Carbothermal Synthesis of Silicon Nitride

Xiaohan Wan

A thesis in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Materials Science and Engineering Faculty of Science

March 2013

THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet

Surname or Family name: Wan First name: Xiaohan Other name/s: Abbreviation for degree as given in the University calendar: PhD School: Materials Science and Engineering Faculty: Science Title: Carbothermal synthesis of silicon nitride

Abstract 350 words maximum

Carbothermal synthesis of silicon nitride Si3N4 followed by decomposition of Si3N4 is a novel approach to production of solar-grade silicon. The aim of the project was to study reduction/nitridation of silica under different conditions and to establish mechanism of silicon nitride formation. Carbothermal reduction of quartz and amorphous silica was investigated in a fixed bed reactor at 1300-1650 °C in nitrogen at 1-11 atm pressure and in hydrogen-nitrogen mixtures at atmospheric pressure. Samples were prepared from silica-graphite mixtures in the form of pellets. Carbon monoxide evolution in the reduction process was monitored using an infrared sensor; oxygen, nitrogen and carbon contents in reduced samples were determined by LECO analyses. Phases formed in the reduction process were analysed by XRD. Silica was reduced to silicon nitride and silicon carbide; their ratio was dependent on reduction time, temperature and nitrogen pressure. Reduction products also included SiO gas which was removed from the pellet with the flowing gas. In the experiments, reduction of silica started below 1300 °C; the reduction rate increased with increasing temperature. Silicon carbide was the major reduction product at the early stage of reduction; the fraction of silicon nitride increased with increasing reaction time. Maximum silicon nitride to carbide ratio (Si3N4/SiC) in the reduction of silica in nitrogen at atmospheric pressure was observed at 1450 °C. Further increase in temperature decreased Si3N4/SiC ratio. Increase in hydrogen content to 10 vol% favoured SiO2 reduction. Further increase in H2 content led to decreased N2 partial pressure, which had a negative effect on the nitridation process. Elevated nitrogen pressure increased silicon nitride yield and stability.When nitrogen pressure was 11 atm, maximum Si3N4/SiC ratio was observed at 1550-1600 °C. Increasing nitrogen pressure increased reduction and nitridation rates and suppressed SiO loss under otherwise the same conditions. Synthesis of silicon nitride proceeded through silicon carbide. In the beginning of the reduction/nitridation process, silica was predominantly reduced to silicon carbide which was converted to silicon nitride. This mechanism of silicon nitride synthesis was supported by kinetic modeling of reduction/nitridation process. Feasibility of production of solar silicon was demonstrated in a study of decomposition of Si3N4 The decomposition rate increased with increased temperature.

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 1968. 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).

…………………………………………………… ……………………………………..…… ……….……………………... Signature Witness 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:

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

ACKNOWLEDEMENTS

I would like to thank Professor Oleg Ostrovski and Dr Guangqing Zhang for their quality supervision. I am appreciative of their constant encouragement and generous support. The comprehensive advices on various aspects of the project and the critical reading of the proof are acknowledged. The scholarship provided by them is one of the important supports to the completion of this project.

It is grateful that Associate Professor Hal Aral supports on sample characterizations. His valuable advices on the project conductions and paper preparation are appreciative.

Technical support from Mr John W. Sharp is one of the critical factors for the completion of the project. His creative and efficient work outcomes overcome lots of problems in the labs. Without his help it is impossible to accomplish the complex experimental works.

The assistance of Dr C. H. Kong on the electron microscope analysis and paper revision, Dr Yu Wang on the X-ray diffraction analysis and the school staff for technical and administration supports are acknowledged.

Thanks are due to Dr Xing Xing and Mr Le Yu for providing plenty of supports on my research life as friends and colleagues.

Finally, my special acknowledgements are for all of my family members for their disinterested and constant support and encouragement.

PUBLICATIONS ORIGINATED FROM THIS PROJECT

(1) Xiaohan Wan, Guangqing Zhang, Hal Aral and Oleg Ostrovski (2011): “Reduction Mechanism of Carbothermal Synthesis of Silicon Nitride” . High Temperature Processing Symposium

(2) Xiaohan Wan, Guangqing Zhang, O. Ostrovski and Hal Aral (2013): “Carbothermal Reduction of Silica in Nitrogen and Nitrogen-Hydrogen Mixture”. INFACON XIII Congress

III

ABSTRACT

Carbothermal synthesis of silicon nitride (Si3N4) followed by its decomposition is a novel approach to production of solar-grade silicon. The aim of this project was to study reduction/nitridation of silica (SiO2) under different conditions and to establish mechanism of Si3N4 formation.

Carbothermal reduction of crystallised and amorphous SiO2 was investigated in a fixed bed reactor at 1300-1650 °C in nitrogen at 1-11 atm pressure and in hydrogen-nitrogen mixtures at atmospheric pressure.

Samples were prepared from silica-graphite mixtures in the form of pellets. Carbon monoxide evolution in the reduction process was monitored using an infrared sensor; oxygen, nitrogen and carbon contents in reduced samples were determined by LECO analyses. Phases formed in the reduction process were analysed by XRD. Silica was reduced to Si3N4 and silicon carbide (SiC); their ratio was dependent on reduction time, temperature and nitrogen pressure. Reduction products also included SiO gas which was removed from the pellet with the flowing gas.

In the experiments, reduction of SiO2 started below 1300 °C; the reduction rate increased with increasing temperature. SiC was the major reduction product at the early stage of reduction; the fraction of Si3N4 increased with increasing reaction time.

Maximum Si3N4/SiC ratio in the reduction of SiO2 at atmospheric pressure was observed at 1450 °C. Further increasing temperature decreased Si3N4/SiC ratio.

Addition of hydrogen into nitrogen promoted silica conversion. Addition of 5 vol% of

H2 significantly increased the rate of reduction of silica. The effect of hydrogen on the kinetics of silica reduction was attributed to formation of methane (CH4) by reacting with carbon, which transferred carbon from graphite particles to the silica particles.

The maximum Si3N4 to SiC ratio was obtained with addition of 10 vol% H2 at 1450 °C after 12-hour reduction. Higher H2 addition led to lower N2 partial pressure with a negative effect on Si3N4 formation.

Increasing nitrogen pressure increased reduction and nitridation rates, and the stability of nitride at higher temperatures. The maximum Si3N4/SiC ratio was observed at 1550-

1600 °C under a N2 pressure of 11 atm. 71.4% of silicon was converted to Si3N4; the rest was mostly SiC with trace residual SiO2 after 1 hour reaction.

Increasing carbon to silica molar ration increased the rate of reduction of silica. With stoichiometric amount of carbon, the reduction was very slow. When C/SiO2 was equal to 4.5, the reduction completed in about 300-minute. Addition of silicon nitride to the graphite-silica mixture promoted reduction of silica and formation of silicon nitride. This indicates that nucleation of silicon nitride is one of the controlling factors in the reaction process.

The reduction rate of silica was slightly increased by increasing gas flow rate to a peak, and decreased with further increase in gas flow rare. It was due to the CO, played the key role on SiO2 reduction, loss by higher gas flow rate. There was no obvious effect on

SiO2 reduction rate by the type of silica used.

Synthesis of Si3N4 proceeded through SiC. In the beginning of the reduction/nitridation process, SiO2 was predominantly reduced to SiC which was further converted to Si3N4. TEM analysis of a silicon carbide sample subjected to partial nitridation showed that silicon diffused out of silicon carbide lattice onto the surface and reacted with nitrogen to form silicon nitride.

This mechanism of Si3N4 synthesis was supported by kinetic modelling of reduction/nitridation process. The reaction model was constructed considering simultaneous conversion of silica to silicon carbide and nitride and further conversion of carbide to nitride. The results showed that silica was first reduced to silicon carbide which was further converted into nitride. Direct conversion of silica into silicon nitride, was much slower than conversion into carbide.

Feasibility of production of silicon was demonstrated in a study of decomposition of

Si3N4. The decomposition rate was higher in hydrogen than in argon, and increased with increasing temperature.

Decomposition of silicon nitride was studied at 1500 °C and 1600 °C in hydrogen and argon. Temperature had a significant effect on the decomposition kinetics. The V decomposition rate in hydrogen was much faster than in argon, which was attributed to the difference between the diffusivity of nitrogen in these two gases.

It was observed that the decomposed silicon nitride had a non-uniform composition; residual Si3N4 particles were trapped in the silicon product which could affect removal of nitrogen from the elemental silicon matrix.

TABLE OF CONTENTS

ORIGINALITY STATEMENT I ACKNOWLEDGEMENTS II PUBLICATIONS ORIGINATED FROM THIS PROJECT III ABSTRACT IV LIST OF FIGURES X LIST OF TABLES XV

CHAPTER 1 INTRODUCTION 1

CHAPTER 2 LITERATURE REVEW 5 2.1 Silicon nitride synthesis 5 2.2 Carbothermal synthesis of silicon nitride 7 2.2.1 Thermodynamics of carbothermal reduction/nitridation of silica 9 2.2.2 Effect of operation parameters on carbothermal reduction/nitridation of silica 12 2.2.2.1 Temperature 12 2.2.2.2 Gas atmosphere 17 2.2.2.3 Partial pressure of nitrogen 20 2.2.2.4 Gas flow rate 21 2.2.2.5 Carbon to silica ratio 24 2.2.2.6 Particle size and specific surface area 26 2.2.2.7 Seeding 29 2.2.3 Reaction mechanism and kinetic models 34 2.2.4 Summary 41 2.3 Objectives of the project 43

CHAPTER 3 EXPERIMENTAL 44 3.1 Materials 44 3.1.1 Silica 44 3.1.2 Graphite 45

VII

3.1.3 Silicon nitride 45 3.1.4 Silicon carbide 45 3.1.5 Gases 45 3.2 Experimental setup 46 3.2.1 Experimental furnace and reactor 46 3.2.2 Gas system 47 3.2.3 Off-gas composition analysis 49 3.3 Experimental procedures 49 3.3.1 Sample preparation 49 3.3.2 Reduction experimental procedure 50 3.3.3 Synthesis of silicon nitride under elevated nitrogen partial pressure 51 3.3.4 Decomposition of silicon nitride 52 3.4 Sample characterization 52 3.4.1 X-ray diffraction analysis 52 3.4.2 LECO analysis 53 3.4.3 SEM analysis 54 3.4.4 EDS analysis 55 3.4.5 FIB-TEM analysis 55 3.5 Calculation of extent of reactions 55 3.5.1 Calculation of the extent of silica reduction 56 3.5.2 Calculation of the extent of nitridation 59 3.5.3 Calculation of the extent of carburization 60 3.5.4 Calculation of the extent of silicon nitride decomposition 61 3.6 Error analysis 62

CHAPTER 4 EXPERIMENTAL RESULTS 63 4.1 Carbothermal synthesis of silicon nitride 63 4.1.1 Effect of temperature 63 4.1.2 Effect of hydrogen addition 68 4.1.3 Effect of gas flow rate 73 4.1.4 Effect of carbon to silica ratio 75 4.1.5 Effect of type of silica 78 4.1.6 Effect of silicon nitride addition 84

VIII

4.1.7 Effect of nitrogen pressure 87 4.2 Decomposition of silicon nitride 92 4.2.1 Effect of temperature on the decomposition of silicon nitride 92 4.2.2 Silicon nitride decomposition in different gas atmospheres 94

CHAPTER 5 DICSUSSION 102 5.1 Thermodynamic analysis of carbothermal synthesis of silicon nitride 102 5.2 Phase development in silicon nitride synthesis 105 5.3 Effect of gas atmosphere and flow rate on reduction/nitridation process 109

5.4 Effect of C/SiO2 ratio and Si3N4 seeding on reduction/nitridation process 114 5.5 Mechanism of carbothermal synthesis of silicon nitride 115 5.6 Kinetic modelling of synthesis of silicon nitride 123

CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER WORK 134 6.1 Conclusions 134 6.1.1 Carbothermal synthesis of silicon nitride 134 6.1.2 Decomposition of silicon nitride 136 6.2 Recommendations for further work 136

REFERENCES 138

LIST OF FIGURES

Figure Title Page

2-1 Standard Gibbs free energy changes for Reactions (2.2)-(2.6) 10

2-2 Equilibrium CO partial pressure for Reaction (2.1) 11

2-3 Equilibrium SiO partial pressure for Reaction (2.2) 11

2-4 Equilibrium CO2 partial pressure for Reaction (2.4) 12

2-5 Effect of reaction temperatures on the nitridation of sample 13

2-6 Diagram showing (-) typical temperature and (+) values as functions of reaction time 14

2-7 SEM micrographs of powder morphology as a function of reaction temperature 15

2-8 Si3N4 formed in the systems SiO2-C-NH3 and SiO2-C-N2 17

2-9 Change in the standard Gibbs free energy for, A: Reaction (2.15); B: Reaction (2.1); C: Reaction (2.18) 19

2-10 Silicon nitride synthesis in 100% N2 20

2-11 Silicon nitride synthesis in the 95 vol% N2 and 5 vol% H2 gas mixture 20

2-12 Effect of gas flow rate on SiO2 conversion 23

2-13 Effect of gas flow rate on the nitridation extent at 1673 K (1400°C) for a time of 7 ks in 95% nitrogen – 5% hydrogen 23

2-14 Effect of C/Silica ratio on nitrogen content in reduction/nitridation of silica at 1673 K (1400°C) for 5 h in nitrogen gas (2000 mL/min) 24

2-15 SEM micrographs of samples after reduction/nitridation with different

C/SiO2 molar ratio 26

2-16 Effect of grain size of SiO2 (a) and C (b) on conversion of silica 27 X

Figure Title Page

2-17 Weight percent Si3N4 vs. surface area of silica 28

2-18 Effect of pellet forming pressure on nitridation of sample 29

2-19 Nitrogen content of synthesized powders vs. Si3N4/SiO2 ratio in the

system SiO2-C-Si3N4 30

2-20 Effect of various seeding additives on the kinetics of carbothermal reduction/nitridation of silica 31

2-21 Effect of seed addition on silica reduction expressed in vol% of total amount of CO evolved 32

2-22 Effect of precursor “seed”/SiO2 ratio on α-Si3N4 crystallite mean number

equivalent circle diameter ( ) 32

2-23 Effect of “seed”/SiO2 ratio on synthesized Si3N4 morphology 33

2-24 Effect of “seed”/SiO2 ratio and excess carbon content on synthesized

product α-Si3N4 surface area 34

2-25 Si3N4 formation vs. time at 1673 K (1400°C) for samples with C/SiO2 ratios 3.5, 6.2 and 16 37

2-26 Arrhenius plot showing temperature dependence of the initial nitridation

rate of SiO2 38

2-27 Experimental data plotted according to shrinking core model for chemical reaction control 39

2-28 Nucleation-growth model for reaction kinetics 40

2-29 Schematic representation of the diffusion mechanism of s single particle 41

3-1 Schematic diagram of the experimental set-up 47

3-2 Schematic diagram of gas system 48

3-3 Calibration of mass flow controllers 48

Figure Title Page

4-1 Temperature-programmed reduction of fumed silica in N2 at atmospheric pressure 63

4-2 Effect of temperature on the reduction of fumed silica in 10 vol% H2 –

90 vol% N2 mixture at atmospheric pressure 64

4-3 Effect of temperature on the conversion of fumed SiO2 in 10 vol% H2 –

90 vol% N2 mixture at atmospheric pressure 65

4-4 XRD patterns of samples reduced in 10 vol% H2 – 90 vol% N2 mixture for 720 minutes 66

4-5 Effect of hydrogen addition in nitrogen on the conversion of fumed SiO2 at 1723 K (1450 °C) 69

4-6 XRD patterns of samples reduced at 1723 K (1450 °C) in the H2 – N2 gas mixture with different hydrogen content for 300 minutes 71

4-7 Effect of gas flow rate on fumed silica reduction in N2 at atmospheric pressure 74

4-8 The peak evolution rate of CO vs. gas flow rate 75

4-9 Effect of carbon to silica ratio on the reduction of fumed SiO2 at 1723 K

(1450°C) in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure 76

4-10 XRD patterns of samples with various C to SiO2 ratio reduced at 1723 K

(1450°C) in 10 vol% H2 – 90 vol% N2 gas mixture at 1 L/min for 120 minutes 77

4-11 Effect of the type of silica on the reduction of SiO2 at 1723 K (1450°C)

in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure 79

4-12 XRD patterns of the samples with different SiO2 types reduced at 1723 K

(1450°C) in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure for 360 minutes 80

XII

Figure Title Page

4-13 XRD patterns of the crystallised silica reduced in pure N2 for 120 minutes 83

4-14 Effct of Si3N4 addition on the reduction of fumed SiO2 at 1723 K

(1450°C) in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure 84

4-15 XRD patterns of the reacted samples with and without addition of silicon nitride 85

4-16 Effect of Si3N4 seeding on the composition of reduction product in carbothermal synthesis of silicon nitride conducted at 1723 K (1450°C)

in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure 86

4-17 XRD patterns of fumed silica reduced in pure N2 at 1100 kPa for 60 minutes 88

4-18 XRD patterns of the samples reduced 1873 K (1600°C) in nitrogen at 700-1100 kPa pressure for 60 minutes 90

4-19 XRD patterns of Si3N4 decomposed in Ar for 240 minutes 93

4-20 Comparison of the XRD patterns of Si3N4 decomposed in Ar and H2 for 240 minutes 95

4-21 The morphology of Si3N4 decomposed in different gas atmospheres at 1873 K (1600°C) 97

4-22 An image of a Si3N4 sample (a) decomposed in H2 (1 L/min) at 1873 K (1600°C) for 240 minutes and EDS spectra at point 26 (b) and point 29 (c) 98

4-23 EDS mapping analysis of relatively small particles in the Si3N4 sample

decomposed in H2 at 1873 K (1600°C) for 240 minutes 100

4-24 EDS mapping analysis of a large particle in the Si3N4 sample decomposed

in H2 at 1873 K (1600°C) for 240 minutes 101

5-1 Equilibrium CO partial pressures vs temperature in Reactions of formation of silicon nitride (Reaction (2.1)) and silicon carbide

XIII

Figure Title Page

(Reaction (2.7)) 103

5-2 XRD patterns of fumed silica reduced in the 10 vol% H2 – 90 vol% N2 gas mixture for various reaction times at 1723 K (1450°C) 107

5-3 Silicon in SiO2, Si3N4, and SiC in reduction/nitridation of fumed silica

in the 10 vol% H2 – 90 vol% N2 gas mixture at 1723 K (1450°C) 108

5-4 Equilibrium nitrogen partial pressure for conversion of SiC to Si3N4 by Reaction (5.7) calculated using data from Chase (1998) 112

5-5 Equilibrium nitrogen partial pressure for SiC to Si3N4 conversion by Reaction (5.7) calculated using data from Binnewies and Milke (2002) 113

5-6 BSE images and EDS element analyses of a sample after 60 minute reduction/nitridation 116

5-7 BSE images and EDS element analyses of a sample after 240 minute reduction/nitridation 117

5-8 XRD patterns of samples obtained in conversion of SiC to Si3N4 118

5-9 SEM/EDS analyses of commercial silicon carbide subjected to nitridation 119

5-10 TEM observation of a partially nitridated SiC particle 121

5-11 Element mapping of the cross section shown in Figure 5-10 122

5-12 Routes of Si3N4 formation in kinetic modelling 124

5-13 Calculated fractions of silicon in different compounds in comparison with experimental data at 1723 K (1450°C) 130

5-14 Calculated fractions of silicon in different compounds in comparison with experimental data obtained at 1600°C in nitrogen at 1100 kPa 132

XIV

LIST OF TABLES

Table Title Page

1-1 Representative impurity levels in MG-Si and SG-Si 2

2-1 Typical properties of commercial Si3N4 powder for structural ceramics 7

2-2 Effect of on the product of reaction with SiO2+C at 1773 K (1500°C) 21

3-1 Specific surface area of raw materials 45

4-1 Effect of temperature on the composition of reduction product in carbothermal synthesis of silicon nitride 67

4-2 Effect of H2 addition on the composition of reduction products in carbothermal synthesis of silicon nitride 72

4-3 Effect of C to SiO2 ratio on the composition of reduction product in carbothermal synthesis of silicon nitride 78

4-4 Elemental composition of reduced samples weight loss, extent of reduction and yield of silicon nitride and silicon carbide in reduction of different types of silica 81

4-5 Elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide in carbothermal synthesis of silicon nitride in nitrogen at 1100 kPa 89

4-6 Elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide in synthesis of silicon nitride in nitrogen at 700-1100 kPa pressure 91

4-7 Effect of temperature on Si3N4 decomposition 94

4-8 LECO nitrogen analysis and extent of decomposition of silicon nitride in different gas atmosphere 96

Table Title Page

5-1 Equilibrium CO2 and SiO partial pressures for Reaction (2.3) at various reaction temperatures calculated using experimental CO concentrations in the off gas 105

5-2 Experimental data obtained in reduction/nitridation experiments at

1723 K (1450°C) in 10 vol% H2-90 vol% N2 gas mixture 129

5-3 Calculated parameters of the kinetic model of carbothermal synthesis of silicon nitride 129

5-4 Experimental data obtained in reduction/nitridation experiments at 1873 K (1600°C) in nitrogen atmosphere at 1100 kPa 131

5-5 Calculated parameters of the kinetic model of carbothermal synthesis of silicon nitride at 1873 K (1600°C) in nitrogen at 1100 kPa 131

XVI

CHAPTER 1 INTRODUCTION

Solar energy has been developed as a promising renewable energy under consideration concerning both limited energy sources and emission of greenhouse gases. The photovoltaic conversion of solar energy directly into electricity, based on silicon solar cells, is believed to be a most competitive way for clean energy production. However, solar grade silicon purity of 99.9999% must be achieved, which can be produced in several technologies with high-cost purification processes. Further broad expansion of photovoltaic generation of electricity is impeded by high cost of silicon production (Huang and Guo, 2007). A significant research effort world-wide is directed towards development of low-cost solar silicon production. This project will study conversion of silica to silicon nitride which can be decomposed with formation of high-purity silicon.

Silicon is categorized into different grades according to its applications. Metallurgical silicon (MG-Si) is commercially produced by carbothermal reduction of silicon dioxide in submerged arc furnace at temperatures 2100-2300ºC (Ceccaroli and Lohne, 2003). Normally metallurgical grade silicon is at least 98% pure; high quality metallurgical silicon has a purity of ~99.5%. Electronic silicon (EG-Si) is produced by chemical purification of metallurgical silicon. In the Siemens Process, metallurgical silicon is reacted with HCl to form trichlorosilane (SiHCl3), which is purified by distillation and then decomposed at temperature over 1000ºC to allow silicon to deposit on a pure silicon rod to form ingots (chemical vapor deposition). Silicon is then crushed and further purified to obtain 99.99999% (7N) pure silicon (Ceccaroli and Lohne, 2003). The process is complex, slow and costly, and the resulting silicon is purer than needed the photovoltaic industry. The purity requirement of solar grade silicon (SG-Si) compared with metallurgical silicon is presented in Table 1-1 (Murray et al., 2006).

The electronic silicon, rejected and recycled from electronic silicon production, is collected and used to manufacture photovoltaic solar cells with cost of about US$50/kg.

Table 1-1 Representative impurity levels in MG-Si and SG-Si Concentration (ppmw) Impurity MG-Si SG-Si B 40 0.2 Cr 50 3 Cu 40 6 Fe 2000 20 P 20 0.2 Ti 200 0.4

Substantial efforts have been made in recent years in developing technologies for low cost solar silicon production (Woditsch and Koch, 2002). Solar grade silicon can be produced in three main routes.

1) Direct production of high purity silicon by carbothermal reduction using high purity silica and carbon (Aulich et al., 1982; Yashiyagawa et al., 1985; Rustioni et al., 1985). Although it is possible to produce pure silicon in a laboratory (Schulze et al., 1984; Strake et al., 1988; Sakaguchi et al., 1989; Aratani at al., 1991; Ishizaki et al., 1992), there are problems of contamination by furnace handling, tapping, lining, and other sources (Rustioni et al., 1985). The products still need complex refining to achieve the solar silicon purity (Nepomnyaschikh et al., 2004; Abdyukhanov, 2001; Nepomnyaschikh et al., 2002; Karabanov, 2002).

2) Conversion of metallurgical silicon into silane or halosilane compounds (with chlorine, fluorine or bromine) which are processed into pure silicon. These processes are derived from Siemens Process further for electronic silicon production, with modifications to reduce the production costs.

3) Purification of metallurgical silicon to solar grade silicon by melting and then directional solidification. This method is considered to be the most cost efficient for removing metal impurities such as Fe, Ti and Cu; however, P and B are not removed. In the NEDO purification process (Yuge et al., 2001), phosphorous is removed from molten metallurgical silicon by electron beam melting under vacuum, then silicon is

subjected to the first directional solidification, plasma melting to remove boron and carbon using a reactive gas such as hydrogen-water vapor mixture, second directional solidification, then crushing and cleaning.

Further intensive expansion of the photovoltaic industry requires solar silicon at US$6- 12/kg (Murray et al., 2006). Although purification of metallurgical silicon has been considered the most promising process, cost reduction is limited by complexity of removal of B and P which proceeds through their diffusion from bulk silicon to the liquid surface and removal from the surface. These slow processes cannot be avoided in existing technologies.

The process investigated in this project includes conversion of silica to silicon nitride and decomposition of silicon nitride with formation of silicon. This process targets minimizing or avoiding P and B in silicon metal with significant economy and technological benefits.

Investigations on silicon nitride synthesis will be primarily carried out as the major part of this thesis, followed by examination on decomposition of silicon nitride forming elemental silicon.

A German patent of 1896 described the production of Si3N4 by the carbothermal reduction of SiO2 (Melner, 1896). Silicon nitride was developed in a search for fully dense, high strength and high toughness materials in the 1960s and 1970s (Riley, 2000). An original driver for its development was to replace metals with ceramics in advanced turbine and reciprocating engines to give higher operating temperatures and efficiencies. Typically, silicon nitride has been extensively studied as structural ceramic material for high temperature application due to its excellent high temperature properties: high strength and hardness, resistance to creep, oxidation, and thermal shock (Gleiter, 1989; Jack, 1979; Mocolm, 1983). Currently there are several routes available to produce silicon nitride. Commercial silicon nitride is manufactured by three main approaches: direct nitridation, Si-diimide processing and carbothermal reduction/nitridation. Besides these processes, synthesis by vapor, laser and plasma reaction are under investigation or at a pilot stage (Hofmann et al., 1993). Details of these three main routes are presented in Chapter 2. Decomposition of silicon nitride was primarily studied by Batha and

Whitney (1973). Study of the use of silicon nitride to produce solar grade silicon has been reported by Murray et al. (2006).

Solar silicon production proposed in this project was based on the usage of high purity raw materials and efficient solid state conversion of silicon dioxide to silicon nitride with low level of impurities. However, behaviour of impurities in the reduction/nitridation reaction was beyond the scope of this research.

The project studied carbothermal synthesis of silicon nitride from crystal and amorphous silica; effects of operation parameters on synthesis of Si3N4 including temperature, gas composition, seed addition. Data obtained in the experimental study of conversion of silica to silicon nitride were used for kinetic modeling and further understanding of reaction mechanisms. The project also examined feasibility of production of silicon by decomposition of silicon nitride.

CHAPTER 2 LITERATURE REVIEW

2.1 Silicon nitride synthesis

Commercial silicon nitride is synthesized in three routes: direct Si nitridation process, Si-diimide process, and carbothermal reduction/nitridation of silicon dioxide.

(1) Direct Si-nitridation

The overall reaction of Si-nitridation is presented as

( ) ( ) ( ) (2.01)

The reaction is exothermic with a value of -733 kJ/mol and the standard Gibbs free energy change of -205 kJ/mol at 1643 K (1370° C). The reaction proceeds slowly because of the build-up of a protective product layer on the exterior of the silicon. The silicon is not converted to nitride readily below its melting point, 1683 K (1410°C), hence a two stage reaction schedule is normally employed. A typical reaction process might include nitridation for 24 to 50 hours at 1623 K (1350°C) followed by 10 to 24 hours nitridation at 1723 K (1450° C) (Parr and May, 1967). The final product formed contains 75 % of α-phase and 25% β-phase. A phase transformation from α- to β-phase occurs via a solution-precipitation mechanism that results in a fibrous microstructure (Durham et al., 1988). Another problem associated with the nitridation of silicon is the presence of contaminants in the final product. Commercial silicon of metallurgical grade is usually on the order of 98% pure and includes various metallic impurities (Durham et al., 1988).

(2) Si-diimide

Si-diimide process was developed as the long-known but little-used silicon amide approach. Application of this route depends on the availability of low cost silicon tetrachloride (SiCl4).

( ) ( ) ( ) ( ) (2.02)

Silicon diimide (Si(NH)2) decomposes to amorphous silicon nitride (Si3N4), which is converted to the α – Si3N4 by heating in nitrogen at 1673 K (1400°C)- 1773 K (1500°C).

( ) ( ) ( ) ( ) (2.03)

The Si-diimide process was the second-most-important route for commercial production (Riley, 2000). The use of ammonia is favored both thermodynamically and kinetically and results in the production of a very fine amorphous powder. The quality of the Si3N4 fine powder is better than that produced by the direct nitridation of silicon, however corrosion, impurities from raw materials in the system SiCl4-NH3, high cost, and toxicity of diimide are disadvantages.

(3) Carbothermal reduction/nitridation of silica

The synthesis of silicon nitride powders by reaction between silica, carbon, and nitrogen was the earliest applied method to produce Si3N4, which has been well known since a German patent was granted to Mehner (1896). The overall reaction is

( ) ( ) ( ) ( ) ( ) (2.04)

Powders manufactured in this way display a high α-phase content (>98%), uniform particle size, and a low amount of metallic impurities. The major impurity, oxygen, can be reduced to less than 1 wt% (Szweda et al., 1981). Common practice involves the reaction of silicon dioxide and carbon in a flow of nitrogen at temperatures between 1573 K (1300ºC) and 1773 K (1500ºC ). Reaction times are typically 5 to 10 hours. It has been demonstrated that the formation of Si3N4 by this method could be enhanced through Si3N4 seeding (Hofmann et al., 1993). This practice resulted in improved yields and excellent control of particle morphology. Depending on abundant sources of silica and carbon with very high purity and fine particle size from commercial source (Cochran et al., 1994), the reaction by carbothermal reduction/nitridation of silica is expected to produce high purity, fine silicon nitride powder (Komeya and Inoue, 1975). The synthesis of silicon nitride powder by the carbothermal reaction using silica, carbon, and nitrogen is the economically most attractive way.

At present, silicon nitride is mainly used as structural ceramics. Property comparison of commercial silicon nitride synthesized by three main routes for ceramics is presented in Table 2-1. (Hofmann et al., 1993) Silicon nitride formed from Si-diimide has very high purity, but is very expensive. Si3N4 from carbothermal reduction/nitridation route has the highest oxygen and impurity content. However, several additional requirements should be met from industrial point of view, such as cost of investment, use of hazardous raw materials, waste disposal, and safety aspects of the process. Si-diimide process requires a special infrastructure for handling SiCl4 and ammonia. The reaction also has safety problem since it is strongly exothermic reaction, as well as waste problem. Si-nitridation method employs cheap and harmless raw material, but has a serious waste problem, due to the additional acid wash step after milling with steel balls.

Table 2-1 Typical properties of commercial Si3N4 powder for structural ceramics (Hofmann et al., 1993) Property Production route Si-nitridation Si-diimide Carbothermal

O2 (mass-%) 1.6 1.2 2.1

N2 (mass-%) 38.7 39.3 39.7

Cfree (mass-%) 0.04 0.02 0.06

Sifree (mass-%) 0.04 0.05 0.04 Metallic impurities (mass-%) 0.04 0.01 0.15

α-Si3N4 (%) 96 97 100 spec. surface (m2/g) 15-20 10-12 13 grain size (µm) 0.78 0.38 1.21

2.2 Carbothermal synthesis of silicon nitride

Carbothermal reduction/nitridation of silica is generally conducted in the nitrogen atmosphere; it includes reduction reaction of SiO2 and nitridation reaction forming

Si3N4. The overall reaction can be written as

( ) ( ) ( ) ( ) ( ) ( )

This main reaction consists of a series of simultaneous and sequential chemical reactions, which could be divided into two stages – generation of silicon monoxide

(reduction of SiO2) and formation of silicon nitride. Silicon monoxide is generated in two ways. At the beginning of the reduction reaction SiO is generated by solid-solid Reaction (2.2) at the contact points between carbon and silica particles. After appearance of CO from Reaction (2.2), SiO can also be produced by solid-gas Reaction

(2.3), which becomes the major reaction for SiO production. CO2 produced by Reaction (2.3) is reduced into CO by the Boudouard Reaction (2.4). Silicon nitride can be formed by Reactions (2.5) and (2.6).

( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( ) ( )

Reduction/nitridation process can also include formation of, silicon carbide by Reaction (2.7), which was observed by several investigators (Komeya and Inoue, 1975; Lee and Cutler, 1977; Miller et al., 1979; Siddigi and Hendry, 1985) in the early stages of nitridation. Vlasova et al. (1995) suggested the SiO generation from reaction between

SiC and SiO2 (Reaction (2.8)).

( ) ( ) ( ) ( ) ( )

Dijen (1992) reported the Si3N4 formation by Reaction (2.9).

( ) ( ) ( ) ( ) ( ) ( )

According to Durham et al. (1991) and Vlasova et al. (1995) when partial pressure of oxygen is greater than , silicon oxynitride ( ) could be formed by Reaction (2.10). The presence of oxygen could be due to the Reaction (2.11) according

8 to Rahman (1989) or Reaction (2.12) between the reactor (alumina crucible) and reactant reported by Durham et al. (1991).

( ) ( ) ( ) ( ) ( ) ( )

( )

( ) ( ) ( )

( )

( ) ( ) ( )

Synthesis of Si3N4 produced through carbothermal reduction/nitridation process can be affected by various parameters, such as reaction temperature, gas atmosphere, total pressure of reaction gas as well as partial pressure of certain gas species, gas flow rate, surface area of SiO2 and C, ratio of C/SiO2, seeding in the precursor, and others. Effect of these parameters on the Si3N4 synthesis is reviewed in Section 2.2.2.

2.2.1 Thermodynamics of carbothermal reduction/nitridation of silica

Liou and Chang (1995) reported thermodynamic evaluations of the carbothermal reduction/nitridation reaction using thermodynamic data from work (Chase at al., 1985). Standard Gibbs free energies for Reactions (2.2) – (2.6) are presented in Figure 2-1. It can be seen that SiO gas formation is more thermodynamically favored by solid-solid Reaction (2.2) than the gas-solid Reaction (2.3) in the temperature range. However,

Reaction (2.3) becomes only way of SiO2 reduction when the contact points disappear due to the assumption of SiO2 and C particles. Since the standard Gibbs free energy of Reaction (2.2) and (2.3) are all positive over the reaction temperature range, the formed CO must be effectively removed. Otherwise, a higher temperature is needed for promoting reaction at a reasonable rate. This standard Gibbs free energy change diagram also shows that the nucleation Reaction (2.5) is favorable over the entire reaction temperature range; while, the growth Reaction (2.6) is not feasible at temperatures above ≈1500°C.

Figure 2-1 Standard Gibbs free energy changes for Reactions (2.2)-(2.6)

Equilibrium partial pressure of CO for Reaction (2.1) is high over the temperature range (Figure 2-2). As shown in Figure 2-3 the equilibrium partial pressure of SiO for

Reaction (2.2) is also significant, 𝑖𝑂 =0.53 atm at 퐶𝑂 =0.001 atm and 1700 K

(1427°C), indicating a great potential for carbothermal reduction of SiO2. The equilibrium 𝑖𝑂 increases with increasing temperature and decreasing CO partial -6 -16 pressure. The equilibrium partial pressure of CO2 is small, 10 – 10 atm presented in

Figure 2-4. Excess carbon maintains CO2 partial pressure low, what is crucial for SiO2 reduction.

3.5

3 PN2 = 1 atm 2.5 PN2 = 0.9 atm

PN2 = 0.8 atm 2 PN2 = 0.7 atm Pco(atm)

PN2 = 0.6 atm 1.5 PN2 = 0.5 atm 1

0.5

0 1500 1600 1700 1800 1900 2000

Temperature (K)

Figure 2-2 Equilibrium CO partial pressure for Reaction (2.1)

1.0E+03

1.0E+02

1.0E+01

1.0E+00

1.0E-01

1.0E-02 Psio(atm) 1.0E-03 Pco = 0.001 atm 1.0E-04 Pco = 0.01 atm 1.0E-05 Pco = 0.1 atm

1.0E-06 Pco = 1 atm

1.0E-07 1500 1600 1700 1800 1900 2000 Temperature (K) Figure 2-3 Equilibrium SiO partial pressure for Reaction (2.2)

1E-06 Pco = 0.001 atm Pco = 0.01 atm Pco = 0.1 atm Pco = 1 atm 1E-07

1E-08

1E-09

1E-10 (atm)

2 1E-11 CO P 1E-12

1E-13

1E-14

1E-15

1E-16 1500 1600 1700 1800 1900 2000 Temperature (K)

Figure 2-4 Equilibrium CO2 partial pressure for Reaction (2.4)

2.2.2 Effect of operation parameters on carbothermal reduction/nitridation of silica

2.2.2.1 Temperature

Conversion of SiO2 to Si3N4 is very sensitive to the reaction temperature. Liou and Chang (1995) explored the effect of temperature in the range from 1623 K (1350°C) to

1748 K (1475°C) on the degree of conversion of SiO2 ( ), which was defined by Equation (2.13).

( )

Here indicates initial weight of SiO2, and is the SiO2 weight after reduction.

Results presented in Figure 2-5 reveal that the SiO2 conversion rate increases dramatically at the reaction temperature above 1698 K (1425°C), which is explained by more effective silicon monoxide actually produced by SiO2 reduction at this 12 temperature range. This is consistent with Zhang and Cannon (1984), who reported formation of a significant amount of SiO when the reaction temperature is greater than

1673 K (1400°C). Additionally Liou and Chang (1995) observed the SiO2 reduction rate slowed down after a period of time when the temperature was higher than 1723 K (1450°C). They illustrated this phenomenon that the reactant surface was gradually covered with the solid product and subsequently results in the decline of reduction rate during the process. They also reported the limit temperature of 1773 K (1500°C). When temperature exceeded 1773 K (1500°C), SiC as byproducts maybe apparently formed.

Figure 2-5 Effect of reaction temperatures on the nitridation of sample

Ekelund and Forslund (1992) employed CO evaluation to characterize the carbothermal process in Figure 2-6. They reported that silica reduction commenced at about 1523 K

(1250°C) and the 퐶𝑂 peak indicating the maximum reaction rate was achieved at about 1623 K (1350°C).

Figure 2-6 Diagram showing (-) typical temperature and (+) values as functions of reaction time

Durham et al. (1991) reported only partial conversion of silica to α-Si3N4, and Si2N2O in 5 hours reaction when temperature was lower than 1773 K (1500°C). There was complete SiO2 conversion to α-Si3N4 when temperature was in the range of 1773 K (1500°C) – 1823 K (1550°C). And when temperature was higher than 1873 K (1600°C), β-SiC was produced. Morphology of samples subjected to reduction/nitridation at different temperatures is presented in Figure 2-7. Unreacted SiO2 was observed throughout the sample at 1623 K (1350°C) (Figure 2-7(a)). Discrete particles of α-

Si3N4 were observed with a mean particle size of ~2.5 to 3 μm (Figure 2-7(b) and (c)). For the sample after reaction at 1873 K (1600°C) (Figure 2-7(d)) the structure became less uniform and was associated with β-SiC formation. They conclude that the highest temperature for α-Si3N4 formation was in the range of 1773 K (1500°C) to 1823 K (1550°C).

Figure 2-7 SEM micrographs of powder morphology as a function of reaction temperature: (a) 1623 K (1350°C), (b) 1773 K (1500°C), (c) 1823 K (1550°C), and (d) 1873 K (1600°C).

Boundary temperature due to SiC formation is of particular interest to experimentalists synthesizing Si3N4 by carbothermal reduction/nitridation approach. The formation rate of Si3N4 increases quickly with temperature up to a certain maximum critical boundary. At the temperature higher than the boundary SiC becomes predominant product since SiC formation is more thermodynamically favored at higher temperatures. Additionally

Reaction (2.14) becomes significant due to the high temperature instability of Si3N4 in the presence of carbon (Rahman and Riley, 1989).

( ) ( ) ( ) ( ) ( )

SiC was theoretically determined to be in thermodynamic equilibrium with Si3N4 at 1 atm of N2 at a temperature of 1713 K (1440°C) (Licko et al., 1992) for the Reaction (2.14).

According to this evaluation in the most of investigations there should not be Si3N4 synthesized above this temperature. However, the boundary in practice was reported by investigators up to 1823 K (1550°C). Licko et al. (1992) designed comparison experiments with seeds in their precursor at temperatures of 1823 K±10 K

(1550°C±10°C) and 1863 K±10 K (1590°C±10°C). The authors reported α-Si3N4 as a dominant product with trace β-Si3N4 and β-SiC in the 1823 K±10 K (1550°C±10°C) run, while β-SiC as a dominant product in the 1863 K±10 K (1590°C±10°C) run. They suggested that Si3N4 was formed in sites where activity of carbon is lower than unity.

They also conducted comparison experiments with the mixture of 60 mass% of α-Si3N4 and 40 mass% of carbon as the precursor, at the temperature of 1783 K±10 K (1510°C±10°C) and 1823 K±10 K (1550°C±10°C). After 4 hours nitridation, there was no phase change in the 1783 K±10 K (1510°C±10°C) run, while 25 mass% of β-SiC was observed in the 1823 K±10 K (1550°C±10°C) run. They concluded that the theoretical boundary of 1713 K (1440°C) for Si3N4 formation should be shifted to 1783 K (1510°C)

– 1823 K (1550°C). Therefore, Si3N4 formation was possible up to about 1823 K (1550°C) when a sufficient supply of nitrogen and efficient removal of CO were secured.

The boundary temperature, studied by different researchers varies as follows 1723 K (1450°C) (Lee and Cutler, 1977), 1773 K (1500°C) (Cho and Charles, 1991), 1783 K (1510°C) – 1823 K (1550°C) (Licko et al., 1992), and 1863 K (1590°C) (Hendry and Jack, 1975). These differences may be due to kinetic considerations (Lee and Cutler, 1977), impurities in reaction system (Komeya and Inoue, 1975; Szweda et. al., 1981; Zhang and Cannon, 1984; Mori et. al., 1983; Siddiqi and Hendary, 1985; Perea, 1987; Figusch and Licko, 1987; Durham et. al. 1991), and error margins in experimental methods.

2.2.2.2 Gas atmosphere

Generally carbothermal synthesis of silicon nitride is conducted in nitrogen gas atmosphere (Hendry and Jack, 1975; Komeya and Inoue, 1975; Durham et al., 1991; Eklelund and Forslund, 1992; van Dijen and Vogt, 1992; Vlasova et al., 1995; Demir et al., 2007). Some experimental works have been carried out with different gas atmosphere including a gas mixture with 95% nitrogen and 5% hydrogen (Rahman and Riley, 1989; Alcala and Criado, 1992; Alcala et al., 2001; Real et al., 2004; Ortega et al., 2008), and ammonia instead of nitrogen (Bartnitskaya et al., 1983; Hanna et al., 1985; Durham et al., 1988).

Durham et al. (1988) compared synthesis of Si3N4 in ammonia (NH3) with that in N2 atmosphere; their results are presented in Figure 2-8.

Figure 2-8 Si3N4 formed in the systems SiO2-C-NH3 and SiO2-C-N2

According to their experimental results, there was no advantage associated with using

NH3 at 1573 K (1300°C), however, as the temperature was increased, a much greater conversion of SiO2 to Si3N4 was observed. Synthesis of Si3N4 in ammonia can be presented by the following overall reaction:

( ) ( ) ( ) ( ) ( ) ( ) ( )

This reaction can proceed by Reactions (2.16) and (2.17) via silicon monoxide (SiO), which is generated by silica reduction by carbon (Reaction (2.2)), or by direct 17 reduction/nitridation of SiO2 by NH3 according to Reaction (2.18), which provides another path for Si3N4 formation.

( ) ( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( )

In the operational temperature range of 1573 K (1300°C) to 1773 K (1500°C), NH3 is dissociated primarily to molecular nitrogen (N2) and hydrogen (H2) with small amounts of NH and NH2, all of which react readily with SiO to produce Si3N4 (Ovsyanikov et al.,

1978). Hydrogen derived from NH3 dissociation, acts as reducing reagent (Reaction (2.19)).

( ) ( ) ( ) ( ) ( )

In view of the proposed reaction mechanism Durham et al. (1988) reasoned that the promoted Si3N4 yield was most likely resulted from an increase in the production of SiO due to the supplementary nitridation and reduction capacity from the NH3 and H2, respectively. The authors compared the change in the standard Gibbs free energy ( ) of the system of SiO2-C-NH3 with that of SiO2-C-N2 presented in Figure 2-9. The overall Reaction (2.15) between SiO2, C, and NH3 (labeled as A) is thermodynamically favored at temperatures higher than about 1300 K (1027°C). However, it should be recognized that under Si3N4 synthesis conditions of high temperature in the presence of carbon, H2O obtained from Reaction (2.18) is eliminated immediately by Reaction (2.20). Combining Reaction (2.20) with Reaction (2.18) gives Reaction (2.15).

( ) ( ) ( ) ( ) ( )

Figure 2-9 Change in the standard Gibbs free energy for, A: Reaction (2.15); B: Reaction (2.1); C: Reaction (2.18)

Thermodynamic equilibrium calculations by Badrak et al. (1982) also proved higher concentration of SiO in gas phase when NH3 is used, which facilitates production of

Si3N4. Bartnitskaya et al. (1982) demonstrated that the dissociation of NH3 resulted in a more “active” form of N2 which should lead to improved Si3N4 yield.

Alcala et al. (1992, 2001, 2004, and 2008) studied synthesis of silicon nitride in the 95% nitrogen and 5% hydrogen gas mixture. Positive influence of hydrogen addition on carbothermal reduction/nitridation process was demonstrated by Rahman and Riley

(1989) by comparison the extent of nitridation ( ) between reaction gas with 100% N2 and with 95% N2 and 5% H2, shown in Figure 2-10 and Figure 2-11, respectively. The authors reasoned that SiO generated by additional reduction capacity derived from H2 (Reaction (2.19)) promoted silicon nitride synthesis.

Figure 2-10 Silicon nitride synthesis in 100% N2

Figure 2-11 Silicon nitride synthesis in the 95 vol% N2 and 5 vol% H2 gas mixture

2.2.2.3 Partial pressure of nitrogen

Effects of partial pressure of nitrogen on carbothermal synthesis of silicon nitride were studied in two aspects: nitridation rate, and stability of silicon nitride under higher reaction temperature.

Henry and Jack (1975) examined carbothermal reduction/nitridation at 1773 K (1500°C) with varied controlled by composition of the N2 - Ar gas mixture. Their results, presented in Table 2-2, show that the weight loss and reduction product, characterized by the ratio of α-Si3N4 to β-Si3N4 (%α) were not affected by in the range 0.5-1 atm

In the case of less than 0.5 atm weight loss was slightly greater and SiO2 was partially converted into SiC. When partial pressure of N2 reached 0.1 atm, SiO2 was fully converted into SiC rather than Si3N4. Their results were consistent with calculated

for Reaction (2.14) that minimum 0.45 atm of was necessary to avoid formation of SiC.

Table 2-2 Effect of on the product of reaction with SiO2+C at 1773 K (1500°C)

Reaction (atm) Wt% loss Product % α 1.0 71 27 0.8 69 29 0.6 66 31 0.5 71 27 0.1 80 100% SiC

Ekelund and Forslund (1992) reported investigations of the effect of nitrogen pressure at elevated up to 6 MPa. Increase in N2 pressure to 2 MPa resulted in the greatest yield of Si3N4. They explained that greater promote Reactions (2.5) and (2.6) of

Si3N4 formation. On the other hand, the nitridation rate also decreased when the increased beyond this optimum for thermodynamic and kinetic reasons. Temperature of about 1723 K (1450°C) represents the boundary between Si3N4 and SiC (Reaction (2.14)) (Komeya and Inoue, 1975; Lee and Cutler, 1977) at standard conditions (1 atm pressure); increase in above 1 atm will enhance the stability of Si3N4, and promote conversion of SiC to Si3N4. Therefore the temperature boundary of Si3N4 formation will be increased due to the higher nitrogen partial pressure (Ekelund et al. 1986).

2.2.2.4 Gas flow rate

Increase of the gas flow rate is expected to increase the silica conversion rate by purging carbon monoxide (CO) from reaction system and keeping partial pressure of carbon monoxide ( 퐶𝑂 ) low (Rahman and Riley, 1989). It was reported by Ekelund and

Forslund (1992) that even a small amount of CO (~1%) produced from the reaction obstructed the formation of Si3N4. Higher gas flow rate also reduces resistance caused by the external mass transfer, and makes the overall reaction controlled by chemical kinetics (Liou and Chang, 1995).

Liou and Chang (1995) studied effects on external mass transfer by changing gas flow rate at reaction temperature of 1723 K (1450°C). They observed that when gas flow rate increased from 300 mL/min to 400 mL/min, the rate of SiO2 conversion increased significantly (Figure 2-12); however, further increase in the gas flow rate in the range

400-700 mL/min had no visible effect on SiO2 conversion ( ) as shown in Figure 2-12. This result indicates that the external mass transfer through the gas boundary layer provides a negligible resistance to the progress of reaction when the gas flow rate larger than the critical value.

Rahman and Riley (1989) studied silica reduction/nitridation in 95 vol% N2 – 5 vol%

H2 gas mixture at 1673 K (1400°C). They reported that at gas flow rates of 1020 mL/min 퐶𝑂 appears to be kept adequately low for the SiO2 reduction and nitridation to

Si3N4. Figure 2-13 shows that reaction achieved the highest extend of nitridation ( ) at this critical value, which is defined as the ratio of Si3N4 weight to total weight of sample. The result is consistent with measurements of Van Dijen et al. (1985).

The critical value of the gas flow rate for specific reaction system, as the minimum gas flow rate keeping sufficiently low 퐶𝑂, can be affected by several factors; including starting materials, reducing gas composition, reactor design and others.

Figure 2-12 Effect of gas flow rate on SiO2 conversion (temperature, 1723 K (1450°C);

C/SiO2 molar ratio, 6; pellet-forming pressure, kPa; pellet diameter, 4 mm; initial grain sizes, silicon dioxide and carbon <45 μm)

Figure 2-13 Effect of gas flow rate on the nitridation extent at 1673 K (1400°C) for a time of 7 ks in 95% nitrogen – 5% hydrogen

2.2.2.5 Carbon to silica ratio

According to the overall Reaction (2.1) stoichiometric molar ratio of carbon to silicon dioxide (C/SiO2) is 2; however, for complete conversion of silica, carbon should be taken in excess. Excess carbon ensures sufficient particle contact between SiO2 and C (Durham et al., 1991), improves SiO generation (Komeya and Inoue, 1975), avoids formation of Si2N2O (Reaction (2.10)) (Komeya and Inoue, 1975; Vlasova et al., 1995), provides active sites for the Si3N4 crystals nucleation (Liou and Chang, 1995) and accommodates other catalysts which improve Si3N4 yield (Dijen et al., 1992).

Komeya and Inoue (1975) investigated the effect of C/SiO2 molar ratio on the Si3N4 formation rate from 2 to 60 (Figure 2-14). A significant enhancement of nitrogen content was observed when C/SiO2 molar ratio raised from 2 to 20. Authors explained the enhancement on Si3N4 formation rate by increased SiO production according to

Reaction (2.2), which contributes to Si3N4 formation by Reactions (2.5) and (2.6).

Excess carbon also provides more active sites for Si3N4 nucleation (Reaction (2.5)), which was experimentally demonstrated by Zhang and Cannon (1982).

Figure 2-14 Effect of C/Silica ratio on nitrogen content in reduction/nitridation of silica at 1673 K (1400°C) for 5 hours in nitrogen gas (2000 mL/min)

Komeya and Inoue (1975) examined the effect of the C/SiO2 molar ratio in the range 2- 20 on the production phase composition using XRD analysis. Samples reduced at 1673

K (1400°C) for 5 h with low C/SiO2 ratio (2-5) contained lower nitrogen, residual silica and a small amount of Si2N2O, while XRD spectra of samples produced with higher

C/SiO2 ratio (20), had sharper α-Si3N4 diffraction peaks with no peaks for silica or

Si2N2O. Silicon oxynitride at low C/SiO2 molar ratio was also observed by Vlasova et al.

(1995), who investigated the influence of molar ratio of C/SiO2 in the range from 2 to 3 on reaction products. Their experiments were implemented in a N2 flow in the temperature range from 1473 K (1200°C) to 1723 K (1450°C). The main phases identified were α- Si3N4, β- Si3N4 and Si2N2O when the ratio of C to SiO2 was 2.

Amount of α- and β-phase Si3N4 increased while Si2N2O disappeared with extended exposure time. For the samples with C/ SiO2 = 3, α-Si3N4 formation was dominant. At

1723 K (1450°C) and 300 minutes reaction time, the α-Si3N4 content was 96%, furthermore, pure Si3N4 was only formed during long heat-treatment exposures at 1723 K (1450°C). Excess of carbon secured low oxygen partial pressure avoiding formation of Si2N2O (Durham et al., 1991).

Liou and Chang (1995) investigated the effect of C/SiO2 molar ratio from 3 to 7 at reaction temperature of 1723 K (1450°C). The results demonstrated significant enhancement of SiO2 conversion, when the C/SiO2 molar ratio raised from 3 to 4, and minor effect when the ratio was greater than 4. Durham et al. (1991) reported that a

C/SiO2 molar ratio in excess of 4 would be required in order to obtain complete SiO2 conversion. Influence of C/SiO2 molar ratio on morphology of samples after reduction/nitridation reactions at 1773 K (1500°C) is shown in Figure 2-15 (Durham et al., 1991). At low C content, unreacted SiO2 remains; at C/SiO2 values in excess of 7, discrete Si3N4 particles form, hexagonal in shape. However, the shape and size of these

Si3N4 particles were not significantly affected by the amount of excess carbon.

Figure 2-15 SEM micrographs of samples after reduction/nitridation with different

C:SiO2 molar ratio (sol-gel powder, 1773 K (1500°C) reaction temperature, graphite crucible): (a) 2:1, (b) 4:1, (c) 7:1, and (d) 10:1 (Durham et al., 1991).

The residual carbon in the final product was removed by oxidation reaction in air at about 873 K (600°C) for several hours (Komeya and Inoue, 1975). Duration of the process must be minimized to prevent the re-oxidation of the Si3N4 product.

2.2.2.6 Particle size and specific surface area

The effect of surface area of silicon dioxide and carbon powders was investigated by Zhang and Cannon (1984); Durham et al. (1991); Licko et al. (1992); Ekelund and

Forslund (1992); Liou and Chang (1995); Weimer et al.( 1997). Commonly, decreased

SiO2 and C particle size lead to higher surface area which improves both SiO2 conversion ( ) shown in Figure 2-16 and Si3N4 yield (wt% Si3N4) presented in Figure

2-17. The improved SiO2 conversion is due to increased contact area between C and

SiO2 particles, and the promoted Si3N4 yield is because of increased carbon surface area for Si3N4 nucleation according to Reaction (2.5).

(a) (b)

Figure 2-16 Effect of grain size of SiO2 (a) and C (b) on conversion of silica

(temperature, 1723 K (1450 °C); gas flow rate, 600 mL/min; C/SiO2 molar ratio, 6; pellet forming pressure, kPa; pellet diameter, 4

mm; initial grain sizes, SiO2 and C < 45 μm) (Liou and Chang, 1995)

The effect of grain size on the SiO2 conversion decreased when the particle sizes of both

SiO2 and C were less than 53 μm. Formation of Si3N4 was affected by the SiO2 surface area as shown in Figure 2-16; however, the effect of carbon size (and, therefore, surface area) was much stronger. These results were consistent with findings by Durham et al. (1991), Ekelund and Forslund (1992) and Inoue and Komeya (1975).

Figure 2-17 Weight percent Si3N4 vs. surface area of silica. In sample B1, carbon with a 2 surface area of 680 m /g was used; in samples B2 and C lamp black was used (surface area 34 m2/g); carbon in sample A had a surface area of 55 m2/g (Zhang and Cannon, 1984)

Liou and Chang (1995) studied formation of Si3N4 in pellets formed under different pressure; they found there was a slight increase of yield of Si3N4 with increasing pressure of pellets formation (Figure 2-18).

Figure 2-18 Effect of pellet forming pressure on nitridation of sample (temperature,

1723 K (1450 °C); gas flow rate, 600 mL/min; C/SiO2 molar ratio, 6; pellet

diameter, 4 mm; initial grain sizes, SiO2 and C < 45 μm)

Particle size also affects Si3N4 morphology. Rahman and Riley (1989) used pyrolysed rice husk (PRH) in the carbothermal reduction/nitridation of silica. According to the morphology of produced samples they concluded that the coarser raw material powders lead to growth of whiskers due to their lower tap densities; the finer powders with higher packing density yield smaller, hexagonal silicon nitride grains. They explained their results that the smaller particles provide a higher density of nucleation sites for

Si3N4 growth.

2.2.2.7 Seeding

Seed additive promotes silicon nitride formation rate and decreases particle size of produced silicon nitride. The main reason why seeding improves carbothermal 29 reduction/nitridation process is based on the hypothesis that Si3N4 nucleation can only be achieved during the initial stage of reaction with supersaturation of SiO, resulting in a burst of nucleation followed by growth at lower supersaturation of a reactant. Experimental results by Inoue et al. (1982), Licko et al. (1992), Hofmann et al. (1993) and Weimer et al. (1997) indicated that nucleation of Si3N4 was a much slower process than the crystal growth, and the introduction of seeds avoided the need for an incubation period increasing greatly the reaction rate. In addition, particle size of synthesized Si3N4 decreased with increased ratio of seed/SiO2, and the Si3N4 crystallites were more uniform.

Inoue and Komeya (1982) introduced Si3N4 powder into the graphite-SiO2 mixture, which reacted at 1673 K (1400 °C) for 5 hour; the results showed great increment in

Si3N4 yield in term of nitrogen content (wt%) shown in Figure 2-19. The authors also reported the formation of homogeneous and smaller Si3N4 grains derived from seeded precursor compared with the Si3N4 synthesized without seed, which is consistent with Weimer’s results (1997) presented in Figure 2-23.

Figure 2-19 Nitrogen content of synthesized powders vs. Si3N4/SiO2 ratio in the system

SiO2-C-Si3N4 (Inoue and Komeya, 1982)

Licko et al. (1992) studied effect of Si3N4 seeds with various surface areas on SiO2 conversion, which was characterized by CO evolution plotted in Figure 2-20. There 2 was a remarkable difference between the effect of coarse Si3N4 seed (<2 m /g) and finer 2 Si3N4 seeds (≥8 m /g) on the silica reduction. For the finer seeds they suggested that the 2 surface area of Si3N4 seed of about 1.5-3.5 m /g in the precursor is sufficient to effectively consume the SiO evolved, and then promote the SiO2 conversion to Si3N4.

Licko et al. (1992) also explored the effect of seeding powders contents on SiO2 conversion to the intermediates and products which is shown in Figure 2-21. The SiO2 conversion was characterized by V(CO)/%, where 100% CO corresponded to the completion of Reaction (2.1). The greatest SiO2 conversion increments was observed between 0 - 5 wt% seeding addition.

Figure 2-20 Effect of various seeding additives on the kinetics of carbothermal

reduction/nitridation of silica (10 mass% of seed added to the SiO2-C 2 mixture, 1783 K (1510 °C), 250 mL/min of N2): (— • • —) <2 m /g, (• • • •) 8.0 m2/g, (— — —) 20.9 m2/g, (—) 22.3 m2/g

Figure 2-21 Effect of seed addition on silica reduction expressed in vol.% of total

amount of CO evolved. Mass% of seeds: (—) 0, (— • — • —) 5, (———)

10, (—— • ——) 20.

Weimer et al. (1997) investigated an effect of precursor with various “seed”/SiO2 ratio on synthesized Si3N4 morphology. After removal of unreacted free carbon, the powders were characterized by mean number equivalent circle diameter ( ) shown in Figure

2-22. There was a dramatic change in of synthesized Si3N4 from no seed to

“seed”/SiO2 of 0.04. The size of Si3N4 kept dropping till the ratio of “seed”/SiO2 achieved around 0.3, and then became stable even though the seed addition was increased to have ratio of “seed”/SiO2 equal to 0.6

Figure 2-22 Effect of precursor “seed”/SiO2 ratio on α-Si3N4 crystallite mean number

equivalent circle diameter ( ) (5 h at 1773 K (1500 °C))

The morphology was observed by Weimer et al. (1997) using SEM imaging analysis demonstrated in Figure 2-23. Crystallite size decreased with seed addition what was explained by nucleation and growth of higher number of crystallites, which leads to smaller product Si3N4 crystallite. They also compared the sizes of synthetic Si3N4 particles derived from the precursor with different sizes of seed. Their results indicated that bigger seed size lead to larger product particles. Therefore they suggested that could be controlled by the addition of seed to the precursor in the proper concentration.

Figure 2-23 Effect of “seed”/SiO2 ratio on synthesized Si3N4 morphology: (a) no seed;

(b) “seed”/SiO2 = 0.05; (c) “seed”/SiO2 = 0.1; (d) “seed”/SiO2 = 0.2.

Additionally Weimer et al. (1997) investigated the combined effect of varying “seed” and excess C content on product surface area. According to their results shown in

Figure 2-24 an increase in excess C content and a decrease in “seed”/SiO2 ratio both lead to a higher Si3N4 surface area.

Figure 2-24 Effect of “seed”/SiO2 ratio and excess carbon content on synthesized

product α-Si3N4 surface area (5 h at 1773 K(1500 °C))

2.2.3 Reaction mechanism and kinetic models

The carbothermal reduction/nitridation of silica involves solid-solid reaction and solid- gas reactions. Generally the reactant system is considered to be an agglomerate of small grains; silicon dioxide and carbon are uniformly dispersed and intimately contact with surrounding grains. Reduction of SiO2 initially occurs at contact points between SiO2 and C aggregates according to Reaction (2.2), which generates SiO and CO. Further reduction of SiO2 continuously proceeds by CO (Reaction (2.3)). CO2 generated in Reaction (2.3) is reduced to CO by Reaction (2.4). Once SiO is formed the nucleation of

Si3N4 occurs on the surface of C according to the Reaction (2.5). After Si3N4 nuclei are formed, growth of Si3N4 crystals will be achieved by the gas phase Reaction (2.6).

Zhang and Cannon (1984) proposed the mechanism of Si3N4 formation which includes the burst of nucleation in the initial stage of the process. First of all, based on the morphology observation of Si3N4 hexagonal prisms, which were single crystals, they proposed Reaction (2.6) for growth of Si3N4. The Reaction (2.5) requires Si3N4 crystals growth only at the carbon surface, which will not lead to the hexagonal prism morphology. Then they suggested a heterogeneous Si3N4 nucleation on either carbon or

SiO2 surface, since their results demonstrated that the Si3N4 particle size was inversely related to the C and SiO2 surface area, and the yield of Si3N4 was also promoted by 34 increased C surface. The nucleation is directly related to the surface area of the initial powders, which was consistent with the results of Inoue et al. (1982). During the very early stage of the process, the gas atmosphere in the powder bed is N2 just as in the furnace. However, as CO and CO2 are given off as the products, the CO concentration in the powder bed increases and differs from that in the furnace. Due to the increase 퐶𝑂 the equilibrium 𝑖𝑂 commence to drop according to the Reaction (2.2), which means a lower supersturation of the SiO. Then Si3N4 nucleation may end, even though the growth reaction may continue at the lower SiO concentration. Once the growth of Si3N4 stops due to the decreased SiO concentration, formation of CO will be reduced and the SiO concentration will be increased again until a steady-state value is achieved. On the basis of this assumption the authors calculated the limiting 퐶𝑂of 0.23 atm value at 1673

K (1400 °C) by remaining 𝑖𝑂 at about atm and at 0.769 atm making the total pressure of 1 atm. Zhang and Cannon (1984) predicted that the nucleation stopped when 퐶𝑂 exceeded this value. They also made an estimate, that nucleation is complete after < 0.003% Si3N4 has formed. Therefore, they proposed that a burst of

Si3N4 nucleation occurs followed by growth. The model was mathematically built up to fit the experimental data by firstly assuming the rate of growth at the Si3N4 interface is constant or

⁄ ( )

Here D is a characteristic diameter and is a constant. The volume of a particle is

, where is a geometric constant.

( )

Where V is a varying volume of a constant number of particles; N, formed during nucleation. The mass fraction of Si3N4 at any time is

( )

Here ρ is the theoretical density of Si3N4 ( ) and m is the instantaneous total mass of SiO2 and Si3N4 at time t and varies as SiO2 converts to Si3N4 in the following manner 35

( ) ( ⁄ )

Here is the initial mass of powder. Substitution of Equation (2.25) into Equation (2.23) and integration is carried out as Equations (2.26) and (2.27).

( )

∫ ∫

( ⁄ )

( )

⁄

According to their experimental results the values of the constants were estimated.

⁄

⁄ ( ) ⁄( )

⁄ ⁄ ( ⁄ )

The predicted (theoretical) curve fits the experimental data very well for the initial stage of the reaction (Figure 2-25) and up to 5 hour; further reaction slows down and the nitridation curve departs from theoretical one. The authors explained that the slower region is probably due to the decreased production of SiO.

Figure 2-25 Si3N4 formation vs. time at 1673 K (1400 °C) for samples with C/SiO2 ratios 3.5, 6.2 and 16. The theoretical curve was obtained by the Equation 2.27.

Liou and Chang (1995) proposed Equation (2.28) for the overall nitridation reaction rate and calculated the activation energy of for the initial reaction stage using the Arrhenius plot (Figure 2-26)

( ) ( )

Here is the initial reaction rate of SiO2 ( ), is the preexponential factor, E is the activation energy ( ), R is the gas constant ( ), and T is the reaction time (K). Zhang and Cannon (1984) reported the average activation energy for the SiO formation 401 kJ/mol, while Klinger at al. (1966) found a higher value of 489 kJ/mol. Liou and Chang (1995) suggested that the average activation energy value for the SiO formation is approximately equal to the activation energy of the overall nitridation reaction. The authors proposed that the overall reaction may be controlled by the reduction of SiO2 to SiO.

Figure 2-26 Arrhenius plot showing temperature dependence of the initial nitridation

rate of SiO2 (Liou and Chang, 1995)

Liou and Chang (1995) also used a shrinking core model (Szekely et al., 1976) to represent the reaction between a porous pellet and gas under the condition where the pellet offers no resistance to the diffusion of the reacting species. An assumption was made that the pellet consists of a number of small grains. The grains are ordinarily nonporous, and the reaction front of each grain maintains its original spherical shape. They obtained the following relationship between conversion and reaction time:

⁄ ( ) ( )

Where

√ ( ) ⁄ ( ) √

Here, is the conversion of SiO2 (%), is the rate constant ( ), is the reaction time ( ), is the initial weight of SiO2 in the pellet ( ), is the density of SiO2 in the pellet ( ), N is the number of SiO2 grains in the pellet, is the molecular weight of SiO2 ( ), is the rate constant ( ). The predicted conversion-time relationship derived from Equation (2.27) is compared with the experimental data in Figure 2-27.

Figure 2-27 Experimental data plotted according to shrinking core model for chemical reaction control (Liou and Chang, 1995)

Weimer et al (1997) suggested that nucleation and growth are commonly combined into a single mechanism called nucleation kinetics developed by Avrami (1939, 1940, and 1941). Tompkins (1976) demonstrated that the Erofeyev (1946) approximation of Avrami’s expression (Equation 2.31) is adequate for fitting most kinetic data of the nucleation type.

( ) ( ) ( )

Here t is the time (s), n is the Avarmi constant, X is the fraction conversion, and k is the rate constant defined by Equation (2.32)

( ⁄ ) ( )

-1 -1 Where is the preexponential factor (s ), R is the universal gas constant (8.314 J mol K-1), T is the temperature (K), and E is the activation energy (J mol-1). Weimer and his colleague (1997) applied a three-level factorial surface response experimental design to quantify the effect of temperature (1648 K (1375°C) to 1698 K (1425°C)) and reaction time (1 to 5 hour) on reaction kinetics. Their data were fit by multiple linear regressions (Figure 2-28) to Equations (2.31) and (2.32), getting an activation energy of , an Avrami constant of n = 0.8, and a preexponential factor of

Figure 2-28 Nucleation-growth model for reaction kinetics (predicted vs. experimental)

The activation energy of is closed to the value for the formation of SiO. Thus the authors suggested that the overall Reaction (2.1) is controlled by the reduction of SiO2. Since the reduction at the contact points of C and SiO2 is fast, the rate-controlling step is most likely the gas phase Reduction (2.3). They also proposed the growth to be between homogeneous parabolic and rod-like due to the n value.

Ortega et al. (2008) considered the diffusion through the layer of the reaction product at the stage of silicon nitride formation. Since the nitridation rate decreases along with the gradually increasing Si3N4 layer thickness, the authors believe the reaction will be mainly controlled by the diffusion of the gaseous reactants through the product layer.

They employed a model that Si3N4 growth from each contact to form a shell which advances from the outside toward the interior during the nitridation process. The schematic representation of the diffusion mechanism of a single particle is demonstrated in Figure 2-29.

Figure 2-29 Schematic representation of the diffusion mechanism of s single particle

2.2.4 Summary

Carbothermal reduction/nitridation of silicon dioxide is most efficient method for synthesis of silicon nitride.

Carbothermal reduction/nitridation of silica commences with the solid-solid reaction at the contact points of carbon and silica, in which SiO and CO are formed. Silicon nitride is formed by nucleation on carbon and growth by SiO reaction with CO and nitrogen. The growth reaction of silicon nitride is thermodynamically feasible (under standard conditions) at temperatures lower than about 1723 K (1450 °C).

A study of effects of operation parameters on kinetics of silicon nitride synthesis can be summarized as follows:

1) Reduction of SiO2 can be accelerated by increasing reaction temperature.

2) SiC is more stable than Si3N4 at high temperatures.

3) Addition of hydrogen to nitrogen has a positive effect on carbothermal reduction/nitridation. The use of ammonia promotes both reduction and nitridation processes.

4) Increased nitrogen pressure enhanced the yield and stability of Si3N4 at higher

temperatures, which increased SiO2 reduction rate.

5) Increased gas flow rate promoted the conversion of SiO2 to Si3N4 by removal of CO.

6) Excess of C provided greater reduction capability, more nucleation sites for

Si3N4, low oxygen partial pressure. Residual C was usually burned off in the air at 600 °C.

7) Smaller grain size of both SiO2 and C promoted SiO2 conversion to Si3N4. Carbon particle size had a stronger effect, since C played an important role not

only in the SiO2 reduction but also the Si3N4 nucleation.

8) Si3N4 whiskers were observed when coarse raw materials were used, and

hexagonal Si3N4 crystals were prepared by finer precursors.

9) Seeding improved the Si3N4 formation rate. The synthetic Si3N4 particle size decreased with increased seed content.

Although investigations of carbothermal synthesis of silicon nitride have been conducted for decades, there are contradictions and inconsistencies in the literature. Particularly, different temperature ranges for stability of silicon carbide and nitride were reported. Mechanism for silicon nitride formation needs further understanding. Few investigations were conducted under elevated nitrogen pressure. This study will focus on the effects of operational parameters on carbothermal synthesis of silicon nitride and mechanism of its formation.

2.3 Objectives of the project

Objective of this project is to investigate the process of silicon nitride synthesis by the carbothermal reduction/nitridation, and to demonstrate feasibility of silicon production by decomposition of silicon nitride. The project will study effects of different operation parameters on silicon nitride synthesis. Reaction mechanism will be elucidated based on results of thermodynamic and kinetic analysis. Kinetic modeling will be developed to describe the silicon nitride formation.

CHAPTER 3 EXPERIMENTAL

Carbothermal synthesis of silicon nitride was examined in a fixed bed reactor under gas atmospheres of nitrogen, hydrogen-nitrogen, and argon-nitrogen mixtures. The off-gas composition was detected and recorded by a gas analyzer. The reacted samples were characterized by phase identification, element composition analysis and morphology observation. Raw materials used in the investigation, experimental set-up, experimental procedures, and sample characterization are described in this chapter.

3.1 Materials

In the process of carbothermal synthesis of silicon nitride, high purity chemicals of silica and graphite fine powders were mixed together as precursor materials. The process was conducted in ultra high pure nitrogen and nitrogen/hydrogen mixture with specific composition. Aiming to identifying the reaction mechanism, a commercial silicon carbide was employed in nitridation experiments. In the attempts to decompose silicon nitride to prepare silicon, both commercial and synthesized silicon nitride were examined in argon/hydrogen atmosphere. Detailed materials specifications are presented as follows

3.1.1 Silica

Crystallized and fumed silica chemicals used in the experiments of this investigation was from Sigma-Aldrich Chemical Company, Inc. Crystallized silica (Catalogue No. 342890) has the particle size of less than 325 mesh and purity > 99.6%. Fumed silica (Catalogue No. 381276) has surface area of 175-225 m2/g and purity > 99.8%. The chemicals may contain adsorbed moisture and CO2 which were removed by calcination at 900 °C.

3.1.2 Graphite

Graphite powder used in the experiments was supplied by Sigma-Aldrich Chemical Company, Inc. (Catalogue No. 282863). The graphite powder is synthetic with particle size less than 20 µm and purity > 99.99%.

The actual specific surface area (SSA) of the crystallized (C) and fumed (F) silica (S), and graphite (Gr) was measured before making the raw materials. And the SSA of the raw mixtures (C+Gr and F+Gr) with C/SiO2 of 4.5 were also measured before reaction and presented in Table 3-1.

Table 3-1 Specific surface area of raw materials C.S. F.S. Gr C.S. + Gr F.S. + Gr SSA (m2/g) 5.8008 467.0752 14.2988 8.3586 128.192

3.1.3 Silicon nitride

Silicon nitride was supplied by Sigma-Aldrich Chemical Company, Inc. (Catalogue No. 334103). Its particle size is less than 1 μm and purity is > 99.9%.

3.1.4 Silicon carbide

Silicon carbide employed to synthesize silicon nitride was purchased from Sigma- Aldrich Chemical Company, Inc. (Catalogue No. 357391). The particle size is about 400 mesh. The purity reported by assay test is ≥ 97.5%.

3.1.5 Gases

Nitrogen, hydrogen and argon were used in this investigation. Nitrogen was used to synthesize silicon nitride. Hydrogen was added into reaction atmosphere to improve silica conversion to silicon nitride. Argon was used to form silicon carbide and then to explore the conversion from silicon carbide to silicon nitride. All of these three gases

45 were supplied by Air Liquide Australia limited, as compressed gases in gas cylinders.

They were all extra high purity gases with typical impurities of H2O <3ppm, O2 <2ppm, and total hydrocarbon <0.5ppm according to their specifications.

3.2 Experimental setup

The experimental setup for carbothermal synthesis of silicon nitride includes (1) fixed bed reactor and furnace with temperature controlling system, (2) gas system and (3) off- gas composition analysis system. The details for each part are presented in the following sections.

3.2.1 Experimental furnace and reactor

The experiments of carbothermal synthesis of silicon nitride were conducted in a fixed bed reactor in a vertical tube electric furnace with MoSi2 heating elements (Ceramic Engineering, Australia). The schematic of the experimental reactor is presented in Figure 3-1. A sample pellet was loaded in the inner tube with the inner diameter of 8.6 mm. It was supported by an alumina plug, which was fixed by an alumina pin via the holes drilled at the bottom of the inner tube. The sample temperature during reaction was measured by a type B thermocouple which was protected against reaction gas atmosphere by an alumina thermocouple sheath. The inner tube holding the sample was inserted into the outside recrystallized alumina sheath with the inner diameter of 19 mm. The inner tube, outside tube and the thermocouple sheath were fixed by special metal fitting and sealed with O-rings. Reaction gas was introduced into reaction system from the top of the inner tube, flowed through pellet sample and plug, and then left the reactor from the gap between the outside and inner tubes. After assembling of reactor with sample pellet, the reactor was inserted into the hot zone of the vertical electric high temperature furnace to heat up to target temperature. The furnace used in this investigation is limited to a maximum heating temperature at 1973 K (1700 °C).

Figure 3-1 Schematic diagram of the experimental set-up

3.2.2 Gas system

Gases used in the experiments were introduced into reactor via a gas system shown in Figure 3-2. The gas system consists of an electronic box with four channels and so the capability to control four Brooks mass flow controllers which can provide the reaction system with different gas or gas mixture with desired composition at controlled flow rate. Gas flow rates from the Brooks mass flow controllers at various setting points were determined using a Defender 520 Primary Flow Meter at room temperature. Then the measured flow rates were converted to the values under standard condition (STP) which are plotted in Figure 3-3.

Figure 3-2 Schematic diagram of gas system

2000

QAr= 16.276S + 60.489 1800 N2 R² = 0.9986 1600 H2

1400 Ar QH2= 19.736S - 3.3547 R² = 0.9999 1200

1000

800 QN2 = 15.572S - 15.309

Flow Flow rate (Q),mL/min 600 R² = 0.9989

400

200

0 0 10 20 30 40 50 60 70 80 90 100

Setting (S) % of full scale

Figure 3-3 Calibration of mass flow controllers

3.2.3 Off-gas composition analysis

The off-gas composition was simultaneously and continuously monitored and recorded by an ABB AO 2020 Continuous Gas Analyzer connected with a computer. The concentration of CO can be detected in two different ranges, 0-3,000 ppm and 0-10 vol%. The maximum concentration of CH4 can be accurately detected is 2000ppm. The measurement range of CO2 is 0-1 vol%. Data of gas concentrations were detected and recorded every 10 seconds automatically by the system.

3.3 Experimental procedures

Generally, silica and graphite were mixed with specific ratios. Then the powder mixtures were pressed into pellets using cylindrical steel die. A sample pellet was loaded into the reactor, and treated under specific conditions to synthesize silicon nitride. The reacted sample pellet was then characterized to evaluate the extent of reaction. In the case of the decomposition of the synthesized Si3N4, residual C was removed by oxidation in air. And then the Si3N4 powder was loaded in a recrystallized Al2O3 crucible, and decomposed to Si. Details of experimental procedures are illustrated in the following sections.

3.3.1 Sample preparation

The raw materials used in the carbothermal silicon nitride synthesis were the mixtures of silica and graphite with specific molar ratios. Weighed SiO2 and C powders were added into a plastic jar with steel milling balls and mixed with distilled water. 0.3 wt% of CMC was added to stabilize the slurry. The mixture was rolled on rolling machine for 8 hours to ensure a homogeneous composition. Then, distilled water was removed by heating mixture up in an oven at about 393 K (120 °C) for several hours till there was no weight loss. The dried mixture was rolled on rolling machine for a few minutes to separate balls with loose powder.

Weighed mixture was pressed by an ENERPAC hydraulic press into a pellet with dimension of 8 mm in diameter and about 9 mm in length. Generally, of 49 the fumed silica and graphite mixture or of the quartz and graphite mixture was weighed (pellet weight varies with C/SiO2 ratio), and packed into a cylindrical steel die with dimension of 8mm in inner diameter and 15mm in length. The mixture was pressed by the hydraulic press at 20 kN (approximate Pa) for 2 minutes. For the experiments of SiC nitridation and Si3N4 decomposition, weighed powder samples were directly packed into an alumina crucible and loaded at the bottom of the outside tube of the reactor. The dimensions of the crucible are: the top outside diameter 15 mm, the bottom outside diameter 12 mm, the height 24.5 mm, and the thickness 1 mm.

3.3.2 Reduction experimental procedures

Loading of sample pellet and assembling of the reactor were conducted at room temperature. A sample pellet was loaded to the bottom of the inner alumina tube, supported by an alumina plug held by an alumina pin via holes drilled through the wall of the inner tube. Then, the inner tube was inserted into the outside tube; the inner and external tubes were then fixed together by metal fittings. Gas was introduced into the reactor from the top of the inner tube, flowed downwards via sample pellets and plug till the bottom of the tube, escaped from the gap between the inner tube and the outside tube, and then flowed upwards to the gas outlet at the top of the reactor. Off-gas flowed through two dust filters with pore sizes of 5.0 μm and 0.65 μm in sequence, then to

ABB AO 2020 gas analyzer, and the concentrations of CO, CO2 and CH4 in it were detected. Soap water was applied in each single run after reactor assembling to ensure no gas leaking from the fitting joints. When the furnace temperature reached the target value, the reactor assembly was directly inserted into the furnace to a position that the sample was located in the isothermal zone. Generally it took about 15 minutes to heat sample up from room temperature to the target temperature. The temperature of the sample pellet was monitored by a thermocouple with its joint about 2 mm above the pellet. Reaction was stopped after desired treatment duration by directly raising the reactor assembly up above the furnace hot zone. Reacted samples were cooled down under the same reaction gas atmosphere. The pellets were removed from the reactor when the thermocouple indicated sample pellet temperature was less than 473 K (200 °C) to avoid reoxidation of the product when it was exposed to air. After stopping 50 the flowing gas, the reactor was disassembled and the reacted sample pellet was taken out.

In some experiments, residual graphite was removed by oxidation reaction. The reacted pellet was manually ground into powder by mortar and pestle, spread to the bottom of an alumina ceramic boat as a thin layer, and then heated by a muffle furnace up to 923 K (650 °C) in air till no weight loss.

3.3.3 Synthesis of silicon nitride under elevated nitrogen partial pressure

Experiments on synthesis of silicon nitride were conducted under nitrogen partial pressures higher than 1 atm (101 kPa) in a vertical graphite furnace chamber (10 atm maximum). A sample pellet was supported by a piece of graphite foil and a graphite pin through the graphite tube reactor. Then the graphite tube reactor was inserted into the centre of the cylindrical chamber from the top of the furnace. An external thermocouple to detect the sample temperature was inserted from furnace bottom about 2 mm below the pellet. And then the cooling water pipes were connected with the top part of the furnace. After the reactor system was assembled, the furnace chamber was evacuated to approximately 10-3 torr residual pressure for a few minutes. Simultaneously the furnace was switched on and the sample pellet was heated up to the target temperature. When the temperature reached the target value, nitrogen was introduced into the reactor from the top of tube, flowed through the pellet and graphite foil, and then escaped from the tube bottom. The nitrogen pressure was controlled by a gas pressure regulator installed on the N2 gas cylinder which was adjusted up to 1100 kPa. And the nitrogen flow in the reactor was controlled by a needle valve connected to the tube at the bottom of the reactor. After the desired reaction duration, the experiment was stopped by switching the furnace power off. Reaction gas was kept flowing till the temperature was lower than 363 K (90 °C). Then the cooling water was switched off before sample pellet was taken out.

3.3.4 Decomposition of silicon nitride

Both commercial and synthesized silicon nitride was used to examine its decomposition.

The Si3N4 powder was packed into a recrystallized alumina crucible which was then loaded down to the bottom of the outside reactor tube. A thermocouple used to indicate the temperature of sample powder was installed with its joint about 2 mm above the upper edge of the crucible. Ar and H2 were used individually as flowing gas to remove the released N2. After the decomposition experiments, the crucible was taken out of the outside tube, and the reacted sample powder was removed from the crucible. The reacted powder was manually mixed by mortar and pestle before further characterization. Samples used in SEM/EDS analyses were not ground.

3.4 Sample characterization

The extent of reduction of silica can be obtained by the data of CO and CO2 concentrations recorded by the continuous gas analyzer. X-ray Diffraction (XRD) analysis was used to identify phases obtained in reacted samples. LECO oxygen/nitrogen/carbon determinators provided quantitative elemental composition of the reduced samples which was used to calculate the content of each phase in the forms of SiO2, Si3N4, SiC, and C. Morphology observation was carried out by scanning electron microscope (SEM) analysis, and Energy Dispersive X-ray (EDS) element composition analysis was employed to identify the chemical composition of particles. The Focused Ion Beam – Transmission Electron Microscope (FIB-TEM) analysis was employed to observe the interface of SiC and Si3N4. Details are illustrated in the following sections.

3.4.1 X-ray diffraction analysis

X’Pert PRO Multi-purpose X-ray Diffraction system (MPD system) was used to identify phases in samples. Reacted sample pellets were manually ground by mortar and pestle into fine and homogeneous powder. Then the powder was packed into a steel circular sample holder. The diameter and depth of the hole holding powder sample were

15 and 1 mm, respectively. A glass slide was used to press the powder surface to ensure that the surface of the sample powder at the same level as the edge of the sample holder. The setting parameters are listed as followings.

 X-ray source: copper Kα radiation with λ = 1.54 Å

 Voltage = 45kV; Current = 40mA

 Scan range (°): 10 ~100

 Scan step size (°): 0.0262606

 Time per step (s): 51.00

 Net time per step (s): 49.47

 Scan speed (°/s): 5.15×10-3

 No. of steps: 3427

 Total time (h:m:s): 00:11:53

 Divergence slit (°): ½

 Diffraction beam (°): 1

3.4.2 Analysis of oxygen, carbon and nitrogen content

The content of oxygen and nitrogen in the reacted sample were determined by LECO TC-436 DR or TC 600 Nitrogen/Oxygen Determinator. The content of carbon was determined by LECO SC-444 DR Carbon/Sulphur or TruSpec Carbon/Nitrogen Determinator. Reacted sample pellets were manually ground into homogeneous powder for analysis.

The Nitrogen/Oxygen determinator has the capability to measure the contents of oxygen and nitrogen simultaneously. The powder sample (approximate 0.04-0.06 g) was filled into tin capsule, gently sealed, then loaded into nickel basket. Analyzing system was purged by high purity helium (He) for 15 – 20 minutes before measurement to ensure 53 the system was stabilized. System checking was conducted before sample measurement in order to ensure no leak in the system. Then the basket with capsule filled by sample was loaded from the top of the furnace, dropped into a graphite crucible. The graphite crucible was heated up to a temperature higher than 3000 °C. The oxygen released from samples reacted with carbon from crucible (extra graphite powder added into crucible in the high oxygen concentration analysis) forming carbon monoxide and carbon dioxide which were detected by infrared detectors and converted to the oxygen content in the analyzed samples. To analyze the nitrogen content released from a sample, the gas from the CO/CO2 detectors was passed through a catalytic converter to convert CO into CO2, and then CO2 was removed in an alkaline adsorption tube. Then the nitrogen content was detected by a thermal conductivity detector.

Similarly, in the case of Carbon/Sulphur Determinator about 0.1g sample powder was weighed and loaded into an alumina crucible. The sample was heated by an induction furnace and the carbon was burnt forming CO and CO2, which were detected by infrared detectors and converted to the carbon content in the analyzed sample.

3.4.3 SEM analysis

Scanning electron microscope (SEM) instruments (Hitachi 3400-I, and NanoSEM 230) were employed to observe the morphology of samples. Reacted sample pellets were observed by SEM to explore not only shape and size of particles, but also relative position of different particles. Both pellet and powder samples were observed in two forms: loose powder, and mounted and polished samples. Phases in various reaction stages were observed, which contributed to the investigation on the mechanism of reaction in SiO2-C-N2 system. To mount samples for SEM observation, sample pellets/particles were put at the middle bottom of a cylindrical plastic mould with diameter of 25 mm. Epoxy resin was loaded into the mould. Then the mould with sample and resin was vacuumed for 20-30 minutes to remove bubbles, and keep overnight to allow the resin to set. Then the mounted samples were ground by alumina sand paper and polished by alumina suspension on polishing pad to get smooth surface. The reason to apply alumina sand paper and suspension was to avoid the contamination of the samples by SiC from diamond sand paper and suspension, since SiC was an

54 intermediate product and consumed in further nitridation process. The polished samples were placed in an oven heating up to about 343 K (70 °C) to remove the moisture. Then samples were sputtering coated with a thin layer of Cr to achieve good conductivity for SEM/EDS analysis.

3.4.4 EDS analysis

Energy-dispersive X-ray spectroscopy instruments (Hitachi 3400-I, and NanoSEM 230) were applied to analyze element distribution and composition in particles. The polished samples for the SEM observation were also analyzed by EDS. Samples without mounting and polishing were also analyzed by EDS point analysis, which helped to distinguish different particles in conjunction with the information of EDS mapping results, SEM morphology observation, and XRD phase identification.

3.4.5 FIB-TEM analysis

The site specific Transmission Electron Microscope (TEM) specimens on the interface of SiC and Si3N4 were prepared using a FEI xT Nova Nanolab 200 Dualbeam Focused Ion Beam (FIB) with the ex-situ lift-out method. Then the cross section obtained was analyzed by the TEM analysis, which was carried out in a Philips CM 200 FEGTEM equipped with a Bruker Esprit EDS system and operating at 200 kV. The element line scanning and mapping analysis identified the phases and the relative location contributing to the current work.

3.5 Calculation of extent of reactions

The mixture of silica and graphite was reacted under the nitridation gas atmosphere to synthesize silicon nitride. Under the experimental conditions in this investigation, there was silicon carbide formed in the reaction system. Gaseous products carbon monoxide, carbon dioxide, silicon monoxide were formed, and methane could be also formed in

55 the presence of hydrogen. There were Si3N4 and SiC as reaction products as well as residual SiO2 and C.

The weight loss during reactions was caused by the formation of CO, CO2, SiO, and

CH4 which were removed by the flowing gas. In consideration of all reactions in SiO2-

C-N2 system, the progress of the carbothermal reduction/nitridation process was characterized by four parameters: (1) Reduction (conversion) of silica ( 𝑖𝑂 ) was obtained based on the oxygen content (O wt%) of the product analyzed by LECO oxygen determinator. (2) Yield of silicon nitride ( 𝑖 ) was calculated based on the nitrogen content (N wt%) of the product obtained from the LECO nitrogen determinator.

(3) The loss of silicon ( 𝑖𝑂) in the form of SiO was worked out by the weight loss and mass balance of silicon. (4) The difference of total silicon in the solid product and those as SiO2 and Si3N4 were assigned as SiC. Content of free carbon was calculated by the carbon content (C wt%) determined by LECO carbon determinator minus that in SiC.

The calculations were based on the assumptions that the only phase containing oxygen in the product was residual SiO2; Si3N4 was the only phase containing nitrogen in the final product; and SiC is the only Si-containing intermediate phase in the solid product. These assumptions were consistent with the results of XRD phase identification.

For decomposition of Si3N4 to prepare Si, the extent of decomposition ( ) was calculated based on nitrogen content determined by LECO nitrogen determinator.

The details of calculations are presented in following sections.

3.5.1 Calculation of the extent of silica reduction

The process of silicon nitride synthesis by carbothermal reduction/nitridation commences with the reduction of silica ( 𝑖𝑂 ). To calculate the extent of reduction, it is assumed that the flow rate of reactant gas equals to that of product gas. The error caused by the slight difference between the flow rates of inlet gas and off gas is negligible.

Overall reaction of carbothermal synthesis of Si3N4, Reaction (2.1), was considered in the calculations. The content of CO detected by the gas analyzer indicates the rate of the

56 silica reduction. Moreover, SiO2 is also reduced by CO resulting in CO2 shown as

Reaction (2.3). Part of the CO2 is purged by flowing gas before further reaction with solid C forming CO. Therefore the CO2 detected by the gas analyzer is also counted to the SiO2 reduction calculation. In the current work the extent of silica reduction ( 𝑖𝑂 ) is also named as the extent of silica conversion (to intermediate (SiO) and products

(SiC/Si3N4)) in the following chapters, since conversion of SiO2 commences from reduction forming SiO. Due to the difficulty in detecting SiO loss rate, it is not considered in the calculations. Instead, the calculated final extent of reduction from CO and CO2 contents in the off gas is scaled to that reached in the final reduction product calculated from product composition, as illustrated bellow.

( ) ( ) ( ) ( ) ( ) (2.1)

( ) ( ) ( ) ( ) (2.3)

SiO2 conversion 𝑖𝑂 is calculated by integrating all the CO and CO2 evolved in the reaction duration (t). The contents ( ) of CO and CO2 in vol% are multiplied by the gas flow rate (GFR) to determine the volumes of CO and CO2 produced in unit reaction time. Corresponding amount of oxygen (O, mol) in unit reaction time (t), i.e. oxygen removal rate, is calculated. Integration of the oxygen removal rate from sample with reaction time from time 0 to t gives the total amount of oxygen removed from the reaction system, which is related to the amount of converted SiO2. This calculation is expressed in Equation (3.1), and can be implemented by the Microsoft Excel spreadsheet functions.

𝑖𝑂 ∫ ( 퐶𝑂 퐶𝑂 ) (3.1) 𝑂 𝑖

where, 𝑖𝑂 – unscaled extent of reduction (conversion) of SiO2 based on the CO and

CO2 contents from gas analyzer, %;

퐶𝑂 - CO content in off gas, vol%;

퐶𝑂 - CO2 content in off gas, vol%;

GFR - gas flow rate, L/min;

22.4 - the volume for 1 mol ideal gas under standard condition, L/mol;

𝑂 𝑖 - initial amount of oxygen in SiO2 in unreduced sample, mol;

t - reaction time, min.

In the above calculation, the loss of oxygen in SiO is not considered, which explains why 𝑖𝑂 is called unscaled extent of reduction.

Based on the assumption that all oxygen detected by LECO oxygen analysis exists in the form of residual SiO 2, the conversion of SiO2 ( 𝑖𝑂 ) is defined by Equation (3.2).

𝑖𝑂 ( 𝑂 𝑖 𝑂 퐶𝑂) (3.2) 𝑂 𝑖

where, 𝑖𝑂 - conversion of SiO2 at the end of reaction, %;

𝑂 퐶𝑂 - final amount of oxygen in the reacted sample, mol.

The final conversion of silica ( 𝑖𝑂 ) is calculated by Equation (3.3).

𝑂 𝑖𝑂 (3.3) 𝑂 𝑖 𝑖 where, 𝑂 - weight fraction of oxygen in reduced sample, wt%;

- mass of sample pellet after reaction, g;

𝑂 - molecular weight of O, g/mol;

𝑖 𝑖 - initial amount of SiO2 in the unreduced sample pellet, mol.

The above conversion of SiO2 is more accurate than that from integration of CO and

CO2 contents in the off gas, and is used to calibrate the final point of the 𝑖𝑂 curve obtained by Equation (3.1). Then, the data of extent of reduction of SiO2 at different time are corrected using a scale factor equal to 𝑖𝑂 𝑖𝑂 at the end of reduction.

3.5.2 Calculation of the extent of nitridation

Calculation of the extent of nitridation is based on the assumption that all nitrogen detected in reacted sample exists in the form of silicon nitride, since no other phase containing nitrogen is detected by XRD. Weight fraction of Si3N4 is defined by Equation (3.4).

𝑖 𝑖 ( ) (3.4)

where, 𝑖 - weight fraction of Si3N4 in a reacted sample, wt%;

- mass of sample pellet after reaction, g;

- weight fraction of nitrogen in a reacted sample, wt%;

𝑖 - molecular weight of Si3N4, g/mol;

- molecular weight of nitrogen, g/mol.

The yield of silicon nitride ( 𝑖 ) is defined by Equation (3.5) as the amount of Si converted into Si3N4 divided by the total amount of Si from initial SiO2, and is determined by Equation (3.6)

𝑖 𝑖 𝑖 (3.5) 𝑖 𝑖

( 𝑖𝑂 퐶) (3.6) 𝑖 𝑖

where, 𝑖 𝑖 – final amount of Si distributed into Si3N4, mol;

𝑖 - mass of sample pellet before reaction, g;

𝑖𝑂 - molecular weight of SiO2, g/mol;

퐶 - molecular weight of C, g/mol;

R - molar ratio of C/SiO2.

Normally in literature synthesized Si3N4 was characterized in the way of nitrogen weight fraction ( ) or Si3N4 weight fraction ( 𝑖 ). Regarding to the definition of

Si3N4 weight fraction ( 𝑖 ) by Equation (3.4), which is only proportional to the nitrogen weight fraction ( ). The Si3N4 yield identified by Equation (3.5) and determined by Equation (3.6). 𝑖 is not only proportional to the but also the ratio of the final pellet weight ( ) to the initial pellet weight ( 𝑖) (weight loss) for raw materials with specific C/SiO2 ratio (R). Si3N4 yield ( 𝑖 ) actually indicates the efficiency of Si conversion to Si3N4 and is employed in the current work to characterise the productivity of Si3N4.

3.5.3 Calculation of the extent of carburization

As detected by XRD, there was silicon carbide formed as intermediate product in the nitridation experiments. Thus the extent of silica carburization to silicon carbide is necessary to be counted. Since there always is excess carbon in the reacted samples and the carbon content analyzed by LECO analysis gave the total carbon content, the amount of silicon carbide cannot be calculated directly. Four elements exist in a reacted sample: Si, C, O, and N, which are assumed to exist in three compounds, SiO2, Si3N4 and SiC. Residual free carbon is also present in the reduced samples. Therefore, weight fraction of element silicon ( 𝑖) can be determined directly by Equation (3.7).

𝑖 𝑂 퐶 (3.7)

Thus, the content of silicon in the reacted sample can be determined by Equation (3.8).

𝑖 ( 퐶 𝑂) (3.8) 𝑖

where, 𝑖 - final amount of silicon in a reacted sample, mol;

퐶 - weight fraction of carbon in reacted sample, wt%;

𝑂 - weight fraction of oxygen in reacted sample, wt%;

𝑖 - molecular weight of silicon, g/mol.

Loss of silicon in the form of SiO ( 𝑖𝑂) can also be calculated through the total carbon content by Equation (3.9). Weight fraction of SiC can be determined by Equation (3.10).

𝑖 𝑖 𝑖 𝑖𝑂 (3.9) 𝑖 𝑖

𝑖퐶 𝑖퐶 ( 𝑖 𝑖 𝑖𝑂 𝑖 𝑖 ) (3.10)

where, 𝑖퐶 - molecular weight of silicon carbide, g/mol;

𝑖 𝑖𝑂 – final amount of Si distributed in residual SiO2, mol.

The extent of carburization is evaluated as yield of silicon carbide ( 𝑖퐶), which is defined (Equation 3.11) as amount of silicon converted into silicon carbide divided by total amount of silicon from reactant. It is calculated according to Equation (3.12).

𝑖 𝑖퐶 𝑖퐶 (3.11) 𝑖 𝑖

𝑖퐶 𝑖퐶 (3.12) 𝑖 𝑖 𝑖퐶 where, 𝑖 𝑖퐶 – Final amount of Si distributed into SiC, mol.

3.5.4 Calculation of the extent of silicon nitride decomposition

Synthesized silicon nitride by carbothermal reduction/nitridation process is decomposed to prepare silicon, which provides a novel potential to prepare solar grade silicon.

Extent of Si3N4 decomposition is defined as Equation (3.13). The nitrogen weight fraction is determined by LECO nitrogen analysis.

𝑖 (3.13) 𝑖

Where, - initial nitrogen weight fraction of Si3N4 powder, wt%;

- nitrogen weight fraction in final reacted sample, wt%

3.6 Error analysis

The rate and extent of reduction were calculated on the basis of CO concentration in the off-gas. Final extents of reduction, nitidation and carburisation were found from oxygen, nitrogen and carbon concentrations in the sample after reaction.

Error in measurement of CO concentration in the off gas included error in the measurement of CO concentration using IR sensor ABB 2008 (0.1%) and error in measurement of the gas flow rate using Brooks flow meters (0.5%). Total error in the measurement of CO concentration was below 1%.

Errors in measurement of oxygen, nitrogen and carbon in the sample after reaction using LECO method were 1.6, 1.7 and 0.9% correspondingly. A detail error analysis in measurement of the rate and extent of reduction/nitridation in carbothermal reduction of metal oxides in nitrogen atmosphere was considered by S.A Rezan in his PhD thesis Synthesis of Titanium Oxycarbonitride by Carbothermal Reduction and Nitridation of Rutile and Ilmenite Concentrates (UNSW, 2010).

Reproducibility of silica reduction was tested in four parallel experiments at 1500oC in nitrogen with gas flow rate 1 L/min. Discrepancy of experimental curves was within 5%.

CHAPTER 4 EXPERIMENTAL RESULTS

4.1 Carbothermal synthesis of silicon nitride

4.1.1 Effect of temperature

The effect of temperature on carbothermal synthesis of silicon nitride was tested in temperature-programmed and isothermal experiments. In these experiments, fumed silica - graphite mixture with C/SiO2 molar ratio of 4.5 was employed.

In the temperature-programmed experiments, which were run in pure nitrogen, the temperature was ramped from 1573 K (1300 °C) to 1873 K (1600 °C) at 3 °C/min. The temperature profile and the change in CO content in the off gas during reaction process are presented in Figure 4-1. CO evolution characterises silica reduction to silicon nitride and silicon carbide.

4000 1650

3500 1600 CO content 3000 temperature 1550

C) °

2500 (

1500 2000 1450

1500 Temperature CO CO content (ppm) 1400 1000

500 1350

0 1300 0 30 60 90 120 150 Reaction time (min)

Figure 4-1 Temperature-programmed reduction of fumed silica in N2 at atmospheric

pressure. C/SiO2 molar ratio was 4.5; temperature ramping rate was

3°C/min; N2 flow rate was 1L/min.

As shown in Figure 4-1, SiO2 conversion had already commenced at a low rate before temperature reached 1573 K (1300 °C). CO evolution rate increased with increase in temperature to 1873 K (1600 °C) at which the sample was held for one hour. Then the rate of reduction started to decrease gradually in the isothermal stage along with consumption of silica.

Isothermal experiments were conducted in the N2-H2 gas mixture (10 vol% hydrogen) at 1 atm pressure with gas flow rate 1L/min. Figure 4-2 compares the contents of CO in the off gas of different experiments at different temperatures. The conversion of SiO2 with time at different temperatures is presented in Figure 4-3.

9000

1425°C 1450°C 8000 1475°C 1500°C 7000

6000

5000

4000 CO CO content (ppm)

3000

2000

1000

0 0 60 120 180 240 300 360 420 480

Reaction time (min)

Figure 4-2 Effect of temperature on the reduction of fumed silica in 10 vol% H2 – 90

vol% N2 mixture at atmospheric pressure. C/SiO2 molar ratio was 4.5; gas flow rate was 1L/min.

100%

90%

80%

70%

60% (%)

2 50%

1425°C 40%

1450°C Conversion Conversion SiO of 30% 1475°C 20% 1500°C

10%

0% 0 1 2 3 4 5 6 7 8 9 10

Reaction time (h)

Figure 4-3 Effect of temperature on the conversion of fumed SiO2 in 10 vol% H2 – 90

vol% N2 mixture at atmospheric pressure. C/SiO2 molar ratio was 4.5; gas flow rate was 1L/min.

It can be seen from Figures 4-2 and 4-3 that the rate of SiO2 reduction increased with increasing reaction temperature. About 7 hours was necessary to obtain complete SiO2 reduction at 1698 K (1425 °C). The completion time decreased to 5 hours at 1723 K (1450 °C ). This time further decreased to approximately 3 hours and 2 hours at 1748 K (1475 °C ) and 1773 K (1500 °C), respectively.

The XRD patterns of samples reduced in isothermal experiments at 1673 K (1400 °C) – 1823 K (1550 °C) for 12 hours are presented in Figure 4-4.

Figure 4-4 XRD patterns of samples reduced in 10 vol% H2 – 90 vol% N2 mixture for

720 minutes. Fumed silica was used with C/SiO2 molar ratio of 4.5; gas flow rate was 1 L/min.

Fumed silica was not totally amorphous; weak crystallised SiO2 (quartz) peaks were observed on the XRD spectra. Silica peaks were not seen in XRD spectra of reduced samples; peaks for α-Si3N4, β-Si3N4, SiC and excess graphite were detected. α-Si3N4 was the dominant phase of silicon nitride; the fraction of β-Si3N4 in nitrides showed a tendency of decrease with increasing temperature. The relative intensity of β-SiC peak at 2θ = 35.63° was also observed to decrease with increasing temperature.

Figure 4-4 shows that the intensity of the graphite peak at 2θ = 26.50° dropped with increasing reaction temperature. Graphite was consumed by the reduction reactions and reaction with hydrogen. Alumina tubes were not dense enough to exclude oxygen penetration into the alumina reactor, which also reacted with graphite. Blank experiments demonstrated that the background CO level increased with increasing temperature.

Table 4-1 presents elemental composition of reduced samples (results of LECO analysis), weight loss of the samples during reaction, calculated extent of reduction of silica and the yields of Si3N4, SiC and SiO. The weight loss was calculated by the mass difference between initial and reacted sample pellet ( 𝑖 ) divided by the mass of initial pellet ( 𝑖):

𝑖 (4.1) 𝑖

Table 4-1 Effect of temperature on the composition of reduction product in carbothermal synthesis of silicon nitride

Temperature Weight loss O N C (°C ) (%) (wt%) (wt%) (wt%) (%) (%) (%) (%) 1400 55.0 0.66 22.8 32.5 98.9 62.7 16.9 19.3 1450 61.1 0.66 28.6 20.9 99.1 68.0 9.8 21.2 1500 68.1 0.54 31.6 8.5 99.4 61.7 14.7 23.0 1550 72.0 0.48 31.7 5.15 99.5 54.3 16.6 28.6

Weight loss increased with increasing temperature. According to the overall Reaction

(2.1) of Si3N4 formation, the theoretical weight loss in 100% conversion of silica to silicon nitride equals to 32.7% for the mixture with C/SiO2 molar ratio of 4.5. A theoretical weight loss of 49.1% corresponds to 100% conversion of SiO2 to SiC according to Reaction (2.7). The experimental results of weight loss shown in Table 4-1 were much higher than the theoretical values. Typically weight loss is due to loss of SiO,

CO, and CO2 gases purged by flowing gas. SiO loss ( ) increased with increase in temperature from 1673 K (1400 °C) to 1823 K (1550 °C). At a higher reduction temperature, SiO had a higher partial pressure due to increased SiO2 reduction and SiO formation rates resulting in more SiO escaped from a pellet by diffusion and purged away by flowing gas. Weight loss was also caused by oxidation of carbon by oxygen penetrated through alumina reactor. The carbon content dropped from 32.5% at 1673 K (1400 °C) to 5.15% at 1823 K (1550 °C), which was consistent with the decline of C peak (2θ = 26.50°) intensity shown in Figure 4-4.

The SiO2 conversion ( ) was close to 100% for all reduced samples after 12 hours reduction. This residual oxygen (< 0.7 wt%) in the reduced samples, although being attributed to SiO2 in Table 4-1, was probably from contamination of the reduced products by oxygen in air when the samples were taken from the reactor.

The nitrogen content (N wt%) increased with increasing temperature. However silicon nitride yield is not directly in proportion to the N wt% value, as the sample weight changed. The greatest Si3N4 yield of 68.0% was obtained at 1723 K (1450 °C); it decreased to 54.3% with increasing temperature to 1823 K (1550 °C), which is consistent with thermodynamic analysis (Si3N4 is less stable at high temperatures). At a lower temperature of 1673 K (1400 °C), the level of of 62.7% was slightly lower than that reached at 1723 K (1450 °C) due to slower rate of Si3N4 formation. The change in SiC yield followed a trend opposite to that of Si3N4. At 1723 K (1450 °C), the greatest yield of Si3N4 and the lowest yield of SiC of 9.8% was attained.

4.1.2 Effect of hydrogen addition

It was reported that carbothermal reduction of stable oxides of manganese, chromium 68 and titanium in hydrogen is faster than in inert gas atmospheres [Kononov, R., 2008; Dewan, M., 2009; Ostrovski, O. and Zhang, G., 2010; Rezan, S.A., 2011]. The effect of hydrogen addition on the carbothermal synthesis of silicon nitride was examined by adding H2 into N2 from 5 vol% to 50 vol%. The raw material used was fumed silica – graphite mixture with C/SiO2 molar ratio of 4.5. The experiments were carried out at

1723 K (1450 °C) at atmospheric pressure. The gas flow rate was 1L/min. The SiO2 conversion with time is presented in Figure 4-5.

100%

90%

80%

70% (%)

2 60%

50%

40% Conversion Conversion SiO of

30% 50 vol% H2 40 vol% H2

20% 30 vol% H2 20 vol% H2 10 vol% H2 5 vol% H2 10% 0 vol% H2 0% 0 30 60 90 120 150 180 210 240 270 300

Reaction time (min)

Figure 4-5 Effect of hydrogen addition in nitrogen on the conversion of fumed SiO2 at

1723 K (1450 °C ). C/SiO2 molar ratio was 4.5; gas flow rate was 1L/min.

According to Figure 4-5, addition of hydrogen significantly enhanced the reduction kinetics of SiO2. Compared to that in pure nitrogen (0 vol% H2), after 300 minutes reaction, SiO2 conversion was increased from 53% in pure nitrogen to 94% in 5 vol%

H2- 95 vol% N2 mixture, that is an increase by more than 40% by addition of 5 vol% H2.

The rate of SiO2 reduction increased with further increase in H2 content from 5 vol% to 50 vol%.

The XRD patterns of the reduced samples presented in Figure 4-5 are compared in Figure 4-6. The cristobalite peak (2θ = 21.95°) was only detected in the sample reduced in pure nitrogen. The XRD patterns of the samples reduced in hydrogen containing gases are all similar except the significant decrease in the strength of the graphite peak at 2θ = 26.50°, which can be explained by reaction of graphite with hydrogen with formation of methane. The rate of methane formation increased with increase in H2 content in the gas mixture. Weak Si3N4 peaks were detected in the sample reacted in pure nitrogen. The relative intensity of silicon nitride peaks increased with addition of 5 vol% of H2, and then did not change much with further increasing hydrogen content. α-

Si3N4 was the main phase of silicon nitride, although β-Si3N4 was also detected. β-SiC peaks at 2θ = 35.63° were observed in all the reacted samples.

The elemental composition of above samples, weight loss, extent of reduction and yield of silicon nitride and silicon carbide are presented in Table 4-2.

In Table 4-2, the extent of reduction of silica, yield of silicon nitride and weight loss in the experiment with a gas mixture containing 10% hydrogen are lower than the corresponding values of the experiment at 1723 K (1450 °C) in Table 4-1. These differences are due to the difference in reaction time; the experiments in Table 4-1 were reacted for 720 minutes while those in Table 4-2 were reacted for 300 minutes.

Figure 4-6 XRD patterns of samples reduced at 1723 K (1450 °C ) in the H2 – N2 gas mixture with different hydrogen content for 300 minutes. Fumed silica was

used with C/SiO2 molar ratio of 4.5; gas flow rate was 1L/min.

Table 4-2 Effect of H2 addition on the composition of reduction products in carbothermal synthesis of silicon nitride

H2 Weight loss O N C (vol%) (%) (wt%) (wt%) (wt%) (%) (%) (%) (%) 0 29.8 18.9 1.8 44.9 52.7 7.89 42.8 2.0 5 49.4 2.07 14.9 41.9 96.3 46.4 34.2 15.7 10 53.3 1.07 19.3 37.4 98.2 55.2 23.1 19.9 20 57.8 0.95 19.9 32.6 98.6 51.6 26.7 20.3 30 62.1 0.95 20.4 28.9 98.7 46.3 28.4 24.0 40 67.4 0.91 22.0 21.6 98.9 43.8 28.6 26.5 50 69.2 0.88 19.5 21.1 99.0 35.8 37.2 26.0

There was a significant increase in the weight loss, from 29.8 % to 49.4 % when 5 vol% of hydrogen was added into nitrogen. The weight loss kept increasing with increased H2 content. The theoretical weight loss of Reaction (2.1), 32.7%, was lower than all the experimental values with addition of H2. The experimental weight losses were even greater than the theoretical weighlt loss (49.1%) for SiC formation by Reaction (2.7).

The increased weight loss was partially attributed to the increased SiO loss ( ).

Besides SiO, CO, and CO2 purged away by flowing gas, consumption of carbon by reacting with H2 to form methane also contributed to the significant weight loss; the carbon content (C wt%) in the reduced samples decreased from 44.9% in the case of pure nitrogen to 21.1% in the case of 50 vol% H2 addition. This result is consistent with the XRD pattern demonstrated in Figure 4-6 in that the relative intensity of C peak at 2θ

= 26.50° sharply decreased with increasing H2 content. The nitrogen and silicon nitride contents in the reduced sample increased with increasing hydrogen content in the gas mixture from 0 vol % to 40 vol% and decreased when hydrogen content increased further to 50 vol%. Increasing hydrogen content in the gas mixture was beneficial to the kinetics of silica reduction, however, it also decreased the nitrogen partial pressure which was ditrimental to the formation of silicon nitride. Similar results were reported by Henry and Jack (1975). They used N2 – Ar gas to control the nitrogen partial pressure, and reported 0.5 atm of was the minimum to ensure the high Si3N4 yield.

The greatest yield of Si3N4 ( ) was obtained at 10 vol% of H2. This will be further discussed in Chapter 5.

The highest yield of silicon carbide of 42.8% was obtained in the experiment in pure nitrogen. When 10 vol% of H2 was added to the gas, the silicon carbide yield of 23.1% was the lowest. Increasing H2 content above 10 vol% resulted in increase in up to

37.2%. The best selectivity of Si3N4 to SiC ( ⁄ ) was obtained at 10 vol % of

H2 addition.

4.1.3 Effect of gas flow rate

In literature it was reported that keeping the CO partial pressure lower than ~1% by increasing gas flow rate ensured the formation of Si3N4 (Ekelund and Forslund, 1992). The effect of gas flow rate in this work was investigated in the range 0.6 to 1.4 L/min, using fumed silica – graphite mixture with the molar ratio of C to SiO2 of 4.5. The experiments were implemented at 1723K (1450°C) in nitrogen under atmospheric pressure for 360 minutes. The pellet weight was kept approximately the same in all experiments.

Figure 4-7 presents the change of CO content in the off gas with time. All the CO evolution curves have the same shape, the only difference being the height of the peaks. CO concentration in the off gas decreased with increasing gas flow rate. The peak evolution rate of CO at various gas flow rate was compared in Figure 4-8. It increased slightly from 5.0 to 5.4 ml/min with increase in gas flow rate from 0.6-0.8 L/min, then decreased with further increasing gas flow rate. This indicates that the SiO2 reduction rate was slightly promoted by greater gas flow rate till 0.8 L/min by removing CO away.

However, CO plays an important role in the reduction of SiO2 in experiments without adding H2. The SiO2 reduction rate decreased when the CO concentration was too low, which was caused by higher gas flow rate. This effect will be further discussed in Chapter 5.

The XRD patterns of samples reduced with different gas flow rate were similar. α-Si3N4 was the main phase of the synthesized silicon nitride. A strong SiC peak at 2θ = 35.63° was observed. There was no obvious difference in the carbon peak at 2θ = 26.4°. 73

9000

8000 0.6L/min 0.7L/min 7000 0.8L/min 1.0L/min

6000 1.1L/min 1.2L/min 5000 1.3L/min 4000 1.4L/min CO CO content (ppm) 3000

2000

1000

0 0 30 60 90 120 150 180 210 240 270 300 330 360

Reaction time (min)

Figure 4-7 Effect of gas flow rate on fumed silica reduction in N2 at atmospheric

pressure. C/SiO2 molar ratio was 4.5; temperature was 1723K (1450°C); reaction time was 360 minutes.

CO peak evolution rate

Peak evolution Peak evolution rate CO of (ml/min) 1

0 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

Gas flow rate (L/min)

Figure 4-8 The peak evolution rate of CO vs. gas flow rate.

4.1.4 Effect of carbon to silica ratio

The stoichiometric carbon to silica ratio is 2.0 for the overall silicon nitride synthesis Reaction (2.1). The effect of carbon to silica ratio was investigated using fumed silica mixed with graphite in the molar ratios of 2.0, 4.5, 8.0, and 12.0. All experiments were run at 1723 K (1450°C) in the 10 vol% H2 – 90 vol% N2 gas mixture at atmospheric pressure. The gas flow rate was 1 L/min. The CO evolution curves are presented in Figure 4-9.

For the same weight of raw materials with different C/SiO2 ratios, the amount of SiO2 differed significantly leading to lower CO content in the cases with higher C/SiO2 ratios.

Thus the CO evolution was related to a unit amount of SiO2 to compare SiO2 reduction rate. Figure 4-9 shows a significant increase of CO evolution rate with increasing

C/SiO2 ratio. In the experiment of stoichiometric ratio, the CO evolution peak was low and broad. The CO evolution peak became stronger and sharper with increased C

75 content; the time to reach the highest CO evolution rate decreased from 11.8 min for

C/SiO2 ratio of 2 to 10.0 minutes in the experiment with C/SiO2 equal to 12.0. SiO2 reduction rate increased with increase in carbon to silica ratio.

0.030

C/SiO2=2.0 0.025 C/SiO2=4.5

C/SiO2=8.0

0.020 C/SiO2=12.0 ·min) 2

0.015

0.010

0.005 CO CO evolution rate (mol / mol SiO

0.000 0 10 20 30 40 50 60 70 80 90 100 110 120 Reaction time (min)

Figure 4-9 Effect of carbon to silica ratio on the reduction of fumed SiO2 at 1723K

(1450°C) in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure; gas flow rate was 1L/min.

The XRD patterns of the reduced samples are presented in Figure 4-10. The elemental composition of the reduced samples, weight loss, extent of reduction and yields of silicon nitride and silicon carbide are given in Table 4-3. The extents of reduction and nitridation of experiment with C/SiO2 ratio equal to 4.5 were lower than corresponding values in similar experiments in Tables 4-1 through 4-2 due to the short reaction time, 120 minutes in Table 4-3.

In the case with stoichiometric C/SiO2 ratio equal to 2, after 120 minutes reduction, cristobalite was still the major silicon compound in the sample; the oxygen content was

31.1 wt%, and the SiO2 conversion was only 18.7%. Silica was mainly converted into β-

SiC. Although the calculated yield of silicon nitride was 4.0%, the nitride phase was practically undetectable by XRD.

Figure 4-10 XRD patterns of samples with various C to SiO2 ratio reduced at 1723K

(1450°C) in 10 vol% H2 – 90 vol% N2 gas mixture at 1 L/min for 120 min.

In the reduced sample with C/SiO2 = 4.5, only a minor cristobalite peak was detected at

2θ = 21.95°; β-SiC became stronger; the spectrum also contained peaks of α-Si3N4 and

β-Si3N4. LECO data in Table 4-3 show that the conversion of silica increased to 72.5% corresponding to the residual oxygen content of 11.9 wt% in the reduced sample. SiC 77 was still the major reduction product with a yield of 43.9%, and the Si3N4 yield was 25.9%.

Table 4-3 Effect of C to SiO2 ratio on the composition of reduction product in carbothermal synthesis of silicon nitride

C/SiO2 Weight loss O N C ratio (%) (wt%) (wt%) (wt%) (%) (%) (%) (%) 2.0 13.8 31.1 0.9 31.8 18.7 4.0 14.6 0.2 4.5 40.0 11.9 5.2 47.6 74.6 19.2 41.5 13.9 8.0 38.0 0.48 11.0 59.9 98.5 57.5 40.5 0.6 12.0 37.1 0.44 9.6 70.3 98.3 66.0 31.9 0.4

In the experiments with C/SiO2 ratio greater than 4.5, SiO2 practically was completely reduced in 120 minutes. No silica peaks were detected in the reduced samples with

C/SiO2 ratio of 8.0 and 12.0, and the detected residual oxygen content was less than 0.5 wt%. The calculated silica conversion was more than 98%; the reduction products include carbide and nitride. The nitride yield increased to 57.5 and 66.0% with corresponding decrease of SiC yield to 40.5 and 31.9% along with increasing C/SiO2 ratio to 8.0 and 12.0.

The amount of silicon lost as SiO ( ) in Table 4-3 was very low in comparison with experiments of C/SiO2 ratio of 4.5 after complete reduction. This decreased loss of SiO is attributed to faster reaction of SiO with carbon in samples with high C/SiO2 ratios. As a result, the SiO formed in reduction was almost totally converted into products silicon carbide and nitride before diffusing into bulk gas phase and being purged away from the pellets.

4.1.5 Effect of the type of silica

Two types of chemical silica were used as raw materials in this investigation, crystallised and fumed silica. Figrue 4-11 compares the CO evolution rates in reduction of different silica-graphite mixtures with C/SiO2 molar ratio of 4.5. Both experiments

78 were run at 1723K (1450°C) in a gas mixture of 10 vol% H2 – 90 vol% N2 at atmospheric pressure. The gas flow rate was 1 L/min.

Figure 4-11 Effect of the type of silica on the reduction of SiO2 at 1723K (1450°C) in

10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure. C/SiO2 ratio was 4.5; gas flow rate was 1L/min.

According to Figure 4-11, there was little difference between the reduction of the two types of silica. The difference between the times reaching peak value of CO evolution rate from both silica was about one minutes. The peak values of CO evolution rate also were about the same. CO evolution in reduction of fumed silica terminated at 300 minutes, and that of crystallised silica terminated at around 340 minutes.

The XRD patterns of the reduced samples are presented in Figure 4-12. XRD spectra of reduced samples contained peaks of silicon carbide β-SiC, silicon nitrides α-Si3N4 and

β-Si3N4 and graphite. It is noted from XRD patterns that the relative content of β-Si3N4 was higher in the sample reduced from crystallised silica.

Figure 4-12 XRD patterns of the samples with different SiO2 types reduced at

1723K (1450°C) in 10 vol% H2 – 90 vol% N2 mixture at atmospheric

pressure for 360 minutes. C/SiO2 molar ratio was 4.5; gas flow rate was 1L/min.

Table 4-4 presents the elemental composition of the reacted samples weight loss, extent of reduction and yield of silicon nitride and silicon carbide.

Table 4-4 Elemental composition of reduced samples weight loss, extent of reduction and yield of silicon nitride and silicon carbide in reduction of different types of silica

Type of Weight loss O N C silica (%) (wt%) (wt%) (wt%) (%) (%) (%) (%) Crystallised 50.3 1.81 17.1 34.7 96.8 52.0 38.6 6.2 silica Fumed silica 55.0 0.87 19.1 32.9 98.6 52.6 32.2 13.8

The conversion of fumed silica was above 98%, while crystallised silica had a slightly lower conversion. There was no obvious difference in silicon nitride yield ( ) after

360 minutes reaction. The yield of silicon carbide ( ) in reduction of fumed silica was 6.4% lower than that of crystallised silica. The Si3N4 to SiC ratio ( ⁄ ) was greater in the reduction of fumed silica. The majority of experiments in investigation of carbothermal synthesis of silicon nitride were carried out using fumed silica.

The weight loss in the experiment with fumed silica (55.0%) was slightly greater than that with crystallised silica (50.3%). This difference can be attributed to two factors.

Firstly, the SiO2 conversion ( ) was higher with fumed silica (98.6%) than with crystallised silica (96.8%), which resulted in higher consumption of C and removal of oxygen as gaseous CO. Secondly, there was higher SiO loss in the experiment with fumed silica ( =13.8%) than that with crystallised silica (6.2%).

The significant difference in SiO loss between fumed and crystallised silica was due to the difference of porosity between the pellets of silica-graphite mixtures with different types of silica. The significant difference in the specific surface area (SSA) between these two types of silica leads to difference in specific surface area of the SiO2 – C mixtures. As shown in Table 3-1, the SSA of the crystallised silica - graphite mixture 2 with C/SiO2 molar ratio of 4.5 equals to 8.36 m /g, while similar fumed silica - graphite mixture has a SSA of 128.2 m2/g. Fumed silica is highly porous and of low density. A greater diffusion rate of SiO gas species is expected from the reaction sites within pellets to the bulk gas flow for fumed silica compared to crystallised silica.

It should be noted that fumed silica started to crystallise to cristobalite above 847K (574°C). Furthermore, sintering also took place which decreased the specific surface area of the reactant. This explains why the reaction behaviour of fumed silica was not significantly different from that of crystallised silica.

Crystallised silica was also subjected to phase transformation during heating and reduction which was studied in reduction experiments in the temperatures range from 1673K (1400°C) to 1773K (1500°C). The experiments were conducted in pure nitrogen at atmospheric pressure for 120 min. The gas flow rate was 1L/min. The molar ratio of

C/SiO2 was 4.5. XRD patterns of the reacted samples are presented in Figure 4-13.

As shown in Figure 4-13, the crystallised silica in the sample before reduction was identified as quartz. After 120 minutes reduction at 1673K (1400°C), the quartz peak significantly decreased, accompanying with the appearance of a cristobalite peak at 21.95°, showing the transformation of quartz to cristobalite. The quartz peak further decreased with increasing reduction temperature and finally disappeared at 1773K (1500°C). The intensity of the cristobalite peak increased with increasing temperature to 1723K (1450 °C), and then decreased with further heating to 1773K (1500 °C) due to enhanced reduction.

Silica was reduced to silicon carbide and silicon nitride β-Si3N4. The product peaks were weak at all the reaction temperatures due to the short reaction time. XRD peaks of

α-Si3N4 were not observed in Figure 4-13, because of their relative low strength in comparison with those of β-Si3N4 when crystallised silica is used, as shown in Figure 4- 12.

Figure 4-13 XRD patterns of the crystallised silica reduced in pure N2 for 120 minutes.

The C/SiO2 molar ratio was 4.5; the gas flow rate was 1L/min.

4.1.6 Effect of silicon nitride addition

Silicon nitride nucleation can affect the rate of the carbothermal synthesis of silicon nitride. In this case, it can be expected that addition of Si3N4 will accelerate the synthesis process. The effect of addition of Si3N4 on the synthesis of silicon nitride was investigated in reduction of SiO2-C- Si3N4 mixture with C:SiO2:Si3N4 ratio of 4.5:1:0.05.

The fumed silica was reduced at 1723K (1450°C) in 10 vol% H2 – 90 vol% N2 gas mixture at atmospheric pressure. The gas flow rate was 1 L/min. Reaction time was 360 minutes. The CO evolution rates are presented in Figure 4-14.

0.008

seeding 0.007 no seeding

0.006

·min) 0.005 2

0.004

0.003

0.002

CO CO evolution rate (mol/mol SiO 0.001

0.000 0 30 60 90 120 150 180 210 240 270 300 330

Reaction time ( min)

Figuer 4-14 Effct of Si3N4 addition on the reduction of fumed SiO2 at 1723K (1450°C)

in 10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure. Gas flow

rate was 1 L/min; the C/SiO2 molar ratio was 4.5 without seeding; the

molar C/SiO2/Si3N4 ratio in the seed-containing mixture was 4.5/1/0.05; reaction time was 360 minutes.

SiO2 reduction rate was promoted by Si3N4 seeding. For unit amount of SiO2, CO evolution rate in reduction of seeded mixture was greater than that of non-seeded 84 mixture. Correspondingly, the time to reach complete reduction of silica became shorter as shown in Figure 4-14.

The XRD patterns of the reduced samples with and without addition of silicon nitride, shown in Figure 4-15, are similar. From relative peak strength, the reduced sample without seeding contained more silicon carbide.

Figure 4-15 XRD patterns of the reacted samples with and without addition of silicon nitride. Reaction was carried out at 1723K (1450°C) in 1 L/min 10 vol%

H2-90 vol% N2 mixture for 360 min; C/SiO2 molar ratio was 4.5 without

seeding; C/SiO2/Si3N4 was 4.5/1/0.05 in the mixture with seeding.

The experiments with materials with and without addition of silicon nitride were stopped at different reaction times. The reduced samples were analysed by LECO. 85

Calculated conversion of silica, yields of silicon carbide and nitride are shown in Figure 4-16.

100%

90%

80%

70%

(%)

SiC 60% / y

4 Si3N4 yield seeding N 3 Si 50% Si3N4 yield no seeding / y

2 SiC yield seeding SiO

x 40% SiC yield no seeding

30% SiO2 conversion seeding SiO2 conversion no seeding 20%

10%

0% 0 60 120 180 240 300 360 420 480 540 600 660 720 780 840 900

Reaction time (min)

Figure 4-16 Effect of Si3N4 seeding on the composition of reduction product in carbothermal synthesis of silicon nitride conducted at 1723K (1450°C) in

10 vol% H2 – 90 vol% N2 mixture at atmospheric pressure. Gas flow rate

was 1 L/min; C/SiO2 molar ratio was 4.5 without seeding; C/SiO2/Si3N4 in the mixture with seeding was 4.5/1/0.05.

In the experiments with seeded raw material, SiO2 was completely reduced in approximate 240 minutes, while reduction of silica without seeding took about 300 minutes.

In Figure 4-16, the SiC yield increased at approximately the same rate within the first 120 minutes of reduction. Then, SiC content commenced to decrease for both cases. However, the SiC content in the seeded sample decreased much faster than that without seeding. It is also noted that the silicon nitride yield curves deviated from the beginning of reduction. The above results indicate that seeding promoted formation of silicon nitride by reduction as well as conversion of SiC to Si3N4.

4.1.7 Effect of nitrogen pressure

It can be expected from the silicon nitride synthesis Reaction (2.1), that increasing nitrogen pressure promotes silicon nitride formation. It is also expected that the Si3N4 formation temperature can be raised, since the stability of Si3N4 will increase with increasing nitrogen pressure. In this work, effect of nitrogen pressure on carbothermal synthesis of silicon nitride was investigated in the range 700-1100 kPa (7-11 atm).

Experiments with fumed silica mixed with graphite in the molar ratio of C/SiO2 of 4.5 were carried out in nitrogen with the gas flow rate of 1L/min. Experiments were conducted in the temperature range of 1723K (1450°C) to 1923K (1650°C) for 60 minutes. The XRD patterns of samples reduced at 1100 kPa nitrogen pressure are presented in Figure 4-17. Table 4-5 shows the elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide.

The reduction rate of SiO2 increased with increasing temperature. The XRD pattern of the reduced sample at 1723K (1450°C) shows a strong cristobalite peak at 2θ = 21.95°, which decreased with increasing temperature, and totally disappeared at 1873K (1600°C). According to Table 4-5, the sample reduced at 1723K (1450 °C) contained 19.3 wt% of oxygen which corresponds to an extent of reduction of 39.3%. The oxygen content decreased to 13.1 wt%, while the extent of reduction increased to 64.1% at 1773K (1500 °C). Further increasing temperature to 1823K (1550 °C) decreased oxygen content to 2.0 wt%, and silica conversion increased to 95.3%. The data also confirmed that the conversion of silica was complete in 60 minutes at 1873K (1600°C) and 1923K (1650 °C).

Figure 4-17 XRD patterns of fumed silica reduced in pure N2 at 1100kPa for 60

minutes. The molar C/SiO2 ratio was 4.5; gas flow rate was 1L/min.

Table 4-5 Elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide in carbothermal synthesis of silicon nitride in nitrogen at 1100 kPa. Weight Temperature O N C loss (°C) (wt%) (wt%) (wt%) (%) (%) (%) (%) (%) 1450 18.5 19.3 4.4 49.5 39.3 23.6 12.5 3.2 1500 29.3 13.1 8.0 48.5 64.1 37.3 21.7 5.1 1550 40.0 2.0 17.4 47.0 95.3 69.2 14.9 11.2 1600 42.7 0.30 18.8 47.1 99.3 71.4 13.3 14.6 1650 45.5 0.33 4.1 56.2 99.4 13.3 70.2 15.9

Both α-Si3N4 and β-Si3N4 were detected in the reduced samples by XRD, with α-Si3N4 being the main phase of synthesised silicon nitride. Silicon carbide was in the form β-

SiC. Si3N4 peaks were observed in all experiments after 60 minutes reaction; their strength increased with increasing temperature from 1723K (1450°C) to 1873K (1600°C). The SiC peak decreased with increasing temperature from 1723K (1450°C) to

1873K (1600°C). At 1923K (1650°C), however, the Si3N4 peaks were significantly weaker, while SiC peak became strong.

The yield of silicon nitride was 23.6% in reduction at 1450°C (Table 4-5). It increased to 37.3% at 1500 °C, 69.2% at 1550 °C, and reached the maximum (71.4%) at 1873K (1600°C). The yield of SiC at 1450 °C was relatively low, 12.5% due to a low silica conversion. It increased to 21.7% at 1500 °C, however, decreased with further increasing temperature until 1600 °C at which the yield was 13.3%. Further increasing the reaction temperature to 1923K (1650°C) reversed the above trend. Up to 70.2% of Si was converted to SiC, while silicon nitride yield was only 13.3%.

The weight loss in reduction increased from 18.5% to 45.5% with increasing temperature from 1723K (1450°C) to 1923K (1650°C). The loss of SiO also increased from 3.2% to 15.9%. Increase in weight loss was consistent with greater SiO2 reduction rate and SiO2 conversion ( ) caused by increasing reaction temperature.

The effect of nitrogen pressure ( ) in the range of 700 to 1100 kPa was investigated at 1873K (1600°C). The XRD patterns of reacted samples are presented in Figure 4-18. Table 4-6 shows elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide.

Figure 4-18 XRD patterns of the samples reduced 1873K (1600°C) in nitrogen at 700-

1100 kPa pressure for 60 minutes. Fumed silica was used with C/SiO2 molar ratio of 4.5; gas flow rate was 1 L/min.

Reduction of silica was complete for all the nitrogen pressures used. The major difference between the samples reduced under different nitrogen pressures was the 90 distribution of silicon between nitride and carbide: the yield of silicon nitride decreased from 71.4% to 43.0% while the yield of silicon carbide increased from 13.3% to 45.4% when the nitrogen pressure was decreased from 1100kPa to 700kPa. The greatest selectivity of Si3N4 to SiC was obtained in the case of 1100kPa of nitrogen pressure.

Table 4-6 Elemental composition, weight loss, extent of reduction and yields of silicon nitride and silicon carbide in synthesis of silicon nitride in nitrogen at 700- 1100 kPa pressure. Weight O N C loss (kPa) (wt%) (wt%) (wt%) (%) (%) (%) (%) (%) 700 46.1 0.37 12.0 50.1 99.2 43.0 45.4 10.8 900 43.1 0.27 16.2 48.8 99.4 61.0 25.5 12.9 1100 42.7 0.30 18.8 47.1 99.3 71.4 13.3 14.6

The weight loss slightly decreased from 46.1% to 42.7% while the SiO loss increased from 10.8% to 14.6% with increase in nitrogen pressure from 700 kPa to 1100 kPa. The decreased weight loss along with increasing nitrogen pressure was attributed to higher nitride yield at higher pressure which increased the content of nitrogen in the reduced sample.

Increase of SiO loss can be related to the increased CO partial pressure with increasing total pressure. This factor suppresses the reduction of SiO vapour to final products and so increased the fraction of SiO lost in the off gas.

4.2 Decomposition of silicon nitride

As suggested in this project, synthesis of silicon nitride by carbothermal reduction and nitridation followed by decomposition can be an attractive way to produce solar grade silica. This section presents preliminary results obtained in a study of decomposition of high purity chemical Si3N4.

( ) ( ) ( ) (4.2)

Decomposition of silicon nitride was studied at different temperatures in argon and hydrogen.

4.2.1 Effect of temperature on the decomposition of silicon nitride

The effect of temperature on the decomposition of silicon nitride was studied in argon at 1823K and 1873K (1550°C and 1600°C). The gas flow rate was 1L/min; duration of experiments was 240 minutes. Based on the appearance of samples after reaction, only top layer of Si3N4 was decomposed. The reacted samples were ground to form uniform mixtures, and then analysed by X-ray diffraction. The XRD patterns of original Si3N4 samples and samples after decomposition are presented in Figure 4-19.

The original chemical silicon nitride was mainly α-Si3N4. Small amount of β-Si3N4 peaks were observed. After 240 minutes reaction, α-Si3N4 still was the main phase, although β-Si3N4 peaks were enhanced due to the high temperature reaction (phase transformation took place as β-Si3N4 is more stable than the α phase). Silicon peaks were also detected which were much stronger in the sample decomposed at 1873K (1600 °C) in comparison with the sample decomposed at 1823K (1550 °C).

The extent of decomposition ( ) defined by Equation (3.13) in CHAPTER 3, was calculated based on LECO nitrogen analysis and presented in Table 4-7. increased from 3.86% to 14.22% when the reaction temperature was increased from 1823K (1550°C) to 1873K (1600°C).

Figure 4-19 XRD patterns of Si3N4 decomposed in Ar for 240 minutes. The gas flow rate was 1L/min; the reaction temperature was 1823K and 1873K (1500°C and 1600°C).

Table 4-7 Effect of temperature on Si3N4 decomposition

Temperature (°C) N ( ) ( )

Si3N4 41.50 - 1550 39.90 3.86 1600 35.60 14.22

4.2.2 Silicon nitride decomposition in different gas atmospheres

Nitrogen is released as gaseous product of silicon nitride decomposition according to Reaction (4.2). Based on the observation that decomposition took place in the surface layer of the sample powder, it can be expected that internal diffusion of nitrogen out of the packed silicon nitride powder can contribute to the decomposition rate control.

Therefore, the diffusivity of nitrogen is important to removal of N2. Effect of gas atmospheres (argon and hydrogen) on silicon nitride decomposition was examined at 1873 K (1600°C). Gas flow rate was 1L/min; reaction time was 240 minutes.

The XRD patterns of Si3N4 samples after reaction in Ar and H2 are presented in Figure 4-20. The intensity of Si peaks in the sample decomposed in hydrogen was significantly higher than the sample decomposed in Ar. The molecular weight of H2 is less than that of Ar, resulting in a greater diffusivity of N2 in H2 gas atmosphere than in argon. In the sample decomposed in H2, the residual silicon nitride was present almost totally as β-

Si3N4; α-Si3N4 peaks were very weak.

Table 4-8 shows the elemental composition of the samples presented in Figure 4-20.

More than half of Si3N4 (51.3%) was decomposed in H2 after 240 minutes, which was a significant improvement compared to decomposition in Ar. The result is consistent with the expectation that high N2 gas diffusion in hydrogen promotes Si3N4 decomposition by keeping a low N2 partial pressure within the layer of nitride powder.

Figure 4-20 Comparison of the XRD patterns of Si3N4 decomposed in Ar and H2 for 240 minutes. The reaction temperature was 1873K (1600°C); gas flow rate was 1L/min.

Table 4-8 LECO nitrogen analysis and extent of decomposition of silicon nitride in different gas atmosphere

Sample N, ( ) , ( )

Chemical Si3N4 41.50 - Decomposed in Ar 35.60 14.2

Decomposed in H2 20.20 51.3

The morphology of samples decomposed in Ar and H2 are presented in Figure 4-21.

Different magnifications were selected, since the residual Si3N4 and produced Si particle sizes varied at different conditions. The melting point of silicon is 1687K (1414°C). Si obtained from decomposition of Si3N4 was liquid at the reaction temperature. The liquid Si solidified into solid particles when decomposition was stopped by decreasing reaction temperature. Si3N4 particles were small and uniform before decomposition (Figure 4-20(a)). After 240 minutes decomposition, Si particles with various sizes can be observed. The Si particles were covered by residual Si3N4 particles. Larger Si particles were observed in samples decomposed in H2.

Si3N4

(a) (b)

Si Si3N4

(c)

Figure 4-21 The morphology of Si3N4 decomposed in different gas atmospheres at

1873K (1600°C). (a) Chemical Si3N4, (b) decomposed in Ar, (c)

decomposed in H2. Reaction time was 240 minutes; gas flow rate was 1L/min.

Figure 4-22 presents the morphology of the Si3N4 sample decomposed in H2 at 1873K (1600°C) for 240 minutes. The EDS spectra at points 26 and 29 were also presented in the figure. Large particles represented by points 26 and 27 were confirmed to be elemental Si, while small particles represented by points 28 and 29, containing both N and Si, were confirmed to be residual Si3N4. The weak Cr peak in each spectrum came from the chromium coating on the sample surface.

(a)

(b)

(c)

Figure 4-22 An image of a Si3N4 sample (a) decomposed in H2 (1 L/min) at 1873K (1600°C) for 240 minutes and EDS spectra at point 26 (b) and point 29 (c).

The samples were also mounted in resin, polished, and then analysed by EDS mapping. Figures 4-23 and 4-24 show the elemental distribution of Si, C and N in particles of different size. They clearly show that fine particles of undecomposed Si3N4 were included in the Si particles. These included Si3N4 were more difficult to decompose further, as molten Si blocked nitrogen removal from the particles.

(a) (b)

(e)

Figure 4-23 EDS mapping analysis of relatively small particles in the Si3N4 sample

decomposed in H2 at 1873K (1600°C) for 240 minutes. The gas flow rate

was 1L/min. (a) BSE image of decomposed Si3N4 sample; (b) mapping of all selected elements, (c) Si, (d) C, (e) N.

100

(a) (b)

(e)

Figure 4-24 EDS mapping analysis of a large particle in the Si3N4 sample decomposed

in H2 at 1873K (1600°C) for 240 minutes. The gas flow rate was 1L/min.

(a) BSE image of decomposed Si3N4 sample; (b) mapping of all selected elements, (c) Si, (d) C, (e) N.

101

CHAPTER 5 DISCUSSION

5.1 Thermodynamic analysis of carbothermal synthesis of silicon nitride

In carbothermal synthesis of silicon nitride, both silicon nitride and carbide are formed.

The standard Gibbs free energy changes for the overall reactions of Si3N4 and SiC formation Reactions (2.1) and (2.7) calculated using data from Chase (1998) are presented by Equations (5.1) and (5.2), respectively.

( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

According to Equation (5.1), the standard Gibbs free energy change for Reaction (2.1) is equal to zero at 1820K (1547°C). At a higher temperature, conversion of silica to silicon nitride is expected to be spontaneous at 퐶𝑂=1 atm and =1 atm (standard state). In practice, the CO partial pressure ( 퐶𝑂) is much lower than the standard pressure (1 atm). Thus Reaction (2.1) can start at a lower temperature. According to results obtained in the temperature-programmed reduction of fumed silica presented in Figure 4-1, SiO2 conversion to Si3N4 commences before reaction temperature achieves 1573K (1300°C). Similarly, reduction of silica to silicon carbide (Reaction (2.7)) under standard o conditions ( 퐶𝑂 =1 atm) spontaneously starts at 1778 K (1515 C). Actually, silicon carbide can be synthesized at a lower temperature since the CO partial pressure is lower than 1 atm.

The equilibrium CO partial pressures in Reactions (2.1) and (2.7) are compared in

Figure 5-1. Since the gas phase in majority of experiments contained 10 vol% H2,

=0.9 atm was selected in calculations of CO partial pressure. Calculations were conducted in the temperature range from 1673K (1400°C) to 1873K (1600°C), which corresponds to experimental conditions.

102

Reaction (2.1) with PN2=0.9 atm

Reaction (2.7)

1 Pco(atm)

0.1 1650 1700 1750 1800 1850 1900

Temperature (K)

Figure 5-1 Equilibrium CO partial pressures vs temperature in Reactions of formation of silicon nitride (Reaction (2.1)) and silicon carbide (Reaction (2.7)).

The equilibrium CO partial pressures of both reactions are significant for the atmospheric total pressure. At temperatures below 1433 °C, calculated 퐶𝑂( ) is slightly higher than 퐶𝑂( ) . It means that at these temperatures Si3N4 is thermodynamically a more favourable product than SiC, although the difference in equilibrium partial pressure of CO is only 0.0313 atm at 1400 °C. At temperatures above 1433 °C, 퐶𝑂( ) is smaller than 퐶𝑂( ), indicating that SiC is more favourable to form than Si3N4 at these temperatures. At 1450 °C, 퐶𝑂( ) is by 0.0281 atm higher than 퐶𝑂( ). Compared to equilibrium CO partial pressures for Reactions (2.1) and

(2.7), the difference between 퐶𝑂( ) and 퐶𝑂( ) is quite small. This explains that there was not a sharp boundary temperature for a preferential formation of Si3N4 or SiC under given experimental conditions. At 1550 °C, although 퐶𝑂( ) = 0.989 atm is smaller than 퐶𝑂( ) = 1.46 atm, the potential to form Si3N4 is still significant. It should be noted that the actual CO partial pressure cannot be higher than 1 atm for the reaction 103 conducted under atmospheric pressure. The equilibrium CO partial pressure higher than 1 atm indicates a strong thermodynamic potential for reaction to occur.

SiO2 reduction starts with formation of SiO by solid-solid Reaction (2.2). At the contact points between C and SiO2, reduction takes place resulting in gaseous intermediate SiO and CO. The standard Gibbs free energy change for this reaction (Equation (5.3)) equals to zero at 2024K (1751°C). At higher temperature, the reduction of silica by graphite will be spontaneous at 퐶𝑂 = 1 atm and 𝑖𝑂 = 1 atm (standard state). In practice, the reduction takes place at a lower temperature, since CO and SiO partial pressures are much lower than unity atm. As it has been mentioned before, CO was detected when the reaction temperature reached 1573K (1300°C). It means Reaction (2.2) has already started at this low temperature.

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

The reduction of SiO2 by C at the contact points has its significant potential in the relative short reaction time.

Further reduction of SiO2 proceeds by CO as presented by Reaction (2.3). CO is regenerated by the Boudouard reaction (2.4).

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( ) ( )

The CO2 equilibrium partial pressure is calculated by Reaction (2.4) using experimental value of CO partial pressure in experiments at different temperatures, and is presented in Table 5.1. Then the equilibrium 𝑖𝑂 can be calculated by Equation (5.6)

( ) 퐶𝑂 ( ) 𝑖𝑂 퐶𝑂

104

Table 5-1 Equilibrium CO2 and SiO partial pressures for Reaction (2.3) at various reaction temperatures calculated using experimental CO concentrations in the off gas

Temperature(K) CO concentration(ppm) (atm) (atm) (atm) 1673 4020 0.0040 0.0604 2.78×10-9 1723 7042 0.0070 0.1390 6.07×10-9 1773 9275 0.0093 0.3899 7.78×10-9 1823 12719 0.0127 0.9896 1.07×10-8

In the experiment at 1723K (1450°C), the maximum CO partial pressure of 0.007 atm was observed; the equilibrium SiO partial pressure at this temperature was 0.139 atm, which shows a high potential for the reaction to take place. The equilibrium 𝑖𝑂 increased with increasing temperature.

5.2 Phase development in silicon nitride synthesis

As it was discussed above, there is no significant difference in the equilibrium CO partial pressure between reactions of formation of Si3N4 and SiC. This explains that SiC peaks were detected in the sample reduced at 1673K (1400°C) and Si3N4 peaks were observed in the sample produced in experiment at 1823K (1550°C) for 720 minutes (Figure 4-2 and Table 4-1). For shorter reaction time of 300 minutes SiC phase dominated in all cases (Figure 4-4). Regardless a type of silica, SiC was the main product after 360 minutes reaction (Figure 4-10). Even weaker Si3N4 peaks were detected after 120 minutes reaction (Figure 4-11). These results indicate that SiC formation cannot be avoided by controlling temperature under given experimental conditions. In the early stage of carbothermal synthesis of silicon nitride, SiC dominates the reaction product; by extending reaction (nitridation) time, Si3N4 yield can be enhanced.

105

A series of experiments was designed to explore the phase evolution during carbothermal reduction/nitridation process. Fumed silica was mixed with graphite mixture with C to SiO2 molar ratio of 4.5. Experiments were conducted at 1723 K

(1450°C) in 10 vol% H2 – 90 vol% N2 gas mixture with the total flow rate of 1 L/min. Samples obtained in the reduction/nitridation process for different time were subjected to XRD analysis, which results are presented in Figure 5-2. In the early stage, both α-

Si3N4 and β-Si3N4 peaks were detected, but β-SiC was the main crystal phase in the reaction. Reduction was incomplete. SiC peaks strengthened with increasing reaction time till about 360 minutes and then decreased with further extension of reaction to 840 minutes, while Si3N4 peak intensity kept increasing during the whole reduction/nitridation process.

Figure 5-3 presents the silicon distribution in each phase of reaction at which the reaction products are shown in Figure 5-2. In 300 minutes reaction, SiO2 was almost depleted. Si3N4 was formed from the begnning and kept increasing after the consumption of SiO2. At the end of 840 minutes reaction, 82.4% of Si was converted into Si3N4. SiC was synthesized from the commencement of reaction. Formation rate of

SiC prevailed over Si3N4 untill approximately 120 minutes. Then content of SiC started to drop amost to zero at the end of 840 minute reaction. It is apparent that SiC is the only source of Si contributing to the Si3N4 formation after the depletion of SiO2.

Therefore, conversion of SiC to Si3N4 can be presented by Reaction (5.7). It is thermodynamically favoured at lower temperature, which explains the greatest Si3N4 selectivity obtained at 1723K (1450°C) (Table 4-1). At 1673K (1400°C), complete SiO2 reduction requires longer reaction time, which leads to shorter residual time for SiC to

Si3N4 conversion. At temperatures higher than 1723K (1450°C), SiO2 reduction is accelerated by the increased temperature, but the conversion of SiC to Si3N4 is slower leading to lower Si3N4 to SiC ratio.

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

106

Figure 5-2 XRD patterns of fumed silica reduced in the 10 vol% H2 – 90 vol% N2 gas mixture for various reaction times at 1723K (1450°C).

107

100%

90%

80%

70%

60% /SiC /SiC (%) 4 N 3 50% /Si 2 Si-SiO2 % 40%

Si in Si in SiO Si-Si3N4 %

30% Si-SiC %

20%

10%

0% 0 60 120 180 240 300 360 420 480 540 600 660 720 780 840 Reaction time (min)

Figure 5-3 Silicon in SiO2, Si3N4, and SiC in reduction/nitridation of fumed silica in the

10 vol% H2 – 90 vol% N2 gas mixture at 1723K (1450°C).

It should be clarified that there are two common SiC crystal structures – α-SiC and β- SiC. The SiC crystal structure involved in the calculation was the cubic β-SiC, also called moissanite, since all synthesized SiC was in this form according to the XRD analysis. This experimental observation is consistent with thermodynamic data on α-SiC - β-SiC transformation provided by the NIST-JANAF thermochemical tables (pages 648-649) (Chase, 1998), which indicate that β-SiC is more stable up to 2000 K (1727°C).

108

Si3N4 exists in two hexagonal crystal structures – α-Si3N4 and β-Si3N4 (Turkdogan,

1958). The α-Si3N4 is considered to be a metastable low-temperature phase which transforms to β-Si3N4 (stable modification) at higher temperatures. The formation and stability of α-Si3N4 was largely controlled by kinetic factors (Riley, 2000). The monotropic transformation temperature from α to β phase has been reported to be in the range of 1573-1723 K (Forgeng, 1958). Thompson (1967) and Priest (1973) suggested that the transformation of α-Si3N4 to β-Si3N4 was impurity controlled, and later shown readily to take place in the presence of a liquid phase through a reconstructive transformation (Brook, 1978). It was reported in literature (Riley, 1989; Dijen, 1992; Weimer, 1997) that the carbothermal reduction/nitridation process produces high α-

Si3N4 content. The experimental data obtained in this work also indicate predominance of the α-Si3N4 phase, which is in agreement with literature.

Hillert et al. (1992) predicted that silicon oxynitride Si2N2O is a stable silicon compound at 1723K (1450°C) with nitrogen partial pressure of 0.5 atm when the oxygen partial pressure is above 10-19 atm. Under experimental conditions with surplus carbon, oxygen partial pressure was lower than 10-19 atm. This is why no silicon oxynitride phase was observed. Moreover, in the presence of carbon, silicon carbide was formed which was a more stable compound than oxynitride.

5.3 Effect of gas atmosphere and flow rate on reduction/nitridation process

Carbothermal reduction of oxides proceeds through the gas phase. Thus, Reaction (2.7) for SiC formation can be presented as a sum of Reactions (5.9) and (5.11):

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( ) ( )

( ) ( ) ( )

Reduction of silica involves carbon transfer from a graphite particle to silica in the form of CO and oxygen transfer from silica to graphite in the form of CO2. Equilibrium 109 partial pressure of CO2 for Reaction (5.9) is very low, for 퐶𝑂 = 0.1 atm, at temperatures -6 1450-1600°C, it is in the range (3.5-4.2)×10 atm. At such low CO2 partial pressure, the reduction rate can be controlled by the mass transfer of CO2 from silica to graphite and chemical reaction of carbon gasification (Kononov, 2008; Dewan, 2009; Ostrovski and Zhang, 2010; Rezan, 2011).

As presented in Figure 4-8, the rate of SiO2 reduction in nitrogen was slightly affected by the gas flow rate. The peak SiO2 reduction rate appeared at 0.8 L/min which can be considered the optimal flow rate under fixed other conditions. This phenomenon can be explained by above mechanism of silica reduction. According to Reaction (5.9), CO, as an actual reductant of silica, has a positive effect on the reduction rate. This requires that CO has a high partial pressure in the reaction system, and this can be achieved by reducing the gas flow rate of nitrogen. However, if the CO concentration is approaching its equilibrium, the driving force of the reaction will be too small and the reduction will stop. As a result, the reaction system needs to maintain an appropriate CO partial pressure which provides proper driving force (CO content is low enough from equilibrium value) and also high enough reductant content.

Addition of hydrogen to the gas phase can promote diffusion of gas species and enhance the reaction kinetics. However, this effect is small and cannot explain significant increase of reduction rate when a small fraction of hydrogen was added (Fig. 4-5). H2 addition can change the reduction mechanism. Hydrogen can be directly involved into silica reduction according to Reactions (5.13) and (5.15).

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

Equilibrium H2O partial pressure for Reaction (5.13) increases with increasing temperature and hydrogen partial pressure and decreasing SiO partial pressure. Assuming SiO partial pressure in the range from 0.01 to 0.1 atm and fixing the

110

-6 hydrogen partial pressure at 0.1 atm, equilibrium pH2O is in the range of 3 x 10 to 3 x -5 10 atm at 1550 °C, which is comparable with CO2 partial pressure for Reaction (5.9).

Hydrogen reacting with carbon forms methane by Reaction (5.17) which can reduce silica by Reaction (5.19):

( ) ( ) ( ) ( )

( ) ( ) ( )

( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

The equilibrium 퐶 of Reaction (5.17) increases with decreasing temperature and increasing hydrogen partial pressure. In the experimental temperature range, calculated -4 methane partial pressure ( =0.5 atm) is (1-2)x10 atm, which is significantly higher

(by more than one order in magnitude) than the equilibrium 퐶𝑂 for Reaction (5.9) and pH2O for Reaction (5.13).

H2 addition reduces N2 partial pressure when the total pressure is kept constant. The nitrogen partial pressure is a significant factor for Si3N4 formation. It is seen in Table 2 that the highest Si3N4 yield and selectivity of Si3N4 to SiC were obtained with 10 vol% of H2 addition. H2 content higher than 10 vol% led to decreased nitrogen partial pressure. The SiC yield increases while Si3N4 yield decreases with increasing H2 content in gas beyond 10 vol%.

The equilibrium nitrogen partial pressure ( ) for Reaction (5.7) of nitridation of SiC to Si3N4 at different temperatures can be calculated from the standard Gibbs free energy change for this reaction presented by Equation (5.8), by combining Equations (5.21) and (5.22).

( ⁄ ) ( )

( )

111

The equilibrium for Reaction (5.7) as a function of temperature is plotted in Figure 5-4.

The equilibrium increases with increase in temperature, meaning that conversion of

SiC to Si3N4 by Reaction (5.7) is easier at lower temperature. At 1673K (1400°C),

Reaction (5.7) spontaneously starts when the is higher than 0.64 atm. When temperature increases to 1923K (1650°C), the has to be greater than 6.92 atm to ensure Si3N4 formation by Reaction (5.7). At the experimental temperature 1723K

(1450°C), the equilibrium of 1.08 atm is required for formation of silicon nitride.

Nevertheless, Si3N4 was detected in the samples produced at this temperature, as shown in Figures 4.4, 4.5 and Table 4-2. A possible reason is that the thermodynamic data are not accurate at high temperatures. The data used for the plot in Figure 5-4 were taken from Chase (1998). A different plot for the equilibrium vs temperature was calculated using data of Binnewies and Milke (2002) which is shown in Figure 5-5.

(atm)

4 2 N P 3

0 1650 1700 1750 1800 1850 1900 1950

Temperature (K)

Figure 5-4 Equilibrium nitrogen partial pressure for conversion of SiC to Si3N4 by

112

Reaction (5.7) calculated using data from Chase (1998).

There is a significant difference in the equilibrium nitrogen partial pressures using these two sources of thermodynamic data. According to Binnewies and Milk (2002) the

Reaction (5.7) of conversion of SiC to Si3N4 is at equilibrium at 1723K (1450°C) when

is 0.42 atm. The equilibrium at 1773K (1500°C) is about 0.7 atm. This explains o observations that Si3N4 was formed at temperatures above 1723K (1450 C) as presented in Figure 4-3 and Table 4-2.

Another effect of addition of hydrogen to nitrogen can be activation of conversion of silicon carbide to silicon nitride by the following nitridation reaction.

( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( )

3.0

2.5

2.0

(atm)

2 1.5 N P

1.0

0.5

0.0 1650 1700 1750 1800 1850 1900 1950

Temperature (K)

113

Figure 5-5 Equilibrium nitrogen partial pressure for SiC to Si3N4 conversion by Reaction (5.7) calculated using data from Binnewies and Milke (2002).

The effect of nitrogen partial pressure on formation of silicon nitride was investigated by raising nitrogen pressure up to 11 atm (1100kPa) at different temperatures which results are presented in Section 4.1.7. The rate of carbothermal Si3N4 synthesis was significantly improved by raising nitrogen pressure, as shown in Table 4-6. The key point is that the elevated nitrogen pressure increases the Si3N4 stability at high temperatures (Figures 5-4 and 5-5); therefore, silicon nitride at elevated nitrogen pressure can be synthesised at higher temperatures than at 1 atm. Increase in temperature accelerated the SiO2 reduction and shortened the Si3N4 formation time.

At 1923K (1650°C), has to be above 7 atm (700 kPa) according to Figure 5-4 (2.7 atm according to Figure 5-5) to ensure the conversion of SiC to Si3N4. However, increase in the temperature from 1873K (1600 °C) to 1923K (1650°C) decreased the

Si3N4 yield as shown in Table 4-6. This reduction of nitridation rate with increasing temperature is most likely due to reduced driving force or the difference between actual

N2 pressure of 11 atm and equilibrium N2 pressure.

5.4 Effects of C/SiO2 ratio and Si3N4 seeding on reduction/nitridation process

According to the overall Reaction (2.1) of the carbothermal synthesis of silicon nitride, stoichiometric carbon to silica molar ratio is 2. Excess carbon is always required to achieve complete nitridation. As shown in Figures 4-7, 4-8 and Table 4-4, excess carbon promoted reduction by increasing the contact area between SiO2 and C, which directly contributes to the SiO2 reduction at the early stage of the process.

Improvement of Si3N4 synthesis by increasing C/SiO2 ratio is in agreement with literature data reported by Komeya and Inoue (1975) and Zhang and Cannon (1982). Komeya and Inoue (1975) and Zhang and Cannon (1982) assumed the nucleation of SiC on the C surface was the rate limiting-step in the silicon carbide synthesis.

114

Although Si3N4 yield is enhanced by excess carbon, removal of residual carbon after reduction is needed. Normally oxidation of reacted sample at about 873K (600°C) in air is employed to remove residual carbon. It brings the risk of re-oxidation of synthesised

Si3N4. The residual carbon is a crucial factor to the process of following decomposition of Si3N4 to prepare Si, since the produced Si will react with C forming SiC. Thus too high C to SiO2 ratio should be avoided. The C/SiO2 ratio of 4.5 was employed in almost all experiments.

Addition of Si3N4 seeds to the silica-graphite mixture promoted silicon nitride synthesis (Figures 4.13-4.15). In literature, effect of seeding was explained by the assumption that nucleation of silicon nitride was the limiting step of carbothermal synthesis of silicon nitride (Inoue et al., 1982; Licko et al. 1992; Hofmann et al. 1993; and Weimer et al. 1997). However, analysis of XRD spectra of samples produced with and without seeding (Figure 4.14) shows similar strong SiC peaks for both samples. Before the depletion of SiO2, the SiC formation rate was almost the same as without seeding

(Figure 4.15). It means formation of SiC was not affected by Si3N4 addition in the raw material. It also indicates that Si3N4 growth is one of the rate-limiting steps, which is consistent with the conclusion of Weimer et al. (1997).

5.5 Mechanism of carbothermal synthesis of silicon nitride

According to the results presented in Chapter 4, carbothermal synthesis of silicon nitride proceeds through silicon carbide as the intermediate compounds.

Change in phase composition in the process of reduction/nitridation and silicon partitioning between SiO2, Si3N4, and SiC give a convincing evidence of SiC conversion to Si3N4 especially after total consumption of SiO2.

The mechanism of silica conversion to silicon nitride was investigated further by SEM/EDS analysis of samples in the process of silica reduction/nitridation. Reduction/nitridation experiments were conducted at 1723K (1450°C) in the 10 vol%

H2 – 90 vol% N2 gas mixture with flow rate 1L/min.

115

Figure 5-6 (a) - (e) show the BSE image and EDS mapping of a sample reduced at

1450 °C for 60 minutes. Si3N4 content in this sample was only about 8% (Figure 5-3). There is a well identified SiC shell seen around the graphite particle with thickness of

3.0-4.0μm. After 240 minutes reaction, conversion of SiC to Si3N4 is seen in Figure 5-

7 (a) - (d). The thichness of consumed SiC is 1.0-1.5μm. Si3N4 nucleated on the SiC layer and grew in the direction outside the layer. Graphite particle was consumed from ~4.0μm to 1.5~2.0μm.

SiC conversion to Si3N4 was also examined in additional experiments in which SiO2 was first reduced to SiC in an Ar atmosphere at 1723K (1450°C) for 300 minutes, then nitridized by introducing N2-H2 gas mixture for another 300 minutes. Figure 5-8 shows the XRD patterns of the sample at different stages of reaction. The XRD pattern of a sample subjected to simultaneous reduction/nitridation is also presented for comparison.

(a) (b) SiC

resin

116

(e)

Figure 5-6 BSE images and EDS element analyses of a sample after 60 minute reduction/nitridation. (a) BSE image, (b) Si, (c) C, (d) N, (e) O.

(a) (b)

SiC

Si3N4 SiC

C SiC

Si3N4

Figure 5-7 BSE images and EDS element analyses of a sample after 240 minute reduction/nitridation. (a) BSE image, (b) Si, (c) C, (d) N.

According to Figure 5-8, reduction of silica to SiC in Ar was close to completion after 117

300 minutes; only a weak peak of cristobalite was observed in the XRD spectrum. After 300 minutes of nitridation, significant amount of silicon nitride was formed. These results confirmed conversion of SiC to Si3N4. However, the rate of SiC conversion to

Si3N4 seemingly was slower in comparison with the rate of Si3N4 formation in the simultaneous reduction/nitridation process. The ratio of the heights of main peaks of nitride to carbide was lower than that in the XRD pattern of the sample obtained by simultaneous reduction/nitridation. This difference in nitridation rate may suggest two possible parallel routes of Si3N4 formation: direct nitridation from SiO2 and nitridation of SiC. This is further examined by kinetic modelling in Section 5.6.

Figure 5-8 XRD patterns of samples obtained in conversion of SiC to Si3N4.

118

Development of the interface between Si3N4 and SiC was examined in direct nitridation of commercial silicon carbide under conditions described above. SEM images of commercial SiC and nitridized SiC particles are presented in Figure 5-9 (a) and (b)).

After nitridation, hexagonal Si3N4 crystals were observed on the surface of SiC particles

SEM/EDS analysis of nitridised samples in Figure 5-9 (c)-(g) confirmed that Si3N4 crystals were originated around SiC particles.

(a) (b)

SiC

Si3N4

(e) (f)

119

(g)

SiC

Si3N4

Figure 5-9 SEM/EDS analyses of commercial silicon carbide subjected to nitridation: (a) SEM image of commercial SiC particles, (b) SEM image of nitridized commercial SiC, (c) SEM image of polished nitridized SiC, (d) Si, (e) C,

(f) N, (g) EDS element mapping analysis of Si3N4 shell around SiC

The cross section of the nitridized SiC sample was analysed by FIB-TEM. The TEM image of the cross section and the elemental distribution along a line in TEM image are shown in Figure 5-10. Figure 5-11 presents the distribution of Si, N and C in the same cross section. It can be seen that silicon nitride was formed on the surface of silicon carbide. There were some channels within the carbide crystal and free carbon layers within carbide and nitride phases. According to these results, Si atoms diffused from the

SiC lattice to the surface of SiC particle through the channels and reacted with N2 forming Si3N4. Free carbon was left within the SiC particle forming the layers; with the progress of the conversion, the carbon layers within carbide phase became in Si3N4 phase.

120

(a) resin Si3N4

C channel SiC SiC

(b)

Figure 5-10 TEM observation of a partially nitridated SiC particle. (a) TEM image showing the line scan position; (b) TEM line scan showing elemental distribution along the line.

121

(a)

(b)

(c)

122

Figure 5-11 Element mapping of the cross section shown in Figure 5-10. (a) Si; (b) N; (c) C.

It can be suggested from experimental results and observations, and thermodynamic analysis, that silicon nitride synthesis includes three parts: reduction of SiO2 to SiO, formation of Si3N4 and SiC, and conversion of SiC to Si3N4.

Reduction of SiO2

( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( )

Formation of Si3N4 and SiC

( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( )

Conversion of SiC to Si3N4

( ) ( ) ( ) ( ) ( )

5.6 Kinetic modelling of synthesis of silicon nitride

Formation of silicon nitride in carbothermal reduction is considered to occur via two parallel routes as illustrated in Figure 5-12.

123

SiO2

SiC (2) Si3N4

Figure 5-12 Routes of Si3N4 formation in kinetic modelling

One route follows Reaction (2.7) to form SiC and then Reaction (5.7) to synthesize

Si3N4. Another parallel route is direct formation of Si3N4 from SiO2 by Reaction (2.1). Reactions in these two routes are as follows.

SiO2SiCSi3N4 route:

( ) ( ) ( ) ( ) ( )

SiO2Si3N4 route:

( ) ( ) ( ) ( ) ( ) ( )

Assuming the rates of reactions are of the first order with respect to SiO2 and SiC and considering that carbon was taken with surplus, the reaction rates R for Reactions (2.7), (5.7) and (2.1) can be expressed by Equations (5.25) to (5.27).

( 𝑖𝑂 ) ( )

( ) 𝑖퐶

( ) ( ) 𝑖𝑂

Where, - reaction constant;

𝑖𝑂 – consumption (conversion) of SiO2, %;

𝑖퐶 – yield of SiC, %; 124

- partial pressure of N2, atm;

– reaction order with respect to N2 in Reaction (5.7);

– reaction order with respect to N2 for Reaction (2.1).

SiO2 conversion ( 𝑖𝑂 ), SiC yield ( 𝑖퐶), and Si3N4 yield ( 𝑖 ), are presented in term of the amount of Si in consumed (converted) SiO2, produced Si3N4 and SiC divided by total amount of Si in initial SiO2. They are identified in Section 3.5.

) SiO2 consumption rate of ( , in formation of both SiC and Si3N4, can be presented by combining Equations (5.25) and (5.27):

𝑖𝑂 ( ) ( ) ( ) 𝑖𝑂 𝑖𝑂 where, is a factor which takes into account SiO2 consumption due to SiO loss assuming that the rate of SiO loss is in proportion to the rate of SiC formation. The rates of SiO loss and SiC formation both are in proportion to the partial pressure of SiO vapour inside a pellet.

)   Formation rate of Si3N4 ( , which is resulted from both SiO2 SiC Si3N4 and

SiO2 Si3N4 routes, can be expressed by Equation (5.28) by combining Equations (5.26) and (5.27).

𝑖 ( ) ( ) 𝑖퐶 𝑖𝑂

The net rate of formation of SiC ( ) can be obtained by combining Reactions (5.25) and (5.26):

𝑖퐶 ( ) ( ) 𝑖𝑂 𝑖퐶

125

Conversion of SiO2:

Equation (5.28) for SiO2 conversion can be rewritten as Equation (5.31).

𝑖𝑂 ( ) ( ) ( ) 𝑖𝑂 𝑖𝑂

Here, and . Equation (5.31) can be simplified further:

𝑖𝑂 ( ) ( )( ) 𝑖𝑂

𝑖𝑂 ( ) ( ) 𝑖𝑂

where . Equation (5.33) can be re-arranged as:

𝑖𝑂 ( )

𝑖𝑂

Before the reaction, 100% of Si present as SiO2 ( 𝑖𝑂 ), and when all SiO2 is exhausted, the conversion of SiO2 equals to 100% ( 𝑖𝑂 ). Integration of Equation

(5.34) with the initial condition (at t = 0, 𝑖𝑂 ) gives Equation (5.35).

( ) ( 𝑖𝑂 )

Therefore, change in the SiO2 conversion ( 𝑖𝑂 ) with reaction time (t) can be presented by Equation (5.36).

𝑖𝑂 ( ) ( )

Yield of SiC:

SiC is an intermediate compound in the SiO2SiCSi3N4 route. Equation (5.30) can be modified by introducing and using Equation (5.37) for 𝑖𝑂 :

𝑖퐶 ( )

𝑖퐶

126

Equation (5.37) is a linear equation which in general form can be presented as Equation (5.38).

( ) ( ) ( ) ( )

where is the first order differential of y with respective to x.

Solution for such type of equation follows Equation (5.39).

( ) ( ) ∫ ( )

( ) ∫ ( ) ( ) ( ) where ( ) ; y and x are equivalent to ySiC and t, respectively.

In the present case, is a function of t written as Equation (5.40)

( ) ( )

= constant.

Equation (5.39) can be modified as follows:

( 퐶 ) ( 퐶 ) ∫ ( 퐶 ) 𝑖퐶

( 퐶 ) ( 퐶 ) 퐶 ( ) 𝑖퐶 ∫

퐶

( 퐶 ) ( 퐶 ) ( ) 𝑖퐶

퐶 𝑖퐶

( )

𝑖퐶

where, 퐶

127

When t = 0, 𝑖퐶= 0, then . Then

( )

𝑖퐶 ( )

Yield of Si3N4:

Substituting and into Equation (5.29) leads to the following equation:

𝑖 ( ) ( ) 𝑖퐶 𝑖𝑂

Using Equation (5.36) for ( 𝑖𝑂 ) and Equation (5.42) for 𝑖퐶 gives Equation (5.44)

𝑖 ( )

( )

Integrating Equation (5.44) gives:

( )

𝑖 ( ) ( )

The Si loss in the form of SiO, defined as 𝑖𝑂, can be estimated by deducting 𝑖퐶 and

𝑖 from 𝑖𝑂 using the experimental data, and can be calculated using Equation (5.46).

𝑖𝑂 ( )( ⁄ )( ) ( )

Calculation of parameters and verification of kinetic model using experimental data

The experimental data of carbothermal reduction/nitridation for varied reaction time were used to verify the reaction model and to determine the kinetic parameters of the model. The reduced samples were analysed for their contents of O, N and C by LECO, from which 𝑖𝑂 , 𝑖퐶, 𝑖 and 𝑖𝑂were calculated. 128

Table 5-2 lists the experimental data obtained in reduction/nitridation experiments at

1723 K (1450°C) in the 10 vol% H2-90 vol% N2 gas mixture at 1 atm.

Table 5-2 Experimental data obtained in reduction/nitridation experiments at 1723 K

(1450°C) in 10 vol% H2-90 vol% N2 gas mixture

Time (h) (%) (%) (%) (%) 0 0 0 0 0 1 50.7 28.3 8.1 14.3 2 74.6 41.5 19.2 13.9 3 91.7 37.2 37.5 17.0 4 96.0 31.4 43.7 20.9 5 98.9 28.6 49.9 20.4 6 98.8 28.9 55.9 14.0 10 98.7 8.09 74.1 16.5 14 98.4 0.43 82.4 15.6

The kinetic parameters of the reaction model can be estimated by minimisation of a sum of squared errors (difference between calculated and experimental values). The Microsoft Excel solver function was used to calculate the optimum values of the reaction rate constants which are presented in Table 5-3.

Table 5-3 Calculated parameters of the kinetic model of carbothermal synthesis of silicon nitride -1 -1 -1 -1 , h , h , h , h 1.2550 0.7176 0.5081 0.2474 0.0800

Using the parameters listed in Table 5-3, the following expressions were obtained for the conversion of SiO2, yield of Si3N4, yield of SiC and the loss of Si in the form of SiO:

𝑖𝑂 ( ) ( )

𝑖퐶 ( ) ( )

𝑖 ( )

𝑖𝑂 ( ) ( )

129

Calculated silicon distribution between SiO2, SiC, Si3N4 and SiO in comparison with experimental data is presented in Figure 5-13.

The kinetic model describes experimental data well. The reaction rate constant for -1 conversion of SiO2 to SiC (Reaction (2.7)), = 0.5081 h . It is much greater than the complex reaction rate constant of direct conversion of SiO2 to Si3N4 (Reaction (2.1)), -1 = 0.0800 h . This indicates that conversion of SiO2 to SiC is much faster than direct conversion to Si3N4 Kinetic parameters were also obtained using experimental data obtained in experiments with nitrogen pressure of 1100 kPa, which are listed in Table 5-4. The calculated parameters are presented in Table 5-5. Figure 5-14 presents the calculated distribution of silicon between different compounds in comparison with experimental data.

100

SiO2 conversion 60 SiC yield Si3N4 yield 50 SiO loss calculated SiO2 conversion 40 calculated SiC yield calculated Si3N4 yield

Conversion/Yield Conversion/Yield (%) 30 calculated SiO loss

0 0 60 120 180 240 300 360 420 480 540 600 660 720 780 840 900

Reaction time (min)

Figure 5-13 Calculated fractions of silicon in different compounds in comparison with experimental data at 1723 K (1450°C)

130

Table 5-4 Experimental data obtained in reduction/nitridation experiments at 1873 K (1600°C) in nitrogen atmosphere at 1100 kPa

Time (min) (%) (%) (%) (%)

0 0.00 0.00 0.00 0.00 10 84.79 78.86 0.19 5.74 20 98.99 72.35 18.95 7.68 30 99.17 52.61 34.82 11.73 40 99.32 41.35 44.75 13.22 50 99.15 33.62 51.53 14.00 60 99.18 30.55 54.46 14.17 80 99.35 32.57 55.47 11.31 120 99.56 17.12 64.77 17.67

Table 5-5 Calculated parameters of the kinetic model of carbothermal synthesis of silicon nitride at 1873 K (1600 °C) in nitrogen at 1100 kPa -1 -1 -1 -1 , h , h , h , h 1.1458 0.2417 0.2109 0.0.167

Using the parameters listed in Table 5-5, the following expressions were obtained for the conversion of SiO2, yield of Si3N4, yield of SiC and the loss of Si in the form of SiO:

𝑖𝑂 ( ) ( )

𝑖퐶 ( ) ( )

𝑖 ( )

𝑖𝑂 ( ) ( )

131

, determined by multiplying to gives 0.2416, which is much greater than of

. This also indicates that conversion of SiO2 to SiC is much faster than to

Si3N4. Even though high nitrogen pressure of 1100 kPa was applied in this specific condition, almost all SiO2 was converted to SiC.

SiO2 conversion SiC yield 0.9 Si3N4 yield SiO loss

calculated SiO2 conversion calculated SiC yield 0.8 calculated Si3N4 yield calculated SiO loss

0.7

0.6

0.5

0.4

Conversion/Yield Conversion/Yield (%) 0.3

0.2

0.1

0 0 10 20 30 40 50 60 70 80 90 100 110 120

Reaction time (min)

Figure 5-14 Calculated fractions of silicon in different compounds in comparison with experimental data obtained at 1823 K (1600 °C ) in nitrogen at 1100 kPa

It can be concluded from the kinetic analysis that formation of SiC from SiO2 is much faster than that of Si3N4 from SiO2 in the carbothermal Si3N4 synthesis process. Almost all SiO2 was converted to SiC; silicon nitride was formed by nitridation of silicon carbide. 132

Therefore, the carbothermal synthesis of silicon nitride can be carried out using two- stage process. At the first stage, quartz is reduced to silicon carbide at temperature elevated temperature (above 1723K (1450°C)). Reduction can be conducted in hydrogen. The second stage involves conversion of SiC to silicon nitride at 1723 K (1450 °C) or below in nitrogen gas atmosphere at elevated nitrogen pressure.

133

CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER WORK

6.1 Conclusions

Synthesis of silicon nitride by carbothermal reduction of silica was investigated in nitrogen and nitrogen-hydrogen mixtures. The effects of temperature, gas composition, pressure and flow rate, carbon to silica ratio, seed addition on reduction/nitridation and silica type on reactions were examined. A preliminary study of decomposition of silicon nitride was also carried out. Reduction/nitridation sequence and mechanisms were established using reaction modelling. The major findings in this project were summarized as follows.

6.1.1 Carbothermal synthesis of silicon nitride

(1) Carbothermal reduction of silica commenced below 1573K (1300°C). The rate of reduction of silica increased with increasing temperature.

(2) In the process of carbothermal reduction in nitrogen and N2-H2 gas mixture, silica was converted to silicon nitride and silicon carbide which were simultaneously formed in the experimental temperature range 1673 K – 1923K (1400 °C – 1650 °C).

The highest Si3N4/SiC ratio was observed at 1723K (1450 °C) in the N2-H2 gas mixture with 10 vol% H2.

(3) Addition of hydrogen into nitrogen promoted silica conversion. Addition of 5 vol% of H2 significantly increased the rate of reduction of silica. The maximum Si3N4 to

SiC ratio was obtained with addition of 10 vol% H2 at 1450 °C after 12-hour reduction.

Higher H2 addition led to lower N2 partial pressure with a negative effect on Si3N4 formation.

(4) Increasing nitrogen pressure increased reduction and nitridation rates, and the stability of nitride at higher temperatures. The maximum Si3N4/SiC ratio was observed at 1550-1600 °C under a N2 pressure of 11 atm. 71.4% of silicon was converted to

134

Si3N4; the rest was mostly SiC with trace residual SiO2 after 1 hour reaction.

(5) Increasing carbon to silica molar ration increased the rate of reduction of silica.

With stoichiometric amount of carbon, the reduction was very slow. When C/SiO2 was equal to 4.5, the reduction completed in about 300 min.

(6) Addition of silicon nitride to the graphite-silca mixture promoted reduction of silica and formation of silicon nitride. This indicates that nucleation of silicon nitride is one of the controlling factors in the reaction process.

(7) Gas flow rate in the experimental range slightly promoted the reduction rate of silica to a peak value, then impeded SiO2 reduction with further increase. This illustrates that carbon monoxide plays important role in silica reduction, particularly in the case of none reducing gas introduced by gas atmosphere.

(8) The effect of silica type was also negligible although the morphology and surface areas were quite different between the amorphous and crystallized silica examined in this project. At the experimental temperatures, amorphous silica crystallized to cristobalite phase, which reducibility was similar to that of quartz.

(9) A reaction model was constructed considering simultaneous conversion of silica to silicon carbide and nitride and further conversion of carbide to nitride. The results showed that silica was first reduced to silicon carbide which was further converted into nitride. Direct conversion of silica into silicon nitride, was much slower than conversion into carbide.

(10) The effect of hydrogen on the kinetics of silica reduction was attributed to formation of CH4 by reacting with carbon, which transferred carbon from graphite particles to the silica particles.

(11) TEM analysis of nitridation of silicon carbide sample showed that silicon diffused out of silicon carbide lattice onto the surface and reacted with nitrogen to form silicon nitride.

135

6.1.2 Decomposition of silicon nitride

Decomposition of silicon nitride was studied at 1500 °C and 1600 °C in hydrogen and argon. Temperature had a significant effect on the decomposition kinetics. The decomposition rate in hydrogen was much faster than in argon, which was attributed to the difference between the diffusivity of nitrogen in these two gases.

6.2 Recommendations for further work

Comprehensive experimental investigation has been carried out on carbothermal synthesis of silicon nitride in nitrogen and hydrogen – nitrogen gas mixtures. In combination with thermodynamic analysis and kinetic modeling, this study provided a fundamental understanding of the carbothermal reduction/nitridation process. Work on decomposition of silicon nitride was preliminary to demonstrate the feasibility of the process. The following further investigation of carbothermal synthesis of silicon nitride and its decomposition is recommended:

(1) The current work showed that the nitrogen pressure had a significant effect on the kinetics of silica reduction and selectivity of silicon nitride formation. However, the experimental investigation was limited to pure nitrogen. Further work will study the effect of hydrogen addition under different pressures. (2) More detailed investigation on the decomposition kinetics of silicon nitride is needed to study the effect of temperature in a broad range, use of vacuum, and pelletised silicon nitride, and to develop further understanding of the effect of internal diffusion of nitrogen on the decomposition rate. (3) In the current investigation of synthesis of silicon nitride and its decomposition, samples of silica and silicon nitride chemicals were examined. Further investigation will study quartz ores as a silica source and the decomposition of silica nitride synthesised in the laboratory. (4) The behavior of impurities, especially of boron and phosphorus in the synthesis 136 and decomposition processes needs to be investigated for the production of high purity silicon.

137

REFERENCES

Abdyukhanov, I. M. (2001): “Development of the basis of technology for production of metallurgical-grade silicon of increased purity for terrestrial solar power engineering”. Rossiiskii Khimicheskii Zhurnal, vol. 45, no. 5-6, pp. 107-111.

Alcala, M. D.; Criado, J. (1992): “Application of constant rate thermal analysis (CRTA) to the synthesis of silicon nitride by carbothermal reduction of silica”. Journal of Thermal Analysis, vol. 38, pp. 313-319.

Alcala, M. D.; Criado, J. M.; Real, C. (2001): “Influence of the experimental conditions and the grinding of the starting materials on the structure of silicon nitride synthesized by carbothermal reduction”. Solid State Ionics vol. 141-142, pp. 657-661.

Amick, J. A. (1982): “Purification of Rice Hulls as a Source of Solar Grade Silicon for Solar Cells”. Journal of Electrochemistry Society, vol. 129, pp. 864-66.

Aratani, F.; Sakaguchi, Y.; Yuge, N.; Ishizaki, M.; Kawahara, T. (1991): “Production of solar grade silicon by carbothermic reduction of high purity silica”. Nippon Kinzoku Gakkai Kaiho, vol. 30, no. 5, pp. 433-435.

Atkins, P. (1994): Physical Chemistry. Oxford, Melbourne, Tokyo: Oxford University Press, 5th ed.

Aulich, H. A.; Eisenrith, K. H.; Urbach, H. P.; Grabmaier, J. G. (1982): “Preparation of high purity starting materials for the carbothermic production of solar grade silicon”. Proceedings-Electronchemical Society, vol. 82, no. 8, pp. 177-182.

Avrami, M. (1939): “Kinetic of Phase Change, I”. Journal of Chemical physics, vol. 7, p. 1103.

Avrami, M. (1940): “Kinetic of Phase Change, II”. Journal of Chemical physics, vol. 8, p. 212.

Avrami, M. (1941): “Kinetic of Phase Change, I”. Journal of Chemical physics, vol. 9, p. 177. 138

Badrak, S. A.; Bartnitskaya, T. S.; Baryshevskaya, I. M.; Kosolapova, T. Ya.; Pikuza, P. P.; Trusov, B.G.; Turov, V. P. (1982): “Thermodynamic Calculations for the Assessment of Possible Ways of Formation of Silicon Nitride”. Powder Metallurgy and Ceramics, vol. 20, no. 9, pp. 66-72.

Bartnitskaya, T. S.; Pikuza, P. P.; Lugovskaya, E. S.; Kosolapova, T. Ya. (1982): “Formation of Silicon Nitride from Silicon Oxide in a Stream of Ammonia”. Inorganic Materials, vol. 18, no. 10, pp. 1728-1732.

Bartnitskaya, T. S.; Pikuza, P. P.; Timofeeva, I. I.; Lugovskaya, I. I.; Kosolapova, T. Y. (1983): “Formation of silicon nitride from silica”. Powder Metallurgy and ceramics, vol. 22, no.7, pp. 523-527.

Batha, H. D.; Whitney, E. D. (1973): “Kinetics and Mechanism of the Thermal

Decomposition of Si3N4”. Journal of the American Ceramic Society, vol. 56, no. 7, pp. 365-369.

Binnewies, M.; Milke, E. (2002): “Thermochemical Data of Elements and Compounds”. Wiley-VCH Verlag GmbH, Weinheim.

Brook, R. J.; Messier D. R.; Riley F. L. (1978): “The α/β Silicon nitride phase transformation” Journal of Materials Science, vol. 13, pp. 1199-1205.

Ceccaroli, B.; Lohne, O. (2003) in Handbook of Photovoltaic Science and Engineering, ed. Luque A. and Hegedus S. Chichester, U.K.: Wiley, 2003, pp. 153-204.

Cochran, G. A.; Conner, C. L.; Eisman, G. A.; Weimer, A.W.; Carroll, D. F.; Dunmead, S. D.; Hwang, C. J. (1994): “The Synthesis of a High Quality, Low Cost Silicon Nitride Power by the Carbothermal Reduction of Silica”. Key Engineering Materials, vol. 89-91, pp. 3-8.

Chase, M. W., National Institute of Standard and Technology (U.S.) (1998): “NIST- JANAF thermochemical tables”. Journal of physical and chemical reference data. Monograph; no. 9. Washington, D.C.: American Chemical Society; Woodbury, N.Y.: American Institute of Physics for the National Institute of Standard and Technology

139

Cho, Y.; Charles,J. (1991): “Synthesis of nitrogen ceramic powders by carbothermal reduction and nitridation. Part 1 silicon nitride”. Materials Science and Technology, vol. 7, p. 289.

Demir, A.; Tatli, Z.; Caliskan, F.; Kurt, A. O. (2007): “Carbothermal Reduction and Nitridation of Quartz Mineral for the Production of Alpha Silicon Nitride Powders”. Materials Science Forum, vol. 554, pp. 163-168.

Durham, B. G.; Murtha, M. J.; Burnet, G. (1988): “Si3N4 by the Carbothermal Ammonolysis of Silica”. Advanced Ceramic Materials, vol. 3, no. 1, pp. 45-48.

Durham, S. J. P.; Shanker, K.; Drew, R. A. L. (1991): “Carbothermal Synthesis of Silicon Nitride: Effect of Reaction Conditions”. Journal of American Ceramic Society, vol. 74, pp. 31-38.

Ekelund, M.; Forslund, B.; Niklewski, T.; Eriksson, G.; Hallgren, G.; Johansson, T.;

Hatcher, M. (1986): “Establishment of the Si3N4/SiC Phase Boundary in the Si-C-O- N System”. The Electrochemical Society’s 169th Meeting, May 4-9, no. 392, pp. 567- 573.

Ekelund, M.; Forslund, B. (1992a): “Carbothermal Preparation of Silicon Nitride: Influence of Starting Materials and Synthesis Parameters”. Journal of American Ceramic Society, vol. 75, no. 3, pp. 532-539.

Ekelund, M.; Forslund, B. (1992b) “Reactions Within Quartz-Carbon Mixtures in a Nitrogen Atmosphere”. Journal of the European Ceramic Society, vol. 9, pp. 107- 119.

Eriksson, G. (1975): “Thermodynamic Studies of High – Temperature Equilibria. XII. SOLGASMIX, a Computer Program for Calculation of Equilibrium Compositions in Multiphase Systems.” Chemica Scripta, vol. 8, pp. 100-103.

Erofeyev, B. V. (1946): “A Generalized Equation of Chemical Kinetics and Its Application in reaction Involving Solids,” C.R. Acad. Sci. URSS, vol. 52, p.511.

140

Fevre, A.; Murat, M. (1975): “Analyse Theorique des Lois Cinetiques Couramment Utilisees en Thermoanalyse Pour L’etude Des reactions Solid-gaz”. Journal of thermal analysis, vol. 7, p. 429.

Figusch, V.; Licko, T. (1987): “Synthesis of Silicon Nitride Powder by Carbothermal Nitriding of Silica”. High Tech Ceramics. Edited by Vincenzini P., Elsevier, Amsterdam, Netherlands, p.517.

Gleiter, H. (1989): “Nanocrystalline Materials”. Progress of Material Science, vol. 33, no. 4, pp. 223-315.

Gotor, F. J.; Macias, M.; Ortega, A.,; Criado, J. M. (1998): “Simultaneous Use of Isothermal, Nonisothermal, and constant Rate Thermal Analysis (CRTA) for Discerning the Kinetics of the Thermal Dissociation of Smithsonite”. International Journal of Chemical Kinetics, vol. 30, p. 647.

Hanna, S.; Mansour, N.; Taha, A.; Abd-Allah, H. (1985): “Silicon carbide and nitride from rice hulls iii-formation of silicon nitride”. British Ceramic Transactions, vol. 84, pp. 18-21.

Hendry, A.; Jack, K. H. (1975): “The Preparation of Silicon Nitride from Silica”. Special Ceramics, vol. 6, p. 199.

Hisao, Y.; Makoto, S. (1999) Patent JP 11287794, 4

Hofmann, H.; Vogt, U.; Kerber, A.; van Dijen, F. (1993): “Silicon Nitride Powder from Carbothermal Reaction”. Material Research Society Symposium Proceeding, vol. 287, pp. 105-120.

Huang, Y.; Guo, H.; Huang, J.; Shen, S. (2007): Gongneng Cailiao, vol. 38, no. 9, pp. 1397-1399, 1404. http://www.theage.com.au/articles/2004/01/27/1075087993151.html

Inoue, H.; Komeya, K.; Tsuge, A. (1982): “Synthesis of Silicon Nitride Powder from Silica Reduction”. Communication of the American Ceramic Society, p. C-205.

141

Ishizaki, M.; Kawahara, T.; Yoshiyagawa, M. (1992): “Production process of solar grade silicon by reduction of high purity silica”. NSG Technical Report, vol. 5, pp. 48-56.

Jack, K. J. (1979): “Review – Sialons and Related Nitrogen Ceramics”. Journal of Material Science, vol. 11, pp. 1135-38.

Johansson, T. (1985): “Process for the Production of Silicon Nitride,” U.S. Patent. No. 4530825, July 23.

Karabanov, S. M. (2002): “New technologies for solar grade silicon production”. Materials Research Society Symposium Proceedings, vol. 716, pp. 495-499.

Kijima, K.; Setaka, N.; Ishii, M.; Tanaka, H. (1982): “Preparation of Si3N4 by Vapor- Phase Reaction”. Journal of The American Ceramic Society-Discussion and Notes, vol.56, no. 6, p. 346.

Komeya, K.; Inoue, H. (1975): “Synthesis of the Alpha Form of Silicon Nitride from Silica”. Journal of Materials Science, vol. 10, no. 7, pp. 1243-1246.

Lee, J. G.; Cutler, I. B. (1977): “Reactions in the SiO2-C-N2 System”. Nitrogen Ceramics, pp. 175-180.

Lee, M. R.; Russell, S. S.; Arden, J. W.; Pillinger, C.T. (1995): “Nierite (Si3N4), a New Mineral from Ordinary and Enstatite Chondrites”. Meteoritics, vol. 30, no.4, pp. 387- 398.

Levenspiel, O. Chemical Reaction Engineering. New York, Chichestre, Brisbane, Singapore, Toronto: Johu Wiley and Sons, 2ed.

Licko, T.; Figusch, V.; Puchyova, J. (1992): “Synthesis of Silicon Nitride by Carbothermal Reduction and Nitriding of Silica: Control of Kinetics and Morphology”. Journal of the European Ceramic Society, vol. 9, pp. 219-230.

Liou, T. H.; Chang, F. W. (1995): “Kinetics of Carbothermal Reduction and Nitridation of Silicon Dioxide/Carbon Mixture”. Industrial and Engineering Chemistry Research, vol. 34, pp. 118-127.

142

Liou, T.; Chang, F. (1996): “The nitridation kinetics of pyrolyzed rice husk”. Industrial and Engineering Chemistry Research, vol. 35, no. 10, pp. 3375-3383.

Mehner, H. (1896): “Verfahren zür darstellung von Nitriden,” German Patent 88999, 30 September.

Miller, P. D.; Lee, J. G.; Culter, I. B. (1979): “The Reduction of Silica with Carbon and Silicon Carbide”. Journal of the American Ceramic Society, vol. 62, no. 3-4, pp. 147- 149.

Mishra, P.; Chakraverty, A.; Bauerjee, H. D. (1985): “Production and Purification of Silicon by Calcium Reduction of Rice Husk White Ash”. Journal of Material Science, vol. 20, pp. 4387-4391.

Mocolm, L. J. (1983) Ceramic Science for Materials Technologist, p.107. Chapman and Hall, New York.

Mori, M.; Inoue, H.; Ochiai, T. (1983): “Preparation of Silicon Nitride Powder from Silica”. Progress in Nitrogen Ceramics. Edited by Riley R.L. Martinus Nijhoff, Amsterdam, Netherlands, p.149.

Mu, J.; Leder, F.; Park, W. C.; Hard, R. A.; Megy, J.; Reiss, H. (1986): “Reduction of phosphate ores by design of a rotary kiln system”. Metallurgical Transaction B: Process Metallurgy, vol. 17B, no. 4, pp. 861-868.

Murray, J. P.; Flamant, G.; Roos, C. J. (2006): “Silicon and solar-grade silicon

production by solar dissociation of Si3N4”. Solar Energy, vol. 80, pp. 1349-1354.

Nepomnyaschikh, A. I.; Fedosenko, V. A.; Eremin, V. P.; Krasin, B. A. (2002): “Low cost multicrystalline silicon for solar cells”. Silicon for the Chemical Industry VI, vol. 17, no. 21, pp. 191-196.

Nepomnyaschikh, A. I.; Zolotaiko, A. V.; Krasin, B. A.; Eliseev, I. A. (2004): “Direct production of multicrystalline solar silicon from high purity metallurgical silicon”. Silicon for the Chemical Industry VII, vol. 21, no. 24, 299-306.

143

Ortega, A. (2002): “The Kinetics of Solid State Reactions Toward Consensus, Part 3. Searching for consistent Kinetic Results: SCTA vs. Conventional Thermal Analysis”. International Journal of Chemical Kinetics, vol. 34, no. 4, p. 223.

Ortega, A.; Alcala, M.; Real, C. (2008): “Carbothermal synthesis of silicon nitride

(Si3N4): Kinetics and diffusion mechanism”. Journal of Materials Processing Technology, vol. 195, pp. 224-231.

Ovsyanikov, A. A.; Oliver, D. Kh.; Polak, L. S. (1978): “Kinetics of Ammonia Decomposition in a Plasmachemical Reactor. I. Experimental Investigation”. High Energy Chemistry, vol. 12, no. 12, pp. 51-57.

Parr, N. L.; May, E. R. W. (1967): “Preparation, microstructure, and mechanical properties of silicon nitride”. British Ceramic Society, vol. 7, p.81.

Perea, D. S. (1987): “Conversion of Precipitated Silica from Geothermal Water to Silicon Nitride”. Journal of Materials Science, vol. 22, p.2411.

Perez-Maqueda, L. A., Criado, J. M., Real, C., Subrt, J., Bohacek, J. (1999): “The use of constant rate thermal analysis (CRTA) for controlling the texture of hematite obtained from the thermal decomposition of goethite”. Journal of Materials Chemistry, vol. 9, pp.1839-1845.

Priest, H. F.; Burns, F. C.; Priest G. L.; Skaar, E. C. (1973): “The oxygen content of alpha silicon nitride”. Journal of American ceramic society, vol. 56, no. 7, pp. 395.

Rahman, I. A.; Riley, F. L. (1989): “The Control of Morphology in Silicon Nitride Powder Prepared from Rice Husk”. Journal of the European Ceramic Society, vol. 5, pp.11-22.

Real, C.; Alcala, M. D.; Criado, J. M. (2004): “Synthesis of Silicon Nitride from Carbothermal Reduction of Rice Husks by the Constant-Rate-Thermal-Analysis (CRTA) Method”. Journal of the American Ceramic Society, vol. 78, no. 1, pp. 75- 78.

Riley, F. L. (2000): “Silicon Nitride and Related Materials”. Journal of the American Ceramic Society, vol. 83, no.2, pp. 245-265.

144

Rustioni, M.; Pirazzi, R.; Tincani, M.; Margadonna, D.; Pizzini, S. (1985): “Advances in solar silicon production”. Proceedings-Electrochemical Society, vol. 85, no. 9, pp. 288-303.

Sakaguchi, Y.; Yuge, N.; Baba, H.; Suhara, S.; Fukai, M.; Aratani, F. (1989): “Production process development of high purity silicon for solar cells and status at Kawasaki Steel Corp”. Kawasaki Seitetsu Giho, vol. 214, no. 4, pp. 329-334.

Schulze, F. W.; Fenzl, H. J.; Geim, K.; Hecht, H. D.; Aulich, H. A. (1984): “Progress on the carbothermic production of solar-grade silicon using high-purity starting materials”. Conference Record of the 17th IEEE Photovoltaic Specialists Conference, pp. 584-587.

Scriven, L. E. (1959): “On the dynamics of phase growth”, Chemical Engineering Science, vol. 10, no. 1, p.1.

Siddigi, S. A.; Hendry, A. (1985): “The influence of iron on the preparation of silicon nitride from silica”. Journal of Material Science, vol. 20, no. 9, pp. 3230-3238.

Strake, B.; Schulze, F. W.; Aulich, H. A. (1988): “Production of solar grade silicon by carbothermic reduction of silica”. Erzmetall, vol. 41, no. 3, pp. 126-131.

Szekely, J.; Evans, J. W.; Sohn, H. Y. (1976): “Reaction of Porous Solids”. in Gas-Solid Reaction; Academic: New York.

Szweda, A.; Henry, A.; Jack, K. H. (1981): “The preparation of silicon nitride from silica by sol-gel processing”. Proceedings of the British Society, vol. 31, pp.107-118

Thompson, D. S.; Pratt, P. I. (1967): “The structure of silicon nitride” Science of Ceramics, vol. 3. Edited by Stewart, G. H. Academic Press, New York, pp. 33-51

Tomkins, F.C. (1976): “Decomposition Reactions”. Treatise on Solid State Chemistry, vol.4, Reactivity of Solids, pp. 206-12. Edited by N. B. Hannay.Plenum Press, New York.

Turkdogan, E.T.; Bills P.M.; Tippett, V.A. (1958): “Silicon nitrides: some physic- chemical properties”. Journal of applied chemistry, vol. 8, issue 5, pp. 296-302.

145

Van Dijen, F. K.; Metselaar, R.; Siskens, C. A. M. (1985): “Reaction-rate limiting steps in carbothermal reduction processes”. Journal of the American Ceramic Society, vol. 68, no. 1, pp. 16-19.

Van Dijen, F. K.; Vogt, U. (1992): “The Chemistry of the Carbothermal Synthesis of α-

S3N4: Reaction Mechnism, Reaction Rate and Properties of the Product”. Journal of the European Ceramic Society, vol. 10, pp. 273-282.

Vlasova, M. V.; Bartnitskaya, T. S.; Sukhikh, L. L.; Krushinskaya, L. A.; Tomila, T. V.;

Artyuch, S. YU. (1995): “Mechanism of Si3N4 Nucleation during Carbothermal Reduction of Silica”. Journal of Materials Science, vol.30, pp. 5263-5271.

Weimer, A. W.; Eisman, G. A.; Susnitzky, D. W.; Beaman, D. R.; McCoy, J. W. J. (1997): “Mechanism and Kinetics of the Carbothermal Nitridation Synthesis of α- Silicon Nitride”. Journal of the American Ceramic Society, vol. 80, no. 11, p. 2853.

Woditsch, P.; Koch, W. (2002): “Solar grade silicon feedstock supply for PV industry”. Solar Energy Materials & Solar Cells, vol. 72, pp. 11-26.

Yashiyagawa, M.; Tamenori, H.; Shimomura, T.; Ishizaki, M.; Hattori, A.; Yokoe, K.; Noguchi, N.; Horigome, T. (1985): “The preparation of high-purity silica for solar- grade silicon”. Proceedings of the 20th Intersociety Engergy Conversion Engineering Conference, vol. 3, pp. 3.466-3.469.

Yuge, N.; Abe, M.; Hanazawa, K.; Baba, H.; Nakamura, N.; Kato, Y.; Sakaguchi, Y.; Hiwasa, S.; Aratani, F. (2001): “Purification of Metallurgical-Grade Silicon up to Solar Grade”. Progress in Photovoltaics: Research and Applications, vol. 9, pp. 203- 209.

Zhang, S.; Cannon, W. (1984): “Preparation of Silicon Nitride from Silica”. Journal of the American Ceramic Society, vol. 67, no. 10, p. 691.

146