Melting Properties and Viscosity of Fluorine-free Mould Flux for Steel Continuous Casting and its Stability

Transcription

1 The University of New South Wales Faculty of Science School of Materials Science and Engineering Melting Properties and Viscosity of Fluorine-free Mould Flux for Steel Continuous Casting and its Stability Submitted for the degree of Doctor of Philosophy by Lin Wang March 2017

2 PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet Surname or Family name: Wang First name: Lin Other name/s: Abbreviation for degree as given in the University calendar: PhD School: Materials Science and Engineering Faculty: Science Title: Melting Properties and Viscosity of Fluorlne-free Mould Flux for Steel Continuous Casting and Its Stablllty Abstract 350 words maximum: (PLEASE TYPE) In the steel continuous casting process, mould flux plays important roles to ensure high quality of steel products. Conventional commercial mould fluxes usually contain fluorides to obtain proper physicochemical properties. The emission of hazardous components from these fluxes, however, causes severe health and environmental issues. Therefore, the replacement of fluorides with more benign components in the mould flux is a research area of significant interest for the steel continuous casting. Among the substitutes, B 2O 3 together with Na 2O could be a promising candidate. The aim of this study is to investigate high-temperature physicochemical properties, including melting properties, viscosity, and stability of boron-containing fluorine-free fluxes from simple 4-component to complex 10-component flux systems. Melting properties of fluxes were determined using the hot stage microscopy method. Viscosity was measured using rotating cylindrical viscometer and structure of fluxes was studied using Raman spectroscopy. The effects of CaO/SiO 2 mass ratio and the concentrations of different components on flux melting properties and viscosity were found to be quite complex. The relationship between flux composition and property was discussed by correlating with the flux structure determined by Raman spectroscopy and phase precipitation predicted from FactSage. A model using the back propagation neural network was developed to describe the viscosity of fluorine-free mould fluxes in this work. Calculated viscosity using the developed model was in a good agreement with the experimental data. Based on this model, the effects of temperature and composition on viscosity of fluorine-free fluxes were calculated and discussed. The evaporation of some fluorine-free fluxes was investigated using thermogravimetric analysis. Kinetic analysis of the evaporation processes demonstrated that external mass transfer contributed to the rate of evaporation. Significant evaporation was observed when the fluxes contained both B 2O 3 and Na 2O, which was attributed to the formation of highly volatile NaBO 2. Variation of the CaO/SiO 2 ratio, however, did not change the flux evaporation rate. The fluorine-free mould fluxes were compared with the commercial fluorine-containing fluxes based on these physicochemical properties. The results of this investigation provide a further understanding of fluorine-free boracic mould fluxes which lays a sound foundation for fluorine-free mould flux development. Declaration relating to disposition of project thesis/dissertation I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).... j<j /r7?. / ' l... Signature Witness Signature Date The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research. FOR OFFICE USE ONLY Date of completion of requirements for Award:

3 COPYRIGHT STATEMENT 'I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act I retain all proprietary rights, such as patent rights. I also retain the right to use in future works ( such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International (this is applicable to doctoral theses only). I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.' Signed Date 3-v I O _5 _ I 1-7 AUTHENTICITY STATEMENT 'I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.' Signed _..;---- Date -4-0_I_ I7 l? 3 / 11

4 ORIGINALITY STATEMENT 'I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.' Signed Date...}. /. }..!..!--f... lll

5 Abstract Continuous casting is a major operation in the steel production. In this process, mould flux plays important roles to ensure high quality of steel products. To obtain proper physicochemical properties, conventional commercial mould fluxes usually contain 2-15 wt% fluorides. The emission of HF, SiF4, NaF etc. from fluorine-containing fluxes, however, causes equipment corrosion, environment pollution, and health hazards. Therefore, the replacement of fluoride with more benign components in the mould flux is a research area of significant interest for the steel continuous casting. Among the substitutes, the combination of B2O3 and Na2O could be a promising candidate. The aim of this study is to investigate high-temperature physicochemical properties, including melting properties, viscosity, and stability of fluorine-free fluxes. Boroncontaining fluorine-free mould fluxes were used in this work, covering broad flux systems of simple 4-component system, CaO-SiO2-Al2O3-B2O3, to more complex 10- component system, CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2. Melting properties of fluxes (softening temperature Ts, hemispherical temperature Th, and fluidity temperature Tf) were determined using the hot stage microscopy method. Viscosity was measured using rotating cylindrical viscometer and structure of fluxes was studied using Raman spectroscopy. The effects of CaO/SiO2 mass ratio and the concentrations of different components (e.g., B2O3, Na2O, TiO2, MgO) on flux melting properties and viscosity were found to be quite complex. The relationship between flux composition and property was discussed by correlating with the flux structure determined by Raman spectroscopy and phase precipitation predicted from FactSage. As direct measurement of viscosity of multi-component systems in a broad range of temperatures and compositions is an onerous work and has some limitations, a model iv

6 using the back propagation neural network was developed to describe the viscosity of fluorine-free mould fluxes in this work. Calculated viscosities using the developed model were in a good agreement with the experimental data. Based on this model, the effects of temperature and composition on the viscosity of fluorine-free fluxes were calculated and discussed. The evaporation of some fluorine-free mould fluxes was investigated using thermogravimetric analysis. Kinetic analysis of the evaporation processes demonstrated that external mass transfer from the flux/gas interface to the main gas stream contributed to the rate of evaporation. Significant evaporation was observed when the fluxes contained both B2O3 and Na2O, which was attributed to the formation of highly volatile NaBO2. Variation of the CaO/SiO2 ratio, however, did not change the flux evaporation rate. The fluorine-free mould fluxes were compared with the commercial fluorine-containing fluxes based on these physicochemical properties. The results of this investigation provide a further understanding of fluorine-free boracic mould fluxes which lays a sound foundation for fluorine-free mould flux development. v

7 Acknowledgements Firstly I would like to express my heartfelt thanks to my supervisor Associate Prof. Jianqiang Zhang, who has offered me UNSW Tuition Fee Scholarship (TFS) and given me the precious opportunity to study in his group. His invaluable guidance and enlightening advice have always led me in a correct direction. His expertise in the field of mould flux along with his innovative ideas and thoroughness in solving problems have played an important role in the completion of this work and my development as a researcher. I would like to express my great gratitude to my co-supervisor Emeritus Prof. Oleg Ostrovski for his suggestions and supports during the project. Working under the guidance of great metallurgists like him gave me the opportunity to improve my understanding of the field of metallurgy. I am grateful to Prof. Yasushi Sasaki, who is the visiting professor in our group. Not to mention his advice and unsurpassed knowledge of mould fluxes, I have been stimulated from his industriousness in research. I would also thank technical staff for their help and support for experimental work. Here, special thanks to Mr. John Sharp, Dr. Rahmat Kartono, Mr. Bill Joe and Dr. George Yang for equipment maintenance and experiment suggestions, Dr. Anne Rich for Raman spectroscopy, and Mr. Dany Kim and Mrs. Jane Gao for I.T. issues. I am deeply grateful to Dr. Chen Zhang and Mr Dexiang Cai for their helpful assistance for my total 5-month experimental experience in Baosteel Group Corporation Research Institute. vi

8 I sincerely appreciate the companionship of my lab mates in our group, especially Mr. Jian Yang, Dr. Yaru Cui, and Dr. Xiaoli Yuan. Their pleasant accompany provided a wonderful working atmosphere during my PhD studies. Moreover, I would like to thank all my family members, especially my parents Mr. Zhiyu Wang and Mrs. Guilan Yan, and my fiancéxingzhou Li for their love, patience and spiritual encouragement in pursuing my PhD degree. Last but not least, this project was made possible by funding from Baosteel through the Baosteel-Australia Joint Research Centre, Abel Metal Services, and Australian Research Council (ARC Linkage Project LP ). vii

9 Table of Contents Thesis/Dissertation Sheet Copyright statement Originality statement Abstract Acknowledgements Table of Contents List of Tables List of Figures List of Publications Journal publications Proceedings publications i ii iii iv vi viii xiv xviii xxvi xxvi xxvii Chapter 1 - Introduction 1 Chapter 2 - Literature review Mould flux for continuous casting Chemical composition and structure of mould flux Functions and types of mould flux Defects associated with mould flux Requirements of mould flux for different steel grades Fluorine-free mould flux 14 viii

10 2.2 Melting properties of mould flux Definition of melting properties Measurement of melting properties Influence of components on melting properties Viscosity of mould flux Definition of viscosity Influence of components on viscosity Break temperature Measurement of viscosity Viscosity models Evaporation of fluorine-free mould flux Evaporation of mould flux Measurement of evaporation Evaporation kinetics Summary and project outline 55 Chapter 3 - Experimental procedure Sample preparation Melting properties measurement Viscosity measurement Evaporation measurement 61 ix

11 3.5 Raman spectroscopy 63 Chapter 4 - Fluorine-free mould flux design and flux equilibrium phase calculation Compositions of designed fluxes SiO2-CaO-Al2O3-B2O3 fluxes SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) fluxes Complex 8- and 10-component fluxes Summary and conclusion 78 Chapter 5 - Melting properties and viscosity of SiO2-CaO-Al2O3-B2O3 fluxes Melting properties Viscosity, break temperature and activation energy Raman spectroscopy Relationship between characteristic temperatures and phase composition Viscosity and structure of molten flux Modelling of viscosity Summary and conclusion 96 Chapter 6 - Melting properties and viscosity of SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) fluxes Melting properties SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes 99 x

12 6.1.3 SiO2-CaO-Al2O3-B2O3-MgO fluxes Viscosity, break temperature and activation energy SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes Raman spectroscopy SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes Relationship between characteristic temperatures and phase composition SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes Viscosity and structure of molten flux SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes Summary and conclusion 130 Chapter 7 - Melting properties and viscosity of multicomponent fluxes Melting properties 133 xi

13 7.2 Viscosity, break temperature and activation energy Comparison between fluorine-free fluxes with commercial fluxes Summary and conclusion 147 Chapter 8 - BP neural network model for viscosity results Motivation for developing BP neural network model Modelling theory Principle of BP neural network Viscosity modelling structure Validation of the BP neural network model Comparison of the BP neural network model with other models Analysis of factors affecting viscosity based on modelling results Influence of temperature Influence of CaO/SiO2 ratio Influence of components (B2O3, Na2O, TiO2, MgO) Summary and conclusion 169 Chapter 9 - Evaporation of fluorine-free mould fluxes Weight loss of fluxes Evaporation rate and activation energy Rate controlling steps of evaporation Evaporation Process 179 xii

14 9.3.2 Rate Controlling Steps of Evaporation Summary and conclusion 189 Chapter 10 - Conclusions and future work Conclusions Recommendation for further work 194 References 195 Appendix I Compositions, properties and equilibrium phases of two industrial fluorine-containing mould fluxes 208 xiii

15 List of Tables Table 2-1 Composition range (wt%) of typical commercial mould fluxes. [11] 4 Table 2-2 Values of Λi used in the calculation of Λ. [31, 33] 8 Table 2-3 Carbon equivalent for different steel grades. 12 Table 2-4 Physicochemical properties of mould fluxes by steel grades. 14 Table 2-5 Values of α i and η 0i for three groups of oxide. [134, 141] 43 Table 2-6 Parameter values of Zhang s model. [92] 46 Table 3-1 Purity of chemicals used in present study. 57 Table 4-1 Designed composition of mould fluxes (wt%). 65 Table 4-2 Measured compositions of mould fluxes (wt%). 67 Table 5-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3 fluxes. 83 Table 5-2 Parameters derived from calculated equilibrium phase fractions of CaO-SiO2- Al2O3-B2O3 fluxes. 87 Table 5-3 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3 fluxes. 90 Table 5-4 Parameters σ and for different models in calculation of viscosity of the SiO2-CaO-Al2O3-B2O3 fluxes. 94 Table 6-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O fluxes. 104 xiv

16 Table 6-2 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-TiO2 fluxes. 105 Table 6-3 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-MgO fluxes. 107 Table 6-4 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-Na2O fluxes. 114 Table 6-5 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-TiO2 fluxes. 117 Table 6-6 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-MgO fluxes. 120 Table 6-7 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. 123 Table 6-8 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. 127 Table 6-9 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 129 Table 7-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-TiO2 fluxes. 138 Table 7-2 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-MgO fluxes. 140 Table 7-3 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-TiO2-MgO fluxes. 142 xv

17 Table 7-4 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O- MnO-ZrO2 fluxes. 143 Table 7-5 Parameters derived from calculated equilibrium phase fractions of CaO-SiO2- Al2O3-Na2O-B2O3-TiO2-MgO-Li2O-MnO-ZrO2 fluxes. 146 Table 8-1 Composition (wt%), CaO/SiO2 ratio, optical basicity, temperature and viscosity of fluorine-free mould fluxes. 156 Table 8-2 Parameters σ and for different models for calculation of viscosity of the fluorine-free mould fluxes. 161 Table 9-1 Chemical compositions of selected fluxes for evaporation experiment (wt%), liquidus temperature ( C) and viscosity (Pa s) measured at 1400 C. 172 Table 9-2 Evaporation rates k (s -1 ) of fluxes. 177 Table 9-3 Calculated EA for fluxes with different Na2O contents. 178 Table 9-4 Calculated vapour pressures PA (Pa) for major gaseous species. 181 Table 9-5 Maximum evaporation rates kmax (s -1 ) for fluxes. 182 Table 9-6 Parameters ε A k and σa for different gaseous species. 185 Table 9-7 Values of κt/εa-ar, Ω A Ar and estimated diffusion coefficients of gaseous species in Ar gas (10-4 m 2 /s). 185 Table A-1 Chemical compositions of commercial fluorine-containing fluxes (wt%). 208 Table A-2 Melting properties and viscosity of commercial fluorine-containing mould fluxes. 208 xvi

18 Table A-3 Parameters derived from calculated equilibrium phase fractions of commercial fluorine-containing fluxes. 210 xvii

19 List of Figures Figure 2-1 Schematic illustration of mould flux in the mould. 3 Figure 2-2 Depolymerisation of flux due to addition of network breaking cations. [17] 5 Figure 2-3 Images of typical stages during the melting process. [28] 19 Figure 2-4 The relationship between viscosity and temperature. [109] 24 Figure 2-5 The break temperature as a function of viscosity at 1300 C for mould flux. [36] 25 Figure 2-6 Schematic diagram of the capillary viscometer. [12] 28 Figure 2-7 Schematic diagram of the falling sphere method. [12] 29 Figure 2-8 Schematic diagram of rotating cylinder method. [99] 30 Figure 2-9 Schematic diagram of oscillating cylinder method. [9] 32 Figure 2-10 Schematic diagram of inclined plane method. [123] 33 Figure 2-11 Slag ribbon length (L) as a function of fluidity (1/η). [122] 34 Figure 2-12 Schematic diagram of parallel plates viscometer. [125] 35 Figure 2-13 Deformation of the sample. [124] 36 Figure 2-14 Measurable viscosity range of different measurement techniques. 36 Figure 2-15 TGA furnace for the measurement of gas vaporisation. [59] 50 Figure 2-16 Schematic diagram of single hot thermocouple technique. [163] 52 Figure 2-17 Schematic of the evaporation process. [151] 53 Figure 2-18 A characteristic of off-gassing from an industrial mould flux. [60] 54 xviii

20 Figure 3-1 Schematic diagram of high-temperature microscopy. 58 Figure 3-2 Images of flux morphology change during the melting process: (a) original height, and heights of (b) softening temperature, (c) hemispherical temperature, and (d) fluidity temperature. 59 Figure 3-3 Schematic diagram for viscosity measurement. 61 Figure 3-4 Schematic diagram of evaporation measurement. 62 Figure 3-5 Temperature profile for the evaporation experiment at 1400 C. 63 Figure 4-1 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-B2O3 fluxes using FactSage: (a) 0.7 CaO/SiO2, 5 wt% B2O3; (b) 0.7 CaO/SiO2, 7 wt% B2O3; (c) 0.7 CaO/SiO2, 9 wt% B2O3; (d) 0.7 CaO/SiO2, 11 wt% B2O3; (e) 1.0 CaO/SiO2, 5 wt% B2O3; (f) 1.0 CaO/SiO2, 7 wt% B2O3; (g) 1.0 CaO/SiO2, 9 wt% B2O3; (h) 1.0 CaO/SiO2, 11 wt% B2O3; (i) 1.5 CaO/SiO2, 5 wt% B2O3; (j) 1.5 CaO/SiO2, 7 wt% B2O3; (k) 1.5 CaO/SiO2, 9 wt% B2O3; and (l) 1.5 CaO/SiO2, 11 wt% B2O3. 72 Figure 4-2 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3 fluxes using FactSage. 74 Figure 4-3 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o fluxes using FactSage. 75 Figure 4-4 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(4 wt%)tio2 fluxes using FactSage. 76 Figure 4-5 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o-(4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o fluxes using FactSage. [61] 77 xix

21 Figure 5-1 Effects of (a) CaO/SiO2 ratio and (b) B2O3 content on Ts, Th, and Tf of SiO2- CaO-Al2O3-B2O3 fluxes. 81 Figure 5-2 Effects of (a) CaO/SiO2 ratio and (b) B2O3 content on viscosity of SiO2- CaO-Al2O3-B2O3 fluxes. 82 Figure 5-3 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3 fluxes with different (a) CaO/SiO2 ratio and (b) B2O3 content in the frequency range cm Figure 5-4 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-B2O3 fluxes using FactSage: (a) 0.8 CaO/SiO2, 6.5 wt% B2O3; (b) 1.0 CaO/SiO2, 6.6 wt% B2O3; (c) 1.3 CaO/SiO2, 4.7 wt% B2O3; (d) 1.3 CaO/SiO2, 6.4 wt% B2O3; (e) 1.3 CaO/SiO2, 8.4 wt% B2O3; and (f) 1.5 CaO/SiO2, 6.7 wt% B2O3. 86 Figure 5-5 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3 fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) B2O3 contents.90 Figure 5-6 Relationship between lnη, EA and NBO/Si with different (a) CaO/SiO2 ratios, and (b) B2O3 contents. 92 Figure 5-7 Calculated viscosity using different models against experimental data. 96 Figure 6-1 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. 99 Figure 6-2 Effects of (a) CaO/SiO2 ratio and (b) content of TiO2 on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. 100 Figure 6-3 Effects of (a) CaO/SiO2 ratio and (b) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 101 xx

22 Figure 6-4 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. 103 Figure 6-5 Effects of (a) CaO/SiO2 ratio and (b) content of TiO2 on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. 105 Figure 6-6 Effects of (a) CaO/SiO2 ratio and (b) content of MgO on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 106 Figure 6-7 Raman spectra of the quenched fluxes with different (a) CaO/SiO2 ratios, and (b) Na2O contents in the frequency range cm -1 of SiO2-CaO-Al2O3-B2O3- Na2O fluxes. 108 Figure 6-8 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3-TiO2 fluxes with different (a) CaO/SiO2 ratios, and (b) TiO2 contents in the frequency range cm Figure 6-9 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3-MgO fluxes with different (a) CaO/SiO2 ratios, and (b) MgO contents in the frequency range cm Figure 6-10 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-Na2O fluxes using FactSage: (a) 0.8 CaO/SiO2, 8 wt% Na2O; (b) 1.0 CaO/SiO2, 8 wt% Na2O; (c) 1.3 CaO/SiO2, 6 wt% Na2O; (d) 1.3 CaO/SiO2, 8 wt% Na2O; (e) 1.3 CaO/SiO2, 9 wt% Na2O; and (f) 1.5 CaO/SiO2, 8 wt% Na2O. 113 Figure 6-11 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes using FactSage: (a) 0.8 CaO/SiO2, 4 wt% TiO2; (b) 1.0 CaO/SiO2, 4 wt% TiO2; (c) 1.3 CaO/SiO2, 2 wt% TiO2; (d) 1.3 CaO/SiO2, 4 wt% TiO2; (e) 1.3 CaO/SiO2, 6 wt% TiO2; and (f) 1.5 CaO/SiO2, 4 wt% TiO xxi

23 Figure 6-12 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-MgO fluxes using FactSage: (a) 0.8 CaO/SiO2, 2wt% MgO; (b) 1.0 CaO/SiO2, 2 wt% MgO; (c) 1.3 CaO/SiO2, 2 wt% MgO; (d) 1.3 CaO/SiO2, 3.5 wt% MgO; and (e) 1.5 CaO/SiO2, 2 wt% MgO. 119 Figure 6-13 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on lnη and EA of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. 121 Figure 6-14 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-Na2O fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) Na2O contents. 122 Figure 6-15 Relationship between lnη, EA and NBO/Si. 124 Figure 6-16 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) TiO2 contents. 126 Figure 6-17 Relationship between lnη, EA and NBO/Si of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. 127 Figure 6-18 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-MgO fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) MgO contents. 129 Figure 6-19 Relationship between lnη, EA and NBO/Si of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 130 Figure 7-1 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of TiO2 on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes. 134 xxii

24 Figure 7-2 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes. 135 Figure 7-3 Effects of (a) CaO/SiO2 ratio, (b) content of TiO2 and (c) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes. 136 Figure 7-4 Effects of CaO/SiO2 ratio on Ts, Th and Tf of (a) SiO2-CaO-Al2O3-B2O3- Na2O-TiO2-MgO-Li2O and (b) SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO- ZrO2 fluxes. 137 Figure 7-5 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of TiO2 on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes. 138 Figure 7-6 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of MgO on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes. 139 Figure 7-7 Effects of (a) CaO/SiO2 ratio, (b) content of TiO2 and (c) content of MgO on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes. 141 Figure 7-8 Effects of CaO/SiO2 ratio on the viscosity-temperature curves of (a) SiO2- CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and (b) SiO2-CaO-Al2O3-B2O3-Na2O-TiO2- MgO-Li2O-MnO-ZrO2 fluxes. 143 Figure 7-9 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-Na2O-B2O3- TiO2-MgO-Li2O-MnO-ZrO2 fluxes using FactSage with different CaO/SiO2 ratios of (a) 0.8; (b) 1.3; and (c) Figure 8-1 Schematic structure of a typical neuron. 151 Figure 8-2 Schematic structure of BP neural network model. 152 Figure 8-3 Structure of BP neural network for viscosity modelling. 154 xxiii

25 Figure 8-4 The sigmoid function curve. 155 Figure 8-5 Comparison between the calculated viscosity using neural network model and the experimental data. 158 Figure 8-6 Comparison between experimental and calculated viscosity using different models: (a) Riboud model; (b) Urbain model; (c) Koyama model; (d) Iida model; and (e) NPL model. 160 Figure 8-7 Influence of temperature on viscosity of fluxes with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 162 Figure 8-8 Influence of CaO/SiO2 ratio on flux viscosity at (a) 1300 C and (b) 1400 C. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 164 Figure 8-9 Influence of the B2O3 content on viscosity of fluxes at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(5-9 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 165 Figure 8-10 Influence of Na2O content on viscosity of fluxes at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: (7-11) wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 166 Figure 8-11 Influence of TiO2 content on flux viscosity at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, (0-6) wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 167 xxiv

26 Figure 8-12 Influence of MgO content on flux viscosity at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, (0-5) wt% MgO and 1 wt% Li2O. 169 Figure 9-1 Repeated weight loss measurements of Flux 11 at 1400 C. 172 Figure 9-2 The weight loss of fluxes with varied Na2O contents of (a) 0 wt%, (b) 6.2 wt%, (c) 7.9 wt%, and (d) 9.1 wt%, as a function of time at different temperatures with CaO/SiO2 ratio of Figure 9-3 The influence of Na2O content on the weight loss of fluxes at different temperatures of (a) 1300 C, (b) 1325 C, (c) 1350 C, and (d) 1400 C. 175 Figure 9-4 The influence of CaO/SiO2 ratio on the weight loss of fluxes at 1300 C. 176 Figure 9-5 Arrhenius plots for fluxes with varying Na2O contents. 178 Figure 9-6 A comparison of measured and estimated evaporation rates. 187 Figure 9-7 The influence of weight loss on (a) vapour pressure and (b) liquidus temperature of Flux 11 calculated using FactSage. 188 Figure A-1 Calculated equilibrium phase fractions of commercial fluorine-containing (a) Flux C1 and (b) Flux C2 using FactSage. 209 xxv

27 List of Publications Journal publications 1. Lin Wang, Jianqiang Zhang, Yasushi Sasaki, Oleg Ostrovski, Chen Zhang, and Dexiang Cai. Stability of Fluorine-Free CaO-SiO2-Al2O3-B2O3-Na2O Mold Fluxes. Metallurgical and Materials Transactions B - Process Metallurgy and Materials Processing Science, 2017, 48: Lin Wang, Chen Zhang, Dexiang Cai, Jianqiang Zhang, Yasushi Sasaki, and Oleg Ostrovski. Effects of CaO/SiO2 Ratio and Na2O Content on Melting Properties and Viscosity of SiO2-CaO-Al2O3-B2O3-Na2O Mold Fluxes. Metallurgical and Materials Transactions B - Process Metallurgy and Materials Processing Science, 2017, 48: Lin Wang, Yaru Cui, Jian Yang, Chen Zhang, Dexiang Cai, Jianqiang Zhang, Yasushi Sasaki, and Oleg Ostrovski. Melting Properties and Viscosity of SiO2-CaO- Al2O3-B2O3 System. Steel Research International, 2015, 86(6): Yaru Cui, Lin Wang, Jian Yang, Jianqiang Zhang, Yasushi Sasaki, and Oleg Ostrovski. Phase Equilibria of Fluoride-Free Boracic Mould Flux for Steel Continuous Casting. Steel Research International, 2015, 86(6): Jian Yang, Yaru Cui, Lin Wang, Yasushi Sasaki, Jianqiang Zhang, Oleg Ostrovski, and Yoshiaki Kashiwaya. In-Situ Study of Crystallization Behavior of a Mold Flux Using Single and Double Hot Thermocouple Technique. Steel Research International, 2015, 86(6): xxvi

28 Proceedings publications 1. Lin Wang, Jianqiang Zhang, Yasushi Sasaki, Oleg Ostrovski Chen Zhang, and Dexiang Cai. Stability of Fluorine-Free Mould Fluxes SiO2-CaO-Al2O3-B2O3-Na2O for Steel Continuous Casting. Proceedings of the 10th International Conference on Molten Slags, Fluxes and Salts, TMS. 2016: Seattle, America. 2. Lin Wang, Yaru Cui, Jian Yang, Chen Zhang, Dexiang Cai, Jianqiang Zhang, Yasushi Sasaki, Oleg Ostrovski. Melting Properties and Viscosity of the SiO2-CaO- Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) System. Proceedings of the 6 th International Congress on the Science and Technology of Steelmaking 2015: Beijing, China. xxvii

29 Chapter 1 - Introduction Continuous casting is a predominant operation in the steel production nowadays. In this process, mould flux is continuously added on top of the mould, covering the free surface of liquid steel. The molten flux then infiltrates into the interface between mould wall and steel strand, forming a thin slag film which consists of liquid and solid layers. The liquid layer next to the steel strand surface acts as a lubricant and the solid layer close to the mould controls heat transfer. Accordingly, mould flux plays indispensable roles in determining the quality of steel products in steelmaking process. Commercial mould fluxes usually contain 2-15 wt% fluorides to ensure proper melting properties, viscosity and heat transfer through the formation of cuspidine (Ca4Si2O7F2). The emission of HF, SiF4, NaF etc. from fluorine-containing fluxes, however, causes equipment corrosion, environment pollution, and health hazards. Therefore, the replacement of fluoride with more benign components in the mould flux is a research area of significant interest for the steel continuous casting. Among the substitutes, B2O3 is a possible one and the combination of B2O3 and Na2O leads to the formation of boracic phase which might be an appropriate candidate to replace cuspidine. The aim of this project is to investigate high-temperature physicochemical properties of varied types of fluorine-free boron-containing mould fluxes, focussing on melting properties, viscosity and flux evaporation. The composition design of the fluxes is based on the matrix system of CaO-SiO2-Al2O3- B2O3 with varied additions of Na2O, TiO2, MgO, Li2O, MnO and ZrO2, from simple 4- component, to 5-, 6-, 8-, and further to 10-component systems. Such a design makes it possible to examine systematically the effects of different components on melting 1

30 properties and viscosities of these flux systems. Melting properties of fluxes (softening temperature Ts, hemispherical temperature Th, and fluidity temperature Tf) were determined using the hot stage microscopy method. Viscosity was measured using rotating cylindrical viscometer and structure of fluxes was studied using Raman spectroscopy. As direct measurement of viscosity of multi-component systems in a broad range of temperatures and compositions is an onerous work and has some limitations, a model using the back propagation neural network was developed to describe the viscosity of fluorine-free mould fluxes in this work. Possible evaporation of boron-containing compounds could be a limitation in the industrial use of these fluorine-free fluxes. The evaporation of high-volatile components of fluxes changes the chemical compositions of mould fluxes, and therefore their physicochemical properties. The changes of mould flux properties can lead to unstable heat transfer and insufficient lubrication, resulting in severe surface defects. Consequently, knowledge of kinetics and mechanism of evaporation of mould fluxes containing both B2O3 and Na2O are important for the development of the fluorine-free mould fluxes. However, limited publications in this area showed that the evaporation is very complex, depending on the flux composition and temperature, and the mechanism of the evaporation has not been established. In this thesis work, the evaporation of selected fluorine-free fluxes was investigated using thermogravimetric analysis, and the rate determining step of evaporation was determined. The results of this investigation will provide a further understanding of fluorine-free boron-containing mould fluxes which will benefit for fluorine-free mould flux development. 2

31 Chapter 2 - Literature review 2.1 Mould flux for continuous casting In recent years, continuous casting process has become the dominant operation for producing steel due to its high productivity, efficiency and improved quality of products. [1-3] During the process, mould flux is continuously added on the top of the mould, covering the free surface of liquid steel. Mould flux on the top of the steel strand protects the molten steel from oxidation and absorbing inclusions. [4] The flux then infiltrates into the interface between strand and copper mould and forms a slag layer consisting of a liquid slag film (around 0.1 mm thick) and a solid slag layer (around 2 mm thick). [5] The liquid layer of the flux next to the steel strand surface acts as a lubricant and the solid layer close to the copper mould controls heat transfer. [6] As a result, mould flux plays important roles in determining the quality of steel products in steelmaking process. [6, 7] The schematic diagram of typical mould flux in the mould is shown in Figure 2-1. Figure 2-1 Schematic illustration of mould flux in the mould. 3

32 2.1.1 Chemical composition and structure of mould flux Mould flux is a blend of fly ash or synthetic fluxes, minerals, slag components, and carbon. [8, 9] Table 2-1 shows the conventional commercial compositions of mould flux. The main components are SiO2, CaO, Na2O, and CaF2. The CaO/SiO2 mass ratio of commercial mould flux is from 0.7 to 1.3 with fluoride and carbonaceous materials additions. [10] Table 2-1 Composition range (wt%) of typical commercial mould fluxes. [11] Component SiO2 CaO CaF2 Al2O3 Na2O Content Component MgO BaO SrO Li2O Fe2O3 Content Component K2O B2O3 MnO TiO2 C Content Physicochemical properties of the mould fluxes are principally determined by their structure, which can be explained by two different theories: molecular and ionic theories. In molecular view, it is assumed that a liquid slag is composed of individual oxides (SiO2, Al2O3, CaO etc.) and CaF2. This proposition leads to the effect of addition of one component characterised by the activity of other individual component. [12] However, electrical conductivity studies have shown that the conduction mechanism of mould flux is predominantly ionic, so molten flux should be ionic in nature. [13] In ionic theory, molten flux is composed of three ionic groups: cations, anions, and anion complexes. [14] The structure of mould flux is mainly based on tetrahedral anion 4

33 complexes SiO4 4-, where each Si 4+ ion is surrounded by four O 2-. Each of the anion O 2- is connected to the other Si 4+ (O 0, named bridging oxygen), forming a three-dimensional network. Except for SiO4 4-, Bockris et al. [15] also proposed the existence of discrete units Si3O9 6-, Si4O12 8-, Si6O15 6-, Si8O20 8-, Si9O21 6- that could exist in liquid silicate melts. As mentioned above, structure of mould fluxes at liquid state includes different types of anionic tetrahedral groups. [13] In molten flux, each component plays a different role in the silicate network. The network is broken when the cations M x+ (e.g. Na + or Ca 2+ ) are introduced. The cations break silicate chains forming non-bridging oxygens O - and free oxygens O 2- that are not bound to cations Si 4+ but to the network breakers (shown in [16, 17] Figure 2-2). Figure 2-2 Depolymerisation of flux due to addition of network breaking cations. [17] Al2O3 is classified as an amphoteric in the SiO4 4- network, as it either joins to the network or occupies the holes between the SiO4 4- tetrahedron in the same way as CaO. When Al2O3 content is low, Al 3+ substitutes for Si 4+ to form the network, increasing the degree of polymerisation. [18] On this occasion, Al 3+ should be located close to other 5

34 cations to maintain the local charge balance. With the increase in Al2O3 content, tetrahedral AlO4 5- decomposes to form Al 3+ or octahedral AlO6 9-, and this new formed octahedral complex unit is expected to act as a network modifier. [19] Fe2O3, in low concentrations, acts as a network modifier, while in higher proportions, Fe 3+ could be incorporated into the chain silicate, similarly to Al 3+. At low TiO2 content, Ti 4+ serves predominantly as a network modifier, but as a network former at high TiO2 content. [20] When Ti 4+ occupies the location of Si 4+, the stability of network structure groups weakens. [21] Introduction of B2O3 into the mould flux leads to the formation of B-O-Si or B-O-B flow units instead of Si-O-Si. The coordination of boron in respect of oxygen can be classified into four-fold coordinated boron (BO4 5- ) or three-fold coordinated boron (BO3 3- ) groups. [21-24] In general, components such as SiO2 and B2O3 are considered as network formers, whereas components such as CaF2, CaO, MgO, FeO, Na2O and K2O work as network modifiers. [25] Al2O3 and Fe2O3 are in the amphoteric category. Network formers have the nature to build up network which is regarded as the basis. Whereas network modifiers tend to break network and, in this way, modify specific properties of mould flux. Knowing the relationship between physicochemical properties and composition is extremely important for mould flux design in order to meet the requirements for given grade steel. Properties of fluxes are greatly affected by the degree of polymerisation. [26] [3, 6, 12, 27] In order to express the flux structure, different parameters have been developed. The simplest parameter is the basicity, which is a convenient characteristic of mould fluxes used in industry. [21, 28, 29] Basicity (B) is expressed as the mass ratio of network [12, 27] modifier CaO to network former SiO2: 6

35 B = W CaO /W SiO2 (2-1) where Wi is the weight percent of individual component i. As other major cations in the fluxes are not taken into account, the above definition is not complete. Therefore, other basicity indices have been defined in which different basic oxides have different influences and weighs. One example is: [3] B = (W CaO + 1.4W MgO )/(W SiO W P2 O 5 ) (2-2) Duffy et al. [30] found that the shifts in frequency of the absorption band are associated with the 6s-6p transaction observed in the ultraviolet region of the spectrum, which is related to the basicity of the fluxes. The frequency shift is usually expressed in terms of optical basicity (Λi): Λ i = Electron donor power of i Electron donor power of CaO (2-3) Λi of each component is shown in Table 2-2, which represents the power of component i to donate a negative charge. The optical basicity (Λ) of the fluxes can be calculated as: Λ = X in i Λ i X i n i (2-4) where n i is the number of oxygen in the component i, and Xi is the mole fraction of i. Λ has been reported to provide a global measure of the concentrations of O, O - and O 2- present in the melt. [31] Therefore, the optical basicity could be considered to provide an alternative measure of the degree of polymerisation. However, Λ does not take into account the fact that some of the cations in Al2O3-containing fluxes are required for charge balancing duties; hence an adjustment has been made to Λ to compensate for the cations used for charge balancing, and this is referred to Λ corr. [32] 7

36 [31, 33] Table 2-2 Values of Λi used in the calculation of Λ. Component K2O Na2O BaO SrO Li2O CaO MgO Al2O3 Λ i Component TiO2 SiO2 B2O3 P2O5 FeO Fe2O3 MnO CaF2 Λ i The number of non-bridging oxygen per the number of tetrahedral coordinated ions is called NBO/T, which indicates another measure of the depolymerisation of silicate melt besides the basicity and optical basicity. [34] This parameter is used more often than basicity by some authors, and it is probably the most used measure of the degree of polymerisation. For mould fluxes, the NBO/T ratio can be calculated as: [6] NBO T = Y NB /X T (2-5) Y NB = 2X CaO + 2X MgO + 2X MnO + 2X FeO + 2X Na2 O + 2X K2 O + 6(1 f)x Fe2 O 3 2X Al2 O 3 2fX Fe2 O 3 (2-6) X T = X SiO2 + 2X Al2 O 3 + X TiO2 + 2X P2 O 5 + 2fX Fe2 O 3 (2-7) where f is the mole ratio of Fe 3+ as network former. In practice, f = 0 for most mould fluxes. [14] Functions and types of mould flux [6, 34, 35] Mould flux is expected to fulfil the following functions: (1) To protect the steel meniscus from oxidation by insulating the steel from the atmosphere; (2) To provide thermal insulation to prevent solidification of the steel surface; 8

37 (3) To absorb inclusions floating up from the steel into the molten slag pool; (4) To provide lubrication of the steel shell; (5) To provide uniform heat transfer across the infiltrated slag layer between the steel strand and the mould. Mould flux is supplied in different types: powders, granulated, extruded and expanding granules. [34] Mould flux, in the form of powders, is inexpensive. But inhaled fine powders could cause health problems, and some inhomogeneity in supplies can occur because fines tend to move to the bottom of the container. Granulated materials are produced by spray drying, and extruded powders by extrusion. These were introduced in order to provide better quality control and minimise health hazards. Expanding granules contain an expanding agent on heating, which alters the shape and reduces the flow ability of the flux particles on the top of the mould. Fritted powders were introduced in order to improve the uniformity of chemical composition Defects associated with mould flux Some defects and problems that are related to the mould flux may occur in the steel continuously casting process, e.g. longitudinal cracks, corner cracks, breakouts, depth of oscillation marks and chemical reactions between molten steel and environment. [1] In some cases, slag and sintered powder build up on the slag rim; these agglomerations are referred to as slag ropes and bears, which have the effect of reducing the effective width of the mould/strand channel and thus tend to reduce the amount of infiltration of slag into the channel, leading to poor lubrication. 9

38 Sticker breakouts Sticker breakouts occur when the solidified shell is broken in or out the mould and, as a consequence, the liquid steel cannot be contained by the solidified shell. [10] There are probably several causes of sticker breakouts. However, all of them arise from lack of lubrication and creation of a shell with a low mechanical strength, such as those formed [1, 36] with high carbon steels. Possible causes of sticking problem are: formation of pseudo-meniscus, which contains carbon-rich agglomerates and blocks the local flow of the slag into the mould/strand gap; increase of the slag viscosity due to Al2O3 enrichment; variation of the liquid steel level and oscillating system in poor conditions; and interrupted lubrication by deficient mould flux supply. [1] Therefore, the solution for this defect is to ensure sufficient lubricity and thermal insulation. [10] Slag entrapment Slag entrapment could be associated with flow conditions and the physical properties of the liquid slag at steel meniscus level. [10, 37] The main causes of this defect are associated with a high flow speed of the liquid steel at meniscus level. These conditions generate important forces that promote the entrapment of slag drops in the liquid steel. The viscosity and surface tension of the liquid slag are the primary physical properties related to the phenomenon of slag entrapment. When the slag entrapments are large, they can interfere with the normal heat flow, producing a thinner and weaker steel shell. Therefore, the risk of breakout increases when the product leaves the mould. 10

39 Longitudinal cracking Longitudinal cracks have been classified into two types: [38, 39] gross cracks (up to 400 mm long) associated with longitudinal depression, which is usually associated with casting problems (e.g. poor mould level control) for all steel grades; subsurface and shallow cracks found when casting peritectic grades of micro-alloyed and non-microalloyed steels, which are frequently difficult to be detected. Particularly in peritectic steel, longitudinal cracking is the prevalent defect, which is caused by 4% mismatch in thermal shrinkage coefficient during the δ-γ phase transformation. [36] And in this situation, the thermal stresses can only be relieved by cracking. In Fe-C phase diagram, the range of carbon contents in steel is between 0.08 and 0.12 wt%, but in practice, other elements also have an effect on the range of carbon contents. [40] The solution for avoiding this defect is to keep steel shell as thin and uniform as possible. This can be achieved primely by reducing the horizontal heat transfer. The mould flux with high heat resistance is the tool to minimise the crack tendency and these tendencies decrease at higher flux consumption because the film thickness increases. [41] Oscillation marks Oscillation marks appear as several depressions on the surface of the strand, due to the oscillation of the mould. [3] These marks can be a cause of further defects and cracks. Although the main reason for these types of defects is the mould oscillation, but some of the mould powder properties can intensify these defects and their depth. Increasing 11

40 viscosity and mould powder consumption can directly increase the depth of oscillation marks. [42] Requirements of mould flux for different steel grades As mentioned above, defects associated with different steel grades are different; the requirements of mould flux vary correspondingly. On the whole, steel could be divided into three grades in terms of equivalent carbon concentration: low-carbon steel, medium-carbon steel, and high-carbon steel. Carbon equivalent for different steel grades are shown in Table 2-3. Carbon equivalent, Cp, for low alloy steels is defined as: [43] C p = W C W Mn W Ni 0.1W Si 0.04W Cr 0.1W Mo 0.7W s (2-8) Table 2-3 Carbon equivalent for different steel grades. Steel grade Low-carbon steel Medium-carbon steel High-carbon steel Cp, % < > Requirements of mould flux for low-carbon steel Low-carbon steel exhibits good high-temperature mechanical properties, while cracks are not usually a serious problem. [35] However, high production rates require increased casting speed and relatively low casting temperature, accordingly it is sensitive to sticker breakout. [44] Consequently, the mould flux for low-carbon steel should have the properties of low crystallisation temperature, low viscosity and basicity, as well as good inclusion entrapment ability. [45] 12

41 Requirements of mould flux for medium-carbon steel Medium-carbon steel is divided into peritectic steel and other medium-carbon steel according to the carbon equivalent of the steel. If Cp is within 0.08 to 0.12%, this kind of medium-carbon steel is classified as peritectic steel. Specialised mould fluxes for peritectic steel casting are used. When the carbon equivalent of medium-carbon steel is within 0.12 to 0.25%, such medium-carbon steel is classified as normal medium-carbon steel. The properties of common medium-carbon steel should be in between for low carbon steel and peritectic steel. The cracking in medium-carbon steel is higher than low-carbon steel because of increased shrinkage associated with peritectic solidification. [35] The δ-γ phase transformation in medium-carbon steel, especially peritectic steel, will result in thermal stresses which can only be relieved by cracking. Mould fluxes with high solidification temperature have been effective in reducing such surfaces cracks, in the way of reducing heat flow. And also, low viscosity enables the flux to provide adequate lubrication. Therefore, the characteristic features of mould flux for medium-carbon steel compared with those for low-carbon steel, are relatively high basicity and crystallisation temperature, [41] good control of heat transfer, [28] as well as low viscosity. [46] Requirements of mould flux for high-carbon steel The thermal strength of solidifying steel and the casting temperature of high-carbon steel is low compared with other grade steels. High-carbon steel is also sensitive to sticker breakout. So, the mould flux for this grade steel is with relatively low viscosity and basicity, and relatively high amount of addition of free carbon to provide sufficient thermal insulation. [35] 13

42 The design of mould flux for different grades of steels requires the knowledge of physicochemical properties of the fluxes at high temperatures. Physicochemical properties of mould flux for each type of steel grade are summarised in Table 2-4, which have a quite broad range. In general, fluxes used in the continuous casting of steels of different grades have different requirements, lower viscosity and melting temperature for low carbon steel, but higher for other steels. Table 2-4 Physicochemical properties of mould fluxes by steel grades. Steel grade Casting speed (m/min) Viscosity at 1300 C (Pa s) Break temperature ( C) Melting temperature ( C) Low-carbon steel , [47] 0.05, [48] 0.09, [47, 49] , [35] 0.13, [27] 0.3 [36] , [35] 1067, [27] 1133 [36] , [35] 1083, [50] 1088 [27] Mediumcarbon steel , [47-49] 0.06, [51] 0.08, [52] , [35] 0.11, [47] , [9] 0.12, [52] 0.13, [51] 0.14, [53, 54] 0.15, [52] , [36] 0.18 [47] , [35] , [36] 1220, [54] 1258 [53] 1048, [55] 1064, [55] 1089, [52] 1105, [52] , [35] 1150, [50] 1158 [52] High-carbon steel < , [35] 0.08, [36] 0.13, [51] 0.17 [36] 1070, [36] 1080, [36] 1140 [36] [35] Fluorine-free mould flux Functions of fluorides To ensure proper physicochemical properties of mould fluxes, conventional commercial mould fluxes always contain 2-15 wt% fluorides. [11, 56] Fluoride is a strong modifier, 14

43 which plays an important role in decreasing viscosity and solidification temperature, and forming cuspidine (3CaO 2SiO2 CaF2) to control heat transfer and lubrication. [57] The latter is especially important for the steel casting where heat transfer control is important to ensure high surface quality Hazards of fluorides However, fluorine contained in mould flux causes the corrosion of plant equipment. Besides, the emission of HF (formed in humid atmosphere), SiF4, KF, or NaF etc. from the fluxes could form acid rain, pollute groundwater, and cause potential health and safety hazards. [57-59] The fluoride emissions arise from the following equations: [59, 60] Na2O + CaF2 = CaO + 2NaF (g) (2-9) SiO2 + 2CaF2 = 2CaO + SiF4 (g) (2-10) Al2O3 + 3CaF2 = 3CaO + 2AlF3 (g) (2-11) B2O3 + 3CaF2 = 3CaO + 2BF3 (g) (2-12) MgO + CaF2 = CaO + MgF2 (g) (2-13) H2O + CaF2 = CaO + 2HF (g) (2-14) Fluoride substitutes With the objective of decreasing or eliminating fluorine content, several attempts to replace the fluorine in the mould flux are reported. [2, 21, 48] Any substitutes considered to replace fluoride should play similar functions as fluorides do. In general, two ways are taken into account: utilizing components to form crystalline phases such as CaSiTiO5 [28] or Ca11Si4B2O22 [61] to be substitutes for cuspidine which is essential in commercial mould fluxes; [2, 48] using oxides like Na2O, K2O, Li2O, MgO, MnO, B2O3, BaO and rare earth element oxides to substitute the fluorides to modify the negative effects (such as 15

44 increasing melting temperature and viscosity) of the mould flux properties caused by [2, 48] the absence of fluorides. TiO2 could decrease melting temperature and the viscosity in fluorine-free fluxes, which means a good lubricity of the addition of TiO2. [21] But when the content of TiO2 is over 23 wt%, melting temperature will increase due to the high melting point of TiO2. [62] Also, the existence of TiO2 could form different mineral compounds with high melting point (e.g., CaTiO3 or CaSiTiO5) which means a possibility to control heat transfer between mould wall and strand. [28] In the research of Nakada and Nagata, [63] the incubation time of CaSiTiO5 is as short as that of cuspidine in the commercial mould flux. Besides, the crystallisation of CaSiTiO5 does not increase the viscosity of the remained flux and maintain the lubrication. Thus, CaSiTiO5 could replace cuspidine in commercial mould fluxes. However, the addition of TiO2 would lead to a higher crystallisation temperature of mould flux, which tends to decrease heat transfer in the mould. The total thermal resistance increases with increasing the thickness of the crystalline layer. Thus, in order to propose CaO-SiO2-TiO2 fluxes as a candidate for fluorine-free mould flux, it is necessary to increase the thickness of the crystalline layer, i.e. to decrease the incubation time of CaSiTiO5 at high temperature. Besides, part of TiO2 will be reduced by carbon to form a series of reduction products such as Ti2O3, TiO, TiC, TiN and Ti(C, N). Unlike TiO2, which behaves as an amphoteric oxide, generally, Ti2O3 and TiO behave more like basic oxides. Mills research [64] pointed out that fluorine-free mould fluxes containing TiO2 have the risk to increase the rate of stick breakout due to the possible formation of Ti(C, N) during the melting process of mould fluxes. Ti(C, N) 16

45 could also result in a change of opacity and therefore radiative properties of slag, which however, is not quantified. [28] The viscosity and melting properties for fluorine-free mould flux could be improved by using the alkaline metal oxides. [57] The addition of Na2O would enlarge the air gap, and heat resistance would correspondingly increase. [57] Lu et al. [65] have studied the effects of Na2O on the melting properties, viscosity and surface tension of mould flux, and suggested that partial fluorine could be replaced by Na2O, but its content should be lower than 9 wt%. B2O3 is an additive for controlling the physicochemical properties of mould flux. B2O3 results in reducing melting temperature of mould flux remarkably. [2, 21, 35, 66-68] B2O3 increases superheat degree of the melts and then decreases viscosity. [48] In fact, the mobility of the particles in the flux intensifies under higher degree of superheat, and the flux structure is loosened. [69] Also, the crystallisation of CaB2SiO7 and CaAl4B2(SiO4)8 can be the substitute of cuspidine. [61] Bezerra et al. [70] designed a kind of B2O3-cotaining fluorine-free mould flux for high-carbon steel, with a high basicity, achieving adequate values of viscosity, thermal conductivity, and crystallisation. Although, B2O3 reduces viscosity and lowers the melting temperature, [21] it would introduce the crystallisation kinetics problem, because of the fact that B2O3 restrains mould flux crystallisation. [71] As Na2O is regarded as network breaker to promote the crystallisation of mould fluxes, the combination effects of B2O3 and Na2O (or high basicity) can obtain low viscosity and promising crystallinity. [71] Wen et al. [72] developed fluorine-free fluxes using B2O3, Na2O, TiO2, and Li2O to obtain a suitable viscosity for peritectic steels recently. The addition of MgO is useful to stabilise viscosity, [35] while ZrO2 can be added as nucleating agents to accelerate the formation of 17

46 perovskite crystals. [73] Those results obtained in above research can provide fundamental guidance for designing new type of low fluorine or fluorine-free mould fluxes for continuous casting process. 2.2 Melting properties of mould flux Definition of melting properties Melting properties of mould fluxes include softening temperature (Ts), hemispherical temperature (Th) and fluidity temperature (Tf). These three temperatures characterise the melting trajectory of flux in industrial applications. The hemispherical temperature is defined as the melting temperature of the mould flux. [28] Th should be lower than surface temperature of strand in mould bottom ( C) to maintain enough lubrication of the strand. [74] Tf is considered as the temperature at which the viscosity of the flux reaches the value apt to flow into the mould-steel gap. [34] Measurement of melting properties Hot stage microscopy (HSM) is most widely used to determine the melting properties of mould fluxes. [10, 27, 51] Experimental equipment is comprised of a high temperature furnace with accurate high temperature control, video image recording and processing system. The test consists of heating a compacted sample pressed into a cylinder at a controlled rate (e.g. 15 C/min), and monitoring the changes in sample shape and dimension. As shown in Figure 2-3, shapes corresponding to softening, hemisphere and fluidity are specified as 75%, 50% and 25% of its original height, respectively. 18

47 Figure 2-3 Images of typical stages during the melting process. [28] Another technique to determine the melting properties of the powders is the ash fusibility method. [10] In this method, a cone shaped sample is prepared by the test material with a binder which is then placed in a sample holder and inserted into the analyser. Subsequently, the cone is heated with a constant heating rate. Simultaneously, one sensor monitors the variation of the profile of the cones with temperature. At the end of the experiment, the results are presented in form of three characteristic temperatures which are defined according to the morphology adopted by the cone: softening temperature, hemispherical temperature, and fluidity temperature. [10] The definition of these three temperatures is different from that of HSM method. The softening temperature is considered to be that at which the specimen bends, sags or puffs out of shape. [75] The hemispherical temperature is noted when the height of the flux becomes equal to half of its width. [76] The fusion temperature is referred to the temperature when the top of the cone touches the floor or the height of the flux reaches [75, 76] a sixteenth of the width Influence of components on melting properties The influence of oxide components on melting temperature is complicated referring to different flux systems. It was generally accepted that components with low melting 19

48 point contribute to low melting temperature of fluxes, but this rule does not apply for all components. For commercial mould fluxes, the addition of fluoride decreased Th. [35] Increasing CaO/SiO2 ratio raised melting temperature of mould fluxes. [28, 35, 66, 77, 78] However, it was also reported that for CaO-SiO2-Al2O3-B2O3 fluxes, increasing CaO/SiO2 from 2 to 4 decreased Th. [79] Addition of Al2O3 increased Th for most mould fluxes. [35] Recent studies showed that the addition of B2O3 leads to apparently low melting temperature. [2, 21, 35, 66-68] The addition of Na2O decreased Th for both fluorine-containing [35] and fluorine-free [28] fluxes. However, there was also a report that the addition of Na2O from 8 to 10 wt% increased melting temperature of CaO-SiO2-B2O3-Na2O-MgO-TiO2-Li2O-MnO fluxes. [28] The influence of TiO2 on melting temperature depends on the content of TiO2. In most cases, minor addition of TiO2 could increase melting temperature of mould fluxes. [28, 35] However, it was also reported that the influence of TiO2 on Th fluctuates with TiO2 from 3 to 9 wt% for CaO-SiO2-B2O3-Na2O-MgO-TiO2-Li2O-MnO fluxes. [28] For fluorine-containing fluxes, the addition of MgO decreased Th. [35] For some fluorinefree fluxes, however, the addition of MgO from 3 to 8 wt% increased Th slightly. [28] As for other components, the addition of both Li2O [28, 35] and MnO [28, 35, 80, 81] decreased Th. Clearly, the effect of oxide components on melting properties is very complicated, depending on flux systems, and no simple conclusion can be drawn from the literature review. 20

49 2.3 Viscosity of mould flux Definition of viscosity It is well known that viscosity (η) is one of the most important factors which play significant roles in the efficiency of the continuous casting process and surface quality of the final product. Viscosity, as a critical indicator of lubricity, is defined as the internal frictional force when a layer of liquid flux moves over an adjacent layer, normally measured at 1300 C in industry. [82, 83] It is mainly determined by network structure, and also the bond energy of flow units. Therefore, the viscosity of molten flux is regarded as a function of the degree of polymerisation of molten flux, [56] which is affected by the composition and temperature of the flux. [84] Viscosity of the molten flux presents a significant dependence on temperature. This dependence is expressed by Arrhenius equation: [85] η = A A exp ( E A R T ) (2-15) where A A is pre-exponential constant, R is the universal gas constant, J/(mol.K), T is absolute temperature, K, and EA is the activation energies for viscous flow, kj/mol. EA represents the energy barrier for viscous flow, the variations of which suggest changes in the structure characteristics of the molten slag and further reflect the changes of the flow units in the flux. [86] When EA decreases, it suggests that the energy barrier for viscous flow reduces, consequently indicating that there are some simpler silicate network units for viscous flow and the degree of polymerisation decreases. 21

50 However, not all viscosities are described well by Arrhenius law and improved fits are obtained using Weymann [87] (Eq.2-16), Brostow [88] (Eq.2-17) and Vogel-Fulcher- Tammann (VFT) [89-91] relationships (Eq.2-18): η = A W Texp ( E W R T ) (2-16) η = exp (A B + E B R T + B BlogT) (2-17) η = exp (A V + E V R T C V ) (2-18) where AW, AB, AV, BB and CV are constants, EB, EW and EV are the activation energies for viscous flow in each equation, respectively. Reliable viscosity is needed for proper process control where flux acts as a lubricant. [92] Excessively low viscosity is concerned with high amount of slag infiltration and consequently thicker but non-uniform film, which results in a higher potential for cracking. On the contrary, excessively high viscosity will impede slag entrapment and therefore associate with thin flux film, leading to poor lubrication which means a high [10, 51] risk of star cracking and sticker breakout Influence of components on viscosity Understanding the influence of various components on viscosity is essential for developing fluorine-free mould fluxes. The influence of components on viscosity is complex. The effect of fluoride on viscosity of commercial flux has been carried out by [2, 9, 12, 35, 93-96] many researchers and the results showed that fluoride reduces viscosity. CaO is a typical network modifying oxide, leading to the breakdown of the network structure. As a result, increase in the CaO/SiO2 ratio decreased viscosity of fluorinecontaining [12, 35, 97] and some fluorine-free [21, 28, 48, 62, ] fluxes. However, Seok et 22

51 al. [99] found an increase in the viscosity of CaO-SiO2-MgO-FeO fluxes with the increase of CaO/SiO2 ratio over 1.4. The increase in the flux viscosity can be attributed to the formation of solid phases at 1300 C. With the increase of the CaO/SiO2 ratio, melting temperature of fluxes increased. [28, 35, 98, 99] Increasing in the melting temperature promoted the nucleation of solid phase, which increases viscosity. [99] For most mould fluxes, minor addition of Al2O3 increased viscosity slightly. [2, 9, 35, 62, 93] However, further increase in Al2O3 caused a decrease in viscosity. Park et al. [37] found that increase in Al2O3 content from 15 to 20 wt% decreased viscosity for CaO-SiO2-Al2O3 and CaO-SiO2-MgO-Al2O3 fluxes. It was reported that the addition of B2O3 leads to apparently low viscosity of fluorine-containing [68] and fluorine-free [2, 21, 28, 101] fluxes. For some fluxes, the influence of B2O3 was marginal. [21, 68] It was also reported that [2, 93, 102] B2O3 has an increasing effect on viscosity for CaO-B2O3-Na2O fluxes. The study of the effect of Na2O on the viscosity of traditional mould fluxes was carried out by several researchers, and all found that an increase in the Na2O content decreased viscosity. [35, 93] A similar effect of Na2O was also found for fluorine-free mould fluxes. [28, 29, 101, 103, 104] However, viscosity of CaO-SiO2-B2O3-Na2O-MgO-TiO2-Li2O- MnO fluxes was found to decrease with the increase in Na2O content first but increase with further raising Na2O content from 8 to 10 wt%. [28] For fluorine-free fluxes, such as CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO [101] and CaO-SiO2-Al2O3-Na2O- MgO-MnO-Li2O [29] fluxes, the addition of Na2O from 8 to 12 wt% and from 5.5 to 10 [21, 62, 100, ] wt% had no significant influence on viscosity. Both the addition of TiO2 and MgO [28, 29, 101, 103, 105] reduced the viscosity in mould fluxes slightly. Minor addition of Li2O [10, 28, 29, 35, 93, 95, 103] and MnO [12, 28, 35, 80, 96, 108] also decreased viscosity. 23

52 2.3.3 Break temperature Break temperature (Tbr) is defined as the temperature where an abrupt change of viscosity takes place, namely the turning point on the viscosity-temperature curve at which the viscosity increased sharply (shown in Figure 2-4(a)) during the cooling process. Tbr is caused by nucleation and growth of the crystal phases in the heterogeneous fluxes. [10, 109] Tbr is industrially important because it is analogous between the situation occurring in the mould and that in a rotating cylinder viscometer. [69] It is applied to help control the horizontal heat transfer and lubrication between the steel shell and the mould, and consequently affect occurrences of longitudinal cracking and sticker breakout in continuous casting. [36] Therefore, Tbr must adapt to the required casting condition of different grade steels. Figure 2-5 shows a general guide to the selection of mould fluxes for different steel grades in terms of satisfactory Tbr, viscosity and casting speed: [36] for steel grades sensitive to longitudinal cracking upper curve should be used, for sticker sensitive grades lower curve and intermediate values should be used for other grades. Figure 2-4 The relationship between viscosity and temperature. [109] 24

53 Figure 2-5 The break temperature as a function of viscosity at 1300 C for mould flux. [36] Measured Tbr values have been found to decrease with increasing cooling rate. The cooling rate in the mould is as high as the order of 500 C/min, and this may lead to an appreciably lower Tbr than that measured by dynamic viscosity measurements (typically 10 C/min). In various investigations, Tbr is approximately equal to solidification temperature Tsol, [29] but can also vary by up to 80 C in other occasions, which may be related to heterogeneous nucleation in differential thermal analysis (DTA) measurement for Tsol. [2] Tbr, C, can be estimated within the error of ±20 C from chemical composition (Eq.2-19 and Eq.2-20): [36] For dynamic conditions: T br = X Al2 O 3 3.3X SiO X CaO 13.86X MgO 6.46X F 3.21X MnO 9.22X TiO X K2 O 3.2X Na2 O (2-19) For steady state conditions: 25

54 T br = X Al2 O X SiO X CaO 9.88X MgO 17.32X F X Fe2 O X MnO 308.7X K2 O X Na2 O (2-20) Most components reduce Tbr, whereas CaO causes an increase in Tbr. It should be noted that the above equation was derived by chemical analysis. For a small amount of components (e.g., K2O, FeO, MnO, TiO2), the constants in Eqs.2-19 and 2-20 may be subject to large errors. Also the pickup of Al2O3 and MnO during casting should be accounted for. [34, 36] Measured Tbr of ZrO2-containing fluxes is higher than the calculated one of the fluxes without ZrO2 content taken into account using Eq.2-19 by about 10 C for every 1 wt% ZrO2 added. [36] In was reported that an increase in CaO/SiO2 enhanced Tbr. [35, 36, 106] It was also reported that increase in CaO/SiO2 from 0.4 to 0.7 decreased Tbr first but then increased it for CaO-SiO2-Al2O3-Na2O-MgO-MnO-Li2O fluxes. [29] Increase in Al2O3 [35, 36, 110] and B2O3 [2, 35, 36, 69] content decreased Tbr. In some reports, an increase in the Na2O content decreased Tbr, [35] but recent results showed slightly increased Tbr for fluorine-free fluxes. [29] For fluorine-containing fluxes, Tbr was ascended as TiO2 content was increased. [35] Tbr was also lowered as TiO2 content was increased for CaO-SiO2-Al2O3-MgO-TiO2 fluxes. [106] Minor additions of Li2O [28, 29] and MgO [29] decreased Tbr. Furthermore, ZrO2 has been used recently to increase Tbr, since it appears to have a low solubility in the molten flux and the solid particles of ZrO2 then act as nucleation sites for the solidification of the flux. [111] Measurement of viscosity In continuous casting applications, flux viscosity is normally evaluated by using its value at 1300 C. [112] A wide range of techniques has been developed to measure the 26

55 viscosity of mould flux according to their ability to perform absolute viscosity measurements, e.g., capillary method, falling sphere method, rotating cylinder method, [113, 114] oscillating method etc Capillary method Figure 2-6 shows the schematic diagram of a capillary viscometer. In capillary method, after pulling the stopper rod away, the time required for a liquid to flow through a capillary tube is determined, so the viscosity η, Pa s, can be calculated using Poiseuille relationship: [12] η = 1 10 πr4 pt 8vL (2-21) where r, m, is capillary radius; L, m, is capillary length; p, Pa, is constant pressure gradient between the ends of the capillary; v, m 3 /s, is the volume of liquid discharged; and t, s, is time. This equation does not account for a certain amount of kinetic energy remaining in the liquid after it has left the capillary and which has not been expended in overcoming friction. [114] This error can be corrected by the introduction of a correction of factor Hagenbach (H): η = 1 10 (πr4 pt 8vL H) (2-22) H = vρ 8πLt (2-23) where ρ, kg/m 3, is the liquid density. 27

56 Figure 2-6 Schematic diagram of the capillary viscometer. [12] The capillary method has been so far only applied in the medium or low temperature range. [115] Above 1200 C, complications arise in connection with the selection of a suitable crucible and capillary material in terms of dimensional stability and corrosion resistance. However, this method permits, given suitable adjustment, absolute viscosity measurement to be made. [12] Falling sphere method In the falling sphere method, the time for the sphere to fall due to gravity or to be dragged upwards through the melt is measured (shown in Figure 2-7). The viscosity is calculated using modified Stokes Law: [116] η = 2gr2 (ρ k ρ liq ) 9S (2-24) where g, 9.81 m/s 2, is gravitational constant; r, m, is the radius of the sphere; ρk, kg/m 3, and ρliq, kg/m 3, are the densities of the sphere and liquid; and S, m/s is the velocity of descent or ascent of the sphere. 28

57 Figure 2-7 Schematic diagram of the falling sphere method. [12] The effect of thermal expansion of the liquid on the viscosity was accounted by using the equation proposed by Ladenberg: [117] η = 2gr2 (ρ k ρ liq ) 9S[1+2.1(d D)] (2-25) where d, m, and D, m, are diameters of the sphere and crucibles, respectively. Eq.2-25 is applicable when the diameter of the sphere is less than 10% of the diameter of crucible. This method has been used for the measurement of the viscosity of glass and slag melts. The time taken to pull a ball out of the melt with a constant force is measured. [116] However, this method does not permit absolute viscosity measurement, since an apparatus constant must be referred to what is derived from calibration curves using liquids of known viscosity. [113] 29

58 Rotating cylinder method Rotating cylinder method is the most popular method to measure viscosity of fluxes. [18, 21, 69, 96, 99, 118] It consists of two concentric cylinders (Figure 2-8): usually the outer cylinder is a crucible and the inner cylinder is spindle in movement. [9, 119] When a cylinder is rotated, it provides a velocity gradient and the torque developed is measured at different temperatures. This method is a time-consuming one and demands a certain experimental skill. Figure 2-8 Schematic diagram of rotating cylinder method. [99] The viscosity is calculated by the following equation: [119] M η = ( 1 8π 2 nh r2 1 i r 0 2) (2-26) where M, N m, is the torque; n, rev/s is the number revolutions per sec; h, m, is the height of the bob inside the molten flux; and ri, m, and ro, m, are the radius of the inner and outer cylinders, respectively. 30

59 For given experimental conditions, the height and radius of the spindle, as well as the radius of the crucible, are known, viscosity is calculated using the following relationship: [114] η = C 0 M n (2-27) where C0 is the equipment constant, which can be obtained from the calibration runs at both room temperature and high temperatures, of the apparatus using standard viscosity materials. Even though this method is widely applied in measuring viscosity of molten fluxes, it is time-consuming and difficult sometimes. The results could be affected observably by the inhomogeneity of the fluxes Oscillating method Oscillating method can be divided into oscillating cylinder and oscillating plate. In the oscillating cylinder method (shown in Figure 2-9), the logarithmic decrement of the swings is recorded during the pendulum oscillating process, and the viscosity is measured by the following equation: [120] η = 2 λ πρτ [2θ( π )+(λ π )2 + R 4 +R 3 d ] (2-28) where ρ, kg/m 3, is the flux density; τ, is the period of oscillation in empty system; θ is the moment of inertia of oscillating system; λ is the decrement due to damping effect of the liquid; R, m, is the radius of cylinder; and d, m, is the thickness of cylinder. 31

60 Figure 2-9 Schematic diagram of oscillating cylinder method. [9] The oscillating plate method is a relatively new method in which a linear oscillating plate is submerged in the melt. As a result, there is a retarding force proportional to the viscosity of the fluid. When reaching a steady state, it records the amplitude of oscillation in air (φ A ) and in the melt (φ). This viscosity is derived from: [121] ηρ = G [ ρ A ρ 1]N (2-29) where N is a constant, and G the constant load cell determined in calibration experiments. Oscillating viscometers are usually used for determining relative values of viscosity and are calibrated with standard viscosity liquids before use Inclined plane method The inclined plane method has been used by some laboratories to estimate viscosities of molten fluxes because of the simplicity of operation. [7, 122, 123] In this method, a mass of of powders (e.g. 15 g) is placed in a graphite crucible and then is melted at a specific 32

61 temperature (e.g C). The melted flux is maintained at that temperature for 15 mins in order to achieve homogeneity. Then, the melt is poured and quenched onto an inclined plane (shown in Figure 2-10). The length (L) of the ribbon is measured to have an estimation of the mould powder fluidity. The length of the slag ribbon (L) formed has been found to be a function of the reciprocal viscosity. [122] One of the expressions is: [123] η = C 1 exp ( C 2 L ) (2-30) where C1 and C2 are constants, which are determined by the setup of the apparatus. Figure 2-10 Schematic diagram of inclined plane method. [123] Experimental trials in the laboratory indicated good reproducibility of results and relationship between ribbon lengths (L) and (1/η) for viscosities over 0.1 Pa s (see Figure 2-11). The scatter was lower for more viscous slags. This method can be applied for slags with viscosities in the range 0.15 to 0.6 Pa s within ±15% of experimental values. [122] 33

62 Figure 2-11 Slag ribbon length (L) as a function of fluidity (1/η). [122] Parallel plates method The parallel plate method has been used for measuring high viscosities of glasses. [124] In the parallel plates method, the sample to be tested is retained between two horizontal plates and is compressed axially by driving the plates together as depicted in Figure

63 Figure 2-12 Schematic diagram of parallel plates viscometer. [125] As shown in Figure 2-13, when no-slip exists, the radial velocity of glass in contact with the plates is zero, and is a maximum at mid-height. When perfect slip exists, the radial velocity is a function of radius only. Under these two conditions of slip, the viscosity is expressed as: η(no slip) = 2πM s gh 5 s 3V(dh dt)(2πh 3 + V) (2-31) η(perfect slip) = M s gh 2 s 3V(dh dt) (2-32) where M and g are the applied load and the gravity acceleration, respectively; h and V are the sample height and the sample volume, respectively; and dh/dt is the deformation or sag rate. 35

64 Figure 2-13 Deformation of the sample. [124] Peculiar difficulties associated with the measurement of the viscosity of molten slags over a wide temperature range arise from the simultaneous presence of a very wide viscosity range, poor heat conducting properties of the liquid, invariable presence of small bubbles in the liquid, or the very high temperatures at which the experiments must be carried out. [126, 127] Viscosity range of iron and steel making fluxes are within the 0.01 to 8 Pa s (shown in Figure 2-14). For this range of viscosity, the most popular method adaptable is concentric rotating cylinder method. [128] Figure 2-14 Measurable viscosity range of different measurement techniques. 36

65 2.3.5 Viscosity models Understanding the viscosity behaviours is significant to know the viscous state of slag melts and investigating the various phenomena of molten slags in high-temperature processes. Therefore, there have been several efforts to determine, predict or model the parameters that can influence the viscosity of fluxes. In recent years, the study of flux viscosity has focused on the development of new models upon experimental data, as reliable measurement of viscosity at high temperature is both difficult and time-consuming. [ ] A number of models have been developed on the basis of a large quantity of experimental data. [127, 129, 132] These models were commonly expressed in the form of Arrhenius, [32, 133, 134] [87, 101, Weymann-Frenkel, 132, ] Brostow, [88] and Vogel-Fulcher-Tammann [139] equations. Estimation of viscosity of fluorine-containing and fluorine-free mould fluxes was carried out using original or modified Riboud model, [21, 34, 62, 101, 110, 135] Koyama model, [133, 140] Urbain model, [136] Iida model, [134, 141] NPL model, [12, 32] [92, 108, 142] and Zhang model Riboud model In 1981, Riboud et al. [135] developed the simplest model for calculating the viscosities of mould fluxes. Riboud model classifies the chemical components of slag into five groups and sums of their mole fractions Xi: (1) Network formers: X SiO2 = X SiO2 + X P2 O 5 + X TiO2 + X ZrO2 ; (2) Network breakers: X CaO = X CaO + X MgO + X FeO + X BO1.5 + X Fe2 O 3 + X MnO + X NiO + X CrO + X ZnO + X Cr2 O 3 ; (3) Al 2O 3 : X Al2 O 3 = X Al2 O 3 ; (4) CaF 2 : X CaF2 ; 37

66 (5) Na 2O : X Na2 O = X Na2 O + X K2 O + {X Li2 O}. If the mould flux contains constituents which are not given in these groups, it should be added to the appropriate category, e.g. Li2O can be added to the group Na 2O. The temperature dependence of viscosity is expressed by Weymann equation (Eq.2-16), and AW and EW are calculated by the following equation: [135] A W = exp( X CaO X CaF X Na2 O 35.76X Al2 O 3 ) (2-33) E W R = X CaO 46356X CaF X Na2 O X Al2 O 3 (2-34) Riboud and co-workers present the equation twice in their paper, but with a different order of the digits in the fourth term in Eq.2-34: 39519X Na2 O and 39159X Na2 O [135] probably because of typo. Riboud model was preliminary developed in order to calculate the viscosities of conventional mould fluxes; besides, it is proved to be applicable for a wide range of different metallurgical fluxes. [34] This model applies for the following content range: SiO2 (28-48 mol%), CaO (13-52 mol%), Al2O3 (0-17 mol%), CaF2 (0-21 mol%), and Na2O (0-27 mol%). [34] It is suitable for predicting low viscosity values below 0.5 Pa s; the deviation is around 30-35%. [32] Furthermore, this empirical model did not properly predict the viscosity of some multicomponent slags containing other fluorides such as NaF or MgF2. [124] 38

67 Koyama model Koyama et al. [133] investigated Riboud s classification in predicting the viscosities of mould fluxes for continuous casting of steels. They determined numerical factors for each flux constituent by using experimental data for mould flux of different viscosity. Based on Arrhenius equation, they developed the following equations for the viscosity of industrial mould flux for steel casting: [140] η = 0.1A K exp(b K /T) (2-35) A K = exp( 6.1X CaO 12.1X MgO + 6.3X CaF2 19X Na2 O 24.2X Al2 O ) (2-36) B K = X CaO 9259X SiO X CaF X Na2 O X Al2 O X Li2 O (2-37) Kim model Kim et al. [143] also derived similar equations as follows: η = A Kim exp(b Kim /T) (2-38) A Kim = exp( 0.046X SiO2 0.07X CaO 0.041X MgO 0.185X Al2 O X CaF X B2 O ) (2-39) B Kim = X SiO X CaO X Al2 O X Na2 O 176.1X CaF X Li2 O X B2 O 3 (2-40) Koyama and Kim models tend to predict higher viscosities than the experimental values for commercial fluorine-containing mould fluxes, the deviation being +40 and +50%, respectively. [32] 39

68 Urbain model In 1987, Urbain proposed a coherent model to estimate the viscosity of complex slags based on the CaO-SiO2-Al2O3 system. [136] According to this model, the oxides inside a melt are classified into different categories: network formers XG, network modifiers XM and amphoteric XA: (1) Network formers: X G = X SiO2 + X P2 O 5 (2) Network modifiers: X M = X CaO + X MgO + X Na2 O + X K2 O + 3X CaF2 + X FeO + X MnO + 2X TiO2 + 2X ZrO2 (3) Amphoteric: X A = X Al2 O 3 + 3X Fe2 O 3 + X B2 O 3 Urbain normalised the values of XG, XM and XA by dividing them by ( X FeO1.5 + X TiO2 + X ZrO2 + X CaF2 ) to get the values of XG*, XM* and XA*. Using the values for XG*, XM* and XA*, empirical equations are derived to calculate constants AU and BU in the Eq.2-41 which is used in Urbain model to calculate viscosity. η = A U Texp(10 3 B U T) (2-41) ln A U = B U (2-42) where η is viscosity in Poise; parameter B is influenced by both the ratio α = X M (X M +X A ) and X G : B U = B 0 + B 1 X G + B 2 (X G ) 2 + B 3 (X G ) 3 (2-43) where B 0 = α α 2 (2-44) B 1 = α α 2 (2-45) B 2 = α α 2 (2-46) 40

69 B 3 = α α 2 (2-47) Recently, Dong et al. [62] proposed a modified Urbain model, which provides a better estimation of the viscosity for a certain range of slag composition. In this modified model, the TiO2 is classified as network former, which is more reasonable Iida model Iida model was proposed to predict the viscosities of industrial mould fluxes for steel casting, on the basis of relationships between the network parameter and the reciprocal of the basicity index. [134, 141] This model is based on the Arrhenius equation which divides oxides into three groups: network formers, network modifiers and amphoterics. The expression for calculating the viscosity is expressed as: η = A η 0 exp(e B i ) (2-48) where A and E are parameters set by adjustments to experimental data: A = T T 2 (2-49) E = T (2-50) η 0 is the viscosity of the melted components not forming network: n η 0 = i=1 X i (2-51) η 0 i where η 0 can be approximately calculated using the following expression: i η 0 i = [M i (T m ) i ]1/2 exp(h i H i = 5.1(T m ) i 1/2 RT) 2/3 (V m ) exp[hi i R(T m ) i ] (2-52) (2-53) 41

70 where M is the formula weight; V m is the molar volume at the melting temperature T m; R is the gas constant. X is the mole fraction and the subscript i refers to the component. Basicity index B i can be readily calculated with: B i = (α i W i ) B (α i W i ) A (2-54) where αi is specific coefficient; A and B represent acidic oxide and basic or fluoride oxide, respectively; and Wi is weight percentage. Table 2-5 gives the value for αi and η 0 i of different compositions. For fluxes containing amphoterics, modified basicity index B i can be expressed as: [134] B i = (α i W i ) B +α Fe W 2O3 Fe 2O3 (α i W i ) A +α Al W 2O3 Al 2O3 +α W TiO2 TiO 2 (2-55) where α i is the modified specific coefficient indicating the interaction of the amphoteric oxide with other components in slags, which is dependent on basicity index and Wi. The value of α i widely varies with slag composition and temperature. The value for α Al2 O 3 for SiO2-Al2O3-CaO-MgO system can be given by reference. [134] In comparison with Riboud model (±43.3%), Koyama model (±33.2%), and Urbain model (±101.5%), Iida model provides the most accurate predictions of the viscosities of CaO-SiO2-Al2O3-MgO fluxes. [134] The probable uncertainties in the predicted values using Iida model are of the order of ±26.6%, close to uncertainties of viscosity measurements for molten industrial slags. [134] 42

71 141] Table 2-5 Values of α i and η 0 for three groups of oxide.[134, i Component αi η 0 i at 1573 K (mpa s) η 0 i at 1623 K (mpa s) η 0 i at 1673 K (mpa s) η 0 i at 1723 K (mpa s) η 0 i at 1773 K (mpa s) SiO Acid oxide TiO B2O Al2O CaO MgO K2O Basic oxide or fluoride Na2O Li2O MnO FeO Fe2O CaF Model proposed by Zhang [108, 142, Recently, a structurally based viscosity model has been proposed by Zhang et al. 144] to describe the viscosity of oxide melts as functions of both temperature and composition. The model makes use of the temperature compensation effect which relates the pre-exponential factor A to the activation energy E of Arrhenius equation: [108] lna = k(e ) (2-56) 43

72 where values and are deducted from the viscosity variation of pure SiO2 melts with temperature. Parameter k can be obtained by the regression analysis of the viscosity data for a specific binary system MxOy-SiO2. For a multicomponent system ƩMxOy-SiO2, it is assumed that the value of parameter lna can be derived from the linear addition of values for the various binary systems MxOy-SiO2 with the weights of the renormalized mole fractions of oxides MxOy. Thus, it follows that the parameter, k, can be calculated by Eq.2-56 based on the parameter ki which is regressed from MxOy- SiO2 binary system: k = i,j SiO 2 (x i k i ) i,j SiO2 x i (2-57) E can be calculated as follows: E = n OSi +α Al n OAl + α i n Oi + α Al,i n OAl,i + α i j Si n OSi i + α Al,i n j OAl,i (2-58) where n Oi are the mole numbers of different types of oxygen ions; and the parameter α, describes the deforming ability of bond around the corresponding unit. In the denominator of Eq.2-58, O Si, O Al, O i, O Al,i, O i j Si, and O Al,i represent the bridging oxygen bonded with Si 4+, oxygen bonded with Al 3+ which is not charge compensated, free oxygen close to metal cation i, bridging oxygen bonded with compensated Al 3+, nonbridging oxygen bonded with Si 4+, and non-bridging oxygen bonded with Al 3+, respectively. αi, the viscosity activation energy for non-glass-forming oxides, is proportional to Tm 1.2 (where Tm is the melting point). Therefore, the parameters for different oxides are inversely proportional to their melting point. According to the value of αfe, the value of αi can be calculated as: 44

73 α i = ( T 1.2 m,feo ) α T Fe (2-59) m,i In this model, an assumption of complete bridge breaking has been used. Values of model parameters are shown in Table 2-6. Besides, five assumptions are made in the calculation of the numbers of the different types of oxygen: [144] (1) When there are several basic oxides in an Al2O3-containing melt, there is a strict order for which cations carry out the charge-compensation of the Al 3+ with the highest priority being given to the cations with the lowest field strength. Priority for charge-compensation is in the order: K + > Na + > Li + > Ba 2+ > Sr 2+ > Ca 2+ > Mn 2+ > Fe 2+ > Mg 2+. Only when the cation with the highest priority is exhausted, will a cation with a lower priority be used to charge compensate the Al 3+. (2) The equilibrium constant for the charge-compensation reaction of MxOy with Al2O3 is infinite. (3) When the mole fractions of basic oxides are larger than those for Al2O3, the numbers of bridging oxygen for tetrahedral AlO4 5- and SiO4 4- are equivalent, and the numbers of non-bridging oxygen bonded to Al 3+ and Si 4+ are proportional to the numbers of tetrahedral AlO4 5- and SiO4 4- present. (4) The equilibrium constant for the reaction of a free oxygen (from the remaining basic oxide) with a bridging oxygen (associated with either a compensated Al 3+ or a Si 4+ ) to generate non-bridging oxygen is infinite. [145] (5) For systems containing several basic oxides, the numbers of different types of oxygen ions can be calculated by the random mixing rule. 45

74 Table 2-6 Parameter values of Zhang s model. [92] i ki 10 5 i i α Si αi αal,i α Al,i Ti Al Ca Mg K Na Li Mn Fe Fe(III) CaF NPL model Mills and Sridhar [32] have developed a model (named NPL model) based on the optical basicity, which is generally applicable without limitation on slag composition. The viscosity is expressed following Arrhenius equation: η = A NPL exp(b NPL T) (2-60) where A NPL and B NPL are functions of Λ corr : ln ( B NPL 2.88 ) = Λcorr (2-61) lna NPL = (Λ corr ) Λ corr (2-62) Λ corr = x in i Λ i x i n i (2-63) 46

75 where x i is the mole fraction of component i; n i is the number of oxygen atoms in the molecule. In prediction of viscosity of mould flux, NPL model provides reasonable values, the deviation being slightly less than that for Riboud and Koyama model, below 0.4 Pa s. [32] KTH model In 1994, a viscosity model has been developed at the Division of Metallurgy, Royal Institute of Technology (KTH) from the theory of absolute reaction rate. [137] The viscosity of molten flux η can be expressed as follows: η = hn A V exp (ΔG ) = hn Aρ RT M exp (ΔG RT ) (2-64) where h is Planck s constant; NA is Avogadro s number; V is molecular volume; R is the gas constant; T is temperature; ρ is density; M is molecular weight; and G* is the Gibbs free energy of activation for viscosity, which is described by the following equation in the case of multicomponent slag: ΔG = m i=1 x i G i + G mix (2-65) where G i is the Gibbs activation energy of pure component i, x i is mole fraction of component i and the term G mix is an extra term that takes into account the contribution of the mutual interaction between different species due to the mixing of the components. [137] Model based on Moynihan s method An alternative method to compute the viscosity of mould flux was used by Benavidez et al.. [7] This method is based on Moynihan s method [146] that uses the initial and final temperatures, corresponding to the glass transition, determined by DTA or DSC results. 47

76 Assuming that at very high temperatures the viscosity approximates at the same value for all liquids, the following viscosity expression for a heating rate of 10 C/min was deduced: [7] logη = [0.147(T T g)/(t g) 2 Δ(1/T g )]+1 (2-66) where η is the viscosity in Pa s; T is the temperature in K; Tg, K, is the initial temperature of glass transition temperature; and T g is the final temperature of the glass transition; and Δ(1/T g ) = 1/T g 1/T g. The Moynihan s method is useful to estimate the viscosity of mould flux in the range C by a single DTA run, without any knowledge of the mould flux chemical composition. [7] A review of models based on the chemical composition showed that the minor differences between the estimated values by both the Iida and Riboud models and those determined experimentally ranged between ±25% and ±30%. [147] Both Riboud and Urbain models can be used for mould flux, but Riboud one is preferable and provides more reasonable estimate. [12, 147] These models can provide useful information to solve industrial problems regarding process control. [56] For the B2O3-containing fluorine-free mould fluxes, however, none of the existing models was applicable. [62, 92, 129, 148] Several trials were conducted to modify abovementioned models, such as modified Riboud model for CaO-SiO2-Al2O3-B2O3-Na2O- TiO2-MgO [21] and CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO [101] fluxes. Nevertheless, these modified models are still not accurate for the B2O3-containing mould fluxes examined in this study. Therefore, it is important to develop a new and more powerful model for the viscosity prediction, correctly reflecting the complex 48

77 effects of temperature, chemical composition, and structure (the degree of [92, 149] polymerisation) of the fluxes. 2.4 Evaporation of fluorine-free mould flux Evaporation of mould flux Evaporation of mould flux is directly related to its vapour pressure. For commercial mould flux, fluorides are lost in the vapour phase to some extent, due to the high vapour pressure of fluoride compounds in the flux during the casting operations. [59] Replacement of fluoride with environmentally friendly constituents to achieve the same functions as fluoride-containing mould flux is essential to the continuous casting process. It was suggested that the combination of B2O3 and Na2O is an appropriate substitute for CaF2, providing low viscosity and melting points, and acceptable crystallisation and heat transfer of mould fluxes. [2, 48, 150] However, B2O3, as one of substitutes for fluoride in the mould flux, has a low melting temperature and its vapour has a relatively high pressure at high temperatures. [151] Besides, high volatility of sodium- and boron-containing compounds can be a limitation in the industrial use of the Na2O-containing boracic fluxes. [ ] Therefore, a lot of efforts are made to work out the extent of volatility, factors affecting the volatility, and the effect of evaporation on the flux properties Measurement of evaporation Thermogravimetric technique Evaporation of mould flux and other metallurgical fluxes in different gas atmosphere is usually measured by thermogravimetric analysis (TGA). [59, 151, 153, 155, 156] Figure 2-15 shows an example of the experimental apparatus. At first, the flux sample was heated to 49

78 a pre-set temperature with a given heating rate, which is usually the maximum heating rate for the apparatus. The sample was kept at the experimental temperature for a period of time, and then cooled down. During the measurement, the weight loss of the flux sample is recorded. In the case of TGA measurements with high gas flow rates, it is significant to stabilise the gas flow rate to obtain stable data, as the sampling error can become large due to vibration of the crucible. [155] Figure 2-15 TGA furnace for the measurement of gas vaporisation. [59] Knudsen cell mass spectrometer Knudsen cell mass spectrometry was developed to measure vapour pressures of alloys and slags. [ ] In this method, vapour pressure Pi, Pa, is measured as ion current Ii, A, which is proportional to the pressure: [161] P i = T S i I i (2-67) 50

79 where T, K, is the temperature of neutral species (i.e. of the Knudsen cell at the time of its evaporation) and Si is the device dependent constant which includes such factors as ionisation cross-section and efficiency of the ion detector Single hot thermocouple technique The qualitative study of the evaporation of mould fluxes could be performed using single hot thermocouple technique (SHTT). [162] In SHTT, heating and cooling rates were up to ±30 C/s by the system control and up to ±220 C/s by the manual control. A B-type (Pt-30%Rh/Pt-6%Rh) thermocouple with 0.5 mm diameter was used to measure and control temperature. Before the measurement, the weight of hot thermocouple was measured. A schematic view of the apparatus is shown in Figure The details of the apparatus were presented elsewhere. [61] During the measurement, a flux sample (approximately 4 mg) was mounted on the tip of the thermocouple at a given temperature, cooled with the highest cooling rate of the apparatus. The weight of flux sample with thermocouple was measured afterwards. Since the amount of the sample is small, another way to determine the extent of evaporation is to analyse the change of samples chemical compositions using scanning electron microscope analysis (SEM) coupled with dispersive X-ray spectroscopy (EDX). [162] 51

80 Figure 2-16 Schematic diagram of single hot thermocouple technique. [163] Evaporation kinetics As TGA measurement is widely applied to investigate the volatility of mould fluxes, here the possible evaporation process of mould flux in this measurement is discussed, which could include four main steps: [59] (1) Evaporation reactions in the molten flux; (2) Internal mass transfer of the anions and cations to the surface of the molten flux; (3) External mass transfer of gaseous species from the flux/gas interface out of the crucible to the main gas stream; (4) For the fluxes with high distance from the liquid surface to the bottom of the crucible, transport of bubbles from bulk liquid slag to the slag/gas interface through the liquid boundary layer. 52

81 Figure 2-17 Schematic of the evaporation process. [151] Chemical reaction There is sufficient research of evaporation of fluorine-containing flux. [59, 60, 164] Some of the features of the evolution of fluoride containing mould fluxes in dry environment are shown in Figure Formations of NaF and KF are at lower temperatures than those of SiF4 and AlF3. [59] For humid environment, the emission of gaseous species contains HF, SiF4, NaF, AlF3 etc.. [59, 60, 164] For boron-containing fluorine-free fluxes, formation [151, 153, 154, 157] of gaseous species can be NaBO2, Na, O2 and B2O3. 53

82 Figure 2-18 A characteristic of off-gassing from an industrial mould flux. [60] Mass transfer in the molten flux Whether liquid mass transport contributes to evaporation of the fluxes depends on flux compositions and experiment conditions. The evaporation mechanism in a ternary CaF2- SiO2-CaO slag system indicated that the liquid phase mass transport of SiO4 4- controlled the evaporation rate in part. [164] Increasing CaO/SiO2 ratio decreases the viscosity of the flux as a result of the flux depolymerisation. It is expected in accordance with Stokes- Einstein equation that diffusion coefficients of flux constituents increase with the increase in the CaO/SiO2 ratio. [165] In some research, the influence of liquid mass transfer is marginal. A negligible effect of CaO/SiO2 ratio from 0.83 to 1.25 and 0.77 to 1.18 on the evaporation of CaO-SiO2-Al2O3-B2O3-Na2O and CaO-SiO2-Al2O3-B2O3- Na2O-TiO2 mould fluxes was observed in work [151]. However, additions of TiO2 or ZrO2 increased the evaporation rate of the fluxes; therefore, in this condition, the evaporation could be influenced by liquid mass transfer. [151] The mass transfer in the liquid flux 54

83 gradually became a controlling step after a long time of the evaporation reaction for CaO-SiO2-Al2O3-Na2O fluxes was reported by Li et al.. [153] External mass transfer in the gaseous phase The transport of gaseous vapour from the surface of the molten flux to the bulk gas through gaseous boundary layer needs to be considered. When the external mass transfer contributes to the rate control, the evaporation of the flux could be influenced by the types of main stream gas. In Shimizu s research, [166] the increases in the rate of weight loss for both CaF2-SiO2-CaO and commercial mould fluxes were observed by switching the stream gas from N2 to He at C. Similarly, Kashiwaya and Cramb [155] found that the sequence of the evaporation rate of NaF in different stream gases was He>N2>Ar. However, Zhang et al. [151] and Li et al. [153] assumed that the transport of gaseous phase may not be the controlling step under their experimental conditions. 2.5 Summary and project outline Development of the fluorine-free mould fluxes is important for decreasing environmental impact, equipment corrosion, and health hazards brought about by the presence of fluorine in the conventional mould fluxes in steel continuous casting process. Among all the substitutes, the combination of B2O3 and Na2O could be a promising candidate to replace the fluorides and achieve required physicochemical properties of fluxes for continuous casting. However, there is very limited experimental work in this research area. This project aims to examine high-temperature physicochemical properties of B2O3-containin fluorine-free mould fluxes, including 55

84 melting properties, viscosity, and stability of fluorine-free fluxes. Specific objectives of this project are: (1) To select composition range of fluorine-free mould fluxes based on thermodynamic calculation; (2) To study melting properties and viscosity of fluorine-free mould fluxes and to investigate relationship between these properties and flux composition by correlating with the flux structure; (3) To develop an applicable model to describe the viscosity of fluorine-free moulds fluxes; (4) To investigate the stability of fluorine-free mould flux and analyse the determining step for flux evaporation. The results of this project will provide a further understanding of fluorine-free mould fluxes and their physiochemical properties, which will lead to the development of environmentally friendly mould fluxes for steel continuous casting. 56

85 Chapter 3 - Experimental procedure 3.1 Sample preparation Fluxes were prepared using reagent grade CaCO3, SiO2, Al2O3, B2O3, Na2CO3, TiO2, MgO, Li2CO3, MnCO3 and ZrO2 with some carbonates being substitutes for oxides due to their stability in air. The purity of these chemicals is shown in Table 3-1. Chemical compositions of fluxes and their selection will be presented in Chapter 4. Table 3-1 Purity of chemicals used in present study. Chemical CaCO3 SiO2 Al2O3 B2O3 Na2CO3 Purity, wt% Chemical MgO MnCO3 Li2CO3 TiO2 ZrO2 Purity, wt% > The blending of components was fully ground in an agate mortar for 20 min and then placed in a high-purity graphite crucible. When the temperature of heating furnace reached 1400 C, the crucible was inserted into it and held for 20 min. After that, the crucible was then removed and the sample was poured onto a steel plate at room temperature. Subsequently, the bulk flux sample was crushed and ground, forming homogenised fine powders for use. 3.2 Melting properties measurement Melting properties of fluxes were determined using hot stage microscopy method. Experimental equipment (model II-AP, Japan) is composed of a high-temperature tube furnace with an accurate high-temperature controller, video image recorder and 57

86 processing system (shown in Figure 3-1). The temperature gradients in the melting furnace for the high-temperature microscopy are within 2 C. The test consisted of heating a compacted cylinder flux (Φ3 3 mm) at a controlled rate (15 C/min), and monitoring the changes in sample shape and dimension. As shown in Figure 3-2, softening temperature Ts, hemispherical temperature Th, and fluidity temperature Tf are defined as the temperatures at which the height of sample reached 75%, 50% and 25% of the original value, respectively. Figure 3-1 Schematic diagram of high-temperature microscopy. 58

87 Figure 3-2 Images of flux morphology change during the melting process: (a) original height, and heights of (b) softening temperature, (c) hemispherical temperature, and (d) fluidity temperature. 3.3 Viscosity measurement Flux viscosity was measured using a rotation viscometer (model ZC-1600, China). The experimental setup for the viscosity measurement is shown in Figure 3-3, which consists of a rotating system, a heating system, a measuring system and a data analysis system. The temperature was controlled within ±2 C with a proportional integral differential controller and a Pt/Rh thermocouple (B-type), which was placed just beneath the bottom of a high purity graphite crucible with 40 mm inner diameter and 59

88 160 mm inner height. Both positions of thermocouple and crucible were in the uniform temperature zone of the furnace. The temperature difference between the bottom of the crucible and molten flux is within 10 C, which has been calibrated before measurement. The hot zone with a constant temperature is 20 mm longer than the depth of the molten flux in the furnace. The fluctuation of temperature within the constant temperature zone is within 3 C. The spindle used for measurement was made of Mo with a diameter of 15 mm and a total height of 30 mm. Calibration of the viscometer was conducted using standard castor oil. Pre-melted flux of 140 g was used in the viscosity measurement. Nitrogen gas was used as the protective gas through the whole experimental process to prevent the oxidation of the crucible and the spindle. The rotating speed of the spindle was kept at 12 rpm. When the temperature reached 1400 C, the sample was held in the furnace for about 20 min to ensure that the flux was completely melted and homogeneous. Then the spindle was carefully immersed into the molten slag. Subsequently, the viscosity was continuously measured while the temperature of the flux decreased with a controlled speed of 5 C/min. The viscosity obtained at 1400 C can therefore be considered as a static measurement, while those measured at subsequent lower temperatures during cooling process are dynamic measurements. Fluxes can be presented in the form of Newtonian or non-newtonian fluxes. [7,37] In the case of non-newtonian molten flux, such as liquidsolid coexistent flux, the apparent or effective viscosity is sometimes used. In this research, we focus on Newtonian flux state and use a common term viscosity to represent both cases. After the viscosity measurement, the sample was reheated to 1300 C, and poured into a copper crucible at room temperature. 60

89 Figure 3-3 Schematic diagram for viscosity measurement. 3.4 Evaporation measurement The evaporation experiments were conducted using a STA449-F1 calorimeter (NETZSCH Instruments, Germany) in the isothermal mode at 1300, 1325, 1350 and 1400 C in Ar atmosphere. Not all temperatures were applied for all samples due to the limitation on using the apparatus for high-temperature experiments. The schematic diagram of the apparatus is shown in Figure 3-4. The Pt/Rh crucible with 6 mm inner diameter and a volume of 84 μl was employed as the container for the flux. The flux sample for each measurement was approximately 30 mg. 61

90 Figure 3-4 Schematic diagram of evaporation measurement. The chamber was evacuated and purged with argon gas for 300 s to ensure a steady gas flow. The gas flow rate was kept constant at 70 ml/min for the duration of the experiment, which is the maximum gas flow rate possible in the measurement. The temperature profile of evaporation experiment at 1400 C is shown as an example in Figure 3-5. The sample was firstly heated to a pre-set temperature with a heating rate of 50 C/min, the maximum heating rate for the apparatus. After that, the sample was kept at the experimental temperature for 60 min. During the experiment, the weight change of the sample was recorded every 0.1 s. 62

91 Figure 3-5 Temperature profile for the evaporation experiment at 1400 C. 3.5 Raman spectroscopy The structure of as-quenched fluorine-free fluxes from 1400 C was studied by Raman spectroscopy using Renishaw invia Raman spectrometer. The measurements were performed at room temperature in the frequency range of cm 1 using excitation wavelength of 514 nm argon ion laser with 1800 l/mm grating. The beam spot size is approximately 1.5 μm in diameter. The Raman shift was collected with backscattering geometry using a 50 microscope objective lens. After measurements, the normalised Raman spectra were deconvoluted based on Gaussian distribution using WiRE 3.4 software. 63

92 Chapter 4 - Fluorine-free mould flux design and flux equilibrium phase calculation The fluorine-free mould fluxes used in this work are based on the SiO2-CaO-Al2O3- B2O3 system with the addition of other different oxide components, forming 4-, 5-, 6-, 8- and 10-component systems, respectively. The purpose of this design is to understand the effects of different oxide components on flux structure and physiochemical properties, from simple system to more complex multi-component systems. The selection of composition range of fluorine-free fluxes is based on literature review but also equilibrium flux phase calculation using thermochemical software FactSage. This chapter lists all compositions of designed fluxes for the test in this work, from simple 4- component to complex 10-component systems. The equilibrium phase compositions of some selected fluxes are also calculated, and the effects of CaO/SiO2 ratio and individual oxide component on phase formation are determined. This information is useful for flux property prediction and experimental result evaluation in the following chapters. 4.1 Compositions of designed fluxes Total 51 fluxes were prepared, based on the base system of CaO-SiO2-Al2O3-B2O3 with varied additions of Na2O, TiO2, MgO, Li2O, MnO and ZrO2, from simple 4-component, to 5-, 6-, 8-, and further to 10-component systems. The designed and measured compositions of all fluxes are shown in Tables 4-1 and 4-2, respectively. The concentrations of B2O3 and Li2O were determined by inductively coupled plasma (ICP) analysis and all others were analysed by X-ray fluoroscopy (XRF). Varied extents of 64

93 loss of components were observed for B2O3, Na2O, MgO and Li2O due to possible evaporation of these components during flux preparation. Table 4-1 Designed composition of mould fluxes (wt%). Flux No. C/S CaO SiO2 Al2O3 B2O3 Na2O TiO2 MgO Li2O MnO ZrO

94

95 *C/S is referred as CaO/SiO2 weight ratio. Table 4-2 Measured compositions of mould fluxes (wt%). Flux No. C/S CaO SiO2 Al2O3 B2O3 Na2O TiO2 MgO Li2O MnO ZrO

96

97 *C/S is referred as CaO/SiO2 weight ratio. 4.2 SiO2-CaO-Al2O3-B2O3 fluxes As shown in Table 4-1, for this 4-component system, the CaO/SiO2 ratio was selected in the range of and B2O3 content in the range of 5-9 wt%. This selection was based on the results of equilibrium phase calculation applied for some selected compositions of this base system listed in Table 4-3, where CaO/SiO2 ratio varies from 0.7 to 1.5 and B2O3 from 5 to 11 wt%. Phase equilibria of the designed CaO-SiO2- Al2O3-B2O3 fluxes with different CaO/SiO2 ratios are shown in Figure 4-1. Solidus temperature (Tsol) and liquidus temperature (Tliq) are the minimum and maximum temperatures at which the liquid and solid phase coexist, respectively. [167] Both are 69

98 thermodynamic quantities and could be determined by the thermodynamic phase diagram of fluxes. Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq derived from Figure 4-1 are also listed in Table 4-3. It was found that Tsol increased with the increase of CaO/SiO2 ratio from 0.7 to 1.5. The value of Tliq increased with increasing CaO/SiO2 ratio from 0.7 to 1.0, and declined when the CaO/SiO2 ratio increased further to 1.5 (Table 4-3). Increase in B2O3 content had no significant influence on Tsol but reduced the value of Tliq for fluxes with a fixed CaO/SiO2 ratio. Phase equilibrium calculation revealed that CaSiO3 was the main phase at low CaO/SiO2 ratio of 0.7. Increase in concentration of B2O3 from 5 to 11 wt% had little influence of the predominant phase (CaSiO3), promoted the formation of Ca2B2SiO7, and inhibited CaAl2Si2O8 (Figures 4-1a-d). When the CaO/SiO2 ratio changed from 0.7 to 1.0, the main crystal phase remained CaSiO3 with an increased weight fraction; the influence of B2O3 content at this CaO/SiO2 ratio showed the same influence as that of fluxes with CaO/SiO2 ratio 0.7 (Figures 4-1e-h). The crystallisation phenomenon of fluxes with CaO/SiO2 ratio of 1.5 is quite different from that of fluxes with lower CaO/SiO2 ratios (0.7 and 1.0). With the increase in CaO/SiO2 ratio from 1.0 to 1.5, the main crystal phase changed from CaSiO3 to Ca3Si2O7, especially for fluxes with lower B2O3 concentrations (5 and 7 wt%). Increase in B2O3 content from 5 to 11 wt% of fluxes with this CaO/SiO2 ratio decreased the content of Ca11B2Si4O22 and Ca3Si2O7, but facilitated the formation of Ca2B2O5 and CaSiO3. When B2O3 content was up to 11%, almost no Ca11B2Si4O22 formed in the flux. Among the above-mentioned hightemperature phases, Ca11B2Si4O22 was proposed as a substitute for cuspidine in fluorinecontaining fluxes, as it tends to precipitate in the low temperature zone that is similar to 70

99 cuspidine in the conventional fluorine-containing mould fluxes. [53, 168] Therefore, taking into account of both appropriate Tliq and dominant high-temperature phases, the optimal composition range of CaO-SiO2-Al2O3-B2O3 fluxes is CaO/SiO2 ratio within and B2O3 content within 5-9 wt%. 71

100 Figure 4-1 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-B2O3 fluxes using FactSage: (a) 0.7 CaO/SiO2, 5 wt% B2O3; (b) 0.7 CaO/SiO2, 7 wt% B2O3; (c) 0.7 CaO/SiO2, 9 wt% B2O3; (d) 0.7 CaO/SiO2, 11 wt% B2O3; (e) 1.0 CaO/SiO2, 5 wt% B2O3; (f) 1.0 CaO/SiO2, 7 wt% B2O3; (g) 1.0 CaO/SiO2, 9 wt% B2O3; (h) 1.0 CaO/SiO2, 11 wt% B2O3; (i) 1.5 CaO/SiO2, 5 wt% B2O3; (j) 1.5 CaO/SiO2, 7 wt% B2O3; (k) 1.5 CaO/SiO2, 9 wt% B2O3; and (l) 1.5 CaO/SiO2, 11 wt% B2O3. 72

101 Table 4-3 Composition (wt%), solidus temperature Tsol, liquidus temperature Tliq, and dominant high-temperature phases of SiO2-CaO-(3 wt%)al2o3-b2o3 fluxes. Flux No. C/S B2O3, wt% Tsol, C Tliq, C Dominant high-temperature phases D CaSiO3, SiO2, CaAl2Si2O8 D CaSiO3, SiO2 D CaSiO3, SiO2 D CaSiO3, SiO2, Ca2B2SiO7 D >1400 D D D D D D D CaSiO3, Ca2B2SiO7, CaAl2Si2O8, Ca2Al2SiO7 CaSiO3, Ca2B2SiO7, CaAl2Si2O8, Ca2Al2SiO7 CaSiO3, Ca2B2SiO7, CaAl2Si2O8, Ca2Al2SiO7 CaSiO3, Ca2B2SiO7, CaAl2Si2O8, Ca2Al2SiO7 CaSiO3, Ca11B2Si4O22, Ca3Si2O7 CaSiO3, Ca11B2Si4O22, Ca3Si2O7, Ca2B2O5, Ca2Al2SiO7 CaSiO3, Ca2B2O5, Ca3Si2O7, Ca11B2Si4O22, Ca2B2SiO7 CaSiO3, Ca2B2O5, Ca2B2SiO7, Ca3Si2O7,Ca11B2Si4O22 The equilibrium phase diagram of the designed CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3 fluxes calculated using FactSage for varying CaO and SiO2 concentrations is shown in 73

102 Figure 4-2. The influences of CaO/SiO2 ratio on dominant phases have been described above. Zero phase fraction (ZPF) curve of Ca11B2Si4O22 is marked with solid square in the phase diagram, which shows the formation boundaries of this phase. If Ca11B2Si4O22 is the targeted phase, CaO/SiO2 ratio should be over 1.2 (marked by dashed lines in Figure 4-2), which is in the range of CaO/SiO2 ratio shown in Table 4-1. Figure 4-2 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3 fluxes using FactSage. 4.3 SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) fluxes The addition of Na2O into SiO2-CaO-Al2O3-B2O3 fluxes is to promote the crystallisation of mould fluxes. [71] However, preliminary experiments showed that when the concentration of Na2O is over 11 wt%, the crystallisation experiment of the fluxes is too fast to be conducted. Therefore, the range of Na2O content is proposed within 7-9 wt%. The equilibrium phase diagram of the designed CaO-SiO2-(3 wt%)al2o3- (7 wt%)b2o3- (9 wt%) Na2O fluxes calculated using FactSage for different CaO/(CaO+SiO2) mass 74

103 ratio is shown in Figure 4-3. The dominant effect of addition of Na2O into CaO-SiO2- Al2O3-B2O3 system is the expansion of the Ca11B2Si4O22 area and decreased Tliq. When CaO/SiO2 ratio is located within the range from 1.2 to 1.4, the fluxes showed the lowest Tliq. In this thesis work, Na2O content was selected in the range of 7-11 wt%. Figure 4-3 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o fluxes using FactSage. The addition of TiO2 into CaO-SiO2-Al2O3-B2O3 fluxes is suggested to form possible Ti-containing phases such as CaSiTiO5, to control heat transfer between mould wall and strand. [28] However, the addition of TiO2 would lead to higher crystallisation temperature, and the formation of CaTiO3 could deteriorate the lubricity of the fluxes. [63] The range of TiO2 content is proposed within 2-6 wt%. The equilibrium phase diagram of the designed CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3-(4 wt%)tio2 fluxes calculated using FactSage is shown in Figure 4-4. The addition of TiO2 into CaO-SiO2-Al2O3- B2O3 flux moved the formation of Ca11B2Si4O22 towards a higher CaO/SiO2 ratio (1.3). 75

104 Minor addition of MgO into the fluxes decreased melting properties [35] and viscosity [28, 29, 101, 103, 105]. However, there is research showed that increase in MgO increased crystallisation temperature and caused unstable heat transfer in industrial practice. [169] The addition of MgO is within the range from 1 to 5 wt%. Because of this consideration, TiO2 (2-6 wt%) and MgO (1-5 wt%), as shown in Table 4-1, were added to investigate the effects of their additions. Figure 4-4 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(4 wt%)tio2 fluxes using FactSage. 4.4 Complex 8- and 10-component fluxes The equilibrium phase diagram of the designed 8-component SiO2-CaO-(3 wt%)al2o3- (7 wt%)b2o3-(9 wt%)na2o-(4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o fluxes calculated using FactSage for varying CaO concentrations is shown in Figure 4-6. ZPF curves of CaSiO3, merwinite (Ca3MgSi2O8) and Ca11B2Si4O22 are highlighted in the phase diagram. Similar to SiO2-CaO-Al2O3-B2O3 fluxes (Figure 4-2), CaSiO3 is favoured to 76

105 precipitate at low concentration of CaO. With the increase in CaO content, merwinite and Ca11B2Si4O22 form in replacement of CaSiO3. When CaO/SiO2 ratio is above 1.4 (0.43 wt% CaO), Tliq increases significantly, and Ca11B2Si4O22 forms prior to merwinite with the decrease of temperature. The designed composition range (CaO/SiO2 ratio from 0.9 to 1.4) marked in Figure 4-6 shows relatively low Tliq. When the range of CaO/SiO2 ratio is within , Tliq reaches the lowest value. As the analyses of equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3 show that the weight fractions of Ca11B2Si4O22 increased with the increase of CaO/SiO2 ratio (Figure 4-1), this ratio should also affect 8-component SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o- (4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o and 10-component SiO2-CaO-(3 wt%)al2o3- (7 wt%)b2o3-(9 wt%)na2o-(4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o-(2 wt%)mno-(1 wt%)zro2 fluxes, which will be investigated using Fluxes shown in Table 4-1. Figure 4-5 Phase diagram of designed SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o-(4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o fluxes using FactSage. [61] 77

106 4.5 Summary and conclusion Thermodynamic calculation provides the guidance for boron-containing fluorine-free mould flux composition selection. The compositions for test fluorine-free mould fluxes were designed. The major findings are summarised as follows: (1) For CaO-SiO2-Al2O3-B2O3 fluxes, Tliq rose with increasing CaO/SiO2 ratio from 0.7 to 1.0, and declined with the further increase of CaO/SiO2 ratio to 1.5. Increase in B2O3 content from 5 to 11 wt% decreased Tliq monotonically. With the formation of the targeted phase of Ca11B2Si4O22, the composition range of CaO-SiO2-Al2O3-B2O3 fluxes is CaO/SiO2 ratio within and B2O3 content within 5-9 wt%. (2) The addition of Na2O into CaO-SiO2-Al2O3-B2O3 system expanded the Ca11B2Si4O22 area, while the addition of TiO2 content had the opposite influence. With the various influencing factors taken into account, the matrix of 5- component flux should be CaO-SiO2-Al2O3-B2O3-(9 wt%)na2o/(4 wt%)tio2/(3 wt%)mgo. (3) By taking into account of both low Tliq and possible formation of Ca11B2Si4O22 phase, the matrix range of CaO/SiO2 ratios for SiO2-CaO-(3 wt%)al2o3-(7 wt%)b2o3-(9 wt%)na2o-(4 wt%)tio2-(3 wt%)mgo-(1 wt%)li2o fluxes is within 1.2 and 1.3. (4) Total 51 fluorine-free mould fluxes, covering simple 4-component system, to 5- and 6-component, and further to complex 8- and 10-component systems, were designed based on above thermodynamic phase calculation and the composition range for the formation of Ca11B2Si4O22 phase which was reported to be the 78

107 phase to replace cuspidine. The composition selection was designed for the variation of CaO/SiO2 ratio and other addition oxide compositions. 79

108 Chapter 5 - Melting properties and viscosity of SiO2-CaO-Al2O3-B2O3 fluxes According to Chapter 4, the fluorine-free mould fluxes investigated in this thesis are all based on SiO2-CaO-Al2O3-B2O3 with the addition of some other oxide components. Therefore, it is necessary to investigate the melting properties and viscosity of this base system first. The compositions of total 6 fluxes of SiO2-CaO-Al2O3-B2O3 are listed in Table 4-2, where Fluxes 1, 2, 4 and 6 are with a fixed B2O3 content but varied CaO/SiO2 mass ratio from 0.8 to 1.5, and Fluxes 3, 4 and 5 are with a fixed CaO/SiO2 mass ratio but varied B2O3 contents from 4.7 to 8.4 wt%. 5.1 Melting properties Figure 5-1 shows the effects of CaO/SiO2 ratio and B2O3 addition on the softening, hemispherical and fluidity temperatures of the SiO2-CaO-Al2O3-B2O3 fluxes (Fluxes 1-6 in Table 4-2). With the increase of CaO/SiO2 from 0.8 to 1.0, Th increased slightly. Afterwards, Th decreased from 1304 C to 1160 C, with CaO/SiO2 ratio increasing from 1.0 to 1.5. Change in the B2O3 concentration from 4.7 to 8.4 wt% had no significant effect on Th, fluctuating in the range of C. Changes of Tf and Ts with varied CaO/SiO2 ratio and B2O3 content showed the similar trend as that of Th. 80

109 Figure 5-1 Effects of (a) CaO/SiO2 ratio and (b) B2O3 content on Ts, Th, and Tf of SiO2- CaO-Al2O3-B2O3 fluxes. 5.2 Viscosity, break temperature and activation energy Figure 5-2 shows the viscosity change of the SiO2-CaO-Al2O3-B2O3 fluxes with different CaO/SiO2 ratios and B2O3 concentrations as a function of temperature. As a whole, the decrease in temperature increases flux viscosity. The break temperature Tbr, defined in Section 2.3.3, was determined by the turning point on viscosity-temperature curve at which the viscosity increased sharply, and the results are shown in Table 5-1. There was no such a sharp viscosity change for Flux 5 (CaO/SiO2 ratio 1.3, B2O3 concentration 8.4 wt%) and therefore no corresponding Tbr was given. The influence of CaO/SiO2 ratio on Tbr seems a bit complex. When CaO/SiO2 ratio increased from 0.83 to 1.0 (Fluxes 1 and 2), Tbr increased. Afterwards, Tbr decreased sharply (Fluxes 2, 4 and 6). B2O3 addition from 4.7 to 6.4 wt% (Fluxes 3 and 4) reduced Tbr. 81

110 Figure 5-2 Effects of (a) CaO/SiO2 ratio and (b) B2O3 content on viscosity of SiO2- CaO-Al2O3-B2O3 fluxes. Effect of temperature on viscosity of molten flux can be described by the Arrhenius equation according to Eq When the temperature is above Tbr, EA is constant for a fixed composition and can be acquired from the slope of the lnη-1/t. [48] Viscosity at 1400 C and activation energy derived from the viscosity-temperature curves are also listed in Table 5-1. There was no EA deduced for Flux 3 because there were no enough points for linear fitting. Increase in CaO/SiO2 from 0.8 to 1.3 (Fluxes 1, 2 and 4) decreased viscosity, and with further increase in the CaO/SiO2 ratio from 1.3 to 1.5 (Fluxes 4 and 6) viscosity kept constant at 1400 C. Increase in CaO/SiO2 from 0.8 to 1.3 (Fluxes 1, 2 and 4) decreased EA first and then increased EA slightly. A further increase in the CaO/SiO2 ratio from 1.3 to 1.5 (Fluxes 4 and 6) decreased EA. Increase in B2O3 concentration from 4.7 to 8.4 wt% in the flux with the CaO/SiO2 ratio 1.3 (Fluxes 3 to 5), decreased both viscosity and activation energy. 82

111 Table 5-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3 fluxes. Flux No. η (1400 C ), Pa s Tbr, C EA, kj/mol Raman spectroscopy Raman spectra in the frequency range of 500 to 1600 cm -1 of quenched fluxes are shown in Figure 5-3. A distinct hump in the frequency region between 500 and 750 cm - 1 (marked by dashed lines in Figure 5-3) and a large hump with several peaks between 750 and 1100 cm -1 was observed in all spectra. The increase in CaO/SiO2 ratio from 0.8 to 1.5 shifted the position of the low frequency hump from 621 to 671 cm -1 (Figure 5-3a). With a fixed CaO/SiO2 at 1.3, change of B2O3 content did not change the position of this low frequency hump (Figure 5-3b). Two small envelopes in the range of cm -1 and cm -1 were also found for all spectra. However, for Flux 1 with CaO/SiO2 ratio 0.8 and B2O3 concentration 6.5 wt%, the envelope in the range of cm -1 was too small to be identified. 83

112 Figure 5-3 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3 fluxes with different (a) CaO/SiO2 ratio and (b) B2O3 content in the frequency range cm Relationship between characteristic temperatures and phase composition Melting properties (Ts, Th, and Tf) are closely related to the formation and melting of solid phases in flux heating process. A sharp increase in the flux viscosity at Tbr is a result of precipitation of solid phase in flux cooling process. Figure 5-4 shows calculated equilibrium phases of Fluxes 1-6 using FactSage. Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq are listed in Table 5-2. The value of Tliq slightly increased from 1379 to 1399 C with increasing CaO/SiO2 ratio from 0.8 to 1.0, and decreased to 1315 C when the CaO/SiO2 ratio increased further to 1.5. Increase in B2O3 content from 4.7 to 6.4 wt% decreased Tliq slightly (Table 5-2). Ts, Th and Tf are located between Tliq and Tsol. In general, Tbr is lower than Tliq, and in most occasions, Tliq can give certain guidance for the estimation of Tbr. As Tbr has different physical interpretation with Tf, the solid fraction of molten slags at Tf might be greater than that at Tbr, and therefore, the value of Tbr could be higher than that of Tf. This is true in this work where a higher Tbr was observed than Tf (Fluxes 2, 3, 4 [10, 27] and 6), which is also consistent with other reports. 84

113 The increase of the CaO/SiO2 mass ratio stimulates crystallisation, while B2O3 addition has the opposite effect. [61, 150] When the CaO/SiO2 ratio changed from 0.8 to 1.0, the main crystal phase remained CaSiO3 (Figures 5-4a,b); further increase in CaO/SiO2 ratio from 1.0 to 1.5, the main crystal phase changed from CaSiO3 to Ca3Si2O7 (Figures 5-4b,d,e), which is consistent with other reports. [58, 71] This implies that the number of bridging oxygen atoms bonded to Si decreases with the increase in CaO/SiO2 ratio. Increase in concentration of B2O3 from 4.7 to 8.4 wt% (Figures 5-4b,c,d) decreased the content of Ca11B2Si4O22 and Ca3Si2O7, but promoted the formation of Ca2B2O5 and CaSiO3. The introduction of B2O3 promotes the formation of small anionic cluster B-O- B and increases the number of bridging oxygen atoms attached to Si. No Tbr was obtained for Flux 5 which could be related to a possible glassy phase formation during the solidification. The same tendency in viscosity change was also reported by Shu et al. [21] when the concentration of B2O3 was high. This observation implies that the increase in B2O3 concentration restrains crystallisation, which has been confirmed in work [71]. 85

114 Figure 5-4 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-B2O3 fluxes using FactSage: (a) 0.8 CaO/SiO2, 6.5 wt% B2O3; (b) 1.0 CaO/SiO2, 6.6 wt% B2O3; (c) 1.3 CaO/SiO2, 4.7 wt% B2O3; (d) 1.3 CaO/SiO2, 6.4 wt% B2O3; (e) 1.3 CaO/SiO2, 8.4 wt% B2O3; and (f) 1.5 CaO/SiO2, 6.7 wt% B2O3. 86

115 Table 5-2 Parameters derived from calculated equilibrium phase fractions of CaO-SiO2- Al2O3-B2O3 fluxes. Flux No. Tsol, C Tliq, C Dominant high-temperature phases CaSiO3, Ca2B2SiO CaSiO3, Ca2B2SiO7, Ca2Al2SiO CaSiO3, Ca3Si2O7, Ca11B2Si4O CaSiO3, Ca3Si2O7, Ca11B2Si4O22, Ca2B2O CaSiO3, Ca2B2O Ca3Si2O7, CaSiO3, Ca11B2Si4O22, Ca2B2O5 5.5 Viscosity and structure of molten flux Viscosity of liquid silicate fluxes is related to the structure of the liquid system. The flux structure is presented as a partially ionic liquid with groups of anions of various degrees of polymerisation. In the fluxes, silicon-containing units are expected to play a major role in the network formation. Silicate fluxes consist of interconnected networks of tetrahedral SiO4 4-. Cations Al 3+ can substitute for Si 4+ to form tetrahedral AlO4 5-, with electric charge compensation by cations. [92] However, concentration of alumina in mould fluxes ( wt%) was much lower than that of silica ( wt%). Therefore silicon-containing units are expected to play a major role in the network formation. Different types of anionic units (SiO2, Si2O5 2-, Si2O6 4-, Si2O7 6-, and SiO4 4- ) could coexist in the melt. [13] It follows from the Raman spectra that with the increase of CaO/SiO2 ratio from 0.8 to 1.5, the band in low frequency range is shifted to a higher frequency. 87

116 The band in the range of cm -1 is generally regarded as Si-O-Si bending modes. The shifting of the low frequency Raman bands to higher frequencies could be the result of the decrease in the Si-O-Si bond angle or the decrease of polymerisation. [84, 170] The increase in the B2O3 concentration from 4.7 to 8.4 wt% had no apparent effect on the shift of this band. The shift of the line does not depend on the species of alkali or alkaline earth elements but on the concentration of the basic oxide. [171] The band located around 1450 cm -1 is assigned to B-O - stretching vibration in BO3 3- trigonal units attached to other large borate groups in high frequency region. [ ] Humps in the range of 800 to 1300 cm -1 were fitted by five Gaussian peaks with R 2 > 99.9% using Gaussian deconvolution method (shown in Figure 5-5). The positions of the peaks in the range of 800 to 1300 cm -1 are around 860 cm -1, 910 cm -1, 960 cm -1, 1050 cm -1 and 1190 cm -1, respectively. The large hump between 800 and 1100 cm -1 reflects the stretching vibration of various structural units, including Q 0 (SiO4 4- ) monomers, Q 1 (Si2O7 6- ) dimers, Q 2 (SiO3 2- ) chains or rings, and Q 3 (Si2O5 2- ) sheets. The small envelope between cm -1 represents Q 4 (SiO2), fully polymerised threedimensional units. [82, 84] The superscript of Q represents the number of bridging oxygen. [82, 84] The frequencies associated with the Q i species remain essentially [170, 176] unchanged, which is in agreement with other reports. The molar fractions of various silicate structural units Q i were estimated by calculating the area fractions of various Gaussian peaks as follows: X Q i = A Q i 4 i=0 A Q i (5-1) where X Q i is the molar fraction of Q i, A Q i is the integrated area of Gaussian peak corresponding to Q i. 88

117 The fractions of different silicate units are listed in Table 5-3. For all fluxes, Q 2 was the dominant unit (except Flux 1 where Q 2 and Q 3 are with a similar fraction), while the fraction of Q 4 was quite low. No Q 1 unit was identified for Fluxes 1 (CaO/SiO2 ratio = 0.8) and 2 (CaO/SiO2 ratio = 1.0), and no Q 4 unit was distinguishable in the spectra for Flux 1 (CaO/SiO2 ratio = 0.8). Kline et al. [176] also found no Q 1 for CaO-SiO2-B2O3 fluxes with low CaO/SiO2 ratio. A relatively low concentration of alumina in mould fluxes examined in this paper and, therefore, a small amount of AlO4 5- units had a minor effect on the silicate units. It was shown in work [165] that alumina contributes primarily to the formation of Q 4, with some effect on Q 2 and Q 3 units. [115] With the increase of CaO/SiO2 ratio from 0.8 to 1.5, fractions of Q 0 and Q 1 increased, while that of Q 3 decreased, indicating that the degree of polymerisation reduced with increasing CaO/SiO2 ratio. This change in the Raman spectra was consistent with the results reported for the CaO-SiO2-B2O3 system. [176] The increase of the CaO/SiO2 ratio promoted reforming of complex network units and lowered the viscosity of the flux system. With increasing B2O3 concentration in the flux with the fixed CaO/SiO2 ratio of 1.3, the fractions of Q 0 and Q 2 decreased, while Q 1 increased. The same effect of B2O3 content was also observed in Sun and Zhang s research. [177] However, a change in the fraction of dimers Q 3 with B2O3 concentration was inconspicuous. Depolymerisation of silicate fluxes can be characterised by non-bridging oxygen per silicon NBO/Si, which can be determined from the Raman spectrum as: [178] NBO/Si = 4X Q 0 + 3X Q 1 + 2X Q 2 + X Q 3 (5-2) The values of NBO/Si are also listed in Table 5-3. NBO/Si increased with the increase in the CaO/SiO2 ratio from 0.8 to 1.5 at a fixed B2O3 content (6.6 wt%), while it 89

118 decreased slightly with increasing B2O3 content from 4.7 to 8.4 wt% at a fixed CaO/SiO2 ratio (1.3). Therefore, the increase in CaO/SiO2 ratio in the flux leads to the flux depolymerisation, while increase in B2O3 content increases the bridging oxygen connected to Si. Figure 5-5 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3 fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) B2O3 contents. Table 5-3 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3 fluxes. Flux No. X Q 0 X Q 1 X Q 2 X Q 3 X Q 4 NBO/Si NBO/T

119 Figure 5-6 presents plots of viscosity lnη and activation energy EA vs NBO/Si with different CaO/SiO2 ratios and B2O3 contents. For fluxes with a fixed B2O3 content (6.6 wt%), increase in CaO/SiO2 ratio increased the NBO/Si. With the increase of NBO/Si, both viscosity η and activation energy EA decreased. Reduction of the energy barrier for viscous flow with increasing CaO/SiO2 ratio is related to depolymerisation of the flux. For fluxes with a fixed CaO/SiO2 ratio (1.3), however, measured NBO/Si decreased with the increase in B2O3 content from 4.7 to 8.4 wt%. With the increase of NBO/Si, both viscosity η and activation energy EA increased. A decrease in the flux viscosity with the addition of B2O3 was also reported by Fox et al. [2] in CaO-SiO2-Al2O3-B2O3- MgO-Na2O-Li2O fluxes and Shu et al. [21] in CaO-SiO2-Al2O3-B2O3-MgO-Na2O-TiO2 fluxes. On the contrary, an observation of increase in viscosity with increasing B2O3 concentration was reported by Tandon et al., [102] but occurred in a quite different system of Na2O-CaO-B2O3. The lack of SiO2 in this system could change the role of B2O3 as the only network former in the flux. A similar effect of B2O3 on activation energy was also observed in work [179]. Viscosity of molten flux is not only related to the degree of [56, 84] polymerisation of molten flux, but also concerned with composition of the flux. Because the bond energy B-O (787 kj/mol) [180] is apparently higher than those of Si-O (600 kj/mol) [165] and Al-O (485 kj/mol), [180] the introduction of B2O3 into the SiO2- CaO-Al2O3 system leads to the formation of B-O-Si or B-O-B flow units instead of Si- O-Si. The coordination of boron in respect of oxygen can be classified into four-fold coordinated boron (BO4 5- ) and three-fold coordinated boron (BO3 3- ). As BO3 3- is the major boron-containing unit in the molten flux, the planar triangular structure is looser than that of tetrahedral BO4 5-, and therefore the depolymerisation of the flux increases with the increase of B2O3. [173] However, the increase of NOB/Si with the increase of 91

120 B2O3 indicates the presence of a small quantity of BO4 5- tetrahedral units in the flux. The tetrahedral BO4 5- is preferentially connected to near Si atoms and introduced into the silicate networks because of the avoidance effect of BO4 5- group; therefore the uniformity of the networks could be lowered followed by a lower strength of the whole structures, which inhibits the formation of complex 3-dimentional structure. [177, 181, 182] It is thus not surprising that a decreasing in flux viscosity was observed with increasing B2O3 content, while the degree of polymerisation of the flux increased with the increase of B2O3 content. Figure 5-6 Relationship between lnη, EA and NBO/Si with different (a) CaO/SiO2 ratios, and (b) B2O3 contents. Effect of alumina on the flux structure can be taken into account of non-bridging oxygen per tetrahedral coordinated cation NBO/T which can be calculated from the following equation: [13] NBO/T = 2X CaO 2X Al 2O3 X SiO 2 +2X Al2O3 (5-3) where Xi is the mole fraction of component i. Values of calculated NBO/T are listed in Table 5-3. NBO/T increases with the increase in the CaO/SiO2 ratio from 0.8 to 1.5 at a fixed B2O3 content (6.6 wt%) and has little change with increasing B2O3 content from 92

121 4.7 to 8.4 wt% at a fixed CaO/SiO2 ratio (1.3). In general, the value of calculated NBO/T is a bit higher than that of measured NBO/Si. These two parameters show the similar trend in their change with the increase of CaO/SiO2 ratio. However, with the increase of B2O3 content, calculated NBO/T stayed almost same, while measured NBO/Si decreased, which indicates that not all B2O3 in the flux are BO3 3- trigonalrelated units, there could be some percentage of BO4 5- tetrahedral units. 5.6 Modelling of viscosity Viscosity of fluxes under examination was calculated using Riboud model, Koyama model, Urbain model, Iida model and NPL model. A comparison of the experimental data with the viscosity calculated using different models is shown in Figure 5-7. Only data at 1400 C were used for modelling. Agreement between calculated and experimental viscosities of fluxes was assessed using parameters σ and, which are defined by the following equations: σ = 1 N ((η N i=1 i) cal (η i ) mea ) 2 (5-4) = 100 N N (η i ) cal (η i ) mea (η i ) mea i=1 (5-5) where (η i ) cal and (η i ) mea are the calculated and measured viscosities for the flux i, and N is the number of flux samples. Parameters of σ and characterise the absolute and relative differences between calculated and measured viscosities, respectively. These values corresponding to different models were calculated and are listed in Table 5-4. As shown in Table 5-4, Iida model and Riboud model create the highest σ and, indicating a significant deviation of the calculation from the measurement. Although both Koyama model and Urbain model give the lowest σ and, both calculated data, 93

122 however, are highly scattering. It should be mentioned that Urbain model was used successfully for systems in which the content %SiO2+%Al2O3+%CaO < 50 wt%. [136] Table 5-4 Parameters σ and for different models in calculation of viscosity of the SiO2-CaO-Al2O3-B2O3 fluxes. Model Riboud model Urbain model Koyama model Iida model NPL model Modified NPL model σ, Pa s The NPL model described experimental flux viscosity well but was not tested in application to the B2O3 containing fluxes. [32] Calculation of viscosity of fluxes, examined in this work using the original NPL model, showed that the original NPL model is not applicable for B2O3-containing fluxes (Figure 5-7 and Table 5-4). To get good agreement between calculated and experimental data, the optical basicity of B2O3 was adjusted as it was done for CaF2 in work [32] by changing Λ B2 O 3 from 0.01 to 2.00 with the step size of 0.01 and obtaining an optimal Λ B2 O 3 for the best fit of viscosity. Two cases were considered. In the first case, B2O3 was assumed to play a similar role as Al2O3, and in this way x CaO in Eq.2-63 was replaced by x CaO = x CaO x Al2 O 3 x B2 O 3. The best fit of viscosity obtained for Λ B2 O 3 is 0.21 with σ = 0.19 Pa s and = 30.81, showing no significant improvement. This result suggests that above assumption that B2O3 plays the same role as Al2O3 could not be correct in this work. Therefore, the second case was considered where x CaO was replaced by normalised value x CaO = 94

123 x CaO x Al2 O 3. The best fit of viscosity with the deviation of 0.19 Pa s was obtained for Λ B2 O 3 = Such low value of optical basicity is unrealistic. Another approach to modifying NPL model was to modify the coefficient in the Eqs.2-61 and Based on the assumption that B2O3 formed BO3 3- structural units, and Al2O3 formed AlO4 5- units by the following processes: B 2 O 3 + 3CaO 2(BO 3 3 ) + 3Ca 2+ (5-6) Al 2 O 3 + 5CaO 2(AlO 4 5 ) + 5Ca 2+ (5-7) On the basis of the experimental data in the present study, the following equations were established to calculate parameters A and B in NPL model (denoted as A and B, in the modified model): ln ( B 1.05 ) = 4.65 (5-8) 1000 Λcorr lna = (Λ corr ) Λ corr (5-9) It can be seen in Table 5-4 that the predicted viscosity by the modified NPL model was in a reasonable agreement with experimental data with the lowest and the mean deviation of 0.07 Pa s. This modified NPL model was applied to the measured results of the same system using electrical vibrating method in Ar atmosphere by Akberdin et al. [183] at 1450 C (shown in Figure 5-7). For low viscosity slags (0.5 Pa s), this model fits well, but not very successful for high viscosity ones (0.7 Pa s and 1.0 Pa s). This difference could be caused by the measurement error which is hard to control in high temperature experiments. 95

124 Figure 5-7 Calculated viscosity using different models against experimental data. 5.7 Summary and conclusion The melting properties and viscosity of SiO2-CaO-Al2O3-B2O3 fluxes were investigated as functions of CaO/SiO2 mass ratio and the concentration of B2O3. The major findings are summarised as follows: (1) Increase in CaO/SiO2 from 0.8 to 1.5 increased Th slightly and then decreased Th, while addition of wt% B2O3 had no significant influence on Th. Tbr increased first and then decreased with the increase of CaO/SiO2, while increasing the content of B2O3 lowered Tbr. (2) Increase in CaO/SiO2 from 0.8 to 1.3 decreased viscosity, and a further increase in the CaO/SiO2 ratio from 1.3 to 1.5 had no further effect on viscosity at 1400 C; increase in CaO/SiO2 from 0.8 to 1.5 decreased EA in general. Increase in B2O3 concentration from 4.7 to 8.4 wt% in the flux with the CaO/SiO2 ratio 1.3 decreased both viscosity and activation energy. 96

125 (3) Viscosity was analysed in relation to the flux structure. The results revealed that there are five different silicate units of Q 0, Q 1, Q 2, Q 3, and Q 4 in the range of 800 and 1300 cm -1. Measured NBO/Si demonstrates the depolymerisation of the slag structure with increasing CaO/SiO2 ratio, while the addition of B2O3 increased the degree of polymerisation slightly. B2O3 apparently acted as a weak network former compared with SiO2. (4) Varied existing models were applied to describe the viscosity results, but none of them was successful. A modified NPL model was then developed which showed a reasonable agreement with the experimental data. 97

126 Chapter 6 - Melting properties and viscosity of SiO2-CaO-Al2O3-B2O3- MxOy (MxOy: Na2O, TiO2, or MgO) fluxes In this chapter, 5-component fluxes of SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) were investigated. Total 17 flux compositions were selected and are listed in Table 4-2. For SiO2-CaO-Al2O3-B2O3-Na2O fluxes, Fluxes 7, 8, 10 and 12 are with a fixed Na2O content (8 wt%) but varied CaO/SiO2 ratio from 0.8 to 1.5, and Fluxes 9, 10 and 11 are with a fixed CaO/SiO2 ratio (1.3) but varied Na2O contents from 6.2 to 9.1 wt%. For SiO2-CaO-Al2O3-B2O3-TiO2 fluxes, Fluxes 13, 14, 16 and 18 are with a fixed TiO2 content (4 wt%) but varied CaO/SiO2 ratio from 0.8 to 1.5, and Fluxes 15, 16 and 17 are with a fixed CaO/SiO2 ratio (1.3) but varied TiO2 contents from 2 to 5.9 wt%. For SiO2-CaO-Al2O3-B2O3-MgO fluxes, Fluxes 19, 20, 21 and 23 are with a fixed MgO content (2 wt%) but varied CaO/SiO2 ratio from 0.8 to 1.5, and Fluxes 21 and 22 are with a fixed CaO/SiO2 ratio but different MgO contents of 2.2 and 3.5 wt%, respectively. 6.1 Melting properties SiO2-CaO-Al2O3-B2O3-Na2O fluxes The effects of CaO/SiO2 ratio and Na2O content on Ts, Th, and Tf of the SiO2-CaO- Al2O3-B2O3-Na2O fluxes are illustrated in Figure 6-1. The value of Th decreased from 1211 C to 1094 C with the increase of CaO/SiO2 ratio from 0.8 to 1.0, but increased back to the same level (1211 C) when further increasing CaO/SiO2 ratio to 1.5 (Figure 6-1a). Change in Na2O concentration from 6.2 to 9.1 wt% with a fixed CaO/SiO2 ratio at 1.3 had a minor influence on Th, which fluctuated in the range of C (Figure 6-1b). 98

127 Two typical commercial fluorine-containing fluxes for low-carbon (Flux C1) and medium-carbon steel (Flux C2) were also analysed for comparison purpose. Their compositions, melting properties, viscosity and equilibrium phases calculated using FactSage are listed in Appendix I. The values of Th of industrial Fluxes C1 (1050 C) and C2 (1150 C) are also shown in Figure 6-1a. The melting temperatures of Fluxes 9-11 with CaO/SiO2 ratio of 1.3 were close to that for the industrial Flux C2 (Figure 6-1). However, the melting temperature of Flux C1 was lower than Th for synthetic Fluxes 7-12; Flux 8 with CaO/SiO2 ratio of 1.0 had the closest Th (Figure 6-1). Fluxes 7 and 12 with CaO/SiO2 ratios of 0.8 and 1.5 had much higher Th than commercial fluxes (Figure 6-1a). Figure 6-1 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes Figure 6-2 shows the effects of CaO/SiO2 ratio and TiO2 content on Ts, Th, and Tf of the SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. Th increased first from 1163 to 1238 C with the increase in CaO/SiO2 ratio from 0.8 to 1.0, then reduced to 1135 C with increasing CaO/SiO2 ratio to 1.3, but increased slightly with further increasing CaO/SiO2 ratio to 99

128 1.5. On the whole, the influence of CaO/SiO2 ratio on Th was complicated with ups and downs with changing CaO/SiO2 ratio (Figure 6-2a). Increase in the TiO2 content from 0 to 4 wt% decreased Th linearly from 1207 to 1135 C; however, Th enhanced to 1194 C significantly when TiO2 content increased from 4 to 6 wt% (Figure 6-2b). The melting temperature of Flux C1 (1050 C) was lower than Th for all fluorine-free TiO2-containing Fluxes 13-18; Th for most fluorine-free TiO2-containing fluxes is higher than that of Flux C2, except Fluxes 16 and 18 with CaO/SiO2 ratio of 1.3 and 1.5, respectively (Figure 6-2). Flux 14 with CaO/SiO2 ratio 1.0 and Flux 17 with TiO2 content 6 wt% both had much higher Th than commercial fluxes (Figure 6-2). Figure 6-2 Effects of (a) CaO/SiO2 ratio and (b) content of TiO2 on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes The effects of CaO/SiO2 ratio and MgO content on Ts, Th, and Tf of the SiO2-CaO- Al2O3-B2O3-MgO fluxes are shown in Figure 6-3. Th of SiO2-CaO-Al2O3-B2O3-MgO fluxes increased slightly from 1195 to 1203 C with CaO/SiO2 ratio from 0.8 to 1.0, then decreased to 1143 C when CaO/SiO2 ratio changed from 1.0 to 1.3, and stabilised 100

129 with further increasing to 1.5 (Figure 6-3a). Change in MgO concentration from 0 to 3.5 wt% reduced Th from 1207 to 1142 C first, but then stabilised it (Figure 6-3b). The hemispherical temperatures of Fluxes 21 and 22 with CaO/SiO2 ratio of 1.3 and Flux 23 with CaO/SiO2 ratio of 1.5 were close to that of the industrial Flux C2 (Figure 6-3); Flux 20 with CaO/SiO2 ratio of 1.0 had much high Th than Flux C2 (Figure 6-1a). The melting temperature of Flux C1 was lower than Th for all synthetic Fluxes (Figure 6-3). Figure 6-3 Effects of (a) CaO/SiO2 ratio and (b) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 6.2 Viscosity, break temperature and activation energy SiO2-CaO-Al2O3-B2O3-Na2O fluxes Viscosity of the SiO2-CaO-Al2O3-B2O3-Na2O fluxes versus temperature for different CaO/SiO2 ratios and Na2O concentrations are plotted in Figure 6-4. The viscosity curves of Fluxes C1 and C2 are also shown in Figure 6-4a. Flux 12 with CaO/SiO2 ratio of 1.5 was not fully liquidised at 1400 C; the η-t curve for this flux was not available. 101

130 Viscosity increased with decreasing temperature. Viscosity of Fluxes 7 and 8 with low CaO/SiO2 ratios of 0.8 and 1.0 sharply increased as a result of the flux solidification (Figure 6-4a). Viscosity-temperature curves for Fluxes 9-11 with CaO/SiO2 ratio of 1.3 exhibited two apparently different stages in the process of solidification (Figure 6-4): the abrupt increase in viscosity in a narrow temperature interval followed by a slower but continuous increase in a rather broad temperature range. The viscosity curve for the fluorine-containing Flux C2 with high CaO/SiO2 ratio of 1.4 had a similar shape as fluorine-free Fluxes 7 and 8 with low CaO/SiO2 ratios of 0.8 and 1. The increase in viscosity of Flux C1 with CaO/SiO2 ratio of 0.9 was not as steep as for C2; the viscosity change in the process of solidification resembled the second stage of the viscosity change for Fluxes 9-11 with CaO/SiO2 ratio of 1.3. Clearly, the effect of CaO/SiO2 ratio on viscosity of fluorine-free and fluorine-containing fluxes was quite different. Values of flux viscosities at 1400 C and 1300 C are shown in Table 6-1. The viscosity of liquid fluxes was low; it was in the range Pa s at 1400 C. Viscosities of fluorine-free fluxes at 1400 C decreased gradually with the increase of CaO/SiO2 ratio from 0.8 to 1.3 (Fluxes 7, 8 and 10). Change in the Na2O concentration from 6.2 to 9.1 wt% at the CaO/SiO2 ratio of 1.3 (Fluxes 9-11) had a minor effect on viscosity. At the lower temperature of 1300 C, viscosity followed the same trend of the decrease with increasing CaO/SiO2 ratio (Fluxes 7, 8 and 10). However, the effect of Na2O concentration on the viscosity with the fixed CaO/SiO2 ratio at 1.3 became significant when Na2O reached 9.1 wt%. The high viscosity at this Na2O content could be because the flux started to solidify at this temperature. 102

131 Values of viscosity of fluorine-free Fluxes 7-11 at 1300 C were higher than those of two commercial fluorine-containing fluxes. Viscosity of Flux C1 with CaO/SiO2 ratio of 0.9 at 1300 C was much higher than viscosity of Flux C2 with CaO/SiO2 ratio of 1.4. Figure 6-4 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. The break temperatures Tbr and calculated values of EA derived from the viscosity curves (Figure 6-4) are also shown in Table 6-1. Tbr reduced slightly from 1195 C to 1175 C with CaO/SiO2 ratio changing from 0.8 to 1.0, but rose to 1273 C with a further increase of CaO/SiO2 ratio to 1.3. An increase in the Na2O content from 6.2 to 9.1 wt% with a fixed CaO/SiO2 ratio at 1.3 increased Tbr from 1265 C to 1360 C. Increase in CaO/SiO2 ratio and Na2O content of fluorine-free fluxes decreased EA slightly. 103

132 Table 6-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol C C SiO2-CaO-Al2O3-B2O3-TiO2 fluxes Viscosity of the SiO2-CaO-Al2O3-B2O3-TiO2 fluxes versus temperature for different CaO/SiO2 ratios and TiO2 concentrations are shown in Figure 6-5. Viscosities of fluxes at 1400 C and 1300 C are listed in Table 6-2. The viscosity of liquid fluxes was low; it was in the range Pa s at 1400 C. Viscosities of TiO2-containing fluxes at both temperatures decreased gradually with the increase of CaO/SiO2 ratio from 0.8 to 1.5 (Fluxes 13, 14, 16 and 18). The reason for that the viscosity at 1300 C of Flux 14 was much higher than others is that Tbr of this flux is higher than 1300 C. Change in the TiO2 concentration from 2 to 6 wt% at the CaO/SiO2 ratio of 1.3 (Fluxes 15-17) decreased viscosity at 1400 C and 1300 C. Similar to the situation of Na2O-containing fluxes, values of viscosity of two commercial fluorine-containing fluxes at 1300 C were lower than those of Fluxes

133 Figure 6-5 Effects of (a) CaO/SiO2 ratio and (b) content of TiO2 on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. The break temperatures Tbr and calculated values of EA derived from the viscosity curves of TiO2-containing fluxes (Figure 6-5) are also shown in Table 6-2. Tbr increased with CaO/SiO2 ratio from 0.8 to 1.0 and then declined with the increase in the CaO/SiO2 ratio to 1.5. Change of the TiO2 content from 4 to 6 wt% had no influence on Tbr. EA had the same trend as η with the increase of CaO/SiO2 ratio and TiO2 content. Table 6-2 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-TiO2 fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol

134 6.2.3 SiO2-CaO-Al2O3-B2O3-MgO fluxes Figure 6-6 shows the viscosity of the SiO2-CaO-Al2O3-B2O3-MgO fluxes versus temperature for different CaO/SiO2 ratios and MgO concentrations. Values of viscosity at 1400 C and 1300 C, break temperatures Tbr, and calculated values of EA derived from the viscosity curves of this system (Figure 6-6) are shown in Table 6-3. Viscosity was in the range Pa s at 1400 C. Viscosity η decreased with CaO/SiO2 ratio increase from 0.8 to 1.3 (Fluxes 19, 20, 21 and 23). Tbr increased with the increase of CaO/SiO2 ratio from 0.8 to 1.0, remained no change with increasing CaO/SiO2 ratio from 1.0 to 1.3, but then decreased with further increase of CaO/SiO2 ratio up to 1.5. η decreased from 0.31 to 0.24 Pa s and Tbr declined from 1330 to 1223 C with the addition of MgO from 2.2 to 3.5 wt%. The tendency of activation energy EA is the same as that of viscosity η. Figure 6-6 Effects of (a) CaO/SiO2 ratio and (b) content of MgO on the viscositytemperature curves of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 106

135 Table 6-3 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-MgO fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol Raman spectroscopy SiO2-CaO-Al2O3-B2O3-Na2O fluxes Raman spectra in the frequency range of 500 to 1600 cm -1 are shown in Figure 6-7. Similar to SiO2-CaO-Al2O3-B2O3 fluxes, a distinct hump in the frequency region between 550 and 750 cm -1 (marked by dashed lines in Figure 6-7) and a large hump with several peaks between 750 and 1100 cm -1 were observed in all spectra. The increase of CaO/SiO2 ratio from 0.8 to 1.3 shifted the position of the low frequency hump from 638 to 676 cm -1 (Figure 6-7a). A small envelope between 1100 and 1300 cm -1 (marked by dashed lines in Figure 6-7) was also found for all spectra apart from Flux 7 with CaO/SiO2 ratio 0.8 and Na2O concentration 8 wt%. 107

136 Figure 6-7 Raman spectra of the quenched fluxes with different (a) CaO/SiO2 ratios, and (b) Na2O contents in the frequency range cm -1 of SiO2-CaO-Al2O3-B2O3- Na2O fluxes SiO2-CaO-Al2O3-B2O3-TiO2 fluxes Raman spectra in the frequency range of 500 to 1600 cm -1 of SiO2-CaO-Al2O3-B2O3- TiO2 fluxes are shown in Figure 6-8. Flux 14 with CaO/SiO2 ratio 1.0 was not listed in the figure, because there was crystal structure in the quenched flux sample. Different from SiO2-CaO-Al2O3-B2O3 and SiO2-CaO-Al2O3-B2O3-Na2O fluxes, the envelope starting from 600 cm -1 merged with peaks of silicate network from 800 to 1100 cm -1 with no gap. Two small envelopes around 1186 cm -1 and 1440 cm -1 (marked by dashed lines in Figure 6-8) were also found for all spectra apart from Flux 13 with CaO/SiO2 ratio 0.8 and TiO2 concentration 4 wt%. 108

137 Figure 6-8 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3-TiO2 fluxes with different (a) CaO/SiO2 ratios, and (b) TiO2 contents in the frequency range cm SiO2-CaO-Al2O3-B2O3-MgO fluxes Figure 6-9 displays the Raman spectra in the frequency range of 500 to 1600 cm -1 of SiO2-CaO-Al2O3-B2O3-MgO fluxes, the shape of which is similar to SiO2-CaO-Al2O3- B2O3-Na2O fluxes. The increase of CaO/SiO2 ratio from 0.8 to 1.5 shifted the position of the low frequency hump from 633 to 672 cm -1 (Figure 6-9a), while increase in MgO content from 2.2 to 3.5 wt% has no influence on this peak. A small envelope between 1100 and 1300 cm -1 (marked by dashed lines in Figure 6-9) was also found for all spectra apart from Flux 7 with CaO/SiO2 ratio of 0.8. Besides, a small envelope around 1430 cm -1 was found for all spectra. 109

138 Figure 6-9 Raman spectra of the quenched SiO2-CaO-Al2O3-B2O3-MgO fluxes with different (a) CaO/SiO2 ratios, and (b) MgO contents in the frequency range cm Relationship between characteristic temperatures and phase composition SiO2-CaO-Al2O3-B2O3-Na2O fluxes Figure 6-10 displays the equilibrium phases of the SiO2-CaO-Al2O3-B2O3-Na2O fluxes calculated using FactSage. Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq derived from this figure are listed in Table 6-4. The value of Tliq slightly decreased with increasing CaO/SiO2 ratio from 0.8 to 1.3, and remained almost unchanged when the CaO/SiO2 ratio increased further to 1.5 (Table 6-4). In general, Ts, Th and Tf should be within the range of Tsol and Tliq. Experimental value Tf is higher than the calculated Tliq for Flux 12 with the CaO/SiO2 ratio of 1.5, which can be attributed to inaccuracy in calculations of equilibrium phases in this flux or something else which is still not clear. Possible evaporation of B2O3 and Na2O could lead to the change of flux composition which would also contribute to this deviation. 110

139 Phase equilibrium calculation revealed that CaSiO3 and combeite were the main phases at low CaO/SiO2 ratio of 0.8 (Figure 6-10a). With the increase of CaO/SiO2 ratio from 0.8 to 1.0, minor Ca3Si2O7 and Ca11B2Si4O22 phases replaced part of CaSiO3 and combeite (Figure 6-10b); with CaO/SiO2 ratio increasing from 1.0 to 1.3, Ca3Si2O7 and Ca11B2Si4O22 became dominant high-temperature phases (Figure 6-10d). With further increase of CaO/SiO2 ratio from 1.3 to 1.5, a small amount of Ca3Si2O7 was replaced by Ca2SiO4 and more Ca11B2Si4O22 formed (Figure 6-10f). There was no significant change of dominant phases close to Tliq for Fluxes 7 and 8 (CaO/SiO2 ratio of 0.8 and 1.0), so these two fluxes had close Tbr. Change in the Na2O concentration from 6.2 to 9.1 wt% decreased Tliq from 1262 C to 1227 C (Table 6-4). Change in the Na2O concentration did not change the high temperature phases but altered their fractions (Figures 6-10c,d,e). As a result, there was no significant change in Th, which is consistent with the change of Tliq. The values of Tbr for Fluxes 9-11 were higher than those of their Tliq, which could be caused by possible formation of high temperature phases nucleated under non-equilibrium state during the cooling process of viscosity measurement. There is no direct correlation between the measured Ts, Th, and Tf of the SiO2-CaO- Al2O3-B2O3-Na2O fluxes and the mass fraction of their equilibrium liquid phases. It follows from the phase diagrams in Figure 6-10 that the fraction of the liquid phase at the melting temperature Th is in the range from 0.6 for Flux 2 to 0.76 for Flux 7; the fraction of liquid at the fluidity temperature reached 0.74 for Flux 9 and 0.96 for Flux 7. Obviously, fluxes are not in the equilibrium state in the process of measurement of their melting properties. Melting properties and break temperature are broadly used in industry. These parameters can be determined experimentally, while Tliq is indicative for 111

140 evaluation of the melting properties and initial design of fluxes for the steel continuous casting. The parameters of two fluorine-containing Fluxes C1 and C2, examined in this work are within the ranges of reported commercial mould fluxes (Table 2-4). Fluxes used in the continuous casting of steels of different grades have significant differences. For low carbon steel, Th is in the range of 960 C to 1140 C, [6, 26, 27] η is in the range of 0.04 to 0.30 Pa s, [12, 28, 29] and Tbr changes from 960 C to 1133 C. [6, 24, 27] Fluxes used for casting of medium-carbon steel have higher Th ( C) [6, 26, 30, 31] and Tbr ( C), [6, 24, 32, 33] and η in the range of 0.05 to 0.39 Pa s. [12, 28, 29, 32-34] Th of fluorinefree fluxes with CaO/SiO2 ratio 1.0 and 1.3 studied in this work is within the ranges of these properties of industrial fluxes (Table 2-4). Viscosity η of fluorine-free fluxes is higher than the range of industrial fluxes, except that η of Fluxes 9 and 10 is within the range but close to upper bound of industrial fluxes for medium-carbon steel. Break temperature Tbr of fluorine-free fluxes with CaO/SiO2 ratio 0.8 and 1.0 is within the range for medium-carbon steel; and Tbr of fluxes with CaO/SiO2 ratio 1.3 is above this range. 112

141 Figure 6-10 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-Na2O fluxes using FactSage: (a) 0.8 CaO/SiO2, 8 wt% Na2O; (b) 1.0 CaO/SiO2, 8 wt% Na2O; (c) 1.3 CaO/SiO2, 6 wt% Na2O; (d) 1.3 CaO/SiO2, 8 wt% Na2O; (e) 1.3 CaO/SiO2, 9 wt% Na2O; and (f) 1.5 CaO/SiO2, 8 wt% Na2O. 113

142 Table 6-4 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-Na2O fluxes. Flux No. Tsol, C Tliq, C Dominant high temperature phases CaSiO3, combeite CaSiO3, combeite Ca3Si2O7, Ca11B2Si4O22, CaSiO3, combeite Ca3Si2O7, Ca11B2Si4O22, combeite Ca3Si2O7, Ca11B2Si4O22, combeite Ca3Si2O7, Ca11B2Si4O22, combeite The results shown in Figure 6-4 reveal that the fluorine-free mould fluxes with low CaO/SiO2 ratios of 0.8 and 1.0 behave similarly to the commercial Flux C2 for mediumcarbon steel, with a similar shape of η-t curves and close Tbr. The viscosities of Fluxes 9-11 with a higher CaO/SiO2 ratio of 1.3 exhibit a two-stage increase with the decrease in temperature. Similar shape of η-t curves was also observed for CaO-SiO2-Al2O3- Na2O-B2O3-Li2O mould fluxes with CaO/SiO2 ratio of 1.25 and 1.15 in work [53], but no explanation was given. Two-stage viscosity curves in the solidification of mould fluxes with a relatively high CaO/SiO2 ratio can be related to the transition to a higher order phase region or appearance of one sticky solid which however, lacks evidence and needs further examination SiO2-CaO-Al2O3-B2O3-TiO2 fluxes The equilibrium phases of the SiO2-CaO-Al2O3-B2O3-TiO2 fluxes calculated using thermochemical software FactSage are shown in Figure Calculated Tliq, Tsol and 114

143 dominant phases in the temperature range Tsol-Tliq derived from this figure are listed in Table 6-5. Similar to Na2O-containing fluxes, measured melting properties and break temperature are within the range of calculated Tsol and Tliq. Tliq increased slightly with increasing CaO/SiO2 ratio from 0.8 to 1.0 (Fluxes 13 and 14), and decreased when the CaO/SiO2 ratio increased further to 1.5 (Fluxes 14, 16 and 18). The increase in Tliq with CaO/SiO2 ratio from 0.8 to 1.0 could contribute to the increase of both Th and Tbr. The increase in TiO2 content from 2 to 6 wt% decreased Tliq, but increased both Th and Tbr, showing different a trend of measured Th and Tbr with the calculated equilibrium Tliq. It should be noticed that the CaO-SiO2-Al2O3-B2O3-TiO2 fluxes contain a higher amount of wollastonite (CaSiO3) than the CaO-SiO2-Al2O3-B2O3-Na2O fluxes. The addition of TiO2 to the CaO-SiO2-Al2O3-B2O3 system restrains the formation of Ca11B2Si4O22 to some extent. The main phases at low CaO/SiO2 ratio (0.8) are CaSiO3 and CaSiTiO5 (Figure 6-11a). With the increase of CaO/SiO2 ratio from 0.8 to 1.0, more fractions of melilite (Ca2B(B,Al)SiO7) form, along with CaSiO3 and CaSiTiO5 as high-temperature phases (Figure 6-11b). With CaO/SiO2 ratio rising from 1.0 to 1.3, Tsol increased greatly from 864 to 1096 C, and CaTiO3 (Tm=1975 C) replaced CaSiTiO5 (Tm=1221 C) as the main Ti-containing phase (Figure 6-11d). With further increasing CaO/SiO2 ratio from 1.3 to 1.5, partial CaSiO3 was substituted by Ca3Si2O7 (Figure 6-11f). Increase in TiO2 content from 2 to 6 wt% had an inconspicuous influence on CaSiO3, while it facilitated the formation of CaTiO3 (Figures 6-11b,d,e). 115

144 Figure 6-11 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes using FactSage: (a) 0.8 CaO/SiO2, 4 wt% TiO2; (b) 1.0 CaO/SiO2, 4 wt% TiO2; (c) 1.3 CaO/SiO2, 2 wt% TiO2; (d) 1.3 CaO/SiO2, 4 wt% TiO2; (e) 1.3 CaO/SiO2, 6 wt% TiO2; and (f) 1.5 CaO/SiO2, 4 wt% TiO2. 116

145 Table 6-5 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-TiO2 fluxes. Flux No. Tsol, C Tliq, C Dominant high-temperature phases CaSiO3, CaSiTiO5, melilite, SiO CaSiO3, CaSiTiO5, melilite CaSiO3, Ca3Si2O7, CaTiO3, melilite CaSiO3, CaTiO3, melilite CaSiO3, CaTiO3, melilite CaSiO3, Ca3Si2O7, CaTiO3, Ca11B2Si4O22, melilite SiO2-CaO-Al2O3-B2O3-MgO fluxes The equilibrium phases of the SiO2-CaO-Al2O3-B2O3-MgO fluxes calculated using thermochemical software FactSage are shown in Figure Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq derived from this figure are listed in Table 6-6. Tliq increased slightly with increasing CaO/SiO2 ratio from 0.8 to 1.0 (Fluxes 19 and 20), and decreased when the CaO/SiO2 ratio increased further from 1.0 to 1.5 (Fluxes 20, 21 and 23). Increase in MgO content from 2 to 3.5 wt% (Fluxes 21 and 22) decreased Tliq slightly. As shown in Figure 6-12, at CaO/SiO2 ratio of 0.8, the main phase is wollastonite (major CaSiO3 + minor MgSiO3) with minor melilite and CaMgSi2O6 at temperature close to Tsol (Figure 6-12a). With the increase of CaO/SiO2 ratio from 0.8 to 1.0, the percentage of melilite ((Ca2B(B,Al)SiO7) and Ca2MgSi2O7) and CaMgSi2O6 increased 117

146 (Figure 6-12b). Similar to TiO2-containing fluxes, the increase of both Th and Tbr could be influenced by the increase in Tliq with CaO/SiO2 ratio from 0.8 to 1.0. With CaO/SiO2 ratio increasing from 1.0 to 1.3, Ca3Si2O7 and Ca11B2Si4O22 formed, and Tsol increased from 771 to 1070 C abruptly (Figure 6-12c). With further increase of CaO/SiO2 ratio from 1.3 to 1.5, Ca3Si2O7 became the dominant high-temperature phase instead of wollastonite, and more Ca11B2Si4O22 formed (Figure 6-12e). Addition of MgO content from 2 to 3.5 wt% accelerated the formation of melilite, with no influence on the types of high-temperature phases. 118

147 Figure 6-12 Calculated equilibrium phase fractions of SiO2-CaO-Al2O3-B2O3-MgO fluxes using FactSage: (a) 0.8 CaO/SiO2, 2wt% MgO; (b) 1.0 CaO/SiO2, 2 wt% MgO; (c) 1.3 CaO/SiO2, 2 wt% MgO; (d) 1.3 CaO/SiO2, 3.5 wt% MgO; and (e) 1.5 CaO/SiO2, 2 wt% MgO. 119

148 Table 6-6 Parameters derived from calculated equilibrium phase fractions of SiO2-CaO- Al2O3-B2O3-MgO fluxes. Flux No. Tsol, C Tliq, C Dominant high-temperature phases CaSiO3, melilite, CaMgSi2O CaSiO3, melilite, CaMgSi2O6, CaAl2Si2O CaSiO3, Ca2Si2O7, melilite CaSiO3, Ca2Si2O7, melilite, Ca11B2Si4O Ca2Si2O7, CaSiO3, melilite, Ca11B2Si4O Viscosity and structure of molten flux SiO2-CaO-Al2O3-B2O3-Na2O fluxes The effects of CaO/SiO2 ratio and Na2O content on lnη at 1400 C and EA of the SiO2- CaO-Al2O3-B2O3-Na2O fluxes are shown in Figure As a result of the flux depolymerisation, viscosity and activation energy decreased with the increase in CaO/SiO2 ratio, as shown in Figure 6-13a. However, change in the viscosity of fluxes with the increasing Na2O concentration at 1400 C was non-monotonic. As shown in Figure 6-13b, increase in Na2O content decreased EA slightly. The value of viscosity increased first with the increase in Na2O content from 6.2 to 7.9 wt% but then decreased with further increasing Na2O content to 9.1 wt%. A slight increase in viscosity can be related to the varied extent of evaporation of Na2O which results in the scattering of experimental data. The trend of decreasing viscosity with increasing Na2O content observed in this work was also reported by Wen et al.. [28] 120

149 Figure 6-13 Effects of (a) CaO/SiO2 ratio and (b) content of Na2O on lnη and EA of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. Raman spectroscopy results shown in Figure 6-7 revealed that similar to SiO2-CaO- Al2O3-B2O3 fluxes, with the increase of CaO/SiO2 ratio from 0.8 to 1.3, the band in low frequency range (Si-O-Si bending modes) is shifted to higher frequency. The increase in the Na2O concentration from 6.2 to 9.1 wt% had no apparent effect on the shift of this band. This phenomenon was also observed in the CaO-SiO2-B2O3 system, [176] SiO2- Na2O and CaO-SiO2 systems. [171] Humps in the range of 750 to 1300 cm -1 were deconvoluted into Q 0 (SiO4 4- ), Q 1 (Si2O7 6- ), Q 2 (SiO3 2- ), Q 3 (Si2O5 2- ), and Q 4 (SiO2) (shown in Figure 6-14). The positions of the peaks are around 860 cm -1, 910 cm -1, 950 cm -1, 1040 cm -1 and 1190 cm -1, respectively. [24, 82, 84, ] The molar fractions of various silicate structural units Q i calculated using Eq.5-1 are listed in Table 6-7. For all fluxes, Q 2 was the dominant unit, while the fraction of Q 4 was quite low. No Q 4 unit was distinguishable in the spectra for Flux 7 (CaO/SiO2 ratio = 0.8). Fractions of Q 0 and Q 1 increased, while that of Q 3 decreased with the increase of CaO/SiO2 ratio from 0.8 to 1.3. The increase of the CaO/SiO2 ratio promoted reforming of complex network units and lowered the viscosity of the flux system (Table 6-1). 121

150 The fraction of monomers Q 0 also increased with increasing Na2O concentration in the flux with the fixed CaO/SiO2 ratio of 1.3; however, a change in the fraction of dimers Q 1 with Na2O concentration was inconsistent. Based on Eq.5-2, the measured values of NBO/Si are also listed in Table 6-7. NBO/Si increased with the increase in the CaO/SiO2 ratio from 0.8 to 1.3 at a fixed Na2O content (8 wt%) and with increasing Na2O content from 6 to 9 wt% at a fixed CaO/SiO2 ratio. Therefore, the increase in both CaO/SiO2 ratio and Na2O content in the flux leads to flux depolymerisation. Figure 6-15 presents plots of viscosity lnη and activation energy EA vs NBO/Si; both viscosity η and activation energy EA decreased with increasing NBO/Si. Figure 6-14 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-Na2O fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) Na2O contents. 122

151 Table 6-7 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-Na2O fluxes. Flux No. X Q 0 X Q 1 X Q 2 X Q 3 X Q 4 NBO/Si NBO/T Raman spectra obtained in this work for quenched fluxes did not reflect the formation of AlO4 5- units. Formation of these units involves charge compensation by Na + and Ca 2+ cations. Priority for cations to compensate the charge of the Al 3+ ions, is defined by the lowest coulombic force (I) between the cation and oxygen ion: [92] I = Z M Z O (r M z++r O 2 ) 2 (6-1) where ZO is the charge of oxygen anion, ZM is the charge of M cation, r M z+ and r O 2 are the radii of M z+ and oxygen ions, respectively. Charge-compensation by divalent Ca 2+ is achieved through two O - bonds in two AlO4 5- units, while univalent Na + is bonded with one O - in one AlO4 5- unit. [13] Thus, Na + has a higher priority for charge compensation than Ca 2+, which causes the changes with concentrations of Na2O and CaO for depolymerisation. Only when the cation with the highest priority is exhausted, will a cation with a lower priority be used to charge compensate the Al

152 Figure 6-15 Relationship between lnη, EA and NBO/Si. NBO/T of this flux system can be calculated from the following equation: [13] NBO/T = 2X CaO+2X Na 2O 2X Al 2O3 X SiO 2 +2X Al2O3 (6-2) Values of calculated NBO/T are listed in Table 6-7. NBO/T increases with the increase in the CaO/SiO2 ratio from 0.8 to 1.3 at a fixed Na2O content (8 wt%) and with increasing Na2O content from 6 to 9 wt% at a fixed CaO/SiO2 ratio (1.3). In general, the value of calculated NBO/T is close to that of measured NBO/Si, and these two parameters show the same trend in their change with flux composition SiO2-CaO-Al2O3-B2O3-TiO2 fluxes According to Raman spectroscopy results in Figure 6-8, the band in the range of 600 to 1300 cm -1 was fitted by eight Gaussian peaks with R 2 > 99.9% using Gaussian deconvolution method (shown in Figure 6-16). The peak located around 660 cm -1 was assigned as Si-O-Si bending modes. [84] The peak around 705 cm -1 was assigned to the 124

153 deformation of O-Ti-O or O-(Si, Ti)-O in chain or sheet units or both. [175, 187, 188] The peak around 805 cm -1 reflected the stretching vibration of Ti-O in TiO4 4- monomers. [187, 188] With the increase of CaO/SiO2 ratio from 0.8 to 1.5, the intensity of the peak around 805 cm -1 reduced, while the increase in TiO2 content from 2 to 6 wt% enhanced the formation of this peak (Figure 6-16). This change in the Raman spectra was consistent with the SiO2-Na2O-TiO2, SiO2-CaO-TiO2 and SiO2-CaO-Al2O3-TiO2 fluxes in Mysen s work. [188] In this way, the addition of TiO2 into SiO2-CaO-Al2O3-B2O3 fluxes caused Ti 4+ occupying the position of Si 4+ as the network former. Ti-O-Ti and Ti-O-Si bonds are weaker than Si-O-Si bonds because the Ti 4+ cation is larger than that of Si 4+. [188] Besides, Ti may form some simpler or unstable structural units. [189] Therefore, viscosity η declined with the increase in TiO2 concentration from 2 to 6 wt%. The positions of the peaks around 860 cm -1, 910 cm -1, 960 cm -1, 1050 cm -1 and 1190 cm -1 reflected Q 0, Q 1, Q 2, Q 3 and Q 4, respectively. The molar fractions of various silicate structural units, NBO/Si and NBO/T are shown in Table 6-8. NBO/T of this flux system can be calculated from the following equation: [13] NBO/T = 2X CaO 2X Al 2O3 X SiO 2 +2X Al2O3 +X TiO2 (6-3) For all the fluxes, Q 2 was dominant, and Q 4 was the lowest with no Q 4 observed for Flux 13 (CaO/SiO2 ratio of 0.8). With the increase of CaO/SiO2 ratio from 0.8 to 1.5, fraction of Q 0 increased, while that of Q 3 decreased. The change of molar fractions of Q 1 and Q 2 was not monotone with the increase of CaO/SiO2 ratio. The molar fractions of Q 0 and Q 1 increased, but Q 2 and Q 3 decreased with the increase of TiO2 concentration from 2 to 6 wt%. Similar decreasing effect of TiO2 content on Q 2 and Q 3 was observed on CaO-SiO2-B2O3 -TiO2 fluxes. [175] 125

154 NBO/Si increased with the increase in the CaO/SiO2 ratio from 0.8 to 1.5 at a fixed TiO2 content (4 wt%) and with increasing TiO2 content from 2 to 6 wt% at a fixed CaO/SiO2 ratio (1.3). The relationship among lnη, activation energy EA and NBO/Si are shown Figure In general, both viscosity η and activation energy EA decreased with increasing NBO/Si. Figure 6-16 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) TiO2 contents. NBO/T increased with the increase in the CaO/SiO2 ratio from 0.8 to 1.5 at a fixed TiO2 content (Fluxes 13, 16 and 18) and with increasing TiO2 content from 2 to 6 wt% at a fixed CaO/SiO2 ratio (Fluxes 15-17). In general, the value of calculated NBO/T is a bit higher than that of measured NBO/Si, except that for Flux 13 whose NBO/T is a little less than NBO/Si. 126

155 Table 6-8 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes. Flux No. X Q 0 X Q 1 X Q 2 X Q 3 X Q 4 NBO/Si NBO/T Figure 6-17 Relationship between lnη, EA and NBO/Si of SiO2-CaO-Al2O3-B2O3-TiO2 fluxes SiO2-CaO-Al2O3-B2O3-MgO fluxes Using the same deconvolution method as for Na2O-containing fluxes, the Raman spectrum band in the range of 600 to 1300 cm -1 of SiO2-CaO-Al2O3-B2O3-MgO fluxes 127

156 was de-convoluted, as shown in Figure The positions of the peaks around 860 cm - 1, 910 cm -1, 960 cm -1, 1050 cm -1 and 1200 cm -1 are assigned to Q 0, Q 1, Q 2, Q 3 and Q 4, respectively. The molar fractions of various silicate structural units, NBO/Si and NBO/T are listed in Table 6-9. For Flux 19 (CaO/SiO2 ratio = 0.8), both fractions of Q 2 and Q 3 were remarkable, while the fraction of Q 2 of other fluxes was the only dominant silicate structural unit. Besides, the peak of Q 4 was not observed from the spectra of Flux 19 (Figure 6-18a). There was no Q 1 identified for Fluxes 19 and 20 with CaO/SiO2 ratio of 0.8 and 1.0, respectively. Similar to CaO-SiO2-Al2O3-B2O3 fluxes, with the increase of CaO/SiO2 ratio from 0.8 to 1.5, fractions of Q 0 and Q 1 increased, while that of Q 3 decreased. The fraction of Q 2 increased with increasing CaO/SiO2 ratio from 0.8 and 1.0 (Fluxes 19 and 20), and decreased with further increasing CaO/SiO2 ratio from 1.0 and 1.5 (Fluxes 20, 21 and 23). With increasing MgO concentration from 2 to 3.5 wt% in the flux (Fluxes 21 and 22), fraction of Q 1 enhanced, but fractions of Q 2 and Q 3 reduced. NBO/Si increased with the increase of either CaO/SiO2 ratio ( ) or MgO concentration (2-3.5 wt%). Figure 6-19 displays plots of lnη and EA as a function of NBO/Si, similar to Na2Ocontaining and TiO2-containig fluxes of which both viscosity η and activation energy EA decreased with increasing NBO/Si. As mentioned before, when there are several basic oxides in a melt containing Al2O3, there is a strict order for which cations carry out the charge-compensation of the Al 3+ with the highest priority being given to the cations with the lowest field strength. Priority for charge-compensation of this flux system is in the order: Ca 2+ >Mg 2+. [92] The values of calculated NBO/T are higher than those of NBO/Si. Both NBO/T and NBO/Si follow the same trend with the change of CaO/SiO2 ratio or MgO contents. 128

157 Figure 6-18 Deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-MgO fluxes in the frequency of cm -1 with different (a) CaO/SiO2 ratios, and (b) MgO contents. Table 6-9 Area fractions of peaks obtained from Gaussian deconvolution of Raman spectra of SiO2-CaO-Al2O3-B2O3-MgO fluxes. Flux No. X Q 0 X Q 1 X Q 2 X Q 3 X Q 4 NBO/Si NBO/T

158 Figure 6-19 Relationship between lnη, EA and NBO/Si of SiO2-CaO-Al2O3-B2O3-MgO fluxes. 6.6 Summary and conclusion The melting properties and viscosity of SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: Na2O, TiO2, or MgO) fluxes were investigated as functions of the ratio of CaO/SiO2 ( ) and the concentration of Na2O (6-9 wt%), TiO2 (2-6 wt%), and MgO (2-3.5 wt%). The results showed quite complex effects of CaO/SiO2 ratio and individual components on these properties. The major results are summarised as follows: (1) For Na2O-containing fluxes, change in the CaO/SiO2 ratio from 0.8 to 1.0 decreased melting temperature Th and Tbr, while further increase in this ratio increased Th and Tbr. Na2O content from 6.2 to 9.1 wt% had a minor influence on Th, but raised Tbr. For TiO2-containing fluxes, the influence of CaO/SiO2 ratio on Th was complicated with ups and downs with changing CaO/SiO2 ratio. Tbr increased 130

159 with CaO/SiO2 ratio from 0.8 to 1.0 and then declined with the increase in the CaO/SiO2 ratio to 1.5. Increase in the TiO2 content from 0 to 4 wt% decreased Th, but enhanced it significantly when TiO2 content increased from 4 to 6 wt%. Change of the TiO2 content from 4 to 6 wt% had no influence on Tbr. For MgO-containing fluxes, Th increased slightly with CaO/SiO2 ratio from 0.8 to 1.0, then decreased when CaO/SiO2 ratio changed from 1.0 to 1.3, and stabilised with further increasing to 1.5. Tbr increased with CaO/SiO2 ratio from 0.8 to 1.0, stayed the same with CaO/SiO2 ratio increasing from 1.0 to 1.3, and declined with further increase of CaO/SiO2 ratio up to 1.5. Change in MgO concentration from 0 to 3.5 wt% reduced Th first but then stabilised it. Tbr decreased with MgO concentration from 2 to 3.5 wt%. (2) For all the flux systems, an increase in the CaO/SiO2 ratio led to the decrease in viscosity η and activation energy EA. Change in Na2O concentration from 6 to 9 wt% had no significant influence on η, but reduced EA. With the increase of TiO2 and MgO content, both η and EA decreased. (3) Raman spectroscopy was used for flux structural analyses for all quenched fluxes. Similar to 4-component fluxes, the results revealed that there are five different silicate units of Q 0, Q 1, Q 2, Q 3, and Q 4 in the range of 750 and 1300 cm -1. Measured NBO/Si demonstrated the depolymerisation of the slag structure with increasing CaO/SiO2 ratio; in general, it had the same tendency as calculated NBO/T. Change in the viscosity and activation energy was correlated with NBO/Si; η and EA decreased with increasing NBO/Si. Unlike Na2O and MgO (both are network modifiers), TiO2 worked as a weak network former, 131

160 evidenced by the formation of two peaks around 705 and 805 cm -1 in Raman spectra. (4) The fluorine-free mould fluxes were compared with the commercial fluorinecontaining Fluxes C1 and C2 based on their melting properties and viscosities. Th of all the fluxes is higher than that of Flux C1, but Th of Na2O-containing fluxes with CaO/SiO2 ratio of 1.3 and Th of for TiO2-containing and MgOcontaining fluxes with CaO/SiO2 ratio of 1.3 and 1.5 have close Th to that of Flux C2. Values of viscosity η of all synthetic fluxes are higher than those of commercial ones. Therefore, multi-component flux systems (more than 5 components) need to be developed to further reduce viscosity and meet the requirements for industrial application. 132

161 Chapter 7 - Melting properties and viscosity of multicomponent fluxes In this chapter, B2O3-containing fluxes with 6-, 8- and 10-components were investigated. Total 28 fluxes were selected and their chemical compositions are listed in Table 4-2. The 6-component systems include SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 (Fluxes 24-30), SiO2-CaO-Al2O3-B2O3-Na2O-MgO (Fluxes 31-37), and SiO2-CaO-Al2O3-B2O3-TiO2- MgO (Fluxes 38-44). The flux design for these systems was such that the effects of CaO/SiO2 and last two components are able to be examined. For 8-component SiO2- CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O (Fluxes 45-48), and 10-component SiO2-CaO- Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2 (Fluxes 49-51), only the effect of CaO/SiO2 was investigated. 7.1 Melting properties The effects of CaO/SiO2 ratio, Na2O content and TiO2 content on Ts, Th, and Tf of the SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes are illustrated in Figure 7-1. The value of Th decreased slightly from 1078 C to 1061 C with the increase of CaO/SiO2 ratio from 0.9 to 1.1, but increased to 1092 C when further increasing CaO/SiO2 ratio to 1.3 (Figure 7-1a). Change in the Na2O concentration from 6.6 to 9.2 wt% or TiO2 concentration from 2.1 to 4.2 wt% both decreased Ts, Th, and Tf. However, further increase in their concentration led to no further change in Th and Ts but a slight increase in Tf (Figures 7-1b,c). 133

162 Figure 7-1 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of TiO2 on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes. The effects of CaO/SiO2 ratio, Na2O content and MgO content on Ts, Th, and Tf of the SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes are plotted in Figure 7-2. Similar to SiO2- CaO-Al2O3-B2O3-Na2O-TiO2 fluxes, the value of Th decreased first with the increase of CaO/SiO2 ratio from 0.9 to 1.1, and then increased when further increasing CaO/SiO2 ratio to 1.3 (Figure 7-2a); increase in Na2O concentration from 7.1 to 8.9 wt% and MgO addition from 0.9 to 2.2 wt% decreased Ts, Th, and Tf (Figures 7-2b,c). However, further increase in Na2O and MgO stabilised Th and Ts and slightly increased Tf (Figures 7-2b,c). 134

163 Figure 7-2 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes. The effects of CaO/SiO2 ratio, TiO2 content and MgO content on Ts, Th, and Tf of the SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes are shown in Figure 7-3. Increase of CaO/SiO2 ratio from 0.9 to 1.3, TiO2 concentration from 2 to 6.2 wt% and MgO concentration from 1.0 to 3.7 wt% all decreased Tf, Th and Ts (Figure 7-3), except a slight increase in Tf when MgO increased from 1 to 2.2 wt%. 135

164 Figure 7-3 Effects of (a) CaO/SiO2 ratio, (b) content of TiO2 and (c) content of MgO on Ts, Th and Tf of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes. Unlike the above mentioned 6-component fluxes, Ts, Th, Tf increased with the increase of CaO/SiO2 ratio for 8-component SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and 10-component SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MgO-Li2O fluxes (shown in Figure 7-4), except in the former flux system, a slight decrease in Tf was observed with CaO/SiO2 ratio increase from 1.1 to

165 Figure 7-4 Effects of CaO/SiO2 ratio on Ts, Th and Tf of (a) SiO2-CaO-Al2O3-B2O3- Na2O-TiO2-MgO-Li2O and (b) SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO- ZrO2 fluxes. 7.2 Viscosity, break temperature and activation energy Viscosity of the SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes versus temperature for different CaO/SiO2 ratios, Na2O concentrations and TiO2 concentrations are plotted in Figure 7-5. Viscosity of Flux 25 (CaO/SiO2 = 1.1) was not obtained because of viscometer failure; therefore only two curves were listed for comparison in each figure. Similar to 5-component Na2O-containing fluxes, the viscosity of Fluxes 30 with CaO/SiO2 ratio of 1.3 exhibited a two-stage increase with the decrease in temperature (Figure 7-5a). Values of flux viscosities at 1400 C and 1300 C, break temperatures and activation energies are shown in Table 7-1. No apparent Tbr was obtained for Flux 26 with a lower Na2O concentration of 6.6 wt% (Figure 7-5b) and Flux 29 with a higher TiO2 concentration of 6.2 wt% (Figure 7-5c). Both η and EA decreased with the increase of CaO/SiO2 ratio from 0.9 to 1.3 (Fluxes 24 and 30), Na2O concentration from 6.6 to 10.8 wt% (Fluxes 26 and 27) and TiO2 concentration from 2.1 to 6.2 wt% (Fluxes 28 and 29). 137

166 Figure 7-5 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of TiO2 on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2 fluxes. Table 7-1 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-TiO2 fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol

167 Viscosity of the SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes as a function of temperature for different CaO/SiO2 ratios, Na2O concentrations and MgO concentrations are plotted in Figure 7-6. Values of flux viscosities at 1400 C and 1300 C, break temperatures and activation energies are shown in Table 7-2. The viscosity curve of Flux 37 with the CaO/SiO2 ratio of 1.3 exhibited a two-stage increase with the decrease in temperature, the same as the above-mentioned Na2O-containing fluxes with the same CaO/SiO2 ratio (Figure 7-6a). No Tbr was able to be deduced for Flux 33 with a lower Na2O concentration of 7.1 wt%. Figure 7-6 Effects of (a) CaO/SiO2 ratio, (b) content of Na2O and (c) content of MgO on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-Na2O-MgO fluxes. 139

168 Viscosity η and EA decreased with the increase of CaO/SiO2 ratio from 0.9 to 1.3 (Fluxes 31, 32 and 37) and MgO concentration from 0.9 to 3.6 wt% (Fluxes 35, 32 and 36). With the increase of Na2O concentration from 7.1 to 10.9 wt%, viscosity η remained no change, but activation energy EA decreased (Fluxes 33, 32 and 34). Tbr decreased first from 1095 to 1040 C with the increase of CaO/SiO2 ratio from 0.9 to 1.1, and then increased significantly to 1321 C with the increase of CaO/SiO2 ratio from 1.1 to 1.3. The increase in Na2O content from 8.9 to 10.9 wt% (Fluxes 32 and 34) increased Tbr, while the influence of MgO content from 0.9 to 3.6 wt% (Fluxes 35, 32 and 36) on Tbr was not significant. Table 7-2 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-MgO fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol Viscosity of the SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes as a function of temperature for different CaO/SiO2 ratios, TiO2 concentrations and MgO concentrations are plotted in Figure 7-6. Values of flux viscosities at 1400 C and 1300 C, break temperatures 140

169 and activation energies are shown in Table 7-3. Viscosity η and EA decreased with the increase of CaO/SiO2 ratio from 0.9 to 1.3 (Fluxes 38, 39 and 44) and TiO2 concentration from 2 to 6.2 wt%. η and EA decreased first with MgO content from 1.0 to 2.3 wt%, and then stabilised with further MgO content from 2.3 to 3.7 wt% (Fluxes 42, 39 and 43). Tbr increased first with the increase of CaO/SiO2 ratio from 0.9 to 1.1 and then decreased with further increasing CaO/SiO2 from 1.1 to 1.3. An increase in TiO2 concentration from 2 to 6.2 wt% or MgO concentration from 1.0 to 3.7 wt% decreased Tbr. Figure 7-7 Effects of (a) CaO/SiO2 ratio, (b) content of TiO2 and (c) content of MgO on the viscosity-temperature curves of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes. 141

170 Table 7-3 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-TiO2-MgO fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol The influence of CaO/SiO2 ratio on viscosity was also investigated for the 8-component and 10-component systems, SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and SiO2- CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2 fluxes, respectively (shown in Figure 7-8). Values of flux viscosities at 1400 C and 1300 C, break temperatures and activation energies are listed in Table 7-4. No apparent Tbr was observed for Fluxes 46, 49 and 51. For SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O fluxes, viscosity η stayed the same with the increase of CaO/SiO2 ratio from 0.9 to 1.0 (Fluxes 45 and 46), and then decreased with a further increase of CaO/SiO2 ratio from 1.0 to 1.2 (Fluxes 46-48). Activation energy EA and break temperature Tbr decreased with the increase of CaO/SiO2 ratio from 0.9 to 1.1, and then increased with the further increase of CaO/SiO2 ratio from 1.1 to 1.2. For SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O- MnO-ZrO2 fluxes, η declined with the increase of CaO/SiO2 ratio from 0.8 to 1.3 (Fluxes 49 and 50), and then rose with the further increase of CaO/SiO2 ratio from

171 to 1.5 (Fluxes 50 and 51). EA decreased with the increase of CaO/SiO2 ratio from 0.8 to 1.3. Figure 7-8 Effects of CaO/SiO2 ratio on the viscosity-temperature curves of (a) SiO2- CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and (b) SiO2-CaO-Al2O3-B2O3-Na2O-TiO2- MgO-Li2O-MnO-ZrO2 fluxes. Table 7-4 Parameters derived from the viscosity-temperature curves of SiO2-CaO- Al2O3-B2O3-Na2O-TiO2-MgO-Li2O and SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O- MnO-ZrO2 fluxes. Flux No. η at 1400 C, Pa s η at 1300 C, Pa s Tbr, C EA, kj/mol

172 7.3 Comparison between fluorine-free fluxes with commercial fluxes In comparison with two typical commercial fluorine-containing fluxes for low-carbon (Flux C1) and medium-carbon steel (Flux C2) listed in Appendix I, values of Th of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes are higher than 1150 C, above those for Flux C1 (1050 C) and Flux C2 (1150 C). For Na2O-containing 6-component fluxes, SiO2- CaO-Al2O3-B2O3-Na2O-TiO2/MgO, Th reduces to the range of C, in between the Flux C1 and Flux C2. The addition of Na2O content into the fluxes caused the drop of Tliq significantly (Figures 4-3 and 4-4), and therefore Th decreased. For more complex fluxes, Fluxes have lower Th than 1050 C, Th of Flux 50 with CaO/SiO2 ratio of 1.3 is within the range of C, and Th of Flux 51 with CaO/SiO2 ratio of 1.5 is higher than 1150 C. Similar effect of CaO/SiO2 ratio on increasing melting properties of 8- and 10-component fluxes was also observed for CaO-SiO2-B2O3-Na2O-MgO-TiO2-Li2O-MnO mould fluxes with CaO/SiO2 ratio from 0.6 to 1.15 in work [28]. Values of Tbr of SiO2-CaO-Al2O3-B2O3-TiO2-MgO fluxes are higher than those of Fluxes C1 (1170 C) and C2 (1180 C), while SiO2-CaO-Al2O3-B2O3-Na2O-TiO2/MgO fluxes have lower Tbr than those of commercial fluxes, except Fluxes 30 and 37 both with a high CaO/SiO2 ratio of 1.3. For more complex fluxes, values of Tbr of Fluxes 46, 47 and 48 with CaO/SiO2 ratio of 0.9, 1.1 and 1.2 are lower than those of two commercial fluxes, with Flux 48 has the closest Tbr. Values of the viscosity of designed complex fluorine-free fluxes (Fluxes 45-50) are close to that of Flux C1 (0.19 Pa s at 1300 C), but much higher than that of Flux C2 (0.05 Pa s at 1300 C). Flux 51 with the highest CaO/SiO2 ratio of 1.5 has a much higher viscosity than both commercial fluxes, which should be attributed to its high tendency of crystallisation. [99] 144

173 Phase equilibria of the CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2 fluxes with different CaO/SiO2 ratios are shown in Figure 7-9. Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq derived from this figure are listed in Table 7-5. Crystallisation of this system starts with precipitation of Li2SiO3; however, the amount of this phase is small for all compositions. When the CaO/SiO2 ratio is 0.8 (Figure 7-9a), the main phases are CaSiO3 and combeite. The main Ti-containing phase is CaTiO3, rather than CaSiTiO5 in the CaO-SiO2-Al2O3-B2O3-TiO2 system shown in Figure 6-11a. With the increase in CaO/SiO2 ratio from 0.8 to 1.3 (Figure 7-9b), Ca11B2Si4O22 forms, CaSiO3 is replaced by merwinite (Ca3MgSi2O8), Ca2B2O5 is replaced by Ca3B2O6, and combeite decreases. Ca11B2Si4O22 [53, 168] and combeite [29] were proposed as substitutes for cuspidine of fluorine-containing fluxes. With further increase in the CaO/SiO2 ratio from 1.3 to 1.5, fractions of Ca11B2Si4O22 increased, while those of combeite decreased (Figure 7-9c). Clearly, the phase fractions shown in the equilibrium phase diagram, Figure 7-9, do not reflect the change of Th as functions of flux composition, and neither the change of Tbr and viscosity. The discussion will be similar to the case for 5-component flux systems described in section 6.4. Although these difficulties, it can be concluded based on experimental measurement that multi-component fluorine-free mould fluxes with a controlled CaO/SiO ratio and co-existence of B2O3 and Na2O can be good candidates for the application for at least low carbon steel casting. 145

174 Figure 7-9 Calculated equilibrium phase fractions of CaO-SiO2-Al2O3-Na2O-B2O3- TiO2-MgO-Li2O-MnO-ZrO2 fluxes using FactSage with different CaO/SiO2 ratios of (a) 0.8; (b) 1.3; and (c) 1.5. Table 7-5 Parameters derived from calculated equilibrium phase fractions of CaO-SiO2- Al2O3-Na2O-B2O3-TiO2-MgO-Li2O-MnO-ZrO2 fluxes. Flux No. Tsol, C Tliq, C Dominant high-temperature phases combeite, CaSiO3, CaTiO3, Li2SiO combeite, Ca11B2Si4O22, merwinite, Ca3B2O6, olivine Ca11B2Si4O22, combeite, merwinite, Ca3B2O6, olivine, CaSiO3 146

175 7.4 Summary and conclusion The melting properties and viscosity of multicomponent (6-, 8- and 10-component) fluxes were investigated as functions of the ratio of CaO/SiO2 and the concentration of Na2O, TiO2, and MgO. The results showed complicated effects of CaO/SiO2 ratio and individual components on these properties. The major results are summarised as follows: (1) Values of Th and Tbr of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2/MgO fluxes are generally lower than those of CaO-SiO2-Al2O3-B2O3-TiO2-MgO fluxes, indicating the effect of Na2O on reducing Th and Tbr. Increase in Na2O, TiO2, and MgO had a decreasing effect on Th. With the increase of CaO/SiO2 ratio from 0.9 to 1.3, Th decreased first and then increased for SiO2-CaO-Al2O3-B2O3- Na2O-TiO2/MgO fluxes; all way decreased for CaO-SiO2-Al2O3-B2O3-TiO2- MgO fluxes; and increased monotonically for SiO2-CaO-Al2O3-B2O3-Na2O- TiO2-MgO-Li2O-(MnO-ZrO2) fluxes. (2) For 6-comonent fluxes, viscosity η and activation energy EA decreased with the increase of the CaO/SiO2 ratio from 0.9 to 1.1, Na2O concentration from 7 to 11 wt%, TiO2 concentration from 2 to 6 wt%, and MgO concentration from 1 to 4 wt%. For more complex multi-component system SiO2-CaO-Al2O3-B2O3-Na2O- TiO2-MgO-Li2O-MnO-ZrO2 fluxes, η declined first with the increase of CaO/SiO2 ratio from 0.8 to 1.3 and then increased with further increasing CaO/SiO2 to 1.5. (3) Melting properties and viscosity of fluorine-free multi-component fluxes are compared with the commercial fluorine-containing Fluxes C1 and C2. Values of Th of all six-component fluxes are higher than that of Flux C1, while those for more complex 8- and 10-component fluxes with CaO/SiO2 ratio less than

176 are lower than 1050 C. Na2O-containing 6-component fluxes with CaO/SiO2 ratio less than 1.3 and more complex 8- and 10-component fluxes with CaO/SiO2 ratio less than 1.5 have lower Tbr than those of commercial fluxes. Values of viscosity η of all six-component fluxes are higher than those of commercial ones, while values of viscosity of more complex fluorine-free fluxes with CaO/SiO2 ratio less than 1.5 are close to that of Flux C1. Therefore, further work in developing fluorine-free mould flux could be based on the composition around these fluxes for low-carbon steel and reduce the viscosity of the fluxes to meet the requirements for medium-carbon steel. 148

177 Chapter 8 - BP neural network model for viscosity results As mentioned in the literature part (section 2.3.5), most models are developed for modelling of viscosity for CaO-SiO2 based fluorine-containing slags at the initial stage. In some models, though B2O3 was taken into consideration, but when they are applied for fluorine-free boracic fluxes invested in our research, these models are not applicable. Other researchers tried to modify empirical models, such as Riboud/NPL models, for fluorine-free boracic fluxes, but only applicable in a narrow composition range, not satisfactory for our broad systems. In this chapter, all measured viscosity of B2O3- containing fluorine-free mould fluxes was modelled and validated using the BP neural network model. Based on the results of this modelling, the effects of temperature, CaO/SiO2 ratio, and flux composition on the viscosity of SiO2-CaO-Al2O3-B2O3-Na2O- TiO2-MgO-Li2O-MnO-ZrO2 fluxes were predicted and discussed. 8.1 Motivation for developing BP neural network model Viscosity is one of the most important physicochemical properties of mould fluxes, which ensures proper lubrication between the steel strand and the copper mould during the steel continuous casting. [6, 190] However, an accurate measurement of viscosity is very difficult; the experimental error in the measurement of viscosities of molten slags/fluxes was estimated to be ±25%. [ ] Moreover, a mould flux is a multicomponent system; measurement of viscosity of this system is time-consuming and expensive. Therefore, the prediction of flux viscosity in a broad range of temperatures and chemical compositions is important for the process optimisation. [9] In recent years, a number of models for calculation of viscosity of molten fluxes have been developed based on a large quantity of experimental data. [127, 129, 132] These models were 149

178 commonly expressed in the forms of Arrhenius, [32, 133, 134] [87, 101, 132, Weymann-Frenkel, ] Brostow, [88] and Vogel-Fulcher-Tammann [139] equations. Estimation of viscosity of fluorine-containing and fluorine-free mould fluxes was carried out using original or modified Riboud model, [21, 34, 62, 101, 110, 135] Koyama model, [133, 140] Urbain model, [136] Iida model, [134, 141] NPL model, [12, 32] and Zhang model, [92, 108, 142] etc. For the B2O3- containing fluorine-free mould fluxes, however, none of the existing models were applicable. [62, 92, 129, 148] Several trials were conducted to modify above-mentioned models, such as modified Riboud model for CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO [21] and CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO fluxes, [101] or modified NPL model for CaO-SiO2-Al2O3-B2O3 mould fluxes in Chapter 5. Nevertheless, these modified models are still not accurate for all the multicomponent B2O3-containing fluorine-free mould fluxes examined in this study. Therefore, it is important to develop a new and more powerful model for the viscosity prediction, correctly reflecting the complex effects of temperature, chemical composition, and structure (the degree of [92, 149] polymerisation) of the fluxes. Back propagation (BP) neural network [191, 192] developed by Rumelhart and McClelland [193] is an artificial intelligence approach for mathematical modelling. It is one of the most widely used neural network models. BP neural network is a sophisticated model-building technique, being capable of modelling data represented by non-linear functions. The viscosity of multiple flux system, which is complex and nonlinear, is unattainable to be described by analytical methods or empirical relationships, but can be dealt with by the BP neural network. 150

179 8.2 Modelling theory Principle of BP neural network BP neural network is inspired by the way in which human brain works. A human brain consists of 10 to 14 billion neurons, which are linked together into a complex network. As shown in Figure 8-1, a neuron is composed of axon, synapse, soma and dendrite. An input signal into the dendrites is recognised by the soma. When the intensity of the signal is higher than a certain target value, an output signal with stimulation is conveyed through axon and synapse by sending an electrical signal along the axon connected with other neurons. [191] Figure 8-1 Schematic structure of a typical neuron. Figure 8-2 is the sketch of BP neural network model. It consists of input layers, hidden layers, and output layers. Each set of layers is formed by a number of nodes, and each node represents a neuron. Nodes of the input layers and the hidden layers are connected by the weights W, and nodes of the hidden layers and the output layers are connected by the weights V; n is the number of neurons in input layer; m is the number of neurons in 151

180 hidden layer; and P is the number of neurons in output layer (Figure 8-2 shows only one output layer of neuron). Figure 8-2 Schematic structure of BP neural network model. BP neural network is a multilayered feedforward network based on the error BP algorithm. The information flow is from the training samples into input layer, through the hidden layer, and then to the output layer. In this process, the calculated output is obtained through the operation of corresponding thresholds, active functions and connection weights between nodes. If input signals reach a critical value, they are conveyed through a sigmoid function to the hidden layer. The new signals that attain another critical value in the hidden layer are conveyed also through a sigmoid function to the final output layer, forming a calculated output. After that, the calculated output is compared with the experimental data. If the output error exceeds the set value, the information will be fed back, and the network weights and thresholds are adjusted until the network error finally reaches the minimum requirement. In this study, the BP neural network applied the Vogl high-speed BP algorithm, [194] which has three advantages: firstly, it has a fast learning speed and a high convergence 152

181 rate; secondly, it avoids the network to be at the local minimum; thirdly, the neural network has a good adaptive capacity and has no generalization problem. This BP algorithm contains the concept of batch process and includes two parts: momentum and optimised study speed. The computing formulas are detailed in reference. [194] Viscosity modelling structure The structure of BP neural network model applied for viscosity is shown in Figure 8-3. As shown in Figure 8-3, BP neural network viscosity model has a typical three-layered structure. For the input layers, input variables include temperature, concentrations of different components, parameters of CaO/SiO2 ratio and optical basicity (Λ). CaO/SiO2 ratio is a convenient characteristic of mould fluxes used in industry. [21, 28, 29] Optical basicity characterises the degree of polymerisation. [13, 31] In this particular case, the total number of variables (temperature, CaO/SiO2 ratio, flux component contents, and optical basicity) is 13, so the number of input layers is 13. Similarly, as only the viscosity comes out from the output, the number of variables is therefore 1 for the output. The number of hidden layers was determined in such a way that the best fitting results were achieved with this number. 153

182 Figure 8-3 Structure of BP neural network for viscosity modelling. The sigmoid function is applied for the calculation from both input layers to hidden layers and hidden layers to output layer. This function with an "S" shaped curve (shown in Figure 8-4) often acts as a logistic function. The sigmoid function is defined by the formula: f(t) = 1 1+e t (8-1) where f (t) is the output, and t is the signal. 154

183 Figure 8-4 The sigmoid function curve. 8.3 Validation of the BP neural network model BP neural network model was developed and validated using 70 experimental values of viscosity of fluorine-free mould fluxes CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O- MnO-ZrO2 in this thesis work. Besides the compositions listed in Table 4-1, other 19 fluxes were prepared and values of viscosities of the fluxes were measured using the rotation viscometer. Only viscosity values at the static measurement were used for modelling. Considering modelling development, the compositions, CaO/SiO2 ratios, optical basicities, temperatures, and viscosity of the examined fluxes are reordered and listed in Table 8-1. The first 51 experimental values of viscosity in Table 8-1 were used to develop the BP model and the rest data were used for the validation of the model. 155

184 Table 8-1 Composition (wt%), CaO/SiO2 ratio, optical basicity, temperature and viscosity of fluorine-free mould fluxes. Flux No. C/S Λ CaO SiO 2 Al 2O 3 B 2O 3 Na 2O TiO 2 MgO LiO 2 MnO ZrO 2 T, C η, Pa s

185

186 Calculated viscosities are compared with the experimental data shown in Figure 8-5, which shows that the neural network model reproduces the experimental viscosity values with an acceptable error. Figure 8-5 Comparison between the calculated viscosity using neural network model and the experimental data. 158

187 8.4 Comparison of the BP neural network model with other models The BP neural network model was compared with Riboud, [135] Urbain, [136] Koyama, [140] Iida, [141] NPL, [12] and modified Riboud [21, 101] and NPL models in section 5.6. Viscosities calculated using different models in comparison with experimental data are shown in Figure 8-6, and parameters σ and, are presented in Table 8-2. As shown in Figure 8-6a, values of viscosity predicted using the Riboud model [34, 135] fluctuate around the experimental data with high scattering, while the viscosity values predicted using the modified Riboud model1 and model2 are generally higher than measured results. The viscosity values predicted by Koyama model, [133, 140] Iida model [134, 141] and Urbain model [136] are also generally higher than experimental values (Figures 8-6b-d). The NPL model, [12] however, predicts viscosity of different fluxes at a roughly similar value around 0.2 Pa s, which is unreasonable. The modified NPL model which was developed for 4-component fluorine-free fluxes predicts far higher values than measured data (Figure 8-6e). Results of this analysis showed that the BP neural network model (σ = 0.05 Pa s, = 23.5) is more reliable for calculation of viscosity of fluorine-free fluxes in comparison with all other models. 159

188 Figure 8-6 Comparison between experimental and calculated viscosity using different models: (a) Riboud model; (b) Urbain model; (c) Koyama model; (d) Iida model; and (e) NPL model. 160

189 Table 8-2 Parameters σ and for different models for calculation of viscosity of the fluorine-free mould fluxes. Model Riboud model [135] Modified Riboud Model1 [21] Modified Riboud Model2 [101] Urbain model [136] Koyama model [140] σ, Pa s Model Iida model [141] NPL model [12] Modified NPL model Neural network model σ, Pa s Analysis of factors affecting viscosity based on modelling results Influence of temperature The effect of temperature on viscosity of fluorine-free mould fluxes was predicted based on the developed BP neural network model; the results are shown in Figure 8-7. As expected, the viscosities decrease with temperature in the examined range of temperatures. The 4-component CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3 flux (referred as flux M in Figure 8-7) had the highest viscosity. The addition of Na2O, TiO2 or MgO to the CaO-SiO2-Al2O3-B2O3 system reduced the viscosity; the most significant reduction was caused by the addition of Na2O. All other more complex systems of 6- or morecomponent fluxes had even lower viscosity. 161

190 The effect of temperature on the flux viscosity is more significant for fluxes with a higher CaO/SiO2 ratio. At the highest temperature (1450 C), values of the fluxes with a higher CaO/SiO2 ratio (1.3) are generally lower than those of a lower CaO/SiO2 ratio (1.0). At the lowest temperature (1300 C), the precipitation of solid phases could occur in the flux, which leads to an increase in viscosity. Increase in CaO/SiO2 ratio increased the crystallisation temperature of the fluxes. [26, 58] Therefore, the difference between viscosity values at 1300 C and 1450 C for fluxes with a higher CaO/SiO2 ratio is larger. The melting properties of fluxes with 4- and 5-components are generally higher than those of more component fluxes analysed in Chapters 5-7, indicating the possibility of formation of more solid phases at 1300 C for 4- and 5-component fluxes. Viscosity of 4- and 5-component fluxes was affected by temperature stronger than that of more complex systems (6 or more components). Figure 8-7 Influence of temperature on viscosity of fluxes with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. 162

191 8.5.2 Influence of CaO/SiO2 ratio Figure 6-8 shows the effect of CaO/SiO2 ratio on viscosity of different fluxes at 1300 C and 1400 C predicted by the BP neural network model. At both temperatures, viscosity decreased with the increase of CaO/SiO2 ratio from 0.8 to 1.1 for almost all fluxes. With further increasing CaO/SiO2 ratio from 1.1 to 1.5, the decrease in viscosity slowed down for some fluxes; however the viscosity of some other fluxes increased slightly in this CaO/SiO2 range. The decrease in viscosity with the CaO/SiO2 ratio can be attributed to the decrease in the degree of the flux polymerisation. Structure of siliceous slags and fluxes includes different types of anionic tetrahedral groups. [13] CaO is a typical network modifying oxide, leading to the breakdown of the network structure. As a result, the increase of the CaO/SiO2 ratio promotes the reforming of complex network units and lowers the viscosity of the fluxes. This conclusion is in agreement with other reports on fluorinecontaining [12, 35, 97] and fluorine-free [21, 28, 48, 62, ] fluxes. Cations of Al 3+ substitute for Si 4+ to form tetrahedral AlO4 5-, with electric charge compensation by cations of Na + or Ca 2+. [98] The increase in viscosity with increasing CaO/SiO2 ratio was found for some fluxes when CaO/SiO2 ratio was above 1.1. This phenomenon is more significant at the lower temperature of 1300 C. Seok et al. [99] found a similar influence of CaO/SiO2 ratio on the viscosity of CaO-SiO2-MgO-FeO fluxes when CaO/SiO2 ratio was over 1.4. The increase in the flux viscosity can be caused by the formation of solid phases at 1300 C. [28, 35, 98, With the increase of the CaO/SiO2 ratio, melting temperature of fluxes increases. 99] Increasing in the melting temperature promotes the nucleation of solid phase, which increases viscosity. [99] In this way, the influence of the formation of solid phases on 163

192 increasing viscosity is in competition with the influence of the depolymerisation on decreasing viscosity with the increase of the CaO/SiO2 ratio. Obviously, the tendency of precipitation of solid phases increases at lower temperatures which makes the viscosity increase more significantly. Figure 8-8 Influence of CaO/SiO2 ratio on flux viscosity at (a) 1300 C and (b) 1400 C. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O Influence of components (B2O3, Na2O, TiO2, MgO) The influence of B2O3 on viscosity of fluxes with CaO/SiO2 ratios of 1.0 and 1.3 is shown in Figure 8-9. For the Na2O-free fluxes, the increase of B2O3 from 5 to 9 wt% decreased viscosity. This decrease was more significant for the fluxes with CaO/SiO2 ratio 1.3. For the fluxes containing Na2O, the influence of B2O3 was marginal. The negligible influence of B2O3 content on viscosity was observed in CaO-SiO2-Al2O3-(7-10 wt%)b2o3-na2o-tio2-mgo [21] fluxes and CaO-SiO2-Al2O3-(6-10 wt%)b2o3-na2o- TiO2-MgO-CaF2 [68] fluxes. A decrease in the flux viscosity with the addition of B2O3 on viscosity was reported for CaO-SiO2-Al2O3-(1.5-8 wt%)b2o3-na2o-mgo, [2] CaO-SiO2- Al2O3-(0-7 wt%)b2o3-na2o-tio2-mgo, [21] CaO-SiO2-Al2O3-(2-6 wt%)b2o3-na2o- 164

193 TiO2-MgO-CaF2, [68] CaO-SiO2-Al2O3-(2-8.1 wt%)b2o3-na2o-tio2-mgo-li2o- MnO, [101] and CaO-SiO2-Al2O3-(4-10 wt%)b2o3-na2o-tio2-mgo-li2o-mno-feo [28] fluxes. Tetrahedral BO4 5- groups in the molten flux are preferentially incorporated into the Si-O-Si network forming Si-O-B structure. [177, 179] However, the main structural groups in fluxes with boron are trigonal BO3 3-[177, 179] As a result, the flux viscosity decreases with the increase in the B2O3 content. Figure 8-9 Influence of the B2O3 content on viscosity of fluxes at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(5-9 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. The influence of Na2O on viscosity of different fluxes at 1400 C with CaO/SiO2 ratios of 1.0 and 1.3 is shown in Figure As mentioned above, the introduction of Na2O into CaO-SiO2-Al2O3-B2O3 flux decreased viscosity significantly. The reason is that Na2O is a strong network modifier and the addition of Na2O significantly depolymerises the network. This phenomenon was also observed for CaO-SiO2-Al2O3-(0-15 wt%)na2o, [103] CaO-SiO2-Al2O3-TiO2-(0-8.2 wt%)na2o, [104] CaO-SiO2-Al2O3-(0-5.5 wt%)na2o-mgo-mno-li2o, [29] CaO-SiO2-Al2O3-B2O3-(2-8.1 wt%)na2o-tio2-mgo- Li2O-MnO, [101] and CaO-SiO2-Al2O3-B2O3-(2-10 wt%)na2o-tio2-mgo-li2o-mno- 165

194 FeO [28] fluxes. However, a further addition of Na2O to the flux up to 11 wt% had a minor influence on viscosity. Similar observation was also reported for CaO-SiO2- Al2O3-B2O3-(7-11 wt%)na2o, [98] CaO-SiO2-Al2O3-B2O3-( wt%)na2o-tio2- MgO-Li2O-MnO, [101] and CaO-SiO2-Al2O3-( wt%)na2o-mgo-mno-li2o [29] fluxes. This phenomenon can be related to an increasing extent of evaporation of Na2O with further increase in Na2O content. The introduction of Na2O into boracic fluxes causes the formation of highly volatile NaBO2. [ ] Evaporation of Na2O increased with increasing Na2O concentration. Therefore, when 7-11 wt% Na2O was added to the flux, its actual concentration in the flux could be lower than the targeted one and the effect of Na2O on the flux viscosity was lower than what would be for the Na2O content in this range. Figure 8-10 Influence of Na2O content on viscosity of fluxes at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: (7-11) wt% Na2O, 4 wt% TiO2, 3 wt% MgO and 1 wt% Li2O. Figure 8-11 illustrates the effect of TiO2 on viscosity of different fluxes at CaO/SiO2 ratios of 1.0 and 1.3. In general, calculated viscosity curves reproduce experimental data well. Wen et al. [28] observed a similar influence of TiO2 content on the viscosity of 166

195 CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-FeO fluxes. The addition of TiO2 into CaO-SiO2-Al2O3-B2O3 fluxes partially replaces Si 4+ in ionic units SiO4 4- with Ti 4+ as the network former. In this case, the incorporation of Ti into Si-related structures and formation of simpler or unstable structural units causes the depolymerisation of the flux structure, [21, 62, 189] which weakens the stability of network structure and decreases the viscosity of the fluxes. [21, 107] A similar tendency in the effect of TiO2 on viscosity was observed in CaO-SiO2-(0-20 wt%)tio2, [105] CaO-SiO2-Al2O3-(0-20 wt%)tio2, [105] CaO-SiO2-Al2O3-(23-43 wt%)tio2, [62] CaO-SiO2-Al2O3-(5-35 wt%)tio2, [100] CaO- SiO2-Al2O3-(0-5 wt%)tio2-mgo, [106] CaO-SiO2-Al2O3-(0-10 wt%)tio2-mgo with CaO/SiO2 at 1.0 and 1.2, [107] CaO-SiO2-Al2O3-(15-30 wt%)tio2-mgo, [100] and CaO- SiO2-Al2O3-B2O3-Na2O-(0-10 wt%)tio2-mgo fluxes with CaO/SiO2 at 1.0. [21] Figure 8-11 Influence of TiO2 content on flux viscosity at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, (0-6) wt% TiO2, 3 wt% MgO and 1 wt% Li2O. The influence of MgO on viscosity of different fluxes with CaO/SiO2 ratios of 1.0 and 1.3 is shown in Figure In general, the addition of MgO from 0 to 5 wt% slightly reduced viscosity. A similar influence on viscosity was also reported in CaO-SiO2-(

196 wt%)mgo, [105] CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-(3-8 wt%)mgo-li2o-mno-feo, [28] CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-( wt%)mgo-li2o-mno, [101] and CaO-SiO2- Al2O3-Na2O-(0-4 wt%)mgo-mno-li2o [29] fluxes. However, viscosity kept unchanged for CaO-SiO2-Al2O3-B2O3-Na2O-MgO fluxes with CaO/SiO2 ratio of 1.0, and increased slightly for CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O fluxes with CaO/SiO2 ratio of 1.0, with the addition of MgO content. MgO is a weaker network modifier in comparison with CaO; the introduction of MgO into CaO-SiO2-Al2O3-B2O3 flux decreases the degree of flux polymerisation. The influence of MgO on viscosity of systems with high content of Na2O (9 wt%) is more complex, because the ability of Na2O to depolymerise flux is stronger than that of MgO. [13, 29] However, some extents of MgO loss during flux preparation indicate the possible evaporation of MgO. This could be the reason that the viscosity decreased to a small extent with the addition of MgO content. Different from fluxes investigated in this research, Saito et al. found that the influence of MgO addition into CaO-SiO2-Al2O3 fluxes with CaO/SiO2 of 1.3 on viscosity is much less than that of TiO2, which acts as a network modifier in the flux. [105] The effect of MgO content on the flux viscosity was a bit more remarkable for fluxes with a higher CaO/SiO2 ratio. A similar effect of MgO on viscosity was reported for CaO-SiO2-Al2O3-(0-15 wt%)mgo fluxes with CaO/SiO2 ratio ranging from 0.7 to 1.2. [103] 168

197 Figure 8-12 Influence of MgO content on flux viscosity at 1400 C with CaO/SiO2 ratio of (a) 1.0 and (b) 1.3. M: CaO-SiO2-(3 wt%)al2o3-(7 wt%)b2o3; the amount of other additives: 9 wt% Na2O, 4 wt% TiO2, (0-5) wt% MgO and 1 wt% Li2O. 8.6 Summary and conclusion A neural network model was developed to describe the viscosity of CaO-SiO2-Al2O3- B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2 mould fluxes. Viscosity of mould fluxes calculated using this model was in a reasonable agreement with the experimental data and showed better fitting results than other viscosity models. Based on the BP neural network model, the effects of temperature and flux composition on viscosity were calculated and analysed. The major results are summarised as follows: (1) Viscosity of fluxes decreased with the increase in temperature from 1300 to 1450 C; the influence of temperature was more significant for Na2O-free fluxes than that for Na2O-containing fluxes. (2) An increase in CaO/SiO2 ratio from 0.8 to 1.1 led to a decrease in viscosity; further increase in CaO/SiO2 ratio from 1.1 to 1.5 slightly increased the viscosity of some fluxes which was attributed to the precipitation of solid phases as a result of increasing melting temperature. 169

198 (3) Increase in B2O3 contents (7-11 wt%) slightly decreased viscosity. (4) Addition of Na2O into the fluxes up to 7 wt% significantly decreased flux viscosity, but this decrease slowed down when Na2O addition increased from 7 to 11 wt%. (5) Addition of TiO2 content (0 to 6 wt%) to mould fluxes first marginally increased, and then continuously decreased viscosity. (6) Addition of MgO from 0 to 5 wt% slightly reduced viscosity. 170

199 Chapter 9 - Evaporation of fluorine-free mould fluxes B2O3 and Na2O are key components of fluorine-free mould fluxes for continuous casting, but both are highly volatile which affects the flux stability. The evaporation of CaO-SiO2-Al2O3-B2O3-Na2O mould fluxes in the temperature range from 1300 to 1400 C using thermogravimetric analysis (TGA) was studied in this chapter, focusing on the effects of the Na2O content (6-10 wt%) and CaO/SiO2 ratio ( ). 9.1 Weight loss of fluxes The measured chemical compositions of selected fluxes in Table 4-2 for evaporation experiment are listed in Table 9-1. This table also includes the liquidus temperatures (Tliq) of fluxes calculated using thermochemical software FactSage and their viscosity values (η). Fluxes 4, 9, 10 and 11 have varied Na2O contents from 0 to 9.1 wt% at a fixed CaO/SiO2 ratio of 1.3, while Fluxes 7, 8 and 10 have varied CaO/SiO2 ratios from 0.8 to 1.3 at Na2O contents close to 8 wt%. A typical industrial fluorine-containing flux (C1) for continuous casting of low carbon steel was also measured for a comparison. Repeated experiments for some selected fluxes demonstrated a good reproducibility of the measurements (Figure 9-1). 171

200 Table 9-1 Chemical compositions of selected fluxes for evaporation experiment (wt%), liquidus temperature ( C) and viscosity (Pa s) measured at 1400 C. Flux No. CaO/SiO2 CaO SiO2 Al2O3 B2O3 Na2O MgO Li2O CaF2 Tliq, C η, Pa s Figure 9-1 Repeated weight loss measurements of Flux 11 at 1400 C. 172

201 Figures 9-2 and 9-3 show the influence of temperature and Na2O content on weight loss of the fluxes (CaO/SiO2 ratio fixed at 1.3), respectively. The weight loss was characterised by m/m0, where m is the weight change, g, and m0 is the initial sample weight, g. The evaporation of the industrial fluorine-containing Flux C1 at 1350 C was also measured and compared with the fluorine-free mould fluxes in Figure 9-3c. The fluorine-containing flux showed much higher weight loss than fluorine-free fluxes at 1350 C (Figure 9-3c). In general, the weight loss increased with the increasing temperature (Figure 9-2). For the Na2O-free mould flux, the weight loss was much smaller than for the Na2Ocontaining fluxes, less than 0.005% at all temperatures for 1 h (Figures 9-2a, 9-3a, 9-3c, 9-3d). Adding 6.2 wt% Na2O significantly increased the rate of flux evaporation (Figures 9-3a and 9-3c). Increasing Na2O content from 6.2 to 7.9 wt % only slightly increased the rate of evaporation (Figures 9-3a-c). When the content of Na2O increased from 7.9 to 9.1 wt%, there was almost no change in the evaporation rate at 1300 C (Figure 9-3a), a slightly increased rate at 1350 C (Figure 9-3c), and an early increase but obvious slowdown after 1000 s at 1400 C (Figures 9-2d, 9-3d). As a result, the weight loss at 1400 C for the flux with 9.1 wt% Na2O was even smaller than that at 1350 C after 1 h isothermal process (Figure 9-2d). 173

202 Figure 9-2 The weight loss of fluxes with varied Na2O contents of (a) 0 wt%, (b) 6.2 wt%, (c) 7.9 wt%, and (d) 9.1 wt%, as a function of time at different temperatures with CaO/SiO2 ratio of

203 Figure 9-3 The influence of Na2O content on the weight loss of fluxes at different temperatures of (a) 1300 C, (b) 1325 C, (c) 1350 C, and (d) 1400 C. Figure 9-4 shows the weight loss of fluxes with a fixed Na2O content at approximately 8 wt% but varied CaO/SiO2 ratios from 0.8 to 1.3 at 1300 C. An increase in the CaO/SiO2 ratio from 0.8 to 1.3 had a minor influence on the weight loss. 175

204 Figure 9-4 The influence of CaO/SiO2 ratio on the weight loss of fluxes at 1300 C. 9.2 Evaporation rate and activation energy Evaporation changed the slag chemical composition and as a result affected the evaporation rate in the process of measurements. To mitigate this effect, the evaporation rates were analysed in the first 10 min of the TGA measurements. Evaporation rate k, s -1, was calculated as: k = m i m 0 t i (9-1) m i = m i m 0 (9-2) where mi/m0 is the flux mass change, and mi, g, is the flux mass at time ti, s. Table 9-2 shows the calculated evaporation rates of different fluxes. The evaporation rate changed only slightly with the increase of CaO/SiO2 ratio at the fixed Na2O content (Fluxes 7, 8, 10), while it increased with the increase of the Na2O content at the fixed CaO/SiO2 ratio of 1.3 (Fluxes 4, 9-11). The apparent activation energies for the evaporation process 176

205 were estimated from the Arrhenius plots using data on k obtained at different temperatures: lnk = lna + E A RT (9-3) where A is the pre-exponential factor; EA is the apparent activation energy, kj/mol; R is the gas constant, J/(mol K); and T is the absolute temperature, K. Values of EA were obtained from the slope of lnk vs 1/T (shown in Figure 9-5) and listed in Table 9-3. Addition of 6.2 wt % Na2O to the CaO-SiO2-Al2O3-B2O3 system significantly decreased EA. A further increase in the Na2O content from 6.2 to 9.1 wt % had a minor effect on EA. Table 9-2 Evaporation rates k (s -1 ) of fluxes. Flux No C 1325 C 1350 C 1400 C C No evaporation rates measured under these conditions 177

206 Figure 9-5 Arrhenius plots for fluxes with varying Na2O contents. Table 9-3 Calculated EA for fluxes with different Na2O contents. Flux No. Na2O, wt% EA, kj/mol R 2 * *R 2 is a statistical measure of how close the data are to the fitted regression line. 9.3 Rate controlling steps of evaporation Vaporisation of fluorine-free fluxes observed in this work, was also reported in works [151, 153, 154, 195]. Evaporation rate of fluorine-free boracic fluxes depended on the temperature and the flux composition. Increasing reaction temperature enhanced the flux evaporation. Adding Na2O into the boracic mould flux significantly increased the rate of flux evaporation. However, varied CaO/SiO2 at the fixed Na2O content had no 178

207 effect on the vaporisation. Mechanisms, kinetics of evaporation and the rate controlling steps are discussed below Evaporation Process Experimental temperatures in a study of Na2O-containing fluxes in this work were higher than liquidus temperatures calculated using FactSage (shown in Table 9-1). The calculated liquidus temperature for the Na2O-free flux (Flux 4) was higher than 1300 C, but in-situ observation using a hot thermocouple technique showed that at 1300 C, this flux was in the liquid state. Therefore, all fluxes were liquid at the experimental temperatures. The silicate-based mould fluxes have complex network structures. In the molten state, oxides form anionic and cationic species. Si-related structures are in the form of different types of anionic tetrahedral units (SiO2, Si2O5 2-, Si2O6 4-, Si2O7 6-, and SiO4 4- ) in the melt. [13] Cations of Al 3+ substitute for Si 4+ to form tetrahedral AlO4 5-, with electric charge compensation by cations of Na + or Ca 2+. [98] B2O3 is partially incorporated into the Si-O-Si network resulting in the Si-O-B structure; it also forms trigonal BO3 3- units. [177, 179] Na2O and CaO are typical network modifying oxides, leading to the breakdown of the network structure. Thermodynamic analysis using FactSage showed that major components of the gas phase in equilibrium with fluxes under examination were gaseous NaBO2, Na, O2 and B2O3. Formation of these gaseous species can be presented by the following evaporation [151, 153, 154, 157] reactions: Na2O + B2O3 = NaBO2 (g) (9-4) 2Na2O = 4Na (g) + O2 (g) (9-5) B2O3 = B2O3 (g) (9-6) 179

208 The evaporated species were transported from the molten flux surface out of the crucible to the main gas stream of Ar. Therefore, evaporation process included three main steps: (1) Evaporation reactions following Eqs.(9-4)-(9-6); (2) Internal mass transfer of the anions and cations to the surface of the molten flux; (3) External mass transfer of gaseous species from the flux/gas interface out of the crucible Rate Controlling Steps of Evaporation Chemical reaction The maximum possible rate of evaporation (free evaporation, s -1 ) of gaseous species can be calculated using Hertz-Knudsen equation: [196] k max = 44.3P A (M A T) 0.5 M A S m 0 (9-7) where PA is partial pressure of A (A: NaBO2, Na, O2, or B2O3), Pa; S is the surface area of interface, m 2 ; T is temperature in K; and MA is molecular weight of A, g/mol. Table 9-4 shows the equilibrium vapour pressures of major gaseous species for Na2Ocontaining fluxes calculated using FactSage. The vapour pressure of NaBO2 is the highest for the Na2O-containing fluxes, indicating that it is a dominant volatile species. Reaction presented by Eq.9-4 is therefore, the major reaction for the evaporation in the presence of both Na2O and B2O3. 180

209 Table 9-4 Calculated vapour pressures PA (Pa) for major gaseous species. Flux No. Gaseous species 1300 C 1325 C 1350 C 1400 C NaBO2 (g) Na (g) O2 (g) B2O3 (g) NaBO2 (g) Na (g) O2 (g) B2O3 (g) NaBO2 (g) Na (g) O2 (g) B2O3 (g) NaBO2 (g) Na (g) O2 (g) B2O3 (g) NaBO2 (g) Na (g) O2 (g) B2O3 (g) The maximum rates of evaporation calculated using Eq.9-7 is shown in Table 9-5. Calculated kmax are three orders of magnitude higher than measured evaporation rates k. 181

210 Therefore, contribution of the chemical reactions to the rate control of the evaporation of the fluxes can be ignored. Table 9-5 Maximum evaporation rates kmax (s -1 ) for fluxes. Flux No C 1325 C 1350 C 1400 C Mass transfer in molten flux Experimental results showed that the rate of evaporation was not affected by the change of the CaO/SiO2 ratio at the fixed B2O3 and Na2O contents. As shown in Table 9-1, increasing CaO/SiO2 ratio decreases the viscosity of the flux as a result of the flux depolymerisation. It is expected in accordance with Stokes-Einstein equation that diffusion coefficients of flux constituents increase with the increase in the CaO/SiO2 ratio. [165] If the mass transfer in the liquid flux contributed to the evaporation rate control, the evaporation rate would have been affected by the depolymerisation of the flux caused by the increasing CaO/SiO2 ratio. Therefore, it can be concluded that the internal mass transfer in the liquid phase is not the rate controlling step in the flux vaporisation under experimental conditions in this work. Zhang et al. [151] also observed a negligible effect of CaO/SiO2 ratio from 0.77 to 1.18 on the evaporation of CaO-SiO2- Al2O3-B2O3-Na2O-TiO2 mould fluxes. For CaO-SiO2-Al2O3 flux with high Na2O 182

211 content (20 wt%), Li et al. [153] found that the mass transfer in the liquid flux gradually became a controlling step after a long time of the evaporation reaction (3500 to 7000 s) resulting in the significant decrease of Na2O concentration. Measurements in this work lasted less than 3600 s External mass transfer in the gas phase In the analysis of the external mass transfer, it is assumed that the evaporation reaction reaches equilibrium and the flux composition remains no change. Concentration gradients in the flux are neglected based on the above analysis of the internal mass transfer. In the measurements, once the evaporation starts, gaseous species evaporate from the interface and leave the system with the carrier Ar gas. After a certain time, a steadystate is established in which the concentrations of gaseous species in the bulk Ar gas become constant. [197] The flux of gaseous species diffusing from the flux/gas interface to the top of a platinum crucible can be obtained by calculating the flux of each gaseous component JA, mol/s, which is defined as: J A = m A tm A (9-8) where ma is the mass loss of component A. In order to simplify the problem, it is assumed that there is no convection occurring inside the crucible. Then the flux of the component A can be calculated as: [153] J A S = (P A i P A b )DA Ar LRT (9-9) where i and b indicate the interface and bulk, respectively; and L is a boundary layer thickness which is assumed to be the distance between the surface of liquid flux and the 183

212 crucible top, m. L is estimated by assuming that the density of the fluorine fluxes is ρ = 3.05g/cm 3 i b. P A and P A are the partial pressures at the interface (equilibrium partial b pressure, PA) and in the bulk gas, respectively. P A in the flowing Ar gas atmosphere is assumed to be zero. D A Ar is the diffusion coefficient of gaseous species A through Ar gas, cm 2 /s, which was calculated by Chapman-Enskog equation: [196] D A Ar = T 3/2 2 Ω A Ar ( 1 P σ A Ar + 1 ) 1 2 M A M Ar (9-10) where P is the pressure, atm; M A and M Ar are molecular weights of gaseous species and Ar, respectively; and Ω A Ar is the collision integral for diffusion, which is a function of the dimensionless temperature (or reduced temperature) kt/εa-ar. [198] Parameters σ A Ar, Å, and ε A Ar, J, are parameters of the Lennard-Jones potential which can be estimated from the following equations: [198] σ A Ar = 1 2 (σ A + σ Ar ) (9-11) ε A Ar = (ε A ε Ar ) 1/2 (9-12) The values ε Ar k and σar taken from[198] are K and Å, respectively. The values of εa and σa were estimated by the following empirical equations: [198] ε A k = 1.92T m,a (9-13) 1 3 σ A = 1.222V m,a (9-14) where k is Boltzmann constant, J/K; Tm is the melting temperature, K; and Vm,A is the molar volume at melting temperature, cm 3 /mol. Table 9-6 shows the calculated ε A k and σa using Eqs.9-12 and Vm of B2O3 was regressed from the temperature dependence of molar volume in work [199]. Values of Ω A Ar for different species, calculated κt/εa-ar and D A Ar values are shown in Table

213 Table 9-6 Parameters ε A k and σa for different gaseous species. Gaseous species MA, g/mol ρliq, g/cm 3 Vm, cm 3 /mol Tm, K ε A k, K σa, Å NaBO Na O [198] [198] B2O [199] Table 9-7 Values of κt/εa-ar, Ω A Ar and estimated diffusion coefficients of gaseous species in Ar gas (10-4 m 2 /s). Parameter 1300 C 1325 C 1350 C 1400 C kt/ε NaBO2 Ar Ω NaBO2 Ar D NaBO2 Ar kt/ε Na Ar Ω Na Ar D Na Ar kt/ε O2 Ar Ω O2 Ar D O2 Ar kt/ε B2 O 3 Ar Ω B2 O 3 Ar D B2 O 3 Ar

214 The rate of total weight loss of the molten flux can be calculated using the following equation: m t = M A J A (9-15) By combining Eqs.9-9 and 9-15, evaporation rate can be estimated from the following equation: m/m t = S RTLm M A D A Ar P A (9-16) The calculated evaporation rates for different fluxes, compared with the experimental results, are shown in Figure 9-6. In general, the measured values are lower than the calculated ones, although the difference between them is small. The measured values are close to the calculated ones when the evaporation rate is low. It is suggested that the gaseous phase diffusion from the flux surface to the main gas stream contributed to the rate control. For the fluxes with low rate of evaporation, the external mass transfer played a more important role in the rate control than for those with high rate of evaporation. Zhang et al. [151] and Li et al. [153] in a similar analysis obtained much larger calculated values than the experimental data. Under conditions in their works, vapour pressures of the gaseous species were much higher than those obtained in this work. The gas phase mass transfer was also reported to contribute to the rate of evaporation of CaO-SiO2-Al2O3-Na2O-MgO-CaF2 mould fluxes at C. [166] At a given temperature, the evaporation rate is directly proportional to P NaBO2. With the increase of Na2O content, P NaBO2 increases, while the increase in CaO/SiO2 does not affect P NaBO2. The acceleration effect of Na2O on the evaporation of B2O3-containing fluxes was also [157, 200] reported by others. 186

215 Figure 9-6 A comparison of measured and estimated evaporation rates. The rate of evaporation was calculated using data for the first 10 min reaction. The rate of evaporation slowed down with the increasing exposure time. As mentioned before, flux evaporation leads to the change of the flux composition and therefore flux properties, including the rate of evaporation. After a significant decrease in the contents of volatile species, internal mass transfer is expected to play a more significant role. The change of the rate limiting step for the flux evaporation with increasing reaction time was reported by Li et al. [153]. It should be mentioned that although the FactSage vapour chemistry database is not fully reliable for complex flux systems, for simple 4/5-component fluxes, the weight loss prediction using FactSage in this work was close to the measured results. We do need to measure the composition of the vapour species when we deal with more complex systems in future work. 187

216 The effect of exposure time on the flux evaporation becomes more significant at higher temperature. In this work, the apparent descending trend in the weight loss of Flux 11 was observed after 1000 s at 1400 C (Figures 9-3(d) and 9-4(d)); it can be attributed to the composition change by reducing the contents of both boron and sodium in the flux, which may change the flux melting and evaporation properties. Figure 9-8 shows variation of the vapour pressure and liquidus temperature as a function of weight loss of Flux 11 calculated using FactSage. In these calculations, only evaporation of NaBO2 was considered. With the increase of the flux weight loss, the vapour pressure reduces (Figure 9-7a) and the liquidus temperature rises (Figure 9-7b). Based on above analysis, the external mass transfer decreases according to Eq.9-16, reducing the evaporation rate. When the dimensionless weight loss reached 0.08%, the calculated liquidus temperature approached 1400 C. It means that any further weight loss will lead to the solid precipitation which slows down the evaporation kinetics significantly. Figure 9-7 The influence of weight loss on (a) vapour pressure and (b) liquidus temperature of Flux 11 calculated using FactSage. 188

217 9.4 Summary and conclusion The study of stability of the fluorine-free mould fluxes SiO2-CaO-Al2O3-B2O3-Na2O demonstrated that the evaporation of these fluxes was much slower in comparison with industrial fluorine-containing flux. The rate of the flux evaporation depended on the temperature and flux composition. The weight loss as a result of the flux evaporation increased with the increasing temperature. The evaporation of the boracic fluxes without Na2O at C was less than %. Significant evaporation was observed when the fluxes contained both B2O3 and Na2O, which was attributed to the formation of highly volatile NaBO2. The addition of 6.2 wt% Na2O increased the flux evaporation at 1350 C to above 0.06 %. The increase in the rate of the vaporization slowed down with further increasing Na2O content to 9.1 wt%. Variation of the CaO/SiO2 ratio, however, did not change the flux evaporation rate. The flux evaporation process was analysed by considering chemical reactions and internal and external mass transfers. Major species in the gas phase in the equilibrium with molten fluxes calculated using FactSage included NaBO2, Na, O2 and B2O3. The maximum rate of evaporation reactions calculated using Hertz-Knudsen equation, was much higher than the measured evaporation rate. The calculated rate of the external mass transfer of the gaseous species from the flux/gas interface to the main gas stream was in a reasonable agreement with the measured rate of evaporation. It was concluded that the external mass transfer in the gas phase contributed to the rate control of the flux evaporation. The flux evaporation leads to the change in the flux composition and flux properties, affecting the evaporation rate. 189

218 Chapter 10 - Conclusions and future work 10.1 Conclusions Various types of fluorine-free boron-containing mould fluxes, covering from simple 4- component system, to 5- and 6-component, and further to 8- and 10-component systems, were investigated by measuring their melting properties and viscosity. The flux composition selection was based on thermodynamic phase calculation using FactSage, and made it possible to systematically investigate the effects of CaO/SiO2 ratio and the addition of other oxide components on melting properties and viscosity in different designed flux systems. The stability of some fluorine-free mould fluxes was also investigated. The effects of CaO/SiO2 mass ratio and the concentrations of different components (e.g., B2O3, Na2O, TiO2, MgO) on flux melting properties and viscosity were found to be quite complex. For SiO2-CaO-Al2O3-B2O3 fluxes, increase in CaO/SiO2 increased Th and Tbr first and then decreased both of them, while changing B2O3 content had no significant influence on Th but lowered Tbr. Increase in CaO/SiO2 decreased viscosity first, and a further increase in the CaO/SiO2 ratio had no effect on viscosity; increase in CaO/SiO2 decreased EA in general. Increase in B2O3 concentration decreased both viscosity and activation energy. Results on 5-component SiO2-CaO-Al2O3-B2O3-MxOy (MxOy: 6-9 wt% Na2O, 2-6 wt% TiO2, or wt% MgO) fluxes showed that for Na2O-containing fluxes, change in the CaO/SiO2 ratio from 0.8 to 1.0 decreased Th and Tbr, while further increase in this ratio increased Th and Tbr. Na2O content from 6.2 to 9.1 wt% had a minor influence on Th, but raised Tbr. For TiO2-containing fluxes, the influence of CaO/SiO2 ratio on Th and Tbr 190

219 was complicated with ups and downs with changing CaO/SiO2 ratio. Increase in the TiO2 content from 0 to 4 wt% decreased Th, but enhanced it significantly when TiO2 content increased from 4 to 6 wt%. Change of the TiO2 content from 4 to 6 wt% had no influence on Tbr. For MgO-containing fluxes, Th and Tbr increased with CaO/SiO2 ratio first, and then either decreased or stabilised with further increasing CaO/SiO2 ratio. Change in MgO concentration from 0 to 3.5 wt% reduced Th first but then stabilised it. Tbr decreased with MgO concentration from 2 to 3.5 wt%. For all 5-component flux systems, an increase in the CaO/SiO2 ratio led to the decrease in viscosity η and activation energy EA. Change in Na2O concentration from 6 to 9 wt% had no significant influence on η, but reduced EA. With the increase of TiO2 and MgO content, both η and EA decreased. The results of 6, 8 and 10-component fluxes showed more complicated effects of CaO/SiO2 ratio and individual components on these properties. Values of Th and Tbr of SiO2-CaO-Al2O3-B2O3-Na2O-TiO2/MgO fluxes are generally lower than those of CaO- SiO2-Al2O3-B2O3-TiO2-MgO fluxes, indicating the effect of Na2O on reducing Th and Tbr. Increase in Na2O, TiO2, and MgO had a decreasing effect on Th. With the increase of CaO/SiO2 ratio, Th decreased first and then increased for SiO2-CaO-Al2O3-B2O3- Na2O-TiO2/MgO fluxes; all way decreased for CaO-SiO2-Al2O3-B2O3-TiO2-MgO fluxes; and increased monotonically for SiO2-CaO-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O- (MnO-ZrO2) fluxes. For 6-comonent fluxes, viscosity η and activation energy EA decreased with the increase of the CaO/SiO2 ratio from 0.9 to 1.1, Na2O concentration from 7 to 11 wt%, TiO2 concentration from 2 to 6 wt%, and MgO concentration from 1 to 4 wt%. For more complex multi-component system SiO2-CaO-Al2O3-B2O3-Na2O- 191

220 TiO2-MgO-Li2O-MnO-ZrO2 fluxes, η declined first with the increase of CaO/SiO2 ratio from 0.8 to 1.3 and then increased with further increasing CaO/SiO2 to 1.5. Melting properties and viscosity of fluorine-free fluxes are compared with the commercial fluorine-containing Fluxes C1 and C2. Results showed that all simple flux systems of 4, 5 and 6-component fluxes have higher melting properties than that of Flux C1, and only some of fluxes have close Th to that of Flux C2, for example, Na2Ocontaining 5-component fluxes with CaO/SiO2 ratio of 1.3, and TiO2-containing and MgO-containing 5-component fluxes with CaO/SiO2 ratio of 1.3 and 1.5. Values of viscosity η of all synthetic 5 and 6-component fluxes are higher than those of commercial ones. Na2O-containing 6-component fluxes with CaO/SiO2 ratio less than 1.3 and more complex 8 and 10-component fluxes with CaO/SiO2 ratio less than 1.5 have lower Tbr than those of commercial fluxes. Values of Th for 8- and 10-component fluxes with CaO/SiO2 ratio less than 1.5 are lower than 1050 C, close to that of industrial fluxes. Values of viscosity of these more complex fluorine-free fluxes with CaO/SiO2 ratio less than 1.5 are close to that of Flux C1. Therefore, these fluxes should be good candidates in application for low-carbon steel casting. Raman spectroscopy was used for flux structural analyses for quenched 4 and 5- component fluxes. The results revealed that there are five different silicate units of Q 0, Q 1, Q 2, Q 3, and Q 4 in the range of 750 and 1300 cm -1. Measured NBO/Si demonstrates the depolymerisation of the slag structure with increasing CaO/SiO2 ratio; in general, it had the same tendency as calculated NBO/T. Change in the viscosity and activation energy was correlated with NBO/Si; η and EA decreased with increasing NBO/Si. Unlike Na2O and MgO (both are network modifiers), TiO2 worked as a weak network 192

221 former, evidenced by the formation of two peaks around 705 and 805 cm -1 in Raman spectra. A neural network model was developed to describe the viscosity of CaO-SiO2-Al2O3- B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2 mould fluxes in this work. Viscosity of mould fluxes calculated using this model was in a reasonable agreement with the experimental data and showed better fitting results than all other existing viscosity models. Based on the BP neural network model, the effects of temperature and flux composition on viscosity were predicted based on this model. The stability of the fluorine-free SiO2-CaO-Al2O3-B2O3-Na2O mould fluxes was studied using TGA method. Results showed that the evaporation of fluorine-free mould fluxes was much slower in comparison with industrial fluorine-containing flux. The rate of the flux evaporation depended on the temperature and flux composition. The weight loss as a result of the flux evaporation increased with the increasing temperature. The evaporation of the boracic fluxes without Na2O is marginal, while significant evaporation was observed when the fluxes contained both B2O3 and Na2O, which was attributed to the formation of highly volatile NaBO2. The increase in the rate of the vaporisation slowed down for fluxes with a higher Na2O content of 9.1 wt%. Variation of the CaO/SiO2 ratio, however, did not change the flux evaporation rate. The flux evaporation leads to the change in the flux composition and flux properties, affecting the evaporation rate. The flux evaporation process was analysed by considering chemical reactions and internal and external mass transfers. It was concluded that the external mass transfer from the flux/gas interface to the main gas stream contributed to the rate control of the flux evaporation. 193

222 10.2 Recommendation for further work This research work showed that multi-component fluorine-free fluxes developed are possible candidates to replace conventional fluorine-containing industrial fluxes for low-carbon steel casting. However, further work is necessary to reduce the viscosity of the fluxes by adding some other components to meet the requirements for mediumcarbon steel casting. The BP neural network model developed in this thesis worked well for B2O3-containing fluorine-free mould fluxes. The broader application of this model to different kinds of other B2O3-containing fluxes is recommended. Evaporation of fluorine-free mould fluxes containing both Na2O and B2O3 has been investigated in this thesis. However, some extents of MgO loss during flux preparation indicated a possible evaporation of MgO. Stability of MgO-containing fluorine-free fluxes should be determined for this type of fluxes in future. Also it is suggested to measure the composition of the vapour species when we deal with the evaporation of more complex systems in future work. 194

223 References 1. K. C. Mills, A. B. Fox, Z. Li and R. P. Thackray, Ironmak Steelmak, 2005, vol. 32, pp A. B. Fox, K. C. Mills, D. Lever, C. Bezerra, C. Valadares, I. Unamuno, J. J. Laraudogoitia and J. Gisby, ISIJ Int, 2005, vol. 45, pp A. S. Foundation and A. W. Cramb, The Making, Shaping and Treating of Steel: Casting Volume, 1st ed., AISE Steel Foundation, C. A. M. Pinheiro, I. V. Samarasekera, J. K. Brimacombe and B. N. Walker, Ironmak Steelmak, 2000, vol. 27, pp K. Mills, P. Ramirez-Lopez, P. Lee, B. Santillana, B. Thomas and R. Morales, Ironmak Steelmak, 2014, vol. 41, pp K. C. Mills and A. B. Fox, ISIJ Int, 2003, vol. 43, pp E. Benavidez, L. Santini, M. Valentini and E. Brandaleze, Proc. Mater. Sci., 2012, vol. 1, pp P. Palai, P. P. Sahoo, A. Dey, T. K. Roy and V. V. Mahashabde, Metall. Mater. Trans. B, 2013, vol. 44, pp Z. Elahipanah, M.Sc. Thesis, Royal Institute of Technology, Stockholm, E. Brandaleze, G. Di Gresia, L. Santini, A. Martín and E. Benavidez, Science and Technology of Casting Processes, 1st ed., InTech, 2012, pp B. G. Thomas, Metall. Mater. Trans. B, 2002, vol. 33, pp K. C. Mills: Slag Atlas, 2nd ed., Verlag Stahleisen GmbH, Germany, 1995, pp K. C. Mills, ISIJ Int, 1993, vol. 33, pp K. C. Mills: Slag Atlas, 2nd ed., Verlag Stahleisen GmbH, Germany, 1995, pp J. M. Bockris, J. Mackenzie and J. Kitchener, T Faraday Soc, 1955, vol. 51, pp

224 16. Y. Waseda and J. M. Toguri, The Structure and Properties of Oxide Melts: Application of Basic Science to Metallurgical Processing, 1st ed., World Scientific, K. C. Mills, ISIJ Int, 2016, vol. 56, pp F. Shahbazian, D. Sichen and S. Seetharaman, ISIJ Int, 2002, vol. 42, pp J. H. Park, D. J. Min and H. S. Song, Metall. Mater. Trans. B, 2004, vol. 35, pp B. Mysen and D. Neuville, Geochim Cosmochim Ac, 1995, vol. 59, pp Z. Wang, Q. F. Shu and K. C. Chou, Steel Res Int, 2013, vol. 84, pp J. Y. Park, G. H. Kim, J. B. Kim, S. Park and I. Sohn, Metall. Mater. Trans. B, 2016, vol. 47, pp Y. Kim and K. Morita, ISIJ Int., 2014, vol. 54, pp Z. Wang, Q. Shu and K. Chou, ISIJ Int., 2011, vol. 51, pp Z. Wang, Q. F. Shu and K. C. Chou, High Temp. Mater. Processes, 2013, vol. 32, pp K. Z. Gu, W. L. Wang, L. J. Zhou, F. J. Ma and D. Y. Huang, Metall. Mater. Trans. B, 2012, vol. 43, pp J. Kromhout, Ph.D thesis, Delft University of Technology, Delft, G. H. Wen, S. Sridhar, P. Tang, X. Qi and Y. Q. Liu, ISIJ Int, 2007, vol. 47, pp N. Takahira, M. Hanao and Y. Tsukaguchi, ISIJ Int, 2013, vol. 53, pp J. Duffy and M. Ingram, J Non-Cryst Solids, 1976, vol. 21, pp J. A. Duffy, Geochim Cosmochim Ac, 1993, vol. 57, pp K. C. Mills and S. Sridhar, Ironmak Steelmak, 1999, vol. 26, pp K. C. Mills, Slag Atlas, 2, Verlag Stahleisen GmbH, Germany, 1995, pp K. C. Mills, The Making, Shaping and Treating of Steel: Casting Volume, Pittsburgh, PA, 2003, pp

225 35. J. Moore, R. Phillips and T. Gibbs, Steelmaking Conference Proceedings., 1991, pp S. Sridhar, K. Mills, O. Afrange, H. Lörz and R. Carli, Ironmak Steelmak, 2000, vol. 27, pp K. C. Mills, ISIJ Int, 2016, vol. 56, pp D. Stewart, P. N. Hewitt, L. Peeters, Iron and S. O. C. Steel, 79th Steelmaking Conference Proceedings, Iron & Steel Soc Aime, Warrendale, 1996, pp N. S. Hunter, J. D. Madill, D. J. Scoones, P. N. Hewitt and D. Stewart, Steelmak. Tech. Inter., 1996, pp M. M. Wolf, Rev Metall-Paris, 1994, vol. 91, pp X. Long, S.-p. He, J.-f. Xu, X.-l. Huo and Q. Wang, J Iron and Steel Res Inter., 2012, vol. 19, pp J. Elfsberg and T. Licentiatexamen, Licentiate thesis, Stockholm, Sweden, A. Paul, N. Pradhan, A. K. Ray, P. K. Bhor, S. Mazumdar and J. M. Korath, Scand J Metall, 2000, vol. 29, pp A. Badri, T. T. Natarajan, C. C. Snyder, K. D. Powers, F. J. Mannion, M. Byrne and A. W. Cramb, Metall. Mater. Trans. B, 2005, vol. 36, pp M. Hanao, ISIJ Int, 2013, vol. 53, pp W. L. Wang, K. Z. Gu, L. J. Zhou, F. J. Ma, I. Sohn, D. J. Min, H. Matsuura and F. Tsukihashi, ISIJ Int, 2011, vol. 51, pp J. Cho, H. Shibata, T. Emi and M. Suzuki, ISIJ Int, 1998, vol. 38, pp S. Y. Choi, D. H. Lee, D. W. Shin, S. Y. Choi, J. W. Cho and J. M. Park, J Non- Cryst Solids, 2004, vol. 345, pp J. W. Cho, T. Emi, H. Shibata and M. Suzuki, ISIJ Int, 1998, vol. 38, pp A. Cruz, F. Chavez, A. Romero, E. Palacios and V. Arredondo, J Mater Proc Tech, 2007, vol. 182, pp C. A. Pinheiro, The University of British Columbia, 1997, pp

226 52. S. P. He, Q. Y. Huang, G. X. Zhang, Y. J. Lu and Q. Wang, J Iron Steel Res Int, 2011, vol. 18, pp L. Zhou, W. Wang and K. Zhou, ISIJ Int, 2015, vol. 55, pp L. Zhou, W. Wang, B. Lu, G. Wen and J. Yang, Metals Mater. Inter., 2015, vol. 21, pp J. Romo-Castaneda, A. Cruz-Ramirez, A. Romero-Serrano, M. Vargas-Ramirez and M. Hallen-Lopez, Thermochimica Acta, 2011, vol. 512, pp T. Matsushita, M. Hayashi and S. Seetharaman, Int J Mater Prod Tec, 2005, vol. 22, pp L. J. Zhou, W. L. Wang, J. Wei and B. X. Lu, ISIJ Int, 2013, vol. 53, pp Z. T. Zhang, G. H. Wen and Y. Y. Zhang, Int J. Min. Met. Mater., 2011, vol. 18, pp M. Persson, S. Seetharaman and S. Seetharaman, ISIJ Int, 2007, vol. 47, pp A. I. Zaitsev, A. V. Leites, A. D. Lrtvina and B. M. Mogutnov, Steel Res Int, 1994, vol. 65, pp J. Yang, J. Q. Zhang, Y. Sasaki, O. Ostrovski, C. Zhang, D. X. Cai and Y. Kashiwaya, ISIJ Int, 2016, vol. 56, pp X. J. Dong, H. Y. Sun, X. F. She, Q. G. Xue and J. S. Wang, Ironmak Steelmak, 2013, vol. 41, pp H. Nakada and K. Nagata, ISIJ Int, 2006, vol. 46, pp Q. Wang, Y. Lu, S. He, K. Mills and Z. Li, Ironmak Steelmak, 2011, vol. 38, pp Y. Lu, G. Zhang, M. Jiang, H. Liu and T. Li, J Environ Sci, 2011, vol. 23, pp. S167-S T. W. Zhang, H. M. Wang, G. R. Li and X. Y. Zhang, Adv. Mat. Res., 2012, vol , pp

227 67. H. Wang, T. Zhang, H. Zhu, G. Li, Y. Yan and J. Wang, ISIJ Int, 2011, vol. 51, pp Y. Q. Lu, G. D. Zhang and M. F. Jiang, Fund Chem Eng, Pts 1-3, 2011, vol , pp X. Huang, J. Liao, K. Zheng, H. Hu, F. Wang and Z. Zhang, Ironmak Steelmak, 2014, vol. 41, pp M. C. C. Bezerra, C. A. G. Valadares, I. P. Rocha, J. R. Bolota, M. C. Carboni, I. L. de Mattos Scripnic, C. R. Santos, K. Mills and D. Lever, XXXVII Steelmaking Seminar ABM, Porto Alegre, Brazil, B. X. Lu, W. L. Wang, J. Li, H. Zhao and D. Y. Huang, Metall. Mater. Trans. B, 2013, vol. 44, pp G. Wen, S. Sridhar, P. Tang, X. Qi and Y. Liu, Isij Int, 2007, vol. 47, pp P. Rocabois, J. N. Pontoire, J. Lehmann and H. Gaye, J Non-Cryst Solids, 2001, vol. 282, pp M. Kawamoto, K. Nakajima, T. Kanazawa and K. Nakai, Steelmaking Conference Proceedings., 1992, pp J. M. Camp and C. B. Francis, The making, shaping and treating of steel, ed., Carnegie Steel Company, S. Gupta, R. Gupta, G. Bryant, L. Juniper and T. Wall, Impact of Mineral Impurities in Solid Fuel Combustion, Springer, 2002, pp H. M. Wang, D. Fu, G. R. Li, X. Zhu, Z. Zhao, Y. T. Zhao and H. Zhu, Appl Mech Mater, 2012, vol , pp H. M. Wang, L. L. Yang, H. Zhu and Y. Q. Yan, Adv Mater Proc, Pts 1-3, 2011, vol , pp J. F. Xu, J. Y. Zhang, C. Jie, L. Tang and K. C. Chou, Adv. Mat. Res., 2011, vol , pp W. L. Wang, B. X. Lu and D. Xiao, Metall. Mater. Trans. B, 2016, vol. 47, pp

228 81. H. Zhao, W. Wang, L. Zhou, B. Lu and Y.-B. Kang, Metall. Mater. Trans. B, 2014, vol. 45, pp B. O. Mysen, D. Virgo and F. A. Seifert, Rev Geophys, 1982, vol. 20, pp V. Danek, T. Licko and Z. Panek, Silikaty, 1985, vol. 29, pp B. Mysen and P. Richet, Silicate Glasses and Melts: Properties and Structure, 1st ed., Elsevier, Amsterdam, AE, S. Arrhenius, Biochem J, 1917, vol. 11, pp H. Kim, W. H. Kim, J. H. Park and D. J. Min, Steel Res Int, 2010, vol. 81, pp H. D. Weymann, Kolloid Z Z Polym, 1962, vol. 181, pp. 131-&. 88. W. Brostow, T. Sterzynski and S. Triouleyre, Polymer, 1996, vol. 37, pp H. Vogel, Phys. Z, 1921, vol. 22, pp G. S. Fulcher, Journal of the American Ceramic Society, 1992, vol. 75, pp G. Tammann and W. Hesse, Z Anorg Allg Chem, 1926, vol. 156, pp G. H. Zhang, K. C. Chou and K. Mills, Metall. Mater. Trans. B, 2013, vol. 45, pp S. Ogibayashi, K. Yamaguchi, T. Mukai, T. Takahashi, Y. Mimura, K. Koyama, Y. Nagano and T. Nakano, Nippon Steel Technical Report. Overseas, 1987, pp H. Ito, T. Yanagase, Y. Suginohara and N. Miyazaki, J. Jpn. Inst. Met, 1967, vol. 31, pp H. Kim and I. Sohn, ISIJ Int, 2011, vol. 51, pp K. Y. Ko and J. H. Park, ISIJ Int, 2013, vol. 53, pp M. Persson, M. Görnerup and S. Seetharaman, ISIJ Int, 2007, vol. 47, pp L. Wang, C. Zhang, D. X. Cai, J. Q. Zhang, Y. Sasaki and O. Ostrovski, Metall Mater Trans B,

229 99. S. H. Seok, S. M. Jung, Y. S. Lee and D. J. Min, ISIJ Int, 2007, vol. 47, pp G. H. Zhang, K. C. Chou and J. L. Zhang, Ironmak Steelmak, 2014, vol. 41, pp X. Qi, G. H. Wen and P. Tang, J Iron Steel Res Int, 2010, vol. 17, pp S. Tandon, R. D. Agrawal and M. L. Kapoor, J. Am. Ceram. Soc., 1994, vol. 77, pp S. Sukenaga, N. Saito, K. Kawakami and K. Nakashima, ISIJ Int, 2006, vol. 46, pp J. L. Klug, R. Hagemann, N. C. Heck, A. C. F. Vilela, H. P. Heller and P. R. Scheller, Steel Res Int, 2012, vol. 83, pp N. Saito, N. Hori, K. Nakashima and K. Mori, Metall. Mater. Trans. B, 2003, vol. 34, pp S. J. Nam, Y. B. Kang, S. M. Jung and Y. Sasaki, ISIJ Int, 2013, vol. 53, pp H. Park, J. Y. Park, G. H. Kim and I. Sohn, Steel Res Int, 2012, vol. 83, pp G. H. Zhang, K. C. Chou, Q. G. Xue and K. C. Mills, Metall. Mater. Trans. B, 2012, vol. 43, pp K. Mills, International Conference of Southern African Pyrometallurgy, Cradle of Humankind, South Africa, 2011, pp Z. T. Zhang, G. H. Wen, P. Tang and S. Sridhar, ISIJ Int, 2008, vol. 48, pp B. Jiang, W. Wang, I. Sohn, J. Wei, L. Zhou and B. Lu, Metall. Mater. Trans. B, 2014, vol. 45, pp M. S. Jenkins and M. U. D. o. C. Engineering, Heat Transfer in the Continuous Casting Mould, 1st ed., Monash University,

230 113. R. Brooks, A. Dinsdale and P. Quested, Meas. Sci. Technol., 2005, vol. 16, pp R. E. Aune, M. Hayashi, K. Nakajima and S. Seetharaman, Jom-J Min Met Mat S, 2002, vol. 54, pp I. Sohn and D. J. Min, Steel Res Int, 2012, vol. 83, pp A. Bartels, H. Behrens, F. Holtz, B. C. Schmidt, M. Fechtelkord, J. Knipping, L. Crede, A. Baasner and N. Pukallus, Chem. Geol., 2013, vol. 346, pp R. Ladenberg, Phys, 1907, vol. 22, pp S.-H. Shin, J. W. Cho and S.-H. Kim, J. Am. Ceram. Soc., 2014, vol. 97, pp F. Z. Ji, S. C. Du and S. Seetharaman, Metall. Mater. Trans. B, 1997, vol. 28, pp O. Meyer, Poggendorfs Ann, 1961, vol. 13, pp Z. Morita, T. Iida, M. Kawamoto and A. Mori, Tetsu-to-Hagane(J. Iron Steel Inst. Jpn.), 1984, vol. 70, pp K. Mills, M. Halali, H. Lörz, A. Kinder, R. Promfret and B. Walker, Proceedings of the 5th International Conference of Molten Slags, Fluxes and Salts, Sydney, Australia, A. Dey and S. Riaz, Ironmak Steelmak, 2012, vol. 39, pp W. Kim, Ph.D. thesis, Ecole Polytechnique, Montreal (Canada), B. R. Stanmore and S. Budd, Fuel, 1996, vol. 75, pp W. Y. Kim, X. Q. Yang, L. J. Yan, A. D. Pelton and S. A. Decterov, Calphad, 2011, vol. 35, pp S. Vargas, F. Frandsen and K. Dam-Johansen, Prog Energ Combust, 2001, vol. 27, pp E. W. Washburn, E. E. Libman and G. R. Shelton, The Viscosities and Surface Tensions of the Soda-lime-silica Glasses at High Temperatures: Part I. The Viscosity of Glass at High Temperatures, The University of Illinois, Urbana,

231 129. K. C. Mills, S. Karagadde, P. D. Lee, L. Yuan and F. Shahbazian, ISIJ Int, 2016, vol. 56, pp W. McCauley and D. Apelian, Can. Metall. Quart., 1981, vol. 20, pp K. C. Mills, L. Chapman, A. B. Fox and S. Sridhar, Scand J Metall, 2001, vol. 30, pp L. Zhang and S. Jahanshahi, Metall. Mater. Trans. B, 1998, vol. 29, pp K. Koyama, Y. Nagano, T. Nakano, K. Yamaguchi, S. Ogibayashi and Y. Mimura, T Iron Steel I Jpn, 1984, vol. 24, pp. B251-B T. Iida, H. Sakai, Y. Kita and K. Shigeno, ISIJ Int, 2000, vol. 40, pp. S110-S P. Riboud, Y. Roux, L. Lucas and H. Gaye, Fachberichte Huttenpraxis Metallweiterverarbeitung, 1981, vol. 19, pp G. Urbain, Steel Research, 1987, vol. 58, pp D. Sichen, J. Bygd'en and S. Seetharaman, Metall. Mater. Trans. B, 1994, vol. 25, pp Q. F. Shu and J. Y. Zhang, ISIJ Int, 2006, vol. 46, pp D. Giordano, A. Mangiacapra, M. Potuzak, J. Russell, C. Romano, D. Dingwell and A. Di Muro, Chem. Geol., 2006, vol. 229, pp K. Koyama, Y. Nagano and T. Nakano, Nippon Steel Tech Report, 1987, vol. 324, pp T. Iida, H. Sakai, Y. Kita and K. Murakami, High Temp. Mater. Proc., 2000, vol. 19, pp G. H. Zhang, K. C. Chou and K. Mills, ISIJ Int, 2012, vol. 52, pp J. W. Kim, J. Choi, O. D. Kwon, I. R. Lee, Y. K. Shin and J. S. Park, Proc. 4th International Conference on Molten Slags and Fluxes, Sendai, Japan, 1992, pp G. H. Zhang, K. C. Chou and K. Mills, Metall. Mater. Trans. B, 2013, pp P. L. Lin and A. D. Pelton, MTB, 1979, vol. 10, pp C. T. Moynihan, J. Am. Ceram. Soc., 1993, vol. 76, pp

232 147. K. Mills, A. Fox, L. Chapman and S. Sridhar, Scand. J. Metallurgy, 2001, vol. 30, pp M. Nakamoto, Y. Miyabayashi, L. Holappa and T. Tanaka, ISIJ Int, 2007, vol. 47, pp K. C. Mills, S. Karagadde, P. D. Lee, L. Yuan and F. Shahbazian, ISIJ Int, 2016, vol. 56, pp C.-B. Shi, M.-D. Seo, J.-W. Cho and S.-H. Kim, Metall. Mater. Trans. B, 2014, vol. 45, pp Z. T. Zhang, S. Sridhar and J. W. Cho, ISIJ Int, 2011, vol. 51, pp R. G. C. Beerkens, J. Am. Ceram. Soc., 2001, vol. 84, pp M. Li, T. Utigard and M. Barati, Metall. Mater. Trans. B, 2015, vol. 46, pp V. L. Stolyarova, G. G. Ivanov and S. V. Stolyar, Glass Phys Chem, 2002, vol. 28, pp Y. Kashiwaya and A. W. Cramb, Metall. Mater. Trans. B, 1998, vol. 29, pp Y. Yang, L. Teng and S. Seetharaman, Metall. Mater. Trans. B, 2012, vol. 43, pp M. Asano and Y. Yasue, J Nucl Mater, 1986, vol. 138, pp T. Nagai, M. Miyake, Y. Mitsuda, H. Kimura and M. Maeda, ISIJ Int, 2007, vol. 47, pp M. Asano, T. Kou and Y. Mizutani, J Non-Cryst Solids, 1989, vol. 112, pp L. J. Wang, V. L. Stolyarova, S. I. Lopatin and S. Seetharaman, Rapid Commun Mass Sp, 2009, vol. 23, pp M. Heyrman, C. Chatillon, H. Collas and J. L. Chemin, Rapid Commun Mass Sp, 2004, vol. 18, pp S. Seetharaman, T. Shyrokykh, C. Schröder and P. R. Scheller, Metall. Mater. Trans., 2013, vol. 44, pp

233 163. Y. Kashiwaya, C. E. Cicutti, A. W. Cramb and K. Ishii, ISIJ Int, 1998, vol. 38, pp K. Shimizu, T. Suzuki, I. Jimbo and A. Cramb, Iron Steelmaker, 2001, vol. 28, pp B. Mysen, Eur J Mineral, 2003, vol. 15, pp K. Shimizu, Ph.D thesis, Carnegie Mellon University, Pittsburgh, Q. Shu, Z. Wang, J. L. Klug, K. Chou and P. R. Scheller, Steel Res Int, J. Wei, W. Wang, L. Zhou, D. Huang, H. Zhao and F. Ma, Metall. Mater. Trans. B, 2014, vol. 45, pp S. S. Jung and I. Sohn, Metall. Mater. Trans. B, 2012, vol. 43, pp T. Furukawa, K. E. Fox and W. B. White, J Chem Phys, 1981, vol. 75, pp Y. Iguchi, S. Kashio, T. Goto, Y. Nishina and T. Fuwa, Can. Metall. Quart., 1981, vol. 20, pp T. Yano, N. Kunimine, S. Shibata and M. Yamane, J Non-Cryst Solids, 2003, vol. 321, pp G. Padmaja and P. Kistaiah, J. Phys. Chem. A, 2009, vol. 113, pp B. N. Meera and J. Ramakrishna, J Non-Cryst Solids, 1993, vol. 159, pp Z. Wang, Q. F. Shu and K. C. Chou, ISIJ Int, 2011, vol. 51, pp J. Kline, M. Tangstad and G. Tranell, Metall. Mater. Trans. B, 2015, vol. 46, pp Y. Q. Sun and Z. T. Zhang, Metall. Mater. Trans. B, 2015, vol. 46, pp J. H. Park, ISIJ Int, 2012, vol. 52, pp X. H. Huang, J. L. Liao, K. Zheng, H. H. Hu, F. M. Wang and Z. T. Zhang, Ironmak Steelmak, 2014, vol. 41, pp B. Darwent, Bond Dissociation Energies in Simple Molecules, The Catholic University of America, Washington,

234 181. L. S. Du and J. F. Stebbins, J. Non-Cryst. Solids, 2003, vol. 315, pp V. P. Klyuev and B. Z. Pevzner, Glass Phys. Chem, 2002, vol. 28, pp A. A. Akberdin, G. M. Kireeva and I. A. Medvedovskaya, Russ Metall+, 1986, pp Y. Sun and Z. Zhang, Metall. Mater. Trans. B, 2015, vol. 46, pp K. Zheng, Z. Zhang, L. Liu and X. Wang, Metall. Mater. Trans. B, 2014, vol. 45, pp B. O. Mysen, D. Virgo and C. M. Scarfe, Am. Mineral, 1980, vol. 65, pp K. Zheng, Z. Zhang, L. Liu and X. Wang, Metall. Mater. Trans. B, 2014, vol. 45, pp B. O. Mysen, F. J. Ryerson and D. Virgo, Am Mineral, 1980, vol. 65, pp T. Mukongo, P. C. Pistorius and A. M. Garbers-Craig, Ironmak Steelmak, 2004, vol. 31, pp J. Y. Park, G. H. Kim, J. B. Kim, S. Park and I. Sohn, Metall. Mater. Trans. B, 2016, vol. 47, pp M. Hanao, M. Kawamoto, T. Tanaka and M. Nakamoto, ISIJ Int, 2006, vol. 46, pp M. A. Duchesne, A. M. Bronsch, R. W. Hughes and P. J. Masset, Fuel, 2013, vol. 114, pp D. E. Rumelhart, J. L. McClelland and P. R. Group, Parallel distributed processing, volume 1: Explorations in the Microstructure of Cognition: Foundations, 1st ed., MIT Press, Cambridge, MA, T. P. Vogl, J. Mangis, A. Rigler, W. Zink and D. Alkon, Biological cybernetics, 1988, vol. 59, pp V. Stolyarova, Open Thermodynamics Journal, 2013, vol. 7, pp E. T. Turkdogan, Physical Chemistry of High Temperature Technology,

235 197. J. Szekely and N. J. Themelis, Rate Phenomena in Process Metallurgy, 1st ed., Wiley-Interscience, New York, NY, R. B. Bird, W. E. Stewart and E. N. Lightfoot, Transport Phenomena, ed., John Wiley & Sons, New York, J. D. Mackenzie, T Faraday Soc, 1956, vol. 52, pp G. Kolykov, Stroenie stekla, Proceedings of the Conference on Glass Structure, pp

236 Appendix I Compositions, properties and equilibrium phases of two industrial fluorine-containing mould fluxes All fluorine-free mould fluxes in this work were investigated by comparing their physiochemical properties with the commercial fluorine-containing industrial fluxes, e.g. melting properties and viscosity. Here, two typical commercial fluorine-containing fluxes for low-carbon (Flux C1) and medium-carbon steel (Flux C2) were selected for comparison purpose. The compositions and physical chemical properties of these two fluxes are listed in Table A-1 and Table A-2, respectively. Flux C1 with a lower CaO/SiO2 ratio of 0.9 has a lower Th (1050 C) but a higher η (0.18 Pa s) at 1300 C than those of Flux C2 with a higher CaO/SiO2 ratio of 1.4 (shown in Table A-2). Table A-1 Chemical compositions of commercial fluorine-containing fluxes (wt%). Flux No. CaO/SiO2 CaO SiO2 Al2O3 Na2O MgO Li2O MnO CaF2 C C Table A-2 Melting properties and viscosity of commercial fluorine-containing mould fluxes. Flux No. Th, C η at 1300 C, Pa s C C

237 Phase equilibria of the two commercial fluorine-containing fluxes are shown in Figure A-1. Calculated Tliq, Tsol and dominant phases in the temperature range Tsol-Tliq derived from this figure are listed in Table A-3. Tliq of both fluxes are over 1400 C, indicating that the calculation of multicomponent fluxes could be erroneous. With the decrease of temperature, crystallisation of Flux C1 starts with the precipitation of CaF2, cuspidine (Ca4Si2F2O7) forms at the main crystallisation phase subsequently, and after that merwinite (Ca3MgSi2O8) and combeite (Na2Ca2Si3O9) forms (Figure A-1a). The equilibrium phase of Flux C2 is quite different from Flux C1. The initial crystallisation phases include Ca2SiO4 and CaF2, and then merwinite forms slightly prior to cuspidine (Figure A-1b). The amount of cuspidine is small in comparison with Flux C1. Figure A-1 Calculated equilibrium phase fractions of commercial fluorine-containing (a) Flux C1 and (b) Flux C2 using FactSage. 209