<<

Order Number 8717683

The behavior of silicon-based in mixed oxidation/chlorination environments

Marra, John Edward, Ph.D.

The Ohio State University, 1987

UMI 300 N. Zeeb Rd. Ann Arbor, MI 48106 PLEASE NOTE:

In all cases this material has been filmed in the best possible way from the available copy. Problems encountered with this document have been identified here with a check mark V .

1. Glossy photographs or pages______

2. Colored illustrations, paper or print ______

3. Photographs with dark background

4. Illustrations are poor copy______

5. Pages with black marks, not original copy ______

6. Print shows through as there is text on both sides of page______

7. Indistinct, broken or small print on several pages

8. Print exceeds margin requirements______

9. Tightly bound copy with print lost in spine______

10. printout pages with indistinct print ______

11. Page(s)______lacking when material received, and not available from school or author.

12. Page(s) seem to be missing in numbering only as text follows.

13. Two pages num bered . Text follows.

14. Curling and wrinkled pages

15. Dissertation contains pages with print at a slant, filmed as received

16. Other

University Microfilms International THE BEHAVIOR OF SILICON-BASED CERAMICS IN MIXED

OXIDATION/CHLORINATION ENVIRONMENTS

DISSERTATION

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of the Ohio State University

By

J ohn Edward Marra, B .S ., B .A .

*****

The Ohio State University

1987

Dissertation Committee Approved by

E.R. Kreidler

D.W. Readey Adviser K.T. Faber Department of Engineering

J .D . Cawley ACKNOWLEDGMENTS

I would like to thank my advisor Dr. Eric Kreidler for his ideas guidance and support. I also appreciate his willingness to allow me to pursue my own research interests. Special thanks go to my committee members Dr. Dennis Readey, Dr. Katherine Faber and Dr. James

Cawley for many insightful discussions during the course of this work.

I am grateful to my fellow graduate students who have made the last four years very enjoyable.

A sincere note of appreciation goes to Nate Jacobson and Dennis

Fox of NASA-Lewis for their help and cooperation during the course of this work. Without their support this work never would have been completed.

I would also like to acknowledge my parents, in-laws, and the rest of my family for their constant support and encouragement.

Finally, I would like to acknowledge my wife and best friend,

Lisa. Without her support, both moral and financial, I could not have done this.

Funding for this project was generously provided by the Orton

Foundation and the Army Research Office at Durham, and is gratefully acknowledged.

ii VITA

January 17, 1961 ...... Born - Rockville Centre New York

1983...... B.S. - Ceramic Science B.A. - Chemistry Alfred University Alfred, New York

1983-Present...... Graduate Research Associate Department of Ceramic Engineering The Ohio State University Columbus, Ohio

PUBLICATIONS

"Stresses in Dumet-Soft Seal," Journal of the American Ceramic Society, March 1985.

FIELD OF STUDY

Major Field: Ceramic Engineering TABLE OF CONTENTS

ACKNOWLEDGMENTS...... ii

VITA ...... iii

LIST OF TABLES...... ix

LIST OF FIGURES...... xi

CHAPTER PAGE

I. INTRODUCTION...... 1

II. BACKGROUND LITERATURE...... 4

2.1. Spectrometry...... 4

2.1.1. General...... 4

2.1.2. Free-Jet Expansion...... 5

2.1.3. Beam Formation...... 12

2.1.4. Beam Modulation...... 14

2.1.5. Quadrupole Mass Spectrometry...... 19

2.II. Previous Experimental Studies...... 21

2.II.1. Oxidation...... 22

2.11. 1.1. Si and S iC...... 22

2.11. 1.2. Si3N4...... 25

2. II. 2. Chlorination Reactions...... 27

2.11.2.1. S i ...... 27

2.11.2.2. SiC...... 29

iv 2.II.2.3. Si3N4 ...... 33

2. II. 3 . Mixed Oxidation/Chlorination Reactions...... 33

2.11.3.1. and Superalloys...... 33

2.11.3.2. S i ...... 34

2.11.3.3. SiC...... 37

2.II.4. Silicon Oxychlorides...... 42

2.II. 5. SiCl4 Mass Spectrometry...... 44

III. EQUILIBRIUM CALCULATIONS...... 48

3.1. Introduction...... 48

3. II. Method of Calculating Equilibrium...... 49

3. III. SOLGASMIX-PV Computer Program...... 51

3.III.1. General Information...... 51

3. Ill. 2. Operational Parameters...... 52

3. IV. Systems Examined...... 53

3.V. Discussion of Results...... 54

3.V.I. SiC-HCl System...... 54

3.V.2. Si3N4-HCl System...... 59

3.V.3. Si-HCl System...... 59

3.V.4. Si02-HCl System...... 61

3.V.5. Environments...... 63

3.V.6. Mixed Oxidation-Chlorination Reactions...... 65

3.V.7. SiC-HF System...... 66

3.V.8. Si^N^-HF System...... 69

3.V.9. SiOn-HF System...... 69

v 3.V.10. SiC-H20 System...... 70

3.V.11. Si3N4-H20 System...... 70

3.V.12. Si02-H20 System...... 73

IV. EXPERIMENTAL PROCEDURE...... 74

4.1. Samples and Atmospheres...... 74

4.1.1. Samples...... 74

4.1.1.1. Silicon ...... 74

4.1.1.2. ...... 77

4.1.1.3. Silicon...... 80

4.1.1.4. Silica...... 80

4.1.2. Experimental Conditions...... 80

4.1.2.1. Gas Mixtures...... 80

4.1.2.2. Temperatures...... 82

4.II. Sample Preparation...... 82

4.III. Mass Spectrometry...... 84

4.111.1. Experimental Set-Up...... 84

4.111.2. Experimental Procedure...... 87

4.111.2.1. System Calibration...... 87

4.111.2.2. Signal Intensity Calculations...... 87

4.111.2.3. Ionization Efficiency Curves...... 88

4.111.2.4. Standard Test Method...... 88

4.111.2.5. Decay Experiments...... 91

4.111.2.6. Pre-Oxidized Samples...... 92

4.IV. Thermogravimetric Analysis...... 93

vi 4.IV.1. Experimental Configuration...... 93

4.IV.2. Experimental Procedure...... 95

V. RESULTS AND DISCUSSION...... 96

5.1. Mass Spectrometry...... 96

5.1.1. System Calibration...... 96

5.1.2. Ionization Efficiency Curves...... 96

5.1.3. Standard Test Method...... 107

5.1.3.1. Effect of Gas Composition at 950 ° C...... 107

5.1.3.2. Effect of Temperature...... 117

5.1.4. Decay Experiments...... 124

5.1.5. Pre-Oxidized Samples...... 129

5.1.6. Silicon Oxychlorides...... 148

5.1. 6.1. Initial Identification...... 148

5.1.6.2. Mechanism of Formation...... 154

5.II. Thermogravimetric Analysis...... 157

5.11.1. Effect of Gas Composition at 950 °C ...... 157

5.11. 1.1. SiC Materials...... 157

5.11. 1.2. Si and Si3N 4...... 162

5 . II . 2. Effect of Temperature...... 164

5.II.3. Effect of Gas Flow Rate ...... 164

VI. MODELS...... 168

6.1. Physical Models...... 168

6.1.1. "Pure" SiC...... 168

6.1.2. SiC Containing Excess C ...... 170

vii 6.1.3. SiC Containing Excess S i ...... 174

6. II. Mass Transport...... 176

6.11.1. Theory...... 176

6. II. 1.1. Laminar Flow Past a Flat Plate...... 176

6.11.2.2. Determination of Constants...... 181

6.11.2. Correlation Between Calculated and Observed Rates 182

6. II. 2.1. SOLGASMIX-PV Results...... 182

6.11.2.2. Mass Spectrometric Results...... 185

VII. CONCLUSIONS...... 189

VIII. SUGGESTIONS FOR FUTURE WORK ...... 193

8.1. Mass Spectrometry...... 193

8.1.1. Equipment Modifications...... 193

8.1.2. Future Studies...... 195

8.II. Si-C-O-Cl System...... 196

REFERENCES...... 199

APPENDIX A ...... 208

APPENDIX B ...... 219

APPENDIX C ...... 220

viii LIST OF TABLES

TABLE PAGE

2.1 Appearance Potential And Relative Intensity Data for ...... 46

2.2 Monoisotopic Spectra for Silicon Tetrachloride...... 47

3.1 Typical Data for the Si-C-H-Cl System...... 56

3.2 Comparison of Free Energy Minimization Data...... 62

4.1 Typical Chemical Analysis of Norton NC203 Hot-Pressed SiC. 75

4.2 Typical Chemical Analysis of Norton NC430 Densified SiC... 76

4.3 Typical Chemical Analysis of Sohio Hexoloy Sintered a-SiC. 78

4.4 Trace Element Analysis of Norton NC132 Hot-Pressed SigN^. . 79

4.5 Set Flow Rates...... 81

5.1 SiClx+ Appearance Potential Data...... 103

5.2 Relative Intensity Values for Major for Spectra of SiCl^ at Room Temperature and Si in 2% CI2 at 960 °C... 106

5.3 Amount of Volatile Product Produced for Various SiC Materials as a Function of Gas Stream Composition...... 110

5.4 Amount of Volatile Product Generated During the Reaction of Si, Si02, and SigN^ as a Function of Gas Stream Composition...... 115

5.5 Effect of Temperature on the Amount of Volatile Product Generated by Reaction of Norton NC203 SiC with Various Gases...... 119

ix Effect of Temperature on the Amount of Volatile Product Generated by Reaction of Sohio Hexoloy SiC with Various Gases...... 120

Steady State Values Achieved During Decay Experiments.... 130

Steady State Values Achieved After Exposure of Pre-Oxidized Samples to a 2% Chlorine Environment...... 133

Parabolic Oxidation Rate Constants for Various Types of (from reference 93)...... 146

Calculated Scale Thickness as a Function of Oxidizing Condition...... 147

Intensities of High Mass Molecules Shown in Figure 5.22 .. 150

Calculated Isotopic Abundances...... 152

Comparison of Observed and Calculated Isotopic Abundances. 155

Effect of Gas Composition on Rate of Weight Loss at 950 °C for Various Types of Silicon Carbide...... 158

Effect of Gas Composition on Rate of Weight Loss for Silicon and Silicon Nitride at 950 °C...... 163

Effect of Temperature on Rate of Weight Loss in 2% Cl 2-Ar Gas Stream...... 165

Comparison of Linear Rate Constants Determined Experimentally and From the S0LGASMIX-PV Calculation Results...... 184

Comparison of Linear Rates Determined Experimentally with Those Determined from Mass Spectrometric Analysis.... 187

x LIST OF FIGURES

FIGURE PAGE

2.1 Free flow field and shock system (from ref. 6) ...... 7

2.2 Graphical representation of isentropic expansion equations for an ideal gas (from reference 9) ...... 9

2.3 Expansion history of (from reference 9) ...... 10

2.4 Kantrowitz and Grey high intensity molecular beam source compared to conventional source (from reference 12)...... 13

2.5 Effect of Sampling orifice-to-skimmer distance on intensity of molecular beam (from reference 9)...... 15

2.6 Modulation system wiring diagram (from reference 5) ...... 18

2.7 Schematic of quadrupole mass filter (from reference 22).. 20

2.8 Oxidation of SiC, effect of pressure at 1300 °C: A, 9 x 10*3 ; B, 4 x 10'2; C, 1 x 10'1 ; D, 5 x 10 '1 Torr (from ref. 27)...... 24

2.9 Matched integral free energy diagram for the Si-C-Cl system (from reference 39)...... 31

2.10 Ellingham plot of the Si-C-Cl-H system (from reference 39)...... 32

2.11 Chlorine concentration profiles for samples prepared by oxidation of Si at 1100 °C in 0.3% CI2/O2 ambients (from ref. 51)...... 36

2.12 spectrum showing existence of silicon oxychloride (after reference 61)...... 38

xi 2.13 Variation of weight loss as a function of different types of low-cost SiC materials (from reference 63)...... 40

2.14 Effect of gas stream composition on the corrosion of low-cost sintered a-SiC (from reference 63)...... 41

3.1 Comparison of data obtained for the Si-C-H-Cl system...... 57

3.2 Graphical representation of data for the Si-N-H-Cl system...... 60

3.3 Graphical representation of data for the Si-O-H-Cl system...... 64

3.4 Partial pressure of SiCl^ versus percent in the SiC-O-Cl-Ar system...... 67

3.5 Graphical representation of data for the Si-C-H-F system...... 68

3 .6 Graphical representation of data for the Si-O-H-F system...... 71

4.1 Schematic diagram of gas mixing system...... 83

4.2 Schematic diagram of high-pressure mass spectrometer sampling system...... 85

4.3 Schematic diagram of thermogravimetric testing apparatus...... 94

5.1 Calibration plot for mass spectrometer output reading...... 97

5.2 Ionization efficiency curves for; a, SiCl^ at room temperature, and b. Si in 2% Cl2 in Ar at 960 °C...... 99

5.3 Ionization efficiency curves for an ion undergoing; a. simple, and b. complex ionization (from reference 91). 100

5.4 Typical spectra for; a. SiCl^ at room temperature, and b. Si in 2% Cl2 -Ar at 960 ° C ...... 105

5.5 Typical spectrum for Norton NC203 hot-pressed SiC in 2% Cl2 at 950 °C...... 108

xii 5.6 Effect of gas stream composition on the surface morphology of Norton NC430 siliconized SiC as prepared (a) and after reaction at 950 °C in 2% CI2 (b), 1% CI2, x% O 2 where x = 2 (c), 4 (d), 10 (e), and 20 (f)...... 112

5.7 Effect of gas stream composition on the surface morphology of Sohio Hexoloy a-SiC as prepared (a) and after reaction at 950 C in 2% CI2 (b), 1% Cl , x% O 2 where x - 1 (c), 2 (d), 4 (e), and 10 (e)...... 113

5.8 Effect of gas composition on the surface morphology of GTE AY 6 SijN^ reacted at 950 °C for 30 minutes in; a. 2% CI2 and b. 1% CI2, 1% O 2...... 118

5.9 Effect of temperature on the amount of volatile product produced for Norton NC203 hot-pressed SiC ...... 122

5.10 Effect of temperature on the amount of volatile product produced for Sohio Hexoloy sintered a-SiC...... 123

5.11 Decay of SiCl^ signal as a function of time in oxygen for Norton NC203 SiC...... 125

5.12 Decay of SiCl^ signal as a function of time in oxygen for Sohio Hexoloy SiC ...... 126

5.13 Decay of SiCl^ signal as a function of time in oxygen for Norton NC430 SiC ...... 128

5.14 Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Norton NC203 hot-pressed SiC...... 132

5.15 SEM photomicrographs of pre-oxidized Norton NC203 after cooling...... 135

5.16 Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Sohio Hexoloy SiC...... 136

5.17 SEM photomicrographs of Sohio Hexoloy a-SiC oxidized for 16 hours in pure oxygen at 950 °C before (a) and after (b) exposure to chlorine...... 138

5.18 SEM photomicrographs of Sohio Hexoloy a-SiC oxidized for 21 hours in pure oxygen at 1200 °C before (a) and after (b) exposure to chlorine...... 139

xiii 5.19 Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Norton NC430 siliconized SiC ...... 141

5.20 SEM photomicrographs of oxide grown on Norton NC430 in 100% 0« for; a. 16 hours at 950 °C, and b. 21 hours at 1200 °C ...... 142

5.21 Comparison of data obtained for samples pre-oxidized in 20% O 2 at 950 °C for 30 minutes...... 144

5.22 Mass spectrum of Norton NC203 SiC reacted in 1% CI2, 1% O 2 at 950 °C showing existence of high mass molecules...... 149

5.23 Sample weight versus time for Norton NC203 hot-pressed SiC as a function of gas stream composition at 950 °C ...... 159

5.24 Effect of temperature on the rate of weight loss for Norton NC203 and Sohio Hexoloy SiC materials in 2% C ^ - A r ...... 166

6.1 Physical model describing corrosion mechanism of "pure" SiC exposed to a CI2-O2 gas stream...... 171

6.2 SEM photomicrograph (A) and X-ray fluorescence map of distribution (B) in a polished section of Sohio Hexoloy a-SiC (from reference 94)...... 172

6.3 Physical model describing corrosion mechanism of high carbon activity SiC exposed to a CI2-O2 gas stream...... 175

6.4 Schematic diagram describing corrosion mechanism of high Si activity SiC when exposed to; a. a CI2 gas stream, and b. a CI2-O2 gas stream...... 177

6.5 Schematic diagram of boundary layer...... 179

xiv CHAPTER I

INTRODUCTION

The recent developments in advanced ceramic materials have made

these materials candidates for applications previously restricted to metals and superalloys. In particular, ceramic/ ceramic composites as well as monolithic bodies, are being used in the construction of

advanced heat engines. Ceramic materials are attractive for heat

engine applications due to the fact that they are able to withstand higher temperatures than their metallic counterparts. - based engines are limited to temperatures in the neighborhood of

1000°C^, whereas ceramic engine parts are able to operate at

temperatures in excess of 1350 °C.

The increased operating temperature, coupled with the lower

of ceramic components make the ceramic engine more efficient

than conventional metallic and superalloy engines. In addition, the

raw materials typically used in the production of these engine

components (silicon carbide and silicon nitride) are relatively

inexpensive. This implies that as production technology advances, the

cost of ceramic engine parts may well be less than superalloy parts used in the same situations.

During the operation of any heat engine, the engine components are

often exposed to severe environments. In addition to severe stress

1 situations, heat engine components are often exposed to oxidative and/or corrosive gases. The behavior of ceramic components in these environments will dictate the extent of their use in heat engine applications. If the ceramic component is not able to withstand exposure to the gaseous environment, surface flaws are likely to be formed in the material. These flaws will probably degrade the physical integrity of the ceramic, and may ultimately to 2 catastrophic failure .

Until recently, not much work has been done to further the understanding of the reactions between corrosive gases and ceramic materials for advanced engine applications. In particular, the mechanisms of the corrosion of Si-based advanced ceramic materials in chlorine environments is not well known. Engines used in marine environments ingest significant amounts of chlorine and the behavior of ceramics in these environments must be studied. In addition to marine-based turbine engines, SiC ceramics are currently being used as heat exchanger tubes in the of Zr and Ti and the remelting of

Al. These environments are typically composed of oxygen and various halide gases.

The reaction of ceramic materials with chlorine environments is likely to produce a variety of corrosion products. The possible products include both volatile and condensed species. The goal of this research was to unambiguously identify and analyze the corrosion products formed during the exposure of Si-based ceramics to chlorine- containing environments.

The reactions between the gas phase and the ceramic substrate have been analyzed using a high pressure mass spectrometric sampling system. This configuration consists of a differentially pumped vacuum system which allows gases at atmospheric pressure to be analyzed by a mass spectrometer which operates under high vacuum. The construction of the sampling system is such that both volatile and condensible species may be analyzed in-situ without altering the compositon of the reaction stream. The unambiguous identification of corrosion products, coupled with thermodynamic calculations and thermogravimetric experiments run under identical conditions enables the corrosion mechanisms to be quantified to an extent unobtainable by using thermogravimetric experiments alone. CHAPTER II

BACKGROUND LITERATURE

2.1. Mass Spectrometry.

2.1.1. General.

Mass spectrometry is a common technique for the analysis of gaseous molecules. In mass spectrometry a stream of gas is ionized and the various ions are separated by virtue of their mass-to-charge

(m/e) ratios. A variety of mass spectrometers are commercially available. The different systems utilize different methods of ionization as well as different methods of separation. Different systems possess differing degrees of sensitivity. The sensitivity of a mass spectrometer is typically represented as the highest molecular weight value that may be resolved by the instrument. Resolution varies from approximately 200 for time-of-flight instruments to upwards of 10,000 for the most sophisticated magnetic sector instruments. Detailed descriptions of mass spectrometry may be found in a variety of texts (see for instance reference 3) and therefore, the mechanics of the various techniques will not be presented here.

The technique used in this research, known as quadrupole mass spectrometry, will be discussed in detail in section 2.1.5.

Although mass spectrometry is a valuable technique for the analysis of gaseous molecules, its application to in-situ monitoring

4 5

of high pressure, high temperature gas streams is severely limited.

Conventional mass spectrometers normally operate under very high

- 7 - 8 vacuum conditions, typically in the range of 10 to 10 torr.

Therefore, in order to analyze the products generated from reactions

occurring under atmospheric conditions (1 atm =760 torr), the

pressure of the gas stream must be reduced by approximately 10 orders

of magnitude. This reduction of the gas stream pressure must be

accomplished without altering the chemical composition of the stream

itself. In order to accomplish both of these objectives the technique

of free-jet expansion must be employed.

2.1.2. Free-Jet Expansion.

If a gas passing through an orifice has a mean free path (A)

significantly less than the orifice diameter (d), the Knudsen number

of the gas, Kjj = A/d, is much less than one. This property, coupled with a pressure drop in excess of 2 orders of magnitude, results in an

isentropic expansion of the gas stream^. This expansion is

terminated by an abrupt transition to collisionless flow. This type

of expansion, when allowed to occur without constraint on the

downstream side of the orifice, is known as free-jet expansion.

When a gas stream passes through an orifice to a low pressure

region, several types of gas flow may result. These include;

effusive, transitional, slip, and continuum flow. The type of gas

flow which occurs is governed by the Knudsen number. For free-jet

expansion processes continuum flow is desired. This type of flow 6 . o minimizes boundary layer effects, and is produced when < 10

This condition must be met in order to insure that the sample being analyzed has not been altered by the boundary layer that exists near the orifice. This boundary layer is common when the orifice is not in thermodynamic equilibrium with the gas stream. A non-equilibrium situation normally exists in high pressure, high temperature sampling.

As the free jet expands past the orifice into the low, but finite, pressure region, the flow field of the supersonic jet is bounded by a shock system consisting of a barrel shock and a Mach disk or normal shock front. This flow field and shock system are schematically represented in figure 2.1. At the Mach disk, the gas density of the free-jet is equal to the gas density of the residual gases in the vacuum chamber. At positions beyond the Mach disk collisions occur which decrease the intensity of the gas stream"*.

The location of the Mach disk is determined from the equation^:

^ - 0.67Do(Po/P)0 -5 (2.1) where is the distance from the orifice to the Mach disk, D 0 is the orifice diameter, PQ is the pressure of the source gas stream, and P is the pressure of the gas in the region immediately downstream of the orifice. Analysis of equation (2.1) shows that as the pressure in the region downstream of the orifice is decreased, the distance between the orifice and the Mach disk is increased. This implies that when the downstream pressure is decreased, the jet may travel a longer distance before it is interrupted by the normal shock front. J E T BOUNDARY

BARREL SHOCK

STREAMLINE

ee now *nd shock syst, ft*Oflj tef The aerodynamics of the free-jet have been studied extensively^.

The expanded gas stream has properties which are related to the properties of the source gas stream and the flow Mach number, M. The

Mach number is defined as the ratio of the local flow velocity, V, to the local in the medium, c, that is:

M = V/c (2.2)

The density, pressure, and temperature of the expanded gas stream are related to the Mach number and the source gas conditions (denoted by O the subscript o) by the isentropic expansion equations :

n_ 1 + (7 - 1) M 2 jV ( l - 7 ) (2.3)

JL 1 + (7 - 1) M 2 (2.4)

T_ _ £ 1 + (7 - 1) M 2 j '1 (2.5) To where 7 is the specific heat ratio for the gas. These equations assume a perfect gas in complete equilibrium.

Work by Stearns and his co-workers^ shows graphical representations of these equations for ideal gases (figure 2.2) as well as the actual expansion history of an argon gas stream (figure

2.3). These figures show that significant reductions in the temperature, pressure, and density of the gas stream occur within a few microseconds of the gas passing through the orifice. This implies that the chemical reactions occurring within the source gas stream cease within a few microseconds of sampling, thereby insuring that the gas analyzed is essentially representative of the source conditions. 9

Temperature ratio

Pressure ratio

Figure 2.2 - Graphical representation of isentropic expansion equations for an ideal gas (from reference 9). iue 2. Figure

P/Po, N/No, or T/To Epninhsoyo ro (rm eeec 9). reference (from Argon of history Expansion - 10~1 10 0 0 i ______

1 i ______Time Time Time ty6ec) for To = 300 K 300 = To for ty6ec) Time EXPANSION HISTORY OF ARGON 2 tyeec) 2 1 i ______for To = 2000 K 2000 = To for 3 Density ratio Density Pressure ratio Pressure Tem perature ratio perature Tem I ______Do = 0.025 cnt 0.025 = Do 4 I ______5 6 . L 11

Although the free-jet expansion technique insures the analysis of a representative sample, it is not without its problems. Extreme care must be taken to understand and avoid the problems that may arise when sampling with a free-jet.

One problem typically encountered is the depletion of low molecular weight species from the gas stream at large distances from the sampling orifice. Stern et al.^® studied the separation of gases in a supersonic jet. By studying various -argon mixtures they found that a first power dependence on molecular weight exists.

The extent of low mass separation in a given sampling system may be ascertained by sampling a gas stream of pre-determined composition.

A second problem typically encountered in free-jet expansion sampling of high pressure gas streams is homogeneous nucleation and condensation. Greene and Milne^ have observed this phenomenon during the sampling of a 5 atmosphere pure argon gas stream at 300 °K. They observed several of argon. This polymerization is caused by a supersaturation effect. As the gas expands past the orifice, its pressure decreases more slowly than the equilibrium vapor pressure of the condensed species. This to supersaturation of the gas stream and ultimately homogeneous nucleation. The extent of this sampling problem may be determined by varying the experimental conditions and examining the corresponding effect on the amount of complex species observed. 12

2.1.3. Beam Formation.

Once a free-jet has been generated, it must be formed into a molecular beam before it can be analyzed. The formation of a molecular beam from a gas stream that is undergoing free-jet expansion

12 is straightforward. Early work conducted by Kantrowitz and Grey established a technique for creating a high intensity molecular beam from a free-jet. Their technique differed from conventional molecular beam work in that they employed a nozzle type source and first slit rather than the "plate and pinhole" sources typically used by earlier researchers. The difference between their high intensity source and a conventional system is schematically represented in figure 2.4. The use of a conical first slit, known as a skimmer, increases the velocity of the gas flowing through the system, resulting in a greater beam intensity.

In order to form a supersonic beam, the skimmer must be designed to meet certain stringent requirements. The interior angle of the skimmer must be made large enough such that scattering of the beam resulting from molecules striking the interior walls is minimized.

The exterior angle must also be constructed such that deflected molecules may be pumped away from the skimmer area. If the amount of deflection is sufficiently large, a normal shock may form, which would destroy the supersonic flow. The Kantrowitz design utilizes a skimmer with an exterior angle of approximately 60° and an interior angle of

50°. This design has become standard in the construction of molecular 13

CONVENTIONAL MOLECULAR BEAM

GAS SUPPLY FIRST SLIT SECONO SLIT

TM

TO NOZZLE TO VACUUM TO VACUUM EXHAUST PUMP PUMP PUMP HIGH INTENSITY MOLECULAR BEAM

Figure 2.4 - Kantrowitz and Grey high intensity molecular beam source compared to conventional source (from reference 12). 14 beam systems.

The position of the skimmer relative to the sampling orifice is also a critical parameter in the formation of a high intensity

□ molecular beam. Stearns studied changes in the intensity of an argon beam with varying sampling orifice-to-skimmer distances. The results of this analysis (figure 2.5) indicate that the intensity of the beam is maximized when the distance from the sampling orifice to the skimmer is 80-100 times the orifice diameter. This implies that for a typical orifice (d=0.025 cm) the skimmer should be placed 2.5 centimeters (1 inch) upstream from the orifice.

After passing through the skimmer opening, the gas passes through a series of collimating slits which serve to decrease the amount of background noise caused by the scattering of molecules in the beam.

The sensitivity of a typical molecular beam system is approximately one part per million. To improve the sensitivity of the system, the molecular beam must be modulated using a beam chopping mechanism located downstream of the skimmer.

2.1.4. Beam Modulation.

The use of modulation or "chopping" of a molecular beam to enhance the signal in a mass spectrometer has been investigated by a number of researchers. ^ ’9 ’1 3 '1 5 Milne and Green^ suggest that using a modulated or pulsed beam, together with a phase-sensitive detector

(i.e. a lock-in amplifier), results in significant improvements in the signal. Among the advantages are; (1) improved signal to noise ratio, 15

At yon source qas S D0 - 0.025 cm S. mE Ds ■ 0.07c cm c Pn in torr c%>

& c 300,

500.

20 40 to 80 100 120 140 ItO

Figure 2.5 - Effect of sampling orifice-to-skimmer distance on intensity of molecular beam (from reference 9). 16

(2 ) discrimination against background ("residual") and scattered gases, and (3) discrimination and identification of reaction products of the beam gas with the ion source.

Successful implementation of beam modulation requires that a method of chopping the beam be coupled with the detector output circuitry of a phase sensitive device. In practice this is readily accomplished by chopping the beam with a vibrating reed, a tuning fork, or a motor driven sector and using a commercially available lock-in amplifier to isolate the proper frequency of the chopper.

The lock-in detector commonly used to accomplish molecular beam modulation consists of a narrow band amplifier in conjunction with a phase sensitive detector. During operation the amplifier has its passband locked to the frequency of the chopping mechanism. The

signal generated passes through an amplification circuit and an

initial noise reduction occurs. A reference signal is generated by

the beam chopper and is sent to the lock-in amplifier. In addition,

the chopper chops the molecular beam at the same frequency as the

reference signal. The chopped beam is detected by an multiplier and this modulated signal is also sent to the lock-in

amplifier. When the reference signal is compared to the signal of the molecular beam at the phase sensitive detector, and when the beam and

reference signals are in phase, a net dc signal arises. Any noise present in the signal appears as an ac fluctuation and is filtered

out. In this manner the signal to noise ratio is greatly improved. A 17 schematic diagram of the modulation system wiring is given in figure

2 . 6 .

The selection of the optimum chopping frequency is critical in order to avoid serious errors in the beam modulation process. The two main considerations necessary to obtain proper modulation have been

13 9 discussed by Fite and are summarized by Stearns et al . The first consideration is that the chopping frequency must be adequately high to insure that the Fourier components of the fluctuations of the background gas pressure are minimal. In other words, the chopping frequency should be fast compared to the time necessary for pump-out of non-condensible species introduced into the detector region.

Second, the chopping frequency should be low enough to maintain coherence of the beam. Coherence of a pulsed beam is easily destroyed by velocity spreading.

For the special case of a supersonic molecular beam, Miller^ has shown that the second consideration described above is negligible. He suggested that for beams with sufficiently high Mach numbers (> 5) there is a narrow distribution of velocities. As a result of this narrow spread the coherence of the beam is insured even when the chopping frequency is high. Therefore, only the first consideration need be considered when dealing with a supersonic molecular beam system.

It is difficult to quantitatively establish a criterion for the optimum chopping frequency that satisfies the requirements described 18

Has* Spectrometer

Recorder Output Oscilloscope Output

Spectrum Trigger

Osc illo6cope

External Trigger

Channel A

Recorder

Channel 1

Lock-in Amplifier Channel 2

Signal Reference Output ___ 1 !

to Chopper Drlver

Figure 2.6 - Modulation system wiring diagram (from reference 5). 19 above based solely on the system pumping time constant. The selection depends on the amount of discrimination desired and is often best accomplished through a repetitive experimental process. One method for experimentally determining a suitable frequency has been described

15 by Milne and Greene and involves observing the ratio between the signal for an unstable species and the signal for a stable species as a function of chopping frequency. Through this technique, they found that a frequency of 15 Hz resulted in adequate beam modulation in their Bendix mass spectrometer system.

2.1.5. Quadrupole Mass Spectrometry.

The sampling system previously described has been coupled with a variety of mass analyzers, ranging from existing time- of-flight instruments'* ’ to more modern quadrupole mass

9 19 21 filters ,x . The function of any mass analyzer in the sampling of high pressure gases is to ionize the molecular beam that enters the mass filter and separate the various ions by virtue of their mass-to-charge ratios. The majority of high temperature-high pressure sampling systems currently in operation use quadrupole mass analyzers due to their moderate cost, relatively high resolution, and easy adaptation to molecular beam systems.

Quadrupole mass spectrometry uses four parallel rods arranged such that their axes lie on the corners of a square. A typical quadrupole system is schematically represented in figure 2.7. The ion beam is 20

°.°oFcTOi • 0 °

SAMPLE QUADRUPOLE INLET MASS FILTER ION SOURCE

Figure 2.7 - Schematic of quadrupole mass filter (from reference 22). 21 accelerated into the center of the rod arrangement along the z-axis.

One pair of diagonally opposed electrodes is held at a positive dc voltage and the other pair at a negative dc voltage. An alternating frequency (rf) voltage is superimposed on the first pair of electrodes, and a second rf voltage (180° out-of-phase with the first) is applied to the second pair. The majority of the ions introduced into the filter oscillate with an increasing amplitude and eventually strike one of the electrodes. However, one particular m/e ratio, determined by the rf potential and the frequency, may pass completely through the analyzer and be detected. The mass detected is related to the frequency, f, the rod separation, r0, and the accelerating voltage, V, by the equation:

m = 0.136Vf2/r02 (2.6)

By scanning the rf frequency or dc potential a mass spectrum may be obtained. A distinct advantage of this type of system is that quadrupole instruments produce a linear mass spectrum as compared to the squared mass scans produced by time-of-flight instruments. A disadvantage associated with quadrupole instrumnets is the decrease in sensitivity as the mass increases.

2.II. Previous Experimental Studies.

Before beginning a discussion of the mass spectrometric identification of volatile ceramic corrosion products, it is appropriate to briefly discuss the simple oxidation and chlorination of these materials. 22

2.11.1. Oxidation.

2.11.1.1. Si and SiC.

The oxidation behavior of Si-based ceramics has been widely

93 9fi studied . The number of classic papers describing the behavior of

Si and SiC in oxidizing environments is staggering. Therefore, any attempts to discuss the great number of achievements made in recent years would be fruitless. In light of this fact, only a few of the classic papers on oxidation will be discussed here.

The behavior of Si-based ceramics in oxidizing atmospheres is complicated by the fact that two distinct regions of oxidation behavior exist. In regions of suitably high oxygen partial pressure

_ 3 (> 10 atm) passive oxidation occurs. In this process, the oxygen present reacts with the Si or SiC substrate to form a silica layer according to the reactions:

SiC(s) + 2 ° 2 (g) s S i 0 2 (s) + C 0 2 (g) <2 -7>

S i (s) + ° 2 (g) = S i 0 2 (s)

This reaction results in a finite weight gain in the sample. The kinetics of this reaction have been found by numerous researchers to

follow a parabolic rate law.

As the partial pressure of oxygen in the system is decreased, a

transition from the passive oxidation regime to a regime of active oxidation is observed. This transition has been studied extensively

27 by Gulbransen and his coworkers . They studied the oxidation behavior of SiC at 1300 °C for a variety of oxidizing conditions. 23

They conducted experiments in 9 x 10”^, 0.04, 0.1, and 0.5 Torr oxygen. The results of their analysis are presented in figure 2.8.

In this figure, weight loss implies an active oxidation process:

“ =(.) + ° 2 (g) » S“>(g) + C°

The presence of CO in the active oxidation reaction gas was verified by mass spectrometric analysis of the condensed product gas stream.

Weight gain, on the other hand, indicates that a passive oxidation process is occurring by equation 2.7. Figure 2.8 indicates that passive oxidation of SiC occurs between 0.04 and 0.1 Torr oxygen pressure at 1300 °C.

Gulbransen et al., also conducted research on the effect of temperature on the active oxidation process. In these experiments, O they maintained an oxygen pressure of 9 x 10 Torr while varying the temperature from 1250 to 1400 °C. While active oxidation occurred for all of these conditions, the rate of weight loss was considerably lower at 1250 °C than it was at 1400 °C. In fact, at temperatures below 1250 °C they suggest that a transition to the passive oxidation regime may occur.

Perhaps the most classic paper on the active oxidation regime is

OO that presented by Hinze and Graham . They studied the behavior of both Si and SiC in regions where various oxidation processes occur.

In particular they studied the behavior of Si in the temperature range of 1400 to 1500 °K with the oxygen partial pressure varying from

10"'’ to 10"^ atm and also for SiC in the range of 1400 to 1650 °K and 24

c 3 | 0.5

• | as

% i M i£ L 0

Him, mirt

Figure 2.8 - Oxidation of SiC. effect of pressure at 1300 °C: A, 9x 10 ; B, 4 x 10 ; C, 1 x 10' ; D, 5 xlO Torr (from ref. 27). 25

1 0 " 5 to 1 0 '^ atm.

For the case of Si they observed a severe active oxidation process:

Si(s) + 1/2 0 2 (g) = Si0(g) (2.10) at these low oxygen partial pressure and elevated temperatures. The morphology of the oxide scale formed on the surface of the sample was described as thick and spongy. This oxide scale was believed to be

formed by condensation of the products of reaction of the SiO vapor with reactant oxygen on the sample surface.

They observed less severe active oxidation in the case of SiC.

In fact their work suggests that in order to achieve active oxidation

in this system at 1750 °K, the partial pressure of oxygen must be

lower than 1 0 ’^ atm.

2.II.1.2. Si3 N4 .

Since the early 1960's there has been a tremendous interest in

the use of Si3N4 for high-temperature structural ceramic applications.

As a result of this interest, the behavior of Si3N 4 in oxidizing

environments has also been analyzed by a variety of researchers^

33 The materials studied range from powders to chemically vapor

i) / deposited samples . As in the case of SiC, a detailed review of the

current literature on the oxidation of Si3N4 would be pointless.

Therefore, only a brief summary of the oxidation behavior of silicon

nitride will be presented here.

Unlike Si and SiC there exists mainly one regime of oxidation in 26 the case of Si^N^ under normally obtainable conditions. This regime is passive and SiC>2 is produced on the silicon nitride substrate by the reaction:

Si3N 4 + 3 02 = 3 Si02 + 2 N 2 (2.11)

However, the oxidation behavior of silicon nitride materials is not as simple as described by the above equation. The role of trace elements and densification aids in the oxidation process is complex. These trace elements migrate to the surface of the material and result in the formation of many different phases in the materials. A detailed description of the role of impurities in oxidation of Si^N^ is not necessary for this work.

The composition of the oxide scale may also be quite complex.

29 Tripp and Graham determined that the oxide scale consists mainly of a- and enstatite. They also noticed a dramatic increase

in the rate of oxidation as the temperature was increased above 1450

°C. They attribute this increase to a melting of the oxide scale followed by an increase in transport of oxygen and subsequently a more

30 rapid rate of oxidation. Kiehle and his coworkers also noticed this

increase, however their analysis showed the existence of cristobalite, enstatite, akermanite, , and diopside in the oxide scale formed.

Although the active oxidation of Si^N^ is a rarely encountered circumstance, this brief summary would be even less complete if no discussion of this subject was presented. Active oxidation of Si3N4 27 occurs in a manner greatly different from the process occurring in

SiG. In Si^N^ the active oxidation process occurs by a reaction

32 between the Si3N4 substrate and the growing oxide film :

Si3N4 + 3 Si02 = 6 SiO + 2 N 2 (2.12)

However this reaction proceeds at a very slow rate even as the temperature is increased to 1400 °C and the oxygen partial pressure is maintained at a very low level.

2.II.2. Chlorination Reactions.

2.II.2.1. Si.

The high temperature reaction of chloride or chlorine and silicon is widely employed in the production of -state

35 electronic devices . Although this process is important in the production of high quality electronic components, very little literature is available regarding this reaction. However, Lin has studied this system along with the reaction between HC1 and silica.

Lin used a flow reactor coupled with a quadrupole mass spectrometer in order to directly analyze the gaseous corrosion

37 38 products ’ . H e studied the reaction between solid silicon test samples and 8 % HC1 in at temperatures varying from 800 to

1200 °C. The reaction between HC1 and Si produces SiCl4 :

Si(s) + 4 HCl(g) * SiCl4 (g) + 2 H 2 (g) (2.13)

Upon electron impact, the SiCl4 generated by the reaction fragments to form SiCl+ , SiCl2+ , and SiCl3+ ions. Lin's data indicate that this fragmentation process occurs to a greater extent as the temperature of 28 the reaction is increased.

Lin also made appearance potential measurements on this system at 800 and 1200 °C. The appearance potentials of SiCl+ , SiCl2+ ,

SiCl^*, and SiCl^+ determined were 16.0, 12.6, 12.5, and 11.5 ± 0.7 eV respectively at 800 °C. At 1200 °C, on the other hand, the appearance potentials determined were 1 2 .0 , 1 0 .0 , 1 2 .2 , and 1 2 . 0 ±

0.7 eV respectively. He suggests that the similarity in the potentials of SiCl2+ , SiCl-j"*", and SiCl^+ at 800 °C indicates that all of these species are formed during the reaction. Lin also suggests that the lower appearance potentials for SiCl+ and SiCl2+ at 1200 °C are a result of increased fragmentation of these species as the temperature is raised.

Lin also examined the reaction between Si02 and a 100 % HC1 gas stream at temperatures ranging from 1200 to 1400 °C. This interaction produces SiCl^ by the reaction:

Si02 (g) + 4 HCl(g) - SiCl4(g) + 2 H 2 0 (g) (2.14) he observed very little reaction in this system until the temperature reached 1300 °C. Above 1300 °C he observed extremely small quantities

*4* • *4" + of SiCl , SiCl2 , and SiCl^ , along with measurable amounts of oxygen.

The quantities of SiCl2+ observed, although small, were greater than the amounts of SiCl^ and 0 2 + measured. He explains this observance of oxygen and lack of SiCl^ formation by the following reactions:

Si°2 (s) + 3 HCl(g) * SiCl3(g) + 1/4 0 2 (g) + 3/2 H 2 0 (g) (2.15)

Si02(s) + 2 H C l (g) * SiCl2(g) + 1/2 0 2 (g) + H 2 0 (g) (2.16) 29

These reactions explain the existence of oxygen, however they do not explain why the reaction to form SiCl^ does not occur, although equilibrium calculations predict SiCl^ to be the most prevalent species produced. Lin suggests that perhaps the kinetics to form a

Si-Cl species with a large number of Cl atoms (SiCl^ and SiClg) are unfavorable when compared to the kinetics of formation for SiCl and

SiCl2 .

2.II.2.2. SiC.

The behavior of SiC in chlorine containing gases (i.e.- Cl2 and

HC1) has not been widely studied. In fact, only papers on the thermodynamics of this system exist in the current literature. Jeffes and Alcock studied this system from the viewpoint of vapor transport OQ of these materials by chlorine containing gases . They considered the general reaction:

\ Y y + 1/2(ax + by)Cl2 * x XCla + y YClb (2.17) as a means of producing single of high melting compounds.

In order for this process to be efficient the following criteria must be met:

1. The chlorides must be volatile,

2. The chlorides must have adequately large heats of formation, and

3. The partial pressures of XCla and YClb must be approximately in the stoichiometric ratio of the compound ^Yy.

In light of these required criteria they examined the partial pressures of the volatile species produced in great detail. In 30

particular they looked at the reaction:

SiC + 4C12 ** SiCl4 + CC14 (2.18) as a means to transport SiC for subsequent growth. They present the results of their analyses in the form of matched integral free energy diagrams. The results from this system (figure 2.9) indicate that very large differences exist between the free energies of formation for SiCl4 and CC14 . This diagram indicates that very small amounts of CC14 are encountered during the equilibrium of Cl2 and SiC. The results also indicate that the pressures of SiCl4 and

CCl^ are far from equal. This implies that transport of SiC by chlorine is a highly inefficient process.

As a result of the undesirable transport of of SiC by chlorine,

Jeffes considered the use of HC1 as a transport gas. The transport reaction of interest in this situation is:

SiC + 4 HC1 ** SiCl4 + CH4 (2.19)

An Ellingham diagram describing the Si-C-Cl-H system is shown in figure 2.10. From this diagram they conclude that transport of SiC by

HC1 should occur with reasonable efficiency between 550 and 950 °C.

This reaction will, however, produce a considerably lesser amount of

SiCl4 than in the case of Cl2 because hydrogen has a comparable affinity for chlorine as has silicon^®. This implies that an excess of HC1 must be maintained if an adequate amount of SiCl4 is to be generated for efficient transport to occur. 31

IS

• IS

10 CCI,

Figure 2.9 - Matched integral free energy diagram for the Si-C-Cl system (from reference 39). 32

T»mp«rotur#*C 9 0 0 1000

20 AO* ( kcol.) 30

4 0

9 0

6 0

6 0

Figure 2,10 - Ellingham plot of the Si-C-Cl-H system (from reference 39). 33

2.11.2.3. Si3 N4 .

The behavior of Si^N^ in the presence of chlorine has not been investigated. There appears to be no major studies of Si3N4 and CI2 in the current literature.

2.11.3. Mixed Oxidation/Chlorination Reactions.

2.II.3.1. Metals and Superalloys.

The behavior of metals and superalloys in high temperature environments containing both oxygen and chlorine has been studied by various researchers^"^. Several researchers have made use of a high pressure mass spectrometer, as well as conventional thermogravimetric analysis, to study these reactions.

The combination of the mass spectrometric identification of the volatile species and the thermogravimetric analysis results allows for

the clarification of the roles of the volatile species during these mixed-oxidation chlorination reactions^"5 . Of particular interest

to the work conducted in this study is the work done by Jacobson,

McNallan and Lee^ on the behavior of cobalt in mixed oxygen-chlorine gases. They studied the behavior of cobalt in mixtures of 1% CI2 , x%

O 2 in Ar at 650 °C. The amount of oxygen was varied from 1 to 50 percent.

The weight change behavior of the cobalt in 1% CI2 , 1% O 2 was quite complex in this case, showing an initial weight gain followed by a linear weight loss. This puzzling behavior was explained, in part, by coupling the TGA analysis observations with the volatile 34 products identified with the mass spectrometer. They concluded that at low oxygen contents the principal corrosion product is C0 CI2 vapor and mass transfer in the gas phase is the rate controlling process.

In contrast to this observation they observed that the were the major product formed during reaction of cobalt with gas streams containing larger amounts of oxygen. In the high oxygen content gases the condensed phase cobalt chlorides were formed below the oxide scale. The escape of C0 CI2 formed by the vaporization of this condensed phase chloride, even though covered by an oxide layer, was found to be rate controlling.

They also made use of the quantitative identification capabilities of the mass spectrometer to examine the existence of oxychloride compounds in the Co-O-Cl system. These species have been suggested as important players in the mixed corrosion of metals in oxygen-chlorine environments. Their work showed that no stable oxychlorides were formed during the reaction of Co with mixtures of oxygen and chlorine.

2.II.3.2. Si.

The oxidation of silicon in chlorine-containing ambients is an important process for the production of films for Si-based electronic components. As a result of this vast interest, the role of chlorine in this process has been extensively researched^In this system the gaseous chlorine is incorporated into the growing SiC>2 film. The chlorine has been postulated to segregate primarily at the 35

Si-SiC>2 interface^ ’ . With the popularization of secondary ion mass spectrometry (SIMS), the actual distribution of the chlorine in the

51 55 oxide film has been studied in great detail . These studies have shown that a finite quantity of chlorine is distributed throughout the oxide, with an increasing amount near the Si-SiC>2 interface (see figure 2.11). This observation has been confirmed by means of Auger sputtering profile (ASP) studies'^’

Numerous researchers have studied the of the Si-

SiC>2 interface, and observed a separation of the growing SiC>2 layer from the Si substrate. Hess and McDonald"^ attributed the separation to stresses induced during cooling. This theory has been disputed by

Monkowski and his coworkers'*'*Their work suggests that the separation is due to the formation of gas bubbles which are trapped at

the interface. They suggest that the chlorine actually reacts with

the SiC>2 layer to produce a silicon oxychloride phase. This phase

subsequently becomes gaseous and is essentially trapped at the

interface and forms gas bubbles. The composition of the silicon-

oxygen-chlorine phase is disputed. Tsai et al."*^ suggest the

existence of a phase with the composition of Si2 0 gCl2 . They arrived

at this composition as a result of electron diffraction patterns which

show a noncrystalline phase of lower Si and 0 concentrations than

those determined for normal SiC>2 . They predicted that the silicon oxychloride results from a phase separation process often seen

in glasses**®. Kirabayashi and Iwamura**^ suggest that the 36

10«

160 «in. . c i o * 1 i C o

CIO20

Lc a o

§ i o ,B

U

j o 1 7 0 100 200 300

Depth in Oxide Film (nm)

Figure 2.11 - Chlorine concentration profiles for samples prepared by oxidation of Si at 1100 °C in 0.3 % CI2 /O2 ambients (from ref. 51). 37 composition of the phase is Si20Clg. They claim to have observed the compound by means of . However, analysis of their results (figure 2.12) alone does not provide a convincing argument for their prediction.

The structure of the silicon oxychloride compound formed has been

62 postulated by Kriegler . He suggests that the chlorine is

substitutional for oxygen, and serves to create non-bridging bonds in

the same manner as conventional network modifiers. He does not present convincing evidence of the actual existence of this phase.

The actual existence of such a phase is generally disputed.

2.II.3.3. SiC.

The behavior of SiC ceramics in mixed oxygen-chlorine

environments has not been as extensively studied as the behavior of metals, superalloys, and Si in similar environments. Only McNallan

and his coworkers at the University of Illinois at Chicago have

63 studied this area . They conducted a brief series of

thermogravimetric analyses at 900 °C on low-cost SiC samples in various flowing gases. They compared these results with a set of

thermodynamic equilibrium calculations for the Si-O-C-Cl system.

Four types of low-cost SiC materials were used in their studies.

The materials ranged from sintered a-SiC samples containing as a

aid (denoted samples A and B) to hot-pressed materials

containing A1 as a densification aid (F) to siliconized SiC which

contained 15% free Si (D). They found tremendous variation in the 38

WAVELENGTH 00 9 10 12 14 16 16 (820 2530 i 1111

in D 4

1600 1400 1200 1000 800 600 600 500 400 WAVE NUMBER (CM-1)

Figure !.12 - Infrared spectrum showing existence of silicon oxychloride phase (after reference 61). weight change versus time behavior of these four materials in a 2%

CI2 , 2% C>2 - Ar gas stream at 900 °C. These results are graphically presented in figure 2.13. The behavior of the two sintered materials which contain boron (A,B) are very similar, that is the slope of the lines is nearly equal. The siliconized material (D) and the hot- pressed material (F) differ greatly from the sintered samples. These two materials show much lower rates of weight loss, that is the slopes of these lines approach zero. They cite this behavior as very surprising and offer no real explanation for this occurrence.

The effect of gas composition on the rate of weight loss is shown in figure 2.14. The rates of weight loss for the sintered a-SiC tested vary significantly depending on the composition of the gas stream. The highest rate of weight loss was observed for the sample tested in the 2% CI2, 2% O 2 - Ar gas stream followed by nearly equal rates for the samples heated in 2% CI2 - Ar and 1% H 2 , 2% Cl2 - Ar.

They observed little or no corrosion for the materials tested in 2%

CI2, 20% O 2 - Ar and 15% HC1 - Ar.

They also express surprise for the accelerated corrosion rate in the 2% CI2 , 2% O 2 gas. They reason that the reaction of small amounts of O 2 and CI2 with SiC produces a Si02 layer separated from the SiC substrate by a finite layer of free C. This free-C layer reacts with the small amount of oxygen present to form CO or CO2 . This carbon layer may lower the rate of corrosion due to the fact that CI2 has a lower affinity for C than Si. Therefore, removal of this pseudo- 40

U. e

0.05.0 00 0.0 120 15 0 18.0 TIME (HR)

Figure 2.13 - Variation of rate of weight loss as a function of different types of low-cost SiC materials (from reference 63). 41

e s

o »*K2-»CU-*7«Xr K

< x

e

o

O

0.0 S O t o 0 0 12 0 ISO 18.0 TIME (HR)

Figure 2.14 - Effect of gas stream composition on the corrosion of low-cost sintered a-SiC (from reference 63). 42 protective layer may create new SiC surface for reaction with the gas.

The lack of reaction in the case of the 20% O 2 gas stream may then be assumed to result from the formation of a protective layer on the SiC substrate more rapidly than in the case of small oxygen content gas streams. This more rapid formation of the protective Si02 layer may prevent the formation of the C-rich layer.

In their discussion of figure 2.14 they fail to mention why so little weight loss was observed for the sample heated in the 15% HC1 flowing gas stream.

2.II.4. Silicon Oxychlorides.

As discussed previously, compounds composed of silicon, oxygen, and chlorine have been suggested as the cause of bubbling of oxide films grown on silicon in the presence of both oxygen and

55 58 59 chlorine ’ ’ . However the existence of these compounds has only been reported in this system once^. The existence of silicon oxychloride compounds of the general formula SinOn _-^Cl2n+ 2 , was first reported by Schumb and Holloway^ in 1941. They prepared the silicon oxychloride compounds by passing a mixture of oxygen and chlorine over heated silicon.

The condensed products from this reaction were distilled to separate SiCl^ from the various silicon oxychloride compounds. The resulting mixture was fractionally distilled and the contents of Si,

Cl, and 0 in each fraction were determined using a gravimetric technique^. In this techniques the chlorine was isolated by 43 formation of chloride and the Si was isolated by precipitation of SiC>2 from the mixture. The amount of oxygen present was determined by difference. By using this technique they determined boiling points of Si^OClg, Si302Clg, Si404Clg, Si40 3Cl^Q, Si304Cl^2 >

^ 6 ^ 5 ^ 1 4 ’ an(* ^ 7 ^ 6 ^ 1 6 '

In 1947, Schumb and Stevens^ produced the first two members of

the SinOn _^Cl2n + 2 series by the partial hydrolysis of silicon

tetrachloride in dilute anhydrous diethyl ether solution by means of

the addition of moist ether. The reactions governing this hydrolysis are:

2 SiCl4 + H20 * Si2OCl6 + 2 HC1 (2.20)

3 SiCl4 + 2 H20 ** Si302Cl8 + 4 HC1 (2.21)

Further work by Schumb and Stevens^ on the partial hydrolysis of silicon tetrachloride provided information on the mechanism of silicon oxychloride formation. They proposed the formation of an intermediate trichlorosilanol by the reaction:

(C2H5)20-H-0-H + Cl-SiCl3 a* Cl3Si-0H +

HC1 (C2H 5)20HC1 + HC1 (2.22)

This trichlorosilanol subsequently reacts with silicon tetrachloride

to form a silicon oxychloride compound:

Cl3Si-0H + Cl-SiCl3 ** Cl3Si-0-SiCl3 + HC1 (2.23)

Higher mass silicon oxychlorides are then formed by reaction of

existing silicon oxychloride compounds with trichlorosilanol:

Cl3Si-0H + Cl-SiCl2-0-SiCl3 = Si302Clg + HC1 (2.24) 44

68 Beattie and McQuillan also analyzed the first member of the

Sin0 n _-^Cl2n + 2 series, Si2 0 Clg, produced by a gas phase reaction of

SiCl4 and vapor. This gas phase hydrolysis occurs by the reactions:

SiCl4 + H20 ** Cl3Si-OH + HC1 (2.25)

2 SiCl4 + H20 a Si2OCl6 + 2 HCl (2.26)

They report infrared analysis of the condensed products, however, no

such results are presented in their communication.

The infrared properties of trichlorosilanol, Cl-jSiOH, and the

first member of the silicon oxychloride series, Si2 0 Clg, were

69 investigated by Rand . Using silicon oxychlorides prepared by hydrolysis of SiCl4 he found that absorption maxima occurred at 2.22 and 2.70 microns. These maxima are a result of a non-hydrogen bonded hydroxyl associated with the trichlorosilanol molecule. He also

found an absorption maxima at 6.48 microns associated with the hexachlorodisiloxane molecule, Si2 0 Clg.

2.11.5. SiCl4 Mass Spectrometry.

The mass spectrometry of silicon tetrachloride has been widely

investigated. Perhaps the pioneering work is that done by Vought in

1946^®. He studied the dissociation of SiCl4 by electron impact

ionization in great detail. Vought used a mass spectrometer system

to determine appearance potentials of the various SiClx+ ions, as well

as a typical spectrum at 75 eV ionization potential. The results of his analysis are summarized in Table 2.1. 45

Vought gives probable processes for the formation of each of these ions from SiCl^. Naturally, the mechanism for formation of

SiCl4+ from SiCl4 is a simple ionization process:

SiCl4 + e' ** SiCl4+ + 2 e' (2.27)

The energy necessary to complete this process is simply the ionization potential of SiCl4> The remaining ions are subsequently formed by dissociation of SiCl4 followed by ionization of the dissociated product. Consequently the energy to produce the most dissociated ion

(SiCl+) is greater than the energy necessary to produce ions from molecules which have been dissociated to a lesser extent.

71 Agafanov and his coworkers examined the mass spectrum of silicon tetrachloride using an ionizing energy of 50 eV. The monoisotopic spectrum from their studies is compared to that of Vought

79 and also to that obtained by Svec at 90 eV and that obtained by

Ban^ at 70 eV in Table 2.2. 46

Table 2.1

Appearance Potential and Relative Intensity Data For Silicon Tetrachloride (from ref. 70).

Ion Relative Intensity (75 eV) Appearance Potential (eV)

SiCl4+ 56 11.6 ± 0.2

SiCl3+ 100 12.9 ± 0.2

SiCl2+ 4.1 18.4 ± 0.3

SiCl+ 13 20.5 ± 0.3 47

Table 2.2

Monoisotopic Spectra of Silicon Tetrachloride

Relative Intensity

Ion 50 eV71 Vought7® Ban73 90 eV

SiCl+ 23.8 13 19 31.1

SiCl2+ 6.9 4.1 8 7.8

SiCl3+ 1 0 0 1 0 0 1 0 0 1 0 0

SiCl + 53.4 56 52 60.5 CHAPTER III

EQUILIBRIUM CALCULATIONS

3.1. Introduction.

In order to thoroughly study the high temperature equilibria occurring in ceramic systems, a multitude of condensed and gaseous species must be considered. However, incorporation of a variety of species severely complicates the calculation of the equilibrium phase assemblage. To solve such a complex system, computer programing techniques have proven to be most useful.

The majority of recent authors have based their programs on the early work of Brinkley^’ ^ and White et al.^. The Brinkley method expresses the abundance of a given species in terms of the abundance of another arbitrarily chosen (usually the most predominant) species by means of an equilibrium constant. The method described by White, on the other hand, does not make this distinction between the constituent species. By focusing the attention on the chemical potentials of the reacting species, he showed that numerical solutions

are obtainable by minimizing the total free energy of the system.

White's method easily lends itself to standard programing

techniques, and thus, has been more widely applied than the approach

described by Brinkley. It is important to note that when using the

48 49 free energy minimization method the mass of each element must be conserved, and either the pressure or the volume of the system must be held constant.

3.II. Method of Calculating Equilibrium.^

By definition the Gibb's free energy of a system may be written a s :

I xi*i (3.1)

Where: x^ - the number of moles of a substance i . g^ = the chemical potential of the substance i.

The chemical potential, in turn, is given by the expression:

gi = gi° + RTlnai (3.2)

Where: g ^ 0 = the chemical potential of the substance i, in the pure state. R - the gas constant. T = the absolute temperature, a^ = the activity of species i.

Combining equations (3.1) and (3.2), the total Gibb's free energy of the system may be represented as:

RTlnai) (3.3) ^ x i<6 i° +

Treating the system in the ideal case, the activities of the species may be taken as:

1. For the gaseous species:

x i ai " Pi (3.4a) X 50

Where: p^ = the partial pressure of the gaseous species i. X = the total number of moles in the gas phase. P = the total pressure of the system.

2. For the condensed species:

aj_ = 1 (3.4b)

It is convenient at this point to represent the total free energy as the dimensionless quantity, G/RT. Thus, through equations (3.3),

(3.4a), and (3.4b), we may write: 0 1 1 bO

r ♦ri G/RT ’ x i ’ — + In P ■ X *i • all . RT X gases ♦ X (3.5) all RT

Considering a system with m gaseous species and n condensed species, equation (3.5) is given as:

* r on m Si ’ x i ' -■ + InP + In G/RT X Xl1 RT . X 1 i- 1 (3.5a) n Si X Xl RT i-n

The chemical potential g° of a species i may also be expressed by

the relation:

g° - G° - H° 298 f ,298 (3.6) 51

This expression makes it possible to relate the chemical potential to readily obtainable terms.

As mentioned earlier it is imperative that the mass of each element be conserved when performing such equilibrium calculations.

The equation to account for mass balance is given by Eriksson^ as:

m n

I aijx i + I aijxi - bj <3-7> i=l i=l

Where: a^j - the number of atoms of the jcn element in one molecule of the i substance, bj >=■ the total number of moles of element j . j - an integer ranging from one to Z , where Z is the total number of elements.

Through use of the latter equations ((3.5a), (3.6), and (3.7)),

LaGrange's method of undetermined multipliers is used in conjunction with a series of Taylor expansions about the number of moles of each species (chosen arbitrarily) to deduce the equilibrium composition resulting in the minimum free energy value. Using these equations this method readily lends itself to an iterative computer solution technique.

3.III. SOLGASMIX-PV Computer Program.

3.Ill.1. General Information.

The computer program used in this research was originally developed by Eriksson^ in the early 1970's as an alternative to the

7 ft much slower, more complicated HALTAFALL program used by Ingri and

7 Q his coworkers. Eriksson's program was further revised by Besmann 52

in 1977 through the ideal gas law (PV=nRT). This modification allows for the computation of equilibria at constant total gas volume while varying the system pressure. Besmann also noted that the activity coefficient, 7 ^, may be incorporated into the calculations by means of the relationship:

X

The resulting program, SOLGASMXX-PV, was obtained on magnetic tape, and has been used to calculate all the high temperature equilibrium compositions of this study. The program is written in

Fortran and has been formatted to operate on both the IBM 4341/CMS and the VAX 11/780 systems present at the University.

3.Ill.2. Operational Parameters.

In its present form the program is designed to operate with a maximum of 10 elements, 99 substances, and 10 mixtures. It is important to note that the gas phase is considered a mixture and each substance is either a gaseous or condensed species, or a member of a condensed phase mixture.

There are a number of input variables of notable importance in the most recent version of the program. These variables are summarized in, and the complete program listing is attached in Appendix A.

The revised form of the SOLGAS program also contains numerous subroutines for special situations. The two most important of these are the subroutines FACTOR and SPEQUA. FACTOR enables the program to 53

handle non-unit activities (ie- non-ideal solids) and variable stoichiometries, while SPEQUA allows the program to calculate quantities which are derivable from the determined equilibrium composition. In each subroutine the user is responsible for writing the applicable relationships and inserting them into the program through the appropriate variables. The use of these special condition subroutines was not greatly explored in this research.

3.IV. Systems Examined.

The research conducted has proven that the SOLGASMIX-PV program is most useful in studying the high temperature equilibrium relationships present in the systems under consideration. In accordance with the major thrust of this research (Corrosion of Si-

Based Ceramics) exhaustive studies were conducted on a multitude of systems. These included Si, SiC, SigN^, and SiC^ in HCl/Ar, C^/Ar,

HF/Ar, and I^O/Ar atmospheres. In addition Si, SiC, and SigN^ were examined in various C^.C^/Ar and HCl.C^/Ar mixtures.

After considerable time and effort the program was successfully

"debugged" and produced output in the expected format. Various changes were made in the program to correct for errors that resulted from slight system differences. These changes were made such that the operation of the revised program was not. altered from the initial program created by Besmann. Slight modifications have also been noted

(by asterisks) in the program listing to allow data to be input 54

80 directly as it appears in the JANAF tables. (Since the program utilizes thermodynamic data in joules and the JANAF tables present such data in calories, program steps were added to accommodate this difference.)

By means of the existing thermodynamic tables®®"®^, a data base was generated for use with the program. The JANAF tables have been used as the major source for the data base used in the studies.

While they are not the most recent set of tables for a number of compounds, they are the most complete, consistent set of tables, and present data in a convenient form.

As expected, the major problem encountered in the work was the location of reliable thermodynamic data. Unfortunately even the most complete set of tables lacks data on numerous compounds of interest.

However, in the interest of producing results strictly for comparison to experimental work, the runs were performed using the most believable data that is currently available. It should be noted that as more data becomes available, it may easily be added to the data base, and the necessary runs may be repeated.

As mentioned earlier, results have been obtained for a multitude

of systems. The results of these calculations are presented in the next section.

3.V. Discussion of Results.

3.V.I. SiC-HCl System.

The work on this system involved 50 different species, that may 55

be viewed as being candidates for formation in this reaction.

85 Recently Fischman has performed calculations on this system using the same SOLGAS program, which has provided a comparison standard for the present work. Although the majority of Fischman's work concentrated on systems with no Cl present, output given for systems containing both H and Cl provided results in accordance with those generated here.

A typical set of data generated by the program for this system is shown in Table 3.1. Data are given for the equilibrium reaction between 10 moles of SiC and 4 moles of HC1 at four different temperatures (1000, 1300, 1500, and 1700 °K). The program generates the partial pressure of each gaseous species (in a system with a total pressure of one atmosphere) and the molar quantity of each condensed species. The data shown were used to construct a plot of log partial pressure versus temperature for the SiCl„ species. A

Figure 3.1 compares the plot generated using data from the

S0LGASMIX-PV program to a similar one obtained by simply analyzing the reaction:

SiC + x HC1 - SiClx + x/2 H 2 + C (3.9)

Differences in these plots are obvious and are easily explained by the

fact that the SOLGAS program considers all the species that may be present (in minimizing the free energy), and not just the ones taking part in the specific reaction. The consideration of the numerous side

reactions should produce more realistic results. Table 3.1

Typical Data for the Si-C-H-Cl System

Partial Pressure (atm) Species 1000 K 1300 K 1500 °K 1700 K

CC14 (3 (3 13 (3

ch3 1.9 X 1 0 - 9 5.8 X 1 0 ' 8 2 . 1 X 1 0 " 7 5.7 X 1 0 ' 7 1 ■F>

ch4 3.0 X 1 0 ' 2 1.5 X 1 0 ' 3 3.0 X I-* o 8.4 X 1 0 ' 5 1

HG1 9.8 X 1 0 ' 2 3.2 X 1 0 ' 1 4.5 X o 5.2 X 1 0 ' 1

SiCl 9.1 X 1 0 ' 15 6 . 1 X 1 0 ' 10 7.5 X 1 0 ' 8 2 . 6 X 1 0 ' 6

SiCl2 7.5 X 1 0 ' 6 1.3 X 1 0 ' 3 1 . 1 X 1 0 ' 2 4.1 X 1 0 ' 2

SiCl3 2 . 1 X 1 0 ' 3 3.0 X 1 0 ' 2 7.1 X 1 0 ' 2 9.8 X 1 0 ' 2

SiCl4 2.9 X 1 0 ' 1 1 . 8 X 1 0 ' 1 1 . 0 X 1 0 ' 1 4.3 X 1 0 " 2

h 2 5.5 X 1 0 ' 1 4.4 X 1 0 ' 1 3.4 X 1 0 ' 1 2 . 8 X 1 0 ' 1

Values listed below are molar quantities.

a-SiC 0 0 0 0

/3-SiC 9.04 9.21 9.29 9.30

C (graph) ° - 8 7 0.79 0.71 0.70 57 a t o t «>

M ici, BJ D W w w Bj B-

*1C1 H (4 < cu 00 -I o -j ooo 1100 1400 UKrtllATUU |°K)

». SOLGASHIX-PV generated plot. a 4J cd -2 w SiCl Bj o w w w aj -3 SiCl o* ►J < M H BJ < 0. SiCl eo o -5 ______s_ 1400 1600 1800 2000

TEMPERATURE ( K) t>. Plot obtained by considering tne reaction? SiC 4 x BC1 • SiCl, 4 x/2 B2 4 C Figure 3. L - Comparison of data obtained for the Si-C-H-Cl system. 58

Referring to the SOLGAS generated plot it is observed that SiCl^ is predicted to be the most prevalent SiClx species below -1520 °K, with SiClg becoming dominant above this temperature. The graph obtained by considering merely the species specific to the given reaction predicts that SiCl^ will be more prevalent than SiCl^ throughout the temperature range tested. This plot also shows that

SiCl2 will become the most prevalent species above 1660 °K, which is not observed in the computer generated plot.

Table 3.1 also illustrates several interesting properties of SiC during reaction with HCl. At all of the temperatures examined, all of the SiC is present in the form of j9-SiC, which is expected. It is also interesting to note that approximately 10% of the SiC will be volatilized at the temperatures tested. This also is not too surprising. Since no oxygen is present, the protective SiC^ layer

(that acts as a corrosion inhibitor) does not form , and as a result a fairly violent reaction occurs. However, it is very surprising to note that the SiC appears to become more stable as the temperature increases. These results are opposite of those intuitively expected, but are consistent with the decrease in the amounts of the SiCl„ A compounds formed at the elevated temperatures.

In an attempt to examine the effect of argon as a carrier gas, the system was also analyzed with a gas phase consisting of 2% HCl in Ar.

However, since the SOLGAS program assigns equilibrium compositions only, the addition of Ar to the system merely serves to lower the 59 partial pressure predicted for the SiClx species. The dilution of the gas phase does not significantly alter the relative abundance of the equilibrium phase assemblage.

3.V.2. SigN^-HCl System.

Thermodynamic data for this system was more difficult to locate, resulting in the consideration of only 31 species in calculations.

As in the SiC system, a plot has been developed representing log partial pressure versus temperature for the SiClx species (Fig. 3.2).

Comparing Fig. 3.2 with Fig. 3.1, it is observed that the plots obtained using the data from the SOLGAS program are quite similar.

The only major difference being that the SiC^ species becomes more prevalent at the higher temperatures in the Si^N^ system.

Although all of the SijN^ appears in the alpha form in the generated results, it must be noted that this is merely because thermodynamic data was unavailable for the beta form. The data generated also indicates that - 8 % of the sample will be volatilized during the equilibrium reaction. However, contrary to the SiC case, more of the sample is volatilized at the higher temperatures (which is to be expected) .

3.V.3. Si-HCl System.

The equilibrium between Si and HCl is of great importance to the industry. Many materials from this system have been used as starting materials in the production of high-grade silicon, and as a result much research has been conducted regarding the iue . - rpia rpeetto fdt fr h Si-N-H-Cl the for data of representation - Graphical 3.2 Figure

log PARTIAL PRESSURE (atm) -10 SiCl system. 1000 Si Cl SiCl HC1 1200 TEMPERATURE (K) SiCl 601800 1600 60 61 thermodynamics of this system.

The data available have been used to compare the results generated by the SOLGASMIX-PV program to those obtained through the use of the 86 Gordon-McBride free energy minimization program. Herrick and

87 Sanchez-Martinez have recently used a modified version of the

Gordon-McBride program to study this system at various temperatures.

To obtain a "quick" comparison, the SOLGAS program was run under conditions similar to one of the runs reported by Herrick. The data for equilibrium at 1600 °K with a Cl/H ratio of 0.1 are summarized in

Table 3.2. Brief examination of this table proves that the SOLGAS results are comparable to those obtained by Herrick for many of the species considered. The slight discrepancies that are observed may be explained by the fact that slightly different values of the heats of formation were used for these compounds.

88 8Q The differing values used by Herrick appear to be from tables ’ that are more current than those given in the JANAF tables (which were used in the SOLGAS calculations).

The SOLGAS program was also used to determine the equilibrium composition when Si is reacted with stoichiometric HC1. The results of these calculations were very similar to those obtained for the SiC-

HC1 system. That is, SiCl^ was predicted to be the most prevalent

species throughout the test temperatures.

3.V.4. Si02-HCl System.

In order to determine the effect of an oxide coating on the Si- 62

TABLE 3.2 Comparison of Free Energy Minimization Data.

System: Si-HCl in Ar System Pressure: 1 atm Reaction Temperature: 1600 °K C1:H Ratio: 0.1

Partial Pressure (atm) Species Gordon-McBride Program87 SOLGASMIX-PV

h 2 0 .875 0.870

SiCl4 1.43 X 1 0 ' 3 8.57 X 1 0 - 4

SiHCl3 2.46 X 1 0 ' 3 2.45 X 1 0 ' 3

SiH2 Cl 2 3.07 X 1 0 ' 4 1.38 X 1 0 ' 3

SiH3Cl 2 . 0 1 X 1 0 " 5 2.63 X 1 0 ' 2

SiH4 9.00 X 1 0 " 7 1.15 X 1 0 ' 6

HC1 9.02 X 1 0 ' 2 8.82 X 1 0 ' 2

SiCl3 4.38 X 1 0 * 3 1 . 0 2 X 1 0 -2

SiCl2 3.30 X 1 0 ' 2 2.46 X 1 0 ‘2

SiCl 3.14 X 1 0 ' 6 5.21 X 1 0 " 6

ci2 1.43 X 1 0 ' 9 1.37 X 1 0 ' 9

Cl 4.14 X 1 0 ' 6 4.12 X 1 0 ‘6

H 5.03 X 1 0 ' 5 5.04 X 1 0 ' 5

SiH 2 . 0 2 X 1 0 ' 7 2 . 0 2 X 1 0 ' 7

Si 9.36 X 1 0 " 8 9.36 X 1 0 ' 8

Si2 2.49 X 1 0 ' 10 2.49 X lO’10

Si3 5.53 X lO’11 5.55 X lO '1 1 63 based substrates, 36 species were considered in the SiC^-HCl system.

Unfortunately the silicon oxychlorides were omitted from the calculations due to lack of thermodynamic data. These compounds may possibly be formed during the course of the equilibrium reaction between SiC^ and HC1, making their omission a serious shortcoming of these calculations.

The resulting data for this system are plotted as log partial pressure versus temperature in Fig. 3.3. From these results it is observed that SiC>2 does not react as completely with HCl as does SiC or Si-jN^. The partial pressure of HCl in all cases is very nearly 1.0 atmosphere, verifying this observation. The only other gaseous species present in meaningful quantities are H2 , ^0, and Cl

(resulting from the dissociation of HCl).

In contrast to the results previously reported, SiCl^ is the predominant SiClx species formed at all the temperatures studied. It should also be noted that the SiCl^ is present at significantly lower partial pressures than in the SiC and Si^N^ systems. As further proof of the lack of reaction in this system, the data predict that no SiC^ will be consumed during the equilibrium reaction with HCl.

3.V .5. Chlorine Environments.

The effect of reaction in pure chlorine environments was also examined by means of equilibrium calculations. The results of this analysis were very much the same as those obtained by the analysis in an HCl environment. The reason for this unexpected degree of iue . - rpia rpeetto fdt o te Si-O-H-Cl the for data of representation - Graphical 3.3 Figure

log PARTIAL PRESSURE (atm) -10 -12 14 -1 -4 system. SiCl 1000 SiCl HCl SiCl 1200 TEMPERATURE (K) 1400 SiCl 1600 1800 64 65 similarity is easily explained by examining the operation of the program.

The SOLGAS program requires that data for the initial phase assemblage be input as molar amounts of the elements involved. This implies that when an HCl environment is desired, equal amounts of H and Cl must be input. Therefore, when an equal amount of pure chlorine is examined the partial pressures of the SiClvA species are not significantly changed since the HCl system is essentially a mixture of pure hydrogen and pure chlorine.

3.V.6. Mixed Oxidation-Chlorination Reactions.

The mixed oxidation of metals in chlorine containing environments has been studied extensively by means of a high pressure mass spectrometer^"*"^. These processes are also of considerable interest to the materials studied in this research. The materials studied here are candidates for applications where the environments contain both corrosive gases and oxygen. The effect of increasing the oxygen content of the corrosive gas stream on the equilibrium composition was analyzed by means of the SOLGAS program.

As discussed earlier, the presence of oxygen in the gas phase results in the formation of a finite amount of Si0 2 on the surface of the material. The calculations performed in this study, as well as numerous experimental works, have shown that this oxide layer is a corrosion inhibitor. Therefore, it is expected that increasing the oxygen content of the gas phase should decrease the amount of 66 corrosion products generated. Equilibrium calculations were performed in order to determine the extent of the corrosion suppression.

Silicon, SiC, and SijN^ were analyzed in CI2 -O2 and HCI-O2 mixtures. In each case the amount of chlorine in the gas was held fixed at 1% in a balance of Ar. The amount of oxygen present was varied as 0, 1, and 5% of the total gas phase. The resulting partial pressure of the SiCl^ species was monitored as a barometer of the extent of corrosion. The results obtained for this analysis are typified in figure 3.4 where the partial pressure of SiCl^ (normalized to the partial pressure of CI2 ) is plotted versus oxygen content of the gas phase for the Si-C^.C^/Ar system. This plot shows that the amount of SiCl^ generated decreases rapidly with increasing oxygen content. In fact, when the oxygen content of the gas phase is

increased above 1 %, the amount of SiCl^ formed is in the part per million range and will most likely not be observed in the mass

spectrometer work.

3.V.7. SiC-HF System.

Thermodynamic data was obtained for 51 species that may possibly be formed during the equilibrium reaction in this system. A

compilation of the meaningful partial pressure data for the system

containing 10 moles of SiC and 4 moles of HF is shown as a log partial pressure versus temperature plot (Fig. 3.5).

By referring to Fig. 3.5 it is obvious that the SiF^ species

dominates at all the temperatures examined. The data generated using 67

O PREDICTED AT 10OO K

- PREDICTED AT 1300 K ■

O

O

P(SiCI ) I ____ 5 O P(CI )

o

o

o 0.0 1 .o 2.0 3.04.0 5 .0 6.0 Percent Oxygen

Figure 3.4 - Partial pressure of SiCl^ versus percent oxygen in the SiC-O-Cl-Ar system. Figure 3.5 - Graphical representation of data for the Si-C-H-F system. Si-C-H-F the for data of representation - Graphical 3.5 Figure

log PARTIAL PRESSURE (atm) -10 SiF. SiF, 1000 SiF,

1200 TEMPERATURE (K)

1400 SiF

1600

o- — 1800 68 69 the SOLGASMIX-PV program predicts that relatively large amounts of H 2 and HF gases will be observed in this system, indicating an amount of

HF in excess of that necessary to complete the reaction. It was also observed that as the temperature increases, the amount of carbon present also increases. This may be attributed to the fact that as the amount of SiFx (due to the volatilization of Si) increases, there is a greater depletion of Si in the SiC sample, thus causing it to become carbon-rich.

3.V.8. Si3N4 -HF System.

Studies in this system included 36 species, of which 3 were condensed. As was the case in the SiC system, the SiF^ species was the dominant species at all of the temperatures examined.

The reactions taking place in this system also appear to be more extreme than in the other systems examined. The results generated using the SOLGAS program indicate that nearly one-half of a mole

(-2 0 %) of Si 3N4 will be consumed in this reaction.

3.V.9. SiC^-HF System.

The JANAF tables provide data on 5 condensed and 32 gaseous species that could possibly be formed during this reaction. Contrary to the Si0 2 -HCl system, data was obtained for the SiOX2 compound.

However, this data is still somewhat suspect since the heat of formation was calculated by comparison to related compounds.

As expected, a relatively large amount of SiC>2 (-1/2 mole) was consumed during this reaction. It appears that the majority of the Si 70 volatilized goes into the formation of SiF^. A plot of log partial pressure versus temperature for this system (Fig. 3.6) shows that

SiF4 is indeed the most predominant species.

3.V.10. SiC-l^O System.

Studies in this system included 8 condensed and 40 gaseous

species. It was obvious from the results obtained that the major

gaseous product formed is H 2 , resulting from the preferential reaction

of oxygen with silicon. The SOLGAS results also predict that

relatively large quantities of CO, HnO, and SiO will also be % ^ observed when the system is examined using the mass

spectrometer.

The data obtained for the condensed species in this system are

interesting and appear to give the expected results. On the average

two moles of SiC are consumed during the reaction. The Si that has been volatilized combines with the oxygen that is present (from the

H 2 O) to form the protective Si0 2 layer that has been confirmed to

exist. It is important to note that at temperatures below 1700 °K the

results predict that a carbon- rich layer will form between the SiC

and the Si0 2 as a result of Si-depletion. At temperatures greater

than 1700 °K, the carbon will also volatilize to form various carbon containing gases (predominantly CO).

3.V.11. Si3N 4 -H20 System.

The runs conducted on this system considered 44 distinct species,

of which 7 were condensed. Results for the equilibrium reaction Figure 3.6 - Graphical representation of data for the Si-O-H-F system. Si-O-H-F the for data of representation - Graphical 3.6 Figure

log PARTIAL PRESSURE (atm) -10 -12 14 -1 4 - 1000 SiF. SiF. 1200 TEMPERATURE (K) 4018001600 1400 SiF. H : & : 2 ° 71 72 between 2.5 moles of Si3N^ and 4 moles of indicate that in this atmosphere, a major transition occurs in the silicon nitride sample.

According to the data generated, all of the Si^N^ will be transformed to SiC>2 (in the form of or cristobalite) or silicon oxynitride (812^0) . While the presence of 312^0 has been verified

in Si^N^ samples following oxidation, it is misleading to believe that a complete transformation will occur. The results generated by the

SOLGAS program are confusing in this case and care must be taken to

avoid misinterpretation.

The thermodynamic data entered for this run indicates that the

likelihood of formation of 8 1 2 ^ 0 is greater than that for Si^N^.

This represents a problem in the future calculations, particularly in

the input amount of each material. As discussed earlier, the SOLGAS

input guide calls for data to be entered as elemental amounts and not

simply the molar quantity of each compound. In other words, data for

Si^N^ is entered as initial moles of Si(reference) and N 2 and not as

moles of SijN^. Thus, when 812^0 is considered in the calculations,

it is formed preferentially to Si^N^ when the program combines the

initial amounts of Si and N.

In reality a total transformation would not be expected to occur.

After heating Si^N^ in an oxygen-containing atmosphere, a 812^0

surface layer would most likely be formed, while the interior of the

sample would remain as silicon nitride. 73

3.V.12. Si02 -H20 System.

The equilibrium composition for this system was calculated using thermodynamic data on 22 separate compounds. Results were obtained for the equilibrium reaction between 5 moles of Si02 and 4 moles of h 2 o.

It was obvious from the calculations that little or no reaction occurs when Si02 is heated in the presence of H 2 0. There is no change in the number of moles of the silica sample. Further proof of the inert behavior of Si02 in this environment is given by the fact that nearly all of the gas present is in the form of water vapor. The only other gases present in meaningful quantities are H 2 and 0 2 , which result from the dissociation of H 20 . CHAPTER IV

EXPERIMENTAL PROCEDURE

4.1. Samples and Atmospheres.

4.1.1. Samples.

4.1.1.1. Silicon Carbide.

Four different types of SiC were utilized during this study. The preparation methods and properties of the materials were distinctly different. The first material was Norton Noralide NC203 SiC*. This material is hot-pressed and contains approximately 1.5% Al as a densification aid. A typical analysis of this material, as supplied by the manufacturer, is shown in Table 4.1.

The second type of SiC was also obtained from the Norton Company.

This material, Noralide NC430, is a sintered form of SiC that is densified after sintering by impregnation with liquid silicon. The densified samples contain approximately 1 0 weight percent free silicon. A chemical analysis of this material, as provided by the manufacturer, is given in Table 4.2.

The third type of SiC tested was also fabricated by a sintering process. This material, Sohio Hexoloy, differed from the other

'fc The addresses of sample and equipment suppliers are summarized in Appendix B of this document.

74 75

Table 4.1

Typical Chemical Analysis of Norton NC203 Hot-Pressed SiC

Al 1.48% Ti 0.01 Fe 0.24 V 0.02 B 0.005 Co 0.09 W 2.5 Ca 0.05 0 2 1.6 Mg 0.05 SiC 94-96 76

Table 4.2

Typical Chemical Analysis of Norton NC430 Densified SiC

SiC 88.5%

Si 10.6

Fe 0.4

Al 0.1

B < 5 0 ppm 77 sintered specimen in that densification was achieved by adding boron and carbon to the alpha-silicon carbide powder. An average chemical analysis of this material is given in Table 4.3.

The final form of SiC tested was a single crystal material.

These samples were taken from a larger crystal formation that was a by-product of the Atcheson process.

4.I.1.2. Silicon Nitride.

Four types of silicon nitride were also considered in this study.

The four types included three hot-pressed materials which contained different amounts and types of densification aids. The first material considered was Norton NC132 SigN^. This material is hot-pressed using a maximum of 1 weight percent MgO as a pressing aid. The final material is composed of beta-silicon nitride with minor amounts of silicon oxynitride (Si2

The second hot-pressed material, Norton NCX-34, is hot-pressed with the addition of 8 weight percent This material also contains trace amounts of oxygen present in the form of SiC^.

The final hot-pressed sample, GTE AY-6 , is prepared with the addition of 6 weight percent yttria and small amounts of A^Og.

In addition to these hot-pressed samples a chemically vapor deposited silicon nitride was also considered. This material was obtained from the Union Carbide Company and contains trace level (<0.5 w/o) impurities of C, 0, H, Al, As, B, and Fe. Table 4.3

Typical Chemical Analysis of Sohio Hexoloy Sintered a-SiC

Free C 0.5-3.0 w/o

B* 0.42

Al* 0.43

Either B or Al is added to the powder used to form the final piece along with C as a densification aid. 79

Table 4.4

Trace Element Analysis of Norton NC132 Hot-Pressed SigN^

Mg 0.4-0.6 w/o

Al 0 .2 -0 .3

Fe 0 .2 -0 .4

Ca 0.01-0.03

Mn 0.05

B 0.003

W 1 .5-2.0 80

4.1.1.3. Silicon.

A specimen of single crystal silicon was used as a reference material for the experiments conducted in this study. This material was obtained from the Crysteco Company and was 99.99% pure.

The material tested contained only ppm impurities of phosphorous, calcium, and .

4.1.1.4. Silica.

A sample of silica was also examined in the mass spectrometer experiments. This material was optical quality (99.9% pure) Vitreosil from the American Company.

4.1.2. Experimental Conditions.

4.I.2.1. Gas Mixtures.

The behavior of the samples previously described was analyzed in a variety of gas mixtures. The gases used included pre-mixed compositions of 2% CI2 in argon and 2% HCl in Ar. Various oxygen- chlorine gas mixtures were also used. These mixtures contained one percent chlorine and the oxygen content was varied from 1 to 20 percent. The mixing of the gases was accomplished by means of a digital metering system. The source tanks used in the mixing included

2% CI2 in Ar, pure O 2 , and pure Ar. The set flow rates of each gas type, to achieve a total flow rate of 400 cubic centimeter per minute

(hereafter ccm) into the furnace, are shown in Table 4.5.

In order to achieve proper mixing the accuracy of the digital flow meters was verified by monitoring the travel of a soap bubble in 81

Table 4.5 Set Flow Rates

Flow Rate (ccm) Gas Mixture 2% Cl2/Ar Ar Total ° 2

1 % Cl2 'l% O 2 2 0 0 4 196 400

1 % Cl2 *2 % O 2 2 0 0 8 192 400

1 % C ^ - 4 % O 2 2 0 0 16 184 400

1 % Cl2 -1 0 % O 2 2 0 0 40 160 400

1 % Cl2 '2 0 % O 2 2 0 0 80 1 2 0 400 82 a buret and the gas mixture was passed through a cylinder filled with glass . A schematic diagram of the mixing system is shown in figure 4.1.

4.1 .2.2. Temperatures.

The test temperature was varied from approximately 700 to 1025 degrees centigrade. The upper limit of the test temperature was 1050

°C due to the use of quartz as a furnace tube material.

4.II. Sample Preparation.

The samples used in both the mass spectrometer and thermogravimetric tests were flat plates with approximate dimensions of 13 x 6 x 1 mm. The plates were cut from manufacturer supplied billets using a low speed cut-off saw. Care was taken to insure that all plate samples for a given series of experiments were cut from the same billet. This eliminated the effect of lot variations.

After the plates were cut from the larger billet a small hole (2 mm) was drilled at a position near the top edge of the sample using a

Penwalt Airbrasive 6500 SiC-grit blaster. The resulting rough hole was then cleaned and enlarged using a high speed diamond drill.

Immediately before testing the samples were lightly ground using a 15 fim diamond wheel. This grinding operation served to eliminate any surface roughness resulting from the cutting/drilling operation and also to remove any unwanted surface oxide layer.

Following the grinding process the samples were ultrasonically 83

£ }

Gla s s 5% H 2 t B e a d s CZZ3K Furnac e

Digital Control

2Z Cl2 02 Ar

Figure 4. L - Schematic diagram of gas mixing system. 84 cleaned in a soap solution followed by followed by isopropyl alcohol. This sequential washing procedure resulted in a clean, film- free surface.

4.III. Mass Spectrometry.

4.III.1. Experimental Set-Up.

The theory and principles dictating the construction of the mass

Q spectrometer system have been described in detail elsewhere . The theory has also been summarized in section 2 .1 . of this document, therefore, only a general description of the equipment will be given here.

A schematic cross-section of the mass spectrometer sampling system located at NASA-Lewis is shown in figure 4.2. During an experiment the sample is positioned close to the sampling cone of the

system using a Pt hook on the end of an alumina tube. During runs on pure Si the platinum was isolated from the silicon by means of a small

diameter AI 2 O 3 tube. This isolation served to prevent the formation

of a low melting Si-Pt eutectic composition. The gas mixture entered

the cold end of the quartz furnace tube and traveled through the hot

zone of the furnace and past the sample.

Volatile species generated by the reaction of the hot-gas and the

sample entered the differential pumping system through a knife-edged

0.022 cm orifice in the Pt-Rh sampling cone. As the gas passes

through the cone into stage I of the pumping system free-jet expansion of the gas stream occurs. The pressure in stage I of the vacuum 85

/-QUADRUPOLE M ASS FILTER

1000 L /S

CHOPPER 2400 U S 1 0 0 0 0 1 /S '-S K IM M E R “ -SAMPLING ORIFICE SAMPLE- ‘““THERMOCOUPLE

FURNACE

QUARTZ TUBE ALUMINA ROD ALUMINA THERMOCOUPLE TUBE

O RINGS

CAS INLET— SLIDING JOINT STAINLESS CYLINDER

Figure 4.2 - Schematic diagram of high-pressure mass spectrometer sampling system. 86 system is maintained at approximately 1 0 "^ atm through the use of two

Leybold-Hereaus DK200 roughing pumps and two 10" liquid nitrogen trapped diffusion pumps.

The expanded gas stream then passes through a 0.076 cm orifice in a skimmer cone into stage II of the vacuum system and is formed into a supersonic molecular beam. The pressure in stage II is maintained at 1 0 "^ atm by means of a 6 " liquid nitrogen trapped diffusion pump. As the molecular beam travels through the second stage of the vacuum system it is modulated by a chopper operating at constant frequency. This modulation allows the separation of the source and background signals.

The molecular beam of gas subsequently travels through a 0.159 cm opening in a collimating cone into the third and final stage of the differential pumping system. The pressure in stage III of the system

- 8 is maintained at 10" atm using a 4" liquid nitrogen trapped diffusion pump.

The modulated molecular beam, which has been stepped down in pressure from the initial value of 1 atm to 1 0 '^ atm by the differential pumping system, then travels into a quadrupole mass filter. The mass spectrometer acts to separate the constituents of the gas stream by virtue of their differing mass-to-charge ratios.

The Extranuclear component mass spectrometer system was maintained at

- 8 a pressure of 10 atm by means of an ion pump. 87

4.III.2. Experimental Procedure.

4.111.2.1. System Calibration.

Before the corrosion experiments were initiated, the mass spectrometer sampling system was tuned to achieve the maximum possible signal intensity. The tuning process was accomplished by locking-in on the nitrogen signal from laboratory air and maximizing the ion signal by translating the sampling orifice using four thumb screws located on the axial positions of the sampling orifice flange. Once the nitrogen ion signal had been maximized care was taken not to alter the position of the sampling cone.

The sensitivity of the mass spectrometer sampling system was determined by analyzing a gas stream of industrial grade oxygen at room temperature. Oxygen gas cylinders contain impurities of at the part-per-million level. Therefore, detection of Xe during the sampling of the O 2 gas stream insured sensitivities on the order of 1 ppm.

Mass values generated by the system were calibrated using an oxygen gas mixture containing 1 0 0 0 ppm each of xenon, , argon, and krypton. The measured masses of the peaks were compared to the known masses and a plot of known value versus measured value was constructed. This plot served as a correction factor for the determination of actual mass-to-charge ratios.

4.111.2.2. Signal Intensity Calculations.

The relative intensity of the ion signal generated was calculated 88

from the peaks obtained on the mass spectrometer trace using the equation:

I = V/aRy (4.1)

Where: V ■= sum of voltage readings over all of the major isotopes a = ionization cross-section determined from additivity rules R - input resistance of instrument 7 - gain of instrument.

4.III.2. 3. Ionization Efficiency Curves.

Ionization efficiency curves, plots of ion signal intensity versus energy of the ionizing , were obtained for silicon tetrachloride at room temperature and also for silicon in 2 % C ^ - A r at

960 °C. In the case of SiCl^, Ar was bubbled through silicon tetrachloride liquid to produce an inert gas stream saturated with silicon tetrachloride. The 2% C^-Ar gas stream, on the other hand, entered the furnace at a rate of 400 ccm and was allowed to react with the hot silicon sample.

The ionization efficiency curves were obtained by slowly decreasing the energy of the ionizing electrons from the initial value of 30 eV while monitoring the intensity of the ion signal. The signal was monitored at all integral values of the ionizing electron energy.

4.III.2.4. Standard Test Method.

In the standard mass spectrometric test method the prepared samples were raised directly into the gas mixture of the desired 89

composition at the temperature of interest. The gas mixtures used are shown in Table 4.5

The test samples were kept in the cold end of the furnace until a stable gas signal (i.e.-Cl2+ ) was obtained on the mass spectrometer trace. When a stable Cl2+ ion intensity was obtained the sample was rapidly raised into the hot zone of the furnace. Maintaining the sample in the cold end of the furnace until testing began prevented the formation of an undesirable surface oxide film on the specimen.

This precaution was particularly important when the behavior of the materials in pure chlorine was studied.

After the sample had been raised into the hot zone of the furnace the peak intensity of SiCl^+ was monitored. After approximately ten minutes in the flowing gas atmosphere the silicon tetrachloride signal reached a stable level. Once past this transition , the spectrum was taken. In the early experiments for a given system, the entire mass range (m/e — 50 to 500) was scanned. These tests established the regions of the mass-to-charge ratio scale where ion signals were observed. Following these identification runs only the regions where peaks had previously been observed were monitored. This greatly reduced the scan time necessary to collect data on an entire spectrum. In cases where unexpected results occurred, that is unexpected absence of peaks, the entire mass region was scanned in order to ascertain the existence of any previously unseen species. The Cl2 + signal was also monitored at various times during the course of the experiment. The intensities of the chlorine ion signal provided an indication of the stability of the gas flow and also served as a normalizing factor for the product ion signals from experiment to experiment.

Following completion of the experiments the samples were lowered into the cold end of the furnace and flow of the corrosive gas stream was stopped. When the gas lines were clear of corrosive gases the furnace was lowered and the sample was rapidly removed from the furnace. The reacted sample was then immediately placed in a dessicator and subsequent exposure to air was minimized. The labeled samples were saved for later microstructural and surface scale analysis.

During the course of certain reactions the sampling orifice became clogged with corrosion products or SiC^ formed by reaction of

SiCl^ with oxygen. Although the time of the runs was kept short in order to avoid this inconvenience, in systems where an extreme reaction occurred it was unavoidable. When it was necessary to clean the orifice during the course of an experiment, the reaction was halted (that is the sample was lowered and the gas flow was stopped).

The furnace was then lowered and the orifice was cleaned with a 10% HF solution. When the sampling orifice was sufficiently cleared of any blockage the furnace was raised and the experimental run was resumed.

Care was taken to insure that data was representative during these 91

interrupted runs by comparing the chlorine and silicon tetrachloride signals obtained both before and after the sampling orifice cleaning.

4.III.2.5. Decay Experiments.

Experiments were also conducted using the mass spectrometer in order to monitor the effect of oxygen on the corrosion process.

Specifically the intensity of the silicon tetrachloride signal was monitored as a function of the time after oxygen was admitted into the gas stream.

In these tests the sample was raised directly into a stable flow of 2% CI2 in Ar at the temperature of interest (950 °C). The intensity of the silicon tetrachloride corrosion product signal was monitored and allowed to reach a steady-state value. Once the SiCl^+ signal had reached a stable value the appropriate amount of oxygen was admitted into the gas stream. Care was taken to insure that the transition to the oxygen containing gas stream occurred rapidly so that the total 400 ccm flow into the furnace was maintained.

After the oxygen was admitted into the furnace the intensity of

the silicon tetrachloride signal was monitored as a function of time

in the oxygen-containing atmosphere. This signal was normalized to

the chlorine ion signal that was monitored at various times throughout

the course of the experiment.

Following the prescribed exposure to the oxygen-containing gas,

the samples were quickly removed from the furnace and stored in a 92 dessicator.

4.III.2.6 . Pre-Oxidized Samples.

Tests were also conducted on samples that had been exposed to various oxidizing conditions before they were analyzed in the mass spectrometer. The samples tested in this manner included Norton

NC203 hot-pressed SiC, Norton NC430 siliconized SiC, Sohio Hexoloy sintered SiC, and pure Si. The testing of these pre-oxidized samples was performed in a 2 % C^-Ar gas stream in the manner described in section 4.III.2.4.

Three different sets of oxidizing conditions were used to grow oxide films on the samples. The first pre-oxidized samples were prepared by exposing the specimens to a 20% 0 2 ~Ar gas stream for 30 minutes at 950 °C. This oxidation was performed in the mass spectrometer furnace and allowed for transition to a chlorine containing gas without exposing the samples to air or allowing the samples to cool. In this test the transition to the 2% CI2 gas was made immediately after the prescribed oxidation period. The intensity of the silicon tetrachloride ion signal was monitored as a function of

the time in the chlorine atmosphere. This signal was normalized to

the chlorine signal that was monitored at various periods throughout

the course of the experiment.

The second set of pre-oxidized samples tested in the mass

spectrometer were prepared by exposing the sample to a 1 0 0 % O 2

atmosphere for 16 hours at 950 °C. This oxidation was performed in a 93

closed atmosphere Lindberg tube furnace. The gas stream was allowed

to flow past the samples at a rate of approximately 10 ccm. The

cleanliness of the oxidizing atmosphere was insured by using a new

alumina furnace tube and sample boat as well as by passing the pre- purified gas stream through a drierite filter column.

After the designated pre-oxidation period the samples were removed from the furnace and placed in a dessicator until testing.

The testing of these samples was identical to the first set of pre-

oxidized test coupons.

The third set of pre-oxidized test samples were prepared in the

same manner as the second set. These samples, however, were oxidized

in the flowing 100% O 2 gas stream for 21 hours at 1200 °C.

4.IV. Thermogravimetric Analysis.

4.IV.1. Experimental Configuration.

A schematic diagram of the thermogravimetric apparatus located at

NASA-Lewis is shown in figure 4.3. The apparatus consists of a gold- plated Cahn RH electrobalance coupled with a quartz tube furnace. The

gas mixtures used were prepared by electronically metering the

appropriate amounts of premixed 2% C^-Ar, pure O 2 , and/or pure Ar in

the amounts shown in Table 4.5. The sample was held in the hot-zone

of the vertically translatable tube furnace, in the vicinity of the

control thermocouple, by a Pt hang wire.

The analog signal from the balance was recorded on a 94

Balance

5Z H. ' Recorder

Digital Control Pt Wire Computer

Furnace Sample

ontrol TC

Quartz Tube 2Z Cl Furnace Control

Figure 4 .3 - Schematic diagram of thermogravimetric testing apparatus. 95 conventional strip chart recorder (Allen Instruments Model 2125M).

This signal was also digitized and recorded on magnetic tape by a

Hewlett-Packard Model 85 data acquisition computer coupled to a

Hewlett-Packard computational computer.

4.IV.2. Experimental Procedure.

The samples tested were prepared in the manner described in section 4.II. The samples were placed in the furnace tube in vicinity of the control thermocouple. This was done with the furnace lowered such that the samples were not exposed to oxidizing conditions. A stream of 5% ^ - A r was allowed to flow past the sample as the furnace was raised into position. This prevented any unwanted oxide from forming on the sample surface as the sample was being brought to temperature. After the sample was at the appropriate temperature the gas flow was quickly changed from the reducing hydrogen atmosphere to the gas composition of interest. This rapid transition minimized any surface oxide formation on the sample. A counter stream of argon flowed through the balance mechanism to prevent any corrosive gases from entering it.

The weight change of the sample was monitored as a function of time. The digitized signal was used to calculate a rate of weight change. This rate was calculated using data from the first two hours of the experiment. Two or three runs were performed for each set of conditions in order to insure reproducibility of results. CHAPTER 5

RESULTS AND DISCUSSION

5.1. Mass Spectrometry.

5.1.1. System Calibration.

A known gas mixture of argon (m/e=39.9), krypton (m/e=83.8), neon

(m/e=20.2), and xenon (m/e=131.3) in oxygen was used to calibrate the digital mass read-out of the mass spectrometer system. The spectrum obtained for this mixture was used to construct a plot of actual value versus meter value (figure 5.1). The resulting plot is nearly linear.

As a result of this approximate linearity, the curve was used to obtain constants for the equation:

y - y x - m (x - x x) (5.1)

where: m = slope of the actual vs. meter reading plot x-^ = meter value corresponding to known mass y-^ x = meter value corresponding to unknown mass y

From this equation the value of any unknown mass peak could easily be determined from the digital value recorded. The meter value (-70.7) of

70 the CI2 peak was used for the x^ value in equation 5.1.

5.1.2. Ionization Efficiency Curves.

The ionization efficiency curves (plots of signal intensity vs. energy of the ionizing electrons) determined for SiCl^ at room

96 iue . - airto lt o as pcrmtrotu reading. output spectrometer mass for plot - Calibration 5.1 Figure

ACTUAL VALUE (A M U ) o in o o o “ T o . 0 10 10 200. 150. 100. 50. O. EE VLE A U) (AM VALUE METER AIRTO PLOT CALIBRATION 97 98 temperature and Si in 2% C^-Ar at 960 °C are shown in figure 5.2.

These curves present data for all of the SiClx ions detected. There are distinct differences in the shapes of the curves observed for the various ions. These different curve shapes result from the fact that there are different types of ionization processes that may occur. In general, either simple or dissociative ionization may occur.

In simple ionization a parent molecule is ionized by an electron according to the equation:

MX + e' = MX+ + 2 e' (5.2)

The energy to accomplish this process is simply the ionization potential of the MX molecule.

For the case of dissociative ionization to form an MX+ ion the governing equations are:

MXy « MX + (y-1) X (5.3)

MX + e' = MX+ + 2 e' (5.2)

The energy necessary to produce an MX+ ion from an MXy molecule is equal to the sum of the dissociation energy of the MXy molecule, the

of the MX molecule plus some finite amount of excess energy. Normally the energy necessary to complete this process increases as the value of y increases.

For the case of simple ionization of a parent molecule, the

ionization efficiency curves are typically shaped like the curve illustrated in figure 5.3a. That is, no ion signal is observed until a certain threshold value is reached. Above this value a steadily iue . - oiainefcec uvs o; . il t room at SiCl^ a. for; curves efficiency Ionization - 5.2 Figure Intensity (Arbltrofy Units) 1*. 1*. SICI + 0 SICI ♦ SICI 0 SICI N IS. IS. oniai r y erg n E n izatio n S 20 2. 4 2. . . . 4 I. S 20 2 4 2. S 30 32. 3 0. 3 2S. 26. 24. . 22 0. 2 IS. IS. 14. 2. 3 O. S S. 2 26. 24. 22. 0. 2 IS. eprtr, n . i n2 Cl 2% in Si b. and temperature, B A (eV) (eV) 3 •/ w o e c SICI* 0 oniai r y erg n E n izatio n Io 2 nA a 90 °C. 960 at Ar in (eV)

99 + w ENERGY OF IONIZING ELECTRONS ELECTRONS IONIZING OF ENERGY ION INTENSITY iue , - oiainefcec uvs o n o negig a. undergoing; ion an for curves efficiency Ionization - 5,3 Figure (eV) ipe adb cmlx oiain fo eeec 91). reference (from ionization complex b. and simple, I + Observed Observed Observed LCRN VOLTS ELECTRON j f I • lj Fragment 'Parent 'Fragment Parent 100 101 increasing ion signal is observed with increasing ionization energy.

This value continues to increase until a maximum value is reached and then begins to slowly decrease.

The shape of the curve for an ion formed by complex or dissociative ionization is distinctly different from those formed by

simple ionization. Ions formed by these processes typically show

sharp upward breaks or long tails as the energy of the ionizing

electrons is increased (figure 5.3b) rather than the gradual increase

to a maximum value.

The ionization efficiency curves of the SiCl+ , SiCl2+ , and SiClj"*"

ions formed from SiCl^ at room temperature (figure 5.2a) are

characteristic of those typically found for dissociated ions. This plot shows a sharp upward break in the curve for SiClg* and long tails

for the SiCl+ and SiCl2 + curves. The curve for SiCl^+ , on the other hand, shows the shape characteristic of an ion formed by simple

ionization of a parent molecule. These results are expected since only SiCl^ is present as a parent molecule in this system, and

therefore only this ion should show simple ionization behavior.

The curves generated during the reaction of Si and 2% CI2 at 960

°C (figure 5.2b) are distinctly different than those reported for

SiCl^. In this case, both the SiCl^ and SiCl^+ curves closely

resemble that of a parent molecule shown in figure 5.3a. Only the

curves for SiCl+ and SiCl2+ show the long tails characteristic of ions

formed by a dissociative ionization process. This implies that both 102

SiClg+ and SiCl^+ exist as parent molecules in this case, and consequently that both SiClg and SiCl^ are formed during the reaction of Si and CI2 at 960 °C.

The intersection of the linear portion of the ionization efficiency curve with the x-axis of the plot yields the appearance potential of the ion. The values of the appearance potentials of the

ions also provide information pertaining to the origin of the ion.

The appearance potentials for the ions in both systems studied in this work are summarized in Table 5.1. This table also lists appearance

36 potentials for these ions as determined by Lin during the reaction

of Si and HC1 and by Vought^® for SiCl^ at room temperature. The

agreement between the data obtained in this research for the case

of SiCl^ and that obtained by Vought under similar conditions is very good.

In both experiments conducted in this research the appearance potential for SiCl^+ is less than that determined for SiClg+ , SiCl2 + ,

and SiCl+ . However, the magnitude of the difference between the values for SiCl^+ and SiCl3+ is lower in the case of Si in 2% Cl2 than

it is in the case of SiCl^ at room temperature. This smaller

deviation implies that a portion of the SiClj"*" observed during the

reaction of Si and CI2 is from simple ionization of parent molecules.

As previously discussed the appearance potential determined is

simply the energy necessary to produce that ion. Therefore, if the

ionization potentials of the ions are known they may be compared with Table 5.1

SiClx+A. Appearance Potential Data (eV).

Lin S t u d y ^ This Study Species Vought70 800 °C 1200 °C SiCl4 Si/2% Cl

SiCl+ 20.5 ± 0.3 16.0 1 2 . 0 19.1 20.7

SiCl2+ 18.4 ± 0.3 1 2 . 6 1 0 . 0 18.6 18.2

SiCl3+ 12.9 ± 0.2 12.5 1 2 . 2 12.9 13.1

SiCl4+ 1 1 . 6 ± 0 . 2 11.5 1 2 . 0 1 1 . 2 12.4 104 the appearance potentials determined for that ion to provide further as to the origin of the ion. Unfortunately the ionization potentials of SiCl^ and SiCl^ are not known. This fact makes it difficult to ascertain the origin of the ions from the appearance potential data alone.

The fragmentation pattern of the silicon tetrachloride may also be used in conjunction with the mass spectrum obtained for the actual corrosion experiments in order to determine the origin of the ions observed during the corrosion studies and also their relative amounts in terms of the amount of SiCl^. A typical spectrum obtained for

SiCl^ at 30 eV is shown in figure 5.4a. This spectrum is distinctly different than a spectrum obtained for Si reacted with CI2 at 960 °C

(figure 5.4b). Table 5.2 gives the isotopic intensities of all of the

SiClx ions observed in these systems.

The ratio of the SiCl+ , SiCl2+ , and SiCl-j+ ion intensities to the intensity of the SiCl^+ peak provides information on the origin of the ions. The ratio of the intensity of the SiCl^+ ion to the intensity of the SiCl3+ group in the silicon tetrachloride study determines the relative amount of SiCl

(0.47) is less than the ratio value obtained for the Si in 2% CI2 case

(0.62). This larger value provides evidence that a portion of the

SiCl-j+ observed in the reaction of Si with CI2 results from simple ionization of the parent molecule SiCl-j as well as from fragmentation of SiCl^. These ratios also indicate that the amount of SiCl^ 105

eo. 7 0 . eo. eo. 100. i 10. 1 2 0 . 130.140. iso. leo. 1 7 0 . 100,

60. 70. 00. 00. 100. 110. 120. ISO. 140. 150. 160. 170. 180, m / a

Figure 5.4 - Typical spectra for; a. SiCl4 at room temperature, and b. Si in 2% Cl2 -Ar at 960 °C. 106

Table 5.2

Relative Intensity Values for Major Isotopes for Spectra of SiCl4 at Room Temperature and Si in 2% Cl2 at 960 °C.

Species m/e (AMU) SiCl4 at 25 °C Si-2% Cl2

SiCl+ 63 55 100 65 21 40

SiCl2+ 98 25 18 100 18 13 102 4 3

SiCl3+ 133 100 71 135 98 79 137 34 29 139 5

SiCl4+ 168 36 34 170 47 46 172 23 25 174 6 6 107 generated during the reaction of Si with CI2 is more than double the amount of SiCl3 generated.

Applying the same type of analysis to the intensities of the

SiCl+ and SiCl2 + ions observed in the Si-C^ system indicates that these ions are formed by fragmentation processes. Therefore, they are assumed to be artifacts of the ionization process and are ignored as actual corrosion products.

The monoisotopic intensities, which are summations of the voltages of all the major isotopes for a given molecule, for the SiCl^ at room temperature using an ionizing voltage of 30 eV are in good agreement with the values reported in the literature (Table 2.2).

These values are 32.3, 19.7, 100, and 47.6 for the SiCl+ , SiCl2+ ,

SiCl3+ , and SiCl^+ ions respectively.

5.1.3. Standard Test Method.

5.1.3.1. Effect of Gas Composition at 950 °C.

A typical spectrum obtained for a sample of Norton NC203 SiC in a

2% C ^ - A r gas stream is shown in figure 5.5. These data are significantly different from the data shown in figure 5.4b for the Si sample reacted in 2% CI2 . Data of this type (in graphical form) is cumbersome and difficult to analyze. The discussion in section 5.1.2. indicates that SiCl^ is the major gaseous species formed during the reaction of Si with chlorine. Therefore, for the sake of ease of data reduction, the amount of volatile corrosion product formed by the reactions examined in this research will be represented Figure 5.5 - Typical spectrum for Norton NC203 hot-pressed SiC in 2% in SiC hot-pressed NC203 Norton for spectrum Typical - 5.5 Figure

Intensity (Arbitrary Units) o N co r> o o o o * o CM t

108 . 109 in terms of the amount SiCl^+ signal observed normalized to the intensity of the Cl2+ signal.

Results of this type for the four different types of silicon carbide studied are presented as a function of gas stream composition in Table 5.3. In general, increasing the amount of oxygen in the reactant gas stream decreases the amount of SiCl^+ observed. This decrease in the amount of corrosion product presumably results from the formation of a protective oxide scale on the surface of the sample. Although this general rule holds basically true, the amount of product observed for any one gas stream composition varies

significantly from sample to sample.

In a 2% C^-Ar gas stream the NC203 produces the most volatile

corrosion product followed by the NC430, Hexoloy, and the single­

crystal material. However, this sequence changes upon introduction of

1% C>2 into the gas stream. In the case of a 1% Cl2 , 1% 02 gas stream

the Hexoloy produces the most volatile corrosion product followed by

the NC203, NC430, and the single, crystal SiC samples. In fact, the

intensity of the SiCl^+ signal for the Hexoloy samples is slightly

greater in this gas stream than in the stream containing 2% CI2 . This

indicates an acceleration in the corrosion process upon introduction of oxygen to the system. As the oxygen content of the gas stream is

increased above 1%, the Hexoloy samples continue to show the largest amount of volatile product generated.

Once the oxygen content is increased above 4% to 10% no volatile 110

Table 5.3

Amount of Volatile Product Produced for Various SiC Materials as a Function of Gas Stream Composition at 950 °C.

ISiC14/XCl2

Gas Mixture Single Crystal Norton NC203 Hexoloy Norton NC430

2% Cl 10'3 X 2 1.1 X 9.4 x 10‘3 2.4 10‘3 4.8 X 10'3

1% ci2 , 1% o2 4.7 x 10‘4 2 . 0 x 10'3 2.5 X 10'3 9.7 X 10'4

1% ci2 i 2 % 02 5.0 x 10’4 7. 5 x io-4 1.8 X 10'3 7.2 X IO '4

1% Cl2, 4% 0 2 1.5 x IQ '4 1.3 x IQ '4 7.2 X IO ’4 1.4 X IO '4

1% Cl2 , 10% o2 - - 2.5 X 10’5 1.3 X 10'5

1% Cl2 , 2 0 % 0 2 - - 9.4 X 10‘6 1.5 X IO' 5 Ill products are observed in the mass spectrometer trace for the NC203 and single crystal SiC materials. Only the Hexoloy and NC430 results indicate that a slight corrosion process is still occurring. The magnitudes of the values indicate that this process is still more severe in the Hexoloy system. This data indicates that the oxide layers formed on the Hexoloy and NC430 samples in a 10% O 2 gas stream are not as protective as the layers formed on the NC203 and single crystal samples. In other words, oxygen contents greater than 20% are necessary in order to sufficiently reduce the corrosion process in the case of Hexoloy and NC430 to the point where no products are observable in the mass spectrometer.

The general effect of gas stream composition on the morphology of the sample surface is illustrated in figure 5.6. This figure shows scanning electron micrographs of the NC430 material both before and after reaction in a variety of gas streams. As the oxygen content of the gas stream is increased the extent of the surface attack decreases. In the high oxygen content gas streams (10 and 20%) the sample closely resembles the as-prepared sample indicating that the formation of a protective oxide layer has inhibited the corrosion reaction. It is important to note that this layer is thin, and does not significantly alter the overall morphology of the sample surface.

Interesting morphologies were also observed for the case of the

Sohio Hexoloy material (figure 5.7). This sample showed a great deal of attack in a 2% CI2 environment (b). The light grains in this 112

50 /jm

Figure 5.6 - Effect of gas stream composition on the surface morphology of Norton NC430 siliconized SiC as prepared (a) and after reaction at 950 °C in 2% Cl 2 (b), 1% Cl2, x% 02 where x- 2 (c), 4 (d), 10 (e), and 20 (f). Figure 5.7 - Effect of gas stream composition on the surface morphology of Sohio Hexoloy a-SiC as prepared (a) and after reaction at 950 °C in 2% Cl2 (b), 1% Cl2 , x% 02 where x- 1 (c), 2 (d), 4 (e), and 10 (f). 114 photomicrograph represent free carbon. The holes around the free carbon grains represent a consumption of the carbon grains by reaction with oxygen impurities in the gas stream. The micrograph for the sample reacted in 1 % C ^ , 1 % C>2 shows the existence of SiC>2 fume on the sample surface. This fume is presumably produced by the reactions; active oxidation of the sample:

SiC + 0 2 ** SiO(g) + CO (5.4)

followed by a reaction of the SiO gas with reactant oxygen to form

Si0 2 solid:

S1 0 (g) + ^ ° 2 (g) * S 1 0 2 (s) <5 -5>

The fact that active oxidation occurs implies that although the total

oxygen partial pressure is 0 . 0 1 atm, the partial pressure at the

sample surface must be significantly lower than this value. The free

carbon present in the samples will keep the oxygen partial pressure

low by consumption of the reactant oxygen to from C O ^ .

Figure 5.7 also indicates that increasing the oxygen content of

the gas stream reduces the extent of surface attack.

The data obtained for the silicon, silica, and silicon nitride

samples are summarized in Table 5.4. The information obtained for the

Si material is particularly interesting. It shows a very high rate of

reaction in the 2% CI 2 environment. In fact, clogging of the orifice

occurred fairly rapidly in this system as a result of the tremendous

amount of volatile product formed. Introduction of 1% C>2 into this

system greatly inhibited the corrosion reaction. In this environment 115

Table 5.4

Amount of Volatile Product Generated During the Reaction of Si, Si0 2 , and SijN^ as a Function of Gas Stream Composition at 950 °C.

„ . . ISiC14/IC12 Gas Composition Silicon Silica Silicon Nitride

2% Cl2 4.3 x 10 ‘2 1.1 x 10 ' 5 4.5 x IO "5

1% Cl2 , 1% 0 2 3.0 x 1 0 " 5 116 the intensity of the SiCl^+ signal decayed to an immeasurable value within five minutes after raising the sample into the reactant gas stream. This lack of reaction implies that a very protective oxide layer forms on the Si samples within 5 minutes even in the low oxygen-content gas streams.

The data obtained for the SiC>2 and Si-jN^ are very similar. In these reactions, SiCl^+ is only visible during the first few minutes after raising the sample into the gas stream. The volatile product is present in fairly small quantities even during this first few minutes and indicates that silica and silicon nitride are relatively unaffected by exposure to chlorine gas.

This lack of reaction is thermodynamically predicted for the SiC^ sample but is not expected in the case of Si^N^. The results on the behavior of silicon nitride compared to silicon carbide in 2 % CI2

(Table 5.3) insinuate that there is a more protective oxide layer or glassy phase present on the silicon nitride samples which prohibits

the reaction from occurring. This oxide or glass layer is apparently more impervious than the oxide layer formed on SiC. This explanation no is consistent with the recent findings of Schlichting which show a lower rate of oxidation for Si^N^ when compared to that reported for

SiC. He suggests that an oxide film always exists on both of these materials under normal conditions. The oxidation of these materials is controlled by transport of oxygen through this existing oxide film.

Therefore, slower rates of oxidation indicate less efficient transport 117 through the oxide film, or a film that is more resistant to penetration by gas molecules. The morphologies of the GTE AY 6 silicon nitride samples reacted in 2 % CI2 and 1 % CI2 , 1 % O2 are shown in figure 5.8. Both samples show a very smooth surface and bright spots, indicating the likely presence of a glassy silicate phase.

The behavior of the various Si-based materials was also examined

in a gas stream consisting of 2% HCl in Argon. No reaction was observed between the SiC, Si, or Si-jN^ samples in this gas stream.

This lack of reaction was also somewhat surprising as the

thermodynamic equilibrium calculations predict a fairly vigorous

reaction. This observation is, however, consistent with the lesser

amounts of SiCl^ produced in this system reported from thermodynamic

<}Q / /■» /• « analyses ’ and the lack of reaction observed by McNallan in

similar corrosion studies on SiC in HCl. Although the previously

reported work does not give estimated partial pressures of the product

gases, the amount of SiCl^ produced may be beneath the sensitivity

level of the mass spectrometer.

5.1 .3.2. Effect of Temperature.

The effect of reaction temperature on the amount of volatile product observed was studied for the Norton NC203 and Sohio Hexoloy

SiC samples. The data for these experiments are summarized in Tables

5.5 and 5.6 respectively. In general, increasing the furnace temperature increases the amount of reaction product observed.

Comparison of these two sets of data indicates that at all 118 a I

Figure 5.8 - Effect of gas composition on the surface morphology of GTE AY 6 SigN^ reacted at 950 °C for 30 minutes in; a. 2% CI2 and b. 1 %C1 2 , 1 % 0 2 . 119

Table 5.5

Effect of Temperature on the Amount of Volatile Product Generated by Reaction of Norton NC203 SiC with Various Gases.

ISiCl4/IC12

Temperature (°C) Gas Mixture 700 800 950 1025 ,-2 2 % ici2 2.4 x 10" 5 5.2 x 10 '4 9.4 x 10 ' 3 1.7 x

1% ci2 , 1% o2 - 1 . 6 x 1 0 " 4 2 . 0 x 1 0 ' 3 -

»-3 1% ci2 , 2 % 0 2 - 1 . 1 x 1 0 ‘4 7.5 x 10 ‘4 2 . 0 x

1% ci2 , 4% 0 2 - 3.3 x 10' 5 1.3 x 10 ’4 3.3 x

1% ci2 f 10% o2 - -. 3.8 x >-5 120

Table 5.5

Effect of Temperature on the Amount of Volatile Product Generated by Reaction of Sohio Hexoloy SiC with Various Gases.

ISiC14/IC12

Temperature (°C) Gas Mixture 700 800 950 1025 CSI 1 1— 1 o ro 2% 'Cl2 3.9 x 10'5 4.6 X IO’4 X 10'3 1.0 X

1% ci2 , 2 % 02 3.3 X 10‘4 1.8 X 10'3 1.1 X 10'2

1% ci2 , 4% 02 6 . 6 X 10'5 7.2 X 10'4 3.5 X 10'3

1% ci2 , 10% o2 2.0 X 10'5 2.5 X 10'5 8.5 X 10'5 121 temperatures tested more corrosion product was observed for the NC203 than Hexoloy in a 2% C^-Ar gas stream. However, as discussed in section 5.I.3.1. adding oxygen to the gas stream has a greater effect on the amount of reaction product produced in the case of the NC203 material. That is, the amount of SiCl^+ detected decreases more rapidly as oxygen is added to the system in the case of NC203 than in the case of Hexoloy.

This fact is also illustrated in figures 5.9 and 5.10 which show the log of the signal intensity versus the inverse of the reaction temperature for the NC203 and Hexoloy samples respectively. The data obtained for the NC203 show nearly linear behavior and large differences between the lines for the different gas compositions. The slopes of the lines seems to decrease slightly with increasing oxygen content of the gas stream. These decreasing slopes indicate a smaller temperature dependence for gas streams containing a larger amount of oxygen.

The effect of oxygen content on the Hexoloy samples is less dramatic. Figure 5.10 shows very little difference between the lines for the 2% CI 2 and 1% CI2 , 1% O 2 environments. In the case of Hexoloy the oxygen content of the gas stream is not significant until the oxygen content of the gas stream is increased to 4%. The data for Hexoloy shows a more dramatic reduction in the slope of the lines as the oxygen content of the gas stream is increased. This indicates a lesser temperature dependence with increasing oxygen 122

o r “

\% Cl 2% 0

O

O T-

O

O 7 . 0 8.0 9 . 0 10.0 1 1.01 1 0 4/ T (K_,>

Figure 5.9 - Effect of temperature on the amount of volatile product produced for Norton NC203 hot-pressed SiC. 123

I O

1% Cl 2% 0

\% Cl 4% 0 N I o 1 0 %

l(SiCl ) ? ------4_ O

♦ Io

*> Io r * 7.0 8.0 9.0 10.0 11.0 10 4/T (K-1)

Figure 5.10 - Effect of temperature on the amount of volatile product produced for Sohio Hexoloy sintered a-SiC. 124 gas content. In fact, at a level of 10% 0£ the effect of increasing temperature is minimal.

5.1.4. Decay Experiments.

The results of the decay experiments were measured using the normalized SiCl^+ intensity as a function of time in the oxygen containing atmosphere. Figure 5.11 shows the results obtained for the

Norton NC203 hot-pressed SiC. The general behavior of the material is the same in all of the gas compositions examined. In all cases a decrease to a steady-state value occurs fairly rapidly (within 15 minutes of admitting oxygen gas into the system). The steady-state value achieved decreases as the amount of oxygen in the reactant gas stream increases, corresponding to the formation of a more coherent and impermeable protective layer. The data represented by the in figure 5.11 represents a repeat of the experiment in the 1% CI2 ,

2% O 2 atmosphere. This repeat experiment indicates very good reproducibility in the data collected.

Figure 5.12 shows the data collected for the Hexoloy SiC material. Data are presented for the entire duration of the run (~2 hours) and also as an expanded view of the first 20 minutes of the reaction. The data collected for the 1% CI21 1% O 2 and 1% CI2 , 2% O2 cases show an initial increase in the signal intensity value

following admission of oxygen to the gas stream, rather than the

fairly rapid decrease to a stable value observed in the NC203 case.

This initial increase in the amount of volatile product observed 125

I O IX Cl

« IX Cl 10X 0 I o

Ksici) ? '(cit> 2

♦ io

A Io 0.0 25.0 50.0 75.0 100.0 Time (minutes)

Figure 5.11 - Decay of SiCl^ signal as a function of time in oxygen for Norton NC203 SiC. 126

or* — ft 6 , - i* 6, ix Cl 2X 0 ix a t - 2x ot IX Clt - 4X 0( 4X 0 — IX Cl - IOX 0 IX Cl IOX 0 o

o

o 0.0 30.0 90.060.0 120 0 0.0 S O 10.0 IS.O 20.0 Time (minutes) Time (minutes)

Figure 5.12 - Decay of SiCl^ signal as a function of time in oxygen for Sohio Hexoloy SiC. 127 represents an acceleration of the corrosion process upon introduction of this small amount of oxygen to the gas stream.

The data for these low oxygen content gases, along with the data generated for the 1% CI2 , 4% O 2 environment, also does not show the rapid approach to steady state values observed in the case of NC203.

The data collected show a slightly increasing tendency as the reaction time increases. Only the data collected for the 10% O 2 case exhibit the rapid decrease to a steady state value. The increasing tendency of the signal intensity indicates that the reaction rate is increasing as time in the oxygen containing atmosphere is increasing.

This implies that rather than growing a protective oxide on these samples, a stripping away of the surface material occurs to expose /:o fresh surface for reaction. As discussed by McNallan this process most likely occurs due to the presence of free carbon in these samples. This free carbon is consumed by the reactant oxygen to form CO or CO2 . This consumption results in the exposure of fresh

SiC surface for the chlorination reaction.

The data do however show that as the oxygen content of the gas stream increases, the steady-state value achieved decreases. This indicates that there is passivity due to formation of a more protective oxide layer as the amount of oxygen in the reactant gas stream increases.

Figure 5.13 shows the data collected for the Norton NC430 siliconized SiC. Again this data shows a decrease in the intensity of 128

o I X Cl r * -2X0

— \% Cl 4 % 0

O 2 0 X 0

O 0.0 20.0 4 0 . 0 6 0 . 0 Time (i linutes)

Figure 5.13 - Decay of SiCl^ signal as a function of time in oxygen for Norton NC430 SiC. 129 the SiCl^+ signal with increasing oxygen content of the gas stream.

The plot also indicates that in all cases except the 1% O 2 environment the signal decreases to a steady-state value within 10 minutes of oxygen introduction. The data for the 1% CI2 , 1% O 2 shows a gradual increase in the signal intensity during the first 1 0 minutes of reaction followed by a steady decrease after this transition time.

This indicates that an initial increase in the corrosion rate occurs followed by the decrease associated with passive oxidation of the sample surface. This seems to indicate that a certain transition time must be overcome in order for sufficient oxidation to occur.

The steady-state values achieved for this system are compared to the values obtained for NC203 and Hexoloy in Table 5.7. This table shows that the Hexoloy sample generally has the highest steady state value. Comparing this table to Table 5.3 shows that the steady state values achieved during the decay experiments are somewhat higher than those obtained by the standard test method.

This increase may be attributed to the pre-reaction that took place in 2% CI2 for the decay experiments. This reaction likely leads to a slight increase in the sample surface area and consequently a slightly higher steady state value.

5.1.5. Pre-Oxidized Samples.

The pre-oxidized samples were reacted in 2% C^-Ar at 950 °C using the standard test method. The data collected from these experiments was used to construct plots of normalized intensity as a Table 5.7

Steady State Values Achieved During Decay Experiments

ISiC14/IC12

Gas Composition Norton NC203 Sohio Hexoloy Norton NC430

1 % Cl_21 O 2 2.92 x 1 0 ' 3 6.62 x 1 0 ' 3 2.91 x 1 0 ' 3

2 % 0 2 1 . 1 0 x 1 0 ' 3 2.47 x 1 0 ‘3 2.59 x 1 0 ' 3

4% 0 2 1.30 x IO'4 1.37 x 1 0 ‘3 7.76 x IO '4

10% o2 1 . 2 2 x 1 0 ' 4 1.93 x 1 0 ' 5 9.31 x 1 0 ‘ 5 131 function of time in the corrosive environment. The data obtained for the Norton NC203 SiC samples are presented in figure 5.14. These data show an increase in the amount of volatile corrosion product generated with increasing time in the corrosive environment followed by a leveling off to a steady-state value. The steady state values achieved for the different oxidizing conditions are summarized in

Table 5.8. These values decrease as the time of oxidation, oxygen content of the gas stream, and oxidizing temperature are increased.

The sample that was oxidized for 21 hours at 1200 °C has the lowest steady state value, however this value is not significantly different from the value obtained for the sample that was oxidized for 16 hours at 950 °C.

An interesting feature of figure 5.14 is the more rapid rise to this steady state value in the case of the samples that had been oxidized for the longer periods of time in the more oxidizing conditions. In the 40 minutes that the samples were reacted in the 2%

CI2 environment, the SiCl^+ intensity for the sample oxidized for 30 minutes in the 2 0 % 0 £ gas stream does not appear to have reached a

stable value . The other samples, on the other hand have reached a

steady value after approximately 10 minutes in the corrosive

environment.

This difference may be explained by analyzing the differences in

oxidizing heat treatment for the various samples. The samples

oxidized for 30 minutes in the mass spectrometer furnace were not 132

O r » 20 % 0. 30 min.. 950 C

950 C

1200 C

O

l(S iC la) ■

O

O 0.0 10.0 20.0 30.0 40.0 Time (minutes)

Figure 5.14 - Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Norton NC203 hot-pressed SiC. Table 5.8

Steady State Values Achieved After Exposure of Pre- Oxidized Samples to a 2% Chlorine Environment.

ISiCl4/^Cl2 10'*>

Oxidizing Cond. NC203 Hexoloy NC430 Si

20% 02, 950 °C 16.1 64.4 101.1 1.49 30 Minutes

100% 02, 950 °C 2.51 51.6 2.37 16 Hours

100% 02, 950 °C 1.96 10.3 1.75 21 Hours 134 allowed to cool before testing occurred. The other samples were cooled before testing as a result of the transportation of these samples from the clean oxidizing furnace to the mass spectrometer system furnace. This cooling is likely to produce cracks in the oxide layer formed on the samples as a result of mismatch or flaws in the oxide. These cracks and pores provide channels of easy access to the SiC sample, and therefore, a more rapid rise to the steady state value. This explanation may also explain why the differences between the samples formed by oxidation in the clean furnace are so small in spite of the major differences in oxidizing heat treatment. That is, the level achieved may be a result of cracks formed in the surface oxide and not due to differences in the actual

thickness and morphology of the surface oxide film.

Figure 5.15 shows scanning electron micrographs of the pre- oxidized sample before exposure to chlorine. These photomicrographs provide evidence of the non-uniformity of the surface oxide layer formed.

The data obtained for the Sohio Hexoloy sintered a-SiC sample is presented in figure 5.16. As in the case of the NC203, a rapid rise to a steady state value is observed in this material. This rise is also more rapid for the cooled samples than for the sample that had not been allowed to cool. The steady-state values are very much different than those observed for the NC203 sample (Table 5.8). The values ultimately achieved for the 30 minute, 20% 02 and the 16 hour, 135

Figure 5.15 - SEM photomicrographs of pre-oxidized Norton NC203 after cooling. 136

O 20 X 0 . 30 mirt., 950 C

« o r*

O r*

o 0.0 10.0 20.0 3 0 .0 4 0 .0 Time (minutes)

Figure 5.16 - Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Sohio Hexoloy SiC. 137

100% O 2 oxidations are nearly equal. This indicates that there is similar behavior between the oxides grown in these environments when exposed to chlorine. This similarity may be due to similar morphologies or simply a coincidence in that the cooled samples have a cracked surface oxide layer which allows transport of chlorine through the oxide layer in a manner similar to that occurring in the uncooled specimens.

Figure 5.17 shows that fairly large cracks and porous regions exist in the oxide grown for 16 hours in pure oxygen at 950 °C. This figure also shows that after exposure to chlorine, the oxide layer has cracked to an even greater extent. In fact, a severe blistering of the oxide occurs in this situation indicating an attack of the interfacial area by chlorine.

Figure 5.16 shows that a decrease in the steady-state value achieved is not observed until the oxide has been allowed to grow for

21 hours at 1200 °G. This large difference in the oxides grown in the

100% O 2 environment may indicate a substantial increase in the kinetics of the oxidation reaction upon increasing the temperature from 950 to 1200 °C.

The oxide grown for a longer period of time at this higher temperature is not as severely cracked as (figure 5.18) upon cooling as that for the less severe oxidizing condition. Even after chlorine exposure, this oxide is not as severely blistered as that shown in figure 5.17. This indicates a lower degree of interfacial attack or a 138

Figure 5.17 - SEM photomicrographs of Sohio Hexoloy ct-SiC oxidized for 16 hours in pure oxygen at 950 °C before (a) and after (b) exposure to chlorine. 139

V /

o V k - u 0 $ ' k *

• «?> 1

Figure 5.18 - SEM photomicrographs of Sohio Hexoloy a-SiC oxidized for 21 hours in pure oxygen at 1200 °C before (a) and after (b) exposure to chlorine. 140 more coherent oxide film.

Figure 5.19 shows the data obtained for the pre-oxidized Norton

NC430 siliconized SiC samples. Again this data shows a slightly more rapid rise to a level value for the samples that had been allowed to cool than for the sample that was exposed to a rapid transition from the oxidizing environment to the corrosive environment. However, the difference between the steady-state values achieved in this system is much larger than that observed in the NC203 or Hexoloy systems. This indicates a more substantial difference in the oxide grown on the

NC430 with differing oxidizing conditions.

The steady-state value achieved for the 30 minute oxidation is very high, in fact higher than the values obtained for the NC203 or

Hexoloy samples (Table 5.8). This result indicates that the oxide scale formed under these conditions on Norton NC430 is somewhat less protective than that formed on NC203 or Hexoloy.

The steady state values for the oxides grown for longer periods of time are lower than those observed for the NC203 or Hexoloy. This implies that under these conditions the oxide surface layer formed on the NC430 is more protective than that formed on the other samples.

This large difference may be due to the large amount of free Si (-10%) that is present in these samples. Silicon oxidizes much more rapidly than SiC, even in low-oxygen content gas streams, and may result in the formation of more protective surface layers on these samples.

Electron micrographs (figure 5.20) of the pre-oxidized Norton 141

I O r-

« I o r*

o r-

o 0.0 10.0 20.0 30.0 40.0 Time (minutes)

Figure 5.19 - Growth of SiCl^ signal as a function of time in chlorine for pre-oxidized Norton NC430 siliconized SiC. Figure 5.20 143

NC430 siliconized SiC indicate the oxide layer grown on these samples under the severe oxidizing conditions is more continuous than that grown on NC203 or Hexoloy.

Data was also collected for a sample of pure Si that was oxidized in the mass spectrometer tube furnace for 30 minutes at 950 °C in a

20% O 2 gas stream. The results for this system are compared to those generated for the NC203, Hexoloy, and NC430 samples in figure 5.21 and

Table 5.8. These results show that the stable value achieved for

silicon is much lower than the stable values achieved for the SiC

samples. This indicates that a much more protective and impervious

oxide layer is formed on Si under these conditions than on SiC. This behavior is consistent with the tremendous decrease observed in the

amount of volatile corrosion product generated after oxygen is

introduced into the Si-C^ system (see Table 5.4).

Attempts were made to measure the thickness of the oxide scale by

electron microscopy of the sample cross-sections. However, due to the

nature of the samples, these attempts were unsuccessful. Fortunately,

the behavior of the different SiC materials in pure, dry oxygen in the

temperature range of 1200 to 1500 °C has recently been reported by

93 Tressler and Costello . They investigated several materials

including the Norton NC203 hot-pressed and the Sohio Hexoloy sintered

SiC samples examined in this research. They also analyzed single

crystal SiC formed as a by-product of the Acheson process, single

crystal silicon, and a controlled nucleation thermal deposition (CNTD) 144

I O f- Norton NC203 H ot-Pressed SiC Sohio Hexolloy Sintered SiC Norton NC430 Siliconized SiC Pure Silicon

O

i(s«cia) »(Clt)

O

o

0.0 10.0 20.0 30.0 40.0 Time (minutes)

Figure 5.21 Comparison of data obtained for samples pre-oxidized in 20% O 2 at 950 °C for 30 minutes. 145

SiC. The CNTD material was found to contain approximately 10% free silicon and therefore is similar to the Norton NC430 siliconized SiC analyzed here.

They measured oxide thicknesses using ellipsometry and profilometry and converted the data to parabolic rate constants using

94 the Deal-Grove model. The rate constants determined for these materials at 1200 °C and extrapolated at 950 °C are shown in Table

5.9. This table clearly shows that the highest oxidation rate for the

SiC materials was observed for the sample containing free silicon. As discussed previously, the free silicon increases the oxidation susceptibility of the material. Of the two remaining polycrystalline

SiC samples, the hot-pressed material has the higher parabolic rate constant. This is consistent with the argument presented here and the lower steady-state value reported for NC203 (Table 5.8). This difference in the oxidation characteristics has been explained by the higher additive levels present in the hot-pressed samples. These higher additive levels effectively decrease the viscosity of the oxide film, increasing the diffusivity of the oxidizing species across it, thereby increasing the oxidation rate^.

Applying these rate constants to the oxidizing conditions employed here gives oxide thicknesses shown in Table 5.10. The different thickness values calculated correspond to the different steady-state values achieved in the corrosion experiments (Table 5.8). Table 5.9

Parabolic Oxidation Rate Constants for Various Types of Silicon Carbide (from reference 93). o Rate Constant (nm /min)

Material 950 °C (extraplotated) 1200 °C

Single Xtal Si 100.48 712 ±25.4

Single Xtal SiC 46.99 346 ± 47

SiC with free Si 29.37 344 ±53.4

Norton NC203 13.07 220 ± 82

Sohio Hexoloy 8.94 175 ± 87 147

Table 5.10

Calculated Oxide Scale Thicknes as a Function of Oxidizing Condition.

Thickness (ran)

Material 950 °C, 0.5 h 950 °C, 16 h 1200 °c, 21 h

Single Xtal Si 55 311 947

Single Xtal SiC 38 212 660

SiC with free Si 30 168 658

Norton NC203 20 112 526

Sohio Hexoloy 16 93 470 148

5.1.6. Silicon Oxychlorides.

5.I.6.1. Initial Identification.

An entire spectrum (m/e = 50 to 350) for the Norton NC203 SiC -

1% CI2 , 1% O 2 system at 950 °C is shown in figure 5.22a. An actual spectrum from this system is shown in Appendix C. This spectrum shows groups of peaks with very low intensity values at m/e =

250, 280, and 310. An enlargement of the intensity scale in this region is presented in figure 5.22b. This enlargement shows that small groups of peaks do exist in these regions and also that these groups exhibit the isotopic abundances typical of chlorine-containing compounds. Table 5.11 lists the mass-to-charge ratios of the peaks observed in this instance, their intensity values normalized to the

1 4- largest group in the entire spectrum ( SiClg ), and also the

intensity values normalized to the largest peak in each group.

The existence of high mass molecules in systems containing

silicon, oxygen, and chlorine at high temperatures has been postulated

55 58 59 by a number of researchers^ . These high molecular weight

species were believed to be silicon oxychloride compounds, containing various ratios of silicon, chlorine, and oxygen. The exact

stoichiometry of a compound is difficult to determine from its molecular weight alone, even when the elements composing the compound are well known. In order to determine the exact stoichiometry of a compound from its molecular weight determined by mass spectrometry, it

is often necessary to look at the isotopic configuration of that 149

2 SO . 3 0 0 . 3 5 0 .

rt

c D s •* I. <

I) c «

o

120, 150, 180, 210, 240. 270, m / e

Figure 5. 22 - Mass spectrum of Norton NC203 SiC reacted in 1% Cl2 , 1% C>2 at 950 °C showing the existence of high mass molecules. 150

Table 5.11

Intensities of High Mass Molecules Shown in Figure 5.22.

m/e Species m/e

Si20Cl5 247 0.86 58 249 1.49 100 251 0.96 64 253 0.38 26

Si2OCl6 282 0.18 56 284 0.32 100 286 0.26 81 288 0.14 44

si3oci6 310 0.18 86 312 0.21 100 314 0.13 62 151 molecule. This was accomplished in this study by using a computer program^ which allowed the determination of the molecular isotopic distribution of any given species based on stored elemental distributions. This program provided a means of matching the isotopic distributions determined experimentally with those predicted from the molecular formula. Therefore, in order to determine the isotopic distribution of a molecule, its stoichiometry must first be speculated.

The previous work on these species®^’^ suggests that a of compounds with the general formula SinOn_-^Cl2 n + 2 exists. In light of this fact, the first member of this series,

Si2 0 Clg, served as the starting point for the initial determination of isotopic abundances. The isotopic abundances for this molecule and the first fragment of this molecule are presented in Table 5.12.

Comparison of the abundances and mass-to-charge ratios presented in Tables 5.11 and 5.12 indicates that the group of peaks at « 250 results from Si2 0 Cl3+ and the group at » 280 corresponds directly to

Si20Clg+ . Table 5.11 and figure 5.22b indicate that the highest mass peak observed is 28 mass units above the Si20Clg group. Elemental silicon has a molecular weight of 28 which implies that this molecule is Si20Clg containing an extra Si, or SijOClg. The existence of such a molecule has not been previously discussed in the literature and is

somewhat surprising. Therefore, many combinations of silicon, oxygen, and chlorine were considered in the computer program in order to find 152

Table 5.12

Calculated Isotopic Abundances

Species m/e VI*

S l^OClc^ 247 59 249 100 251 69 253 25

Si20Clg 282 50 284 100 286 85 288 39

Si3OCl6 310 49 312 100 314 87 153 any other combinations of silicon, oxygen, and chlorine falling at this mass. However, the isotopic distribution proves that this compound is most likely Si3 0 Clg. It is curious to note that the observed isotopic distribution does not correspond to that calculated as closely in this case.

Increasing the sensitivity of the system to its maximum possible value resulted in the observance of groups of peaks in the neighborhood of 345, 215, 180, and 155 atomic mass units. These groups correspond to the molecules Si3 0 Cl^+ , Si2 0 Cl^+ , Si2 0 Cl>3+ , and

SiOCl3+ respectively.

These fragmentation patterns may be used to deduce the molecular structure of these high mass compounds. From the existing literature and the fragmentation of the Si2 0 Clg compound it is relatively easy to determine a structure for this compound. It appears that this molecule consists of a silicon-oxygen-silicon backbone with chlorine atoms attached to the silicons. This suggests a molecular structure o f :

Cl Cl I I Cl — Si— 0— Si — Cl I I Cl Cl

The peaks observed corresponding to Si3 0 Clg+ and Si3 0 Cly+ are somewhat more difficult to assign a molecular structure to. These compounds seem to be fragments of the parent molecule Si3

Cl Cl Cl I I I Cl — Si — Si — 0— Si — Cl I I I Cl Cl Cl

It is not terribly surprising that the parent molecule SigOClg is

not observed in the mass spectrometer trace. The earlier work on

SiCl^ has shown that it is a relatively easy process to remove a Cl

atom from the molecule to form SiClg. Therefore, it should be equally

or more likely to fragment a Cl atom from SijOClg. The amount of

fragmentation occurring may be sufficient to produce intensities of

the parent molecule that are below the limits of sensitivity of the

mass spectrometer system.

5.I,6.2. Mechanism of Formation.

In order to determine the mechanism of formation of the silicon

oxychloride compounds, an experiment was conducted at room temperature

in which a gas stream consisting of 50% O2 in Ar was flowed over SiCl^

liquid. A wide variety of high mass molecules were observed in this

experiment. The most intense molecules observed and their isotopic

abundances are illustrated in Table 5.13. Comparing these results 155

Table 5.13

Comparison of Observed and Calculated Isotopic Abundances

Species m/e Observed Calculated

Si3OCl7 345 35 41 347 89 96 349 100 100 351 67 59 353 30 21

si3oci6 312 100 100 314 55 87 316 20 42

Si2OCl6 282 48 50 284 100 100 286 82 85 288 36 39 290 10 11

Si2OCl5 247 58 59 249 100 100 251 74 69 253 26 25

SiOCl3 151 100 100 153 55 35 155 20 4 156 with those discussed previously suggests that the silicon oxychloride compounds observed are a result of a gas phase reaction between the silicon tetrachloride that is produced by the reaction of the chlorine gas with the silicon-containing compound and the reactant oxygen gas.

A likely chemical reaction to form these molecules may be written a s :

4 SiCl4 + 0 2 ** 2 Si2OCl6 + 2 Cl2 (5.6)

The reaction may also proceed by the formation of an intermediate compound or compounds. No likely intermediate compounds were observed during the mass spectrometric analysis, and therefore speculation of a possible intermediate phase is not beneficial. It also should be noted that although the evidence of a gas phase reaction to form the silicon oxychlorides exists, it does not totally preclude the formation of these compound by surface reaction of the hot Cl2/02 gas stream with the hot Si-containing material.

A group of peaks was also observed at m/e — 150 for the room

temperature reaction. This group of peaks most likely corresponds to

the trichlorosilanol compound, SiClgOH, which is formed by reaction of

SiCl^ and water impurities by reaction (2.25).

The evidence of the gas phase reaction to produce the silicon

oxychloride compounds also raises questions as to the origin of the

Si^OClg compound. Both the Si^Od^"*" and SigOCly+ species were

observed in the experiment where an oxygen gas stream was reacted with

silicon tetrachloride vapor. Again, their isotopic abundances do not 157 closely agree with the predicted values. A chemical reaction between the SiCl^ and O 2 to form SigOClg is difficult to imagine. Therefore, it seems possible that this compound may be an artifact of combination of Si20Clij+ and SiClg in the ionization chamber of the mass spectrometer.

5.II. Thermogravimetric Analysis.

5.11.1. Effect of Gas Composition at 950 °C.

5.11.1.1. SiC Materials.

In much the same manner as the mass spectrometer tests, the effect of gas stream composition on the rate of weight change of the samples was examined. For the sake of conserving time, the gas streams examined were limited to 2% CI2 in Ar and 1% CI2 , x% O 2 in Ar where x = 2, 4, or 10. Table 5.14 lists the results obtained for the

SiC samples as a function of gas stream composition at 950 °C. Figure

5.23 is a graphical representation of the sample weight versus time in the reactive gas stream for the Norton NC203 SiC.

Figure 5.23 clearly shows that the addition of oxygen to the

NC203-C12 system severely decreases the rate of weight loss observed.

This plot indicates that when the oxygen content of the gas stream has been raised to 4% the rate of weight loss is minimal, indicating that the corrosion reaction has ceased.

The shape of the curve for the 2% CI2 environment is particularly intriguing. Instead of a constant linear behavior, the slope of the line (rate of weight loss) seems to decrease as the time of chlorine 158

Table 5.14

Effect of Gas Composition on Rate of Weight Loss at 950 °C for Various Types of Silicon Carbide

k {mg/cm -hr} Gas Stream Norton NC203 Sohio Hexoloy Norton NC430

2% Clo 6.927 ± 0.496 3.591 ± 0.8874 11.07 ± 0.50

1% Cl2, 2% 02 1.316 ± 0.045 3.461 ± 0.344 0.5275 ± 0.1674

1% Cl2, 4% 02 0.067 ± 0.019 0.156 ± 0.0407 0.0380 ± 0.0146

1% Cl2, 10% 02 0.028 ± 0.013 0.024 ± 0.0129 iue .3 Sml egtvru tm frNro C0 hot-pressed NC203 Norton for time versus weight Sample - 5.23 Figure SPECIFIC WEIGHT(MC/CM*2) -60 -30 -70 -50 -40 -20 -10 - - - - - X120/r 101A 950C 0120710A 0126710B 1XC1/2X02/Ar l%Cl/4X02/Ar - — i a a ucino a sra cmoiina 90 °C. 950 at composition stream gas of function a as SiC lXCl/10Z02/Ar 012971 012971 OB lXCl/10Z02/Ar X1Agn 141B NC203 011471OB 2XC1/Argon TIME(HOURS)

o 159 160 exposure increases. This of the surface may be a result of oxidation of the surface by O 2 impurities in the gas stream

(approximately « 0.1%) or more likely by the formation of a Si- depleted (i.e.- C rich) region near the sample surface. This carbon layer formed may actually be "pseudo-protective". The affinity of CI2 for C is significantly less than that for Si which would lead to a lower rate of surface consumption for a surface that is more rich in C

than stoichiometric SiC. Visual and microscopic analysis of the

samples showed a dark, powdery substance on the sample surface.

Scanning Auger Microprobe (SAM) analysis of the surface confirmed the

existence of carbon.

The data obtained for the Hexoloy samples (Table 5.14) shows a much less dramatic effect on the rate of weight loss when small

amounts of oxygen are added to the gas stream. In fact, within the

error of the experiment, there is no significant difference between

the rate of weight loss observed in the 2% CI2 environment and that

observed in an environment containing 1% CI2 and 2% C^. Only when

the oxygen content of the gas stream is increased to 4% or greater is

there a significant decrease in the rate of weight loss.

This data is consistent with that obtained by the mass

spectrometer. The results of both these analyses indicate that

Hexoloy is more resistant to oxidation in low oxygen content gas

streams. This resistance to oxidation is most likely caused by the

fact that the free carbon present in this material reacts with the 161 oxygen in the gas stream. This reaction maintains a lower partial pressure of oxygen near the sample surface which prevents the oxide from forming. This lack of oxidation allows the chlorination reaction to continue even in the presence of small quantities of oxygen. The data also indicates that greater amounts of oxygen are necessary to form an oxide film on the sample surface which significantly reduces

63 the extent of the chlorination reaction. McNallan obtained similar results for his low-cost samples containing excess carbon.

Table 5.14 also lists data obtained for the Norton NC430 siliconized SiC samples. These samples exhibit a severe rate of weight loss in the 2% CI2 environment at 950 °C. This rate is nearly double the rate observed for the NC203 SiC and nearly 4 times the rate

seen for the Hexoloy under similar conditions. This high rate of weight loss may be attributed to the large amount of free Si present

in the NC430. The mass spectrometer results (Table 5.4) indicate that

Si is severely attacked by CI2 . Therefore, excess Si in the sample will be preferentially attacked leading to an accelerated rate of weight loss compared to that observed for stoichiometric SiC.

The addition of 2% O 2 to the reaction environment severely

retards the rate of the corrosion reaction in the case of NC430. The

rate of weight loss observed for this material is significantly lower

than that observed for NC203 or Hexoloy under these conditions. This

result may also be attributed to the excess of silicon which is present in the NC430. The mass spectrometer experiments show that 162 introduction of oxygen to the reactant gas stream results in a stoppage of the chlorination reaction of Si. Therefore, the excess

Si present in this sample is more rapidly oxidized than the SiC and reduces the sample surface area which is available for reaction with the chlorine gas.

5.II.1.2. Si and Si3N4 .

Pure silicon and GTE AY6 Si3N4 were also examined by the thermogravimetric technique at 950 °C. The results of these analyses are presented in Table 5.15. The information obtained for the Si sample is in good agreement with the data collected in the mass spectrometer experiments. These results indicate that a very severe corrosion reaction occurs in the case of Si in 2% CI2 . The rate of weight change determined is nearly an order of magnitude greater than that observed for Hexoloy. However, as observed in the mass spectrometer work, addition of oxygen to the silicon-chlorine reaction has a large and immediate effect on the rate of reaction. The rate of weight loss in the environment consisting of 1% CI2 and 2% O 2 is minimal. This indicates that a very protective oxide layer is rapidly

formed on the Si even when the oxygen content of the gas stream is

fairly low.

The data obtained for the GTE AY6 Si3N4 is also consistent with

the mass spectrometer results. A very small rate of weight loss was

observed for this material in an environment containing 2% CI2 . This

lack of reaction is supported by the inability to observe SiCl4+ in Table 5.15

Effect of Gas Stream Composition on Rate of Weight Loss for Silicon and Silicon Nitride at 950 °C. o k {mg/cm -hr) Gas Stream Silicon GTE AY6 Si3N4

2% Cl2 25.91 ± 3.37 0.03911

1% Cl2 , 2% 02 0.01855 164 the mass spectrometer, and is most likely a result of the existence of a very impervious surface oxide film on these samples.

5.11.2. Effect of Temperature.

The effect of temperature on the rate of weight loss was also briefly examined for the NC203 and Hexoloy samples in a 2% C^-Ar environment. In addition to 950 °C, the behavior of these samples was studied at 800 and 1100 °C. The results of these test are presented

in'Table 5.16 and in figure 5.24 where they are plotted as log rate versus inverse temperature.

As intuitively expected, increasing the reaction temperature results in ax> increase in the rate of weight loss observed. For all of the temperatures tested the Hexoloy sample shows a rate of weight loss lower than that observed for the NC203. At 950 and 1100 °C the rate of weight change observed for the NC203 is approximately twice that of the Hexoloy. While neither set of data appears to fit a straight line, the slopes of the "best-fit" lines appear similar. This indicates that the effect of temperature on the rate of weight loss in a 2% CI2 environment is similar for both materials. There are not enough data points available to determine the of this process from the lines presented.

5.11.3. Effect of Gas Flow Rate.

In order to determine the rate controlling mechanism in the

corrosion process a TGA test was also performed using a gas stream Table 5.16

Effect of Temperature on Rate of Weight Loss in a 2% Cl£ - Ar Gas Stream.

O k {mg/cm -hr) Temperature (°C) Norton NC203 Sohio Hexoloy

800 0.6710 ± 0.2766 0.511 ± 0.089

950 6.927 ± 0.496 3.591 ± 0.887

1100 12.62 ± 0.610 5.327 ± 0.089 166 < No r * Norton NC203

- Sohio Hexolloy

O T“

O 6.0 7.0 8.0 9.0 10.0 104/T (K"1)

Figure 5.24 - Effect of temperature on the rate of weight loss for Norton NC203 and Sohio Hexoloy SiC materials in 2% C^-Ar. 167 with a reduced flow rate. For this experiment a sample of Sohio

Hexoloy was reacted in a 100 ccm flow of 2% CI2 at 950 °C. The results of this analysis showed a rate of weight loss of 0.2640 ±

0.0753 mg/cm^hr as compared to the 3.5950 ± 0.8874 rate determined in a 400 ccm flow. This decrease in the rate of weight loss indicates that the reaction is controlled by transport of reactants or products through the gas phase and not by surface reaction. CHAPTER VI

CORROSION MODELS

6.1. Physical Models.

The large variations that were observed in the behavior of the different SiC samples were somewhat unexpected. These differences seem to be a manifestation of the different processing routes for these materials. In particular, the processing additives used to aid in the densification of these materials are major players in the mixed oxidation/chlorination reaction. In order to further understand the role of these minor constituents it is useful to develop certain physical models which represent the corrosion mechanism on a very simplistic scale. Such models are discussed in the first section of this chapter.

6.1.1. "Pure" SiC.

The results presented in chapter 5 indicate that the behavior of the single crystal material and the Norton NC203 hot-pressed sample are very similar. As a result of this similarity and the availability of the NC203 material, it was studied in detail as a form of SiC closely resembling a pure, stoichiometric material.

In the mixed oxidation/chlorination of this material under the conditions of these experiments two distinct chemical reactions are

168 169 likely to occur:

SiC + x/2 Cl2 * siclx + c C6-1)

SiC + 3/2 0 2 * Si02 + CO (6.2)

These two reactions are competing, and consequently lead to an extremely complex situation. However, certain observations may be made from the data collected with regard to the unbalanced, overall reaction:

SiC + 02 + Cl2 ** SiClx + Si02 + C0X (6.3)

Both the mass spectrometric and thermogravimetric analyses

indicate that Si02 is not attacked by chlorine gas. Therefore, the

formation of a silica layer on the silicon carbide material will result in an inhibition of the corrosion reaction and will lead to a

low rate of weight loss and lack of SiCl species in the mass A spectrometer trace.

The data indicates that the corrosion reaction continues, although somewhat inhibited, even after a significant amount of oxygen

(~ 4%) is introduced into the system. That is, a noticeable SiClx signal appears on the mass spectrometer trace and a measurable weight loss is observed in the thermogravimetric analysis. Both sets of data show that the reaction occurs to a lesser extent as the amount of oxygen in the gas stream is increased. As previously discussed, this decrease in reaction results from the formation of an oxide film on

the sample surface.

The fact that the reaction continues even following the 170 introduction of oxygen into the gas stream indicates that the oxide film allows for the transport of reactant chlorine gas to the SiC material. This implies that the oxide film formed is somewhat porous.

This porosity may result from the liberation of CO which is a product of the oxidation reaction (Equation (6.2)).

A physical model of the "pure" SiC-Cl2-02 system is shown in figure 6.1. This model depicts the C12,02 gas mixture diffusing through the growing oxide film. The gas then may react to form more oxide and the gaseous SiClx corrosion product which is transported back out through the oxide scale. When the oxygen content of the gas stream is high (> 4%) there is no observation of corrosion products in the mass spectrometer and the rate of weight loss is minimal. This suggests that the oxide film becomes more resistant to the transport of gas as the oxygen content of the gas stream is increased.

6.1.2. SiC Containing Excess C.

The Sohio Hexoloy material sintered a-SiC material examined in this research contains a significant amount of free carbon (0.5 - 3

%). An x-ray map showing the distribution of the carbon in a polished section of this material is shown in figure 6.2^. The addition of free carbon to the reactant system further complicates the process by adding another chemical reaction to the system:

C + x/2 02 * C0X (6.4)

This reaction is highly significant in the mechanism of the overall reaction (equation (6.3)) in that it provides another means for the 171

Si Cl

SiO-

SiC Figure 6.1 - Physical model describing corrosion mechanism of "pure" SiC exposed to a CI2 -O2 gas stream. b

Figure 6.2 - SEM photomicrograph (A) and x-ray fluorescence map of carbon distribution (B) in a polished section of Sohio Hexoloy sintered a-SiC (from ref. 97). 173

consumption of the reactant oxygen gas. The reaction of carbon with

oxygen results in a lower partial pressure of oxygen at the surface of

the solid. This decrease in the partial pressure will make the

formation of an oxide layer less likely.

The mass spectrometer and thermogravimetric results show little variation in the corrosion process upon introduction of small amounts

of oxygen (1-2 %) to the Hexoloy-chlorine system. In fact, the mass

spectrometer results show a slight increase in the amount of volatile product observed when 1% O 2 is added to the system. This acceleration

of the corrosion reaction may be attributed to the stripping away of

the free carbon regions to expose fresh SiC surface for the

chlorination reaction. Since chlorine has a higher affinity for Si

than C removal of the carbon should expose more Si to the reactant

chlorine gas, and therefore, increase the amount of SiClx species

observed. However, it should be noted that the degree of acceleration

is small, indicating that the removal of carbon to expose fresh

surface is a process of secondary importance when compared to the lack

of oxide formation in these regions.

The data indicates that only when the oxygen content of the gas

stream is increased above 4% does an appreciable decrease in the

extent of the corrosion reaction occur. This implies that for the

Hexoloy samples, higher oxygen contents are required due to the

initial lowering of the partial pressure resulting from the free

carbon present, before the oxide scale formed is suitably protective 174 to alter the corrosion reaction . Again the fact that corrosion products are still observable in the mass spectrometer indicates that the oxide product layer formed is somewhat porous and allows transport of the reactant gas to the silicon carbide surface and subsequent transport of the reaction products back out through the oxide.

A physical model of this reaction system is presented in figure

6.3. In regions where free carbon is present in appreciable amounts there is a break in the oxide scale formed. This break provides an access area for the chlorination reaction to occur. Higher oxygen contents are required in such a system for the oxide layer formed to overcome the fact that these breaks exist and suitably protect the sample.

6.1.3. SiC Containing Excess Si.

The Norton NC430 SiC examined is produced by sintering a-SiC powder. The porosity in the resulting piece is closed by impregnation with liquid silicon. As a result of this processing scheme the final material contains approximately 10% free silicon. This free silicon is also an important factor in the mechanism of the overall reaction.

However this case differs from the case of free carbon ,in that, instead of preventing the formation of an oxide scale the free silicon promotes it:

Si + 02 ** Si02 (6.5)

Both the mass spectrometric and thermogravimetric analyses illustrate that the behavior of silicon in environments containing Free C

L Low Pn Region u 2

Figure 6.3 - Physical model describing corrosion mechanism of high carbon activity SiC exposed to a CI2 -O2 gas stream. 176

oxygen and/or chlorine is more severe than that of SiC. Silicon is more readily attacked by chlorine and yet more easily oxidized than

SiC.

In light of these observations the results obtained for the high

Si content SiC are not surprising. The large amount of free silicon

present leads to an accelerated rate of attack in a chlorine-

containing environment as a result of preferential attack of the

silicon phase. This excess silicon also leads to a reduced weight

loss rate in mixed oxygen/chlorine environments containing small

amounts of oxygen as a result of the more rapid oxidation of the

silicon.

A schematic diagram of these differences are shown in figure 6.4.

As described, the reaction of this material in high chlorine

content gases results in the preferential attack of the silicon

present (a). The reaction with oxygen on the other hand results in

the rapid formation of an oxide on the Si regions, and decreases

the overall area available for the chlorination reaction (b).

6.II. Mass Transport.

6.11.1. Theory.

6.11.1.1. Laminar Flow Past a Flat Plate.

The rate of chlorination of SiC in argon-oxygen-chlorine mixtures has been assumed to be controlled by mass transfer in the gas phase

through a laminar boundary layer of gas near the corroding

63 interface . For mixed oxidation/chlorination of SiC in parallel 177

SiCI

SiC

CL-O,

SiCI

SiO SiC Si-rich

b

Figure 6. Schematic diagram describing corrosion mechanism of high Si activity SiC when exposed to; a. a CI2 gas stream, and b. a CI2 -O2 gas stream. 178 flow, a stagnant boundary layer is formed adjacent to the sample surface. The rate of chlorination is determined by the diffusion of the SiClx compounds formed by the reaction;

SiC + x/2 Cl2 ** SiClx + C (6.6) through the boundary layer. At a distance S from the sample surface the partial pressure of the SiClx becomes small when compared with that at the sample surface. A schematic diagram of the boundary layer is shown in figure 6.5.

Under steady state conditions Fick's first law states:

J = DSiClx(PSiClx ' PSiClx'>/5RT

Where: Dgidx “ diffusivity of SiClx

T = absolute temperature R - gas constant

PSiClx> ” partial pressure of SiClx at a distance S from the surface PSiClx “ partial pressure of SiClx at the solid-vapor interface

From figure 6.5, ^ s i d x ' ^ PSiClx’ equation (6.7) becomes:

JSiClx “ DSiClxPSiClx/5RT (6-8>

The weight loss observed has proven to be nearly linear and may be described by a linear rate constant, k. This constant gives the weight loss of SiC per unit area per unit time. Dividing this constant by the molecular weight of SiC (Mgic) yields the number of moles of SiC which are consumed per unit time per unit area. From equation (6.6) one mole of SiClx is produced for every mole of SiC consumed. Therefore, we may write: 179 Gas

I f If

Boundary Layer

Figure 6.5 - Schematic diagram of boundary layer. JSiClx = k/MSiC

Equating (6.8) and (6.9):

k = M SiCDSiClxPSiClx/5RT (6.10)

For laminar flow past a flat plate the following expression

i . 98 applies :

hL/D = 0.664 (NSc)1/3(NRe)1/2 (6.n)

Where: h = average film mass transfer coefficient L = sample length parallel to the direction of g flow NSc “ Schmidt Number NRe = Reynolds Number

The quantity hL/D is normally referred to as the dimensionless ma transfer coefficient and is analogous to the dimensionless heat transfer coefficient defined for heat transfer by convection".

By means of this analogy, the diffusion boundary layer is given by:

5 “ D/h (6.12)

Combining equations (6.11) and (6.12):

6 - (NSc)'1/3(NR e )-1/2L / 0 .664 (6.13)

By definition:

NSc = »//D (6.14a)

NRe - VL/iv (6.14b)

Where: V *= uniform gas flow through the furnace v = kinematic viscosity = viscosity/density

Combining equations (6.13), (6.14a), and (6.14b):

S = 1/0.664 (D/i/)1/ 3 (jz/VL)1/ 2 L 181

or: D1/3l,l/6Ll/2 6 ------— (6.15) 0.664

Substituting equation (6.15) into (6.10) we obtain an expression

for the linear rate constant in terms of the partial pressure of the

gaseous corrosion product:

m n 2/3 1/2 SiC SiClx V 0.664 PSiClx (6.16) RT v 1/6 L

For partial pressures in atmospheres this equation gives the linear rate constant in grams per square centimeter per second.

6.II.1.2. Determination of Constants.

The viscosity of the gas stream was assumed to be that of pure

Ar. The viscosity of a pure gas may be written as^®®:

(MT)1/2 r, x 107 - 266.93 _ ------[f_(3)(T*)] (6.17) a 1 (T )] 1

Where: T) - viscosity T - absolute temperature T - reduced temperature = Tk/e M = molecular weight a = collision diameter

fl(2’2)*(T*) *= collision integral as a function of reduced temperature

f^(3 ^(T*) = correction factor

For Ar the molecular weight is 39.944 and the collision diameter and potential diameter are 3.542 and 93.3 respectively^®^. The reduced temperature at 950 °C is 13.1. The collision integral and 182

correction factor at a reduced temperature of 13.1 are 0.79915 and

1.0075 respectively. These values yield a viscosity (calculated from equation (6.17)) of 5.93 x 10"4 g/cm-s.

The interdiffusion coefficient was calculated for SiCl^ diffusing through a boundary layer of Ar. The gaseous diffusion coefficient may 102 be determined from the equation :

t 3/2

°AB - 1-8583 * 1 0 '3 < V M A + VMb) ------2------(6.18) PcrAB fiD,AB

Where: Ma - molecular weight of species A Mg = molecular weight of species B = collision diameter “ 1/2(a^ + <7g)

Og “ collision integral

Substituting the appropriate values into equation (6.18) yields a diffusion coefficient of 0.696 cm /s.

The kinematic viscosity of the gas stream was calculated to be o 0.33 cm /s and the uniform gas flow rate past the sample, for a constant 400 ccm flow into the furnace at 950 °C, was determined to be

7.71 cm/s. These values lead to a relationship between the linear rate constant, k and the partial pressure of the volatile products of:

k = 6.10 x 10'4 -Psiclx (6.19)

6.II.2. Correlation Between Calculated and Observed Rates.

6.II.2.1. S0LGASMIX-PV Results.

The equilibrium partial pressures as determined by the SOLGASMIX-

PV program were used to calculate equilibrium rate constants based on 183 equation (6.19). The partial pressures of the volatile products determined by the SOLGAS program and the corresponding linear rates are compared to the experimental values determined for Norton NC203 hot-pressed SiC in Table 6.1.

Table 6.1 shows that the agreement between the calculated and experimental values is reasonably good for the 2% CI2 and 1% CI2 , 10%

O 2 gas streams. However, the values are always lower in the calculated rates. This difference may easily be explained by the lack of equilibrium in the experimental situation. The flowing gas experimental configuration leads to a constant removal of the reaction products which tends to push the chlorination reaction (equation

(6,6)) to the right. This will tend to produce larger amounts of volatile products than predicted by the calculation.

The fact that the agreement between the calculated and experimental values is better as the oxygen content of the gas stream

is increased is also significant. This fact is a manifestation of the method in which the equilibrium compositions are obtained by the

SOLGAS program. The equilibrium calculations are performed by

inputting the desired number of moles of each element into the program. This implies that a study of the SiC-Cl2-02 system is actually performed by inputting equal molar amounts of Si and C with the appropriate amounts of CI2 and O 2 . The program assumes that the oxygen will combine favorably with the silicon to produce a solid piece of SiC^. This effectively reduces the amount of SiC available 184

Table 6.1

Comparison of Linear Rate Constants Determined Experimentally and From the SOLGASMIX-PV Calculation Results at 950 °C.

k (g/cm2 -s) Gas Mixture PSSiClx (atm> SOLGASMIX-PV Experimental

2% ci2 1.0 x 10'3 6.10 x 10'7 1.92 x 10'6

1% cl2 ’ 2% 02 6.8 x 10'6 4.12 x 10’9 3.66 x 10-7

1% ci2 > 4% 02 2.2 x 10'6 1.34 x 10'9 1.86 x 10'8

1% ci2 > 1 0 % o2 1.7 x 10'6 1.04 x 10'9 7.78 x 10'9 185 for chlorination. The calculated results, in effect, approximate an impermeable oxide film on the SiC substrate. Therefore as the oxide film experimentally formed becomes more impermeable (i.e. - higher oxygen content gas streams), the agreement between the calculated and experimentally determined weight loss rates becomes closer due to the fact that the test piece is essentially Si0 2 (i.e.- the SiC>2 layer is impermeable). In high oxygen content gas streams the minimal weight loss observed and the small amount of SiCl^+ seen in the mass spectrometer may be due to the slight reaction that occurs between

SiC>2 and CI2 .

6.II.2.2. Mass Spectrometer Results.

The ion intensities of the volatile products determined by mass spectrometry are proportional to the partial pressure of the molecules observed. The pressure, p is related to the ion intensity,

I+ through a proportionality factor, and the source temperature,

T:

p - £-I+-T (6.20)

For chlorine equation (6.20) becomes:

PC12 " ^ ‘IC12+ 'T

or;

P - A (6.21)

For the case of chlorination of SiC the results have shown that the predominant products formed are SiClg and SiCl^. However, as a result of fragmentation the intensities of all of the SiClx ions must 186 be considered. Therefore equation (6.20) for the actual volatile products is:

PzSiClx = P' ISSiClx+/T (6-22)

Combining equations (6.21) and (6.22):

PzSiClx “ (IZSiClx+/IC12+^ ' PC12 (6.23)

Where: PC12 ” Parti-al pressure of chlorine - 0 . 0 2 for 2% C ^ - A r gas - 0.01 for 1% CI2 , x% C^-Ar gas

The calculated linear rates determined are compared to the experimental values in Table 6.2.

Considering the semi-quantitative nature of the mass spectrometer partial pressure determination and the simplistic nature of the model, the agreement between the two sets of data is remarkably good. In fact when the variations of both methods (Mass Spectrometry - ± 10% and TGA - ± 5%) are accounted for the values correspond even more closely. This agreement between the pressure values indicates that transport through the boundary layer occurs by a simple mass transport process for the pure chlorination reaction. That is the reaction of

Si-based ceramics with chlorine under the conditions studied is controlled by diffusion of the product SiClx gases through the

stagnant boundary layer.

The values obtained for the oxygen-containing environments also

seem to fit the mass transport model. However, the pressure values used to calculate the the linear rate constants actually take into account the diffusion of reactants and/or products through a growing Table 6.2

Comparison of Linear Rates Determined Thermogravimetrically (TGA) with Those Determined from Mass Spectrometric (MS) Analysis.

k (mg/cm^-s x 10*®)

NC203 Hexoloy NC430 Si Gas MS TGA MS TGA MS TGAMSTGA

2% Cl2 54.9 192 21.6 99.7 32.0 308 270 720

2% 02* 2.47 36.6 5.83 96.1 1.95 14.7 0.52

4% 02* 0.348 1.86 2.25 4.34 0.449 1.05

1 0 % o2* 0.78 0.03 0.69

^ O all mixtures are 1% CI2 , x% C>2 -Ar, all temperatures - 950 C. 188 oxide layer. The fact that the calculated rates agree with the mass transport model but are lower than the rate observed for pure chlorination indicates that a second process is occurring which serves

to hinder the rate of reaction. This secondary process must be the

rate limiting step in the corrosion reaction, and is more than likely

diffusion of the reactant gases or reaction products through the oxide

layer.

It is unclear whether or not the decrease observed is due to

transport of reactants through the oxide to the base material or

transport of products through the oxide to the atmosphere. The process may be controlled by an average of both processes. It is also

difficult to ascertain whether the transport occurs as molecular

transport through the actual oxide structure or simply as transport

through the flaws and cracks in the oxide. Although the morphologies

suggest that the transport most likely occurs through the large flaws

in the oxide layer. Therefore, the specific process controlling the rate of reaction may not be deduced from the data obtained. CHAPTER VII

CONCLUSIONS

The research conducted on the behavior of Si-based ceramics in high-temperature environments containing oxygen and chlorine has resulted in a variety of interesting observations. In

general certain conclusions may be made regarding the behavior

of these materials and the mechanisms of this process.

The most general conclusion is that the reaction of silicon-

containing ceramics with chlorine produces volatile chloride products.

The predominant product formed is SiCl^ under all of the conditions

studied in this research. It has also been observed that this

chlorination reaction occurs more readily in the cases of Si and

SiC than for SiC^ and Si^N^. The lack of reaction in the silicon

nitride system is due to the existence of an impermeable oxide

glass film on these samples which prevents transport of the

reactant chlorine to the Si^N^ base material.

It has also been observed that the vigorous reaction occurring in

the silicon-chlorine system decreases to an immeasurable extent upon

the introduction of oxygen into the system. This decrease is due to

the rapid formation of a very protective silica layer on the sample

surface.

189 190

The effect of oxygen on the silicon carbide-chlorine system depends greatly on the minor phases that are added to the SiC powder during processing. These densification and sintering aids are particularly important when the oxygen content of the gas stream is

low. The small level additives are especially important when they

affect the overall carbon and silicon activities of the silicon

carbide material. Samples containing a higher carbon activity (i.e.-

samples with excess carbon) show a lesser dependence on oxygen

introduction than samples with a higher Si activity. In general the

high carbon activity materials show the smallest rate of reaction in a

chlorine environment and the largest rate of reaction in environments

containing various amounts of oxygen. The samples containing

excess silicon, on the other hand, show a severe rate of attack in

pure chlorine but a very low rate of attack in oxygen-containing

environments.

This large difference is a result of the different reactions

of these materials with oxygen:

C (.) + V 2 0 2 (g) = C0X (s) (7.1)

S1 (s) + °2 (g) “ si02 (s) <7-2)

The reaction of silicon with oxygen (in relatively high oxygen

pressures) produces a protective SiC^ layer whereas the reaction

of carbon with oxygen produces volatile C0X compounds. This

volatilization reaction maintains a low oxygen partial pressure

at the sample surface and prevents a protective oxide from 191

forming on the silicon carbide samples which possess a high C activity. This lack of oxide formation allows the chlorination reaction to continue more readily in the higher oxygen content gases.

The results also indicate that in environments containing only chlorine the rate of reaction is controlled by mass transport of the product gas (SiCl^) through the stagnant boundary layer of argon formed adjacent to the sample. However, when oxygen is added to the system the rate controlling mechanism is transport through the oxide grown on the sample surface. It is unclear whether the controlling mechanism is transport of products or reactants or whether the method is through the oxide structure or transport through the pore/crack system that exists in the growing oxide. In environments containing a high oxygen content (> 10%) the rate controlling step appears to revert to mass transport of the product gas through the boundary layer. In these high oxygen contents it is more difficult to transport reactants or products through the oxide film and consequently the behavior of the material approximates a monolithic silica material.

The mass spectrometer studies have also shown that silicon oxychloride compounds of the general formula,

SinOn _^Cl2 n + 2 are formed during the reaction of Si and SiC with mixtures of oxygen and chlorine. These compounds have been directly analyzed for the first time in this research. The results indicate 192

that these compounds are formed primarily by a gas phase reaction between the reactant oxygen gas and the SiCl^ formed by reaction of

the reactant chlorine with the Si-containing material. A new compound with the molecular formula SigOClg was also discovered. It is unclear whether this compound is an actual reaction product or simply an artifact of the combination of two or more species in the ionization chamber of the mass spectrometer. In any event, this compound is stable and is observable in the mass spectrometer.

No reaction was observed when Si-based ceramics were exposed to a

2% HCl-Ar environment at temperature up to 1050 °C. This lack of reaction is due to the difficulty involved with dissociation of the

HC1 molecule at these temperatures. CHAPTER VIII

SUGGESTIONS FOR FUTURE WORK

8.1. Mass Spectrometry.

8.1.1. Equipment Modifications.

The research conducted has shown that the high pressure mass

spectrometer sampling system is a versatile and valuable piece of

analytical equipment. A great deal of quantitative and qualitative

information may be obtained on the reactions occurring in a given

system in a relatively short period of time using this method. The

technique of free jet expansion gas sampling has proven to yield very

reproducible results and reasonably accurate determinations of the partial pressures of the reaction products.

In light of the value of this equipment, efforts should be made

to improve the system present at the University. The system has shown

reasonable success in the analysis of certain systems that liberate

fairly large amounts of products under non-hazardous conditions. If

the system is to become more versatile certain modifications must be made.

A serious problem with the current system is the manner in which

unreacted gases are vented from the furnace area. The current gas

removal system is inadequate for the removal of toxic gases and

193 194 reaction products from the laboratory. To combat this problem a new venting system must be installed. The design of this vent system should be based around a centrifugal fan which will allow for the rapid and complete removal of hazardous reactant and product gases.

A second means of avoiding the venting problem is to move from

the open system currently being used to a closed flow reactor system.

Closed-end alumina tubes have been machined for use in the flow

reactor configuration. These tubes have a small end wall thickness (~

4 mil) and have a 10 mil hole drilled along the axis of the tube. Use

of these tubes in the flow reactor should produce a better signal than

that obtained earlier"*. In addition to elimination of "spillover" of unreacted gas into the atmosphere the flow reactor configuration allows for a tighter control of the gaseous environment. This tighter control will permit the analysis of reactions that proceed undesirably when exposed to air.

After repeated success of the existing system has been demonstrated funds should be sought for the purchase of a quadrupole mass filter for mating with the current sampling system.

This purchase will increase the sensitivity of the system by a

factor of approximately five and will allow for in line

installation of the mass separator. The current system contains a

90° bend in the flow of the ions formed by ionization of the gas

stream. Elimination of this bend should produce better

sensitivity. 195

8.1.2. Future Studies.

As discussed, the implementation of an improved flow reactor will permit the examination of reactions that will not occur in air. An

example of this type of reaction is the of polymeric

precursors to advanced silicon-based materials. The polycarbosilanes

typically used as starting materials for the production of SiC

decompose rapidly when exposed to oxygen. Therefore to study these

reactions an inert environment of nitrogen or argon must be used.

This environment could be maintained in the flow reactor system.

Little is known about the pyrolysis of these materials, and

consequently a detailed mass spectrometer study of the pyrolysis

products will be useful.

The mass spectrometer sampling system is especially suited for

the analysis of high-temperature gas-solid reactions. Its uses for

the quantitative identification of hot corrosion products are

limitless. A particularly interesting study that is related

to this research is the behavior of advanced turbine engine

component materials in environments containing the impurities

typically found in diesel fuels. These impurities are

typically and vanadium. The effect of sulfur on these

materials may be examined by reacting the samples with SO2

gas. The flow reactor system may be better suited for this

study since its design makes it more efficient in the removal

of hazardous gaseous products than the open system. The 196 effect of vanadium on the corrosion of turbine engine materials may be examined by placing a small of V 2 0 tj below the sample. Raising the temperature of the furnace will result in volatilization of the V 2 0 ij anc* a subsequent

gas-solid reaction.

The recent use of SiC in aluminum reclamation environments also

provides fuel for further study of the mixed oxidation/halogenation

reactions. These environments usually contain which will

limit the protection characteristics of the oxide surface layer.

8 .II. Si-C-O-Cl System.

The research conducted has shown that mixed

oxidation/chlorination of silicon based ceramics produces many

interesting effects. Clearly a great number of unsolved questions

remain. In general the reaction of SiC with mixtures of oxygen and

chlorine may produce two overall reactions:

SiC + 2 0 2 + 2 Cl2 ** Si02 + SiCl4 + C02 (8.1)

SiC + 0 2 + 2 Cl2 ** SiO + SiCl4 + CO (8.2)

The conditions examined in this research dealt mainly with the former

reaction. However, slight evidence of the latter reaction was also

observed.

The active oxidation reaction (8.2) should provide interesting

results. Although two processes are still occurring in this system,

they are not necessarily competitive. That is only volatile products

are observed and passivation of the sample surface should not occur. 197

In order to effectively study this reaction the oxygen content of the reactant gas stream will need to be closely monitored. The oxygen partial pressure may be controlled in a number of ways. In addition to simply paying closer attention to the metering of the gas streams very close control of the oxygen pressure may be obtained by conducting the experiments in a furnace based around a zirconia tube. By placing a known EMF across the wall of the zirconia tube the amount of oxygen transported through the furnace tube and consequently the oxygen content of the gas stream may be controlled.

Along with the examination of the active oxidation situation more work is necessary to better understand the passive oxidation case

(equation (8.1)). Of distinct importance is a better understanding of the oxide layer formed on the sample surface during exposure to the mixed oxidation/chlorination environments. The experimental conditions studied to date have not been conducive to the formation of thick oxide layers on the samples. In most cases the oxide layer formed is difficult to observe. Therefore, experimental conditions should be examined which produce thicker oxide films. This may be accomplished by raising the temperature of the furnace to the 1300 °C level and also paying closer attention to the behavior of silicon in these environments. Silicon has shown more rapid oxidation behavior than silicon carbide and therefore its role as a reference material should be more thoroughly examined. In addition to the more rapid oxidation at high temperatures it should be interesting to analyze the reaction products of the chlorination reaction at higher temperatures. The calculations predict that as the temperature of the reaction is increased above

1300 °C the SiClj species should be present in amounts nearly equal to the SiCl^. However before this question may be investigated the furnace mated to the system will require modification and an alternate sampling cone material may be necessary to prevent degradation of the Pt-Rh cone at these higher temperatures. REFERENCES

1. A.F. McClean, "Overview of ARPA/ERDA/FORD Ceramic Turbine Engine Program," in Ceramics for High Performance Applications - II. Vol. 5, J.J. Burke, E.N. Lenoe, and R.N. Katz, eds., 1-34, Brook Hill Publishing Company (1978).

2. D.W. Richardson, "Evaluation in the U.S. of Ceramic Technology for Turbine Engines", Bull. Am. Ceram. Soc., 64 282 (1985).

3. J. Roboz, "Chapter One - Introduction," in Introduction to Mass Spectrometry: Instrumentation and Techniques. 3-25, John Wiley and Sons, Inc. (1968).

4. T.A. Milne and F.T. Green, "Molecular Beams in High Temperature Chemistry," p. 107 in Advances in High Temperature Chemistry, vol.2, L. Eyring, ed., Academic Press, New York (1969).

5. D.S. Fox, "A High Pressure Mass Spectrometer Sampling System for the Study of Hot Gas Corrosion of Ceramics," M.S. Thesis, The Ohio State University, Columbus, Ohio (1985).

6 . H. Ashenkas and F.S. Sherman, "The Structure and Utilization of Supersonic Free Jets in Low Density Tunnels," in Rarefied Gas Dynamics. Proceedings of the Fourth International Symposium, Vol.2, J.H. de Leeuw, ed., 84-105, Academic Press (1966).

7. J.B. Anderson, "Molecular Beams from Nozzle Sources," in Molecular Beams and Low Density Gas Dynamics. P.P. Wegener, ed., 1-91, Marcel Dekker, Inc. (1974).

8 . A.G. Hansen, Fluid Mechanics. John Wiley and Sons, Inc. (1967).

9. C.A. Stearns, F.J. Kohl, G.C. Fryburg, and R.A. Miller, "A High Pressure Modulated Molecular Beam Mass Spectrometric Sampling System," NASA Technical Memorandum 73726, NASA-Lewis Research Center, Cleveland, Ohio 44135 (1977).

10. S.A. Stern, P.C. Waterman, and T.F. Sinclair, "Separation of Gas Mixtures in a Supersonic Jet. II. Behavior of Helium-Argon Mixtures and Evidence of Shock Separation," J. Chem. Phys., 33[3]

199 200

805 (1960).

11. F.T. Greene and T.A. Milne, "Mass Spectrometric Sampling of High Pressure-High Temperature Sources," Adv. in Mass Spectrometry, 3 841 (1966).

12. A. Kantrowitz and J. Grey, "A High Intensity Source for the Molecular Beam. Part I. Theoretical," Rev. Scientific Instruments, 22 [5] 328 (1951).

13. W.L. Fite, "Selection of Frequency in Modulated Beam Mass Spectrometry," Extranuclear Laboratories, Inc., Res. Note 2.

14. R.A. Miller, "Crossed Molecular Beam Reactive Scattering of Barium and with ," Ph.D. Thesis, Case Western Reserve University, Cleveland, Ohio (1976).

15. T.A. Milne and F.T. Greene, "Mass Spectrometry in Inorganic Chemistry," p. 68 in Advances in Chemistry Series. J.L. Margrave, ed. (1968).

16. F.T. Greene, J. Brewer, and T.A. Milne, "Mass Spectrometric Studies of Reactions in Flames. I. Beam Formation and Mass Dependence in Sampling 1-Atm Gases," J. Chem. Phys., 40[6] 1488 (1964).

17. T.A. Milne, and F.T. Greene, "Direct mass Spectrometric Sampling of High Pressure Systems," in Mass Spectrometry in Inorganic Chemistry. R.F. Gould, ed., Am. Chem. Soc. Advances in Chemistry Series, 72 (1968).

18. J.B. Anderson, R.P. Andres, and J.B. Fenn, "Supersonic Nozzle Beams," Adv. Chem. Phys.,10 275 (1966).

19. H. Dun, B.L. Mattes, and D.A. Stevenson, J. Chem. Phys., 38 161 (1979).

20. S.S. Lin, "Quadrupole Mass Spectrometer for Direct Sampling of Atmospheric Gases and Condensation Products," AMMRC TR 73-9, Army Materials and Mechanics Research Center, Watertown, MA (1973).

21. S.S. Lin, "Detection of Large Water Clusters by a Low RF Quadrupole Mass Filter," AMMRC TR 73-30, Army Materials and Mechanics Research Center, Watertown, MA (1973).

22. R.L. Pecsok, L.D. Shields, T. Cairns, and I.G. McWilliam, "Chapter 16 - Mass Spectrometry of Organic Compounds," in Modern Methods of Chemical Analysis. 316, John Wiley and Sons, Inc., New York (1976). 201

23. G. Ervin, "Oxidation Behavior of Silicon Carbide," J. Am. Ceram. Soc., 41 247 (1958).

24. P.J. Jorgensen, M.E. Wadsworth, and I.B. Cutler, "Oxidation of Silicon Carbide," J. Am. Ceram. Soc.,42 613 (1959).

25. P.J. Jorgensen, M.E. Wadsworth, and I.B. Cutler, "Effects of Oxygen Partial Pressure on the Oxidation of Silicon Carbide," J. Am. Ceram. Soc., 43 209 (1960).

26. P.J. Jorgensen, M.E. Wadsworth, and I.B. Cutler, "Effects of Water Vapor on Oxidation of Silicon Carbide," J. Am. Ceram. Soc., 44[6 ] 258 (1961).

27. E.A. Gulbransen, K.F. Andrew, and F.A. Brassart, "The Oxidation of Silicon Carbide at 1150° to 1400 °C and at 9 x 10 to 5 x 10 Torr Oxygen Pressure," J. Electrochem. Soc.,113[12] 1311 (1966).

28. J.W. Hinze and H.C. Graham, "The Active Oxidation of Si and SiC in the Viscous Gas Flow Regime," J. Electrochem. Soc.,123[7] 1066 (1976).

29. W.C. Tripp and H.C. Graham, "Oxidation of Si-jN^ in the Range 1300° to 1500 °C," J. Am. Ceram. Soc., 58[9-10] 399 (1976).

30. A.J. Kiehle, L.K. Heung, P.J. Gielisse, and T.J, Rockett, "Oxidation Behavior of Hot-Pressed SioNA," J. Am. Ceram. Soc., 58[1-2] 17 (1975).

31. G.N. Babini, A. Bellosi, and P. Vincenzini, "Oxidation of Silicon Nitride Hot-Pressed with Ceria," J. Am, Ceram. Soc., 64[10] 578 (1981) .

32. D.R. Clarke, "Comparison of Reducing and Oxidizing Heat Treatments of Hot-Pressed Silicon Nitride," J. Am. Ceram. Soc., 6 6 [2] 92 (1983) .

33. R.M. Horton, "Oxidation Kinetics of Powdered Silicon Nitride," J. Am. Ceram. Soc., 52[2] 121 (1969).

34. T. Hirai, K. Niihara, and T. Goto, "Oxidation of CVD SioNA at 1550° to 1650 °C," J. am. Ceram. Soc., 63[7-8] 419 (1980).

35. J.M. Blocher, Jr., "Strusture/property/process relationships in chemical vapor deposition CVD," J. Vac. Sci. Tech., 11 680 (1974). 202

36. S.S. Lin, "Mass Spectrometric Studies on High Temperature Reaction Between Hydrogen Chloride and Silica/Silicon," J. Electrochem. Soc., 123[4] 512 (1976).

37. S.S. Lin, "The Mass Spectrometric Investigation of the Chemical Vapor Deposition Process for Alumina", AMMRC TR 75-20, Army Materials and Mechanics Research Center, Watertown, Massachusetts, October (1975).

38. S.S. Lin,"Mass Spectrometric Analyses of Vapor in Chemical Vapor Deposition of Alumina," J. Electrochem. Soc., 122 1405 (1975).

39. J.H.E. Jeffes and C.B. Alcock, "The Production of Refractory Crystals by Vapour Transport Reactions," J. Mat. Sci., 3 635 (1968).

40. 0. Kubaschewski, E. LL. Evans, and C.B. Alcock, p. 221 in Metallurgical Thermochemistry. 5 edition, Pergamon Press, New York (1979).

41. J.M. Oh, M.J. McNallan, G.Y. Lai, and M.F. Rothman, "High Temperature Corrosion of Superalloys in an Environemt Containing both Oxygen and Chlorine," Met. Trans., 17A 1087 (1986).

42. R.P. Viswanath, D. Rein, and K. Hauffe, "High Temperature Corrosion of NiCrAl in Oxygen-Chlorine Mixtures," Werkstoffe Und Korrosion,31 778 (1980).

43. Y. Ihara, H. Ohgame, K. Sakiyama, and K. Hashimoto, "The Corrosion Behavior of Iron in Hydrogen Chloride Gas and Gas Mixtures of Hydrogen Chloride and Oxygen at High Temperatures," Corros. Sci., 23 805 (1982).

44. Y. Ihara, H. Ohgame, K. Sakiyama, and K. Hashimoto, "The Corrosion Behavior of Nickel in Hydrogen Chloride Gas and Gas Mixtures of Hydrogen Chloride and Oxygen at High Temperatures," Corros. Sci., 22 901 (1981).

45. N.S. Jacobson, "Reaction of Iron with Hydrogen Chloride - Oxygen Mixtures at 550 °C," Oxidation of Metals, 26[3-4] 157 (1986).

46. N.S. Jacobson, M.J. McNallan, and Y.Y. Lee, "The Formation of Volatile Corrosion Products During the Mixed Oxidation- Chlorination of Cobalt at at 650 C," Met. Trans. A, 17A 1223 (1986).

47. N.S. Jacobson, "Application of an Atmospheric Pressure Mass 203

Spectrometer to Chlorination Reactions," NASA TM87270, NASA-Lewis Research Center (1986).

48. L.E. Katz, in VLSI Technology. McGraw-Hill Publishing Company, New York, 131 (1983).

49. Y.J. van der Meulen, C.M. Osburn, J.F. Ziegler, "Properties of SiOo Grown in the Presence of HC1 or C^," J. Electrochem. Soc., 122[2] 284 (1975).

50. A. Rhotagi, S.R. Butler, F.J. Feigl, H.W. Kraner, and K.W. Jones, "Chlorine Incorporation in HC1 Oxides," J. Electrochem. Soc., 126[1] 143 (1979).

51. Y.D. Sheu, S.R. Butler, F.J. Feigl, and C.W. Magee, "The Distribution of Cl Impurities in Si0 2 Films Produced by of Si in Cl-Containing Ambients," J. Electrochem. Soc., 133[10] 2136 (1986).

52. J.W. Rouse, C.R. Helms, B.E. Deal, and R.R. Razouk, "Auger Sputter Profiling and Secondary Ion Mass Spectrometry Studies of Si02 Grown in HCI/O2 Mixtures," J. Electrochem. Soc., 131[4] 887 (1984).

53. H. Frenzel and P. Balk, "Structure and composition of the Si/Si0 2 interfacial region on TCE/O2 and CCI4/O 2 oxidized silicon, J.Vac. Sci. Tech., 16 1454 (1979).

54. H.L. Tsai, S.R. Butler, D.B. Williams, H.W. Kraner, and K.W. Jones, "Cl Incorporation at the Si/Si0 2 Interface During the Oxidation of Si in HCI/O2 Ambients," J. Electrochem. Soc., 131[2] 411 (1984).

55. M.D. Monchowski, J.R. Monkowski, I.S.T. Tsong, J. Stach, and R.E. Tressler, "Microstructure Development During the Thermal Oxidation of Silicon in Chlorine Containing Ambients," J. Non-Crystalline Solids, 49 201 (1982).

56. J.W. Rouse, C.R. Helms, B.E. Deal, and R.R. Razouk, "Summary Abtract: Auger Sputter Profiling Studies of Si0 2 Grown in O 2 /HCI Mixtures," J. Vac. Sci. Tech., 18[3] (1981).

57. D.W. Hess and R.C. McDonald, "Investigation of Silicon Etching and Bubble Formation During Silicon Oxidation in HC1- Oxygen Atmospheres," Thin Solid Films, 42 127 (1977).

58. J. Monkowski, J. Stach and R.E. Tressler, " Phase Separation and Passivation in Silicon Oxides Grown in HCI/O2 Ambients," J. 204

Electrochem. Soc., 126[7] 1129 (1979).

59. J. Monkowski, R.E. Tressler, and J. Stach, "The Structure and Composition of Silicon Oxides Grown in HCI/O2 Ambients," J. Electrochem. Soc., 125[11] 1867 (1978).

60. R.H. Doremus, "Chapter 4 - Phase Separation," in Glass Science. John Wiley and Sons, New York, (1973).

61. K, Hirabayashi and J. Iwamura, "Kinetics of Thermal Growth of HC1- O 2 Oxides on Silicon," J. Electrochem. Soc., 120(11] 1595 (1973).

62. R. J. Kriegler, "The Role of HC1 in the Passivation of MOS Structures," Thin Solid Films, 13 11 (1972).

63. M.J. McNallan, S.Y. Ip, S. Saaam, and W.W. Liang, "High Temperature Corrosion of SiC Based Ceramics in Chlorine Containg Environments," p.328 in High Temperature Materials Chemistry - III. Z.A. Munir and D. Cubicciotti, eds., The Electrochemical Sciety Press, Pennington, NJ (1986)

64. W.C. Schumb and D.F. Holloway, "The Oxychlorides of Silicon and the Correponding Ethyl ," J. Am. Chem. Soc., 63 2753 (1941).

65. W.C. Schumb and C.H. Klein, "The Oxybromides of Silicon," J. Am. Chem. Soc., 59 261 (1937).

6 6 . W.C. Schumb and A.J. Stevens, "Partial Hydrolysis of Silicon Tetrachloride," J. Am. Chem. Soc., 69 726 (1947).

67. W.C. Schumb and A.J. Stevens, "The Partial Hydrolysis of Silicon Tetrachloride," J. Am. Chem. Soc.,72 3178 (1950).

6 8 . I.R. Beattie and G. McQuillan, "The Interaction of Silicon Tetrachloride and Water Vapor in the Gaseous Phase," J. Chem. Soc. Lonon, 2072 (1962).

69. M.J. Rand, "Purity Examination of Silicon and Halides by Long-Path Infrared Spectrophotometry," Anal. Chem., 35(13] 2126 (1963).

70. R.H. Vought, "Molecular Dissociation by Electron Bombardment: A Study of SiCl^," Physical Review, 71(2] 93 (1947).

71. I.L. Agafanov, G.G. Devytykh, and N.V. Larin, "Mass Spectrum of Silicon Tetrachloride," Russian J. Inorg. Chem., 8(7] (1963). 205

72. H.J. Svec and G.R. Sparrow,"Mass Spectra of the Silcon Tetrahalides: Metastable Ions," J. Chem. Soc. (A), 1162 (1970).

73. V.S. Ban, "Mass Spectra of Silicon Containing Industrial Gases," Mat. Res. Bull., 10 81 (1975).

74. S.R. Brinkley, Jr., "Note on the Condition of Equilibrium for Systems of Many Constituents," J. Chem. Phys., 14 [9] 563 (1946).

75. S.R. Brinkley, Jr., "Calculation of the Equilibrium Composition of Systems of Many Constituents," J. Chem. Phys., 15 [2] 107 (1947).

76. W.B. White, S.M. Johnson, and G.B. Dantzig, "Chemical Equilibrium in Complex Mixtures," J. Chem. Phys., 28 [5] 751 (1958),

77. G. Eriksson, "Thermodynamic Studies of High Temperature Equilibria," Acta Chemica Scand., 25 [7] 2651 (1971).

78. N. Ingri, W. Kakolowicz, L.G. Sillen, and B. Warnqvist, "High Speed as a Supplement to Graphical Methods - V," Talanta, 14 1261 (1967).

79. T.M. Besmann, "SOLGASMIX-PV, A Computer Progran to Calculate Equilibrium Relationships in Complex Chemical Systems," Oak Ridge National Laboratory Report, ORNL/TM-5775, Oak Ridge, Tennessee 37830 (1977).

80. D.R. Stull and H. Prophet, "JANAF Thermochemical Tables, Second Edition," NBS Publication 37, U.S. Government printing Office, Washington (1971).

81. L.B. Pankratz, "Thermodynamic Properties of Elements and Oxides," U.S. Bureau of Mines Bulletin 672, U.S. Government Printing Office, Washington (1982).

82. C.E. Wicks and F.E. , "Thermodynamic Properties of 65 Elements - Their Oxides, Halides, , and Nitrides," U.S. Bureau of Mines Bulletin 605, U.S. Government Printing Office, Washington (1963).

83. 0. Kubachewski and C.B. Alcock, Metallurgical Thermochemistry. Pergamon Press, New York (1979).

84. M. Bruce Fegley, Jr., "The Thermodynamic Properties of Silicon Oxynitride," J. Am. Ceram. Soc. 64 [9] C124 (1981).

85. G. Fischman, "Appendix II," M.S. Thesis, University of Illinois, Urbana, Illinois (1983). 206

86. R.S. Gordon and B.J. McBride, "Computer Program for Calculation of Complex Chemical Equilibrium Compositions, Rocket Performance, Incident and Reflected Shocks, and Chapman-Jouguet Detonations," NASA Report SP-275, NTIS, Springfield, Virginia (1971).

87. C.S. Herrick and R.A. Sanchez-Martinez, "Equilibrium Calculations for the Si-H-Cl System from 300 to 3000 K," J. Electrochem. Soc. 131 [2] 454 (1984).

88. L.P. Hunt and E. Sirtl, " A Thorough Thermodynamic Evaluation of the Silicon-Hydrogen-Chlorine System," J. Electrochem. Soc. 119 [12] 1741 (1972).

89. M. Farber and R.D. Srivastava, "Enthalpies of formation of the chlorides," J. Chem. Thermodyn. 11 939 (1979).

90. J.B. Mann, "Ionization Cross Sections of the Elements Calculated from Mean-Square Radii of Atomic Orbitals," J. Chem. Phys., 46 1646 (1967).

91. R.T. Grimley, "Chapter 8 - Mass Spectrometry," p. 195 in The Characterization of High-Temperature Vapors, J .L . Margrave, ed., John Wiley and Sons, New York (1967).

92. J. Schilichting, "Oxygen transport through silica surface layers on silicon-containing ceramic materials," High Temperatures - High Pressures, 14 717 (1982).

93. J .A. Costello and R.E. Tressler, "Oxidation Kinetics of Silicon Carbide Crystals and Ceramics: I, In Dry Oxygen," J. Am. Ceram. Soc., 69 [9] 674 (1986).

94. B.E. Deal and A.S. Grove, "General Relationship for the Thermal Oxidation of Silicon," J. Appl. Phys., 36 [12] 3770 (1965).

95. S.C. Singhal and F.F. Lange, "Effect of Alumina Content in the Oxidation of Hot-Pressed SiC," J. Am. Ceram. Soc., 58 [9-10] 433 (1975).

96. R.A. Miller and F.J. Kohl, "Computer Programs for the Interpretation of Low Resolution Mass Spectra: Program for Calculation of Molecular Isotopic Distribution and Program for Assignment of Molecular Formulas," NASA TM X-3564, NASA - Lewis Research Center, Cleveland, Ohio 44135 (1977).

97. N.S. Jacobson and J.L. Smialek, "Hot Corrosion of Sintered a-SiC at 1000 °C," J. Am. Ceram. Soc., 68 [8] 432 (1985). 207

98. W.H. Rohsenomw and H. Choi, p. 148 in Heat. Mass. and Momentum Transfer. Prentice Hall, New York (1961).

99. E.R.G. Eckert and R.M. Drake, Jr., Heat and Mass Transfer. McGraw- Hill Book Company, New York (1959).

100. J.O. Hirschfelder, C.F. Curtiss, and R.B. Bird, Molecular Theory of Gases and Liquids. John Wiley and Sons Inc., New York (1954).

101. R.V. Svehla, "Estimated Viscosities and Thermal Conductivities for Gases at Elevated Temperature," NASA. TR-132, NASA-Lewis Research Center, Cleveland, Ohio (1962).

102. R.B. Bird, W.E. Stewart, and E.N. Lightfooot, Transport Phenomena. John Wiley and Sons, New York (1960). APPENDIX A

»** 80LGA3H1X-PV

THEODORE N. BESMANN OAK RIDGE NATIONAL LABORATORY OAK RIDGE, TN 37*30

CALCULATES EQUILIBRIA AT CONSTANT PRESSURE OR VOLUME. THE PROGRAM IE MODIFICATION Or SOLGASMIX (G. ERIKSSON, CHEHICA SCRIPTA., S (197S) 10

IMPLICIT REAL'S (A-H.O-X) COMMON A(99,20), A0(99,20), AKT(99), AKTF(99), B(20,99), SG(99), Pl<40), PTOT, T, TEXT(99), V<99), Vr(99),YTOT(40), V, $KH(«0), L, MO.Nl.KA, MB, MP(20), ML(20), MP, MS, N0(99), NP,NV,NW DIMENSION PEr(99), OE(99,<), HP(99), IEL(20), IFH(IS), IGT(6), $IOK(99), XI(99),T1TLE(10),RL(20),OY(99,6), B0(20) OPEN(UNlT-5,TYPE-'OLD',PILE-'SOL.DAT') OPEN(UNIT-6,TYPE-'NEM',riLE-'SOL.OUT') H • DLOGU0.D+00) IN-5 IOUT-6 Mf(1) - 1 26 PTOT - 1. DO «» X - 1, 40 40 KH(K) - 0 READ (IN,240) (TITLE(I),1-1,10) 240 rORMAT (10A0) WRITE (IOUT.230) (TITLB(I), 1-1,10) 230 FORMAT ( .20X.10A0) READ (IN,102) L, MP, HR, (ML(M), H - 1, HP) 102 FORMAT (4012) WRITE (IOUT,250) L,MP,MR,(NL(N),N-1,NP) 250 FORMAT (>0NO. Or ELEMENTS • ',I2,5X,'NO. OF MIXTURES - '.I2.5X, 1'NO. OF INVARIANT SOLIDS - ',12/,'ONO. OF SPECIES PER MIXTURE - ', 210(12,',')) IP (MP .SO. 1) GO TO 146 DO 112 N - 2, MP MF(H) - HL(M-l) ♦ 1 132 HL(H) - ML(M-l) ♦ ML(H) 146 Ml - ML(NP) ♦ 1 MS - ML(MP) ♦ MR READ (IN,260) (EL(I),1-1,L) 260 FORMAT (20A4) DO 13 I - 1, MS XI(I1 - 0. READ (IN,260) (A(I,J),0-1,L) 260 FORMAT (16FS.0/16F5.0) DO 33 J - 1, L 33 A0(X,J) - A(I,J) READ (IN,200) NV,V 200 rORMAT (11,9X,E10.0) READ (IN,102) (IEL(

208 209 DO 267 N-1,6 267 GY(I,N)»0. DO 266 N-l.MGT 266 GY(I,IGT(N))-GE(I,N) WRITE (IOUT,300) (TEXT(I), (GY(I,N),N-1,6 ),I-1,MS) 300 FORMAT ('OSPECIES',6X,'A',HX,'B',14X,'C',14X,'D',14X,'E',14X,'F', l/,99{*0',A8,3X,6(E12.5,3X),/)) 310 WRITE (IOUT,230) (TITLE(I ),1-1,10) WRITE (IOUT,340) (BL(I),I-1,L) 340 FORMAT ( '0 ',20X,'SUBSCRIPTS ON ELEMENTAL SYMBOLS OF EACH SPECIES', 1/,'0',11X,20(A4,2X)) DO 350 I-1,MS 350 WRITE (IOUT,360) TEXT(I),(A(I,J), J-1,L) 360 FORMAT ( '0 ',A8,lX,20(F5.2,lX)) 42 READ (IN,103) T WRITE (IOUT,122) T 122 FORMAT (4H0T -, F8.2, 2H It/) IF (MGT .EQ. 0) GO TO 82 DO 84 I - 1, MS 84 G(I) - 0. DO 81 N - 1, MGT IF (IGT(N) .LT. 6) TP • T**(IGT(N) - 3)/8.31433 IF (IGT(N) .EQ. 6) TP - DLOG(T)/8.31433 DO 81 I - 1, MS 81 G( I) - G (I ) -I- GE(I,N)*TP GO TO 8 82 DO 47 I - 1, MS READ (IN,103) FEF(I) 47 G(I) - (FEF(I) + 1000.*HF(I)/T)/8.31433 IF (MOK .EQ. 0) GO TO 4 DO 23 N - 1, MOK I - IOK(N) G( I ) - -H*FEF(I) DO 23 J - 1, L K - IEL(J) 23 G(I) - G(I) + A(I,J)/A(K,J)*G(K) 4 WRITE (IOUT,230) (TITLE(I ),1-1,10) WRITE (IOUT,106) (TEXT(I), HF(I), FEF(I), G(I), I - 1, MS) 106 FORMAT (15X, 5HHF298, 8X, 3HFEF, 9X, 4HG/RT/(' ' ,A8,F12.0,2F12.3)) GO TO 8 37 READ (IN,103) PTOT 103 FORMAT (8E10.0) 8 READ (IN,102) KVAL1 GO TO (40,40,40,7,34,42,37,85,26,1), KVALl 40 READ (IN,102) NPKT READ (IN,102) (KH(J ), J - 1, L) 7 DO 164 J - 1, L N - KH(J) GO TO (161,162,163,161,162,163), N 161 READ (IN,103) (B(J,N), N - 1, NPKT) GO TO 164 162 READ (IN,103) B(J,1) IF (NPKT .EQ. 1) GO TO 164 DO 160 N - 2, NPKT 160 B(J,N) - B(J,1) GO TO 164 163 READ (IN,103) B(J,1), STEP DO 165 N - 2, NPKT 165 B(J,N) - B(J,N-l) + STEP 164 CONTINUE GO TO (35,39,34,34), KVALl 35 DO 370 I-1,MS 370 READ (IN,103) Y(I) GO TO 166 39 BEST - 0. DO 169 J - 1, L IP (DABS(B(J,1)).GT. BEST .AND. RB(J) .LT. 4) BEST - DABS(B(J,1)) 169 CONTINUE BEST - BEST/DPLOAT(MS) DO 167 I - 1, MS 167 Y(I) - BEST 166 DO 135 H - 1, NP YTOT(H) - 0. HA - MP(H) MB - ML(M) DO 135 I - HA, MB 135 YTOT(M) - YTOT(M) + Y(I) 34 NP - 0 27 NP - NP + 1 DO 75 I - I, MS 75 N0(I) - 1 DO 22 J - 1, L B0(J) - B(J,NP) R - IEL(J) IF (RB(J ) .LT. 4) GO TO 170 PI(J) - (G(R) + H*B0(J))/A(R|J) GO TO 22 170 IP (B0(J) .NE. 0.) GO TO 22 DO 49 I-1,MS IP (A(I,J )*A(R,J ) .LT. 0.) GO TO 22 49 CONTINUE DO 30 I - 1, MS IP (A(I,J) .EQ. 0.) GO TO 30 N0(I) - 0 Y(I) - 0. 30 CONTINUE 22 CONTINUE CALL GASOL WRITE (IOUT,230) (TITLE(I), 1-1,10) WRITE (IOUT,111) T, PTOT 111 FORMAT (4B0T PS.2, 2H R/ 4R P 1PE10.3, 4H ATM/) IP (NV.EQ.0) GO TO 210 WRITE (IOUT,220) V 220 FORMAT (4B V -, 1PE10.3, 2B L/) 210 CONTINUE IP (M0.LE. MP) GO TO 64 WRITE (IOUT,118) 118 FORMAT (/50B TBE EQUILIBRIUM COMPOSITION BAS NOT BEEN OBTAINED) IP (NP - NPRT) 27,8,8 64 IP (M0.EQ. 0) WRITE (IOUT,121) 121 FORMAT (33B TBE SMALL Y-VALUE8 ARE NOT EXACT/) DO 17 J - 1, L R - IEL(J) IP (RB(J) .LT. 4) GO TO 98 XI(R) - 0. DO 50 I - 1, MS 50 XI(R) - XI(R) + A0(I,J)/A0(R,J)*Y(I) GO TO 17 98 IF(AO(R,J).EQ.O.O) GO TO 17 XI(R) - B(J,NP)/A0(R,J) 17 CONTINUE WRITE (IOUT,105) 105 FORMAT (15X,7flX*/MOLE,8X,6HY/MOLE,10X,5HP/ATM,8X,8HACTIVITY) DO 140 M - 1, MP IP (M .GT. 1) WRITE (IOUT,125) 125 FORMAT (41X, 13BMOLE FRACTION) MA - MF(M) MB - ML(M) DO 140 I - MA, MB 140 WRITE (IOUT,124) TEXT(I), XI(I), Y(I), Yr(I), ART(I) 124 FORMAT (' »,A8,4(3X,E12.5)) IP (MR .GT. 0) WRITE (IOUT,120) (TEXT(I), XI(I), Y(I), I - Ml, MS) 211 120 FORMAT (/(' ' ,A8,2(3X,E12.5))) CALL SPEQUA IF (NP - NPKT) 27,8,8 1 CONTINUE CLOSE(UNIT-6) CLOSE(UNIT-5) STOP END SUBROUTINE GASOL IMPLICIT REAL*8 (A-H,0-Z) COMMON A(99,20), A0(99,20), AKT(99), AKTF(99), B(20,99), $G(99), PI(40), PTOT, T, TEXT(99), Y(99), YF(99),YTOT{40), V, $KH(40), L, MO,Ml, HA, MB, MF(20), ML(20), MP, MS, N0(99), NP,NV,NW DIMENSION F(99), IFAS(40), ISOL(40), LX(99), N8UM(500), OPI(20), $R(40,42), YFTOT(40), YX(99), B0(20) MX - 0 BMAX-0. DO 183 J-1,L IF (DABS(B(J ,NP)) .GT. BMAX .AND. KH(J) .LT. 4) BMAX-DABS(B(J ,NP)) 183 B0(J)-B(J,NP) 171 18 - -1 MSUM - 0 NG - 0 55 ISUM - 0 MSA - 0 IF (M8 .LT. Ml) GO TO 47 DO 52 I - Ml, MS IF (Y(I) .EQ. 0.) GO TO 52 MSA - MSA + 1 ISOL(MSA) - I ISUM - ISUM 4 2**(I + MP - Ml) 52 CONTINUE 47 MPA - 0 NG - NG 4- 1 IF (NG .EQ. 501) NG - 1 NSUH(NG) - ISUM YSUH - 0. DO 152 M - 1, MP IF (YTOT(M) .EQ. 0.) GO TO 152 MPA - MPA 4 1 IFAS(MPA) - M NSUH(NG) - NSUM(NG) 4 2**(M - 1) YSUH - YSUM 4- YTOT(M) 152 CONTINUE IF (NSUM(NG) .LT. IS .OR. MPA 4- MSA .GT. L) GO TO 59 GO TO 69 86 NG-NG-1 59 IS - IS ♦ 2 DO 154 N - 1, MPA M - IFAS(N) 154 YTOT(M) - 0. 1SUM - 0 MPA - 1 YTOT(l) - 1. IF (MSA .EQ. 0) GO TO 73 DO 68 N - 1, MSA I - ISOL(N) 68 Y( I ) - 0. MSA - 0 73 IF (IS .EQ. 1) GO TO 74 IT - IS 61 H - 0 71 M - M 4 1 IF (2**M .LE. IT) GO TO 71 IF (M .GT. MP) GO TO 57 212 MPA - MPA + 1 IFAS(MPA) - M YTOT(M) - 1. MA - MF(H) MB - ML(M) DO 150 I - MA, MB IF (NO(I) .EQ. 1) Y( I ) - YSUH 150 CONTINUE GO TO 136 57 I - M + ML(MP) - MP MO-M IF (I .GT. MS) RETURN IF (N0(I) .EQ. 0) GO TO 59 MSA - MSA + 1 ISOL(MSA) - I ISUM - ISUM + 2**(H - 1) Y(I) - YSUM 136 IT - IT - 2**(M - 1) IF (IT .GT. 1) GO TO 61 IF (MPA + MSA .GT. L) GO TO 59 74 IFAS(1) - 1 IF (NSUM(NG) .GT. IS) NG - NG + 1 IF (NG.EQ.SOI) NG-1 NSUM(NG) - IS 69 IF (NG .EQ. 1) GO TO 129 DO 148 R - 2, NG IF (NSUM(NG) .EQ. NSUM(K-l)) GO TO 86 148 CONTINUE IF (MY .LT. 0 .OR. NSUM(NG-l) .EQ. MSUM) NSUM(NG-l) --NSUM(NG-l) IF (NG .EQ. 2) GO TO 129 IF (MY .GT. 0 .AND. NSUM(NG) .EQ.MSUM .OR. NSUN(NG-2) .GT. 0) 1GO TO 129 NSUM(NG-2) - -NSUM (NG-2) IF (NSUM(NG) .EQ. MSUM) GO TO 59 129 MY“0 DO 142 H - 1, MP MA - MF(M) MB - ML(M) IF (YTOT(M) .GT. 0.) GO TO 130 DO 126 I - HA, MB ART(I) - 0. AKTF(I) - 1. Y(I) - 0. 126 YF(I) - 0. GO TO 142 130 DO 45 I - MA, MB IF(Y(I) .GT. BMAX) Y(I)- BMAX/DFLOAT(MS) IF (N0(I) .EQ. 1 .AND. Y(I) .LT. l.E-8) Y(I) - l.E-8 AKTF(I) - 1. 45 LX(I) - 0 M0-H CALL ABER IF (YTOT(N) .EQ. 0.) GO TO 47 142 CONTINUE LSI - L + MPA ♦ MSA LS - LSI - 1 LS2 - LS + 2 134 DMIN - l.E-6 IVAR - 0 IVARJ - HL(1) - MS 16 DO 6 J - 1, LSI DO 6 K - J, LS2 6 R(J,K) - 0. DO 9 N - 1, MPA LI - L + N M - IFAS(N) 213 MA - MF(M) MB - ML(M) DO S I - MA, MB ir (Y(I) .EQ. 0.) GO TO 5 r(I) - G(I ) + DLOG(AKT(I)) R(L1,LS2) - R(L1,LS2) + F(I)*Y(I) DO 77 J - 1, L ir (A(I,J ) .EQ. 0.) GO TO 77 AY - A(I,J)*Y(I ) R(J,L1) - R(J,L1) 4- AY R(J,LS2) - R(J,LS2) 4- AY*F(I) DO 80 K - J, L 80 R(J,K) - R(J,K) 4- AY*A(I,K) 77 CONTINUE 5 CONTINUE DO 9 J ■ 1, L 9 R(J,L82) - R(J,LS2) - R(J,Ll) ir (MSA .EQ. 0) GO TO 63 DO 67 N - 1, MSA I - ISOL(N) K - L 4- MPA 4- N R(K,LS2) - G(I) DO 67 J - 1, L 67 R(J,R) - A(I,J) 63 DO 31 J - 2, LSI N - J - 1 DO 31 R - 1, N 31 R(J,K) - R(K,J) ir (MX .EQ. 0) GO TO 156 DO 131 J - 1, L 131 B0(J) - B(J,NP) DO 172 N - 1, MPA H - IPAS(N) ir (MF(M) .NE. ML(M)) GO TO 172 MA - MF(M) DO 89 J - 1, L IF (A(MA,J) .EQ. A0(MA,J )) GO TO 89 AKVOT - (1. - A(HA,J)/A0(MA,J))*Y(HA) DO 91 K - 1, L If (A(MA,K) .EQ. AO(MA.K)) B0(R) - B0(K) 4- AKVOT*A(MA,K) 91 CONTINUE GO TO 172 89CONTINUE 172 CONTINUE 156 DO 44 K - 1, L IF (KH(K) .LT. 4) GO TO 44 DO 83 J - 1, LSI 83 R(J,LS2) - R(J,LS2) - PI(K)*R(J,K) 44 R(K,LS2) - R(K,LS2) 4- B0(K) DO 10 K - 1, LS IF (KH(K) .GT. 3) GO TO 10 ELMAX » 0. DO 11 J - K, LSI IF(DABS(R(J,K)) .LE. ELMAX .OR. KH(J) 3) GO TO 11 MROW - J ELMAX -DABS(R(J,K)) 11 CONTINUE IF (ELMAX .GT. 0.) GO TO 36 IF (K .GT. L) GO TO 10 IF (B0(K)) 59,10,59 36 IF (MROW .EQ. K) GO TO 13 DO 15 N - K, LS2 RADBYT - R(MROW,N) R (MROW,N ) - R(K,N) 15 R(K,N) - RADBYT 13 KA - K + 1 214 DO 46 J - KA( LSI RKVOT - R(J,K)/R(K,K) DO 46 N - KA, LS2 46 R(J,N) - R{J,N) - RKVOT*R(K,N) 10 CONTINUE DO 20 N - 1, LSI K - LS2 - N IF (KH(K) .GT. 3) GO TO 20 IF (R(K,K) .NE. 0. .AND. R(K,LS2) .NE. 0.) GO TO 62 PI(K) - 0. K - K - L - MPA IF (K .LE. 0) GO TO 20 I - ISOL(K) Y(I) - 0. GO TO 55 62 PI(K) - R( K,LS2)/R(K,R) KA - K - 1 IF (KA .EQ. 0) GO TO 20 DO 58 J - 1, KA 58 R(J,LS2) - R(J,LS2) - PI(K)*R(J,K) 20 CONTINUE IF (IVAR .EQ. 0 .OR. IVARJ .GE. 0 .OR. SLAM .LT. 0.1) GO TO 66 DO 70 J - 1, L IF (DABS(PI(J)) .LT. l.E-8) GO TO 70 IF (DABS(OPI(J)/PI(J)-l.) .GT. DMIN) GO TO 65 70 CONTINUE NR - 0 IF (NG .EQ. 1) GO TO 155 DO 157 K - 2, NG IF (NSUM(NG) .EQ. -NSUM(K-l)) NR - NR + 1 157 CONTINUE 155 DO 143 M - 1, MP IF (YTOT(M) .GT. 0.) GO TO 143 M0-M CALL XBER YFTOT(M) - 0. DO 144 I - MA, MB YFTOT(M) - YFTOT(M) + YF(I) YX(I) - YF(I) AKT(I) - 0. 144 YF(I) - 0. IF (NV.EQ.0) GO TO 326 YA-0. DO 325 I-1,HL(1) 325 YA-Y(I)+YA PTOT-YA*.0821*T/V 326 CONTINUE IF (H .EQ. 1) YFTOT(M) - YFTOT(M)/PTOT 143 CONTINUE 145 DIFM - 1. DO 158 M - 1, MP IF (YTOT(H) .GT. 0. .OR. YFTOT(M) .LE. DIFM) GO TO 158 KA - M DIFM - YFTOT(M) 158 CONTINUE IF (DIFM .EQ. 1.) GO TO 138 IF (NR .EQ. 0) GO TO 159 NR - NR - 1 YFTOT(KA) - 1. GO TO 145 159 MSUM - NSUM(NG) YTOT(KA) - 1. MA - MF(KA) MB - ML(KA) DO 153 I - MA, MB 153 Y{I) - YSUM*YX(I)/YFTOT(KA) 215 GO TO 47 138 IF (MS .LT. Ml) GO TO 51 DirM - 0. DO 54 I - Ml, MS IF (N0(I) .EQ. 0 .OR. Y(I) .GT. 0.) GO TO 54 PIA - -G(I) DO 56 J - 1, L 56 PIA - PIA + A(I,J)*PI(J) IF (PIA .LT. DIFM) GO TO 54 KA - I DIFM - PIA 54 CONTINUE IF (DIFM .EQ. 0.) GO TO 51 Y(KA) - YSUM GO TO 55 51 IF (MX .EQ. 1) GO TO 93 DO 160 N - 1, MPA M - IFAS(N) IF (MF(M) .NE. ML(H)) GO TO 168 MA - MF(M) DO 173 J - 1, L IF (A(MA,J) .EQ. A0(MAfJ)) GO TO 173 MX - 1 GO TO 171 173 CONTINUE 168 CONTINUE 93 IVARJ - 0 GO TO 66 65 DO 60 J ■ 1, L 60 OPI(J) - PI(J) 66 SLAM - 1. DO 12 N- 1, MPA Ll « L + N M - IFAS(N) HA - HF(M) MB - HL(M) DO 12 I • MA, MB IF (Y(I) .EQ. 0.) GO TO 12 PIA - F(I ) - PI(L1) DO 19 J - 1, L 19 PIA - PIA - A(I,J)*PI(J) YX(I) - PIA*Y(I) IF (PIA .GT. SLAM) SLAM - PIA 12 CONTINUE IF (SLAM .GT. 1.) SLAM - 0.999*(SLAM - 0.5)/SLAM/SLAM IF (MSA .EQ. 0) GO TO 72 DO 53 N - 1, MSA I - ISOL(N) K - L ♦ MPA + N IF (PI(K) .LT. 0.) Ll - 0 IF (IVAR .EQ. 0 .OR. PI(K) .GT. -Y(I)) GO TO 53 IF (SLAM .LT. 1. .AND. -PI(K)/YSUM .LT. 1.E8) GO TO 53 Y(I) - 0. GO TO 55 53 Y(I) -DABS(PI(K)) 72 YSUM - 0. DO 128 N - 1, MPA M - IFAS(N) MA - MF(M) MB - ML(M) DO 29 I - MA, MB IF (Y(I) .EQ. 0.) GO TO 29 Y(I) - Y(I) - SLAM*YX(I) IF (Y(I) .LT. l.E-20) Y(I) - 0. 29 CONTINUE H0-H 216 CALL ABER IF (YTOT(M) .GT. 0.) GO TO 128 IT (SLAM .GT. 0.1) MY-MSUM - 1 GO TO 47 128 YSUM - YSUM + YTOT(M) I VAR * IVAR + 1 IF (IVAR .EQ. 25 .OR. IVAR .EQ. 50) DHIN - 100.*DMIN IF (IVAR .EQ. 75) GO TO 59 IF (IVARJ .LT. 0 .OR. Ll .EQ. 0 .OR. SLAM .LT. 1.) GO TO 16 IVARJ - IVARJ + 1 IF (IVARJ .EQ. 10) GO TO 88 DO 3 N - 1, MPA M - IFAS(N) MA - MF(M) MB - ML(M) DO 3 I - HA, MB IF(Y(I) .EQ. 0.) GO TO 3 IF (DABS(YX(I))/Y(I) .GT. l.E-6) GO TO 16 3 CONTINUE 88 DO 149 N - 1, MPA M • IFAS(N) M0-M CALL XBER IF (MA .NE. MB) GO TO 133 DO 92 J - 1, L IF (A(MA,J) .NE. A0(MA,J)) Y(MA) - A(MA,J)/A0(MA,J)*Y(MA) 92 CONTINUE 133 DO 149 I - MA, MB IF (N0(I) .EQ. 0 .OR. Y(I) .GT. 0.) GO TO 149 Y(I) - YTOT(M)*YF(I) IF (Y(I) .LT. l.B-20 .OR. LX(I) .EQ. 1) GO TO 149 IF (M .EQ. 1 .AND. YF(I) .GT. PTOT) Y(I)«YTOT(1)*PTOT IF (M .GT. 1 .AND. YF(I) .GT. 1.) Y(I)-YTOT(M) LX(I) - 1 GO TO 134 149 CONTINUE IF (IVARJ .EQ. 10) M - 0 RETURN END SUBROUTINE ABER IMPLICIT REAL*6 (A-H.O-Z) COMMON A(99,20), A0(99,20), AKT(99), AKTF(99), B(20,99), $G(99), PI(40), PTOT, T, TEXT(99), Y(99), YF(99),YTOT(40), V, $KH(40), L, H0,H1,MA, MB, MF(20), ML(20), MP, MS, N0(99), NP,NV,NW M-M0 YTOT(H) - 0. DO 2 I - HA, MB 2 YTOT(M) - YTOT(M) + Y(I) IF (YTOT(M) .GE. l.E-8) GO TO 151 YTOT(M) - 0. RETURN 151 CONTINUE IF (NV.EQ.0) GO TO 326 YA-0. DO 325 I-1,ML(1) 325 YA-Y(1 )+YA PTOT-YA*.0821*T/V 326 CONTINUE IF (M .EQ. 1) YTOT(l) - YTOT(l)/PTOT DO 127 I - MA, MB 127 YF(I) - Y(I)/YTOT(M) CALL FACTOR DO 141 I - MA, MB 141 AKT(I) - AKTF(I)*YF(I) RETURN ononnooononnnnoo 1CO 3 7 N T I N U E 3 Y{) AKT(I)/AKTF(I)139 YF{I) - 147 IVAR IVAR - 1 + 7 PIA PIA87 « A(I,J)*PI(J) + 3CO 8 N T I N U E CONTINUE1 CONTINUE5 4 CONTINUE2 3 $KH(40), M0,M1,MA, L, PTOT, TEXT(99), PI(40), $G(99), T, YF(99),YTOT(40), Y(99), MB, MF(20), V, ML(20), HP, MS, NP,NV,NW N0(99), $KH(40), M0,M1,MA, L, MB, MF(20), ML(20), MP, MS, NP,NV,NW N0(99), G 9, I4) PTOT,$G( TBXT(99), PI(40), T, YF(99), Y(99), 99), YTOT( V,40)', AKT(I) -DEXP(PIA) Y () YF(I)OYF(I) - MB ML(M) - MA MF(M) - M-MO RETURN CONTINUECONTINUE M0 GO TO (1,2,3,4,5,6,7,8,9,10), COMMONS AXT(99), A0(99,20), U A(99,20), B AKTF(99), R B(20,99), O U T I NFA E RETURN C T O R MA,DO - MB 137 I COMMON ART(99), A0(99,20),A(99,20), AKTF(99), B(20,99), RETURN RETURN RETURN RETURN END END O 3 I MA,DO - MB 139 I O 7 - , L 1, DO - J 87 MA,DO - MB I 38 P L ODM G- A X 1 ( DD L O I G M ( P E T N O T S ) I , 1 O 0 N .O D + Y 0 P 0 ) { 9 9 ) FI - AKT(I) - YF(I) S U B R O U T I NXB E E R I - -G(I) PIA - F (DABS(OYF(I)/YF(I)—1.)IF GO TO .GT. l.E-4) 147 F N() E. .R YI .T 0) GO .GT. 0.) TO (N0(X) .OR. Y(I) 38IF .EQ. 0 IMPLICIT REAL*8 (A-H,0-Z) GO .EQ. 0.) TO (YF(I) 137IP (PIA .GT. PLOG)IF PLOG PIA - IMPLICIT REAL*8 (A-H,0-Z) F (IVARIF CALL 25) .LT. FACTOR IVAR 0 - A 0 - ADA - E 0 V I A T I N GST O I C H I O M E T R ICO C E F F I C I E N T A KA T C F- T Jk|pfp I V I TCO Y E F F I C I E N T N O T I C ET H A TA NA M O U N TOM SU F I X B T S U T R A E NA S C R EWIN E H S E I C RLE TIS H E S TH D SA F A TM E Nl.E-20 R OT H ST E A T E M E N TN U M B E RTH A TCO R R E S P O N Y F - MY - F O L EFR A C T I O NO RPA R T I APR L E S S U R E R E G AT H R EM D I E X DA T S UASET R H P EN A EA U R C M T A I T V B E IM E T YCOFO I R X . E T R F UASO F R I CTH L E I E I , D NWCO EEX T I T M P HVA R P E O S R S S I I I T A O I B O NFO L S NMST E U RTH O S I C ESP TA H E L I C S O IIN ETH OB M S SP E E EN T R O YW N H I Y - AY - M O U N TO SU F B S T A N C E P UE T Q U A LT OZE R O . ( M A X I M U MO N EST O I C H I O M E T R I CCO E F F I C I E NAIS T L L O W E DTO DE V I A T E ) . m JkrTIVTTV 217 n n n n 1 0 CO1 0 N T I N U E 8 CO8 N T I N U E 9 CO9 N T I N U E 6CONTINUE 7 CO7 N T I N U E $KH(40), M0,M1,MA, L, PTOT, TEXT{99), $G(99), PI(40), T, YP(99),YTOT(40), Y(99), MB, MF(20), ML(20), V, MP, MS, N0(99), COMMON END RETURN END RETURN RETURN RETURN RETURN RETURN S U B R O U T I NSP E E Q U A IMPLICIT REAL*8 (A-H,0-Z) D E R I V A B LFR E O MTH EEQ U I L I B R I U MCO M P O S I T I O N . S P E Q UASU AIS B R O U T I NFO E RC A L C U L A T I O NO QU F A N T I T I EW S H I C HA R E ( 92) A(92) AKT(99), A0(99,20),A(99,20), AKTP(99), B(20,99), NP,NV,NW 218 APPENDIX B

Norton Company Industrial Ceramics Division Worcester, Massachusetts

Sohio Engineered Ceramics Niagara Falls, New York

Thermal American Fused Quartz Company Montville, New Jersey

Crysteco Company Wilmington, Ohio

Union Carbide Company Cleveland, Ohio

219 t "'1 - v k r < E 8 > ** .. -■ . — . ; ; r - 0 r : r ^ . +

-1 1 I* I r— CM - - CM 1 C -U ‘ t = in • 1 ID ’ft Ga , v- , ' ' _ “ S S .i ■ — -- J - 'fr ..r ~Z. \ •; 7 7. “■ ■ “■ -- • • :/ " - •: -■ n n n SiCl Ga in-5Ga SiCl C a i n - 10 SiCl Ga in-5Ga

220 249 253

f282 N> Qfc.284 P .286 288

‘310 5^314 314 O' +

3 mV

IZZ