calorimetry

SEM Scanning Electron Microscope

EDS Energy Dispersive Spectroscopy

XRD X-Ray Diffraction

Symbols and Abbreviations

viii

XRF X-Ray Fluorescence Spectroscopy

EPMA Electron Probe Microanalysis

WDS Wavelength Dispersive Spectroscopy

SHTT Single Hot Thermocouple Technique

DHTT Double Hot Thermocouple Technique

CSLM Confocal Laser Scanning Microscope

TTT Temperature -Time -Transformation

CCT Continuous Cooling Transformation

BOF Basic Oxygen Furnace

EAF Electric Arc furnace

C2S Dicalcium Silicate

C3S Tricalcium Silicate

C2AF Calcium Aluminoferrite

C3A Calcium Aluminite

C3MS2 Merwinite

C3MS2 Akermanite

CMS Monticellite

C2F Srebrodolskite

RO Magnesia Wustite

C2FAS Melilite

A/F Mass Ratio of Al2O3 to Fe2O3

C/S Mass ratio of CaO to SiO2

JMA Johnson-Mehl-Avrami model

ix

Contents

Acknowledgments .............................................................................................................. i

Abstract ............................................................................................................................ iii

Samenvatting .................................................................................................................... v

Symbols and abbreviations ............................................................................................ vii

Chapter 1. General introduction ..................................................................................... 1

1.1 Project background ....................................................................................................... 1

1.2 Research objectives ...................................................................................................... 3

1.3 Thesis outline ................................................................................................................ 3

Chapter 2. Literature review: BOF slag hot-stage engineering towards iron recovery

and use as binders ............................................................................................................ 7

Abstract ................................................................................................................. 7

2.1 Introduction ................................................................................................................. 7

2.2 Compositional features of BOF slags ........................................................................... 8

2.3 Added-value applications of BOF slag ......................................................................... 9

2.3.1 Internal use in steel industry ........................................................................... 9

2.3.2 Application as construction materials ........................................................... 10

2.3.3 Other applications ......................................................................................... 10

2.4 Metal recovery from BOF slag ................................................................................... 12

2.4.1 State-of-the-art of carbothermic reduction of the Fe bearing minerals of steel

slag ............................................................................................................... 12

2.4.2 Quality of the reduced metal ........................................................................ 17

2.5 Slag solidification and optimization of the microstructure and minerals of solidified

BOF slag ............................................................................................................... 18

Contents

x

2.5.1 Fundamental study of slag solidification ...................................................... 18

2.5.2 Effect of slag chemistry and cooling rate on the minerals and microstructure

of the solidified BOF slag............................................................................. 26

2.6 Conclusion and outlook .............................................................................................. 32

Chapter 3. Valorization of BOF steel slag by reduction and phase modification: metal

recovery and slag valorization ....................................................................................... 41

Abstract ............................................................................................................... 41

3.1 Introduction ............................................................................................................... 42

3.2 Experimental methods and materials .......................................................................... 43

3.2.1 Slag preparation ............................................................................................ 43

3.2.2 Experimental procedure and characterization ............................................... 44

3.3 Results and discussion ................................................................................................ 44

3.3.1 Effect of C and Al2O3 on metal recovery ..................................................... 44

3.3.2 Effect of Al2O3 on solidification microstructure during reduction ............... 52

3.3.3 Effect of combination of SiO2 and Al2O3 on solidification microstructure

during reduction ........................................................................................... 54

3.4 Conclusions ............................................................................................................... 59

Chapter 4. Effect of Al2O3 addition on mineralogical modification and crystallization

kinetics of a high basicity BOF steel slag ...................................................................... 61

Abstract ............................................................................................................... 61

4.1 Introduction ............................................................................................................... 62

4.2 Experimental methods and thermodynamic calculations ........................................... 63

4.2.1 Materials preparation .................................................................................... 63

4.2.2 Furnace experiments ..................................................................................... 64

4.2.3 In-situ CSLM observation ............................................................................ 64

4.2.4 Thermodynamic calculation ......................................................................... 65

4.2.5 Sample analysis and characterization ........................................................... 66

Contents

xi


4.3.1 Effect of Al2O3 addition on mineralogy of the quenched BOF slag ............. 66

4.3.2 Construction of CCT and TTT diagrams of the Al2O3 modified BOF slag in

air ................................................................................................................ 69

4.3.3 Crystallization sequence during continuous cooling of the slag ................... 76

4.4 Conclusions ............................................................................................................... 79

Chapter 5. Optimization of mineralogy and microstructure of solidified BOF slag

through SiO2 addition or atmosphere control during hot-stage slag treatment ....... 83

Abstract ............................................................................................................... 83

5.1 Introduction ............................................................................................................... 84

5.2 Experiments and thermodynamic modelling .............................................................. 85

5.2.1 Materials preparation and experimental procedure....................................... 85

5.2.2 Characterization of the slag samples ............................................................ 86

5.2.3 Thermodynamic modelling ........................................................................... 86


5.3.1 Influence of SiO2 addition ............................................................................ 87

5.3.2 Influence of oxygen partial pressure ............................................................. 90

5.3.3 Evaluation for lime stabilization of BOF slag with respect to energy

consumption ................................................................................................ 94

5.4 Conclusions ............................................................................................................... 97

Chapter 6. Experimental and mathematical simulation study on the granulation of a

modified BOF steel slag ............................................................................................... 101

Abstract ............................................................................................................. 101

6.1 Introduction ............................................................................................................. 102

6.2 Experimental procedure and mathematical simulation ............................................. 103

6.2.1 Granulation of a modified BOF slag at a pilot scale ................................... 103

6.2.2 Evaluation of the transition and nose temperature of the modified BOF

slag ............................................................................................................. 103

Contents

xii

6.2.3 Characterization .......................................................................................... 104

6.2.4 Fundamentals of Modelling of the Pilot Granulation ................................. 104

6.3 Results and discussion .............................................................................................. 107

6.3.1 Pilot experiment.......................................................................................... 107

6.3.2 Critical Cooling Rate to Vitrify the Modified BOF Slag: in-situ Observation

and Calculation ........................................................................................... 108

6.3.3. Mathematical simulation on the granulation of a modified BOF slag ....... 110

6.4 Conclusions ............................................................................................................. 115

Chapter 7. General conclusions and future work ...................................................... 119

7.1 General conclusions ................................................................................................. 120

7.2 Future work ............................................................................................................. 122

Curriculum vitae .......................................................................................................... 125

List of publications ....................................................................................................... 127

1

Chapter 1

General introduction

1.1 Project background

During the steelmaking process different by-products are produced (gases, dust and slag),

commercialised and/or stockpiled. One of the most important by-products is the Basic

Oxygen Furnace (BOF) slag produced at the steel shop. BOF slag is also commonly known

as LD (Linz-Donawitz) slag. BOF slag is formed during the blowing process as a

consequence of the interaction between oxygen and the dissolved impurities in the pig iron

originating from the blast furnace. Concurrently, Fe is also oxidised. MgO and CaO (e.g.

lime and dolomite) are used as fluxes for the protection of the converter refractory lining

(magnesia/doloma based bricks) and for the removal of P (CaO). Figure 1.1 shows a

schematic representation of the BOF process and slag generation. After tapping the steel,

the BOF slag is tapped and transported directly to the cooling pits or to the hot stage slag

stabilisation plant to produce BOF/LD gravel. Both products are cooled and subsequently

transported (BOF slag and BOF gravel are treated apart) to the sieving installation where

the metallic Fe is recovered.

The EUROSLAG (European Slag Association) members generate approximately 21

million tons steel slag annually in recent years [1]. BOF slag accounts for almost half of

the total volume [2,3]. Storage of the huge amount of BOF slags burdens the environment

and is an economic liability to steel industry. Recycling and valorization of slags towards

added-value applications are therefore of significant importance for the sustainability of

the steel industry.

The chemical composition of the BOF slag depends on the impurities accompanied with

the hot metal. Typically, BOF slags contain 14-29 wt% of Fe, present in iron oxides and

iron-containing minerals, 42-55 wt% CaO and 12-18 wt% SiO2 [4]. The major phases are

dicalcium silicate, calcium aluminate and wustite. “Zero waste” of BOF slag can be

achieved by both metal recovery and slag utilization. Carbothermic reduction has been

confirmed as an effective method to extract metallic Fe from the slag [5], but innovative

Chapter 1. General introduction

2

technology is needed to control metal quality by preventing phosphorus contamination

during the reduction, and to concurrently reutilize the remaining slag.

Figure 1.1. Schematic representation of the generation of BOF slag [1]

According to the chemical and mineralogical nature of BOF slag, it has been recycled in

different fields, e.g. aggregate for road construction [6], metallurgical powder for

desulfurization of steel [7], cement and concrete for preparing building materials [8–11],

P-rich fertilizers in agricultural application [12,13]. Among these possible applications,

constructional applications like hydraulic and cementitious binders appear to be more

interesting since they create higher value.

The hydraulic and cementitious properties are determined by the minerals and

microstructure of the solidified slag, which could be optimized by modifying the slag

chemistry and controlling the slag solidification process (i.e. hot-stage slag engineering).

The research on the solidification of BOF slags is therefore of fundamental importance.

Multiple studies have been made on the thermodynamics and kinetics of the slag

solidification with respect to the influence of cooling rate and slag chemistry [14-16].

Those studies, however, are focused on the slag with low basicity (mostly below 1.5).


3

Available literature is very limited on slags with high basicity (above 3, such as BOF slag)

due to its high melting temperature and rapid crystallization behavior.

As a conclusion of the aforementioned information, it is imperative to recover metal from

BOF slag and reutilize the residue slag in added-value construction applications. The

hydraulic and cementitious property of the solidified slag largely depend on the minerals

and microstructure, which is greatly influenced by the slag chemistry and cooling history.

By modifying slag chemistry and optimizing slag solidification conditions, a favorable

microstructure can be obtained for slag valorization. The present work contributes to “Zero

waste” of BOF slag, which will be achieved by innovation of the hot-stage slag engineering,

with both metal recovery and slag utilization.

1.2 Research objectives

The present PhD research has three major objectives: (A) recover iron from BOF slag with

high purity; (B) tailor slag microstructure and minerals for added-value applications such

as binders; (C) concurrent optimization of metal recovery and slag handling (i.e. recycling

of the residual oxides). To achieve these objectives, lab experiments, pilot industrial trials

and numerical simulations are designed to:

investigate carbothermic reduction of the BOF slag, focusing on metal cleanliness

control (phosphorus content in metal), which is for the added-value application of the

recovered metal.

identify the crystallization behavior/kinetics of the original and modified BOF slag for

different conditions, which is relevant to the slag solidification process optimization.

characterize the influence of the cooling history, slag chemistry, treatment atmosphere

on the end microstructure and minerals of the solidified slag for the slag products with

high added value.

develop a guideline for the hot-stage BOF slag treatment practice for the design of the

end slag product chemistry and cooling process and for the control of the operation

parameters (such as temperature, atmosphere, reductant addition) for both metal

recovery and recycling of the residue oxides, achieving “zero-waste” of BOF slag.

1.3 Thesis outline

The thesis outline is given in Figure 1.2, which contains three parts: (I) background

information of the research topic and a comprehensive review (Chapters 1 and 2), (II)

description of the experimental and simulation work and the discussions on the results


4

(Chapters 3 to 6), and (III) general conclusions and perspectives on the future work

(Chapter 7).

Chapter 1 gives a general introduction to the research background, i.e. an overview of the

generation of BOF slags during converter steelmaking, the significance of valorizing BOF

slags and its possible application routes. The objectives and the outline of the thesis are

described.

Chapter 2 presents a state-of-the-art in the field of BOF slag valorization, which includes

a comprehensive examination of the Fe recovery from BOF slag by carbothermic reduction,

and of the hot-stage slag treatment by manipulating the slag chemistry and cooling paths,

aiming at high added-value applications of the slag. The optimization of the hot-stage

engineering and the outlook for future work are discussed, providing an overview of BOF

slag valorization.

Figure 1.2. Schematic diagram of the thesis outline

Chapter 3 systematically investigates the carbothermic reduction of BOF slag with special

attention on the reduction of P-containing phases (i.e. oxides and compounds). Effects of

Al2O3 and SiO2 additions on the solidification microstructure and mineralogy associated


5

with the reduction processes are also studied. The formation and growth mechanisms of

the extracted metallic phase are proposed, and the mineralogy of the residual slag is

determined. By controlling the additions and the cooling rate, it is proposed to recover a

high-grade metal and simultaneously to utilize the remaining slag into construction

applications.

Chapter 4 discusses the effects of Al2O3 addition on the microstructure modification and

crystallization kinetics of a high basicity BOF steel slag, targeting to improve slag

cementitious properties. Continuous cooling transformation (CCT) and time-temperature-

transformation (TTT) curves have been constructed under continuous cooling and

isothermal solidification conditions, to determine the crystallization characteristic of BOF

slag. The isothermal crystallization kinetics is studied, and the critical cooling rate to vitrify

the slags is experimentally obtained. The crystallization sequence of the slag has been

clarified by integrating the in-situ and post-mortem observations with thermodynamic

calculations.

Chapter 5 studies the influence of basicity (mass ratio of CaO/SiO2) and oxygen partial

pressure on the mineralogical and morphological modification of a typical industrial BOF

slag. The slag basicity (mass ratio of CaO/SiO2) is varied by mixing specific amounts of

SiO2 with the master BOF slag. The original and modified slags are re-melted and solidified

under argon and/or air atmosphere followed by slow cooling. Experimental observations

are compared with the results of thermodynamic modelling. It is found that with decreasing

basicity free lime is eliminated through combination with SiO2 to form dicalcium silicate

(Ca2SiO4). With increasing oxygen partial pressure, wustite is oxidized to hematite, and

then combined with free lime, forming calcium aluminoferrite (C2AF). The effects of

basicity and oxygen partial pressure are finally evaluated with respect to the energy

consumption for the BOF slag valorization. The result provides a precursor basis to prepare

materials for high added-value application.

Chapter 6 presents a pilot scale granulation experiment to obtain a slag with larger

amorphous fraction to enhance the potential as binders for construction application. The

Al2O3 and SiO2 modified BOF slag is melted and water atomized at a pilot scale plasma

smelting installation. The amorphous fraction is measured quantitatively. A mathematical

model is developed to provide an insight into the crystallization behavior during the

granulation process. Temperature profiles of the slag particles with different sizes are

calculated with the aid of COMSOL Multiphysics software. The simulation is validated by

comparing the amorphous fraction obtained from experiments with that from simulation.

The results provide fundamental understanding on the key parameters of the vitrification

process, which should be considered in a practical operation for slag valorization.

In Chapter 7, the overall conclusions of the thesis are summarized. It also gives an outlook

to further work that can be conducted to improve the hot-stage slag engineering and take

the BOF slag research to next level.


6

References

[1] I. Z. Yildirim and M. Prezzi: Adv. Civ. Eng., 2011, vol. 2011, pp. 1–13.

[2] www.euroslag.com, accessed on 25/2/2017.

[3] X. Gao, M. Okubo, N. Maruoka, H. Shibata, T. Ito, and S.-Y. Kitamura: Miner.

Process. Extr. Metall., 2015, vol. 124, pp. 116–24.

[4] H. Motz and J. Geiseler: Waste Manag., 2001, vol. 21, pp. 285–93.

[5] S. R. Story, B. Sarma, R. J. Fruehan, A. W. Cramb, and G. R. Belton: Metall.

Mater. Trans. B, 1998, vol. 29, pp. 929–32.

[6] S. A. Mikhail and A. M. Turcotte: Thermochim. Acta, 1995, vol. 263, pp. 87–94.

[7] G. Z. Ye, E. Burstr, M. Kuhn, and J. Piret: Scand. J. Metall., 2003, vol. 32, pp. 7–

14.

[8] T.S. Zhang, Q.J. Yu, J.X. Wei, J.X. Li, and P.P. Zhang: Resour. Conserv. Recycl.,

2011, vol. 56, pp. 48–55.

[9] G. Wimmer, H. Wulfert, H.M. Ludwig, and A. Fleischanderl: in METEC 2nd Eur.

Steel Technol. Appl. Days, 2015, pp. 1–6.

[10] A. Srinivasa Reddy, R. K. Pradhan, and Sanjay Chandra: Int. J. Miner. Process.,

2006, vol. 79, pp. 98–105.

[11] Y. J. Xue, S. P. Wu, H. B. Hou, and J. Zha: J. Hazard. Mater., 2006, vol. 138,

pp. 261–68.

[12] K. Yokoyama, H. Kubo, K. Mori, H. Okada, S. Takeuchi, and T. Nagasaka: ISIJ

Int., 2007, vol. 47, pp. 1541–48.

[13] M. O. Suk, S. K. Jo, S. H. Kim, K. Y. Lee, and J. M. Park: Metall. Mater. Trans.

B, 2006, vol. 37, pp. 99–107.

[14] J. B. Ferreira Neto, J. O . G. Faria, C. Fredericci, F. F Chotoli, A. N. L. Silva, B.

B. Ferraro, T. R. Ribeiro, A. Malynowskyj, V. A. Quarcioni, and A. A. Lotto: J.

Sustain. Metall., 2016, vol. 2, pp. 13–27.

[15] B. Deo, J. Halder, B. Snoeijer, A. Overbosch, and R. Boom: Ironmak. Steelmak.,

2005, vol. 32, pp. 54–60.

[16] L. Kriskova, Y. Pontikes, L. Pandelaers, Ö. Cizer, P. T. Jones, K. Van Balen, and

B. Blanpain: Metall. Mater. Trans. B, 2013, vol. 44, pp. 1173–84.

7

Chapter 2

Literature review: BOF slag hot-stage engineering towards iron

recovery and use as binders

Abstract

Valorization of basic oxygen furnace (BOF) slag, as an approach to improve the

sustainability of the steel industry and alleviate the environmental impact, has received

wide attention from scientific studies to engineering applications. Up to now, BOF slag

valorization has been mainly focused on Fe recovery and on reutilizing the slag as high

added-value material. The potential to valorize the BOF slag depends crucially on the

understanding of the fundamentals of hot-stage engineering. This review provides a

comprehensive examination of Fe recovery from BOF slag by carbothermic reduction, and

of the optimized solidification by manipulating the slag chemistry and cooling paths.

2.1 Introduction

With the rapid development of the steel industry in the past decades, the volume of slag

has increased tremendously, as a main and inevitable by-product generated during iron and

steelmaking. In recent years, steel slag production is approximately 21 million tonne per

year in the EUROSLAG Association members [1], 90 million tons per year in China [2], 8

million tons per year in the USA [3] and 13.8 million tons per year in Japan [4]. BOF slag

accounts for the major part of the steel slags [1,5]. Disposal of the slag takes huge land

space and burdens the industry. Furthermore, potential leaching of heavy metals, such as

Cr and V from slags contaminates the soil, which is a threat to human’s health [6].

Alternatively, recycling of the BOF slag contributes to the sustainability of steel industry

and alleviates the environmental impact significantly.

BOF slag is generated during the steelmaking process in a converter. After the converter

processing, steel is tapped via the tapping hole and, subsequently, slag with entrapped

metallic particles/droplets is discharged. The BOF slag typically entraps 7-10 wt% metallic

Chapter 2. Literature review

8

iron [7], and therefore recovery of metallic Fe from the slag is profitable. The main

components of BOF slag are CaO, FexO, SiO2, MnO and MgO. Other elements include Cr,

Ti, V, P, S and C [8]. In the past few years, BOF slag has shown its potential to be recycled

for internal use in the steelmaking process, for road construction, cementitious substitutes,

and as agricultural fertilizer [6,9–12]. More recently, BOF slag has been successfully used

to purify waste water by extracting the phosphorus [13,14]. The utilization of BOF slag is

determined by its chemical composition and minerals, which can be manipulated by hot-

stage engineering, e.g. chemical modification and solidification control.

This literature study aims to provide a state-of-the-art of BOF slag valorization, including

metal recovery, slag solidification and heat recovery. After providing brief information on

the chemical and mineralogical features of BOF slag, the main methods to recover metallic

Fe from slags are presented. Kinetic studies of the carbothermic reduction on the steel slags

are summarized. The quality of the recovered metal by carbothermic reduction is discussed.

This is followed by a review on the recent progress of the fundamentals and optimization

of slag solidification. Crystallization kinetics of slags under isothermal and continuous

cooling conditions is discussed based on in-situ studies. Thereafter, the potential of slag

reutilization and heat recovery is presented.

2.2 Compositional features of BOF slags

The main purpose of the BOF process is to convert molten pig iron and steel scrap into

high quality steel. Thereby, huge amounts of slags are generated as a by-product [15,16].

The amount of steel slag from different steel industries is 100–200 kg/t of steel produced

depending on process conditions. The recycling of BOF slag is an important challenge for

steelmakers in order to reduce landfilling, to alleviate the pressure on natural resources (e.g.

metallurgical fluxes for steelmaking, aggregates as construction materials) and to reduce

CO2 emissions from traditional cement production by preparing blended cement and/or

inorganic polymers.

The chemical composition of typical BOF slags is shown in Table 2.1. The main

components are Fe, CaO, and SiO2, MgO and MnO. Fe is normally in the form of metallic

iron (7 to 10 wt%), iron oxide and other iron bearing minerals. Iron can be separated by

applying mineral processing technology and can be recycled in the blast furnace. The high

content of CaO in the slag can be used to substitute for a part limestone as fluxing material

to reduce the steelmaking cost. Many steel companies in Asia utilize BOF slag as a

replacement for limestone. BOF slags, however, contain a high amount of P and S, which

affects the recycling to the iron and steelmaking process, since P and S contaminate the

metal quality seriously. BOF slag is a mixture of different mineral phases. X-ray diffraction

studies (XRD) of BOF slags show that the major phases of BOF slag are dicalcium silicate

(2CaO∙SiO2, in short C2S), dicalcium aluminoferrite and wustite as reported in Table 2.2.

BOF slags contain reactive mineral phases such as C2S, tricalcium silicate (3CaO∙SiO2, in


9

short C3S), free lime (CaO) and periclase (MgO). But the mineral composition and mineral

grain size depend on the chemical composition and cooling path. For example, at higher

basicities (CaO/SiO2+P2O5 > 2.5), C3S and CaO become dominant in steel slag [7]. Slow

cooling of a typical BOF slag can produce larger phase grains than fast cooling [18].

Table 2.1. Chemical composition (wt%) of BOF slags

Region/Country Europe China USA Japan

Slag BOF slag

(low MgO)

BOF slag

(high MgO) BOF slag BOF slag BOF slag

T.Fe 14-20 15-20 17-27 12-23 17.4

CaO 45-55 42-50 45-60 29-51 45.8

SiO2 12-18 12-15 10-15 7-24 11.0

MnO <5 <5 2-6 2-8 5.3

Al2O3 <3 <3 1-5 4.5 1.9

MgO <3 5.8 2.5-10 5-12 6.5

P2O5 <2 <2 1-4 <1.3 1.7

S NR NR 0.2 <0.9 0.06

Ref. [6] [7,17] [3] [4]

NR: not reported. T.Fe: including mass percent of metallic iron and iron in the form of

oxides.

Table 2.2. Major mineral phases present in BOF slags [7]

Mineral phases wt%

Tricalcium silicate (C3S), Ca3SiO5 0-20

Dicalcium silicate (C2S), Ca2SiO4 30-60

Other silicate 0-10

Magnesiumcalcium wustite 15-30

Dicalcium aluminoferrite, Ca2(Fe,Al,Ti)O5 10-25

Magnesium type phases (Fe, Mn, Mg, Ca)O 0-5

Free lime, CaO 0-20

Periclase, MgO 0-5

Fluorite, Ca2F 0-1

2.3 Added-value applications of BOF slag

2.3.1 Internal use in steel industry

Internal use of BOF slag in steel industry includes recycling of metallic iron and oxides.

The metallic iron entrapped can be reclaimed through crushing, sorting, magnetic

separation. The recovered iron can be reused in the steelmaking process. The typical CaO


10

content of BOF slag is higher than 40 wt%, and the slag can be applied as sinter flux by

partially replacing lime [17]. The CaO-rich system can also be used as a metallurgical

powder for the dephosphorization and desulfurization of steel [19]. Das et al. [20] reported

after treatment of a hot metal by BOF slag containing flux, that the phosphorus content in

the metal decreased from 0.09-0.16 wt% to 0.02-0.06 wt%.

2.3.2 Application as construction materials

A large amount of slags has been used as construction materials, such as aggregates and

binders. After steering the slag chemistry and minerals, the strength and durability of slag

can be comparable with natural aggregates. So far, the slag has been successfully reused as

aggregates for road construction after crushing and/or screening [9,10]. In Germany, about

400,000 tons per year is used as aggregate for the stabilization of river banks and river beds

against erosion [6]. The presence of C2S and C3S minerals is at the origin of the

cementitious property of BOF slag. As a result, the slag can be applied in preparing blended

cements [21–23]. By forming a larger fraction of amorphous phase, the slag can be applied

in manufacturing inorganic polymers [24,25].

2.3.3 Other applications

BOF slag can be recycled as P-rich fertilizers by enriching the phosphorus concentration

of the slag [26,27]. The fertilizer can improve the soil quality and be used in agricultural

applications. In recent years, BOF slag is reported as effective for CO2 capture and storage

through hot-stage carbonation. Santos et al. [28,29] investigated the hot-stage carbonation

of BOF and AOD steel slag. They found the CO2 was captured and slag can be stabilized

via the reaction between CaO and CO2. In addition, by mixing with a certain amount of

sand and alkali, the BOF slag can be recycled to prepare glass-ceramics with excellent

physical and mechanical properties [30–32]. More recently, the BOF slag has been

successfully used to purify waste water by absorbing the phosphorus [13,14].

In summary, although various recycling potentials of the BOF slag are revealed, a large-

scale application of the BOF slag has been dedicated to road pavement and internal

utilization as metallurgical powder [33], as given in Figure 2.1. Less than 10 % of the

produced slags has been reutilized as hydraulic and cementitious materials due to volume

instability and poor cementitious properties. Yet, construction applications like hydraulic

and cementitious binders appear to be more interesting since they create higher added value.

The binder property is determined by the minerals and microstructure of the solidified slag,

which can be optimized by modifying the slag chemistry and controlling the slag

solidification process (i.e. hot-stage slag engineering). The understanding of BOF slag

solidification is therefore of fundamental importance. Slag properties can be steered

through slag solidification control to satisfy the requirements of the minerals and


11

microstructure for the binders [18]. Figure 2.2 shows the aspects and interrelationships in

the research field of slag engineering for recycling purposes. Both the microstructure and

final properties are closely related to the processing parameters, i.e. the control of slag

chemistry and its solidification. Thus, it is imperative to understand the influence of

processing parameters (chemical modifiers, cooling path/rate and atmosphere) on the

mineralogy and microstructure of BOF slags.

Figure 2.1. Use of Steel slag in EUROSLAG members in 2012 [1]

Figure 2.2. Aspects and interrelationships in the research field of slag engineering for

recycling purposes. Adapted from Durinck et al. [34].


12

2.4 Metal recovery from BOF slag

BOF slag contains a large amount of iron, present in metallic iron, iron oxides and iron-

containing minerals, which might be considered as a potential resource of metals in cases

of inadequate iron ore supply. Therefore, BOF steel slag is usually subjected to metal

recovery prior to its application outside the steelmaking process. The methods of the slag

processing to recover metallic iron are different, depending on the cooling method,

chemical and mineralogical composition of steel slag, and its targeted application. In

general, post-treatment includes crushing or grinding, screening and magnetic separation,

and sometimes removal of P [26,35–37]. To further recover metal from the slag, smelting

approaches, such as carbon smelting reduction [19,36,38–43] and molten slag oxidation

with O2 or CO2 gas [44–46] have been attempted. This work will present the current

progress in carbothermic reduction fundamentals to recover metal from slag since it forms

the base for pyro-metallurgical approaches.

2.4.1 State-of-the-art of carbothermic reduction of the Fe bearing minerals of steel

slag

Kinetic study aspect of the carbothermic reduction

The reduction reactions between the reductant carbon and the Fe-bearing minerals in the

slag can be expressed by the reactions (2.1) to (2.3) [47–49].

C (s) + CO2 (g) = 2 CO (g) (2.1)

(FeO)slag + CO (g) = Femetal + CO2 (g) (2.2)

(FeO)slag + C (s) = Femetal + CO (g) (2.3)

where reaction (2.3) is resulted from a combination of reactions (2.1) and (2.2). As shown

in Figure 2.3, the overall process can be divided into the following individual steps, from

a kinetic point of view:

(a) chemical reaction at the gas-carbon interface (Boudouard reaction), i.e. reaction (2.1);

(b) gas diffusion of CO from the gas-carbon interface to the gas-slag interface;

(c) solid diffusion of FeO (Fe2+ and O2- ions) from the bulk slag to the gas-slag interface;

(d) chemical reaction at the gas-slag interface, i.e. reaction (2.2);

(e) CO2 gas diffusion from the gas-slag interface to the gas-carbon interface.


13

Figure 2.3. Schematic diagram of carbothermic reduction: carbon gasification and slag

reduction, after Ref [50]

Each step may contribute resistance to the overall reduction of slags. The slowest step

controls the overall reaction, and it is named as the rate-limiting step.

Carbothermic reduction has been confirmed as an effective method to extract metallic Fe

from slag, but so far, the reduction mechanisms and underlying rate-limiting steps are not

fully understood. During past years, intensive investigations have been carried out to

determine the rate-limiting steps for optimizing the reduction process. Table 2.3

summarizes reported studies on carbothermic reduction of iron oxides bearing slag. In these

studies, industrial and/or synthetic Fe-bearing slags have been used as the starting slags.

Different carbon sources such as graphite, coke, or coal have been employed as a reductant.

In most of the studies, the Boudouard reaction and diffusion of FeO in molten slag were

confirmed to be the rate-limiting steps, but it was not the case for all the experimental

observations. Difficulty in determining the rate-limiting step is caused by several aspects

(1) gas diffusion and chemical reaction rates are closely related to gas concentrations of

CO and CO2 at the gas-carbon interface and gas-slag interface, which are difficult to be

measured; (2) the gas concentration varies with both the gas diffusion and chemical

reactions concurrently; (3) complicated slag composition leads to a difficulty of the kinetic

study, e.g. the combined influences of slag oxidation state, iron oxide content, and melt

basicity are complex. A quantitative kinetic model, therefore, has not been developed so

far due to the poor understanding of the reduction mechanism.


14

Table 2.3. Reported studies of carbothermic reduction of the iron oxides bearing slag

Slag system Carbon source Experimental

condition

Rate-limiting

step Ref.

Synthetic slags

with varied FeO

contents

Graphite, coke,

and coal char

Resistance furnace,

Ar atmosphere,

1673K

Boudouard

reaction, solid

diffusion of

FeO

[47,49]

Synthetic slags

with varied FeO

contents

Graphite, coke,

and coal char

Resistance furnace,

Ar atmosphere,

1323 and 1423 K

Boudouard

reaction [51]

Synthetic slags

with varied FeO

contents

Graphite

Resistance furnace,

Ar atmosphere,

1773 to 1873 K

Solid diffusion

of FeO [52]

Nickel-

containing

converter slags

Graphite

electrode

Resistance furnace,

N2 atmosphere,

1473 and 1723 K

Boudouard

reaction [53]

Synthetic

fayalite slags

Graphite disk

and rod

Resistance furnace,

N2 atmosphere,

1523 to 1723 K

Boudouard

reaction [54]

Synthetic cao-

feo-sio2 slags

with varied FeO

contents

Graphite disk

Resistance furnace,

Ar atmosphere,

1623 to 1693 K

Solid diffusion

of FeO [39]

Industrial

copper slag Graphite

Resistance furnace,

N2 atmosphere,

1573 K

Mass transport [55]

Synthetic CaO-

FeO-SiO2 slags

with varied FeO

contents

Solid carbon

Plasma reactor, Ar

atmosphere, 1784

to 1920 K

Solid diffusion

of FeO [56]

Iron-saturated

FeO-CaO-

Al2O3-SiO2

slags

Graphite, coke,

bituminous coal

and anthracitic

coal

Resistance furnace,

Ar atmosphere,

1673 to 1873 K

Boudouard

reaction and

mass transfer

[57]

Synthetic CaO-

FeO-SiO2-

Al2O3 slags

Graphite disc

Resistance furnace,

Ar atmosphere,

1723 K

Boudouard

reaction and

mass transfer

(<5 wt% FeO) [58]

Boudouard

reaction (>30

wt% FeO)

Magnetic field effects on the kinetics of carbothermic reduction

To facilitate the reduction of Fe oxides, the application of a magnetic field has been

attempted. Hay has conducted a pioneering study on magnetic field effects on the reduction


15

of Fe ore [59]. A furnace equipped with a magnetic coil was developed, where the

carbothermic reduction of iron ore was carried out. It was reported that the reduction was

considerably promoted by the magnetic field and fine metallic Fe particles with uniform

size were produced [59]. Although the underlying mechanism of the magnetic field effect

was wrongly attributed to the heat induced by the magnetic coil, this pioneering work

triggers intensive research in this area.

In 1970s, Skorski [60] examined the effect of AC and DC magnetic field (0 to 1.4 kOe)

effects on the reduction of hematite by H2, CH4 and CO below 333 °C. It was found that

the magnetic field effect was more significant in the case of H2 reduction than in the cases

of CH4 and CO reduction. This was considered to be caused by the magnetic properties of

H2 since orthohydrogen attracts the magnetized ore powder [60]. However, Svare [61]

challenged this interpretation and argued that the difference is insignificant between H2,

CH4 and CO with respect to the magnetic properties under Skorski’s experimental

conditions. Considering the considerable difference in magnetic property between product

and reactant species, Peters [62] proposed an alternative mechanism for Skorski’s

experimental observation. Absorption of gas reductant onto the reaction interface is

believed to be the rate-limiting step of the reduction. A smaller molecule size of H2 than

that of CH4 and CO was considered to further accelerate the reduction via faster

adsorption [62]. It was then clarified that the reduction nature of hematite → magnetite →

wustite induces uncertainties for the research. To overcome the uncertainty, Rowe et al. [63]

have systematically investigated the magnetic field effects (~4200 Oe) on magnetite,

wustite, and NiO, respectively under H2 atmosphere at 385 °C (for magnetite), 490 °C (for

wustite), 280 °C (for cobalt oxides) and 295 °C (for NiO). It was found that the reduction

of wustite was slowed down but the reduction of magnetite and cobalt oxides was

accelerated by imposing a magnetic field, and no magnetic field effect on the NiO reduction

was observed [63–66]. After critically examining the conventional gravimetric technique

used in the oxide reduction, Gallagher et al. [67] developed a more reliable method to study

the magnetic field effect by in-situ analysis of the evolved gas during the reduction. Fe2O3,

Fe3O4, Co3O4, NiO oxides were tested under H2 reduction at elevated temperatures, but no

effect of magnetic field on the reduction rate ( < 0.5 pct) was observed. Based on those

experimental results, Rowe and Gallagher [68] have confirmed that the reduction was

mainly influenced by particle movement induced by the magnetic field gradient, particle

size and temperature, etc.

Recently, reduction of iron ore by carbon powder in a static magnetic field (4-10 kOe) from

800 to 1000 °C was investigated by Jin et al. [69]The reduction rate was significantly

increased compared to that without magnetic field and the origins of increasing in the

reduction rate was attributed to the boost of gas diffusion by the external field. Kim et al.

studied the permanent magnetic field (0, 0.7 and 1.72 kOe) effects on the reduction of

magnetite at 500 °C and 750 °C under a mixture of CO and H2. The results demonstrated

that H2 accelerates the reduction rate but CO deteriorates the reduction rate due to the

formation of Fe carbide in the boundary layer. Also, reduction of magnetite powder was


16

facilitated by imposing the magnetic field, but there was no effect on the reduction of

magnetite pellets. The rate-limiting step was deduced to be the formation of the porous Fe

layer that provided the mass transport path for gas. As seen in Figure 2.4, Kim et al. [70]

have proposed a mechanism to describe the reduction kinetics. With imposing a magnetic

field, magnetite powders are aligned along the magnetic field, resulting in a more porous

structure than without the external magnetic field. In the course of reduction without

magnetic field, reduced Fe agglomerates around magnetite and hinders the diffusion of H2

to the magnetite. In contrary, the external field facilitates alignment of the reduced Fe and

leaves a space for the diffusion of H2 to the magnetite. Therefore, the overall reduction rate

of the unreduced magnetite core increases [70].

Figure 2.4. Schematic illustration of the influence of a magnetic field on the reduction of

magnetite powder [70]


17

To conclude, kinetics of the carbothermic reduction of iron oxide containing slags

significantly improves by imposing an external magnetic field. This combined magnetic

treatment with the carbothermic reduction is a promising approach to recover metal from

steel slag due to that fact that the magnetic treatment can not only facilitate the process

kinetics, but also save energy by decreasing the traditional reduction temperature. However,

this has been investigated just at the lab test stage, and the magnetic treatment process has

not been optimized. In addition, since the underlying mechanism of the magnetic field

effect has not been fully understood yet, further work is needed for developing a technically

reliable and economically efficient method to recover metal from the steel slag by using a

combined magnetic treatment.

2.4.2 Quality of the reduced metal

Since phosphorus in the slag is also reduced and dissolved into metallic iron during the

carbothermic reduction, high content phosphorus in the reduced Fe has limited its recycling

in iron and steelmaking process [41]. Due to the low oxygen partial pressure during the

ironmaking process in the blast furnace, phosphorus can hardly be removed from the hot

metal [15]. Therefore an extra dephosphorization in the steelmaking process is required,

leading to high production cost and difficulty with the final steel quality control. Against

such a background, investigation on separation of phosphorus from the slag and

purification of the recovered Fe has been carried out [14,38,71,72]. In BOF slag, it has

been revealed that phosphorus is rich in the dicalcium silicate (C2S) which is a major

phase [38,72–78]. Miki et al. [79] reported a novel method to recover FeO and P from a

steelmaking slag through capillary action. A sintered CaO particle was introduced to the

molten slag (with FeO rich liquid phase and P rich C2S solid phase), and penetration of

FeO rich liquid fraction of the slag into the sintered CaO was facilitated by the capillary

effect. Therefore, the solid C2S phase and liquid FeO rich phase could be effectively

separated. This method is able to recover 87 pct of P2O5 and 90 pct of FeO from

steelmaking slag [79].

Phosphorus distribution ratio (Lp) between slag and liquid Fe after reduction of the BOF

slag has been determined experimentally. Lp is defined as Eq. (2.4).

𝐿𝑃 =(%P)

[%P] (2.4)

where (%P) and [%P] are respectively the phosphorus content in slag and metal [80]. Lp

increases with increasing FeO and CaO contents in the slag and increasing dissolved carbon

in the metal. Among these influencing factors, FeO content in the slag predominantly

affects the phosphorus distribution [27,73,80–82]. Morita et al. used K2CO3 and Na2CO3

to extract phosphorus from Fe-P-C alloys, which was obtained by the carbothermic

reduction of a BOF steelmaking slag at temperature higher than 1600 °C using microwave

irradiation. Na2CO3 was found more effective for extraction of phosphorus from Fe–P–C

alloys than that of K2CO3 owing to the lower evaporation of Na2CO3 [41]. Recently, Kim


18

et al. reported the phosphorus behavior in carbothermic reduction of (Mn, Fe) oxides at

1200 °C under H2 atmosphere. It was confirmed that the phosphorus did not combine with

manganese iron carbide ((Mn,Fe)7C3). Instead, it reacted with Fe and Mn to form

(Fe,Mn)3P [38]. Su et al. carried out reductive heating experiments on BOF slag at 1300–

1600 °C using quartz or serpentine as a basicity modifier, and found that phosphorus was

transferred into the metal phases instead of being released as vapor at temperatures higher

than 1400 °C [71].

To summarize the above, it can be concluded that the distribution of phosphorus in the

metal and slag phase are influenced by the slag chemistry, reduction temperature and flux

composition. It is still a challenge to separate phosphorus from the metal phase during

reduction, which is crucial to the quality of the reduced metal. In order to solve this problem,

it is of interest to further study the phosphorus behavior during the reduction.

2.5 Slag solidification and optimization of the microstructure and minerals of

solidified BOF slag

In order to obtain desirable properties of the cold/solidified slag products, hot-stage slag

engineering in the liquid state has been applied. In more detail, hot-stage slag engineering

can involve: a) additions during the molten state for reduction and separation of a metallic

phase and/or stabilization of minerals (and complementary additions to secure the

dissolution of the materials added); this can be done before, during or after slag tapping;

and b) selection of appropriate cooling paths to deliver the desirable slag product with the

optimized microstructure and minerals [83]. Energy recuperation during cooling is also a

topic of great interest. The drive behind hot-stage slag engineering is the need for slag

products that comply with environmental legislation and possibly, receive higher value in

the market. State-of-the-art concerning the fundamental and optimization of slag

solidification is presented as follows.

2.5.1 Fundamental study of slag solidification

A fundamental study of slag solidification has attracted significant attention in previous

studies. This includes (1) optimization of the microstructure and minerals of slag

product [16,84]; (2) optimization of the hot-stage process to recover heat/energy from

slag [21,85,86]; (3) freeze linings formation to protect the refractory walls in pyro-

metallurgical processes [87–89]; (4) mold fluxes solidification to optimize continuous

casting [90–92]. Considering that slag solidification will determine the microstructure of

the cooled slag, it is necessary to provide fundamental information related to the

solidification characteristics of the slags.

Slag solidification has been primarily studied via air-cooling by experiments resembling

industrial practice [93], or via rapid cooling by water or air granulation [18]. Typically, a


19

laboratory furnace is used to impose a cooling path on a synthetic slag sample, which is

subsequently characterized with respect to mineralogy and microstructure. For a very slow

cooling rate, computational thermodynamics (equilibrium modelling) [94] is able to

reasonably predict slag mineralogy. High cooling rates, applied for the synthesis of special

glasses, allow little processing control and conclusions are usually deduced using post-

mortem analysis and computational thermodynamics (Scheil-Gulliver modelling) [95,96].

In addition to a sufficiently high cooling rate, the glass forming ability of the oxidic liquid

needs consideration as it increases with the fraction of network formers (SiO2, Al2O3) [97].

Thermodynamic modeling of solidification provides an essential tool to investigate the

effects of composition on the mineralogy of the cooled slag. The common approach to

solidification modeling in oxide systems is based on the assumption of global

thermodynamic equilibrium. Evolution of slag minerals versus temperature can be

estimated by thermodynamic calculation. The stability and equilibrium composition of

minerals within slag, at sampling temperature and atmosphere, can be predicted. Moreover,

the amounts of the phases, such as Fe containing phases and C2S as well as free-CaO and

MgO are also calculated.

Figure 2.5 shows the equilibrium phases of a typical BOF slag cooled from 1800 °C to

room temperature in air atmosphere [98]. The slag has a basicity of 3.8 and MgO content

of 9.6 wt%. The thermodynamic calculation was performed by FactSage. The first

crystalline phase to precipitate from the liquid BOF slag is MgO (s), followed by the

crystallization of FeO and CaO. C3S is the first silicate to crystallise at approximately

1450 °C. Below 1270 °C, C3S is transformed to α’-C2S and CaO. At lower temperatures,

FeO and CaO react with O2 to form calcium ferrite (Ca2Fe2O5, in short C2F). Fe is found

in the minerals of C2F and FeO oxide. Since these minerals are precipitated at different

temperatures, complex combinations of crystalline structures with amorphous phases are

expected in BOF slag through slow cooling or rapid granulation.

In the past two decades, the technical development of confocal scanning laser microscopy

(CSLM), single hot thermocouple technique (SHTT) and double hot thermocouple

technique (DHTT) allows to investigate the solidification kinetics of molten slags in-situ,

e.g. the initiation of solidification [99–106], the time evolution of the solid

fraction [90,107–109], crystal growth and morphology [110–112]. Figure 2.6 shows a

schematic diagram of the CSLM set-up. A halogen infra-red heating lamp is placed in the

lower focal position of an elliptical chamber, whose refection is enhanced by coating with

gold to improve the reflection of heat. The working crucible is positioned at the upper focal

position of the chamber, where the heating energy is collected. The melting and

solidification of the sample can be readily observed by a scanning laser microscopy.


20

Figure 2.5. (a) Result of the thermodynamic calculation of a BOF slag with composition

(in wt%): 10.7 FeO, 10.9 Fe2O3, 2.3 Fe, 42.2 CaO, 9.6 MgO, 3.2 MnO, 11.1 SiO2, 1.9

Al2O3, 0.1 Cr2O3, 0.5 P2O5, 1.4 TiO2. (b) Schematic of reactions taking place during BOF

slag cooling [98]


21

Figure 2.6. The schematic diagram of the CSLM [113]

CCT (continuous cooling transformation) and time-temperature-transformation (TTT)

diagrams can be constructed by using the CSLM technique to identify the primary crystal

phase at different compositions and cooling rates. To construct CCT diagrams, molten slag

is cooled at different cooling rates. The crystallization can then be observed in-situ and

recorded as the temperature of the onset crystal precipitation. To obtain the TTT diagrams,

the molten slag is cooled down to a desired temperature rapidly and fixed at that

temperature for a certain duration. Thereafter, the isothermal crystallization can be

observed in-situ, and incubation time at different isothermal temperatures is recorded. By

integrating in-situ observations with SEM-EDS or EPMA-WDS analysis, the crystal

chemistry can be determined. High temperature in-situ X-ray diffraction (XRD) and

differential thermal analysis (DTA) can also assist the identification of the minerals

observed in the CSLM/SHTT/DHTT.

Jung et al. have investigated the crystallization behavior and crystal growth of an FeO-rich

steelmaking slag with different basicities by in-situ CSLM [99]. Figure 2.7 (a) and (b)

present the TTT and CCT diagrams of the slag. The TTT diagram has a double nose shape

and the higher basicity melt exhibits a delayed nose time. It suggests that the high basicity

delays the crystallization. The start of crystallization occurs at a lower temperature

(crystallization temperature) with increasing cooling rate. For the cooling rate ranging from

25 to 100 °C∙min-1, the crystallization temperature decreases with increasing basicity, yet


22

there is not a significant difference for crystallization temperature in the cooling rate above

400 °C∙min-1.

Figure 2.7. (a) CCT and (b) TTT diagrams of CaO–SiO2-13 wt% Al2O3-8 wt% MgO-25

mass pct FetO slags with basicity ranging from 0.7 to 1.08. The triangle, circle and

rectangle represent basicity of 0.7, 0.9 and 1.08, respectively [99]


23

Figure 2.8. CSLM and SEM images of the primary spinel phase in the slag with a basicity

of 1.08 during (a) isothermal cooling at 1200 °C and (b) rapid cooling at 3000 °C∙min-1 [99]

Figure 2.8 (a) and (b) show the crystals observed in the CSLM and SEM, under isothermal

and continuous cooling solidification respectively. Under the isothermal solidification

(Figure 2.7 (a)) at 1200 °C, the primary crystals have faceted and triangular morphologies.

For the rapid cooling at 3000 °C∙min-1, the primary crystals have changed to a dendrite

shape. The crystalline phase is identified as an Fe-rich spinel by the EDS analysis. It is

clear that with a given slag composition, the crystal morphology changes with the

solidification conditions.

Sun et al. investigated the heat recovery from high temperature blast furnace (BF) slag with

a varying amount of Al2O3 addition. Figure 2.9 (a) and (b) present, respectively, the TTT

curves of the samples and the schematic proposal for the multi-stage control of waste heat

recovery based on the measured TTT curves. Since it is necessary to obtain an amorphous

slag with high reactivity for cement production, three treatment phases were proposed for

extracting the heat, where the maximum amorphous fraction can be achieved. In the first

stage (temperature above T1), heat exchange could be kept for a rather long time by storing

the heat/energy in a phase change material (PCM) [114]. In the second stage (temperature

from T1 to T2), the heat extraction duration is limited and the cooling rate should be higher

than the critical cooling rate to avoid crystallization. The third stage (temperature below

T2), as the slag was vitrified through a rapid cooling, allows a long time for heat


24

exchange [115]. Apparently, slag solidification can also be linked with energy recovery

from the slag.

Figure 2.9. (a) TTT curves of the different samples of interest; (b) Schematic diagram of

the multi-stage control of waste heat recovery [115]

The changes in solid fraction during solidification can be evaluated using the pictures

captured with CSLM/SHTT/DHTT. The crystallization kinetics can be described by

relating the solid (crystal) fraction (𝑋) with the crystallization time using the Johnson-

Mehl-Avrami (JMA) model [116–118]. The JMA model is given by Eq. (2.5).

𝑋 = 1 − 𝑒𝑥𝑝{−[𝑘(𝑡 − 𝜏)]𝑛} (2.5)


25

where 𝑘 is a crystallization coefficient with respect to nucleation and growth, t the

crystallization time (s), 𝜏 the incubation time(s), 𝑛 the Avrami exponent that determines

the crystallization mechanisms. By rearrangement of Eq. (2.5), Eq. (2.6) is derived.

ln ln1

1−𝑋= 𝑛 ln 𝑘 + 𝑛 ln(𝑡 − 𝜏) (2.6)

By plotting ln ln1

1−𝑋 vs ln(𝑡 − 𝜏), the values of 𝑘 and 𝑛 can be determined as interception

and slope of the Eq. (2.6), respectively. The relation between 𝑛 and the corresponding

crystallization mechanism (nucleation mechanism and dimensionality of crystal growth)

has been well established and given in Table 2.4 [119].

Furthermore, the activation energy EA (J mol−1) can also be determined by the Arrhenius

Eq. (Eq. (2.7)) [90]

𝑘 = 𝐴 𝑒𝑥𝑝 (−𝐸𝐴

𝑅𝑇) (2.7)

where 𝐴 is the pre-exponential factor, 𝑅 the universal gas constant (J∙K−1∙mol−1), 𝑇 the

absolute temperature (K).

Table 2.4. Value of n for Different Nucleation and Growth Mode. Adapted from Seo et

al. [91]

Crystallization mode n

Constant nucleation rate

3-Dimensional growth 2.5

2-Dimensional growth 2


Instantaneous nucleation


2-Dimensional growth 1


Surface nucleation 0.5

Zhou et al. have studied the crystallization of mold slags with different basicities under

various isothermal temperature and cooling rates using SHTT [90]. By recording the solid

fraction associated with the solidification process the crystallization coefficient 𝑘 and the

Avrami exponent 𝑛 were obtained through Eq. (2.6). With increasing isothermal

temperature and slag basicity, it was found that both 𝑘 and 𝑛 increased. So that growth of

the primary crystal was believed to change from one-dimensional to three-dimensional

growth with a constant number of nuclei (instantaneous nucleation). Also, it was concluded

that the activation energy of the crystallization is lowered with increasing basicity [90].

According to the aforementioned overview, slag solidification/crystallization kinetics have

been investigated with respect to the influence of cooling rate and slag chemistry. Those


26

studies, however, are focused on slag with low basicity (mostly below 1.5). To the best of

the authors’ knowledge, there is no literature reporting the crystallization kinetics of a slag

with high basicity (above 3) due to its high melting temperature and rapid crystallization

behavior. To bridge this gap and meet the needs of BOF slag valorization, efforts will be

made in this thesis on the identification of the crystallization behavior of the original and

modified BOF slags, including crystal morphology, crystal nucleation and growth

mechanism and crystallization sequence.

2.5.2 Effect of slag chemistry and cooling rate on the minerals and microstructure of

the solidified BOF slag

BOF slag has poor hydraulic and cementitious behavior due to the relatively low contents

of C3S and C4AF, which are the main minerals of Portland cement clinker [120–125]. In

addition, the recycling of BOF slags is restricted due to its volume instability caused by the

free lime and periclase. The hydration and carbonation of free lime induces up to 10 pct

swelling [10]. Less than 4 wt% of free lime in the slag is required to avoid the volume

expansion [126]. This section presents the state of the art of the studies on the minerals and

microstructure optimization of BOF slag through hot-sate slag engineering (i.e. slag

composition modification and solidification control).

Table 2.5 summarizes the minerals identified in BOF slags, obtained by slow or rapid

cooling methods. Generally, all BOF slags have a high basicity and high Fe oxides content.

Even when rapidly cooled, in general, BOF slag tends to crystallize due to its chemical

composition. Tossavainen et al. [18] studied the effect of cooling rates on the mineralogy

of BOF slag, finding that the rapidly cooled (granulated) BOF slag exhibited very complex

crystalline structures similar to those of slowly cooled BOF slag. Reddy et al. [23] also

identified a very crystalline structure in quenched BOF slag using XRD analysis.

Precipitated free CaO was not found in the quenched slag due to the lack of time for its

formation upon cooling. Engstrom et al. [98] analyzed slags from a commercial BOF plant

that were cooled at various rates ranging from 0.3 to 500 °C∙s-1 using MgO crucible cooling

and water granulation cooling, in which uniformly fine meta-stable phases, such as C3S

and α-C2S formed during rapid cooling. Recently, Wang et al. [127] investigated the

influence of cooling rate by water quenching, air and furnace cooling on the minerals

formation and Fe recovery of BOF slags. Only a small amount of amorphous slag was

detected by the water quenching, whereas both the metastable and stable minerals were

found in the slowly and rapidly cooled slag. With the slowing down of the cooling, the

yield and iron content in the magnetic concentrate increased due to the higher Fe content

in the wustite solid solution. With different cooling rates, differences were observed in the

particle size distribution and crystal phase contents, including β-C2S, C2F as well as the

free CaO and MgO, which determined the physical properties of the slag after solidification.

Based on previous literature, it is clear that high cooling rates, such as by water/air

granulation or quenching is able to retain some of the high temperature phases, such as α-


27

C2S and Ca3SiO5 (C3S) and to limit the formation of free CaO. Under slow cooling

conditions, these BOF slags from industrial process generate full crystal minerals, with the

presence of free CaO, MgO, C2S, C2F and some other minerals, depending on the slag

composition.

The anticipated effect of cooling rate on the end-microstructure presented in Figure 2.10,

refers to a high basicity BOF slag [127]. From left to right, the cooling rate decreases.

Cooling by water granulation (a) results in a glassy matrix and spindle-shaped a-C2S

particles. The crystalline phases in the air cooling slag (b) are very complex. Traces of C3S

are observed as dark gray lath shapes. Compared with the granulated slag, the area ratio of

the glass matrix in the air cooling slag decreases significantly. The furnace cooling slag in

Figure 2.10 (c) has the lowest cooling rate, and the minerals are fully crystallized with very

big sizes. The glass matrix disappeared and new phases, such as C12A7 (12CaO.7Al2O3),

calcium ferrite, and Mg-Al spinel, start to form. Upon the change of slag composition,

different microstructures are expected when using slow or rapid cooling.

Table 2.5. Mineralogical compositions of BOF slags under different processing conditions

Chemical composition

(wt%)

Scale

/Atmosphere

Cooling

path Minerals Ref.

(normal slag) 52.3 CaO,

15.3 SiO2, 1.1 MgO, 3.1

P2O5, 16.2 T.Fe, (10 f-

CaO)

Pilot/Ambient Slow

cooling

C2F, C3P, β-

C2S, CaO

[23]

Pilot/Ambient Water

quenching

C2F, β-C2S,

C3S, α-C2S,

Fe2O3

(reduced slag) 61.7 CaO,

31.5 SiO2, 2.1 MgO, 0.5

P2O5, 0.4 T.Fe, (<1 f-

CaO)

Pilot/Ambient Slow

cooling γ-C2S, MgO

Pilot/Ambient Water

quenching

β-C2S, C3MS2,

C2AS

47.71 CaO, 24.36 Fe2O3,

13.25 SiO2, 6.37 MgO,

3.04 Al2O3, 2.64 MnO,

1.47 P2O5, 0.67 TiO2 (9.2

f-CaO)

Industrial/

Ambient Air cooling

β-C2S, C2AF,

C2F, CaO, RO,

Fe

[121]

42-55 CaO, 12-18 SiO2, 3-

8 MgO, <2 P2O5, 14-20

T.Fe (<10 f-CaO)

Industrial/

Ambient NR

C2S,C3S, C2F,

RO, f-MgO,

CaO

[122]

Master slag: C/S=4.39

Lab/Ar

Slow

cooling

20 CaO, 21

RO, 27.1 C2AF,

31.8 C2S

[128] SiO2 modified slag:

C/S=2.94

1.9 CaO, 25.3

RO, 15.4 C2AF,

27.1 C2S

SiO2 modified slag: C/S

=2.21

39.9 RO, 15.7

C2AF, 44.4 C2S


28

SiO2 modified slag: C/S

=1.77

34 RO, 14.5

C2AS, 51.5

C2MS

41.7 CaO, 14.8 SiO2, 3.9

MgO, 1.4 P2O5, 8.6 Al2O3,

1.2MnO, 12.1 T.Fe,

C/S=2.82

Lab/NP Air cooling

β-C2S, CT,

C2F, RO

43.4 CaO, 11.1 SiO2, 9.1

MgO, 2.2 P2O5, 1.6 Al2O3,

1.7 MnO, 21.9 T.Fe,

C/S=3.93

β-C2S, C3S,

CT, C2F, RO

43.0 CaO, 8.4 SiO2, 10.1

MgO, 1.5 P2O5, 0.8 Al2O3,

2.6 MnO, 24.2 T.Fe,

C/S=5.14

β-C2S, C3S,

CT, C2F, RO

CaO, SiO2, 30-40 Fe2O3,

10 MgO, 4 Al2O3, C/S=3-

4

Lab/NP Water

quenching

60-70 glass, 25-

35 β-C2S, 1.5-5

MgO. Trace

minerals such

as C2MS, C3A,

CMS2.

[129]

52.4 CaO, 12.8 SiO2, 5.2

MgO, 2.3 P2O5, 18.4 FeO Lab/Air

Slow

cooling

C2F, β-C2S,

CaO, MgO

[130] Air cooling CaO, C2F, β-

C2S, MgO, RO

Water

quenching

C2F, C3S, CaO,

MgO

42.3 CaO, 10.7 SiO2, 6.9

MgO, 1.2 P2O5, 1.2 Al2O3,

24.6 T. Fe

Lab/Ar

Slow

cooling

36 β-C2S, 2C3S,

7 C2F, 35 RO,

17 C12A7

[127]

Air cooling

29 β-C2S, 3C3S,

10 C2F, 31 RO,

27 glass

Splashing

42 β-C2S, 8

MgO, 6 RO, 44

glass

Water

quenching

19 β-C2S, 8

RO, 73 glass

44.7 CaO, 14.2 SiO2, 8.2

MgO, 1.7 P2O5, 3.4 Al2O3,

15.9 T.Fe

Slow

cooling

34 β-C2S, 1C3S,

9 C2F, 37 RO,

15 C12A7

Air cooling

29 β-C2S, 9

C3S, 4 C2F, 28

RO, 30 glass

Splashing

48 β-C2S, 2

MgO, 1 RO, 5

C2F, 43 glass


29

Water

quenching

28 β-C2S, 5

RO, 67 glass

36.5 CaO, 12.8 SiO2, 8.5

MgO, 1 P2O5, 5.1 Al2O3,

6.5 MnO, 29 Fe2O3

Pilot/Ambient

Air cooling

3-10

FeMgSiO4, >50

RO, 20-50 β-

C2S, 3-10

C12A7, 3-10

C3MS2, 3-10

Spinel, <3 glass [131]

Water

quenching

3-10 CFS2, 10-

20 FeO, 20-50

β-C2S, 10-20

C12A7, 20-50

Fe2O3, 3-10

glass

(modified BOF slag) 37.9

CaO, 27.7 SiO2, 6.0

Fe2O3, 7.2 FeO, 0.9 Fe,

10.8 MgO, 2.1 Al2O3, 4.4

MnO, 1.2 P2O5, C/S=1.37

Pilot/Ambient

Slow

cooling

4.0 β-C2S, 3.4

C2AF, 3.2 RO,

29.6 C3MS2, 30

CMS, 6.6

C2MS2, 22.3

amorphous

[132]

Steel ball

cooling

5.7 β-C2S, 3.8

C2AF, 7.8 RO,

47.2 C3MS2,

11.1 CMS,23

glass


CaO, 32.2 SiO2, 2.0

Fe2O3, 7.6 FeO, 0.3 Fe,

9.8 MgO, 4.8 Al2O3, 3.1

MnO, 0.8 P2O5, C/S=1.16

Slow

cooling

4.6 β-C2S,10.9

C3MS2, 55.7

CMS, 17.9

C2MS2, 9.5

C2AS

Steel ball

cooling

34.9 C3MS2,

42.8 CMS, 10.9

C2MS2, 11.5

glass

Water

quenching

37.2 C3MS2,

62.8 glass


CaO, 21.6 SiO2, 3.1

Fe2O3, 13.2 FeO, 0.1 Fe,

10.4 MgO, 10.7 Al2O3, 3.8

MnO, 1.1 P2O5, C/S=1.6

Steel ball

cooling

12.1 β-C2S,

16.5 RO, 37.4

C3MS2, 32.1

C2AF, 1.9 glass

45 CaO, 11.1 SiO2, 9.6

MgO, 2270 ppm P, 10.9

P2O5, 10.7 FeO

Lab/NR

Slow

cooling β-C2S, C2F,

RO, MgO

C3S, α-C2S, RO

[18] Water

quenching


30

35.3 Ca, 21.1 Fe, 5.5 Si,

2.7 Mn, 1.1 Al, 0.52 Mg,

0.40 Ti, 0.32 Cr, 0.17 V Lab/CO2

Carbonation

at 200-

800 °C

CaCO3, C2F,

C2S, Fe2O3,

FeO, Fe2SiO4.

[28] 41.3 Ca, 15.8 Fe, 5.9 Si,

2.9 Mn, 0.62 Al, 0.33 Mg,

0.30 Ti, 0.12 Cr, 0.15 V

23.93 Ca, 19.6 Fe, 35.69

Si, 4.66 Mg, 2.24 Mn,

1.51 Al, 0.31 Na. others

less than 0.2.

Lab/CO2 Granulation

by CO2

CaCO3, C2S,

C2AF, SiO2,

Al2O3, VO,

Ca(OH)2,

Mg(OH)2

[11]

T.Fe: total Fe; NP: Not reported; B is basicity; C: CaO; F: Fe2O3; P: P2O5; RO = FeO; A:

Al2O3; M: MgO; S: SiO2; T: TiO2.

To summarize previous studies listed in Table 2.5, it can be concluded that: (1) SiO2 and

Al2O3 are commonly used to modify the slag composition for the microstructure and

mineral optimization, targeting the added-value application [133]. (2) The addition of SiO2

lowers slag basicity and results in more silicate after solidification, which accommodates

the extra CaO and thereof eliminates the free lime. When the slag basicity (mass ratio of

CaO/SiO2) was reduced to around 1.8, a large amount of bredigite phase is formed, which

incorporates MgO. This bredigite formation should be prevented [134–136] due to its poor

hydraulic behaviour. (3) The addition of Al2O3 eliminated free lime by forming C2AF

mineral [74,128,129,137]. (4) Upon the change in composition, different microstructures

are expected when using slow or rapid cooling. Even when rapidly cooled, BOF slag tends

to crystallize due to its chemical composition (i.e. high basicity C/S > 4). A high cooling

rate retains some of the high temperature phases, refines the crystal grain size and promotes

the formation of glassy phase, but there is no report on forming a complete glassy slag

through increasing the cooling rate of the original slag [23,98,128–131,138,24]. (5) The

reported studies on slag valorization were focused on either material product preparation

or on the metal extraction, there is, however, no concurrent study of both the metal recovery

and slag reutilization. In this thesis, therefore, the effect of SiO2 and Al2O3 on the

mineralogical and microstructural modification of the solidified BOF slag under various

cooling conditions will be investigated. The glass formation ability of the modified slag

will be evaluated with the aim to valorize the modified slag into construction applications.

Furthermore, a potential method to achieve “Zero waste” of BOF slag will be developed in

this PhD work by both metal recovery and slag utilization.


31

Figure 2.10. BSE images of a BOF slag (C/S = 3.1) with different cooling methods, (a)

water granulation, (b) air cooling and (c) furnace cooling. 1-C2S; 2-Glass matrix; 3-RO

phase; 4-C3S; 5-MgAl2O4; 6-Calcium ferrite; 7-12CaO∙7Al2O3; 8-Free CaO; 9-

Periclase [127]


32

2.6 Conclusion and outlook

In this chapter, the different approaches to recover Fe and optimize the solidification of

BOF slags are reviewed. Studies on the reaction kinetics of the carbothermic reduction of

slags are summarized. Also, the recent progress on the fundamentals of the slag

solidification is presented. Moreover, to steer BOF slag for hydraulic and cementitious

applications, effects of chemical composition and cooling paths on slag minerals and

microstructure are discussed.

The main components of BOF slag are Fe, CaO, and SiO2, MgO and MnO. Fe is in the

form of metallic iron (7 to 10 wt%), iron oxide and other iron bearing minerals. The major

phases are dicalcium silicate, calcium aluminate and wustite. The recycling of BOF slag is

an important challenge for the sustainability of steelmaking industry and for environmental

protection.

Carbothermic reduction is an effective method to extract metallic Fe from the slag. Many

investigations have been carried out to study reduction mechanisms for optimizing the

reduction process. But so far, the reduction mechanisms are not fully understood. Kinetics

of the carbothermic reduction of iron oxide containing slag has been significantly improved

by imposing an external magnetic field during the reduction. However, this is still at lab

experimental stage and out of the scope of the present thesis.

The distribution of phosphorus in the recovered metal and slag phase are influenced by

slag chemistry, reduction temperature and flux composition. Phosphorus contaminates the

quality of reduced metal. In order to control phosphorus content in metallic iron, it will be

of great importance to study the phosphorus behavior distribution during reduction.

Chapter 3 will present a carbothermic reduction of BOF slag, aiming to extract a high-

grade iron metal, and concurrently reutilize the remaining slag, aiming to achieve “Zero

waste” of BOF slag.

Since understanding BOF slag crystallization is the precondition to control the slag

solidification process for its microstructure optimization, fundamental study on slag

solidification has attracted significant attention. BOF slag typically has a high melting

temperature and rapid crystallization, leading to a challenge for the study of crystallization

kinetics. To fill this gap and meet the needs of BOF slag valorization, Chapter 4 will focus

on the identification of the crystallization behavior of original and modified BOF slags,

including crystallization time (TTT) and temperature (CCT), crystal morphology and

crystallization sequence.

To improve the hydraulic and cementitious properties of BOF slags, SiO2 and Al2O3 are

commonly used to modify the slag composition for the microstructure and mineral

optimization, targeting the added-value application [127]. Upon the change in slag

composition, different microstructures are expected when using slow or rapid cooling.


33

Even when rapidly cooled, BOF slag tends to crystallize due to its chemical composition

(i.e. high basicity C/S > 4). A high cooling rate promotes the glassy formation of the slag

and refines the crystal grain size. To tailor the microstructure for BOF slag valorization, it

is needed to have a better understanding of the combined effect of slag chemistry and

cooling conditions on the mineralogical and microstructural modification of the solidified

BOF slag. In chapter 5, laboratory experiments on the effects of SiO2 addition and oxygen

partial pressure will be discussed. Furthermore, pilot scale granulation trials are conducted

in Chapter 6 to evaluate the glass formability of BOF slag with a specific SiO2 and Al2O3

additions. A mathematical simulation will be performed to provide a better understanding

of the granulation process.


34

References


[2] J. T. Gao, S. Q. Li, Y. T. Zhang, Y. L. Zhang, P. Y. Chen, and P. Shen: J. Iron

Steel Res. Int., 2011, vol. 18, pp. 32–39.

[3] D. M. Proctor, K. A. Fehling, E. C. Shay, J. L. Wittenborn, J. J. Green, C. Avent,

R. D. Bigham, M. Connolly, B. Lee, T. O. Shepker, and M. a. Zak: Environ. Sci.

Technol., 2000, vol. 34, pp. 1576–82.

[4] http://www.slg.jp/e/slag/character.html, accessed on 1/3/2017.

[5] X. Gao, M. Okubo, N. Maruoka, H. Shibata, T. Ito, and S. Y. Kitamura: Miner.

Process. Extr. Metall., 2015, vol. 124, pp. 116–24.


[7] H. T. Shen and E. Forssberg: Waste Manag., 2003, vol. 23, pp. 933–49.

[8] K. C. Mills and B. J. Keene: Int. Mater. Rev., 1987, vol. 32, pp. 1–120.


pp. 261–68.


[11] M. Morone, G. Costa, A. Polettini, R. Pomi, and R. Baciocchi: Miner. Eng., 2014,

vol. 59, pp. 82–90.

[12] J. Vlcek, V. Tomkova, H. Ovcacikova, F. Ovcacik, M. Topinkova, and V. Matejka:

METALURGIJA, 2013, vol. 52, pp. 329–33.

[13] W. Z. Zhou, Z. S. Huang, C. P. Sun, H. X. Zhao, and Y. Z. Zhang: Bioresour.

Technol., 2016, vol. 214, pp. 534–40.

[14] L. I. Bowden, A. P. Jarvis, P. L. Younger, and K. L. Johnson: Environ. Sci.

Technol., 2009, vol. 43, pp. 2476–81.

[15] E. T. Turkdogan and R. J. Fruehan: in Fundam. Iron Steelmak., 11th ed.,

Association of Iron and Steel Engineers, Pittsburgh, 1998, pp. 160.

[16] I. Z. Yildirim and M. Prezzi: Adv. Civ. Eng., 2011, vol. 2011, pp. 1–13.

[17] H. Yi, G.P. Xu, H.G. Cheng, J.S. Wang, Y.F. Wan, and H. Chen: Procedia Environ.

Sci., 2012, vol. 16, pp. 791–801.

[18] M. Tossavainen, F. Engstrom, Q. Yang, N. Menad, M. Lidstrom Larsson, and B.

Bjorkman: Waste Manag., 2007, vol. 27, pp. 1335–44.


14.

[20] B. Das, S. Prakash, P.S.R. Reddy, and V.N. Misra: Resour. Conserv. Recycl., 2007,

vol. 50, pp. 40–57.

[21] T. S. Zhang, Q. J. Yu, J. X. Wei, J. X. Li, and P. P. Zhang: Resour. Conserv.

Recycl., 2011, vol. 56, pp. 48–55.



[23] A. Srinivasa Reddy, R. K. Pradhan, and S. Chandra: Int. J. Miner. Process., 2006,

vol. 79, pp. 98–105.

[24] A. Rai, J. Prabakar, C.B. Raju, and R.K. Morchalle: Constr. Build. Mater., 2002,


35

vol. 16, pp. 489–94.

[25] M. Chiara, S. Manzi, I. Lancellotti, E. Kamseu, L. Barbieri, and C. Leonelli: Appl.

Clay Sci., 2013, vol. 73, pp. 78–85.


Int., 2007, vol. 47, pp. 1541–48.

[27] M. O. Suk, S. K. Jo, S. H. Kim, K. Y. Lee, and J. M. Park: Metall. Mater. Trans.

B, 2006, vol. 37, pp. 99–107.

[28] R. M. Santos, D. Ling, A. Sarvaramini, M. X. Guo, J. Elsen, F. Larachi, G.

Beaudoin, B. Blanpain, and T. Van Gerven: Chem. Eng. J., 2012, vol. 203,

pp. 239–50.

[29] M. Salman, Ö. Cizer, Y. Pontikes, R. M. Santos, R. Snellings, L. Vandewalle, B.

Blanpain, and K. V. Balen: Chem. Eng. J., 2014, vol. 246, pp. 39–52.

[30] R. D. Rawlings, J. P. Wu, and a. R. Boccaccini: J. Mater. Sci., 2006, vol. 41,

pp. 733–61.

[31] K. Zhang, J. W. Liu, W. C. Liu, and J. K. Yang: Chemosphere, 2011, vol. 85,

pp. 689–92.

[32] E. B. Ferreira, E. D. Zanotto, and L. M. Scudeller: Glas. Sci. Technol., 2002,

vol. 75, pp. 75–86.

[33] J. N. Murphy, T. R. Meadowcroft, and P. V. Barr: Can. Metall. Q., 1997, vol. 36,

pp. 315–31.

[34] D. Durinck, F. Engström, S. Arnout, J. Heulens, P.T. Jones, B. Björkman, B.

Blanpain, and P. Wollants: Resour. Conserv. Recycl., 2008, vol. 52, pp. 1121–31.

[35] K. Horii, N. Tsutsumi, Y. Kitano, and T. Kato: Processing and Reusing

Technologies for Steelmaking Slag, 2013.

[36] N. Menad, N. Kanari, and M. Save: Int. J. Miner. Process., 2014, vol. 126, pp. 1–

9.

[37] N. Y. Ma and J. B. Houser: J. Clean. Prod., 2014, vol. 82, pp. 221–31.

[38] D. Y. Kim, H. S. Kim, and S. M. Jung: ISIJ Int., 2015, vol. 55, pp. 504–12.

[39] J. Mróz: Met. Mat. Trans. B, 2001, vol. 32B, pp. 821–30.

[40] M. Ishikawa: ISIJ Int., 2006, vol. 46, pp. 530–38.

[41] K. Morita, M. X. Guo, N. Oka, and N. Sano: J. Mater. Cycles Waste, 2002, vol. 4,

pp. 93–101.

[42] T. Yamashita, T. Nakada, and K. Nagata: Metall. Mater. Trans. B, 2007, vol. 38,

pp. 185–91.

[43] M. F. Jiang, Y. Y. Cui, D. Y. Wang, Y. Min, and C. J. Liu: J. Iron Steel Res. Int.,

2013, vol. 20, pp. 1–20.

[44] A. Semykina, PhD thesis, Royal Institute of Technology, 2010.

[45] A. Semykina, V. Shatokha, M. Iwase, and S. Seetharaman: Metall. Mater. Trans.

B, 2010, vol. 41, pp. 1230–39.

[46] A. Semykina and S. Seetharaman: Metall. Mater. Trans. B, 2010, vol. 42, pp. 2–

4.

[47] S. R. Story, B. Sarma, R. J. Fruehan, A. W. Cramb, and G. R. Belton: Metall.


[48] S. Y. Sun, PhD thesis, McMaster University, 1997.

[49] B. Sarma, A. W. Cramb, and R. J. Fruehan: Metall. Mater. Trans. B, 1996, vol. 27,

pp. 717–30.

[50] M. Barati and K.S. Coley: Metall. Mater. Trans. B, 2005, vol. 36B, pp. 169–78.


36

[51] R. Corbari, H. Matsuura, S. Halder, M. Walker, and R.J. Fruehan: Metall. Mater.

Trans. B Process Metall. Mater. Process. Sci., 2009, vol. 40, pp. 940–48.

[52] N. Siddiqi, B. Bhoi, R.K. Paramguru, V. Sahajwalla, and O. Ostrovski: Ironmak.

Steelmak., 2000, vol. 27, pp. 437–41.

[53] K. P. El-Rassi and T. A Utigard: Met. Mater. Trans. B, 2000, vol. 31, pp. 1187–

94.

[54] A. Warczok and T.A. Utigard: Can. Metall. Q., 1998, vol. 37, pp. 27–39.

[55] M. Kucharski, T. Sak, P. Madej, M. Wędrychowicz, and W. Mróz: Metall. Mater.

Trans. B, 2013, vol. 45, pp. 590–602.

[56] A. K. Jouhari, R. K. Galgali, P. Chattopadhyay, R. C. Gupta, and H. S. Ray: Scand.

J. Met., 2001, vol. 30, pp. 14–20.

[57] S. L. Teasdale and P. C. Hayes: ISIJ Int., 2005, vol. 45, pp. 642–50.

[58] D. J. Min, J. W. Han, and W. S. Chung: Metall. Mater. Trans. B, 1999, vol. 30,

pp. 215–21.

[59] A. T. Hay, patent, 1872.

[60] R. Skorski: Nat. Phys. Sci., 1972, vol. 240, pp. 15–16.

[61] I. Svare: Nat. Phys. Sci., 1973, vol. 244, pp. 78–79.

[62] C. T. Peters: Nat. Phys. Sci., 1973, vol. 244, p. 79.

[63] M. W. Rowe, R. Fanick, D. Jewett, and J. D. Rowe: Nature, 1976, vol. 263,

pp. 756–57.

[64] M. W. Rowe, S. M. Lake, and R. Fanick: Nature, 1977, vol. 266, pp. 612–14.

[65] M. W. Rowe, D. A. Edgerley, M. Hyman, and S. M. Lake: J. Mater. Sci., 1979,

vol. 14, pp. 999–1001.

[66] D. Aylmer and M. W. Rowe: J. Chem. Phys., 1983, vol. 78, pp. 2094–95.

[67] P. K. Gallagher: J. Chem. Phys., 1981, vol. 75, p. 3847.

[68] M. W. Rowe, P. K. Gallagher, and E. M. Gyorgy: J. Chem. Phys., 1983, vol. 79,

pp. 3534–36.

[69] Y. L. Jin, Z. W. Zhao, B. W. Li, H. X. Dai, X. F. Zhang, patent, 2015.

[70] K. Y. Kim, E. J. Jang, D. Y. Kim, and S. M. Jung: Ironmak. Steelmak., 2016, pp. 1–

11.

[71] T. H. Su, H. J. Yang, Y. C. Lee, Y. H. Shau, E. Takazawa, M.F. Lin, J.L. Mou, and

W.T. Jiang: Steel Res. Int., 2016, vol. 87, pp. 1511–1526.

[72] S. Basu, A. K. Lahiri, and S. Seetharaman: Metall. Mater. Trans. B Process Metall.

Mater. Process. Sci., 2007, vol. 38, pp. 357–66.

[73] C. M. Lee and R. J. Fruehan: Ironmak. Steelmak., 2005, vol. 32, pp. 503–8.


2005, vol. 32, pp. 54–60.

[75] M. D. Johnston and M. Barati: J. Non. Cryst. Solids, 2011, vol. 357, pp. 970–75.

[76] C. Duée, C. Bourgel, E. Véron, M. Allix, F. Fayon, F. Bodénan, and J. Poirier:

Cem. Concr. Res., 2015, vol. 73, pp. 207–14.

[77] Y. Q. Sun and Z. T. Zhang: Metall. Mater. Trans. E, 2016, vol. 3, pp. 114–22.

[78] V. Rubio, M.A. de la Casa-Lillo, S. Aza, and P.N. Aza: J. Am. Ceram. Soc., 2011,

vol. 94, pp. 4459–62.

[79] R. Inoue and H. Suito: ISIJ Int., 2006, vol. 46, pp. 174–79.

[80] N. Maruoka, S. Narumi, and S.Y. Kitamura: ISIJ Int., 2015, vol. 55, pp. 419–27.


37

[81] S. Basu, A. K. Lahiri, and S. Seetharaman: ISIJ Int., 2007, vol. 47, pp. 1236–38.

[82] S. Basu, A. K. Lahiri, and S. Seetharaman: Metall. Mater. Trans. B Process Metall.

Mater. Process. Sci., 2007, vol. 38, pp. 623–30.

[83] F. Engström, Y. Pontikes, D. Geysen, P. T. Jones, B. Björkman, and B. Blanpain:

in 2nd Int. Slag Valoris. Symp., 2011, pp. 231–50.

[84] R. I. Iacobescu, A. Malfliet, L. Machiels, P. T. Jones, B. Blanpain, and Y. Pontikes:

Waste and Biomass Valorization, 2014, vol. 5, pp. 343–53.

[85] M. Barati, S. Esfahani, and T. A. Utigard: Energy, 2011, vol. 36, pp. 5440–49.

[86] G. Bisio: Energy, 1997, vol. 22, pp. 501–9.

[87] K. Verscheure, M. Campforts, F. Verhaeghe, E. Boydens, M. Van Camp, B.

Blanpain, and P. Wollants: Metall. Mater. Trans. B, 2006, vol. 37, pp. 929–40.

[88] M. Campforts, E. Jak, B. Blanpain, and P. Wollants: Metall. Mater. Trans. B, 2009,

vol. 40, pp. 619–31.

[89] M. Campforts, E. Jak, B. Blanpain, and P. Wollants: Metall. Mater. Trans. B, 2009,

vol. 40, pp. 632–42.

[90] L. J. Zhou, W. L. Wang, F. J. Ma, J. Li, J. Wei, H. Matsuura, and F. Tsukihashi:

Metall. Mater. Trans. B, 2011, vol. 43, pp. 354–62.

[91] M. D. Seo, C. B. Shi, J. Y. Baek, J. W. Cho, and S. H. Kim: Metall. Mater. Trans.

B Process, 2015, vol. 46, pp. 2374–83.

[92] C. B. Shi, M. D. Seo, H. Wang, J. W. Cho, and S. H. Kim: Metall. Mater. Trans.

B, 2014, vol. 46, pp. 345–56.

[93] D. Durinck, P. T. Jones, B. Blanpain, and P. Wollants: J. Am. Ceram. Soc., 2008,

vol. 91, pp. 3342–48.

[94] C. W. Bale, P. Chartrand, S. A. Decterov, G. Eriksson, K. Hack, R.Ben. Mahfoud,

J. Melancon, A. D. Pelton, and S. Petersen: Calphad, 2002, vol. 26, pp. 189–228.

[95] D. Durinck, P. T. Jones, B. Blanpain, P. Wollants, G. Mertens, and J. Elsen: J. Am.

Ceram. Soc., 2007, vol. 90, pp. 1177–85.

[96] K. I. Dragnevski, A. M. Mullis, and R. F. Cochrane: Mater. Sci. Eng. A, 2004,

vol. 375–377.

[97] P. Richet, M. Roskosz, and J. Roux: Chem. Geol., 2006, vol. 225, pp. 388–401.

[98] F. Engström, D. Adolfsson, Q. Yang, C. Samuelsson, and B. Björkman: Steel Res.

Int., 2010, vol. 81, pp. 362–71.

[99] S. S. Jung and I. Sohn: Environ. Sci. Technol., 2014, vol. 48, pp. 1886–92.

[100] J. J. Liu, G. Chen, P. C. Yan, B. Blanpain, N. Moelans, and M.X. Guo: Am. J.

Sci., 2014, vol. 402, pp. 1–8.

[101] S. S. Jung and I. Sohn: Metall. Mater. Trans. B, 2012.

[102] S. S Jung and I. Sohn: J. Am. Ceram. Soc., 2013, vol. 96, pp. 1309–16.

[103] M. L. Hu, L. Liu, X.W. Lv, C. G. Bai, and S. F. Zhang: Metall. Mater. Trans. B,

2014, vol. 45, pp. 76–85.

[104] Y. Kashiwaya, T. Nakauchi, K.S. Pham, S. Akiyama, and K. Ishii: ISIJ Int., 2007,

vol. 47, pp. 44–52.


38

[105] Y. Kashiwaya, C.E. Cicutti, and A.W. Cramb: ISIJ Int., 1998, vol. 38, pp. 357–

65.

[106] Z. J. Wang, Y. Q. Sun, S. Sridhar, M. Zhang, M. Guo, and Z.T. Zhang: Metall.


[107] W. W. Xuan, K. J. Whitty, Q. L. Guan, D. P. Bi, and J. S. Zhang: Fuel, 2014,

vol. 137, pp. 193–99.

[108] W. W. Xuan, K. J. Whitty, Q. L. Guan, D. P. Bi, Z. H. Zhan, and J. S. Zhang:

Energy and Fuels, 2014, vol. 28, p. 6627−6634.

[109] L. J. Zhou, W. L. Wang, R. Liu, and B. G. Thomas: Metall. Mater. Trans. B, 2013,

vol. 44, pp. 1264–79.

[110] A. Semykina, J. Nakano, S. Sridhar, V. Shatokha, and S. Seetharaman: Metall.


[111] A. Semykina, J. Nakano, S. Sridhar, V. Shatokha, and S. Seetharaman: Metall.


[112] J. Heulens, B. Blanpain, and N. Moelans: J. Eur. Ceram. Soc., 2011, vol. 31,

pp. 1873–79.

[113] I. Sohn and R. Dippenaar: Metall. Mater. Trans. B, 2016, vol. 47, pp. 2083–94.

[114] T. Nomura, N. Okinaka, and T. Akiyama: ISIJ Int., 2010, vol. 50, pp. 1229–39.

[115] Y. Q. Sun, Z. T. Zhang, L. L. Liu, and X. D. Wang: Energies, 2014, vol. 7,

pp. 1673–84.

[116] M. Avrami: J. Chem. Phys., 1939, vol. 7, p. 1103.

[117] W. A. Johnson and R. F. Mehl: Trans. Am. Inst. Min. Met. Eng, 1939, vol. 135,

pp. 416–42.

[118] H. Yinnon and D. R. Uhlmann: J. Non. Cryst. Solids, 1983, vol. 54, pp. 253–75.

[119] J. W. Christian: The Theory of Transformations in Metals and Alloys, 3rd ed.,

Elsevier Ltd, Oxford, 2002.

[120] H. F. W. Taylor: Cement Chemistry, Thomas Telford Publishing; 2 edition

(January 1, 1997), London, 1997.

[121] J. Waligora, D. Bulteel, P. Degrugilliers, D. Damidot, J. L. Potdevin, and M.

Measson: Mater. Charact., 2010, vol. 61, pp. 39–48.

[122] J. Geiseler: Waste Manag., 1996, vol. 16, pp. 59–63.

[123] P. Y. Mahieux, J. E. Aubert, G. Escadeillas, and M. Measson: J. Mater. Civ. Eng.,

2014, vol. 26, pp. 593–98.

[124] P. Y. Mahieux, J. E. Aubert, and G. Escadeillas: Constr. Build. Mater., 2009,

vol. 23, pp. 742–47.

[125] H.M. Ludwig and W.S. Zhang: Cem. Concr. Res., 2015, vol. 78A, pp. 24–37.

[126] G. Wang, Y.H. Wang, and Z.L. Gao: J. Hazard. Mater., 2010, vol. 184, pp. 555–

60.

[127] D.Y. Wang, M.F. Jiang, C.J. Liu, Y. Min, Y.Y. Cui, J. Liu, and Y.C. Zhang: Steel

Res. Int., 2012, vol. 83, pp. 189–96.

[128] X.R. Wu, P. Wang, L.S. Li, Z.J. Wu, and R.H. Chen: Ironmak. Steelmak., 2011,

vol. 38, pp. 185–88.


39

[129] D. Ionescu, T.R. Meadowcroft, and P.V. Barr: Adv. Cem. Res., 2001, vol. 13,

pp. 21–30.

[130] M. Gautier, J. Poirier, F. Bodénan, G. Franceschini, and E. Véron: Int. J. Miner.

Process., 2013, vol. 123, pp. 94–101.

[131] D. Norval: in South. African Inst. Min. Metall., 2011, pp. 129–44.

[132] J. B. Ferreira Neto, J. O. G. Faria, C. Fredericci, F. F Chotoli, A. N. L. Silva, and

B. B. Ferraro: J. Sustain. Metall., 2016, vol. 2, pp. 13–27.

[133] J. A. Duffy, M. D. Ingram, and L. D. Sommerville: J. Chem. Soc. Faraday Trans.

1 Phys. Chem. Condens. Phases, 1977, vol. 74, pp. 1410–19.

[134] D. Moseley and F. P. Glasser: Cem. Concr. Res., 1981, vol. 11, pp. 559–65.

[135] D. Moseley and F. P. Glasser: J. Mater. Sci., 1982, vol. 17, pp. 2736–40.

[136] L. Kriskova, Y. Pontikes, L. Pandelaers, Ö. Cizer, P.T. Jones, K. Van Balen, and


[137] V. Z. Serjun, B. Mirtic, and A. Mladenovic: Mater. Tehnol., 2013, vol. 47,

pp. 543–50.

[138] P. N. Chaudhary, J. Pal, D. P. Singh, M. Singh, and S. Ghosh: Trans. Indian Inst.

Met., 2004, vol. 57, p. A5.

41

Chapter 3

Valorization of BOF steel slag by reduction and phase modification:

metal recovery and slag valorization

Published in the Journal of Metallurgical and Materials Transactions B, 2017, 48(3):

1602-1612 with minor adjustments, Chunwei Liu; Shuigen Huang; Patrick Wollants; Bart

Blanpain; Muxing Guo

Department of Materials Engineering, KU Leuven, Kasteelpark Arenberg 44, BE-3001

Leuven, Belgium

DOI: 10.1007/s11663-017-0966-0

Abstract

BOF steel slag is a main by-product in steelmaking and its valorization is therefore of

considerable interest, both from a metal recovery and from a residue utilization perspective.

In the present study, the carbothermic reduction of BOF slag was investigated

systematically. The reduction of Fe and P containing phases (i.e. oxide and compounds) is

discussed. Effects of Al2O3 and SiO2 additions on the solidification microstructure and

mineralogy associated with the reduction process were also investigated. The formation

and growth of the extracted metallic phase is discussed and the mineralogy of the residue

slag is determined. We conclude that by controlling the additions under a rapid cooling

condition, it is possible to extract metallic iron as high-grade metal and simultaneously to

utilize the remaining slag into construction applications.

Key words: metal recovery; BOF slag valorization; reduction; phase modification.

Contribution of Chunwei Liu: Shuigen Huang initiated the experiment. Chunwei Liu

performed the experiments, analysed the samples, interpreted the results and wrote the

manuscript. The contribution of the co-authors consisted in discussion of the results and

reviewing of the paper before final publication.

Chapter 3. Reduction

42

3.1 Introduction

Steel slag is classified as BOF (Basic Oxygen Furnace) slag, EAF (Electric Arc Furnace)

slag and LM (Ladle Metallurgy) slag. The EU produced around 20 million tons steel slag

annually in recent years, of which 46 wt% is BOF slag [1]. Recycling and reutilization of

such large amounts of steel slag is of great importance, both for the sustainability of the

metallurgical industry and for the environment. After modification of its chemistry and

solidification mineralogy by means of hot stage engineering, steel slag can be recycled for

internal use in the steelmaking process, for road construction, for cementitious substitutes,

and as agricultural fertilizer [2–8].

Using the slag for constructional or agricultural purposes, however, does not consider the

large amount of Fe present in the slag. Typically, BOF slags contain 14-29 wt% of Fe,

present in iron oxides and iron-containing minerals [9]. “Zero waste” of BOF slag can be

achieved by both metal recovery and slag utilization. However, the difficulty of controlling

the P distribution between slag and the recovered metal prevents the slag from a full

reutilization. Because of P, the recovered Fe has a limited application. Also, P stabilizes β

dicalcium silicate (β-C2S). The complete removal of P from BOF slag leads to

disintegration of the slag due to the transformation of β dicalcium silicate to γ dicalcium

silicate (γ-C2S) during cooling [10]. Thus it is favorable to concentrate P in the oxides

while limit its concentration in metallic Fe. In previous studies, the carbothermic reduction

of slag has been described to recover Cr and/or Fe from stainless steel slag, hot metal

dephosphorization slag and other steel making slag [11–14], without focusing on

improving the purity and controlling the morphology of the metal. Furthermore, the

utilization of the remaining slag by modifying the chemical composition did not receive as

much attention as the metal. Ye et al. studied the reduction of steel slags to recover both

metals (Fe, Mn, V and Cr) and oxide materials. The oxides could be reused as cementitious

materials and/or desulphurization fluxes in secondary metallurgy, but no chemical

modification of the oxides was considered [15]. To utilize the steel slag as aggregates in

road construction, Yang et al. studied the chemical modification of the slag after reduction,

focusing on avoiding the disintegration of slag caused by the transformation of β-C2S to γ-

C2S [16]. However, the concurrent implementation of reduction to achieve high added

value metal products, and chemical improvement of the remaining oxides for application

in cements and geopolymers, has not been shown.

In this work, the carbothermic reduction of Fe oxides and P containing compound and/or

oxides of BOF slags with different amounts of Al2O3 and SiO2 additions was investigated.

The effect of Al2O3 on the size distribution of the reduced Fe was discussed. Based on

experimental observations and thermodynamic calculations, a potential method to recover

low P containing Fe particles with a reasonably homogeneous size distribution is suggested.

Concurrently, the phase modification of residual BOF slag with the aim to produce slag

product with enhanced cementitious property was studied. A potential method to achieve


43

the “zero-waste” concept of the BOF slag, contributing to the sustainability of the

steelmaking industry, is proposed as well.

3.2 Experimental methods and materials

3.2.1 Slag preparation

An industrial BOF slag was collected at different positions in a slag yard to obtain a

representative chemistry of the slag. The samples were ground, followed by milling below

200 µm. Thereafter the powders were thoroughly mixed for homogenization, and then

applied in this study as the master material. Table 3.1 shows the chemical composition of

this BOF slag, as determined by XRF (X-Ray Fluorescence spectroscopy, Panalytical

PW2400). Fe2+ and Fe3+ were measured using chemical titration by potassium dichromate.

10.2 wt% Fe2+ and 10.4 wt% Fe3+ were determined for the starting slag.

Due to the presence of different iron oxides (FeO, Fe3O4 and Fe2O3) in the BOF slag, a

preliminary experiment was performed. An appropriate range of C for reducing the oxides

was determined to be 5-8 wt%. Thus, in the present study, to study the effect of carbon on

the reduction behavior of iron and phosphorus containing compounds, respectively 5, 6, 7

and 8 wt% C (Superior Graphite, Sweden, 6.5 µm) were added to the master slag in the

form of powder. To reveal the effect of SiO2 and Al2O3 additions on phase modification of

the slag and metal formation in the reduction experiments, different amounts of SiO2

(Sibelco, Belgium, 1-40 µm) and Al2O3 (Sasol, Germany, 25 µm) were mixed with the

master slag, keeping the C addition fixed at 7 wt%. Table 3.2 shows the various additions

of SiO2, Al2O3 and C in the master slags. Each mixture was wet-mixed using ethanol in a

multidirectional mixer (Turbula type) for 24 hours. The mixture was then dried by a

rotating evaporator at 65 °C and further dried at 80 °C for 24 hours.

Table 3.1 Chemical composition of the master slag (wt%)

CaO *Fe Fe2+ Fe3+ SiO2 MnO MgO Al2O3 P2O5 TiO2 V2O5

44.5 20.6 10.2 10.4 10.14 4.78 2.18 2.05 2.43 0.89 0.37 *Fe is the total amount of iron in the oxides.

Table 3.2. Different additions for preparing the experimental slags

Parameters of interest Addition (wt%)

C SiO2 Al2O3

Effect of C 5, 6, 7, 8 0 0

Effect of SiO2 and Al2O3 7 0, 1, 2, 3, 4, 5 5, 10


44

3.2.2 Experimental procedure and characterization

Each slag mixture (10 g) was loaded in a high purity magnesia crucible (21 mm ID, 50 mm

H), which was suspended by Mo hooks in a vertical tube furnace (100-250/18, HTRV,

GERO, Germany) under an Ar flow rate of 0.4 L∙min-1. The samples were introduced at

room temperature and held in the furnace at 1600 °C for 1 hour, followed by water

quenching.

The microstructure of the slags was quantitatively analyzed using Field-Emission Electron

Probe Micro Analysis (FE-EPMA, JXA-8530F, JEOL Ltd, Japan) at fixed accelerating

voltage (15 kV) and beam current (15 nA). For the WDS analysis of the light element C,

the analyzing crystal adopted is a layered diffracting element 1 (LED1) at K-alpha line.

Phase identification was achieved by X-Ray Diffraction (XRD, 3003-TT, Seifert,

Ahrensburg, Germany), with 2θ in the range of 10-80° using Cu Kα radiation at 40 kV and

40 mA.

3.3 Results and discussion

3.3.1 Effect of C and Al2O3 on metal recovery

Figure 3.1 shows the Fe and P elemental distribution after reduction by different C

additions. With the additions of 5 and 6 wt% C, most of the Fe is concentrated in the

spherical metallic phase. Some Fe is found in other phases, mainly in magnesia wustite

(RO). This phase is composed of monoxides, such as FeO, MnO and MgO. With increasing

C addition, Fe eventually concentrates entirely in the metallic phase, as the higher C

additions create stronger reduction conditions. In addition, increasing C additions also

favor the transfer of P from slag to metal. As indicated by the color level, with 5 and 6 wt%

C additions, very low concentrations of P can be seen in the metallic phase. At 8 wt% C

addition, the concentration of P in the metallic phase is much higher. Furthermore, free

lime and periclase precipitate during the reduction due to the high basicity of the slags. It

is known that the hydration of free lime and periclase induces about 10 vol%

expansion [17], which should be avoided for construction applications.

In order to acquire accurate data on the reduction behavior of BOF slags, more than 10

metallic grains were analyzed using WDS (Wavelength-Dispersive Spectroscopy). The

dissolved P content in the metallic Fe is given in Figure 3.2 (solid square points). With 5

and 6 wt% C addition, the P content is below 0.01 wt%. With 8 wt% C addition, the P

content increases to above 2 wt%. Clearly the reduction of P-containing compounds and

the dissolution of P in Fe are closely related to the amount of C addition, i.e. the higher the

carbon addition, the higher concentration of phosphorus in the extracted Fe. At 5 wt% and

6 wt% C additions, the low P contents suggest that P-containing compounds cannot be


45

reduced or volatilized after reduction. Kazuki et al. [11] have carried out carbothermic

reduction experiments of steelmaking slag at over 1600 °C using microwave processing,

and proved that FexO is more easily reduced than P oxides. Maruoka et al. [18] studied the

distribution ratio Lp of P between metallic Fe and slag after the reduction of iron ore by

CO, and observed that P is reduced and dissolves in the metallic phase when the FexO

content in the slag is below 10 wt%.

Figure 3.1. Fe and P distribution at various C additions.


46

Figure 3.2. P content in the metallic Fe and P distribution ratio Lp.

Most P in BOF slag is incorporated with Ca2SiO4 (C2S) to form a C2SiO4-Ca3(PO4)2 (C2S-

C3P) solid solution [7]. The structure of C2S-C3P is similar to that of pure C2S, with part of

tetrahedral [SiO44-] groups substituted by tetrahedral [PO4

3-] groups. To maintain the

charge balance, the substitution combines less Ca2+ than that in the pure C2S [19]. During

the reduction of the C2S-C3P solid solution, the P-containing compounds react with C

and/or CO to break the P-O chemical bonds. Thereafter, the released free Ca ions can

combine with silicate and form Ca3SiO4 (C3S). Therefore, with the reduction of P from the

BOF slag, the C3S content is expected to increase. The reduction of P-rich compounds and

the formation of C3S can be written as reaction (3.1-1) to reaction (3.1-4)

5 C + 3 Ca2SiO4 + Ca3(PO4)2 = 3 Ca3SiO5 + 5 CO(g) + 2 [P] (3.1-1)

5 C + 3 Ca2SiO4 + Ca3(PO4)2 = 3 Ca3SiO5 + 5 CO(g) + P2(g) (3.1-2)

5 CO + 3 Ca2SiO4 + Ca3(PO4)2 = 3 Ca3SiO5 + 5 CO2(g) + 2 [P] (3.1-3)

5 CO + 3 Ca2SiO4 + Ca3(PO4)2 = 3 Ca3SiO5 + 5 CO2(g) + P2(g) (3.1-4)

The P distribution ratio Lp is defined as

𝐿𝑃 =(%P)

[%P] (3.2)

where (P) and [P] represent the wt% of P, respectively in slag and metal. Lee and

Fruehan [20] suggested that Lp can be estimated as

log 𝐿𝑃 = −12.24 +2000

T+ 2.5 log(%FeO) + 6.65 log B∗ + 0.13[%C] (3.3)

5 6 7 8

0

2

4

6

8

10

12

14

log

Lp , w

t%

[P], measured

[P], calculated

C addition, wt pct

[P],

wt

pct

-1.5

0.0

1.5

3.0

4.5

6.0

log Lp


47

where T the absolute temperature (K), (%FeO) and [%C] the wt% of FeO in the liquid slag

and C in the metallic Fe, B* the “weighted basicity” defined as B∗ = [(%CaO) +

0.8(%MgO)]/[%SiO2 + (%Al2O3) + 0.8(%P2O5)]. [%C] was measured by WDS, and B∗

was calculated according to the slag composition and (%FeO) was calculated using

FactSage 7.0. For the calculation, all the iron is assumed to be hematite (Fe2O3). FactPS

and FToxid database were applied. Then the P distribution ratio can be obtained. T, [%C],

(%FeO), B* and the calculated values of Lp with Eq. (3.3) are listed in Table 3.3. log Lp is

given as a function of carbon addition in Figure 3.2 by the solid triangles.

Table 3.3. Calculated P distribution ratio Lp.

Carbon addition (wt%) T (K) [%C] B* (%FeO) Lp

5

1873

0.12

2.68

12.1 4.07

6 0.39 6.0 3.32

7 0.48 0.21 -0.41

8 1.09 0.08 -1.37

[%C]: dissolved carbon in metallic Fe, measured by WDS; (%FeO): FeO content in the liquid

slag, calculated by FactSage 7.0.

By considering the mass balance of P and assuming that no P is evaporated during the

reduction, the P content in the metal can be calculated, as shown in the Figure 3.2 (solid

circle points). The P content in the metallic Fe matches well with the experimental

observation where the P-rich phase changes from the slag (mainly C2S-C3P) to metallic Fe

with increasing C addition (Figure 3.1). With increasing C addition, the measured and

calculated P contents in the metallic Fe increase considerably, but the measured increase is

less than the calculated one and this trend becomes more significant with higher carbon

addition. The reason is probably that some P evaporates after the reduction.

The consumption of C by the reduction of iron oxides can be estimated via the reaction

(3.4):

3 C + Fe2O3 = 3 CO + 2 Fe (3.4)

To simplify the calculation, all iron oxides are assumed to be hematite (Fe2O3). According

to the reaction (3.4), the molar ratio of C to Fe2O3 is 3 to completely reduce the Fe2O3. In

the present experiments, 5, 6, 7 and 8 wt% C were used based on the Fe2O3 content in the

master slag. In addition to the reduction of iron oxides, the C can also reduce other oxides,

such as P-containing compounds and Mn-containing compounds. For the current slag,

under the condition of 6 wt% C addition, the P content in the metallic Fe is below 0.01 wt%

and the purity of the extracted Fe is above 98 wt%, implying a promising metallic product.

FactSage calculation suggests that Fe3P and Fe3C precipitate at higher C additions (8 wt%).

Fe3P increases by increasing temperature (precipitating temperature is 1440 °C), but Fe3C

decreases with increasing temperature (disappearing temperature is 1560 °C). To estimate


48

Fe recovery, mass balance calculation based on the measured values was carried out by

considering following assumptions: Fe is in metal, RO (5 wt% C addition) and periclase

phases (6-8 wt% C additions); all MgO is in the periclase phase; Mn is in metal and RO

phases. Also, the Fe recovery can be assessed by FactSage calculation. Figure3.3 shows

the measured and calculated (by FactSage) Fe recovery. By increasing the C additions from

5 to 7 wt%, the Fe recovery increases rapidly, but further increasing C additions has little

effect. Under 5 and 6 wt% C additions, in addition to the metallic Fe and the periclases, Fe

elements also distributes in free lime, which could explain the overestimation of the

measured Fe recovery.

Therefore, to minimize the P content in the metallic Fe while maximizing the Fe recovery

from the present BOF slag at 1873K, the optimized molar ratio of carbon to iron oxides is

suggested to be 3.

Figure 3.4 shows the morphological evolution of Fe particles reduced by 7 wt% C,

respectively with 0, 5 and 10 wt% Al2O3 additions. With increasing Al2O3 addition, the size

of the particles changes significantly. The particles which attach to the amorphous phase

are larger (typical diameter above 10 µm) than the ones which are located in the solid

solutions (typical diameter below 10 µm). The crystals in the amorphous part were

probably precipitated during water quenching.

Figure 3.3. Measured and calculated Fe recovery with different C additions.

5 6 7 8

70

80

90

100

Fe

reco

ver

y,

pct

C addition, wt pct

measured

calculated


49

Figure 3.4. Morphology of metallic Fe recovered with 7 wt% C and (a) 0 wt% Al2O3; (b)

5 wt% Al2O3; (c) 10 wt% Al2O3 additions.

In order to understand the formation and growth mechanism, the Fe particles in each case

were processed by ImageJ, with the aid of which the 2-dimensional size distribution can

be obtained. The particles with a size below 1 µm are neglected to avoid errors due to the

limitation of the software. In order to eliminate the arbitrariness caused by the size of bins

defined manually, the Population Density Function (PDF) [21,22] is adopted. The

frequency in PDF is defined as the normal frequency divided by the bin width and has


50

length –4 units. The functional PDF is applied to compare the size distributions between

different samples (Figure 3.5). In each case, the Fe particle size varies greatly and most of

the Fe particles are below 10 µm. The population density decreases markedly with the size

increment. The largest particle size is increased from 16.5 to 78.5 and 127.5 µm with

increasing Al2O3 addition from 0 to 5 and 10 wt%. Furthermore, the population density of

smaller sized particles (diameter below 10 µm) decreases with increasing Al2O3 addition,

indicating that the number of smaller particles is decreased by Al2O3 addition. The large

range of Fe particle size under higher Al2O3 addition implies a more inhomogeneous size

distribution was induced by Al2O3 addition.

The particles were divided into two classes: one below 10 µm and the other above 10 µm.

Figure 3.6 shows the average diameter of the Fe particles and the liquid fraction with

respect to different Al2O3 additions. Below 10 µm no significant size difference can be

detected. Above 10 µm, however, the size varies significantly with Al2O3 additions. Once

a stable nucleus of an Fe particle is formed, it grows by diffusion of Fe atoms (reduced by

carbon) from the bulk of the slag to the surface of the Fe particle (volume diffusion),

followed by incorporation of the Fe atoms into the units (surface kinetics). Most of the

metallic Fe is believed to be reduced from the Fe-containing RO solid solution and the

C2AF (brownmillerite) phases. As shown in Figure 3.7 (a), two kinds of RO

microstructures in the master bulk slag are found: smaller RO dispersed in the free lime

and larger sized RO separated from other phases.

Figure 3.5. Population density functions of Fe particles under 0, 5 and 10 wt% Al2O3

additions. Largest particle size is 16.5 to 78.5 and 127.5 µm under respectively 0, 5 and

10 wt% Al2O3 additions.


51

Figure 3.6. Average diameter of metallic Fe and liquid fraction with 0, 5 and 10 wt%

Al2O3 additions under 7 wt% C: (a) particles below 10 µm; (b) particles above 10 µm.

The error bars represent the standard error of the mean.

The Fe particles below 10 µm (circles in Figure 3.7 (b)), are considered to be formed from

the dispersed RO in the lime (circles in Figure 3.7 (a)). Their average size is independent

of Al2O3 additions and is quite homogeneous (see the standard deviation indicated by the

error bar in Figure 3.6 (a)). FactSage calculation suggests that at 1600 °C, the liquid

fractions under 7 wt% C addition are 33.5%, 62.6% and 77.5% with 0, 5 and 10 wt% Al2O3

additions (Figure 3.6). Thus the Fe particles grow in the mixture of liquid and solids. As

the Fe particles are always located in the solid solution phase, their growth is probably

controlled by solid diffusion. Due to the small diffusion coefficient of Fe in the slag (10-10


52

to 10-11 m2∙s-1) [23], diffusion-controlled growth is limited to an extreme small scale.

Therefore, the size of the smaller sized Fe particles (<10 µm) is comparable to that of the

dispersed RO in the master slag.

The larger Fe particles (>10 µm) are believed to be formed from the larger-sized RO and

C2AF phases. After Fe has been extracted, the melting temperature of the Fe-deficient slag

is significantly increased. Figure 3.4 indicates that Al2O3 addition enlarges the liquid

fraction in the slag, resulting in more amorphous phase. Due to the fact that the larger Fe

particles are located in the amorphous area which corresponds to the liquid fraction at the

experimental temperature, the growth of Fe particles is probably enhanced by the liquid

slag at high temperature. The liquid fraction at 1600 °C was calculated by FactSage 7.0

(Figure 3.6). The increase of the average Fe diameter shows a similar trend as the increase

of the liquid fraction of the slag at 1600 °C, implying that the particles sizes are indeed

increased by the liquid slag. There are two possible reasons for the liquid-induced growth:

more Fe particle collisions can be caused by the flow of the liquid slag, and the diffusion

and convection of Fe are improved in the liquid phase. The large and inhomogeneously

distributed Fe particles caused by external Al2O3 additions (as given in Figure 3.5) is

therefore attributed to the local inhomogeneous distribution of the liquid slag in the mixture

at high temperature. Although a more quantitative understanding of Fe particle growth is

needed for the commercial fine Fe preparation, based on the above findings, however, it

can be concluded that the Fe particle purity, size and size distribution can be manipulated

through the process control with carbothermic reduction parameters, such as carbon and

slag modifier additions and temperature.

Figure 3.7. (a) Microstructures of the master bulk slag; (b) Microstructure of the reduced

slag with 10 wt% Al2O3 addition

3.3.2 Effect of Al2O3 on solidification microstructure during reduction

To optimize the microstructure of the slag after Fe extraction, the effects of Al2O3 on slag

microstructure and mineralogy were investigated. Figure 3.8 shows the XRD patterns for


53

the tested samples under fixed 7 wt% C, with 0, 5 and 10 wt% Al2O3 additions. The main

crystalline phases are metallic Fe, dicalcium silicate (C2S) and tricalcium silicate (C3S).

The JCPDS 37-1497 database [24] was employed for free lime identification. The

theoretical interplanar distance 2.4059 Å and 1.7008 Å corresponds to 2 theta at 37.34° and

53.86° using Cu Kα radiation. It can be seen that at the 2 theta positions of 37.62° and

54.40°, which we consider to be the characteristic position of lime for our samples, the

slags with 0 and 5 wt% Al2O3 additions present a peak while the one with 10 wt% Al2O3

addition shows no peak. The free lime can be completely removed by more Al2O3 additions.

The shift of 2 theta degree in our sample is believed to be caused by the incorporation of

Fe2+, Mn2+ and Mg2+, whose diameters are smaller than Ca2+. Thereafter the practical

interplanar distance is slightly decreased, thus the corresponding 2 theta degree is shifted

to a bigger 2 theta degree. The decrease of intensity for Fe peaks indicate that the quantity

of metallic Fe is decreased with Al2O3.

Figure 3.9 shows the effect of Al2O3 additions on the microstructural modification of BOF

slag reduced by 7 wt% C. The composition of each phase was measured by WDS. Through

linking each phase detected by XRD, its mineralogy can be identified. It is evident that

most of the Fe particles are spherical. C3S typically has a bigger size (> 30 µm) than other

phases, indicating that it might be the first crystal that precipitates, having the longest time

to grow. No free lime can be observed in the case of 10 wt% Al2O3 addition, which is

consistent with the XRD pattern.

Figure 3.8. Comparison of XRD pattern for samples with 0, 5 and 10 wt% Al2O3

respectively, and 7 wt% C additions


54

Figure 3.9. Microstructural modification by Al2O3 (7 wt% C): (a): 0 wt% Al2O3, (b): 5

wt% Al2O3, (c) 10 wt% Al2O3

There is a phase containing most of the elements in the slag and cannot be linked to the

composition of any crystalline phase. Its unique morphology is like a wall separating each

crystal. It is therefore considered to be amorphous. In each sample, more than 5 points of

the amorphous were measured. It was found that the CaO content increases from 39.9 wt%

to 46.3 wt% by increasing Al2O3 additions from 5 wt% to 10 wt%, suggesting that the

elimination of free lime is caused by dissolving more CaO into the amorphous phase (i.e.

Al2O3 addition increases lime solubility of the slag).

3.3.3 Effect of combination of SiO2 and Al2O3 on solidification microstructure during

reduction

Figure 3.10. Phase modification by different additions of SiO2 associated with (a) 7 wt%

C + 5 wt% Al2O3 additions and (b) 7 wt% C + 10 wt% Al2O3

As described above, with 5 wt% Al2O3 and 7 wt% C additions, free lime cannot be

eliminated (see Figure 3.8). Figure 3.10 shows the phase identification for the slags with


55

combined SiO2 and Al2O3 additions. An extra 1 wt% SiO2 addition is able to remove the

free lime completely. C3S, which is a favored phase for cement production [25], can be

formed under a combination of 5 wt% Al2O3 and no more than 4 wt% SiO2, as shown in

Figure 3.10(a). Yet with 5 wt% SiO2 + 5wt% Al2O3 (5 SiO2 curves in Figure 3.10(a)) or 10

wt% Al2O3 additions (Figure 3.10(b)), no C3S phase has been precipitated. The decrease of

peaks at approximately 34 degree in Figure 3.10 (a) indicates the decrease of C3S by adding

SiO2. By referring to the EPMA observation, the peaks at around 40 degree in Figure 3.10

(b) is considered as the overlap of C2S and C3S. Therefore, with increasing SiO2, C3S is

decreased while C2S is increased due to the decrease of basicity B:

𝐵 =𝐶𝑎𝑂 (𝑤𝑡%)

𝑆𝑖𝑂2(𝑤𝑡%)+𝐴𝑙2𝑂3(𝑤𝑡%) (3.5)

According to Eq. (3.5), B > 1.98 is a tentative criteria to precipitate C3S (at visible extent).

On the other hand, B < 2.59 is required to avoid any free lime. Therefore, the optimized

basicity to maintain C3S and to remove free lime is in the range of 1.98 to 2.59.

Figure 3.11. Microstructural modification by SiO2 and Al2O3 additions with 7 wt% C

addition. “mSnA” represents m wt% SiO2 and n wt% Al2O3 addition. 1 - Fe; 2 - C3S; 3 -

C2S; 4 - MgO; 5 - Amorphous; S - SiO2; A: Al2O3.

Figure 3.11 presents the combined effect of SiO2 and Al2O3 on slag microstructural

modification for the reduction test with 7 wt% C addition. No free lime is observed in the

test, which agrees well with the XRD patterns (see Figure 3.10). The composition of each

phase was characterized by WDS analysis. The crystals could be identified by their XRD

patterns. By comparing samples with the same level of SiO2 addition (Figure 3.11(a) and

(d), (b) and (e), (c) and (f)), the amorphous phase is found to increase with increasing Al2O3


56

addition. By comparing the amorphous phase under the same level of Al2O3 addition

(Figure 3.11(a) to (c), and (d) to (e)), the amorphous phase is decreased with increasing

SiO2 additions.

For a quantitative understanding of the effect of combined SiO2 and Al2O3 additions on the

mineral modification associated with the reduction, the composition of each microstructure

was measured using WDS analysis. To ensure the accuracy, every microstructure was

measured in more than 10 points. A mass balance calculation of Al2O3, CaO and SiO2 was

carried out using the overall composition and the composition of each phase to evaluate

the fraction of different phases. Specifically, Al2O3, CaO and SiO2 are assumed to be in the

amorphous, C3S and C2S phases. Thus the Eqs. (3.6-1) through (3.6-3) can be derived.

𝑚𝐴𝑙2𝑂3= 𝑚𝐴𝑚 × 𝑤𝐴1

+ 𝑚𝐶3𝑆 × 𝑤𝐴2+ 𝑚𝐶3𝑆 × 𝑤𝐴3

(3.6-1)

𝑚𝐶𝑎𝑂 = 𝑚𝐴𝑚 × 𝑤𝐶1+ 𝑚𝐶3𝑆 × 𝑤𝐶2

+ 𝑚𝐶2𝑆 × 𝑤𝐶3 (3.6-2)

𝑚𝑆𝑖𝑂2= 𝑚𝐴𝑚 × 𝑤𝑆1

+ 𝑚𝐶3𝑆 × 𝑤𝑆2+ 𝑚𝐶2𝑆 × 𝑤𝑆3

(3.6-3)

where 𝑚𝐴𝑙2𝑂3, 𝑚𝐶𝑎𝑂, and 𝑚𝑆𝑖𝑂2

are the overall mass percentages of the Al2O3, CaO and

SiO2; 𝑚𝐴𝑚, 𝑚𝐶3𝑆, and 𝑚𝐶2𝑆 the mass percentages of amorphous, C3S and C2S phases; 𝑤𝐴,

𝑤𝐶 and 𝑤𝑆 the mass fractions of Al2O3, CaO and SiO2 in which the subscript indicates the

mass fractions in amorphous (1), C3S (2) and C2S (3) phase. In this way, the mineral

modification by the combined addition of SiO2 and Al2O3 can be obtained. The main

minerals are presented in Figure 3.12.

The amount of the amorphous phase (Figure 3.12(a)) increases significantly by increasing

the amount of Al2O3 addition from 5 wt% to 10 wt%, which is in reasonable agreement

with Figure 3.11. Meanwhile, the amount of amorphous phase decreases with SiO2 addition,

specifically for 10 wt% Al2O3 addition. Due to the fact that the metallic Fe particle size is

decreased by lowering the amorphous fraction, thus the Fe size is expected to decrease with

SiO2 addition. The amount of C3S decreases by increasing the amount of either Al2O3 or

SiO2 addition (see Figure 3.12(b)) due to the decrease of basicity. By adding 10 wt% Al2O3,

C3S is completely eliminated. On the contrary, the amount of C2S increases, both by the

addition of Al2O3 and SiO2 (see Figure 3.12(c)). The increase in the amount of C2S phase

resulting from the addition of SiO2 might be attributed to the decrease of basicity,

consequently more C2S was precipitated from the molten slag at high temperature. The

changes of the amounts of C2S and C3S by SiO2 at the addition level of 5 wt% Al2O3 are

more significant than the changes at the addition level of 10 wt% Al2O3. According to Eq.

(3.5), the basicity change by SiO2 at a lower level of Al2O3 addition is more sensitive than

at a higher level of Al2O3 addition. The changed level of basicity could explain the behavior

of C2S and C3S precipitation.


57

Figure 3.12. Modification of (a) amorphous, (b) C3S, and (c) C2S distribution by the

combined addition of SiO2 and Al2O3


58

Figure 3.13. The ternary phase diagram CaO-SiO2-Al2O3 as calculated at 1600 °C by

FactSage 7.0. The lines indicated by A and S represent the composition modification by

Al2O3 and SiO2, respectively.

The ternary phase diagram CaO-SiO2-Al2O3 was calculated at 1873 K with FactSage 7.0

and is shown in Figure 3.13. To find the position indicating the present composition in the

phase diagram, the slag was simplified to a ternary system of CaO-SiO2-Al2O3 and

accordingly, the composition is normalized. The calculated phase diagram is, where “+”

and “-” represent slags with additions of 5 wt% and 10 wt% Al2O3, respectively. The solid

point shows the composition of the original slag. By adding 5 wt% and 10 wt% Al2O3, as

indicated by the arrows, the composition moves closer to the liquid area (the liquidus is

highlighted by the thick curve in Figure 3.13). Furthermore, the dashed line S and A

respectively represents the modification direction by SiO2 (S) and by Al2O3 (A) additions.

With SiO2 addition, the modified slag moves through region 3 (liquid+C2S+C3S) to region

2 (liquid+C2S), which indicates that SiO2 addition decreases C3S and increases C2S.

Meanwhile, with Al2O3 addition, the modified slag moves through region 3

(liquid+C2S+C3S), region 2 (liquid+C2S) and to region 1 (liquid). It demonstrates that a

limited Al2O3 addition can decrease C3S and increase C2S, but further adding Al2O3

decreases C2S and results in a fully liquid slag. Therefore, the amount of each phase can

be controlled by controlling the additions of SiO2 and Al2O3.


59

3.4 Conclusions

The valorization of a typical BOF slag with high basicity was studied by carbothermic

reduction and slag microstructure modification through combined SiO2 and Al2O3 additions.

The reduction of Fe oxides and P containing compounds was discussed, and optimized

process conditions to recover high-grade metallic iron (> 98 wt%) were proposed. The

formation and growth of the extracted metallic Fe was observed. Due to their excellent

electronic, magnetic and chemical properties, the micron-sized iron particles may have

high-added value applications in preparing magnetorheological fluids, metal injected

molding and microwave absorption materials [26]. The effects of SiO2 and Al2O3 on the

slag microstructure after solidification were investigated. The formation of the amorphous

phases, and C2S and C3S crystals can be controlled by additions, implying potential

applications as cementitious substitutions. The main conclusions can be summarized as:

(1) with increasing C addition, the P-rich phase changes from slag to metal. By controlling

C addition, it is possible to avoid contamination of metallic Fe by P during carbothermic

reduction. It is suggested to keep the molar ratio of C to iron oxides as 3:1, in order to

minimize the P content in the metallic Fe.

(2) the smaller metallic Fe particles (<10 µm) are formed from the dispersed RO in the

lime, and the larger ones (<10 µm) are formed from the bigger sized RO and C2AF. The

growth of smaller particles is dominated by solid diffusion, and the growth of the larger

particles is influenced by liquid fraction of the slag at high temperature (liquid-induced

flow, diffusion and convection). Based on the findings in this work, it can be concluded

that the Fe particle purity, size and size distribution can be manipulated through the process

control with carbothermic reduction parameters, such as carbon, slag modifier additions

(SiO2 and Al2O3) and temperature.

(3) the optimized basicity to precipitate C3S crystal and remove free lime (for high added

value application of the slag) should be controlled in the range of 1.98 to 2.59.

(4) Al2O3 addition enlarges the liquid fraction at high temperature and consequently

decreases the viscosity of the current slag; SiO2 addition effectively stabilizes free lime and

promotes the formation of the C2S phase. Depending on the application purpose, the

microstructure and mineralogy of the slag product can be tailored through the combined

Al2O3 and SiO2 effects.


60

References

[1] İ. Yüksel: Environ. Dev. Sustain., 2016, pp. 1–16.

[2] Y. Xue, S. Wu, H. Hou, and J. Zha: J. Hazard. Mater., 2006, vol. 138, pp. 261–68.

[3] S.A. Mikhail and A.M. Turcotte: Thermochim. Acta, 1995, vol. 263, pp. 87–94.

[4] G. Wimmer, H .Wulfert, H. M. Ludwig, and A. Fleischanderl: in METEC and 2nd

ESTAD, 2015, pp. 15-19.

[5] J. Vlcek, V. Tomkova, H. Ovcacikova, F. Ovcacik, M Topinkova, and V Matejka:

2013, vol. 52, pp. 329–33.


vol. 59, pp. 82–90.


Int., 2007, vol. 47, pp. 1541–48.


[9] H. Shen and E. Forssberg: Waste Manag., 2003, vol. 23, pp. 933–49.



[11] K. Morita, M. Guo, N. Oka, and N. Sano: J. Mater. Cycles. Waste. Manag., 2002,

vol. 4, pp. 93–101.

[12] T. Kim and J. Lee: Mater. Trans., 2011, vol. 52, pp. 2233–38.

[13] M. Ishikawa: ISIJ Int., 2006, vol. 46, pp. 530–38.

[14] E. Shibata, S. Egawa, and T. Nakamura: ISIJ Int., 2002, vol. 42, pp. 609–13.

[15] G. Ye, E. Burstr, M. Kuhn, and J. Piret: Scand. J. Metall., 2003, vol. 32, pp. 7–14.

[16] Q. Yang, F. Engström, B. Björkman, and D. Adolfsson: in 8th Int. Conf. Molten

Slags, Fluxes Salts–MOLTEN 2009.

[17] R. M. Santos, D. Ling, A. Sarvaramini, M. Guo, J. Elsen, F. Larachi, G. Beaudoin,

B. Blanpain, and T. V. Gerven: Chem. Eng. J., 2012, vol. 203, pp. 239–50.

[18] N. Maruoka, S. Narumi, and S. Kitamura: ISIJ Int., 2015, vol. 55, pp. 419–27.



[20] C. M. Lee and R. J. Fruehan: Ironmak. Steelmak., 2005, vol. 32, pp. 503–8.

[21] M. D. Higgins: Am. Mineral., 2000, vol. 85, pp. 1105–16.

[22] M. V. Ende, M. Guo, E. Zinngrebe, B. Blanpain, and In-Ho Jung: ISIJ Int., 2013,

vol. 53, pp. 1974–82.

[23] J. Mróz: Met. Mat. Trans. B, 2001, vol. 32B, pp. 821–30.

[24] E.H. Swanson, H.E., Morris, M.C., Stinchfield, R.P. and Evans: Standard X-Ray

Diffraction Powder Patterns, Section 14, U.S. Department of Commerce, National

Bureau of Standards, Washington, DC, 1962, pp. 49

[25] J. Zach, H. Kminova, O. Horky, and J. Brozovsky: in 8th Int. Conf. Slov. Soc.

Non-Destructive Testing, Conf. Proc., 2005, pp. 375–82.

[26] A. Kordecki and B. Weglinski: Powder Metall., 1990, vol. 33, pp. 151–55.

61

Chapter 4

Effect of Al2O3 addition on mineralogical modification and

crystallization kinetics of a high basicity BOF steel slag

Submitted to Waste Management, Chunwei Liu, Shuigen Huang, Bart Blanpain,

Muxing Guo.

Abstract

Basic oxygen furnace (BOF) steel slag is a main by-product produced during the converter

steelmaking process. The volume instability and fast crystallization of BOF slag limit its

added-value application. This article aims to understand the effect of Al2O3 on the

mineralogical modification and crystallization kinetics of a high basicity BOF steel slag.

Continuous cooling transformation (CCT) and time-temperature-transformation (TTT)

curves have been constructed to determine the crystallization characteristics of BOF slag.

The critical cooling rate to vitrify the slags was experimentally obtained. The

crystallization sequence has been clarified by integrating the in-situ and post-mortem

observations with thermodynamic calculation. The result suggests that by steering Al2O3

addition and cooling rate, BOF slag can be modified to obtain enhanced potential as binder

for added-value applications.

Key words: recycling; BOF steel slag; Al2O3 addition; crystallization kinetics.

Contribution of Chunwei Liu: Chunwei Liu performed the experiments, analysed the

samples, interpreted the results and wrote the paper. The contribution of the co-authors

consisted in discussion of the results and reviewing of the paper before final publication.

Chapter 4. Effect of Al2O3 addition

62

4.1 Introduction

The world crude steel production has increased substantially from 1.25 billion tonnes in

2006 to 1.62 billion tonnes in 2015[1]. Steel slag, as a main by-product during the

steelmaking process, has a commensurate output. The EUROSLAG (European Slag

Association) members produce around 20 million tonnes steel slag annually, among which

approximately 10 million tonnes is BOF slag[2]. Storage of such large amount of slag is a

major issue for both the company and the environment. Therefore it is of great importance

to find and implement added-value applications for steel slag in general and BOF slag in

particular.

By hot stage engineering, the BOF steel slag has been successfully reused in different

applications, such as aggregates for road construction pavement[3,4], cement for building

purpose[5–7] and fertilizer for agricultural application[8]. The high strength and

mechanical resistance of the slag due to its high contents of silicate enables its applications

in civil engineering. Also, the phosphorus in the slag is favorable to nourish plants and

improve soil quality. The volume disintegration of the high basicity (mass ratio of

CaO/SiO2) BOF steel slag, however, raises challenges in its reutilization as added-value

constructional materials. Up to 10 pct expansion is induced due to the hydration and

carbonation of free lime and magnesia[9]. According to the previous work, 4 wt% free lime

is tolerable to avoid general expansion[10]. Several different methods have been carried

out with the goal to chemically stabilize the free lime. Accelerated carbonation treatment

of slag has been demonstrated as an efficient way to eliminate free lime[11–13]. SiO2

addition was shown to be effective by transforming the free lime and dicalcium silicate to

merwinite and akermanite[14,15]. Recently, the present authors reported a lime

stabilization method by cooling the slag in air, where wustite in the slag is oxidized to

hematite which then stabilizes the free lime by forming calcium aluminoferrite[16].

Another issue that limits the added-value application of BOF slag is its fast crystallization

property due to the high mass ratio of CaO/SiO2 in the slag[17]. A stabilized slag that is

mostly crystalline can be mixed with ordinary Portland cement (OPC) into blended

cements [14,18]. By forming more amorphous phase, the slag can be used as an inorganic

polymer (IP)[19]. Compared to blended cements, the compressive strength of IP is similar,

but the production of IP reduces CO2 generation substantially[20]. Therefore, the use of

BOF slag to prepare inorganic polymer may be given priority in terms of economic and

environmental considerations. Rapid cooling is the most common method to vitrify the

slag[21–23]. To elucidate the cooling condition for vitrification, many efforts have been

made to construct CCT and TTT diagrams of slags by using confocal laser scanning

microscopy (CSLM)[24–27], the single hot thermocouple technique (SHTT)[28,29] and

double hot thermocouple technique (DHTT)[30,31]. These CCT and TTT curves provide

fundamental understanding of the crystallization kinetics of the slag during cooling. Jiang

et al. investigated the effect of CaO/Al2O3 mass ratio (ranged from 0.8 to 1.2) on the

crystallization behavior of a CaO-Al2O3 based slag system, where Al2O3 has been reported


63

to facilitate the crystallization and increase the crystallization temperature[32]. We have

however not found any research reported on the crystallization kinetics of a high basicity

BOF slag system.

In summary of the previous studies, it can be concluded that (1) high basicity induces

volume instability and fast crystallization of the BOF slag, which has limited its added-

value application as constructional materials; (2) rapid cooling is a promising method to

vitrify the BOF slag that helps to improve cementitious potential; (3) Al2O3 has been

reported to facilitate the crystallization and increase the crystallization temperature for

lower basicity slags, but its effect on the crystallization kinetics of a high basicity BOF slag

has not been investigated.

In this paper, the role of Al2O3 in the mineralogical modification and crystallization

kinetics of the high basicity BOF slag are investigated. Water quenching has been used to

vitrify the original and Al2O3 modified slags. To reveal the influence of Al2O3 on

crystallization kinetics of BOF slag, in-situ observation of the crystallization process of

both the original and Al2O3 modified slags was performed using CSLM. The critical

cooling rate to vitrify BOF slags is determined. This study suggests that by steering Al2O3

addition and cooling rate, BOF slag can be modified to obtain enhanced potential as binder

for added-value applications.

4.2 Experimental methods and thermodynamic calculations

4.2.1 Materials preparation

A typical BOF steel slag was used as the starting material. The slag was sampled at

different positions from a slag yard. Then the bulk slag was milled to below 200 µm and

mixed thoroughly. Table 4.1 gives the chemical composition range of the master slag,

which was determined by X-ray Fluorescence spectroscopy (Panalytical PW2400).

Concentration of Fe2+ and Fe3+ in the slag were measured using chemical titration by

potassium dichromate.

Table 4.1. Chemical composition of the master slag (wt%)


44.5 20.6 10.2 10.4 10.14 4.78 2.18 2.05 2.43 0.89 0.37

*Fe is the total iron in the oxides.

To modify the chemistry, 5, 10 and 15 wt% Al2O3 (Sasol, Germany, 25 µm) were

respectively added to the master slag in ethanol and each of the mixtures was homogenized


64

in a multidirectional mixer (Turbula type) for 24 h, followed by drying in a rotating

evaporator at 60 °C and further dried in a muffle furnace at 80 °C for 24 h.

4.2.2 Furnace experiments

To study the rapid cooling effect on the original and modified slags, 30 g of each of the

slags was loaded in a high purity magnesia crucible (32 mm ID, 70 mm H), which was

suspended by Mo hooks in a vertical tube furnace (100-250/18, HTRV, GERO, Germany)

under an Ar flow rate of 0.4 L∙min-1. The Ar was pre-purified by passing through a Mg

furnace at 500 °C. The slag was melted and held at 1600 °C for 1 h to homogenize the

composition, followed by water quenching.

4.2.3 In-situ CSLM observation

In-situ observation was performed to investigate Al2O3 effects on the crystallization

kinetics of BOF slag. A technical description and temperature calibration of CSLM has

been reported in our previous work[24,33]. The schematic representation of the sample

holder used in the present study is given in Figure 4.1. A 10×10 mm Pt foil was used as the

holder and the center of the foil was mechanically curved, allowing to hold a small amount

(~ 0.1 g) of slag. The slag was pre-melted at 1600 °C in air for 5 min to homogenize the

chemical composition, release the gas and determine the liquidus temperature of the slag

for the afterward observation. The slag was then re-melted again at the temperature 50-

100 °C higher than the pre-determined liquidus temperature in air. Figure 4.2 (a) and (b)

present a schematic temperature profile used in the construction of the CCT and TTT

diagrams, respectively. To determine the CCT diagram, the molten slag was cooled at

various cooling rate, i.e. from 30 to 1600 °C∙min-1. The onset temperature of the phase

transition (liquid-solid) is defined as the temperature at which the primary crystal was first

observed. The ending point of the phase transition is considered to be at the temperature at

which the crystals stop growing. Actual cooling rate was determined by analyzing the

temperature history after experiments, and is given in Figure 4.2 (a). To investigate

temperature and time dependences of the crystallization under isothermal condition, the

molten slag was cooled rapidly to the temperature of interest and held at that temperature

to observe the isothermal crystallization. The incubation time is defined as the moment

when the primary crystal became detectable. The ending time for the crystallization is

defined as the moment when crystals stop growing.

During the entire process, the sample was continuously monitored and images were

captured at a frequency of 30 frames per minute. ImageJ software was used to analyze the

amorphous and mineralogical fractions in the solidified slags.


65

Figure 4.1. Schematic representation of the slag specimen holder used in CSLM

Figure 4.2. Temperature profile to construct (a) CCT and (b) TTT curves in CSLM

4.2.4 Thermodynamic calculation

To have a better understanding on the crystallization of the original and modified slags,

FactSage 7.0 [34,35] was used to calculate the thermodynamic equilibrium in air using the

databases FactPS and FToxid. “Equilibrium” and “Scheil-Gulliver” solidification modules


66

coupled with FactSage modelling were adopted to investigate the slag crystallization

behavior. The “Equilibrium” module assumes mass diffusion is infinitely fast, so that the

system is at thermodynamic equilibrium. The “Scheil-Gulliver” module, on the other hand,

assumes that infinitely fast diffusion in liquid slag, and no diffusion in the solids. Therefore

in the “Scheil-Gulliver” module, thermodynamic equilibrium is achieved within the liquid

slag, but not in the global system[36].

4.2.5 Sample analysis and characterization

The mineralogical composition of the quenched slags was identified by X-ray Diffraction

(XRD, D2 Phaser, Bruker, Germany), with 2θ in the range of 5-70° using Cu Kα radiation

at 30 kV and 10 mA. The step size was set at 0.02° within 0.6 second. Approximate 10 wt%

ZnO powder (purity 99.9 pct, Sigma-Aldrich BVBA, Belgium) was wet-mixed with the

samples as a reference to identify the amount of amorphous phase. Rietveld refinement

was applied to achieve the quantitative analysis of the XRD pattern. Microstructural

analysis of the slags was done by electron probe microanalysis (EPMA, JXA-8530F, JEOL

Ltd, Japan). The accelerating voltage was 15 kV and beam current was15 nA.


4.3.1 Effect of Al2O3 addition on mineralogy of the quenched BOF slag

Figure 4.3 shows the XRD spectra of the original and Al2O3 modified slags after water

quenching at 1600 °C. The theoretical interplanar distances corresponding to the strongest

diffraction level of lime is dt = 2.41 Å (h, k, l = 2, 0, 0), and the corresponding 2 theta

position is at 37.34° using Cu Kα radiation. Free lime is found in the original slag, with the

2 theta positions shifting to 37.51° (see dashed line) due to the incorporation of FeO, MnO

and MgO in the lime phase. This is in agreement with the previous studies[16,37,38]. Free

lime is completely eliminated in the modified slags. The other phases in the slags are

similar, including brownmillerite (formula Ca2(AlxFe1-x)2O5, in short C2AF, x is in range of

0-0.7))[39], belite (formula Ca2SiO4, in short C2S), monoxide (FeO and/or MgO-based

solid solution, in short RO) and zincite (ZnO). ZnO was intentionally added into the slag

as a standard oxide for mineral/phase quantification. Figure 4.4 shows the quantified

contents of minerals in the original and Al2O3 modified slags. With increasing Al2O3

additions, C2AF and C2S decrease markedly, and RO remains at an approximately constant

level (around 15 wt%). The amorphous phase, however, increases from 2.7 wt% for the

sample with 0 wt% Al2O3 addition to 38.0 wt% for the slag with15 wt% Al2O3 addition.

The elimination of free lime by adding Al2O3 is attributed to the reaction as shown in

reaction (4.1)


67

2CaO + 𝑥Al2O3 + (1 − 𝑥)Fe2O3 = Ca2(Al𝑥Fe1−𝑥)2O5 (4.1)

where x is in the range of 0-0.7[39]. From the stoichiometric consideration, to stabilize 5.6

wt% free lime, the maximum consumption of Al2O3 should be 3.57 wt%. Therefore, 5 wt%

Al2O3 addition is enough to stabilize all the free lime, which has been confirmed in the

present study (see Figure 4.3 and Figure 4.4).

Figure 4.3. XRD spectra of the original BOF slag and Al2O3 modified BOF slags as

quenched at 1600 °C. 5 A, 10 A and 15 A represent 5, 10 and 15 wt% Al2O3 modified slags.

The mass ratio of A/F in the C2AF phase with various Al2O3 addition was determined by

WDS analysis and the result is shown in Figure 4.5. It is clear that the ratio increases

significantly with increasing Al2O3 addition. Figure 4.5 also shows the thermodynamically

calculated liquidus temperature of the slags as a function of Al2O3 addition amount. By

increasing the Al2O3 addition, the liquidus temperature of the slags decreases from 1493 °C

for the original slag to 1249 °C for the 15 wt% Al2O3 modified slag. The decrease of

liquidus temperature is considered to be an important reason for the increase of the

amorphous phase in the quenched slag, as observed in our experimental study (Figure 4.4).

On the other hand, it has been reported that with the increased A/F ratio in C2AF, the

hydration reaction rate of C2AF increases[40]. Therefore, the hydration reaction can be

manipulated by controlling the A/F ratio through Al2O3 addition, tailoring the slag product

property for its high added value application.


68

Figure 4.4. Comparison of the major phases of the original BOF slag and modified slags,

as quantified by Rietveld refinement of XRD spectra. C2AF, C2S and RO represent

calcium aluminoferrite, belite, and monoxide, respectively

Figure 4.5. Change of A/F in the C2AF phase and liquidus temperature of the slags with

different Al2O3 additions

0

10

20

30

40

50

Lime

0 002.7

38.0

24.6

5.6

11.1

AmorphousROC2S

Phas

e co

nte

nt,

wt

pct

Original

5 wt pct Al2O

3 addition

10 wt pct Al2O

3 addition

15 wt pct Al2O

3 addition

C2AF

0 5 10 15

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Mas

s ra

tio

of

Al 2

O3/F

e 2O

3

Al2O

3 addition, wt pct

1250

1300

1350

1400

1450

1500

Meltin

g tem

peratu

re, C


69

Figure 4.6 shows the liquidus surface of the Al2O3-CaO-FeOx ternary system in air. The

composition of C2AF minerals generated from the original and the Al2O3 modified BOF

slags are indicated by the green dots. As shown by the arrows, the C2AF composition

moves towards lower-liquidus area with increasing Al2O3 addition, implying that the liquid

fraction in the slag at elevated temperature is increased with Al2O3 addition. The decreased

liquidus temperature of C2AF with higher mass ratio of Al2O3/Fe2O3 (A/F) is in an

agreement with the finding of the experimental study by Fukuda et al [41].

Figure 4.6. Liquidus surface in the Al2O3-CaO-FeOx system in air. Adapted from ref[42].

The green arrows indicate the evolution of C2AF with different A/F ratios caused by

Al2O3 addition. A: Al2O3; C: CaO; F: Fe2O3.

4.3.2 Construction of CCT and TTT diagrams of the Al2O3 modified BOF slag in air

Crystallization behavior of the original and the Al2O3 modified BOF slags under continuous

cooling and isothermal conditions in air has been observed in-situ using the CSLM. 15 wt%

Al2O3 modified BOF slag is taken as an example to illustrate the crystallization process.

The slag was cooled from 1450 °C with a continuous cooling rate of 2 °C∙s-1. Figure 4.7 (a)

shows the slag before cooling, where the slag is fully liquid. As temperature decreases, a


70

solid crystal appeared at 1204 °C, which is defined as the onset crystallization temperature

under such a cooling rate. The precipitated crystal then started growing, forming a larger

crystal with an equiaxial shape. According to post-mortem analysis of the quenched slag

sample, the primary crystal was identified as the C2S, which is confirmed by FactSage

calculation. As shown in Figure 4.7 (d), with further cooling, the second C2S crystal with

a similar morphology appeared in the observation view and was connected to the primary

C2S crystal. At the temperature of 1169 °C, crystals with a different morphology

precipitated from the melt, as seen in Figure 4.7 (e). Post-mortem analysis identified that

these crystals are melilite (C2FAS), which is confirmed by FactSage calculation. No

significant change was observed after temperature below 1129 °C.

Figure 4.7. Illustration of crystallization of the 15 wt% Al2O3 modified BOF slag with a

continuous cooling rate of 2 °C∙s-1 in CSLM test.


71

Figure 4.8. (a) CCT diagram of the original and the modified BOF slags. The solid and

open symbols represent the starting and ending crystallization temperatures, respectively.

T1, T2, T3 and T4 indicate the calculated liquidus temperature of original, 5 , 10 , and 15

wt% Al2O3 modified BOF slags. The line is drawn as a guide for reading. (b)

Undercooling of the original and modified BOF slags under various cooling rates. 5 A, 10

A, and 15 A indicate the 5, 10, and 15 wt% Al2O3 modified BOF slags, respectively.


72

Using the images captured from the CSLM crystallization experiments, the CCT diagrams

of the different slags (i.e. onset crystallization temperature as a function of time under

continuous cooling conditions) are constructed and given in Figure 4.8 (a). 5 A, 10 A, and

15 A refer to the 5, 10, and 15 wt% Al2O3 modified slags, respectively. In the CCT diagram,

the solid symbols and corresponding open symbols respectively represent the onset and

ending temperature of crystallization under a certain cooling rate. The narrow temperature

range between the onset and ending temperatures for each slag implies rapid crystallization

of the slags. The fast crystallization of BOF slag is not significantly changed by adding

Al2O3. For a given slag, an increase in the cooling rate results in a decrease of the onset

crystallization temperature as indicated by the solid lines in Figure 4.8 (a). The theoretical

crystallization temperatures of slags with different Al2O3 addition were calculated using

FactSage, and the results are given by the horizontal dashed lines in Figure 4.8 (a). The

liquidus temperature of the original and the Al2O3 modified slags are, respectively

1527 °C(T1, original slag), 1450 °C(T2, 5 A), 1400 °C(T3, 10 A), 1365 °C(T4, 15 A). Based

on these calculated liquidus temperature data, the undercooling (temperature difference

between the liquidus temperature and actual solidification temperature) of each slag at

various cooling rates can be estimated as shown in Figure 4.8 (b). The undercooling is

increased with increasing cooling rate and Al2O3 addition. For the original slags, by

increasing the cooling rate from 29.4 ° to 1260 °C∙min-1, the undercooling increased from

80 to 247 °C. While for the 15 wt% Al2O3 modified slag, the undercooling increased from

128 to 277 °C by increasing the cooling rate from 29.4 to 1260 °C∙min-1. According to the

classical nucleation theory, the undercooling implies the resistance of a liquid for solid

precipitate nucleation, which is much larger than the resistance of crystal growth[43]. It

has been reported that Al2O3 can increase the viscosity of the BOF slag at high temperature,

where the slag is mostly liquid[44,45]. The slag melt is assumed under the creeping flow

limit, according to the Stokes–Einstein equation[46], the diffusion constant (D) of a

spherical atom to form a nucleus is calculated as

𝐷 =𝑘𝐵𝑇

6𝜋𝜂𝑟 (4.2)

where 𝑘𝐵 is Boltzmann's constant, T the absolute temperature (K), 𝜂 viscosity of the melt

(Pa∙s), r radius of the atom. The diffusion constant is reversely proportional to the viscosity.

Therefore, a larger driving force (undercooling) is needed to form a critical nucleus in

Al2O3 modified slags.

The TTT diagram with respect to the crystallization behaviour of the present slags under

isothermal condition is presented in Figure 4.9. It is clear that the onset crystallization

temperature of the slags is remarkably decreased by adding Al2O3. For a given slag, the

incubation time of the crystallization decreases significantly with the isothermal

temperature.


73

Figure 4.9. TTT diagram of the original and modified BOF slags. 5 A, 10 A, and 15 A

indicate the 5, 10, and 15 wt% Al2O3 modified BOF slags, respectively. Tn1, Tn2, Tn3, and

Tn4 are the nose temperature of original, 5 A, 10 A and 15 A slags. The line is drawn as a

guide for reading.

According to the measured TTT diagram, the nose temperature (indicated as Tn, at which

the incubation time is shortest) was also decreased substantially with Al2O3 addition, i.e.

from 1467 °C of the original slag to 1150 °C of the slag 15 A (the nose temperature

decreased by 317 °C). The nose temperature can be used to estimate the critical cooling

rate (𝑅𝑐), which is a key parameter to examine the glass formation ability of the slag.

Theoretically, when the cooling rate is larger than 𝑅𝑐 , no crystal can form during the

cooling process[43]. 𝑅𝑐 is therefore of fundamental importance to describe the glass

formation ability. Based on the theory of crystallization kinetics, Uhlmann and Onoratoo

developed a model to estimate the critical cooling rate (𝑅𝑐) for glass formation[47]:

𝑅𝑐 =𝐴𝑇𝑙

2

𝜂𝑛𝑒𝑥𝑝(−0.212𝐵) [1 − 𝑒𝑥𝑝 (

−0.3∆𝑆𝑚

𝑅)]

3/4

(4.3)

where 𝐴= 40,000 J∙m-3∙K-1, 𝑇𝑙 is the liquidus temperature of the slag (°C), 𝜂𝑛 the viscosity

at the nose temperature (Pa∙s), R the universal gas constant (8.314 J∙mol−1∙K−1). B a

parameter to measure a kinetic barrier of forming the critical nucleus[48], and ∆𝑆𝑚 the

melting entropy of the precipitated oxide (J∙mol−1∙K−1). B can be estimated by the entropy


74

of fusion via 𝐵 ≈ 12.6 × (Δ𝑆𝑚/𝑅)[49] . After plugging the values of the relevant

parameters into Eq. (4.3), it can be simplified as


2

𝜂𝑛0.975∆𝑆𝑚[1 − 0.965∆𝑆𝑚]3/4 (4.4)

where 𝑇𝑙 and 𝑇𝑛 can be determined by the CSLM tests, 𝜂𝑛, and Δ𝑆𝑚 can be calculated by

FactSage. The obtained values of these parameters (i.e. 𝑇𝑙, 𝑇𝑛, 𝜂𝑛, and Δ𝑆𝑚) in Eq. (4.4)

are shown in Table 4.2. Thereof the critical cooling rate as a function of Al2O3 addition

amount is calculated as shown in Figure 4.10. Apparently, 𝑅𝑐 decreases dramatically with

adding Al2O3 into the slag, i.e. from 45077 °C∙s-1 of the original BOF slag to 1188 °C∙s-1

of the 15 wt% Al2O3 modified slag. A smaller 𝑅𝑐 with a lower Al2O3 slag implies that the

Al2O3 modified BOF slag is easier to be vitrified.

Table 4.2. Parameters used to calculate the critical cooling rate to produce amorphous slag.

𝑇𝑙 , °C 𝑇𝑛, °C 𝜂𝑛, Pa∙s Δ𝑆𝑚, J∙K−1∙mol−1

Original 1556 1470 0.033 54.0

5 Al2O3 addition 1406 1320 0.126 50.1

10 Al2O3 addition 1318 1230 0.379 49.6

15 Al2O3 addition 1300 1150 1.391 51.2

Figure 4.10. Change in the critical cooling rate of forming fully amorphous slag in air

with different Al2O3 additions


75

Figure 4.11. Change of the key parameters, which are applied for the calculation on the

critical cooling rate of the original and Al2O3 modified BOF slags, as a function of Al2O3

addition amount. (a) liquidus and nose temperature; (b) viscosity and melting entropy

According to Eq. (4.4), the critical cooling rate 𝑅𝑐 (better glass formation ability) is

proportional to the liquidus temperature (𝑇𝑙 ) squared and melting entropy (Δ𝑆𝑚 ), and

reversely proportional to the viscosity of the slag at the nose temperature (𝜂𝑛). As shown

in Figure 4.11 (a), both the liquidus and nose temperature of the slag was decreased by

adding Al2O3, whereas the slag viscosity 𝜂𝑛 increases remarkably (see Figure 4.11 (b)).


76

On the other hand, the melting entropy of slag (Δ𝑆𝑚) firstly decreases and then increases

slightly (see Figure 4.11 (b)). According to the results in Figure 4.11, the combined effect

of Al2O3 (i.e. increase in viscosity and decrease in liquidus temperature) is favorable to

form the glass phase in the slag. Since the changing in Δ𝑆𝑚 with Al2O3 is negligible in the

present work (see Figure 4.11 (b)) it can be concluded that the decrease in 𝑇𝑙 and increase

in 𝜂𝑛 with Al2O3 addition are the origins of the increase in the glass formation ability of the

slag.

4.3.3 Crystallization sequence during continuous cooling of the slag

To have a fundamental understanding of the effects of Al2O3 addition on the crystallization

kinetics, the crystallization sequence during continuous slag cooling was further

investigated. The in-situ observation of the mineral evolution during the solidification

process with a cooling rate of 0.98 °C∙s-1 is shown in Figure 4.12. To identify the mineral

phases, WDS analysis was performed and the microstructure of the slags after solidification

is also demonstrated in the figure.

Figure 4.12 (a-1) and (a-2) respectively exhibit the morphologies of the primary and

secondary crystals (CSLM images) during the solidification of the original BOF slag. They

were identified as C2S and C2AF, as shown in Figure 4.12 (a-3). Under the cooling rate of

0.98 °C∙s-1, the crystallization temperatures of C2S and C2AF are measured to be 1442 °C

and 1288 °C respectively. As shown in Figure 4.12 (a-1), during the crystallization process,

C2S has an equiaxed morphology and grows rapidly, forming a matrix phase of the

solidified slag. C2AF is brighter compared to the surrounding matrix. As presented in

Figure 4.12 (b) and (c), the crystallization sequence of the 5 and 10 wt% Al2O3 modified

BOF slag is similar to that of the original slag, i.e. C2S precipitates followed by C2AF. But

the precipitation temperature of both C2S and C2AF was gradually decreased by Al2O3

addition. Under the cooling rate of 0.98 °C∙s-1, the onset crystallization temperatures of C2S

and C2AF and respectively drop to 1259 °C and 1062 °C in the 10 wt% Al2O3 modified

slag. It should be noted that a grey phase was also identified as melilite in the 10 wt% Al2O3

modified slag after the CSLM test. Due to the small amount of melilite, it was not observed

in the CSLM test but it was identified in the post-mortem analysis (see Figure 4.12 (c-3)).

In general, melilite grains are solid solutions of a series of calcium silicates, including

akermanite (formula Ca2(MgSi2O7)), gehlenite (formula Ca2(Al2SiO7), in short C2AS), and

ferri-gehlenite (formula Ca2FeAlSiO7, in short C2FAS)[50,51]. WDS analysis of the slag

samples confirms that the melilite in the present study is C2FAS. Figure 4.12 (d-1 to 3)

shows the crystallization sequence of the 15 wt% Al2O3 modified BOF slag at the cooling

rate of 0.98 °C∙s-1 in air. The faceted C2S crystals precipitates at 1230 °C, followed by

melilite precipitating from 1176 °C. In the CSLM, melilite is darker than C2S. Compared

with other slags, the onset crystallization temperature of C2S is further decreased with

adding 15 wt% Al2O3, but its morphology has changed from the equiaxed to faceted shape.

Also, the secondary crystal is melilite rather than C2AF. Based on the present observation,


77

it is clear that adding Al2O3 into the slag decreases the crystallization temperature of C2S,

changes the crystal morphology and alters the crystallization sequence of the different solid

crystals during the slag solidification.

Figure 4.12. Crystallization sequence of the (a) original BOF slag, (b) 5 A, (c) 10 A and (d)

15 A slags during continuous cooling condition (0.98 °C∙s-1) in air, as observed using

CSLM (1 and 2) and backscattered electron (BSE) (3). C: CaO; S: SiO2; F: Fe2O3; A: Al2O3.


78

Figure 4.13. Comparison of crystallization sequence and slag mineralogy for original and

modified slags as experimentally determined and simulated by the equilibrium model and

the Scheil-Gulliver model. Ex: experimental results; SG: Scheil-Gulliver model; Eq:

Equilibrium model. C: CaO; S: SiO2; F: Fe2O3; A: Al2O3; M: MgO.

Thermodynamic calculation on the slag solidification process was performed by using

FactSage software. The “Equilibrium” solidification model was used for the solidification

from 1700 to 500 °C, while the “Scheil-Gulliver” model was applied for the solidification

from 1700 °C till the temperature at which the liquid slag disappears. Figure 4.13 presents

comparison of the crystallization sequence and slag mineralogy of the original, 5, 10 and

15 wt% Al2O3 modified BOF slags as experimentally determined and calculated by the

“Scheil-Gulliver” and “Equilibrium” models. ImageJ was used to obtain the mineralogical

fraction of the experimental slag samples. According to the calculation, C2S is the primary

phase in all the cases, regardless of Al2O3 addition. The secondary phase, however, differs

in different models. In the “Scheil-Gulliver” modelling, the secondary phase is,

respectively MgO (for the original BOF slag), C2AF (for the 5 and 10 wt% Al2O3 modified

BOF slags) and C2FAS (for the 15 wt% Al2O3 modified BOF slags). This simulated

crystallization sequence is in a reasonable agreement with the experimental results (see

Figure 14). In the “Equilibrium” modelling, however, the secondary phase is merwinite

(formula Ca3Mg(SiO4)2, in short C3MS2) for the original, 5 and 10 wt% Al2O3 modified

BOF slags, which is not the case with our experimental observation. With respect to the

mineralogical fraction, C2S and C2AF are the main minerals in the experimental results. As

shown in Figure 4.13, C2S and C2AF are the main minerals in “Scheil-Gulliver” model, but


79

C2S becomes a minor mineral in the “Equilibrium” model, where the main minerals are

C2AF, C3MS2 and C2FAS. Therefore, it can be concluded that “Scheil-Gulliver” model has

more correctly predicted the phase constitution than the “Equilibrium” model for the

present experimental conditions. Under the continuous cooling condition, the solid phase

can hardly participate in the reactions occurring in the remaining liquid, so that the

assumption of “Scheil-Gulliver” model is closer to the reality of the BOF slag solidification

under the present experimental condition.

4.4 Conclusions

The influence of Al2O3 additions on the mineralogy and crystallization kinetics of the BOF

slag has been systematically studied through quenching experiments and in-situ CSLM

observation. The mineral changes with Al2O3 additions were revealed. CCT and TTT

diagrams have been constructed to understand the crystallization behavior (e. g. identify

the primary and secondary crystal phase of the slag at different compositions and cooling

rates) of the original and modified BOF slags. The effect of the cooling rate on the

undercooling of slag solidification was discussed. The isothermal crystallization kinetics

was studied, and the critical cooling rate to vitrify the BOF slag was determined based on

the measured TTT diagrams. Also, the crystallization sequence during continuous cooling

has been discussed through CSLM observation combined with slag solidification

modelling. The main conclusions are:

(1) 5 percent Al2O3 addition can effectively remove free lime from the high basicity BOF

slag, and increases the liquid fraction of the slag. With further Al2O3 additions, the

amorphous phase increases while the C2S and C2AF decrease in the solidified slag. The

reason is that the liquidus temperature of the slag is decreased with increasing Al2O3

additions.

(2) Under the current continuous cooling condition, the original and modified BOF slags

crystallize rapidly. Increasing the cooling rate and Al2O3 addition amount can appreciably

increase the undercooling of the slag during solidification. The glass formation ability of

the slag is found to be enhanced by adding Al2O3. This is attributed to that Al2O3 causes a

lowering of the liquidus temperature and an increase in viscosity of the slag at the nose

temperature.

(3) To have a fundamental understanding of the effects of Al2O3 addition on the

crystallization kinetics, the crystallization sequence during continuous slag cooling was

further investigated. C2S crystals precipitated first from the slag in the continuous cooling

condition with a cooling rate of 0.98 °C∙s-1, regardless of Al2O3 addition. Yet the secondary

phase can be C2AF or C2FAS, dependent on slag composition. “Scheil-Gulliver” modelling

predicts a better result than “Equilibrium” modelling in the continuous cooling condition.


80

References

[1] www.worldsteel.org/statistics/statistics-archive/yearbook-archive.html, accssed

on 25/2/2017.

[2] www.euroslag.com/products/statistics, accessed 24/3/2017.

[3] J. M. Manso, J. A. Polanco, M. Losañez, and J. J. González: Cem. Concr.

Compos., 2006, vol. 28, pp. 528–34.


pp. 261–68.


Recycl., 2011, vol. 56, pp. 48–55.

[6] G. Wimmer, H. Wulfert, H.M. Ludwig, and A. Fleischanderl: in METEC 2nd

Eur. Steel Technol. Appl. Days, 2015, pp. 1–6.

[7] P. Xue, A. J. Xu, D. F. He, Q. X. Yang, G. Q. Liu, F. Engström, and B.

Björkman: Constr. Build. Mater., 2016, vol. 122, pp. 567–76.


Int., 2007, vol. 47, pp. 1541–48.

[9] S. A. Mikhail and A. M. Turcotte: Thermochim. Acta, 1995, vol. 263, pp. 87–

94.

[10] G. Wang, Y. H. Wang, and Z. L. Gao: J. Hazard. Mater., 2010, vol. 184,

pp. 555–60.



pp. 239–50.

[12] M. Morone, G. Costa, A. Polettini, R. Pomi, and R. Baciocchi: Miner. Eng.,

2014, vol. 59, pp. 82–90.

[13] M. Salman, Ö. Cizer, Y. Pontikes, R. M. Santos, R. Snellings, L. Vandewalle, B.

Blanpain, and K. Van Balen: Chem. Eng. J., 2014, vol. 246, pp. 39–52.

[14] H. Alanyali, M. Çöl, M. Yilmaz, and Ş. Karagöz: Int. J. Appl. Ceram. Technol.,

2009, vol. 6, pp. 736–48.

[15] T. H. Su, H. J. Yang, Y. C. Lee, Y. H. Shau, E. Takazawa, M. F. Lin, J. L. Mou,

and W. T. Jiang: Steel Res. Int., 2016, vol. 87, pp. 1511–26.

[16] C. W. Liu, M. X. Guo, L. Pandelaers, B. Blanpain, and S.G. Huang: Metall.


[17] D. Ionescu, T. R. Meadowcroft, and P. V. Barr: Adv. Cem. Res., 2001, vol. 13,

pp. 21–30.

[18] A. Rai, J. Prabakar, C. B. Raju, and R. K. Morchalle: Constr. Build. Mater.,

2002, vol. 16, pp. 489–94.

[19] S. Onisei, K. Lesage, B. Blanpain, and Y. Pontikes: J. Am. Ceram. Soc., 2015,

vol. 9, pp. 1–9.

[20] J. Provis and J. Van Deventer: Alkali Activated Materials: State-of-the-Art

Report, RILEM TC 224-AAM, Springer Science & Business Media, 2013.

[21] J. N. Murphy, T. R. Meadowcroft, and P. V. Barr: Can. Metall. Q., 1997,

vol. 36, pp. 315–31.

[22] A. Srinivasa Reddy, R. K. Pradhan, and S. Chandra: Int. J. Miner. Process.,

2006, vol. 79, pp. 98–105.




81

[24] J. J. Liu, G. Chen, P. C. Yan, B. Blanpain, N. Moelans, and M. X. Guo: Am. J.

Sci., 2014, vol. 402, pp. 1–8.

[25] I. Sohn and R. Dippenaar: Metall. Mater. Trans. B, 2016, vol. 47, pp. 2083–94.

[26] Y. Kashiwaya, T. Nakauchi, K. S. Pham, S. Akiyama, and K. Ishii: ISIJ Int.,

2007, vol. 47, pp. 44–52.

[27] Y. Q. Sun, Z. T. Zhang, L. L. Liu, and X. D. Wang: Energies, 2014, vol. 7,

pp. 1673–84.

[28] Y. Q. Sun, H. W. Shen, H. Wang, X. D. Wang, and Z.T. Zhang: Energy, 2014,

vol. 76, pp. 761–67.

[29] L. J. Zhou, W. L. Wang, F. J. Ma, J. Li, J. Wei, H. Matsuura, and F. Tsukihashi:

Metall. Mater. Trans. B, 2011, vol. 43, pp. 354–62.

[30] Y. Kashiwaya, C. E. Cicutti, A. W. Cramb, and K. Ishii: ISIJ Int., 1998, vol. 38,

pp. 348–56.

[31] L. J. Zhou, W. L. Wang, R. Liu, and B. G. Thomas: Metall. Mater. Trans. B

Process Metall. Mater. Process. Sci., 2013, vol. 44, pp. 1264–79.

[32] B. B. Jiang, W. L. Wang, I. Sohn, J. Wei, L. J. Zhou, and B. Z. Lu: Metall.

Mater. Trans. B, 2014, vol. 45.

[33] J. J. Liu, M. X. Guo, P. T. Jones, F. Verhaeghe, B. Blanpain, and P. Wollants: J.

Eur. Ceram. Soc., 2007, vol. 27, pp. 1961–72.

[34] J. Heulens, B. Blanpain, and N. Moelans: Chem. Geol., 2011, vol. 290, pp. 156–

62.

[35] C. W. Bale, P. Chartrand, S. A. Decterov, G. Eriksson, K. Hack, R.Ben.

Mahfoud, J. Melancon, A. D. Pelton, and S. Petersen: Calphad, 2002, vol. 26,

pp. 189–228.

[36] D. Durinck, P. T. Jones, B. Blanpain, P. Wollants, G. Mertens, and J. Elsen: J.

Am. Ceram. Soc., 2007, vol. 90, pp. 1177–85.

[37] I. Nishinohara, N. Kase, H. Maruoka, S. Hirai, and H. Eba: ISIJ Int., 2015,

vol. 55, pp. 616–22.

[38] E. H. Swanson, H. E., Morris, M. C., Stinchfield, R. P. and Evans: Standard X-

Ray Diffraction Powder Patterns, U.S. Department of Commerce, National

Bureau of Standards, Washington, DC, 1962.



[40] S. Elalfi, S. Abdaleem, H. Eldidamony: Indian J. Eng. Mater. Sci., 2001, vol. 8,

pp. 292–96.

[41] K. Fukuda, T. Bessho, K. Matsunaga, and H. Yoshida: 2004, vol. 34, pp. 1535–

40.

[42] V. D. Eisenhuttenleute: Slag Atlas, Verlag Stahleisen GmbH, Dusseldorf, 1995.

[43] D. Turnbull: Contemp. Phys., 1969, vol. 10, pp. 473–88.

[44] S. H. Seok, S. M. Jung, Y. S. Lee, and D. J. Min: ISIJ Int., 2007, vol. 47,

pp. 1090–96.

[45] Z. Z. Liu, L. Pandelaers, P. T. Jones, B. Blanpain, and M. X. Guo: in Adv.

Molten Slags, Fluxes, Salts Proc. 10th Int. Conf. Molten Slags, Fluxes Salts

2016, pp. 439–46.

[46] A. Einstein: Ann. Phys., 1905, vol. 322, pp. 549–60.

[47] D. R. Uhlmann and P. I. K. Onoratoo: in 10 Th Proc. Lunar Planet. Sci. Conf.,

1979, pp. 375–81.

[48] H. Yinnon and D.R. Uhlmann: J. Non. Cryst. Solids, 1981, vol. 44, pp. 37–55.


82

[49] D. R. Uhlmann, H. Yinnon, and C. Y. Fang: in Lunar Planet. Sci. Conf. Proc.,

1982, pp. 281–88.

[50] F. F. Foit, R. L. Hooper, and P. E. Rosenberg: Am. Mineral., 1987, vol. 72,

pp. 137–47.

rameters were used to calculate the apparent activation energy and estimate the

hydrothermal stability of a number of 3Y-TZP ceramics. The conventional method

of

Materials and methods

83

Chapter 5

Optimization of mineralogy and microstructure of solidified BOF slag

through SiO2 addition or atmosphere control during hot-stage slag

treatment

A part of this Chapter has been published in the Journal of Metallurgical and Materials

Transactions B, 2016, 47(6): 3237-3240, Chunwei Liu; Muxing Guo; Lieven Pandelaers;

Bart Blanpain; Shuigen Huang.

Department of Materials Engineering, KU Leuven, Kasteelpark Arenberg 44, BE-3001

Leuven, Belgium

DOI: 10.1007/s11663-016-0809-4

Abstract

Valorization of basic oxygen furnace (BOF) slag is of significant importance for mitigation

of the steel production’s environmental impact. The present work aims to investigate the

influence of SiO2 addition and oxygen partial pressure on the mineralogical modification

of a typical industrial BOF slag. The slag basicity (mass ratio of CaO/SiO2) was varied

from 1.8 to 4.4 by mixing specific amounts of SiO2 with the master BOF slag. The original

and modified slags were re-melted and solidified under argon and/or air atmosphere

followed by slow cooling. The experimental observations were then compared with the

results of thermodynamic modelling to achieve a thorough understanding. With decreasing

the basicity, free lime is eliminated as it forms dicalcium silicate (Ca2SiO4). With

increasing the oxygen partial pressure, wustite is oxidized to hematite, which combined

with free lime to form calcium aluminoferrite (C2AF). The effects of SiO2 addition and

oxygen partial pressure were finally evaluated with respect to the energy consumption for

the BOF slag valorization. The modified slag might be suitable as a precursor for value-

added applications.

Key words: BOF slag; Microstructure; Valorization; Basicity; Oxygen partial pressure;

Contribution of Chunwei Liu: The experiments were performed by Shuigen Huang and

Chunwei Liu. Chunwei Liu analysed the samples, interpreted the results and wrote the

paper. The co- authors assisted in discussion of the results and writing process.

Chapter 5. Effect of SiO2 addition and atmosphere

84

5.1 Introduction

Steel slag, generated during the steelmaking process, is a main by-product in steel industry.

Depending on the steel grade, 100 - 150 kg slag is discharged for producing 1 tonne of

steel [1]. About 20 million tonnes of steel slag are produced annually in Europe [2], of

which BOF slag accounts for almost half of the volume [3]. Storage of the huge amount

of slag burdens the environment and is an economic liability to steel industry. Recycling

and reutilization of slags towards added-value applications are therefore of significant

importance for the sustainability of steel industry.

In the past decades, metallurgical slag valorization has been intensively explored to apply

the slag in various fields, such as landfill liner [4,5], hydraulic binder [6–8], and

fertilizer [9]. Among these, hydraulic binder that can be used for constructional application

appears to be more interesting as they create higher value. However, application of BOF

slag for constructional materials is limited by its volume expansion during natural

aging [10–12]. This dimensional instability is induced by the hydration of free lime and

free magnesia in the slag [11]. During the hydration, free lime and magnesia exhibit around

10 pct swelling [13], leading to the disintegration of bulk slags. Typically, the lime content

in the BOF slag is markedly higher than that of magnesia, thus the lime is the main concern

for the volume instability [10]. Recent research addressed that 4 wt% free lime in BOF slag

is the limit for applying the slag as a construction material [10]. Different methods have

been investigated to stabilize free lime in BOF slag, amongst which, SiO2 and Al2O3 were

reported as effective additions to eliminate the free lime [14–16]. The addition of SiO2

results in more silicate after solidification, which accommodates the extra CaO and

therefore eliminates the free lime, while the addition of Al2O3 produces more C2AF as it

its combines with free lime [15]. Also, Gautier et al. studied the effect of the cooling

method on slag microstructure and mineralogy in air, and found that the free lime content

was decreased in the solidified slag [17]. The elimination of free lime was tentatively

attributed to the suppression of the decomposition of tricalcium silicate by a rapid cooling.

Additionally, hot-stage carbonation of the free lime has been investigated in recent

years [3,11,18]. The mechanism is to stabilize the free lime by forming geochemically

stable calcium carbonate. But size reduction of BOF slag is needed to perform an effective

carbonation, which inhibits the application of this method [11,18]. By summarizing

previous studies, it can be concluded that: (1) recycling BOF slag has shown significant

importance but high value-added application is limited due to the presence of free lime; (2)

SiO2 addition is effective to stabilize the slag, yet quantitative knowledge on the

mineralogical evolution associated with SiO2 addition is needed; (3) although previous

study provided evidence that an oxidizing atmosphere may facilitate slag stabilization, the

underlying mechanisms remain unclear.

The present study aims to provide a fundamental understanding of the effects of SiO2

addition and oxygen partial pressure on BOF slag microstructure and minerology after


85

solidification, targeting a quantitative and effective control of the slag product that is

microstructurally and mineralogically tailored to achieve better hydraulic properties.

Experiments were carried out to assess the influence of basicity by mixing specific amounts

of SiO2 with the master BOF slag. Also, the original and modified BOF slags were re-

melted and solidified under argon and/or air atmosphere followed by slow cooling.

Experimental results were compared with the thermodynamic calculation using FactSage

to reveal the underlying mechanism. Thereafter, the promising methods were evaluated

from the perspective of energy consumption. The findings of this work suggest a novel

approach to optimize the microstructure/mineralogy of the solidified slag through hot-stage

slag engineering, providing a precursor for high value-added applications.

5.2 Experiments and thermodynamic modelling

5.2.1 Materials preparation and experimental procedure

A typical BOF slag was used as the starting material, which was sampled from a slag yard

at different locations and depth from the surface. The slag was then mixed, ground and

milled to powder below 200 µm. The typical chemical composition and mineral

constitution of the BOF slag are listed in Tables 5.1 and 5.2 respectively. Concentration of

Fe2+ and Fe3+ in the slag were measured using chemical titration by potassium dichromate.

The basicity (mass ratio of CaO to SiO2) of the master slag is 4.4.

Table 5.1. Chemical composition of the master slag, wt%


44.5 20.6 10.2 10.4 10.14 4.78 2.18 2.05 2.43 0.89 0.37 * Fe is given as the total iron. Measured by X-ray fluorescence (XRF, Panalytical PW 2400).

Table 5.2. Mineral composition of the master slag

Mineral phase Free

lime Wustite

Calcium

aluminoferrite

Beta dicalcium

silicate

Chemical

formula CaO

RO (FeO-MgO-

MnO) Ca2(Al, Fe)2O5 β-Ca2SiO4

(wt%) 15-22 15-25 25-35 25-35

To investigate the effect of basicity on the solidification microstructure and mineralogy,

100 g of master slag was individually mixed with 5.0, 10.0, and 15.0 grams of SiO2 (Sibelco,

Belgium) in ethanol for 24 hours by using a multidirectional mixer (Turbula type). The

mixtures were dried in a rotating evaporator at 65°C. 10 grams of the mixture with different


86

basicity was loaded in a MgO crucible (inner diameter 15mm, height 40mm) and melted

in a graphite heating furnace (W100/150-2200-50 LAX, FCT Systeme, Rauenstein,

Germany) under vacuum atmosphere at 1600 °C for 1 hour to homogenize the slag

composition, afterwards the slag was cooled down to room temperature inside the furnace

at a cooling rate of 5 °C∙min-1.

To study the effect of oxygen partial pressure, 20 grams of master slag were loaded in MgO

crucibles (inner diameter: 32 mm, height 70 mm). For the tests under Ar, the sample was

placed in a vertical tube furnace (HTRV 100-250/18, GERO, Germany) under Mg-purified

Ar atmosphere with a flow rate of 0.4 l∙min-1. After maintaining at 1400 °C for 1 hour, the

slag was cooled down to room temperature in the furnace at 5 °C∙min-1. The oxygen partial

pressure was measured by an oxygen sensor (Rapidox 2100, Cambridge Sensotec, United

Kingdom) and was maintained below 10-18 atm. For the tests in air, the slag was melted at

1400 °C for 1 hour in a bottom loading furnace (AGNI - ELT 160-02. Spring type), then

cooled down to room temperature in the furnace at 5 °C∙min-1.

5.2.2 Characterization of the slag samples

The slag specimens were mounted in a low viscosity resin (Epofix), ground by silicon

carbide papers and polished with diamond paste. The polished specimens were coated with

carbon for compositional and microstructural analyses, which was achieved by electron

probe micro-analysis (FE-EPMA, JXA-8530F, JEOL Ltd, Japan), equipped with a

wavelength dispersive spectroscopy (WDS) that allows fully quantitative chemical

composition analysis. The beam current and accelerating voltage were set as 15 nA and 15

kV, respectively. After crushing the slags to powder with size below 100 µm, X-Ray

Fluorescence (XRF, Panalytical PW 2400) was used to measure the composition while X-

Ray Diffraction (XRD, 3003-TT, Seifert, Ahrensburg, Germany) was used to identify the

minerals. Quantitative analysis of the XRD pattern was achieved with the aid of the

Rietveld refinement method.

5.2.3 Thermodynamic modelling

Thermodynamic modelling of the phase evolution during solidification were performed by

using FactSage 7.0 associated with FactPS and FToxid databases. The FactSage software

has been successfully used in modelling the slag system in recent years [19–22].

Experimental observation was integrated with the simulation to understand the effect of

oxygen partial pressure on the BOF slag mineralogy.


87


5.3.1 Influence of SiO2 addition

Figure 5.1 presents the quantitative XRD analysis of the slags with different SiO2 additions

treated at 1600 °C under Ar atmosphere by means of Rietveld refinement. Major phases

presented in these slags are calcium aluminoferrite, beta dicalcium silicate (β-C2S), wustite

(RO) and free lime (formula CaO). The formula of calcium aluminoferrite is Ca2(AlxFe1-

x)2O5 (in short C2AF), with x in the range of 0-0.7 [23]. There is 20 wt% free lime detected

in the master slag with a basicity of 4.4. With SiO2 addition of 5, 10 and 15 wt%, the

basicity was accordingly decreased to 2.8, 2.4, and 1.8, respectively. The content of free

lime is significantly lowered with the decrease of basicity, as indicated by the solid line in

Figure 5.1. In the slags with basicity of 2.8, the content of free lime is decreased to 1.9

wt%, which is less than the critical limit (4 wt%) for the slag application as construction

materials [10]. While in the slags with basicity of 2.4 and 1.8, free lime is eliminated

completely. In contrast, the content of C2AF phase is decreased with the addition of SiO2

due to the strong affinity of SiO2 with CaO at high temperature. In the slag with a basicity

of 1.8, bredigite (formula Ca1.75Mg0.25SiO4, in short CMS) indicated by the dashed line

becomes a major phase, and accordingly β-C2S phase is decreased drastically compared to

other minerals. Chemical reactions of the free lime stabilization and bredigite formation

are respectively given in reactions (5.1) and (5.2). According to the mass balance

calculation with reaction (5.1), 0.54 wt% SiO2 is consumed to stabilize 1 wt% free lime.

Thus, stabilization of 20 wt% free lime in the present study consumes 10.7 wt% SiO2. In

practice, however, the actual consumption of SiO2 should be less than 10.7 wt% due to the

fact that CaO can also react with FeO, MnO and MgO and form RO as seen in reaction

(5.3). For this reason, free lime was completely eliminated even in the test with 10 wt%

SiO2 addition (basicity of 2.4).

2 CaO + SiO2 = Ca2SiO4 (5.1)

1.75 CaO + SiO2 + 0.25 MgO = Ca1.75Mg0.25SiO4 (5.2)

CaO + FeO + MnO + MgO → RO (5.3)

According to the mass balance calculation with reaction (5.1), 0.54 wt% SiO2 is consumed

to stabilize 1 wt% free lime. Thus, stabilization of 20 wt% free lime in the present study

consumes 10.7 wt% SiO2. In practice, however, the actual consumption of SiO2 should be

less than 10.7 wt% due to the fact that CaO can also react with FeO, MnO and MgO and

form RO as seen in reaction (5.3). For this reason, free lime was completely eliminated

even in the test with 10 wt% SiO2 addition (basicity of 2.4).


88

Figure 5.1. Quantitative XRD analysis of minerals content in the slag treated at 1600 °C

for 1 hour and cooled at 5 °C∙min-1 in Ar

To have a quantitative understanding of the chemical composition of each phase, the

microstructure of the solidified slag was analyzed using EPMA-WDS. Figure 5.2 shows

the typical microstructures of the slags with different SiO2 additions. The microstructure

of the original slag is shown in Figure 5.2 (a). WDS analysis confirms that free lime is a

CaO-based solid solution which also incorporates FeO, MgO and MnO. The typical CaO

content in the lime phase is above 68 wt%. β-C2S crystal can be observed and constitutes

one of the major phases in the slags, except for the one with a basicity of 1.8 (Figure 5.2

(d)). WDS analysis suggests that the β-C2S phase is typically associated with P2O5 (>2

wt%), forming a solid solution of Ca2SiO4-Ca3PO4, in short C2S-C3P [24–26]. It has been

reported that [PO4]3- acts as a substitution of [SiO4]4− anion groups in the C2S-C3P solid

solution [24]. According to previous studies, incorporation of 0.5 wt% P2O5 can stabilize

β-C2S, and this phase is crucial to prepare Portland cement [27,28]. RO (a bright phase in

Figure 5.2) is another main phase in all the cases. The RO phase is a FeO-based solid

solution with a typical FeO content over 60 wt%, and other oxides found in RO are MgO,

MnO and CaO. It is noteworthy that two typical morphologies of RO can be found: small

(<10 µm) grains dispersed in C2S-C3P and free lime (indicated by circles in Figure 5.2);

larger (>20 µm) RO grains that were separated from other phases. In the slag with basicity

of 1.8, a large amount of bredigite (CMS) was formed (see Figure 5.2 (d)). The chemical

formula of CMS (Ca1.75Mg0.25SiO4) is close to the C2S (Ca2SiO4), with a few CaO

substituted by MgO. But CMS has a distinct structure compared to the C2S, and the

impurities in the slag such as Mn can partially replace Mg [29,30]. It has been reported that


89

the bredigite reacts slowly with water, and the compressive strength of lime-activated

bredigite cubes is too low to be of use for a cementitious material [30]. Therefore, it should

be avoided to generate CMS for utilizing the slag as a binder in constructional application.

Furthermore, both the mineralogical and microstructural analysis show that no free MgO

exists in the slag. WDS analysis confirms that Mg incorporates with RO and C2S phases in

the slags with R=2.4, 2.8 and 4.4. In the case of R=1.8, however, Mg concentrates in RO,

C2S and bredigite. Free MgO can cause slag instability by its hydration, thus, it is preferred

to avoid free MgO for preparing binder materials.

Based on the above findings, it is confirmed that free lime is effectively eliminated by

forming C2S or CMS through SiO2 additions. In the current study, by decreasing the

basicity to 2.8, the free lime has been lowered to a level (1.9 wt%) that satisfies the

requirement of the volume stability for constructional application. Decreasing the basicity

to 1.8, however, leads to a precipitation of CMS phase with poor cementitious property.

Therefore, to stabilize free lime and obtain a better cementitious property of the BOF slag,

the slag should be modified at hot-stage towards a basicity ranging from 1.8 to 2.8.

Figure 5.2. Microstructures of slags treated at 1600 °C for 1 hour and cooled at 5 °C∙min-

1 in Ar. (a) R=4.4 (matrix slag); (b) R=2.8; (c) R=2.4; (d) R=1.8. 1: β-C2S; 2: C2AF; 3:

RO; 4: CaO; 5 CMS.


90

5.3.2 Influence of oxygen partial pressure

The original BOF slag was treated at 1400 °C for 1 hour under a controlled atmosphere,

followed by furnace cooling at 5 °C∙min-1. Figure 5.3 (a) shows the XRD patterns of the

air and Ar-treated slags. The characteristic 2 theta values of free lime are 37.34° and 53.86°

using Cu Kα radiation. In this work, the 2 theta positions of free lime shifted to the

diffraction angles of 37.51° and 54.16° (dashed lines in Figure 5.3 (a)). This is probably

due to the incorporation of MgO and FeO in the lime phase, which is commonly observed

in BOF slag [31]. In the present study, 1.60 and 18.98 wt% MgO and FeO were identified

in the lime phase. The interplanar distance of MgO and FeO are relatively smaller than that

of CaO, so that the peaks shift to larger 2 theta positions [32]. Free lime was detected in

the Ar-treated slag, but was eliminated completely in the air-treated slag. The other major

phases, i.e. C2AF, β-C2S and RO are observed in both the Ar and Air treated slags.

Quantitative analysis of the mineralogical composition was achieved by using Rietveld

refinement method, and the result is shown in Figure 5.3 (b). The predominant phase in the

Ar-treated slag is β-C2S (>40 wt%). But C2AF becomes the predominant phase (>65 wt%)

in the air-treated slag. Consequently, the amount of RO and β-C2S phases in the air-treated

slag is considerably lower than in the Ar-treated slag. Additionally, 3.8 wt% MgO was

found in the air-treated slag.

Figure 5.3. (a) XRD spectra of the slags treated under Ar (top) and air atmosphere

(bottom) (dashed lines represent the characteristic position of free lime); (b) Comparison

of the minerals in the air-treated and Ar-treated slags, as quantified by Rietveld

refinement of XRD spectra

The microstructure of the air and Ar-treated slag is shown in Figure 5.4. The size of C2AF

crystal is much larger in the air treatment condition. The WDS analysis confirms that the

free lime, RO, and periclase are solid solutions. In the air-treated slag, only small RO grains

are observed and those grains are dispersed in the periclase, as shown in Figure 5.4 (a). The


91

RO phase in the Ar-treated slag, as shown in Figure 5.4 (b), appears in two different

morphologies, e.g. big RO grains as a separated crystal, and tiny grains dispersed inside

the lime solid solution (indicated by the circles in Figure 5.4 (b)). Figure 5.5 shows the

phase diagram of CaO-FeOx system in contact with iron. The system is under a low oxygen

partial pressure that is comparable to the present study. The maximum solubility of FeOx

in the lime solid solution is 12 wt% when the system is equilibrated at 1130 °C [33]. WDS

analysis indicates that 19.0 wt% FeOx was incorporated in the overall lime phase, resulting

in the precipitation of FeOx (RO). According to Figure 5.5, FeOx should precipitate from

the lime rich CaO-FeOx solid solution in the temperature range of 1130 to 1045 °C. The

periclase is a MgO-based solid solution with a typical MgO content over 70 wt%. WDS

results indicate that the overall MgO content in the slag treated in air is higher than that in

Ar (5.61 wt% to 3.79 wt%), implying a slight dissolution of the MgO crucible into the air-

treated slag. This is probably due to the increase of Fe2O3 (oxidation of FeO under air

treatment condition), which behaves as an acidic substance and facilitates the dissolution

of MgO. In the industrial practice, there will be less MgO pickup because of the lower

surface to volume ratio and relatively short duration of the treatment.

Figure 5.4. Microstructures of the master slag treated under (a) air and (b) Ar atmosphere.

Circles represent the dispersed RO in lime and MgO matrix. The yellow rectangles

represent local magnification. (1) C2AF; (2) β-C2S; (3) RO (the shining points), (4) CaO

(the grey matrix); (5) MgO (the black matrix). Scale bar is 10 µm. 1: C2AF; 2: β-C2S; 3:

RO; 4: CaO; 5: MgO.


92

Figure 5.5. Phase diagram of CaO-FeOx system in contact with iron, adapted from

Ref [33]. L-liquid; ss-solid solution.

Figure 5.6 (a) and (b) present the phase evolution respectively for air and Ar-treated slags,

as predicted by FactSage equilibrium calculations. The C2S transforms from β-type to γ-

type in the FactSage calculation, since the stabilization of β-C2S by phosphorus is not

considered. To avoid such a discrepancy, all the C2S is given in the form of β-C2S in Figure

5.6. Figure 5.6 (c) indicates that the calculated quantities of major phases agree well with

the experimental results. According to the calculation, the change of the oxygen partial

pressure affects the crystal precipitation sequence. C2S is the first crystal precipitated in

the air-treated slag (at 1605 °C), followed by C3S (at 1435 °C) and C2AF (at 1365 °C). In

the Ar-treated slag, however, CaO will firstly precipitate at 1690 °C, followed by C3S (at

1500 °C), MgO (1440 °C) and C2S (1352 °C), respectively. Durinck et al have reported on

the influence of the oxygen partial pressure on the crystallization sequence in the EAF

slag [34], and found that the C2S precipitation temperature in air-cooled slag was higher

than that in Ar-cooled slag. In the present work, above 1500 °C, 23.1 wt% C2S is generated

in air-treated slag, but no C2S precipitated from the slag melted in Ar until 1352 °C.

According to the calculation, a large amount of free lime is formed by the decomposition

of C3S between 1400 and 1300 °C, as expressed by reaction (5.4).

Ca3SiO5 = CaO + Ca2SiO4 (5.4)


93

Figure 5.6. Phase evolution during solidification as calculated by FactSage under (a) Ar

and (b) air atmosphere; (c) comparison between the experimental and calculated results


94

The amount of C3S generated in the air-treated slag is decreased markedly due to the

decrease in basicity after the formation and precipitation of C2S, therefore the CaO formed

by C3S decomposition is also decreased. Additionally, as indicated by the circles in Figure

5.6 (a) and (b), C2AF starts precipitating at a much higher temperature in air (1370 °C) than

in Ar (1180°C). This means in the case of slag treatment under air, C2AF has more time to

grow after its precipitation during cooling, which explains the larger C2AF crystals

observed in Figure 5.4 (b) than that in Ar-treated slag. Also, the calculated amount of C2AF

is much higher in air than in Ar condition, which agrees with the experimental data. The

large amount of C2AF generated under air condition consumes more Ca2+ and contributes

to the removal of free lime in the slag.

The elimination of CaO, the decrease of the RO content, and the concurrent increase in

C2AF content for the air-treated slag can be better understood from reactions (5.5) and (5.6).

When the slag is heated up in air, Fe2+ in the slag (FeO) is oxidized to Fe3+ (Fe2O3), as

expressed by reaction (5.5).

2 FeO + 0.5 O2 = Fe2O3 (5.5)

Fe2O3 then reacts with CaO at around 1370°C to produce C2AF through the reaction (5.6)

during solidification process,

2 CaO + x Al2O3 + (1-x) Fe2O3 = Ca2(Alx, Fe1-x)2O5 (5.6)

Assuming that the contribution of Al2O3 in reaction (5.6) was negligible and FeO can be

oxidized completely during the treatment, stabilization of 1 mole CaO consumes 1 mole

FeO, i.e. stabilizing 1 wt% CaO needs 1.28 wt% FeO. Therefore, 25.6 wt% FeO is needed

to completely remove 20 wt% of the free lime in the original BOF slag. In the BOF slag,

however, due to the presence of Al2O3 and Mn2O3 (respectively 5.49 and 3.76 wt% in our

study), and the presence of FeO in the lime phase, stabilization of 1 wt% free lime may

require less than 1.28 wt% FeO.

5.3.3 Evaluation for lime stabilization of BOF slag with respect to energy consumption

The current BOF slag stabilization treatment lies in the introduction of additives and/or

increasing oxygen partial pressure to remove the free CaO by reactions towards a stable

matrix of calcium silicates and/or ferrites. Based on the results we obtained in

aforementioned sections, the effective stabilization methods are evaluated with respect to

energy consumption. The energy consumption during the hot-stage engineering is

estimated under assumptions: (1) during treatment, the melt temperature was fixed at

1600 °C, i.e. the energy needed to maintain such a temperature during the treatment is

calculated; (2) 20 tonne slag is charged to a slag pot at 1600 °C, which can be simplified

to a circular truncated cone with upper diameter 3.5 m, lower diameter 2.5 m and height

3.5 m; (3) the treatment time is 10 min.


95

(a) Energy consumed during the stabilization by lowering the slag basicity (through SiO2

addition)

Stabilization of BOF slag by adding SiO2 can effectively decrease the basicity and

eliminate free lime by forming C2S. As suggested in the present study, 5 wt% SiO2 addition

is sufficient to stabilize the free lime. To stabilize 20 tonne slag, 1 tonne SiO2 is injected

for the treatment in 10 min. The energy required for heating up the SiO2 from 20 to 1600 °C

can be calculated by Eq. (5.7)

𝑄1 = 𝐶𝑃𝑠𝑖 ∙ 𝑚𝑠𝑖

∙ ∆𝑇 (5.7)

where 𝐶𝑃𝑠𝑖 is the heat capacity of SiO2 (0.719 J∙g-1∙K-1) [36]; 𝑚𝑠𝑖

is the mass of SiO2 (kg).

The enthalpy changed (∆𝐻1600℃) by the formation of C2S through the reaction of CaO with

SiO2 (reaction (1)) is -1.76×104 J∙mol-1, as calculated by FactSage. The negative sign “-”

represents the reaction is exothermic.

Thus, energy released during the reaction is

𝑄2 = ∆𝐻1600℃ ∙ 𝑛𝑆𝑖𝑂2 (5.8)

where 𝑛𝑆𝑖𝑂2 is the molar number of SiO2 (mol).

The energy loss due to the radiation can be calculated by Eq. (5.9)

𝑄3 = 𝐸 ∙ 𝐴 ∙ 𝑡 (5.9)

where A is the surface area of the slag pot (top and inclined surface, m2), t is the processing

time (s). 𝐸 = 𝜎 ∙ 𝑇4 is the radiation of the slag (W∙m-2), where 𝜎=5.67×10-8 W∙m-2∙K-4 is

the Stefan-Boltzmann constant [37], T is the absolute temperature (K).

The convective contribution to the heat loss is

𝑄4 = 𝛼 ∙ 𝐴 ∙ 𝑡 (5.10)

where 𝛼 = 50 W∙m-2∙K-1 is the convective coefficient of the air [38].

By substituting values of parameters into the Eqs. (5.7) to (5.10), the total energy

exchange 𝑄 during the slag treatment is obtained:

𝑄 = 𝑄1 + 𝑄2 + 𝑄3 + 𝑄4 = 1.79 × 1010𝐽 (5.11)

(b) Energy consumed during the stabilization by oxidizing the slag (through oxygen partial

pressure control)


96

By increasing the oxygen partial pressure, the wustite is oxidized to hematite, and then

stabilizes the free lime by forming C2AF. Similarly, 20 tonne slag (with 15 wt% FeO) is

assumed to be charged in a slag pot at 1600 °C. The energy required for heating up the

oxygen is negligible compared to other energy consumptions. The enthalpy change during

the oxidation from FeO to Fe2O3 can be calculated by FactSage,

Δ𝐻1600℃, = −1.46 × 105 𝐽 ∙ 𝑚𝑜𝑙−1 (5.12)

So that the energy released during the oxidation can be calculated as

𝑄1, = Δ𝐻1600℃

, ∙𝑛𝐹𝑒𝑂

2 (5.13)

where 𝑛𝐹𝑒𝑂 is the molar number of FeO. Due to the Al2O3 content in the slag is much lower

than that of FeO and Fe2O3, χ = 0.1 is assumed in the Ca2(Alx, Fe1-x)2O5. Thus, the enthalpy

change of reaction (6) can be calculated by FactSage,

Δ𝐻1600℃,, = 1.23 × 105 𝐽 ∙ 𝑚𝑜𝑙−1 (5.14)

Thereby the energy change of stabilizing all the free lime becomes

𝑄2, = ∆𝐻1600℃ ∙

𝑛𝐶𝑎𝑂

2 (5.15)

where 𝑛𝐶𝑎𝑂 is the molar number of the free lime. Considering heat loss due to the radiation

and convection, which can be estimated by Eq. (5.9) and (5.10) respectively,

the total energy change 𝑄, in the slag stabilization treatment can be calculated in Eq. (5.16)

𝑄, = 𝑄1, + 𝑄2

, + 𝑄3 + 𝑄4 = 1.34 × 1010𝐽 (5.16)

Compared with SiO2 addition, energy consumption during the stabilization treatment by

oxidizing slag under high oxygen partial pressure is lower (1.34×1010 to 1.79×1010 J). To

have a thorough evaluation, following aspects should also be considered.

(1) From the materials consumption point of view, reducing the basicity of the BOF slag

consumes SiO2 (5 wt% in the present study) while changing oxygen partial pressure

consumes only air. So that stabilization of the BOF slag by oxidation could be more

material-effective.

(2) In view of the practical operation, injection of SiO2 or air during hot-stage treatment

may increase the viscosity due to the temperature drop, deteriorating the reaction kinetics

and the operation. The dependence of reaction kinetics on the slag viscosity and

temperature should be further investigate, i.e. dissolution of SiO2 and reaction between

SiO2 and free lime vs oxidation of FeOx and reaction between Fe2O3 and free lime.


97

(3) SiO2 addition results in more C2S while air atmosphere generates more C2AF. Both C2S

and C2AF phases facilitate the quality improvement of the slag products. The product

quality with respect to volume stability and cementitious property of the stabilized slag

should be further studied.

5.4 Conclusions

Effects of SiO2 addition and oxygen partial pressure on the microstructure and minerology

of the solidified BOF slag has been studied. The mineralogical change over SiO2 additions

was discussed in a quantitative way. FactSage calculation was performed to explore the

mechanisms of oxygen partial pressure on stabilizing the slag. Finally, the methods of SiO2

and/or oxidizing treatments were evaluated by comparing the energy consumption during

hot-stage engineering. The main conclusions are summarized as follows:

(1) Free lime can be transformed to C2S or CMS by lowering basicity through SiO2 addition.

At the basicity of 2.8, free lime can be controlled in a satisfied level to reutilize the slag in

constructional application. With too low basicity (R<1.8 for current case), however, large

amount of CMS with poor cementitious property is formed. Based on the results of this

work, an optimized basicity is suggested in the range from 1.8 to 2.8.

(2) Laboratory experiment and FactSage calculation demonstrate that free lime can be

removed during hot-stage treatment under a high oxygen partial pressure condition. The

reason is attributed to that under high oxygen partial pressure (“air” in current study), FeO

can be oxidized to Fe2O3 and further consumes free lime to form C2AF. To eliminate free

lime completely, the theoretical mass ratio of free lime to FeO should be less than 1.28,

which is satisfied in the as-delivered BOF slag.

(3) Energy efficiency for hot-stage slag stabilization by SiO2 addition and slag oxidation

has been evaluated. The calculation of energy consumption of the slag treatment indicates

that stabilization of the free lime through the slag oxidation is more energy-effective. A

comprehensive understanding is still needed to further evaluate the stabilization

approaches, e.g. the effect of slag viscosity and temperature on the reaction kinetics of the

two methods; comparison of volume stability and cementitious property of the slags

stabilized by the two methods.


98

References


R. D. Bigham, M. Connolly, B. Lee, T. O. Shepker, and M. a. Zak: Environ. Sci.

Technol., 2000, vol. 34, pp. 1576–82.



vol. 59, pp. 82–90.

[4] L. Andreas, S. Diener, and A. Lagerkvist: Waste Manag., 2014, vol. 34, pp. 692–

701.

[5] Inga Herrmann, Lale Andreas, Silvia Diener, and Lotta Lind: Waste Manag. Res.,

2010, vol. 28, pp. 1114–21.

[6] A. Srinivasa Reddy, R. K. Pradhan, and Sanjay Chandra: Int. J. Miner. Process.,

2006, vol. 79, pp. 98–105.



[8] J.N. Murphy, T.R. Meadowcroft, and P.V. Barr: Can. Metall. Q., 1997, vol. 36,

pp. 315–31.


[10] G. Wang, Y.H. Wang, and Z.L. Gao: J. Hazard. Mater., 2010, vol. 184, pp. 555–

60.

[11] R.M. Santos, D. Ling, A. Sarvaramini, M.X. Guo, J. Elsen, F. Larachi, G.


pp. 239–50.

[12] R.I. Iacobescu, A. Malfliet, L. Machiels, P.T. Jones, B. Blanpain, and Y. Pontikes:

Waste and Biomass Valorization, 2014, vol. 5, pp. 343–53.

[13] S.A. Mikhail and A.M. Turcotte: Thermochim. Acta, 1995, vol. 263, pp. 87–94.

[14] C.W. Liu, M.X. Guo, L. Pandelaers, B. Blanpain, and S.G. Huang: in Adv. Molten

Slags, Fluxes, Salts Proc. 10th Int. Conf. Molten Slags, Fluxes Salts, 2016,

pp. 1185–90.

[15] J.B. Ferreira Neto, J.O.G. Faria, C. Fredericci, F.F Chotoli, A.N.L. Silva, B.B.

Ferraro, Tiago R Ribeiro, Antônio Malynowskyj, Valdecir A Quarcioni, and

Andre A Lotto: J. Sustain. Metall., 2016, vol. 2, pp. 13–27.


2005, vol. 32, pp. 54–60.

[17] M. Gautier, J. Poirier, F. Bodénan, G. Franceschini, and E. Véron: Int. J. Miner.

Process., 2013, vol. 123, pp. 94–101.

[18] Muhammad Salman, Özlem Cizer, Yiannis Pontikes, Rafael M. Santos, Ruben

Snellings, Lucie Vandewalle, Bart Blanpain, and Koen Van Balen: Chem. Eng. J.,

2014, vol. 246, pp. 39–52.

[19] Jeroen Heulens, Bart Blanpain, and Nele Moelans: Chem. Geol., 2011, vol. 290,

pp. 156–62.

[20] Y.Q. Sun, Z.T. Zhang, L.L. Liu, and X.D. Wang: Energies, 2014, vol. 7, pp. 1673–

84.

[21] C. W. Bale, E. Belisle, P. Chartrand, S. A. Decterov, G. Eriksson, K. Hack, I. H.

Jung, Y. B. Kang, J. Melancon, A. D. Pelton, C. Robelin, and S. Petersen: Calphad

Comput. Coupling Phase Diagrams Thermochem., 2009, vol. 33, pp. 295–311.


99

[22] C. W. Bale, P. Chartrand, S. A. Decterov, G. Eriksson, K. Hack, R.Ben. Mahfoud,

J. Melancon, A. D. Pelton, and S. Petersen: Calphad, 2002, vol. 26, pp. 189–228.





[25] Xilicate-tricalcium Phosphate, J H Welch, and W Gutt: J. Chem. Soc., 1956,

pp. 1077–83.

[26] X.R. Wu, P. Wang, L.S. Li, Z.J. Wu, and R.H. Chen: Ironmak. Steelmak., 2011,

vol. 38, pp. 185–88.

[27] Surendra N. Ghosh, P. Bhaskara Rao, A. K. Paul, and K. Raina: J. Mater. Sci.,

1979, vol. 14, pp. 1554–66.

[28] H.M. Ludwig and W.S. Zhang: Cem. Concr. Res., 2015, vol. 78A, pp. 24–37.

[29] D. Moseley and F. P. Glasser: Cem. Concr. Res., 1981, vol. 11, pp. 559–65.

[30] D. Moseley and F.P. Glasser: J. Mater. Sci., 1982, vol. 17, pp. 2736–40.

[31] Ippei Nishinohara, Naoki Kase, Hirokazu Maruoka, Shoji Hirai, and Hiromi Eba:

ISIJ Int., 2015, vol. 55, pp. 616–22.

[32] Peter F Lang and Barry C Smith: Dalt. Trans., 2010, vol. 39, pp. 7786–91.

[33] Verein Deutscher Eisenhuttenleute: Slag Atlas, Verlag Stahleisen GmbH,

Dusseldorf, 1995.

[34] Dirk Durinck, Peter Tom Jones, Bart Blanpain, and Patrick Wollants: J. Am.

Ceram. Soc., 2008, vol. 91, pp. 3342–48.

[35] J Murphy, T Meadwcroft, and P Barr: Can. Metall. Q., 1997, vol. 6, pp. 315–31.

[36] S Andersson and L Dzhavadov: J. Phys. Condens. Matter, 1992, vol. 4, pp. 6209–

16.

[37] Lieven Pandelaers, Alexandro D Alfonso, Peter Tom Jones, and Bart Blanpain:

ISIJ Int., 2013, vol. 53, pp. 1106–11.

[38] Edwin N. Lightfoot R. Byron Bird, Warren E. Stewart: Transport Phenomena,

illustrate, John Wiley & Sons, 2007, 2007.

101

Chapter 6

Experimental and mathematical simulation study on the granulation of

a modified BOF steel slag

Submitted to AIChE Journal, Chunwei Liu, Pavel Leonardo Lopez Gonzalez, Shuigen

Huang, Yiannis Pontikes, Bart Blanpain, Muxing Guo

Abstract

Basic oxygen furnace (BOF) steel slag is a major by-product generated from steelmaking

industry. Recycling of BOF slag contributes to the sustainability of the steel industry by

alleviating its environmental impact. Vitrification is an effective method to promote the

cementitious activity of slags with the aim to apply slags in high value-added applications.

In the present study, a SiO2 and Al2O3-modified BOF slag was water granulated at a pilot

scale. The amorphous and mineral fractions were measured quantitatively. The critical

cooling rate to vitrify the modified slag was calculated through the constructed Time-

Temperature-Transformation (TTT) diagrams using in-situ confocal scanning laser

microscope (CSLM) observation. To provide an insight into the crystallization behavior

during the granulation process, a mathematical model was developed. The model was

validated by comparing the amorphous fraction obtained from experiments with that from

simulation. Temperature profiles of the slag particles with varied sizes were calculated with

the aid of COMSOL Multiphysics software. The effect of particle size on the vitrified

fraction was discussed in detail and the temperature gradient from surface to center of the

particle was identified. The results provide fundamental understanding of the vitrification

process, which will help the industrial implementation of slag vitrification.

Key words: BOF Slag; Critical cooling rate; Simulation; Vitrification;

Contribution of Chunwei Liu: Chunwei Liu performed the lab-scale experiments. The

pilot experiments were carried out in ScanArc, Sweden. Chunwe Liu analysed the samples

(Pavel helped in the XRD analysis), interpreted the results and wrote the paper. Shuigen

Huang assisted in building the mathematical model. The contribution of other co-authors

consisted in discussion of the results and reviewing of the paper before final publication.

Chapter 6. Simulation of granulation

102

6.1 Introduction

Steel slag is a main residue produced in the steel industry. Around 100 - 150 kg slag is

discharged for producing 1 tonne steel [1]. Basic oxygen furnace (BOF) slag represents a

major portion of steel slag, and it originates from the external fluxes (burnt lime or dolomite)

and oxidation of impurities in the hot metal during the converter steelmaking process [2].

Associated with the large steel production, approximately 10 million tonnes BOF slag is

generated annually according to the EUROSLAG Association [3]. Storage of the BOF slag

not only causes a substantial financial burden to steel industry, but also triggers

environmental concerns due to the potential leaching of heavy metals, such as V, Cr and

Mn [4].

Recycling of slags as a secondary resource, alternatively, contributes to the sustainable

development of the steel industry and the environment. BOF slag is a CaO-rich silicate

system in general, and it could be re-used in cement applications [5–7], concrete aggregate

[8], road construction [9,10] and as metallurgical fluxing agent [11]. Although high value-

added applications such as cement appear to be more interesting, large-scale application of

the BOF slag has been limited to the metallurgical fluxing agent due to its volume

expansion [12]. The disintegration of bulk slags is caused by the hydration of free lime and

magnesia in the slag [13], which exhibits around 10 pct swelling [9]. Therefore, the

chemical composition and cooling path should be modified to prevent volume instability

of the slag and to improve its hydraulic/cementitious potential. SiO2 and Al2O3 have been

reported as effective modifiers to stabilize the BOF slag [14]. The addition of SiO2 results

in more silicate, which binds the free CaO into calcium silicates. The addition of Al2O3

produces more calcium aluminoferrite (formular Ca2(Al, Fe)2O5, in short C2AF) by

combining with free lime [15–19]. In addition to modifying the chemical composition,

rapid cooling to form glass phase is an effective method to enhance the potential as binder

or added-value applications [20]. Ferreira Neto et al. investigated the effects of SiO2 and

Al2O3 on the cementitious property of a steel slag under different cooling conditions. It was

concluded that SiO2 and Al2O3 additions promotes vitrification of the BOF slag by fast

cooling [21]. However, it has not been determined yet under what conditions the slag can

be vitrified. Moreover, it is not clear how the slag temperature changes during the

granulation process. This information is essential to be able to develop the granulation

process.

The present work aims to investigate the vitrification of an Al2O3 and SiO2-modified BOF

slag through pilot scale granulation and mathematical simulation. The mineralogical

composition of the modified BOF slag after granulation are characterized. The critical

cooling rate to vitrify the slag is obtained through the calculation combined with the

measured TTT diagrams. A mathematical model is developed to quantitatively determine

the temperature profiles during granulation. The model is validated by comparing the


103

measured glass fraction with that of simulation. Thereafter, the effect of particle size on

the vitrification is discussed.

6.2 Experimental procedure and mathematical simulation

6.2.1 Granulation of a modified BOF slag at a pilot scale

An industrial BOF slag was premixed with secondary Al2O3 and SiO2-riched materials at

a mass ratio that was determined by the preliminary lab scale experiments (see Chapters 4

and 5). The scale-up furnace is able to melt 1,500 kg slag at approximately 1400 °C. After

holding for 3 hours to homogenize the chemical composition, the molten slag was tapped

out and immediately granulated by water. The granulated slags with various particle sizes

were eventually collected and analyzed. The chemical composition of the slag after

granulation was measured by X-Ray Fluorescence (XRF, Panalytical PW 2400), as given

in Table 6.1.

Table 6.1. Chemical composition of the granulated slag (in mass pct).

Total Fe CaO SiO2 Al2O3 MgO MnO TiO2 V2O5 ZrO2 Cr2O3

16.17 37.38 17.93 11.89 4.78 2.18 1.09 0.50 0.27 0.19

6.2.2 Evaluation of the transition and nose temperature of the modified BOF slag

In-situ CSLM observation was used to estimate the liquidus (𝑇𝑙 , °C) and solidus (𝑇𝑠, °C)

temperature of the modified slag. The CSLM combines the advantages of confocal optics

and a He-Ne laser, thereby making it possible to observe samples at high resolution at

elevated temperatures. The temperature calibration and technical description of the set-up

were reported in our previous work [22,23]. Figure 6.1 (a) shows the corresponding

temperature profile. The slag was heated up to 1500 °C in air at a rate of 200 °C∙min-1

and kept at that temperature for 5 min to homogenize the slag composition. Then the slag

was cooled to 1200 °C at a rate of 5 °C min-1, followed by a rapid cooling at 200 °C min-

1. Consequently, the liquidus and solidus temperature of the crystallization (𝑇𝑜1 and 𝑇𝑒

1)

can be estimated. The temperature, where the first crystal formation was detected, is

considered to be the liquidus temperature. The temperature from which the crystals stop

growing is considered as the solidus temperature.

Nose temperature, which delineates the minimum time for developing the detectable

crystallinity [24], is a key value to calculate the critical cooling rate to vitrify the melt. To

measure the nose temperature, the TTT diagram was constructed by in-situ CSLM

observation of the slag crystallization. The temperature profile employed to construct the

TTT diagram is shown schematically in Figure 6.1 (b). A tiny amount of slag (~0.1 g) was


104

placed on a Pt holder and heated up at 200 °C min-1 to 1500 °C in air. The molten slag was

kept at 1500 °C for 5 min to homogenize the slag composition. Thereafter, the slag was

cooled rapidly to a desired temperature ranging from 1240 to 1290 °C. The slag was kept

at that temperature and the onset and ending temperature of the crystallization was recorded.

Figure 6.1. Schematic diagram of the temperature profile (a) to determine liquidus and

solidus temperature (b) to measure the TTT diagram

6.2.3 Characterization

To quantitatively measure the amorphous content in the granulated slag, the slag was

milled to a size less than 100 µm and wet-mixed with 10 wt% ZnO powder (purity 99.9

pct, Sigma-Aldrich BVBA, Belgium), which was added as a standard reference. Then

mineral and amorphous fractions in the granulated slag were identified by X-Ray

Diffraction (XRD, D2 Phaser, Bruker, Germany), with 2θ in the range of 5-70° using Cu

Kα radiation at 30 kV and 10 mA. The step size was 0.02° scanning within 0.6 second.

Quantitative analysis of the XRD (QXRD) result was achieved through Rietveld

refinement.

The slag specimens were mounted in a low viscosity resin (Epoxy), ground by silicon

carbide sanding papers and polished with diamond paste. The polished specimens were

coated with carbon for compositional and microstructural analyses by using electron probe

microanalysis (EPMA, JXA-8530F, JEOL Ltd, Japan). The accelerating voltage was set at

15 kV and a beam current was 15 nA used.

6.2.4 Fundamentals of Modelling of the Pilot Granulation

The calculated domain and boundary conditions are schematically illustrated in Figure

6.2. To simplify the calculation, the following assumptions were considered:

(a) A single slag droplet with spherical shape is considered and it is fixed at the center

of the domain;


105

(b) Incompressible water flow is at steady state with a flow rate of 1 m∙s-1. The outlet of

the water flow is at zero pressure.

(c) Liquid water is considered to be opaque to thermal radiation, i.e., no radiation is

considered in the calculation.

(d) No-slip adiabatic boundary conditions are enforced on all external walls of the device.

Figure 6.2. Schematic presentation of the calculated domain and boundary conditions

Based on the above assumptions, the heat transfer in the slag and in water during the

granulation process are respectively governed by Eqs. (6.1-1) and (6.1-2) [25]

𝐶𝑝𝜌𝜕𝑇

𝜕𝑡= 𝑘∇2𝑇 + ∅ (6.1-1)

𝐶𝑝, 𝜌, 𝜕𝑇 ,

𝜕𝑡= 𝐶𝑝

, 𝜌,𝑣∇𝑇 , + 𝑘 ,∇2𝑇 , (6.1-2)

where 𝐶𝑝 (J∙kg-1∙K-1) and 𝜌 (kg∙m3) are, respectively the heat capacity and density of the

slag; 𝑇 (°C) is the temperature of the slag; 𝑡 (s) is the granulation time; 𝑣 (m∙s-1) is the

water flow rate; 𝑘 (W∙m-1∙K-1) is the thermal conductivity of the slag; ∅ (W∙m-3) is the

volumetric heat generation rate of phase transition (crystallization). Accordingly, symbols

with a comma as superscript refer to the corresponding physical property of water. The

temperature field and the heat flux are continuous at the slag/water interface.

Latent heat (𝑄, J) generated during crystallization can be calculated as

𝑄 = ∫ 𝐶𝑝𝑑𝑇𝑇2

𝑇1 (6.2)


106

where T2 and T1 are, respectively the onset and ending temperatures of the crystallization;

the heat capacity of the slag during the crystallization can be obtained through Eq. (6.3)

[25].

𝐶𝑝 =1

𝜌[𝜙𝜌𝑠𝑜𝑙𝑖𝑑𝐶𝑝𝑠𝑜𝑙𝑖𝑑

+ (1 − 𝜙)𝜌𝑙𝑖𝑞𝑢𝑖𝑑𝐶𝑝𝑙𝑖𝑞𝑢𝑖𝑑] (6.3)

where 𝜙 (pct) is the solidified fraction of the slag during the crystalliation process;

𝐶𝑝𝑠𝑜𝑙𝑖𝑑 (J∙kg-1∙K-1) and 𝐶𝑝𝑙𝑖𝑞𝑢𝑖𝑑

(J∙kg-1∙K-1) are the heat capacity of solid and liquid slag

respectively; 𝜌𝑠𝑜𝑙𝑖𝑑 and 𝜌𝑙𝑖𝑞𝑢𝑖𝑑 (kg-1∙m3) are the density of the solid and liquid slag.

Density of the whole partile can be calculated as

𝜌 = 𝜙𝜌𝑠𝑜𝑙𝑖𝑑 + (1 − 𝜙)𝜌𝑙𝑖𝑞𝑢𝑖𝑑 (6.4)

Heat capacities of the liquid and solid slag are obtained from Ref [26]. The parameters

used in the model are shown in Table 6.2.

Table 6.2. Parameters used in the present modelling by COMSOL Multiphysics software.

Phase 1

(Liquid slag)

Density, (ρ, kg∙m-3) 2804

Heat capacity, (Cp, J∙kg-1∙K-1) 1400 [26]

Thermal diffusivity, (α, m2∙s-1) 1.27 × 10-7 [28]

Thermal conductivity, (k, W∙m-1·K-1) 0.4 [29,30]

Phase 2

(Solidified slag)

Density, (ρ, kg/m3) 3650 [31]

Heat capacity, (Cp, J∙kg-1∙K-1) 1200 [26]

Thermal diffusivity, (α, m2∙s-1) 4.6 × 10-7 [32]

Thermal conductivity, (λ, W∙m-1·K-1) 2 [27]

Phase transition Liquidus temperature, °C 1330

Solidus temperature, °C 1260

The Eq. (6.1) was numerically solved by means of the Finite Element (FE) software-

COMSOL Multiphysics. It combines heat transfer with the fluid flow (water jet). This

provides a proper way to evaluate the temperature profiles during water granulation with

respect to various slag particle sizes. The initial conditions are given in Eq. (6.5).

𝑇𝑤𝑎𝑡𝑒𝑟 = 20 °C (6.5-1)

𝑇𝑠𝑙𝑎𝑔 = 1400 °C (6.5-2)

Discretization of the calculated domain into finite elements was performed by an

unstructured meshing.


107


6.3.1 Pilot experiment

The overall amorphous and mineralogical contents in the modified BOF slag after

granulation are given in Figure 6.3. 53.4 wt% amorphous phase was measured. In addition

to the amorphous phase, the main minerals observed in the samples are respectively

dicalcium silicate (Ca2SiO4, in short C2S), tricalcium aluminate (Ca₃Al₂O₆, in short C3A)

and iron oxides (Fe3O4 and Fe2O3). The large fraction of amorphous phase indicates that

granulation is an effective method to partially vitrify the modified BOF slag. Figure 6.4

shows the microstructure of the granulated slag. A nearly fully amorphous granule is shown

in Figure 6.4 (a). In the meanwhile, in other granules, small sized crystalline phases can be

observed and are scattered in the amorphous granule matrix, as seen in Figure 6.4 (b). The

slag granules with spherical shape and small size suggests that the modified BOF slag was

water atomized. Due to the large specific surface area of the small slag droplets, this

granulation may maximize the cooling rate of the slag particles under a certain water flow.

Figure 6.4 (c) and (d) represent the typical morphology of the granulated slag in more detail.

The amorphous phase represents the slag matrix, in which there are crystalline phases

precipitated during the granulation, such as C3A (darker crystal), C2S (concave shape) and

iron oxides (FeOx, bright crystal), as identified by WDS analysis. The large area fraction

of amorphous phase observed in the backscattered electron (BSE) images (Figure 6.4)

agrees well with the QXRD analysis of the slag samples. It is concluded that the slag has

been broken down to small particles and vitrified at a fair level during the water granulation

process.

Figure 6.3. Mass percentage of the phases in the modified BOF slag after granulation, as

quantified with Rietveld refinement

0

10

20

30

40

50

C2AFMgOFe

3O

4+Fe

2O

3C

3AC

2S

Mas

s p

erce

nt,

wt

pct

Phases

Amorphous


108

Figure 6.4. Typical microstructure of the modified slag after granulation. (a) A typical

amorphous granule; (b) A typical crystal-containing granule; (c) Morphology of C3A and

FeO; (d) Morphology of C2S

6.3.2 Critical Cooling Rate to Vitrify the Modified BOF Slag: in-situ Observation and

Calculation

Figure 6.5 shows TTT diagram of the modified BOF slag, where a classical C shape curve

was identified. By decreasing the isothermal temperature, the incubation time for

crystallization was decreased significantly. The critical part of the TTT diagram is the nose

region, which is determined to be at 1240 °C.

In principle, all materials can form glasses during cooling from a liquid state at the moment

when the viscosity of the liquid reaches 1013 Pa∙s. At the corresponding temperature, e.g.

glass transition temperature, the mobility of the structural units is insufficient to be

arranged in a crystal [33]. In this study, 1240 °C is assumed as the glass transition

temperature.


109

Figure 6.5. TTT diagram of the modified BOF slag solidification in air (measured by the

CSLM test). Tn is the nose temperature

When the cooling rate is higher than the critical cooling rate (𝑅𝑐), amorphous phase will

be formed. Otherwise, the crystalline phase will be precipitated [34]. Therefore, 𝑅𝑐 is of

significant importance to evaluate the glass formation ability of a melt. Based on the

fundamentals of the crystallization kinetics, Uhlmann and Onoratoo developed a model to

estimate the critical cooling rate (𝑅𝑐) for glass formation [24]:


2

𝜂𝑛𝑒𝑥𝑝(−0.212𝐵) [1 − 𝑒𝑥𝑝 (

−0.3∆𝑆𝑚

𝑅)]

3/4

(6.6)

where 𝐴 = 40,000 J∙m-3∙K-1 is a constant; 𝑇𝑙 (°C) is the liquidus temperature of the slag; 𝜂𝑛

(Pa∙s) is the viscosity at the nose temperature; B refers to the kinetic barrier to form a critical

nucleus (𝐵 ≈ 12.6 × (Δ𝑆𝑚/𝑅) at a relative undercooling of (∆T/T) of 0.2, where ∆T is the

undercooling of the slag during crystallization) [35,36]; ∆𝑆𝑚 (J∙K−1∙mol−1) is the fusion

entropy of the slag. This model has been successfully applied to estimate 𝑅𝑐 of the slag in

recent years[37,38].

Among those parameters in Eq. (6.6), B and ∆𝑆𝑚 can be determined using FactSage 7.1.

𝑇𝑙 can be estimated by in-situ CSLM observation of the slag. Since 𝜂𝑛 is difficult to be

measured experimentally due to the rapid crystallization characteristic of BOF slag, it is

determined in this work by FactSage calculation using the measured nose temperature of

the slag. Table 6.3 presents values of the parameters employed for determining the critical

cooling rate ( 𝑅𝑐) of the glass formation. 𝑅𝑐 is determined to be 2700 °C∙s-1.


110

Table 6.3. Parameters used to calculate the critical cooling rate to vitrify the modified BOF

slag

𝑇𝑙 , °C 𝑇𝑛, °C 𝜂𝑛, Pa∙s Δ𝑆𝑚, J∙K−1∙mol−1 𝑅𝑐, °C∙s-1

1330 1240 0.295 55.3 2700

6.3.3. Mathematical simulation on the granulation of a modified BOF slag

The present pilot-scale slag granulation was simulated using COMSOL with respect to

the changes in slag granule temperature as a function of granulation time and position

within a slag particle. To vitrify the slag, the slag should be cooled down from 1400 °C

(tapping temperature) to 1295 °C (glass transition temperature) at a cooling rate higher

than 2700 °C∙s-1 (Rc), which takes 0.0387 s. Thus, the slag which could be cooled down to

lower than 1295 °C within 0.0387 s is able to form glass phase and avoid the slag

crystallization. For a particle with a given size, the temperature at 0.0387 s as a function

of position within a slag particle can be calculated via COMSOL. All the particles are

assumed to have a spherical shape and its density change during the phase transition is

ignored. As a consequence, the vitrification fraction (defined by Eq. (6.7)) can be

calculated.

𝜑 =𝑉1

𝑉2 (6.7)

where V1 the volume of the glass phase and V2 the total slag volume.

Figure 6.6 (a) shows the calculated effect of particle size on the vitrification at cooling rate

of 2700 °C s-1. For comparison, the experimental data were also plotted in the figure as

indicated by the brackets. Clearly, the amorphous fraction decreased appreciably upon

increasing the particle size. To validate this mathematical simulation, the granulated slag

from the pilot experiment was sieved and characterized. The amorphous contents in

different size ranges were measured by the QXRD. As seen in Figure 6.6 (a), the

amorphous fractions are 78.2, 51.2, and 20.9, respectively for the particles ranging from

0.08 to 0.8 mm, 0.8 to 2.5 mm and larger than 4 mm. The calculated fractions for the 0.8,

2.5 and 4 mm-sized particles are respectively 81.5, 31.3 and 27.9 pct. Correspondingly, for

the particles below 0.8 mm, the amorphous fractions is above 81.5 pct; for the particles

ranging from 0.8 to 2.5 mm, this fraction is from 31.3 to 81.5 pct; for the particles larger

than 4 mm, this fraction is below 27.9 pct. The simulated results agree well with the

measured ones. Therefore, the present model can be applied to simulate the pilot scale

water granulation of the modified BOF slag.


111

Figure 6.6. (a) Effect of particle size on the amorphous fraction. The values with brackets

are measured result, others are calculated result. (b) Ratio of the vitrified thickness to the

particle size, as COMSOL simulated. “AM” represents the amorphous-dominated region.

As indicated in Figure 6.6 (a), the critical particle size to fully vitrify the slag is calculated

to be 0.5 mm. A larger particle size implies a less than fully amorphous structure. For the

1 mm-sized particles, the amorphous fraction is 74.2 pct, but it drops to 30.3 and 27.9 pct

for particles with diameters of 3 and 4 mm. But the decreasing rate of amorphous fraction

slows down with further increasing particles size. The variation of amorphous fraction with

particle size can be also described by the change of the ratio of the vitrified thickness to


112

particle diameter. As shown in Figure 6.6 (b), for particles smaller than 2 mm, the ratio

decreased rapidly, followed by a slow decrease in the ratio with further increasing particle

size. The amorphous fraction of 2 mm-sized particles is calculated as 40.3 pct. “AM” in

Figure 6.6 (b) indicates the amorphous-dominated region, where the amorphous fraction is

larger than that of crystalline fraction. This particle size effect can be attributed to the heat

transfer. A larger particle implies more heat (both latent and sensible) to be evacuated, and

the heat transfer by conduction inside a larger particle is relatively smaller than that in a

smaller particle due to a lower temperature gradient and a smaller surface to volume ratio.

In order to obtain the amorphous-dominated solid slag, it is suggested to atomize liquid

slag to particle size less than 2 mm.

Figure 6.7. Temperature contours of the granulated slag with a diameter of 3 mm after (a)

0.5 s and (b) 1.0 s, respectively. The slice contains the center of the particle and is

parallel to the water flow direction.

Temperature contours of the water granulated slag can be calculated as a function of

granulation time and particle size. Figure 6.7 (a) and (b) take a 3-mm sized particle

respectively after water quenching of 0.5 s and 1.0 s as an example. The selected slice

passes through the center of the particle and is parallel to water flow direction. The

calculated results in Figure 6.7 suggest that a large temperature gradient forms inside the

slag particle during the granulation. By comparing the contours at 0.5 s and 1.0 s, it is

found that the thickness of lower-temperature range (<1060 °C, indicated by the double-

headed arrows) increases with time. After 0.5 s, the lower-temperature thickness is

approximately 0.35 mm, while it increases to 0.54 mm after 1.0 s. In addition, the

temperature contours are nearly symmetric, meaning that the water flow direction has a

minor influence on the temperature distribution.


113


114

Figure 6.8. (a) Temperature evolution with granulation time at different positions in a 3

mm-sized particle; (b) Temperature profile as a function of positions at 0.5 s after

granulation in a 3 mm-sized particle. The distance is measured from the slag surface.


115

The calculated temperature profiles at different positions in a 3 mm-sized particle are given

in Figure 6.8. Since the temperature contour is almost symmetric regardless of the water

flow direction (see Figure 6.7), only a quarter of the particle is considered. Figure 6.8 (a-

1) shows the temperature evolution within 10 s after granulation at various positions which

are indicated in Figure 6.8 (a-2). The distance is measured from the slag surface. The slag

surface experienced a rapid cooling, and the temperature decreased to less than 20 °C

within 0.01 s (Figure 6.8 (a-1)). The temperature profile, however, varied markedly with

the distance from particle surface. It takes 4.11 s to cool down the slag below 200 °C at the

position of 0.5 mm, while it takes 5.63 and 6.05 s for the slag at the positions of 1.0 and

1.5 mm (center) respectively. It is noteworthy that at the position of 0.5 mm the temperature

firstly increases, and then decreases, as shown by an embedded figure in Figure 6.8 (a-1).

The slight increase of temperature at the beginning of the granulation is attributed to the

latent heat released during crystallization, which can be calculated via Eqs. (6.2) and (6.3).

When the latent heat generation is larger than the heat loss by water cooling, the

temperature is expected to be increased. Figure 6.8 (b) shows the temperature profile from

surface to center of the particle at 0.5 s was given in Figure 6.8 (b). Temperature changes

from 20 to 1397 °C from the surface to 0.9 mm, while it keeps approximately at 1397 to

1407 °C from 0.9 to 1.5 mm (center). Thus, the solidification of the slag can be divided

into rapid cooling and slow cooling regions. A significant temperature difference between

the particle surface and center has been observed in Figure 6.9 (b). It is evident that after

0.5 s water quenching the surface temperature of the granule is comparable to the ambient

temperature (20 °C), but the core of the slag particle is still very hot (T >1295 °C). As a

result, it will be not possible to completely vitrify a 3 mm-sized particle within 0.5 second.

Therefore, it is key to atomize the slag melt into smaller particles (due to the large

surface/volume ratio, balls are preferable) during the granulation process, because a

smaller particle size gives a faster cooling rate. The slag property (e.g. viscosity, surface

tension) and water flow property (e.g. rate, injection direction) might play an important

role in smashing/crushing the slag, but the details need further clarification.

6.4 Conclusions

Pilot scale granulation, in-situ CSLM observation and mathematical simulation have been

performed to investigate the vitrification behavior of a modified BOF slag. After water

granulation, the amorphous fraction and minerals composition of the slag were analyzed

by QXRD. TTT diagram of the modified BOF slag was constructed. The critical cooling

rate to vitrify the modified BOF slag was obtained through FactSage calculation and CSLM

observation. Mathematical simulation was then developed and solved with the aid of

COMSOL Multiphysics. The model was validated by comparing the measured glass

fraction with that of simulation. The effect of particle size on formation of amorphous

fraction was revealed. This study provides a fundamental understanding on granulation

process. The main conclusions are summarized as follows.


116

(1) In the current water granulation at a pilot scale, the SiO2 and Al2O3-modified BOF slag

yields 53.4 wt% amorphous fraction. Microstructural analysis confirms that the amorphous

phase represents the slag matrix, in which there are crystalline phases precipitated during

the granulation, such as C3A, C2S and iron oxides.

(2) The incubation time reaches the shortest at 1240 °C under the isothermal solidification.

The critical cooling rate to vitrify the modified BOF slag is determined to be 2700 °C∙s-1.

(3) The present simulation is valid to predict the temperature evolution during granulation.

Temperature profiles as a function of position within a slag particle were calculated via

COMSOL. The particle size has a key influence on the cooling rate and the critical size to

completely vitrify the modified slag is 0.5 mm. A large temperature gradient from surface

to center of the slag particle was identified.


117

References


R. D. Bigham, M. Connolly, B. Lee, T. O. Shepker, and M. A. Zak: Environ. Sci.

Technol., 2000, vol. 34, pp. 1576–82.

[2] E. T. Turkdogan and R. J. Fruehan: in Fundam. Iron Steelmak., 11th ed.,

Association of Iron and Steel Engineers, Pittsburgh, 1998, p. 160.




Recycl., 2011, vol. 56, pp. 48–55.



[7] A. Srinivasa Reddy, R. K. Pradhan, and S. Chandra: Int. J. Miner. Process., 2006,

vol. 79, pp. 98–105.


pp. 261–68.


[10] J. Murphy, T. Meadwcroft, and P. Barr: Can. Metall. Q., 1997, vol. 6, pp. 315–31.


14.

[12] J. N. Murphy, T. R. Meadowcroft, and P. V. Barr: Can. Metall. Q., 1997, vol. 36,

pp. 315–31.



pp. 239–50.

[14] J. A. Duffy, M. D. Ingram, and L. D. Sommerville: J. Chem. Soc. Faraday Trans.

1 Phys. Chem. Condens. Phases, 1978, vol. 74, pp. 1410–19.


2005, vol. 32, pp. 54–60.

[16] C. W. Liu, M. X. Guo, L. Pandelaers, B. Blanpain, and S. G. Huang: in Adv.

Molten Slags, Fluxes, Salts Proc. 10th Int. Conf. Molten Slags, Fluxes Salts, 2016,

pp. 1185–90.

[17] X. R. Wu, P. Wang, L. S. Li, Z. J. Wu, and R. H. Chen: Ironmak. Steelmak., 2011,

vol. 38, pp. 185–88.

[18] V. Z. Serjun, B. Mirtic, and A. Mladenovic: Mater. Tehnol., 2013, vol. 47,

pp. 543–50.

[19] D. Ionescu, T. R. Meadowcroft, and P. V. Barr: Adv. Cem. Res., 2001, vol. 13,

pp. 21–30.



[21] J. B. Ferreira Neto, J.O.G. Faria, C. Fredericci, F.F Chotoli, A.N.L. Silva, B.B.

Ferraro, T. R Ribeiro, A. Malynowskyj, V. A. Quarcioni, and A. A Lotto: J.

Sustain. Metall., 2016, vol. 2, pp. 13–27.

[22] J. J. Liu, G. Chen, P. C. Yan, B. Blanpain, N. Moelans, and M. X. Guo: Am. J.

Sci., 2014, vol. 402, pp. 1–8.

[23] J. J. Liu, M. X. Guo, P. T. Jones, F. Verhaeghe, B. Blanpain, and P. Wollants: J.

Eur. Ceram. Soc., 2007, vol. 27, pp. 1961–72.


118

[24] D. R. Uhlmann and P. I. K. Onoratoo: in 10 Th Proc. Lunar Planet. Sci. Conf.,

1979, pp. 375–81.

[25] R. B.ron Bird, W. E. Stewart, E. N. Lightfoot: Transport Phenomena, illustrate,

John Wiley & Sons, 2007, 2007.

[26] K. C. Mills: South. African Pyrometallurgy 2011 Int. Conf., 2011, pp. 1–52.

[27] R. Eriksson, and S. Seetharaman: Metall. Mater. Trans. B, 2004, vol. 35, pp. 461–

69.

[28] Y.J. Kang and K. Morita: ISIJ Int., 2006, vol. 46, pp. 420–26.

[29] Y. Kang, K. Nomura, K. Tokumitsu, and H. Tobo: Metall. Mater. Trans. B, 2012,

vol. 43B, pp. 1420–1426.



[31] H. S. Carslaw and J. C. Jaeger: Conduction of Heat in Solids, Clarendon Press,

1959.

[32] I. A. Neacşu, B. Scheichl, H. Rojacz, G. Vorlaufer, M. Varga, H. Schmid, and J.

Heiss: Steel Res. Int., 2016, vol. 87, pp. 720–32.

[33] D. Durinck, F. Engström, S. Arnout, J. Heulens, P. T. Jones, B. Björkman, B.

Blanpain, and P. Wollants: Resour. Conserv. Recycl., 2008, vol. 52, pp. 1121–31.

[34] D. Turnbull: Contemp. Phys., 1969, vol. 10, pp. 473–88.

[35] H. Yinnon and D. R. Uhlmann: J. Non. Cryst. Solids, 1981, vol. 44, pp. 37–55.

[36] D. R. Uhlmann, H. Yinnon, and C. Y. Fang: in Lunar Planet. Sci. Conf. Proc.,

1982, pp. 281–88.

[37] P. Rocabois, J. N. Pontoire, J. Lehmann, and H. Gaye: J. Non. Cryst. Solids, 2001,

vol. 282, pp. 98–109.

[38] Y. Q. Sun, H. W. Shen, H. Wang, X. D. Wang, and Z. T. Zhang: Energy, 2014,

vol. 76, pp. 761–67.

119

Chapter 7

General conclusions and future work

Chapter 7. Conclusions and future work

120

Storage of BOF slags, a main by-product generated during converter steelmaking, burdens

the environment and is an economic liability to steel industry. This thesis aims to optimize

BOF slags with respect to recovery of high quality metal and mineral optimization of the

solidified slags for high added-value applications. Lab and pilot scale experiments have

been designed to understand and optimize hot-stage slag engineering. The lab scale

carbothermic reduction of BOF slags with SiO2 and Al2O3 additions has been carried out

to understand the factors influencing the reduction and to concurrently tailor the

microstructure and minerals of the slags after metal recovery. The crystallization

behavior/kinetics of the original and modified BOF slag has been observed using in-situ

CSLM and the influence of the cooling history, slag chemistry, and atmosphere on the end

microstructure and minerals of the solidified slag has been characterized. Based on the

experimental results, a pilot scale granulation trial was done to improve the amorphous

fraction of the solidified slag for preparing inorganic polymer binders. A mathematic model

was developed for a deeper understanding of the granulation process. According to the

experimental data and simulations, a guideline was developed for the hot-stage BOF slag

treatment practice. Our results suggest that it might be possible to achieve “zero-waste” of

BOF slag, through control of the operation parameters (such as temperature, atmosphere,

reductant addition) for both metal recovery and recycling of the remaining oxides.

7.1 General conclusions

Carbothermic reduction of BOF slag and phase modification

It is found that increasing C addition facilitates the transfer of P from slag to metal. By

controlling C addition, it is possible to avoid contamination of metallic Fe by P during

carbothermic reduction. To minimize the P content in the metallic Fe, the molar ratio of C

to iron oxides should be 3:1.

The formation and growth mechanisms of metallic Fe particle were discussed by post-

mortem analysis of the solidified microstructure. The smaller metallic Fe particles (<10

µm) are formed from the dispersed RO in the lime, while the larger ones (>10 µm) are

formed from the bigger sized RO and C2AF. Solid diffusion dominates the growth of

smaller particles, while the liquid fraction of the slag at elevated temperature controls the

growth of the larger particles. Therefore, Fe particle purity, size and size distribution can

be manipulated through control of carbon addition, modifier addition and processing

temperature.

The C3S phase has excellent hydraulic properties. It can be obtained when the mass ratio

of CaO/(Al2O3+SiO2) is in the range of 1.98 to 2.59. Al2O3 addition enlarges the liquid

fraction at elevated temperature and SiO2 addition stabilizes free lime by the formation of


121

C2S. Therefore, the microstructure and mineralogy of the slag product can be tailored

through the combined control of Al2O3 and SiO2.

Effect of Al2O3 addition on the mineralogy and crystallization kinetics of a high

basicity BOF slag

The influence of Al2O3 addition on the mineralogy and crystallization kinetics of a high

basicity BOF slag was studied through water quenching experiments and in-situ CSLM

observation. The Al2O3 addition can effectively remove free lime from the BOF slag, and

increases the liquid fraction of the slag. The reason is that the melting point of the slag is

decreased with increasing Al2O3 additions. The cementitious property of the quenched slag

can be manipulated by controlling the A/F ratio through Al2O3 addition.

Increasing cooling rate and Al2O3 addition increases the undercooling of the slag during

solidification. The glass formation ability of the slag is enhanced by adding Al2O3. This is

attributed to a decrease in liquidus temperature and an increase in viscosity of the slag at

the nose temperature.

C2S is the primary crystal precipitated in the continuous cooling condition with a cooling

rate of 0.98 °C∙s-1, regardless of Al2O3 addition. The secondary crystal can be C2AF or

C2FAS, which is dependent on the slag composition. Predictions by the “Scheil-Gulliver”

model are in better agreement with the experimental results in the continuous cooling

condition than those of “Equilibrium” modelling.

Mineral and microstructural optimization of the solidified BOF slag through SiO2

addition or atmosphere control during the hot-stage slag treatment

Free lime can be removed by the formation of C2S or CMS by lowering the basicity of the

slag through SiO2 addition. At a basicity of 2.8, free lime can be controlled at a satisfactory

level to reutilize the slag in constructional applications. At a too low basicity (R < 1.8 for

the current case), however, a large amount of CMS with poor cementitious property is

formed. Based on the results of this work, an optimal basicity is suggested to be in the

range from 1.8 to 2.8.

Laboratory experiments and FactSage calculation demonstrate that free lime can be

removed during hot-stage treatment under a high oxygen partial pressure. The reason is

attributed to the fact that under high oxygen partial pressure (“air” in current study), FeO

can be oxidized to Fe2O3 and further consumes free lime to form C2AF. To eliminate free

lime completely, the theoretical mass ratio of free lime to FeO should be less than 1.28,

which was achieved in the as-delivered BOF slag.

Energy efficiency for hot-stage slag stabilization by SiO2 addition and slag oxidation has

been evaluated. The calculation of energy consumption of the slag treatment indicates that

stabilization of the free lime through slag oxidation is more energy-effective. A further

evaluation of the stabilization mechanism is needed.


122

Vitrification of a modified BOF steel slag

Based on laboratory experiments, a pilot granulation of Al2O3 and SiO2 modified BOF slag

was performed. This granulation experiment aims at a better understanding of the

granulation process and the key parameters that control the vitrification of BOF slags. 53.4

wt% of amorphous fraction could be achieved by water granulation. Other main minerals

are C2S, C3A, and iron oxides. In-situ CSLM observation of the crystallization behavior of

the granulated slag indicated that an increasing cooling rate lowers the crystallization

temperature.

A mathematical model was developed and validated with experiments results. A

temperature gradient from the surface to the center of the slag particle was identified

quantitatively. The critical particle size to completely vitrify the slag was evaluated to be

approximately 0.5 mm.

7.2 Future work

Extended study on the carbothermic reduction of BOF slags

This PhD thesis contributes to the realization of a “zero-waste” concept of BOF slag with

the extraction of metallic Fe and concurrently reutilization of the remaining slag.

Concerning the metal recovery aspect, it is concluded that Fe particle purity, size and size

distribution can be manipulated by carbon additions, modifier additions and processing

temperature. However, how to separate the reduced Fe particles from the slag matrix

remains a challenge. In general, the larger Fe particle size suggests an easier separation

from slag. Since the metallic particle size is closely related with the liquid fraction in the

slag, increasing the liquid fraction at elevated temperature will be key to facilitate the

subsequent metal separation. To increase the liquid fraction, compositional modification

to reduce the liquidus temperature of the remaining slag should be studied further.

Effect of oxygen partial pressure on the liquidus and vitrification of the original and

modified BOF slags

Based on the experimental results, a high oxygen partial pressure (air) significantly

facilitates the removal of free lime from the BOF slag and modifies the minerals in the slag.

The mechanism is suggested to be the oxidation of FeO to Fe2O3, which reacts with free

lime to form C2AF. The oxygen partial pressure effect on the liquidus temperature of the

slag, however, has not been quantitatively understood yet, which will result in an

uncertainty of the atmosphere control for the optimization of the hot-stage slag treatment.

In-situ CSLM observation of the melting and solidification of BOF slags under various

oxygen partial pressure is a promising method to quantify the oxygen partial pressure effect.


123

Comprehensive evaluation of the proposed approaches to optimize solidification

In this thesis, different methods to modify slag minerals, e.g. by Al2O3 addition, SiO2

addition and control of oxygen partial pressure, are proposed. Treating the slag under a

high oxygen partial pressure (e.g. by blowing air into the slag) is proven to be more energy-

effective to remove free lime than the addition of SiO2. A comprehensive evaluation of

these different methods is needed for industrial implementation. The

hydraulic/cementitious properties and durability of the treated slag by using different

methods should be further investigated. Cost estimation of these methods should be carried

out via a comparison of the investment (energy, materials and infrastructure) and

operational costs with potential profits (metal and slag products).

125

Curriculum vitae

Chunwei LIU

Date of birth: 1 February 1990

Place of birth: Caoxian, Shandong, China

Nationality: Chinese

Email: [email protected]

Education background

2013-2017, KU Leuven, Belgium

PhD study in Materials Engineering

Topic: BOF slag hot-stage engineering towards iron recovery and use as

binders

2011-2013, Northeastern University, China

Master study in Ferrous Metallurgy

Topic: Deep reduction-separation and classification of low-grade iron ore

2007-2011, Northeastern University, China

Bachelor study in Metallurgy Engineering

Topic: Simulation on the temperature field during continue casting of a carbon

steel

127

List of publications

Peer Reviewed Journal Papers

[1] Liu, C., Lopez Gonzalez, P., Huang, S., Pontikes, Y., Blanpain, B & Guo, M.

Experimental and mathematical simulation study on the granulation of a modified

BOF steel slag. Submitted to AIChE Journal.

[2] Liu, C., Huang, S., Blanpain, B., & Guo, M. Effect of Al2O3 addition on mineralogical

modification and crystallization kinetics of a high basicity BOF steel slag. Submitted

to Waste Management.

[3] Liu, C., Huang, S., Blanpain, B., & Guo, M. Effect of basicity and oxygen partial

pressure on the mineralogy and microstructure of solidified BOF slag. Submitted to

Metallurgical and Materials Transactions B.

[4] Liu, C., Guo, M., Pandelaers, L., Blanpain, B., Huang, S., & Wollants, P. (2017).

Metal Recovery from BOF Steel Slag by Carbo-thermic. BHM Berg- und

Hüttenmännische Monatshefte, 162 (7), 258-262.

[5] Liu, C., Huang, S., Blanpain, B., & Guo, M. (2017). Valorization of BOF steel slag

by reduction and phase modification metal recovery and slag valorization.

Metallurgical and Materials Transactions B, 48 (3), 1602–1612.

[6] Liu, C., Guo, M., Pandelaers, L., Blanpain, B., & Huang, S. (2016). Stabilization of

Free Lime in BOF Slag by Melting and Solidification in Air. Metallurgical and

Materials Transactions B, 47(6), 3237-3240.

[7] Liu, C., Sun, Z., Zheng, L., Huang, S., Blanpain, B., & Guo, M. (2015). Numerical

simulation on magnetic assembled structures of iron-based metallic particles within

MMCs by a homogeneous strong magnetic field. Journal of Physics D, Applied

Physics, 48 (36), 365501.

[8] Zhang, T., Wang, D., Liu, C., Jiang, M., Lu, M., Bo, W., & Zhang, S. (2014).

Modification of inclusions in liquid iron by Mg treatment. Journal of Iron and Steel

Research, International, 21, 99-103.

Full Papers in Conference Proceedings

[1] Liu, C., Guo, M., Blanpain, B., & Huang, S. Effect of Al2O3 addition on the

crystallization of a high basicity BOF slag: perspective of glass forming ability for slag

valorization. 5th International Slag Valorisation Symposium. Leuven, Belgium, April

2017.

[2] Liu, C., Guo, M., Pandelaers, L., Blanpain, B., & Huang, S. Effect of Basicity on

Basic Oxygen Furnace (BOF) Slag Solidification Microstructure and Mineralogy.

Proceedings of the 10th International Conference on Molten Slags, Fluxes and Salts.

Seattle, USA, May 2016.

List of Publications

128

[3] Liu, C., Sun, Z., Huang, S., Blanpain, B., & Guo, M. (2015). Numerical simulation on

the self-assembled structures of colloidal particles through magnetic dipole-dipole

interactions. EPM 2015, International Conference on Electromagnetic Processing of

Materials. Cannes, France, October 2015.

BOF SLAG HOT-STAGE ENGINEERING TOWARDS IRON … · Besides the creative ideas put forward in every...

Documents

Transcript of BOF SLAG HOT-STAGE ENGINEERING TOWARDS IRON … · Besides the creative ideas put forward in every...