PROPERTIES AND PROCESSING OF CHEMICAL VAPOR DEPOSITED ZINC SULFIDE

PROPERTIES AND PROCESSING OF CHEMICAL VAPOR DEPOSITED ZINC SULFIDE

PROPERTIES AND PROCESSING OF CHEMICAL VAPOR

DEPOSITED ZINC SULFIDE

by

John S. McCloy

_________________________

Copyright © John Stuart McCloy 2008

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF MATERIALS SCIENCES AND ENGINEERING

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2008

2

THE UNIVERSITY OF ARIZONA

GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by John S. McCloy entitled “Properties and Processing of Chemical Vapor Deposited Zinc Sulfide” and recommend that it be accepted as fulfilling the dissertation requirement for the

Degree of Doctor of Philosophy

Donald R. Uhlmann Date: 4/ 7/ 08

Barrett G. Potter Date: 4/ 7/ 08

Supapan Seraphin

Randal W. Tustison

Date: 4/ 7/ 08

Date: 4/ 7/ 08

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

Dissertation Director: Donald R. Uhlmann Date: 4/ 7/ 08

3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the copyright holder.

SIGNED: John S. McCloy

4

ACKNOWLEDGEMENTS

So many people have been important in the gathering of data for this project. First and foremost is my long-term research partner Ralph Korenstein. Ralph performed most of the heat treatments, etchings, and many transmission measurements. For a time, Patrick

Hogan also joined our ZnS development team and vestiges of his ideas, particularly of the action of metals with ZnS, can be found in this work. Many of the “old guard” of

ZnS at Raytheon must be thanked as well for useful discussions, including Chuck

Willingham, Rick Gentilman, and Barney diBenedetto. In Massachusetts, Lenn

Kupferberg was helpful with his knowledge of x-ray testing, and Todd Stefanik and C.

Scott Nordahl performed some of the polycrystalline x-ray measurements. In Arizona,

Brian Zelinski was a great sounding board for ideas in all arenas, as well as reading the complete manuscript and offering extremely valuable comments. Eric Fest offered help on scattering models and originally suggested the surface scattering form which appears here. W. Howard Poisl assisted with mechanical testing and powder XRD. Chris

Dudding spent many hours with me on the SEM, and assisted with EDS and transmission measurements. Gavin Buttigieg offered his knowledge of Raman spectroscopy. Phil

Anderson at the Arizona Materials Laboratory assisted with TEM and XRD measurements, and Prof. Supapan Seraphin performed some of the TEM. Prof. B.G.

Potter performed the photoluminescence measurements. Special thanks to Dave Burdick, who laboriously translated Sulfid Tsinka and several seminal papers on ZnS and ZnSe from the original Russian. Thanks to Dan Harris at China Lake for inspiring me to care about infrared windows and the fascination they offer in science and chemistry. Thanks to those who supplied samples of their ZnS material for my testing including Stephen

Jacobs of the University of Rochester for the Chinese ZnS, Klaus Wöhner of Vitron in

Germany, and Shay Joseph of Rafael in Israel. An extra special thanks to the librarians at

Raytheon, Amy Smith, Sharon McCausland, and Erin Rust, who tirelessly found me obscure and difficult papers and books for this research. To my committee, Dr. Donald

Uhlmann, Dr. B. G. Potter, and Dr. Supapan Seraphin from the University of Arizona and

Dr. Randal Tustison from Raytheon, thank you for seeing me through this milestone.

Special thanks to Don for being a great advisor and Randy for being my Raytheon champion. Surely I have forgotten some, so forgive me.

5

DEDICATION

I dedicate this dissertation to my patient, understanding, and ever-supportive family—my wife Christy and my sons Maxwell and Gavin. I could not have done it without Christy’s hard work in doing double parent duty. I thank the boys for dealing with their Dad after long hours in the office. I truly believe this milestone will benefit our family in the long run. I love you all.

6

TABLE OF CONTENTS

TABLE OF CONTENTS..............................................................................................................................6

LIST OF TABLES ........................................................................................................................................9

LIST OF FIGURES ....................................................................................................................................10

ABSTRACT .................................................................................................................................................13

1 INTRODUCTION...............................................................................................................................15

1.1

S

COPE OF

D

ISSERTATION

...............................................................................................................16

1.2

H

ISTORY OF

I

NFRARED

M

ATERIALS AND

Z

INC SULFIDE

.................................................................18

1.3

E

TYMOLOGY

..................................................................................................................................22

1.4

S

UMMARY

......................................................................................................................................23

2 PHYSICS AND CHEMISTRY OF ZINC SULFIDE.......................................................................24

2.1

C

HEMISTRY OF

Z

N

S .......................................................................................................................24

2.2

C

RYSTALLOGRAPHY

......................................................................................................................28

2.2.1

Phase Transformation and Twinning ....................................................................................31

2.3

E

LECTRONIC

S

TRUCTURE

...............................................................................................................35

2.3.1

Electronic Defects and Luminescence ...................................................................................36

2.3.2

Isoelectronic Oxygen .............................................................................................................40

2.4

V

IBRATIONAL STRUCTURE

.............................................................................................................41

2.4.1

Infrared and Raman active modes.........................................................................................43

2.4.2

Hydrogen Impurity ................................................................................................................44

3 COMMERCIAL PROCESSING OF ZINC SULFIDE ...................................................................46

3.1

I

NTRODUCTION

...............................................................................................................................46

3.1.1

Vapor Phase Equilibrium ......................................................................................................48

3.2

C

HEMICAL

V

APOR

D

EPOSITION OF

Z

N

S.........................................................................................51

3.2.1

Homogeneous and Heterogeneous CVD Reactions...............................................................52

3.2.2

Porosity in CVD ZnS .............................................................................................................57

3.2.3

Summary................................................................................................................................59

3.3

H

EAT

T

REATMENT OF

CVD Z

N

S....................................................................................................59

3.3.1

Annealing...............................................................................................................................60

3.3.2

Hot Isostatic Pressing (HIPing) ............................................................................................62

3.3.2.1

Hot Pressing and Sintering of ZnS Powders..................................................................................... 62

3.3.2.2

Hot Isostatic Pressing of CVD ZnS.................................................................................................. 63

3.3.2.3

Creep Mechanisms ........................................................................................................................... 65

3.3.2.4

Effects of HIP on Microstructure and Mechanical Properties .......................................................... 68

3.3.2.5

Effects of HIP on Optical Properties ................................................................................................ 73

3.3.3

Summary................................................................................................................................74

4 RESEARCH APPROACH .................................................................................................................75

5 EXPERIMENTAL DESCRIPTION..................................................................................................78

5.1

S

AMPLE

D

ESCRIPTION AND

P

REPARATION

.....................................................................................78

5.1.1

Samples characterized as-received........................................................................................78

5.1.2

Annealed Samples..................................................................................................................82

5.1.3

Hot-Isostatically Pressed (HIP’d) Samples...........................................................................84

5.1.3.1

Selection of Treatment Types........................................................................................................... 84

7

TABLE OF CONTENTS - Continued

5.1.3.2

Preparation of Samples for HIPing................................................................................................... 87

5.1.3.3

Specific Experimental Series............................................................................................................ 90

5.2

C

HARACTERIZATION

T

ECHNIQUES

.................................................................................................97

5.2.1

Microscopy ............................................................................................................................98

5.2.2

Diffraction .............................................................................................................................99

5.2.3

Spectroscopy........................................................................................................................102

5.2.4

Mechanical and Physical Testing........................................................................................106

5.3

D

ATA MANIPULATION

,

PROCESSING

.............................................................................................109

6 CHARACTERIZATION OF MICROSTRUCTURE AND PHYSICAL PROPERTIES ..........110

6.1

M

ICROSCOPIC

I

NVESTIGATION OF

Z

N

S ........................................................................................110

6.1.1

CVD ZnS..............................................................................................................................111

6.1.2

Hot Isostatic Pressed CVD ZnS...........................................................................................118

6.1.3

Comparison of CVD ZnS and HIP’d CVD ZnS ...................................................................121

6.1.4

Recrystallization ..................................................................................................................123

6.2

A

SSESSMENT OF

M

ETAL

F

OILS

.....................................................................................................135

6.3

D

IFFRACTION

I

NVESTIGATION OF

Z

N

S .........................................................................................142

6.3.1

X-ray diffraction ..................................................................................................................143

6.3.1.1

Polycrystalline sample diffraction.................................................................................................. 143

6.3.1.2

Powder diffraction.......................................................................................................................... 151

6.3.2

Electron diffraction .............................................................................................................155

6.4

P

HYSICAL

P

ROPERTIES

.................................................................................................................161

6.4.1

Density.................................................................................................................................161

6.4.2

Fracture Strength ................................................................................................................162

6.4.3

Chemical Composition ........................................................................................................170

6.4.3.1

X-ray Microanalysis....................................................................................................................... 171

6.4.3.2

Interstitial Gas Analysis (IGA)....................................................................................................... 174

7 CHARACTERIZATION OF OPTICAL PROPERTIES ..............................................................177

7.1

P

HOTOLUMINESCENCE

.................................................................................................................177

7.1.1

Spurious artifacts.................................................................................................................177

7.1.2

As-deposited CVD ZnS ........................................................................................................178

7.1.3

CVD ZnS Core Sections.......................................................................................................181

7.1.4

Hot isostatic pressed ZnS ....................................................................................................188

7.2

T

RANSMISSION

M

EASUREMENTS

.................................................................................................190

7.2.1

Single Crystal ZnS ...............................................................................................................194

7.2.2

Polycrystalline samples, no heat treatment .........................................................................196

7.2.2.1

Cores .............................................................................................................................................. 200

7.2.3

Heat treated samples ...........................................................................................................205

7.2.3.1

Heat treated without metal ............................................................................................................. 206

7.2.3.1.1

Red ZnS.................................................................................................................................... 206

7.2.3.1.2

Standard ZnS............................................................................................................................ 214

7.2.3.2

Heat treated with metal .................................................................................................................. 220

7.2.3.2.1

Pt series .................................................................................................................................... 221

7.2.3.2.2

Co series................................................................................................................................... 223

7.2.3.2.3

Ag series................................................................................................................................... 227

7.2.3.2.4

Others (Fe, Ni, Cu)................................................................................................................... 233

7.2.4

Discussion ...........................................................................................................................238

7.2.4.1

Extrinsic impurity absorptions ....................................................................................................... 240

7.2.4.2

Scattering ....................................................................................................................................... 242

7.2.4.2.1

Rayleigh Scattering .................................................................................................................. 243

7.2.4.2.2

Internal Surface Scattering ....................................................................................................... 248

8 DISCUSSION AND CONCLUSIONS.............................................................................................254

TABLE OF CONTENTS - Continued

8

8.1

W

HAT IS THE NATURE OF STANDARD

CVD Z

N

S?.........................................................................254

8.1.1

What is red ZnS? .................................................................................................................259

8.1.2

What is elemental ZnS? .......................................................................................................261

8.2

W

HAT IS THE NATURE OF TRANSFORMATION TO

M

ULTISPECTRAL

Z

N

S? .....................................263

8.2.1

What is the HIP doing? .......................................................................................................265

8.2.2

What is the metal doing? .....................................................................................................270

8.3

C

ONCLUSIONS

..............................................................................................................................273

8.4

S

UGGESTIONS FOR

F

UTURE

W

ORK

...............................................................................................276

8.5

F

INAL THOUGHTS

.........................................................................................................................279

9 APPENDIX A: DETAILED CRYSTALLOGRAPHY .................................................................281

9.1

S

TRUCTURES OF

Z

N

S ...................................................................................................................281

9.2

P

OLYTYPES OF

Z

N

S......................................................................................................................282

9.2.1

Notations for stacking..........................................................................................................284

9.2.2

X-ray methods and fault modeling.......................................................................................287

10 APPENDIX B: ELECTRONIC STRUCTURE .........................................................................292

10.1

E

LECTRONIC BAND GAP AND ABSORPTION EDGE

..........................................................................292

10.2

B

AND

S

TRUCTURE OF

S

PHALERITE AND

W

URTZITE AT

Z

ONE

C

ENTER

........................................294

10.3

P

OINT

D

EFECTS IN

Z

N

S ................................................................................................................295

10.3.1

Kröger Vink and charged defects ........................................................................................295

10.3.2

Defect Equilibria .................................................................................................................296

11 APPENDIX C: THERMODYNAMICS AND KINETICS .......................................................300

11.1

S

ULFUR

A

CTIVITY AND

G

IBBS

E

NERGY OF

C

REATION OF

Z

N

S ....................................................300

11.2

O

XYGEN IN

Z

N

S: T

HERMODYNAMICS AND

K

INETICS

..................................................................303

11.3

V

APOR

P

RESSURES

.......................................................................................................................309

11.4

D

IFFUSION

....................................................................................................................................310

12 APPENDIX D: ENGINEERING PROPERTIES AND MISCELLANEOUS DATA ............316

12.1

E

LASTIC

P

ROPERTIES

...................................................................................................................316

12.2

H

ARDNESS

...................................................................................................................................317

12.3

T

OUGHNESS

.................................................................................................................................318

12.4

S

TRENGTH AND

W

EIBULL

A

NALYSIS

...........................................................................................319

12.4.1

Fracture testing of ceramics by ring on ring testing ...........................................................319

12.4.2

Weibull Analysis procedures ...............................................................................................320

12.4.3

Definitions of strength .........................................................................................................322

12.4.4

Area effect............................................................................................................................325

12.5

I

MPURITY CONTENT OF

CVD Z

N

S

BY

GDMS ..............................................................................327

13 APPENDIX E: DETAILS OF XRD ANALYSIS.......................................................................329

13.1

A N

OTE ABOUT METHOD

..............................................................................................................329

13.2

T

EXTURE ANALYSIS

.....................................................................................................................330

13.3

N

ONSTOICHIOMETRY

F

ACTOR

......................................................................................................336

13.4

D

ISORDERED

F

RACTION AND

I

111

/I

200

R

ATIO

................................................................................337

13.5

C

RYSTALLITE

S

IZE AND

C

OHERENTLY

D

IFFRACTING

D

OMAINS

..................................................340

13.6

L

ATTICE PARAMETER ANALYSIS

...................................................................................................342

13.7

H

EXAGONAL FRACTION

................................................................................................................347

13.8

S

UMMARY OF

XRD

QUANTITATIVE ANALYSIS

.............................................................................353

14 APPENDIX F: DERIVATION OF THE SURFACE SCATTERING EQUATION ..............356

15 REFERENCES ..............................................................................................................................358

9

LIST OF TABLES

Table 5-1: Materials characterized as-received without further processing .................... 81

Table 5-2: Summary of the different annealing experiments on CVD ZnS .................... 84

Table 5-3: Summary of the different metal treatments for HIP experiments .................. 87

Table 5-4: Special experiments: no metal, low-P HIP, non-contact HIP ....................... 91

Table 5-5: Platinum HIP experiments.............................................................................. 92

Table 5-6: Silver HIP experiments .................................................................................. 93

Table 5-7: Cobalt HIP experiments ................................................................................. 95

Table 5-8: Simultaneous foils, Interrupted HIP, Fe, Ni, and Cu HIP experiments ......... 97

Table 6-1: Summary of the Effects of Different Metals on ZnS ................................... 141

Table 6-2: T-test on biaxial flexure strength of ZnS samples........................................ 164

Table 6-3: ZnS biaxial flexure results............................................................................ 165

Table 6-4: X-ray spectrometry: EDS, 300 seconds, raw data....................................... 172

Table 6-5: X-ray spectrometry: EDS, data normalized to Bridgman crystal................ 173

Table 6-6: Interstitial Gas Analysis first run results for oxygen.................................... 174

Table 6-7: Interstitial Gas Analysis second run results for oxygen and hydrogen ........ 175

Table 6-8: Volume versus surface sites for oxygen in IGA samples............................. 176

Table 7-1: Extinction coefficients and UV edge of Pt foil HIP’d samples.................... 222

Table 7-2: Extinction coefficients and UV edge of 750 °C Ag foil HIP’d samples...... 229

Table 9-1: Some of the more important polytypes of ZnS. ........................................... 286

Table 9-2: Examples of stacking fault types.................................................................. 288

Table 10-1: Room temperature ultraviolet cut-on edge of ZnS..................................... 293

Table 10-2: Defect concentrations of oxygen and hydrogen in ZnS ............................. 299

Table 11-1: Dissociation pressures of ZnO and H

2

O(g) at 650 °C................................ 304

Table 11-2: Some possible reactions in the Zn-O-S-H complex equilibrium ............... 309

Table 11-3: Diffusion data for ZnS................................................................................ 314

Table 12-1: Published values for Young’s modulus (E) and Poisson’s ratio (ν)........... 317

Table 12-2: ASTM standards relating to ceramic strength testing ................................ 320

Table 12-3: Various ways of calculating the probability of failure ............................... 323

Table 12-4: What is meant by “strength” ...................................................................... 323

Table 13-1: Literature ZnS lattice parameters ............................................................... 344

Table 13-2: Measured ZnS Lattice Parameters.............................................................. 345

Table 13-3: Lattice parameters correlated to oxygen concentration only...................... 346

Table 13-4: CVD ZnS groupings by hexagonality ........................................................ 355

Table 13-5: Hexagonality and texture of all x-rayed samples ....................................... 355

10

LIST OF FIGURES

Figure 1-1: Transmission of Candidate Longwave infrared Window Materials ............. 20

Figure 1-2: Band-gap energy versus LO mode frequency for LWIR window materials 22

Figure 2-1: Wurtzite (L) and Sphalerite (R) structures.................................................... 29

Figure 2-2: Energy band structure and density of states for sphalerite and wurtzite ZnS36

Figure 2-3: Phonon density of states for ZnS and calculated dispersion......................... 42

Figure 3-1: Free energy of formation for 2 moles of ZnS or ZnO from the elements..... 47

Figure 3-2: CVD ZnS operation of 60” diameter production furnaces ........................... 52

Figure 3-3: Illustration of the proposed model of ZnS growth........................................ 54

Figure 3-4: Nodular growth on the surface of CVD ZnS away from the mandrel .......... 69

Figure 5-1: Orientation nomenclature for samples taken for CVD ZnS “cores”............. 82

Figure 5-2: HIP fixturing and insulation hood................................................................. 88

Figure 5-3: Typical HIP schedule showing temperature and pressure ............................ 89

Figure 5-4: Schematic of set-up for photoluminescence measurements ....................... 106

Figure 6-1: Hierarchies of structure in CVD ZnS.......................................................... 110

Figure 6-2: Macroscopic features in polished CVD ZnS............................................... 111

Figure 6-3: Micrographs of CVD ZnS perpendicular and parallel to growth direction 112

Figure 6-4: SEM micrographs of standard CVD ZnS showing nanosized lamellae ..... 114

Figure 6-5: TEM images of CVD ZnS at 80k................................................................ 116

Figure 6-6: TEM images of CVD ZnS at 60kx.............................................................. 116

Figure 6-7: TEM images of CVD ZnS at 30kx.............................................................. 117

Figure 6-8: TEM images of CVD ZnS at 20kx.............................................................. 117

Figure 6-9: TEM images of CVD ZnS at 10kx.............................................................. 118

Figure 6-10: Optical and SEM micrographs of Hot Isostatic Pressed CVD ZnS.......... 119

Figure 6-11: TEM images of heat treated ZnS .............................................................. 120

Figure 6-12: CVD ZnS (L) versus HIP’d CVD ZnS (R) at various length scales (1) ... 121

Figure 6-13: CVD ZnS (L) versus HIP’d CVD ZnS (R) at various length scales (2) ... 122

Figure 6-14: Etched HIP’d ZnS samples with partial recrystallization. ........................ 123

Figure 6-15: Etched HIP’d ZnS samples showing threshold for recrystallization ........ 124

Figure 6-16: Etched ZnS samples showing 750 °C treatments ..................................... 125

Figure 6-17: Etched heat treated ZnS samples recrystallized without metal................. 125

Figure 6-18: Etched HIP’d ZnS samples ....................................................................... 127

Figure 6-19: Recrytallization of sample exposed to Cu foil in 900 °C interrupted HIP 128

Figure 6-20: Samples exposed to Pt & Ni in 900 °C interrupted HIP ........................... 130

Figure 6-21: Samples exposed to Cu, Co, & Fe in 900 °C interrupted HIP .................. 131

Figure 6-22: Low temperature interrupted HIP micrographs for metal foils................. 132

Figure 6-23: Comparison of Pt and Cu foil interrupted HIP at 2 temperatures............. 134

Figure 6-24: Recrystallization by copper foil at low temperature ................................. 135

Figure 6-25: SEM of Co foil after treatment of ZnS and corresponding XRD pattern . 136

Figure 6-26: SEM of Fe foil after treatment of ZnS and corresponding XRD pattern.. 137

Figure 6-27: SEM of Pt foil after treatment of ZnS and corresponding XRD pattern... 139

Figure 6-28: SEM of Cu foil after treatment of ZnS and corresponding XRD pattern . 140

Figure 6-29: SEM of Ag foil after treatment of ZnS and corresponding XRD pattern . 141

LIST OF FIGURES - Continued

11

Figure 6-30: X-ray diffraction of polycrystalline specimens of ZnS............................. 147

Figure 6-31: XRD comparisons for CVD versus HIP CVD.......................................... 149

Figure 6-32: XRD comparisons for Silver versus No Metal HIP.................................. 149

Figure 6-33: X-ray powder diffraction of the heat treated group. ................................. 153

Figure 6-34: X-ray powder diffraction of the transparent as-deposited group. ............. 154

Figure 6-35: X-ray powder diffraction of the opaque group. ........................................ 154

Figure 6-36: Powder diffraction versus polycrystalline diffraction for the cores.......... 155

Figure 6-37: Raytheon ZnS 60kx TEM and SAED of highly twinned region .............. 157

Figure 6-38: Raytheon ZnS 40kx TEM and SAED of highly twinned region .............. 158

Figure 6-39: Red ZnS 10kx TEM and SAED of highly twinned region ....................... 158

Figure 6-40: Raytheon ZnS 40kx TEM and SAED of two similar regions................... 159

Figure 6-41: Red ZnS 20kx TEM and SAED showing small variously oriented grains159

Figure 6-42: Princeton Scientific ZnS 40kx TEM and SAED of two regions............... 160

Figure 6-43: Histogram of measured densities of ZnS samples .................................... 162

Figure 6-44: ZnS biaxial flexure probability of failure versus applied stress................ 166

Figure 6-45: Stress-strain curve of fractured samples in compression test ................... 169

Figure 6-46: Images of fractures by compression parallel to CVD ZnS columns......... 170

Figure 6-47: Images of fractures by compression perpendicular to CVD ZnS columns170

Figure 7-1: Photoluminescence in Raytheon CVD ZnS ................................................ 179

Figure 7-2: Photoluminescence in Raytheon Elemental ZnS ........................................ 181

Figure 7-3: Photoluminescence in R&H ZnS core, mandrel side.................................. 182

Figure 7-4: Photoluminescence in R&H ZnS core, mandrel side, comparison ............. 183

Figure 7-5: Photoluminescence in R&H ZnS core, growth side ................................... 184

Figure 7-6: Photoluminescence in R&H ZnS core, growth side, comparison............... 185

Figure 7-7: Photoluminescence in R&H ZnS core, middle ........................................... 186

Figure 7-8: Photoluminescence comparison of all the cores at 10 K ............................ 186

Figure 7-9: Layering in the cores and luminescence difference along growth direction188

Figure 7-10: Photoluminescence in Pt HIP Raytheon multispectral ZnS...................... 189

Figure 7-11: Transmission of ZnO, ZnS, and ZnSe....................................................... 195

Figure 7-12: Band edge transmission of ZnO, ZnS, and ZnSe...................................... 196

Figure 7-13: CVD ZnS Transmission: redZnS, msZnS, eZnS, stdZnS........................ 198

Figure 7-14: ZnS Transmission: stdZnS, hot-pressed powder ZnS, Chinese CVD ZnS199

Figure 7-15: Slab of ZnS cut from the core along the growth direction........................ 201

Figure 7-16: Extinction parallel and perpendicular to the CVD growth direction ........ 205

Figure 7-17: Transmission and photographs of annealed Red ZnS samples ................. 208

Figure 7-18: Transmission at band edge of annealed Red ZnS samples ....................... 210

Figure 7-19: Putative evidence for ZnO precipitates and comparison with water lines 213

Figure 7-20: Transmission of ZnS heat treated without metal: 700 °C & 750 °C......... 215

Figure 7-21: Transmission of ZnS heat treated without metal: 800 °C & 850 °C......... 217

Figure 7-22: Transmission of ZnS heat treated without metal at 900 °C. ..................... 218

Figure 7-23: Transmission curves for CVD ZnS HIP’d with platinum foil .................. 222

Figure 7-24: Transmission of ZnS HIP’d with sputtered cobalt showing absorption ... 225

Figure 7-25: Transmission of ZnS HIP’d with sputtered cobalt showing no absorption 225

Figure 7-26: Transmission of ZnS HIP’d with cobalt foil............................................. 226

12

LIST OF FIGURES - Continued

Figure 7-27: Transmission of ZnS (visible wavelengths) HIP’d with silver foil .......... 228

Figure 7-28: Transmission and micrographs of silver foil HIP experiment samples .... 229

Figure 7-29: Transmission of ZnS (visible wavelengths) HIP’d sputtered silver ......... 230

Figure 7-30: Transmission of ZnS (infrared wavelengths) HIP’d sputtered silver ....... 232

Figure 7-31: Transmission of ZnS HIP’d with iron foil ................................................ 234

Figure 7-32: Transmission of ZnS HIP’d with sputtered nickel.................................... 235

Figure 7-33: Transmission of ZnS HIP’d with sputtered copper and copper foil ......... 237

Figure 7-34: Transmission curves compared for 750 °C, 16 hours with various metals238

Figure 7-35: Transmission of ZnS in the vicinity of the hydride absorption................. 241

Figure 7-36: Extrinsic infrared absorptions in ZnS samples ......................................... 242

Figure 7-37: Rayleigh scattering models for hot-pressed ZnS and CVD ZnS .............. 247

Figure 7-38: ZnS internal surface scattering model schematic...................................... 251

Figure 7-39: Internal surface scattering model of msZnS, eZnS, & standard ZnS........ 253

Figure 9-1: Correspondence of various notations for stacking sequence ...................... 287

Figure 9-2: XRD of polycrystals and possible evidence for short period polytypes ..... 291

Figure 10-1: Approximate intrinsic defect locations in the bandgap of cubic ZnS. ....... 297

Figure 11-1: Activity of sulfur in the hydrogen sulfide reaction ................................... 302

Figure 11-2: Gibbs energy of formation of CVD ZnS from H

2

S. ................................. 302

Figure 11-3: Standard state free energy change per mole for ZnO and H

2

O................. 303

Figure 11-4: Reaction rate of hydrogen with oxygen as function of pressure............... 306

Figure 11-5: Mole fraction and partial pressures of sulfur allotropes in sulfur vapor... 309

Figure 11-6: Vapor pressures of various materials important in the CVD ZnS process 310

Figure 13-1: Comparison of predominant crystallographic texture by sample ............. 336

Figure 13-2: Peak broadening fitting for phase size analysis using Scherrer formula .. 342

Figure 13-3: Example of hexagonality determination and corresponding XRD ........... 352

Figure 13-4: Polytype stacking in single crystal ZnS .................................................... 353

13

ABSTRACT

The structure and properties of chemical vapor deposited zinc sulfide (CVD ZnS) were assessed before and after heat treatments, involving different annealing and hot isostatic pressing (HIPing) profiles. Samples were characterized using optical microscopy, SEM, TEM, electron diffraction, polycrystalline and powder x-ray diffraction, x-ray chemical microanalysis, photoluminescence, ultraviolet through longwave infrared transmission, and mechanical testing.

Before heat treatment, CVD ZnS consists of lamellar twinned structures in 10 to

100 nm layers aggregated into domains which compose grains typically 5 to 10 μm in diameter with an overall crystallographic texture on the { 100 } planes. The scattering behavior of CVD ZnS was investigated and described by a surface scattering model based on internal surface roughness and refractive index variations due to onedimensional stacking disorder. The two to five percent hexagonality measured by x-ray diffraction is believed to form due to oxygen impurities at the twin boundaries which cause nanostructural polytypism and result in differential refractive index and scattering.

CVD ZnS variants in low temperature deposited red ZnS and sulfur precursor elemental

ZnS are examined as well. Color in CVD ZnS is believed to be due to band edge position, probably due to oxygen content, and not directly related to the hydride absorption at 6 μm.

After annealing or hot isostatic pressing above 850 °C for sufficient time, CVD

ZnS recrystallizes and becomes strongly textured on the { 111 } planes. This recrystallization is required to remove stacking disorder, resulting in a structure with less

14 than half a percent hexagonality and low visible scattering. The recrystallization is believed to proceed by diffusing the oxygen at the nano-twin boundaries back into the lattice, thus unpinning the boundaries and allowing them to move and grow into the tabular recrystallized morphology by polytype induced exaggerated grain growth. The presence of active metals like platinum, silver, copper, or nickel during hot isostatic pressing causes a reaction with sulfur and lowers the temperature required for recrystallization. The optical scattering model is consistent in describing standard CVD

ZnS, elemental ZnS, and multispectral recrystallized ZnS as having successively lower birefringence at internal surfaces.

15

1 Introduction

Zinc sulfide (ZnS) has shown unequaled utility for infrared windows that require a combination of longwave infrared (8-12μm) transparency, mechanical durability, and elevated temperature performance. Its unique set of properties lends its usefulness also to electroluminescent phosphors, optical thin films for anti-reflection, and various other opto-electronic applications. High optical quality chemical vapor deposited ZnS windows several millimeters thick transmit visible light and so have received attention as candidates for multi-spectral windows.

ZnS is a wide-bandgap II-VI semiconductor with properties that vary widely with processing conditions. Traditionally, bulk ZnS for infrared windows is manufactured by chemical vapor deposition (CVD) in large reactors. Deposition temperature & mole fractions of the reactants, H

2

S gas and Zn vapor, have a large influence on visible scatter and absorption. These effects have been ascribed to electronic defects in the bandgap, hexagonal phase ZnS, and residual porosity, but the exact mechanisms have never been adequately explained.

Multispectral ZnS is formed by taking traditionally grown polycrystalline CVD

ZnS which is visibly yellow and opaque, then subjecting it to a hot-isostatic press while wrapped in platinum foil. The heat and pressure result in recrystallization of the CVD

ZnS, large grain growth, and a visibly clear product. The kinetics of this post-process, as well as the dependence on the platinum foil, are poorly understood. It is known that starting CVD materials grown under different conditions do not behave identically when subsequently heat treated.

16

1.1 Scope of Dissertation

This project aims to understand the relationship between processing, properties, and microstructure in various grades of chemical vapor deposited zinc sulfide. The processing steps to be explored include the chemical vapor desposition step and any subsequent heat treatment step. The properties to be investigated are primarily related to its application as a window material – optical and mechanical. Finally, the microstructure is probed at various scales and explained in terms of the processing which, in turn, explains some of the observed properties.

Two specific questions emerged early on that had been left unanswered in the literature. What is the source of transmission loss in the visible and infrared for asdeposited CVD ZnS? Given this, why does the transmission improve to near-theoretical values when subjected to hot isostatic pressing in the presence of platinum? To answer these questions, I take the reader through the existing literature on ZnS then delve into the research specific to this dissertation. Finally I interpret the results in terms of the chemistry, crystallographic structure, impurities, and hierarchical structure of ZnS in its various forms.

Chapter 1 sets the stage by reviewing the motivations behind the push for development of chemical vapor deposited ZnS for infrared windows.

Chapter 2 is a tour of the physics and chemistry of zinc sulfide. The chemistry of stoichiometry and primary impurities is addressed first. Then the crystallography of sphalerite, wurtzite, and the polytypes is explained along with issues regarding the hexagonal to cubic phase transition. Next, the electronic band structure, defects, and

17 optical properties including luminescence are summarized. Finally, the infrared and

Raman vibrational properties are reviewed.

Chapter 3 introduces the commercial processing of zinc sulfide. Next, the process of chemical vapor deposition is described, and the material is considered in light of issues with homogeneous reactions and porosity. Finally, heat treatment of CVD ZnS by annealing and hot isostatic pressing is assessed. Issues involving creep, microstructural change, and diffusion in HIPing are addressed.

Chapter 4 is a bridge chapter, restating the research goals in light of the literature just reviewed. It shows the importance of the various aspects of ZnS considered for the first part of the dissertation.

Chapter 5 details the samples and experiments conducted for this dissertation. It closely describes sample preparation and heat treatments including special experimental series designed to answer specific technical questions. This chapter also gives an overview of the types of characterization techniques employed and their standard settings. Microscopy, diffraction, spectroscopy, and mechanical and physical testing are all employed to understand the nature of CVD ZnS both before and after treatment.

Chapter 6 is the first of two chapters on the results of the characterization. This chapter reviews the results from the optical microscopy, scanning electron microscopy, and transmission electron microscopy. Then the crystallographic information obtained by x-ray diffraction and electron diffraction is presented. Finally, physical and chemical properties of density, fracture strength, and chemical composition are presented.

18

Chapter 7 presents the characterization of the optical properties. Results of some photoluminescence measurements are summarized. Then the bulk of the chapter is spent describing the transmission measurements. Many types of ZnS are analyzed, including single crystals, polycrystalline CVD, hot-pressed ZnS, annealed CVD ZnS, and HIP’d

CVD ZnS with and without metals. A detailed discussion follows focusing on issues of extrinsic impurity absorptions, band edge absorption, scattering, and free carrier absorption. A model for explaining scattering behavior in CVD ZnS is offered.

Finally, chapter 8 collates the discussion and implications and offers some conclusions. A thumbnail is offered on the nature of CVD ZnS and its variants the red

ZnS and elemental ZnS. Then the recrystallization with HIPing is described in terms of the temperature, pressure, and metal promoter components of the process. A few suggestions for future work and final thoughts are offered at the end.

A number of appendices have been provided with accessory information which may be useful and interesting to certain readers. This includes additional comments on crystallography, electronic structure, point defects, thermodynamics, and materials property data. Extended discussions are offered on the details of the x-ray diffraction analyses and the fracture strength analyses.

1.2 History of Infrared Materials and Zinc sulfide

The history of zinc sulfide as an infrared material has been recently summarized by

Harris [1] for U.S. development and by McCloy [2] for international development. A very brief history is presented below that describes the initial motivation for the

19 development of chemical vapor deposited ZnS as an infrared material as well as the prior art in its commercial manufacture.

Zinc sulfide has significant advantages over germanium, the thermal imaging material available prior to the development of zinc sulfide bulk optical material.

Germanium transmits light from the short-wave infrared through the long-wave infrared, about 2 μm to 11 μm.

Germanium does not transmit in the visible range (0.4 to 0.7 μm), however, and thus alignment of optical systems is more difficult and shared apertures with visible cameras are not possible. Also, germanium behaves as a metal and so does not allow the transmission of radar frequencies due to its small skin depth and high dielectric constant, which make it a good shielding material but not a good radar transmitting material. Finally, germanium does not perform well as a transmitting material at elevated temperatures, as it has a very large thermo-optic coefficient (change of index of refraction with temperature) and has prohibitively high free carrier absorption above about 100 °C. All of these properties limited the application of germanium to thermal imaging to either ground-based systems or very low speed aircraft systems.

Other materials had been developed in the mid-1960s which sought to replace or augment germanium. Notable among these were the chalcogenide glasses [3]. Around

1964, glasses known as chalcogenides were developed by Texas Instruments and others.

Chalcogens are elements in the period table group VI excluding oxygen, but especially sulfur, selenium, and tellurium. These elements tend to readily form glasses due to their ring or chain structures.

20

The simplest chalcogenide was arsenic trisulfide, but many other chalcogenides were produced based on germanium. Chief among these were the germanium-arsenicsulfide (e.g. Amorphous Materials Inc.’s AMTIR-1; Texas Instruments’ TI-1120) and germanium-antimony-sulfide (e.g. Texas Instruments’ TI-1173; Amorphous Materials,

Inc.’s AMTIR-3) glasses. Selenium or tellurium could be substituted for sulfur if farther infrared transmission was desired and weaker mechanical properties were acceptable[4].

All of the germanium-based chalcogenide glasses suffered to an extent from germanium’s tendency to exhibit free carrier absorption at elevated temperatures, so their use was still limited. Additionally, as glasses they tended to be rather weak and have low thermal conductivity, making them more susceptible to thermal shock. Transmission of various candidate long-wave infrared (thermal imaging infrared) is shown in Figure 1-1.

Figure 1-1: Transmission of Candidate Longwave infrared Window Materials

It is apparent from examination of this that Ge is not suitable for applications requiring visible and IR transmission. Ge, As

2

S

3

, and AMTIR-1 all suffer from free carrier absorption at temperatures of a few hundred degrees Celsius. ZnSe is broadband transmitting but also weaker mechanically than ZnS.

21

A material was desired that exhibited broadband transmission from the visible long into the infrared, did not exhibit the moisture problems suffered by alkali halides, and maintained its properties to at least 300 °C. Other semiconductors were surveyed to see if they could provide this transparency. On the short wavelength side of the transparency region, materials with a small bandgap suffered from absorption in the visible and excessive free carrier absorption at elevated temperatures. On the long wavelength side, transmission into longer infrared wavelengths is enabled by a low frequency longitudinal optical phonon frequency allowed the multi-phonon absorption to begin at longer wavelengths. When these transparency requirements were considered together with erosion durability needed for an external structural window, zinc sulfide appeared as a leading candidate (see Figure 1-2). While other materials such as ZnSe transmit further into the infrared, they do so at the expense of mechanical strength. Thus it became clear from an early stage that ZnS would be an important infrared window material.

22

293 nm

(UV-B)

Weaker mechanically

Multi-phonon absorption due to lattice vibrations

Electronic absorption in VIS-

NIR from small Eg

Free carrier absorption due to bandgap transitions

Figure 1-2: Band-gap energy versus LO mode frequency for LWIR window materials

Band gap energy determines free carrier absorption and visible transmission. LO frequency influences the infrared transmission cutoff since it is related to the Restrahlen resonance. This shows that ZnS is preferred for any applications requiring transmission in the ultraviolet. Materials requiring only visible through infrared transmission could use ZnTe, ZnSe, or CdS. ZnSe and especially ZnTe are very weak mechanically for the same reasons they transmit well in the infrared – weak bonds. CdS has hexagonal structure creates birefringence and excessive scattering in polycrystalline solids.

1.3 Etymology

Zinc sulfide is well known as the primary ore of zinc. The common names for the cubic form of ZnS all come from its superficial resemblance to galena (lead sulfide, PbS) but ZnS does not yield any metal when smelted. It was therefore called “blende” or

“zincblende,” from the German blenden, “to deceive” or “to blind” or “sphalerite” from the Greek σφαλερός (sphaleros) “deceptive, treacherous” [5]. A special white, transparent, or colorless variety of sphalerite from Franklin, New Jersey, and Nordmark,

Sweden is called cleiophane, which is nearly pure ZnS with only traces of cadmium.

23

Mineral sphalerite tends to have a large component of iron and manganese, and some specimens are very black, being called “black jack.” Mineral cleiophane and sphalerite have different colored fluorescence under short-wave and long-wave ultraviolet light, and are of interest to the mineral collector.

Wurtzite is a less common hexagonal form of ZnS, named after the French chemist Charles-Adolphe Wurtz (1817-1884) by C. Friedel when first identified from a

Bolivian silver mine [5]. Mineral hexagonal zinc sulfide containing significant amounts of cadmium is known as pribramite. Hexagonal zinc oxysulfide has been called voltzite or voltzine, though this term has been used to describe a lead oxysulfide as well.

1.4 Summary

Zinc sulfide is a material which has undergone extensive material tailoring. The defense industry continues to drive the efforts to improve the mechanical properties of the bulk material. None of the innovations to date have produced a cost-effective, durable, and multispectral polycrystalline zinc sulfide material. Thus there remains a niche for development in this area, as well as a need for more fundamental scientific understanding of the relevant processing-properties relationship.

24

2 Physics and Chemistry of Zinc Sulfide

This chapter sets the background for all the relevant physics and chemistry of

ZnS, supplemented with additional details in the appendices. It begins with the discussion of the chemistry of ZnS and its relation to other metal sulfides and zinc chalcogenides. Next the crystallography of ZnS is reviewed, including details on its polytypes and the cubic to hexagonal phase transformation. Then the electronic structure is reviewed, including the band structure and intrinsic defects. Luminescence phenomena in ZnS are reviewed including the importance of transition metals, oxygen, and excitons.

Data on the dielectric constant and refractive index of ZnS are summarized, and then the vibrational structure of ZnS is reviewed including both infrared and Raman active phonon modes. The chapter ends with some discussion of the importance of the hydrogen impurity in chemical vapor deposited ZnS and its role in infrared absorption.

2.1 Chemistry of ZnS

Zinc sulfide (ZnS) occurs naturally in the common mineral sphalerite, though the ore form typically contains a large amount of iron, manganese, or cadmium impurities.

Zinc sulfide, whose phase diagram can be found in [6], occurs in two polymorph crystal structures [7]. The first is sphalerite (structure 3C, β FCC phase), also known as zincblende, which is the isotropic cubic structure and the most important structure for optical materials. The second is wurtzite (structure 2H or α HCP phase) which is the anisotropic hexagonal structure and the more high temperature form. Wurtzite single crystals can be formed by quenching from high temperatures above 1020 °C, though they frequently have a large volume fraction of stacking faults. Rhombohedral phases have

25 also been reported, but are most common as higher period polytypes [8, 9]. Two other forms of the zinc sulfide structure exist only under high pressure situations. The rocksalt structure forms from sphalerite or wurtzite at pressures of 12-65 GPa [10, 11], which at higher pressure forms a distorted orthorhombic structure [12].

The phase transition requires rearrangement of ions in the lattice, and the ions must have sufficient mobility to rearrange. This may occur at higher or lower temperatures than what is normally stated as the thermodynamic transition temperatures

[13]. The low temperature cubic phase invariably contains some domains of hexagonal material. Deformed sphalerite specimens with high dislocation density transform to wurtzite at higher temperatures when reheated due to reduced ion mobility. 1020 °C can be seen as the temperature where the ions have vibrational energy sufficient to rearrange the whole crystal when it is not overly disordered. In other words, sphalerite may thermodynamically be able to transform to wurtzite at lower temperatures locally, but kinetically the process is relatively slow until about 1000 °C where there is sufficient thermal energy for rapid rearrangement.

Various metal impurity atoms have been said to stabilize either the cubic or hexagonal phases of ZnS, but one main effect of these metal impurities is to change the solubility of oxygen. This in turn has the larger effect of changing interplanar distances and stabilizing a particular crystallographic polymorph. Of course, the impurity content directly affect the lattice constant, and the relation of lattice constant to concentration of common impurities has been related by Barton and Skinner [14]. Preferential

26 stabilization of a particular lattice requires impurity density (of any kind) ≥0.01 atomic percent [15].

Iron and manganese are the two main impurities in mineral sphalerite. Phase diagram studies of the FeS-ZnS and MnS-ZnS systems have shown that increasing amounts of iron or manganese lower the transformation temperature for hexagonal to cubic transition [16]. Where multiple polytypes are found together in natural ZnS, they have differing concentrations of iron, and degree of hexagonality of polytypes correlate with iron content across source localities [17]. Solid solutions of up to 16 atomic % Fe substituting for Zn have shown increased lattice constant, increased hardness, and decreased infrared transmission [18]. Additions of iron, manganese, and cadmium lower the phase transition temperature to the hexagonal phase.

Additions of Ag and Cu in ZnS have been shown to increase the transition rate to the cubic phase [19]. The proposed mechanism is the exothermic formation of unspecified silver and copper sulfides in the bulk of the powder processed phosphor compact. The energy released on precipitation of Ag-S or Cu-S provides enough energy to produce stable nuclei of the cubic phase of ZnS in wurtzite below the transformation temperature. These cubic ZnS nuclei then grow consuming wurtzite and releasing energy as the more stable phase at low temperature. This energy from the transition serves to further nucleate cubic phase ZnS, resulting in a self-sustaining transformation that proceeds at a rapid rate. A particular sequence of quenching from high temperature followed by annealing was found to be the most effective means of rapid conversion to cubic phase ZnS. Another study on ZnS:Cu suggested that it is the nuclei of Cu-S and

27 not the energy released that starts the phase transformation [20]. This only happens when heating in the presence of H

2

S, whereas heating in air does not induce the transition to cubic, presumably due to the inhibiting effects of oxygen.

Scott and Barnes [21] have claimed that the fugacity of oxygen is not a controlling factor in the sphalerite-wurtzite equilibrium, at least below 900 °C, but rather the fugacity of sulfur. Wurtzite on the average is said to be sulfur-deficient (i.e. zincrich) relative to sphalerite and that the defects are sulfur vacancies, whereas sphalerite is relatively sulfur-rich (i.e. zinc-deficient) and the defects are zinc vacancies, with overall nonstoichiometry estimated at 0.9 atomic percent. Whereas in pure stoichiometric ZnS, the cubic to hexagonal transition is said to occur at 1020 °C, it has been reported that wurtzite can be formed at temperatures as low as 200 °C in a sulfur-deficient environment while sphalerite persists to temperatures above 1240° C in a zinc-deficient environment. Other sources cite the extent of the off-stoichiometry or width of the “line compound” in chalcogenides to be 0.01 to 0.0001 atomic percent [22]. Yet another source states that the maximum amount of excess sulfur in ZnS is 0.035 atomic percent

(8.7 x 10

18

/cm

3

) but only about 0.0035 atomic percent (8.7 x 10

17

/cm

3

) for excess zinc

[15]. Of course these small fractions of off-stoichiometry are stated for thermodynamic equilibrium. Many samples in the literature whose stoichiometry has been reported, such as those made by chemical vapor deposition, were clearly not in chemical equilibrium when they were assessed.

It has been argued that the presence of substitutional oxygen in sulfur sites leads to low-temperature stabilization of the wurtzite structure in ZnS [21]. This is not

28 surprising since the stable form of ZnO at low temperature is wurtzite. The solubility of oxygen in wurtzite ZnS is said to be 0.6 atomic percent [23] while only 0.4 atomic percent in sphalerite [24]. As expected, these numbers differ somewhat between authors and studies (see Appendix E). The effect of adding oxygen to wurtzite ZnS is to shrink the lattice in both the a-axis and the c-axis direction [15]. Dissolved oxygen is said to stabilize the hexagonal polytypes of ZnS by deforming the sphalerite lattice which is predominantly covalently bonded, and adding an additional degree of ionicity from the

ZnO. Adding oxygen to sphalerite, on the other hand, increases the interplanar distances between the closest packed planes (i.e. the (00.1) c-axis in wurtzite which corresponds to the (111) axis in sphalerite).

Kroeger and Dikhoff [23] have pointed out that oxygen acts as a reducing agent for ZnS since it tends to form SO

2

leaving excess Zn in the ZnS. Similarly, sulfur acts as a reducing agent for ZnO. Therefore, as long as both sulfur and oxygen ions are present with zinc, any ZnS and ZnO will remain reduced (i.e. zinc-rich). Stoichiometric compounds of ZnS or ZnO are only obtainable after prolonged treatment with sulfur or oxygen vapor, respectively, in order to remove the other phase.

2.2 Crystallography

Sphalerite and wurtzite are quite similar structures. Both have tetrahedrally coordinated zinc surrounded by four sulfurs and tetrahedrally coordinated sulfurs surrounded by four zincs. These are often shown as tetrahedrons, the stacking of which differs for the two structures. The other ion occupies half of the tetrahedral sites. Both structures can be modeled as close packed spheres with sphalerite being face-centered

29 cubic (FCC) and wurtzite being hexagonal close packed (HCP). Figure 2-1 shows these structures, and their lattice parameters are shown in Appendix E. Considerable variation is found, likely due to unknown amounts of oxygen and metallic impurities. However, the estimated best value for the lattice parameter of sphalerite is 5.4092 angstroms [24].

Bonding in zinc sulfide can be described as partially covalent and partially ionic.

The radius ratios of the two ions is 0.402 for CN=6 (Zn

2+

(CN=6)/S

2-

(CN=6)=

0.74Å/1.84Å) or 0.326 (Zn

2+

(CN=4)/S

2-

(CN=6)= 0.60Å/1.84Å). These values are intermediate between the geometry-limiting configurations described by Pauling for octahedral coordination (CN=6; ratio = 0.414) and tetrahedral coordination (CN=4; ratio

= 0.225) in a strictly ionic solid [25]. The high polarizing power of Zn

2+ and polarizability of S

2-

leads to the ionic character of the bonding. However, ZnS is considerably covalent and its bonding can be[26]. The Madelung constant describing the ratio of Coulomb energy of the Zn-S ion pair in a lattice to that of the isolated pair at the same separation is 1.63806 for sphalerite and 1.64132 for wurtzite [25].

Figure 2-1: Wurtzite (L) and Sphalerite (R) structures

Wurtzite shown with hexagonal basis cell; sphalerite shown with face-centered cubic basis cell. In this figure the dark atoms are Zn and light atoms are S.

Sphalerite does not have a center of symmetry or inversion. Thus the orientations along <111> directions of ZnS

4

tetrahedra are unique. This makes crystals

30 planes }

h

} polar opposites, and

< hkl >

versus

<

h k l

> directions can have different physical and chemical properties. The Zn-S layers can be seen to be a network of permanent dipole moments, which perfectly cancel each other in a perfect crystal.

However, when mechanically distorted, these dipoles give rise to a potential and hence result in piezoelectricity from mechanical tension and compression [27]. Similarly, wurtzite does not have a center of symmetry and it has a polar axis. The permanent dipole moments do not cancel as they do in sphalerite, but result in a single permanent polar axis along the [00.1] c-axis. Hence, wurtzite is not only piezoelectric, but also pyroelectric, developing charge and polarization due to heating and cooling of the crystal

[28].

Besides the two main thermodynamically stable structures of sphalerite and wurtzite, zinc sulfide tends to form a large number of stable interim phases, called

polytypes [9]. Polytypes are crystal structures formed from repeated stacking of twodimensional “modular layers.” They represent one-dimensional disorder compared to the parent structures. Polytypes are distinct from polymorphs, which have well-defined stability regions separated by first order phase transitions. By contrast, polytypes can often exist simultaneously and intergrown in the same crystal and have very similar energy in many cases. Thermodynamically, the main difference among all the polytypes is structural entropy. For zinc sulfide, at least 194 polytypes of the crystal have been identified in three “families” corresponding to the space group symmetry of the structure

[29]. Other important materials exhibiting polytypic structures include silicon carbide and cadmium iodide [30]. For more information on polytypes in ZnS see Appendix A.

31

There is far from agreement in detail as to how the polytypes of ZnS form as intermediate stable or metastable phases between wurtzite and sphalerite. Whether twin faulting is due to deformation, growth, or phase transformation is still debatable and may vary considerably with local impurities, thermodynamic conditions, and material history.

However, most authors are in agreement, either explicitly or tacitly, that the 2H to 3C phase transformation in ZnS occurs by a martensitic type transformation. Only one source consulted [31] claimed that this transformation was reconstructive.

2.2.1 Phase Transformation and Twinning

A displacive (martensitic) transformation has no structural change in the first coordination sphere, no bonds are broken, and requires no diffusion [32]. The reduction in free energy in going from the high temperature phase to the low temperature phase is accounted for by changes in the second and higher coordination sphere resulting in a denser structure with lower specific volume. Transformations of this type usually result in twinning within the martensite phase. Commonly encountered martensitic transformations in ceramics include cubic to tetragonal BaTiO

3

and tetragonal to monoclinic ZrO

2

.

Martensitic phase transformations can be driven by two different forces [33].

Like all phase transformations, the reduction of bulk free energy per unit volume as the parent phase transforms is the primary driving force. However, since the new phase induces a shape change, there are shear and dilational stress terms in the driving force as well. Therefore, martensitic transformations can be induced by applied stress, as is seen with ZnS [34]. Stated another way, a martensitic transformation can be seen as a

32 spontaneous plastic deformation (i.e. shape change) in response to a bulk free energy change and/ or an applied mechanical force.

Since it becomes critical in CVD ZnS processing, the effect of mechanical stresses on the phase transformation of ZnS is further considered. Many have argued that the transformation to a twinned sphalerite structure is accomplished by deformation twinning, either due to mechanical stress or just local thermal strains during growth [35,

36]. Recently, this transformation has been demonstrated on wurtzite nanobelts which begin taking on the sphalerite phase at pressures of 6.8 GPa and are converted completely to the cubic phase at 11.4 GPa as shown by x-ray diffraction [37].

Deformation twinning without any other mechanism has been demonstrated in sphalerite crystals using mechanical indentation with translational shear of 0.4084 times the lattice parameter, with the (111) twin plane, the (110) shear plane, and the [112] shear direction [38]. Twinning set in about the same fraction of the melting point as did slip. Some theories of transformation require a recrystallization and dislocation motion involving ordering of the crystal through dislocation glide at high temperature [36]. This passage of dislocations and its effect on the transformation of the 2H to the 3C phase has been observed directly [39].

Deformation faults were shown to nucleate the 2H to 3C phase transformation at low temperatures (400-750 °C) resulting in a twinned cubic crystal with regions of perfect cubic order with sizes corresponding to initial fault separation distance [27]. If few layers exist between faults, cubic regions grow together to form large domains. If more layers intervene, cubic regions grow together to form “microtwins” at twin

33 boundaries if the orientation is correct, or are separated by a thin hexagonal lamella or fault. This microtwinning has been said to give rise to “false hexagonality” in x-ray diffraction studies [34].

In a few cases it has been observed that application of pressure alone can perfect the cubic crystal, removing twin boundaries and stacking faults. In one case, the transformation was shown to be permanent, due to plastic deformation by the movement of partial dislocations with Burgers vector

(1/6) 112

>

[34]. In another case the effect on stacking faults was elastic and reversible with application of uniaxial pressure [40]. Also, it has been observed that crushing, used to prepare samples for powder x-ray diffraction, or even sieving through a fine metal screen has the effect of reducing the amount of hexagonal phase in a sample [19].

Other observations of twinned sphalerite bear mention. It is not surprising that twinning on {111} ZnS sphalerite planes is so common, since the stacking fault energy is

<6 mJ/m

2

(or <5 meV/atom), considerably smaller than for any other II-VI or III-VI sphalerite structure compound [41]. Twins in mineral ZnS have been observed in TEM to vary from a few angstroms to 100 nm in width and are organized into “blocks” where the {111} slip planes meet with other {111} slip planes [36]. The study of synthetic wurtzite nanobelts has resulted in an assessment of two ways of converting wurtzite to sphalerite, one by the transformation of three layers of wurtzite into a twinned cubic structure, the other involving the transformation of four layers of wurtzite resulting in a stacking fault [37].

34

The effect of impurities on {111} twinning in sphalerite has recently been considered in detail. Chemistry of twin boundaries of sphalerite crystals from Kosovo was studied using TEM, electron energy loss spectroscopy (EELS), and energydispersive spectroscopy (EDS) [42]. Twinning was observed at regular intervals ranging from a few nanometers to several millimeters, and said to be due to growth faults since the crystals are tabular. Using difference spectra to differentiate the small signal of the boundary from that of the surrounding grains, it was found that the twin boundaries were preferentially depleted of sulfur and enriched in zinc, oxygen, manganese, iron, and copper. The depletion of sulfur at the twin boundary along with enrichment in oxygen is consistent with the tendency of hexagonal stacking layers (i.e. the local twin boundary) to form under these conditions [21].

Electronically, twins can have unique behavior. The stacking fault (twin) layer is effectively hexagonal wurtzite. Since hexagonal ZnS has a larger electronic bandgap than cubic ZnS, the cubic region separated by stacking fault becomes a quantum well for electrons[43]. Additionally, since wurtzite is piezoelectric and polar, a static charge will build up on the stacking fault with a sign reversal between the interior and the surface of the fault. Thus the semiconductor nature of ZnS combined with simultaneously occurring structures results in what can only be defined as a semiconductor heterostructure with all of the associated band structure implications. It is to these matters of electronic structure that we now turn.

35

2.3 Electronic Structure

Zinc sulfide is a direct gap II-VI semiconductor with a room temperature bandgap of about 3.8 electron volts. The band structure of the hexagonal phase can be considered as a perturbation of the cubic crystal field. The reduction in symmetry in the hexagonal structure results in additional valence and conduction bands due to the removal of degeneracy (see Appendix B).

Intrinsic zinc sulfide has an electrical resistivity of 10

14

ohm-cm[44], and at very high resistivities, above about 10

10

ohm-cm, zinc sulfide behaves as an insulator.

However, zinc sulfide can be n-type doped with group III elements like aluminum and gallium to provide resistivities as low as 10

2

ohm-cm [45]. The latter method has been investigated for the production of blue light emitting diodes (LEDs) of the Metal-

Insulator-Semiconductor (MIS) type [44]. The luminescent properties of zinc sulfide are highly dependent on controlled doping and removal of impurities.

The band structure of the two forms of ZnS and the corresponding density of states as calculated by the ab initio quantum mechanical Vienna Ab Initio Simulation

Package (VASP) is shown in Figure 2-2. The density of states shown can also be understood in the context of a Linear Combination of Atomic Orbitals (LCAO) model.

The lowest valence band is composed primarily of the S-3s electrons, the middle valance bands have primarily Zn-3d character, and the upper valance bands have S-3p character.

Conduction bands are derived from the zinc 4s electrons.

The difference in band structures between sphalerite and wurtzite can be appreciated by considering the two types of valence band splitting. Spin orbit splitting in

36 sphalerite results in a lower energy split-off band and two hole bands (light hole and heavy hole) which are degenerate at the gamma point in the Brillouin zone. Crystal field splitting removes the degeneracy in the hole bands and results in three valence bands. A schematic of the splitting and its effect on the character of the bands is shown in

Appendix E along with the simplified bands structures around the gamma point, the total angular momenta, and the irreducible representations of the symmetry.

Figure 2-2: Energy band structure and density of states for sphalerite and wurtzite ZnS

As calculated by MedeA VASP using the generalized gradient approximation (GGA) which always underestimates the band gap. Colors in the density of states refer to the orbital quantum number (yellow = s, red = p, blue = d).

2.3.1 Electronic Defects and Luminescence

Luminescence-based analytical techniques are very sensitive to defects and so can provide experimental evidence of electronic effects in both pure and doped ZnS. There are several kinds of important electronic defects in semiconductors like ZnS. Native defects in ZnS include zinc vacancies (V

Zn

) and interstitials (Zn i

), sulfur vacancies (V

S

) and interstitials (S i

), and antisites such as a zinc atom in a sulfur lattice site (Zn

S

).

37

Isoelectronic defects are substitutional defects with the same valence as the atom being replaced. An important impurity isoelectronic defect which is inevitably present in ZnS is oxygen on a sulfur lattice site (O

S

).

II-VI and III-V semiconductors almost invariably form off-stoichiometry, causing electronic defects in the energy gap [46]. As stated previously, the estimated extent of the off-stoichiometry in ZnS, at least in natural sphalerites and wurtzites, is 0.9 atomic percent [21]. In materials with a Zn-rich stoichiometry, having more Zn

2+

ions and fewer

S

2-

ions (and hence empty sulfur lattice sites), sulfur vacancies (V

S

˙˙) or zinc interstitials

(Zn i

˙˙) will create donor levels close to the conduction band. Similarly, materials with Srich stoichiometry will have zinc vacancies (V

Zn

′′) or sulfur interstitials (S i

′′) which will act as acceptors with defect levels close to the valence band. The tendency toward one or the other of these conditions will depend on the partial pressures of the two components.

Equilibrium concentrations of the intrinsic defects will depend on the temperature (i.e. of deposition or subsequent annealing) and activation energy for formation. Growth at high sulfur vapor pressure favors V

Zn

′ and Zn i

˙˙, while growth at large excess zinc vapor pressure favors V

Zn

′′, Zn i x

, V

S x

and Zn i

˙˙ native defects.

ZnS has a thermodynamic tendency to be Zn-rich and thus be an n-type semiconductor due to the relative magnitude of the reaction constants of zinc and sulfur condensation and evaporation [47]. This has been determined by assuming Schottky defects and intrinsic defects and solving for the equilibrium concentration of charged defects versus partial pressures of zinc and sulfur (see Appendix E). The driving force for Zn-richness due to electronic defect equilibrium is in addition to that already

38 mentioned, relating to the chemical tendency for Zn-richness due to the presence of oxygen. There has been considerable disagreement in the literature as to which point defects are responsible for the n-type conductivity in ZnS, depending on whether the assumptions emphasize Schottky or Frenkel disorder.

Some argue that the intrinsic donors come from primarily interstitial Zn atoms and a relatively smaller number of sulfur vacancies [48]. Sulfur vacancy sites when available then tend to be filled by oxygen, especially after heat treatment. The presence of interstitial zinc is supported by self-diffusion experiments and electrical conductivity experiments from annealing in zinc vapor [15]. Researchers who focus on the interstitial zinc defect tend to see Frenkel disorder as particularly important.

The presence of zinc vacancies, even in zinc-rich sphalerites, is supported by emissions from ZnS annealed in sulfur vapor [15]. The only vacancy point defect with optical transitions in the visible part of the spectrum is V

Zn

′, which has a transition from the conduction band to the acceptor level in the blue (~2.7 eV or 460-470nm at 77 K).

Transitions to and from V

S

˙˙, V

S

˙, and V

Zn

′′ all lie in the infrared [15]. Infrared emission bands and their assignments include 820-850nm (V

S

˙˙, V

S

˙, O i

), 950-1100nm (V

Zn

′′,

V

Zn

′), 1200-1300nm (V

S

˙˙, V

S

˙), 1400-1600nm (V

Zn

′′), and 1900-2950nm (V

Zn

′, O i

).

Others see this sulfur vacancy as the primary donor defect in ZnS. Sulfur vacancies, though considered by some to be rare in intrinsic material, are created easily with irradiation, creating the so-called “F

+

center,” which is a charged sulfur vacancy

(V

S

) [43, 49]. Equilibrium analyses considering the Schottky defect as the controlling

39 one in ZnS show that singly charged sulfur vacancies (V

S

) are dominant [50]. High sulfur pressures are therefore necessary for making stoichiometric material.

Liquid nitrogen temperature absorption measurements in neutron-irradiated and

Zn-vapor treated ZnS have identified a number of intrinsic defect bands thought to be related to the sulfur vacancy [51]. These authors doubt the importance of a zinc interstitial since there is evidence that optical effects of zinc Frenkel pairs (i.e Zn i

and

V

Zn

) produced by irradiation are annealed away at room temperature. Very high energy absorptions near the band edge at 3.5 eV (354.7 mm) and 3.2 eV (388.0 nm) are believed to be due to an exciton bound to a neutral donor F center (V

S x

) and an exciton bound to an ionized donor F

+

center (V

S

), respectively. Absorptions at 1.75 eV (709.4 nm) and

2.7 eV (459.8 nm) are thought to be due to hole transitions. Absorptions at 2.3 eV (539.8 nm) and 2.9 eV (428.1 nm), observed by several researchers, are though to be due to the

F

+

center (V

S

) itself. The latter absorption was observed by Lewis et al. [52] at 80 K for

CVD ZnS, in addition to another absorption in very highly colored material at 2.5 eV

(497 nm) which may be related to a zinc hydride defect. The 2.9 eV absorption was also observed at room temperature by Collins et al. [53] in CVD ZnS grown at very high temperatures (950 °C) and the yellow color of this material was said to be due to sulfur vacancies.

Excitons are electron-hole pairs bound by Coulombic forces. Energetically, exciton levels are located just below the conduction band edge, but exist only near the

Brillouin zone center (gamma point) because the electron-hole pair motion is coupled and must conserve momentum. Excitons in ZnS are of the Wannier type, meaning that their

40 orbital radius (i.e. the separation distance of the electron and hole) is much greater than the lattice spacing. Excitons have a binding energy (ground state energy) that keeps the electron and hole separated and is related to the dielectric strength of the medium in which they are present. The binding energy of an exciton determines whether it will be observable at a given temperatures, since thermal energy greater than the binding energy will dissociate the exciton.

2.3.2 Isoelectronic Oxygen

Oxygen is thought to be ubiquitous in as-prepared ZnS which has not been treated by annealing in sulfur vapors. Morozova et al. [15] claim that oxygen is present in ZnS in concentrations greater than 10

18

/cm

3

but more frequently 10

19 to 10

20

/cm

3

. It has been suggested that substitutional solid solutions ZnS·O

S

only form when the material is grown with excess zinc, since the V

S

tend to be filled with oxygen to form O

S

, and the

Zn i

lattice distortion is compensated for by the lattice contraction due to O

S

[54]. In meltgrown crystals, oxygen present in excess of the solubility limit has been shown to segregate into layers parallel to the

{111} planes which have compositional fluctuations with respect to oxygen and a ZnO·S phase [15].

In ZnS, the conduction band can be viewed as consisting of monovalent zinc ions

(Zn

+

) since Zn

2+

+ e

= Zn

+

[23]. A larger Zn

+ ion will distort the lattice and hence will prefer to be associated with a smaller anion like oxygen in a sulfur site. Thus one can think of the isoelectronic defect of oxygen as an electron trap since it captures an electron because of short range Coulomb potential from the difference in electronegativity between the sulfur and oxygen. Because isoelectronic oxygens already have electrons

41 already associated, they tend to capture holes leading to a bound exciton[15]. The character of the luminescence for an isoelectronic center is very different than that for donor-acceptor-pairs since the latter involve long-range Coulombic forces.

Studies of the effects of high pressure on cathodoluminescent spectra have lent insight into changes that occur during hot-isostatic pressing of CVD ZnS. Diffusion of oxygen, which is suspected to congregate at the grain boundaries probably in sulfur sites, is much faster than interstitial zinc diffusion at high pressures [55]. Changes in emission spectra of Zn-rich CVD ZnS after pressure treatment are thought to be due to in-diffusion of oxygen and out-diffusion of interstitial zinc [55]. Undissolved oxygen which had previously accumulated at stacking faults (twin boundaries), where it was optically inactive, moves into the lattice during dislocation motion, where it creates an optically active defect center [15].

ZnO has been known to precipitate out, both in ZnS [56], forming ZnO·S, and in

ZnSe [57], forming ZnO·Se. Absorption spectra of these CVD samples shows characteristic absorption bands in the 5 to 7μm range, which are often masked by the larger 6 μm absorption, currently thought to be due to Zn-H.

2.4 Vibrational structure

The ab initio computed phonon density of states and phonon dispersion as shown for cubic ZnS in Figure 2-3 agrees fairly well with that derived by neutron diffraction

[58]. The computations were done using the MedeA platform with VASP and PHONON packages.

42

The infrared absorption spectrum of CVD ZnS has been investigated in detail due to its usefulness as an infrared window in the 8-14 μm wavelength atmospheric transparency region. Early in its development as a window, the absorption coefficient was measured by laser calorimetry at available wavelengths (2.7 μm, 3.8 μm, 5.25 μm,

9.27 μm, and 10.6 μm) and also deduced from emittance measurements (8 μm, 9 μm, and

10 μm) [59]. More recently, clear grade (HIP’d CVD ZnS) has been measured at Johns

Hopkins Applied Physics Laboratory by broadband Fourier transform infrared (FTIR) spectrometry and laser transmissometry (at 0.633 μm, 3.39 μm, and 10.6 μm) up to 500

°C [60]. A corresponding semi-empirical multi-phonon absorption model has been derived in Hahn et al. [61]. Hemispherical total emittance versus temperature and normal spectral emittance versus wavelength is compiled in Touloukian [62], clearly illustrating the Restrahlen band or longitudinal optical phonon resonance around 30 μm.

Figure 2-3: Phonon density of states for ZnS and calculated dispersion

43

2.4.1 Infrared and Raman active modes

The Brillouin zone critical point phonons for ZnS have been reviewed. Selection rules in sphalerite structures have been elicited using group theory [63, 64]. For phonon assignment in Raman and infrared absorption spectra, it has become common to consider

“characteristic” zone edge phonons rather than critical point phonons. This treatment assumes that for processes involving ≥3 phonons, selection rules (determining symmetry points for allowed transitions) can be ignored. Assignments of spectral features are then simplified as sums and differences of the characteristic phonons [65]. For ZnS, these characteristic zone edge phonons are longitudinal optical (LO) 330 cm

-1

, transverse optical (LO) 295 cm

-1

, longitudinal acoustic (LA) 193 cm

-1

, and transverse acoustic (TA)

83 cm

-1

. These characteristic phonons have been used for calculating the higher-order phonon frequencies observed in the infrared absorption edge of ZnS [66].

The single crystal wurtzite hexagonal phase phonon structure has been probed by absorption spectroscopy from 350 to 775 cm

-1

at room temperature, and the features assigned to sum and difference combinations of two and three characteristic phonons

[67]. Hexagonal and cubic Raman spectra have been compared, showing identical LO phonon frequency (351 cm

-1

) for the two structures and only slightly different TO frequency (276 cm

-1

for cubic versus 273 cm

-1

for hexagonal) [68, 69].

The single crystal cubic sphalerite phase phonon structure has been probed by polarization controlled Raman spectroscopy using both 488 nm and 514.5 nm excitation, and the resulting transition selection rules are discussed [70]. More recently, the Raman scattering of cubic ZnS was extended to isotopic compositions of Zn and S and probed as

44 a function of temperature and pressure at two excitation wavelengths, 514.5 nm and

647.1 nm [71]. It was found that the line width of the transverse optical (TO) phonon varied strongly with isotopic mass and pressure. The intensity of the TO line, which is much weaker than the LO line, is lost to the background at 300 K when excited with the red (647.1 nm) laser due to destructive interference with background phonons. In ZnS nanoparticles, additional transitions ascribable to surface modes of the particles have been reported [72].

2.4.2 Hydrogen Impurity

The effect of hydrogen in many semiconductors has been well-documented, and its amphoteric character (acting sometimes as a donor and sometimes as an acceptor) makes it particularly difficult to study [73, 74]. The bulk of chemical vapor deposited

(CVD) ZnS is grown in the presence of hydrogen in some form, since the primary reaction for creating ZnS gives off hydrogen.

Zn (g) + H

2

S (g) => ZnS (s) + H

2

(g)

The existence of a Zn-H type vibration at 1608 cm

-1

(6.22 μm) was predicted based on similar hydride molecules [75]. In early CVD ZnS it was observed that there was a strong absorption band around 6 μm which was attributed to some kind of zinc hydride

[76]. Later, Lewis et al. [52] described the hydrogen defect in detail, and correlated it to forward visible scattering, color, and the presence of a sulfur vacancy. Conclusively, they measured the evolution of hydrogen from CVD ZnS upon heating using mass spectrometry, correlating the loss of hydrogen to the disappearance of the 6 μm absorption after heating.

45

In the earlier Raytheon CVD work, hydrogen was deliberately added to a deposition run (#ZS-117), resulting in material with more visible scattering than average, and a very deep 6 μm band [76]. In a complementary run (#ZS-118), the normal process gases were supplemented with oxygen, and the material produced was colorless, had almost no 6 μm absorption, and had less scatter than average. Similar results with oxygen had been found for CVD ZnSe by the Raytheon group.

Hydrogen also seems to have other effects in ZnS, specifically with regard to the grain structure. As already indicated, it has been postulated that hydride segregates to the grain boundaries, resulting in luminescence quenching. Additionally, it has been shown that the presence of hydrogen tends to limit the grain size in CVD ZnS [77]. In these experiments, the doping with iodine prevented the incorporation of hydrogen and resulted in larger grained material.

46

3 Commercial Processing of Zinc Sulfide

In this chapter the production of bulk ZnS for infrared windows is discussed.

After an introduction of the relevant thermodynamics, chemical vapor deposition (CVD) of ZnS is summarized, with notes on porosity, color, and oxygen in ZnS. The important processes and mechanisms of heat treatment of ZnS are outlined, including simple annealing and hot isostatic pressing (HIPing). Mechanisms of creep, material deformation, and diffusion are presented with the aim of understanding the microstructural and property changes of CVD ZnS with HIPing.

3.1 Introduction

In order to appreciate the complexities associated with chemical vapor deposition of ZnS and the subsequent heat treatment of the material, one must first examine the characteristics and reactions of the main components of the system (Zn, S, H

2

S, and Ar).

Trace components in the gas system (e.g. H

2

O, H

2

, and O

2

) which provide alternative reaction paths must be considered as well.

It is well known that the presence of hydrogen is a controlling factor for the chemical potential of sulfur, and hydrogen has been widely used in this respect to study the reactions of sulfides in the geochemical field [78], where the reaction is 2H

2

S = 2H

2

+

S

2

. This reaction determines the fugacity (i.e. activity, at low pressure) for sulfur.

Increasing the [H

2

]/[H

2

S] ratio in the gas phase then must lower the partial pressure of sulfur and further prevent the H

2

S from dissociating (see Appendix C).

For ZnS, the reaction of sulfur with zinc to form ZnS, 2Zn + S

2

= 2ZnS, is known as univariant because at constant temperature and pressure it depends only on the activity

47 of sulfu r[79]. The free energy change for the sulfidation of zinc, 2Zn + S

2

= 2ZnS, has been shown by Vaughan [79] to be

Δ

G

°

rxn

,

II

, 25

420

C

Δ

G

°

rxn

,

II

, 420

1200

C

(

kJ

)

(

kJ

)

=

=

Δ

H

Δ

H rxn

,

II

, 25

420

C

rxn

,

II

, 420

1200

C

T

Δ

S rxn

,

II

, 25

420

C

=

T

Δ

S rxn

,

II

, 420

1200

C

537.88

+

0.19T(K)

548.87

+

0.21T(K)

These values agree with those in the thermodynamic properties database in HSC [80].

This database also includes thermodynamic data for the similar formation of ZnO which is plotted for comparison in Figure 3-1. That ZnO has a lower standard state Gibbs energy of formation implies that, other factors being equal, zinc has a preference for oxygen over sulfur.

Figure 3-1: Free energy of formation for 2 moles of ZnS or ZnO from the elements

Two moles was chosen to compare to the published Vaughan equations (see text).

The presence of oxygen in ZnS has been alluded to before in the context of luminescence. Because ZnS forms a solid solution with oxygen, dissolving about 1 mol%O, ZnS-O can be hard to detect, and will not show up as a separated phase in x-ray diffraction [23]. There has been some debate as to whether the weak extrinsic absorption

48 in CVD ZnS around 9.1 μm (1100 cm

-1

) is due to ZnO [59], Si-O [81], or SO

3

2-

[82].

Absorption bands in ZnSe between 850 and 1100 cm

-1 have been attributed to ZnO [57].

Low temperature deposits (≤670 °C) are said to have higher [O] present. Annealing in Zn vapor increases dissolved oxygen and decreases density, while annealing in S vapor reduces dissolved oxygen to 10

18

cm

-3

but does not affect density [15]. Additionally, the

ZnS-O solid solution is said to be noticeable in photoluminescence, as a red shift to the free exciton due to the band gap shrinkage of ZnS from accommodation of oxygen [55].

As already mentioned, Raytheon experimented with deliberate doping of CVD ZnS with oxygen, which produced colorless material with lower visible scatter than material grown using only hydrogen sulfide and zinc. A discussion of the thermodynamics and kinetics of oxygen incorporation in ZnS is provided in Appendix C.

3.1.1 Vapor Phase Equilibrium

Homogeneous decomposition of hydrogen sulfide is severely restricted below 500

°C, and at 800 °C, conversion to the elemental components is only 25% at 110 kPa pressure (825 torr) [83]. Trace amounts of O

2

are known to greatly accelerate the decomposition of H

2

S at high temperature (>~2100 °C) [84]. It has been observed that when H

2

, O

2

, and H

2

S are present in equal volumes, the H

2

S will be oxidized before the

H

2

[85]. However, the reaction mechanisms in the H

2

S-O

2

system are very complex, and at least thirty possible reactions have been identified involving several paths [86, 87].

Sulfur vapor is extremely complex in its constituency, as it exhibits a number of allotropes, the distribution of which varies strongly with temperature (see Appendix C).

At low temperatures, S

8

is the most stable species, with significant fractions of S

7

and S

6

,

49 but above 720 °C, S

2

becomes the most prevalent species [88]. Recently, reaction enthalpies for conversion from S

8 to lower molecular weight forms have been reviewed, comparing experimental and calculated values [89]. S

2

is the biggest mole fraction above

1000 K [90], and individual sulfur atoms are unlikely to be present at temperatures lower than 2000K due to their high enthalphy of formation [89]. However, a different conclusion was reached by Wiedemieir [91] in studying the reaction of ZnS, who stated that sulfur atoms were an important fraction of the equilibrium gas phase above the ZnS reaction from H

2

S and Zn.

Recently, the equilibrium composition of the gas phase above ZnS in the presence of H

2

, H

2

O, and CO

2

has been investigated using a total energy minimization method

[91]. First, the study simulated the equilibrium vapor phase composition above ZnS(s) including all the allotropes of sulfur along with zinc and zinc sulfide. It was found that

Zn and S

2

are the dominant species by several orders of magnitude, and are very close in partial pressures throughout the simulated temperature range (~588 - 2000 K). Next a simulation was done assuming the addition of 10

-3

atm (~1 torr, at 298 K) of H

2

gas, which significantly changed the equilibrium such that the dominant species below 1000

°C are H

2

and H

2

S followed by Zn then HS or S

2

, depending on the temperature. The partial pressure of zinc (p

Zn

) was significantly greater than the partial pressure of sulfur

(p

S2

). Additions of H

2

gas at the 10

-8

atm (~10

-5 torr) level did not significantly alter the equilibrium of the ZnS(s) system alone, but larger concentrations on the order of 1 atm

(~10

3

torr) of H

2

resulted in even more depletion of sulfur gas in the vapor. Similar results were found when adding 10

-8

atm (at 298 K) of H

2

O(g), in that the sulfur was

50 much depleted below about 700 °C, and the major constituents of the vapor in order of concentration were H

2

O, H

2

, SO

2

, Zn, S

2,

and H

2

S with the order differing slightly with temperature. When both H

2

O and H

2

were added at the 10

-3

atm (at 298 K) level, the vapor was nearly all H

2

O and H

2

, with concentrations of H

2

S and Zn being next. Finally, the addition of 10

-8

atm (at 298 K) of H

2

O(g) and H

2 along with 10

-5

atm (~10

-2 torr, at

298 K) of CO

2

resulted in an extremely complex equilibrium consisting of CO

2,

CO, then

Zn and S

2

being the next greatest component and very close in concentration due to sublimation down to about 650 °C when H

2

O(g), H

2

, COS, and H

2

S become important.

In this last case, only at temperatures below 560 °C is there a return to the p

Zn

>> p

S2

.

This illustrates the profound effect of trace impurity gases in vapor equilibria.

Conventional vacuum deposition reactors which do not use careful outgassing techniques can expect to have 10

-8

atm of H

2

O(g) and H

2

from the chamber walls, along with 10

-5 atm of CO

2 in a continuous supply from pumps, furnaces, and graphite crucibles [91]. If the deposition system is dynamic, it is unlikely to be at equilibrium. Defect concentrations will be inhomogeneous and vary spatially in the chamber. Overall, the effect of any H

2

O(g) and H

2

present is to lower the chemical potential of sulfur by forming gases with sulfur, leaving a great excess of zinc in the vapor. Assumptions in this study included a congruently vaporizing ZnS (i.e. Zn and S ratios upon sublimation are equal) and no other solid phases were formed. Already this has limited applicability to the CVD ZnS system, since ZnO and Zn-H species were not considered. However, this study is the type of study which must be performed and compared to experimental mass spectrometer data from actual CVD depositions.

51

3.2 Chemical Vapor Deposition of ZnS

Chemical vapor deposition methods of zinc sulfide can be separated into static and dynamic methods, as well as transport methods and conventional methods[92]. Static methods involve sealing the reactants in a furnace while dynamic methods involve continuously feeding reactants and removing spent gases. Transport chemical vapor deposition requires taking solids of zinc sulfide and combining them with hydrochloric acid vapor to produce zinc chloride and hydrogen sulfide gasses which are held at a high temperature. These are then recombined such that the zinc sulfide solid re-deposits in a bulk solid on substrates at a lower temperature, leaving hydrochloride acid which can be evaporated off. The process works well for growing small single crystals and some polycrystalline material, but is limited in the size of parts which can be grown, and zinc sulfide grown this way tends to have undesirably large grain size.

The standard dynamic chemical vapor deposition process of zinc sulfide as practiced today (see Figure 3-2) involves a liquid zinc metal which combines with hydrogen sulfide gas in a higher temperature deposition area and the growth of ZnS proceeds by the reaction

Zn (g) + H

2

S (g) => ZnS (s) + H

2

(g) .

Hydrogen gas is given off in the process, some of which ends up being incorporated in the solid zinc sulfide lattice, especially at lower deposition temperatures.

ZnS CVD growth is carried out in a hot-wall CVD reactor, with a heated retort of zinc metal, hydrogen sulfide gas injectors, and argon carrier gas for the vapors [93].

Recently complex models of the transport phenomena have been simulated, showing

52 dependence of geometries on deposition rates [94, 95]. Despite this, the details of the reaction kinetics, thermodynamics, and transport phenomena are still poorly understood today.

Figure 3-2: CVD ZnS operation of 60” diameter production furnaces

Former Raytheon CVD ZnS operation, showing graphite female mandrels for large domes

3.2.1 Homogeneous and Heterogeneous CVD Reactions

There is considerable debate as to exactly how the ZnS forms. The relative importance of the homogeneous versus heterogeneous reaction in CVD ZnS has never been resolved. CVD processes in general can be completely homogenous, completely heterogeneous catalytic, or in between [96]. Since CVD reactions generally show deposition of material on all internal surfaces not just those which are meant to catalyze the reaction, this suggests that CVD reactions typically involve at least some homogeneous reaction [97].

53

In a homogenous reaction, ZnS forms in the gas phase and either condenses into a powder or diffuses to the substrate to form a film [97]. Higher gas phase concentrations

(supersaturation) will favor self-collisions of gaseous products and powder formation, while low concentrations favor diffusion and film formation. With a cold-wall CVD process, the diffusion to the substrate is known as thermophoresis and is used in producing silica optical waveguide fiber. The CVD ZnS process, however, is a hot-wall

CVD process. In recent modeling of ZnS growth using transition state theory (TST) to estimate the configuration of transition state complexes between H

2

S and ZnS, Sharifi

[98] assumed that ZnS forms in the gas phase then diffuses and deposits on the substrate.

It has been argued that the gas phase reactions in ZnS are undesirable from an optical standpoint, and that heterogeneous reactions and low growth rates are preferred to produce a more cubic material [76].

Others have proposed that heterogeneous reactions are more important (or at least more desirable) for ZnS, where the hydrogen sulfide adsorbs to the graphite surface then breaks apart leaving the sulfur behind where it bonds to the zinc and releases hydrogen

[99]. This process is depicted in Figure 3-3. This mechanism should be more important at lower temperatures, and would be more affected by the stability of the H

2

S and hence the H

2

content. Note that conditions which favor the highest growth rates do not favor the best intrinsic optical quality (i.e. low hexagonality fraction).

The choice of mandrel material could be considered as a means of catalyst of the surface reaction for the first material to be deposited. Early on, alumina, fused silica, and molybdenum were investigated but were found to be too reactive, effecting either the

54 deposited ZnS material or the reusability of the mandrel [92]. Graphite mandrels with graphite release coatings were found to be the most durable and nonreactive and are still commonly used today. Recently, the use of graphite versus tantalum substrates was investigated to determine any differential effect on optical properties, and none was found [100]. Therefore it can be concluded that the initial mandrel material is relatively unimportant to the quality of the ZnS material, and noncatalytic unreactive substrates are preferred.

Figure 3-3: Illustration of the proposed model of ZnS growth

H

2

S adsorbs to the surface and reacts with Zn. Gas phase reactions also produce ZnS which diffuses to the surface. Possible other reactions involve H or H gas phase ZnH

2

2

, O

. Figure courtesy of C. Willingham of CBWTechservices.

2

and H

2

O, or

Deposition temperatures (i.e. hot-wall mandrel temperatures) can range from 300 to 1200 °C under very low total reactor pressures, on the order of 40 torr (~5 kPa or

~0.05 atm). It was found that grain size of the polycrystalline zinc sulfide was highly dependent on deposition temperature, with higher temperature deposition resulting in larger grain size [101]. Standard CVD zinc sulfide varied in visible color from colorless

55 to yellow to orange to red to brown to even black and was generally opaque (i.e. highly scattering) to the eye.

From the very beginning, it was discovered that CVD growth of bulk ZnS was fraught with process sensitivity [102]. The primary measure of the optical quality of

CVD ZnS has always been its visible color and scatter, with several groups having produced transparent, colorless or near-colorless material without post-processing [53].

However, the literature is inconsistent as to the optimal processing conditions for repeatable production of intrinsically high optical quality CVD ZnS. The main variables affecting the properties of the material have been established, although the physics of the interactions are poorly understood. The deposition temperature, which is the substrate or mandrel temperature of the hot wall reactor, has been varied successfully from as low as

450 °C to as high as 950 °C.

Reactant molar ratios in the range of Zn/H

2

S of 0.1 (very sulfur rich) all the way to 4.5 (very zinc rich) have been tried with generally acceptable long-wave infrared (8-14

μm wavelength) properties for all cases but with highly variable visible to mid-wave infrared (0.4-5 μm wavelength). Reactant flow rates (metered rates of argon carriers for

Zn vapor and H

2

S) seem to have more effect on the deposition thickness uniformity than specifically on the visible transmission. It was shown by Raytheon [103] that areas where the sulfur partial pressure are higher, such as near the H

2

S inlet jets, results in ZnS material which is higher in sulfur and has poorer optical quality (i.e. more scatter). This might suggest that vapor far from the H

2

S injectors would have less sulfur, be more zinc rich, and thus potentially have higher optical quality.

56

“Elemental” zinc sulfide was developed at Raytheon beginning in the mid-1990s because of the concerns about storing large quantities of hydrogen sulfide gas. The process differed from previous CVD zinc sulfide in that the hydrogen sulfide gas was made in-situ by a gas generator from hydrogen gas and sulfur vapor. The H

2

S gas, probably with some unreacted H

2

, was then delivered to the growth chamber along with the zinc vapor. Surprisingly, given deposition conditions very similar to the standard

CVD process practiced commercially, the elemental ZnS product resulted in very pale yellowish but low scatter material. Recently a similar material was reported by a Chinese group grown by separate sources of hydrogen and sulfur [104].

A Russian study on CVD ZnS has investigated the rate-controlling step in the deposition of ZnS from H

2

S over a range of total system pressures (1-72 Torr) and temperatures (550 – 850 °C) Zn-rich reactant stoichiometry [105]. The total rate constant

(k t

), defined by the mass transfer in the gas phase (k g

) and the first-order chemical reaction (k r

) at the substrate were studied. ZnS deposit thickness as a function of distance from the injectors was used to compute the gas phase diffusion coefficient of the limiting species (H

2

S) and the effective mass transport rate constant (k g

) of the heterogeneous reaction. A first order reaction rate constant (k r

) with respect to H

2

S was assumed for the chemical reaction.

The study found that the mass transfer rate constant (k g

) strongly increased with temperature in ZnS, suggesting large activation energy for mass transfer. This is in marked contrast to CVD ZnSe where k g was found to be independent of temperature (i.e. non-activated) [105]. At 750 °C, the total rate constant was (k t

) found to be nearly

57 independent of system pressure from 10-70 Torr, unlike the situation of ZnSe at 645° C.

This pressure independence of the growth rate CVD ZnS suggests that growth is kinetically limited and not boundary-layer diffusion limited as in ZnSe. This chemical control on the growth kinetics of ZnS holds even at temperatures as high as 750 °C for concentrations of the limiting reactant H

2

S gas of 3.5 x 10

-5

mol/L. Growth rates at this temperature over the pressure range studied were 2.3 x 10

3

mm-L/hour-mol and varied less than 25% over this pressure range, well within the error of the measurement of deposition thickness.

In comparing CVD ZnSe and CVD ZnS, the authors state that ZnS deposition is

kinetically controlled over a much larger range of temperatures and reactant dilutions (i.e. total system pressure) than ZnSe. Note that this study focused on deposition rate, concluding that rate was relatively insensitive to total system pressure (which includes argon carrier) and temperature. However, it is likely that optical properties, particularly at short wavelengths, are much more sensitive to these parameters than overall growth rate.

3.2.2 Porosity in CVD ZnS

Recently, an Israeli group has provided a detailed description of the microstructure of CVD ZnS [100]. The authors show that the intrinsic visible-throughmid-wave infrared transparency is improved when going from 650 to 730 °C deposition temperatures (Zn/H

2

S ratios are not specified). Pores were found in decreasing numbers with increased distance away from the mandrel and with increased deposition temperature. Pore size was in the range of 50 to 200 nm, according to scanning electron

58 microscopy (SEM), and varied not with distance from mandrel or deposition temperature, but only with gas flow rate and total system pressure. No significant changes in porosity were noted when using tantalum mandrels versus graphite ones.

These authors suggest that these observations can be explained by a

heterogeneous ZnS reaction model where gases such as H

2

S and H

2

adsorb to the mandrel substrate where ZnS material is deposited around them creating pores [100].

Higher temperatures make the adsorption of gas molecules less likely. Increased temperatures on the growth surface of ZnS versus the mandrel (due to exothermic reactions producing ZnS and lower thermal conductivity of ZnS) are invoked to explain the decrease in porosity at distances away from the mandrel. Since mandrel material did not affect observed porosity, adsorption of the gas was believed to be controlled by the arrival of the gas to the substrate, and so should be influenced only by the pressure and flow rate.

A research group at the IKhVV RAN (Institute of the Chemistry of High-Purity

Substances, Russian Academy of Sciences, Nizhnii Novogorod) where both CVD ZnS and CVD ZnSe are still produced, have published similar studies [106]. CVD ZnS samples were produced under various deposition temperatures, total pressures (of argon carrier gases), and H

2

S concentrations, and density was measured for calculations of porosity [107]. Each of the parameters was only varied by itself while the other two conditions were held constant, so the results may be somewhat misleading. For example, porosity was seen to increase with increasing deposition temperature which is completely opposite to the results presented by the Israelis [100]. However, the role of pressure is

59 clearer, with low total pressures favoring kinetically controlled heterogeneous surface nucleation while higher pressures result in homogenous gas phase reactions. Similarly, kinetics control deposition reactions at low temperatures, mass transport across the boundary layer controls reactions at intermediate temperatures, and thermodynamics generally controls reactions at the highest temperatures [93].

3.2.3 Summary

ZnS is made commercially by low-pressure hot-wall chemical vapor deposition

(CVD) from zinc vapor and hydrogen sulfide gas. Deposition of ZnS probably occurs by a combination of (1) homogeneous gas phase reactions where particles diffuse to the substrate and (2) heterogeneous reactions at the growth surface. The nature of these reactions determines the microstructure, growth rate, and optical quality. The main impurities in CVD ZnS are believed to be oxygen and hydrogen, though the relative quantity and identity of all the gas species in the deposition chamber is not known. The presence of porosity in some CVD ZnS is believed to be related to the adsorption of gas molecules on the growing solid surface and subsequent trapping of gas as material is deposited around them. CVD ZnS thus produced is yellowish and opaque in the visible, requiring additional post-processing to be usable as windows in the visible and nearinfrared.

3.3 Heat Treatment of CVD ZnS

Various heat treatment protocols have been used on CVD zinc sulfide in an attempt to produce clear and colorless material from the visible through the long-wave infrared. These treatments can be separated into those which involved only heating (i.e.

60 annealing) and those which also involved pressure (i.e. hot isostatic pressing).

Additionally, over the years, attempts have been made to understand the impact of furnace atmosphere on material properties such as optical quality and mechanical hardness.

3.3.1 Annealing

Many annealing experiments were done in the early trials of Raytheon[76], and several observations bear repeating. Materials were annealed in air, vacuum, hydrogen, hydrogen sulfide, sulfur, and zinc vapors in an effort to remove the color and reduce the optical scattering in the visible and infrared. Material response to annealing treatments was found to depend on CVD deposition conditions. Type “A” CVD ZnS material had high visible scatter (e.g. run #ZS-39 [102]) was deposited above 750 °C, had no 6 μm absorption, and would not improve upon annealing under any known conditions. Type

“B” material was deposited temperatures < 700 °C (e.g. run #ZS-45 deposited at 550 °C

[102]), had good visible clarity, a large 6 μm absorption, brownish color, and deposition rates of ~0.002 inches per hour.

For annealing studies, samples were place in a quartz tube and the tube was sealed or left open to the atmosphere (in the case of air annealing). In the type “B” materials which improved upon annealing, heat treatment in atmospheres of vacuum, air, zinc, and sulfur all reduced the 6 μm absorption. The best annealing results were with sulfur (at

500 °C for 125 hours) or air (not specified, but either 500 °C for 125 hours or 750 °C for

24 hours)[102]. For these samples, transmission was improved from the visible through the infrared, and the absorption band was completely eliminated. Hydrogen sulfide

61 anneals were attempted later and were found to eliminate the color, but increase the scattering over the starting material [76].

The hexagonal to cubic phase transformation has been studied by annealing powders (starting composition was 20% sphalerite, 80% wurtzite) in vacuum, zinc vapor, or sulfur vapor at various temperatures between 800 and 900 °C [108]. The rate of the transformation was determined by changes in peak heights by x-ray diffraction, monitoring the reduction of the (10.0) and (10.1) wurtzite peaks and the growth of the

(111) sphalerite peak as the material becomes more cubic. Results were modeled using a second-order kinetics expression with terms for fraction unreacted and density of transformed nuclei. Transformations from vacuum and sulfur vapor anneals were said to be controlled by nucleation of new regions of cubic phase. Concentrations of nuclei for phase transformation were found to be independent of temperature. A higher concentration of nuclei was found for annealing in sulfur vapor than in vacuum, explained as being due to adsorption of sulfur on the ZnS surface causing increased transformation rates. Above 0.5 atm of sulfur, the adsorption is saturated and no further increase in transformation rate was achieved.

The rate of transformation in Zn vapor was well-modeled by a first-order rate expression for three-dimensional diffusion in a sphere [108]. The proposed mechanism involved 60 degree rotation of the zinc tetrahedron along the c-axis of wurtzite. This

“diffusion” of three zinc atoms in basal plane of wurtzite, changes the c-axis stacking sequence from ABA to ABC, thus converting the wurtzite structure to sphalerite structure. It can be appreciated that if the levels of excess zinc are too high, this

62 mechanism is not efficient, as shown by the decrease in transformation rate above 0.3 atm of Zn atmosphere.

3.3.2 Hot Isostatic Pressing (HIPing)

Before addressing hot isostatic pressing (HIPing) of CVD ZnS directly, it is useful to examine temperature and pressure parameters in powder hot pressing and sintering. This processing, while very different than HIPing of CVD ZnS, lends significant insight into the effects of temperature and pressure on the phase transformation in ZnS.

3.3.2.1 Hot Pressing and Sintering of ZnS Powders

Various samples of Kodak polycrystalline hot-pressed ZnS (IRTRAN) with average grain size of 1.9 μm were subjected to uniaxial compression strains at temperatures just above and just below the phase transition temperature (~1020 °C)

[109]. (Note that these samples were originally hot-pressed from powders when they were made by Kodak, but this particular study was conducted by (re-) hot pressing these already dense polycrystalline samples.) When samples were hot-pressed above the transformation temperature (1050 °C), there was evidence of superplasticity, an influence on the transformation kinetics, and an exponential increase in the hexagonal phase content of the resulting material. Hot pressing below the transformation temperature

(975 °C) produced no additional hexagonal phase (IRTRAN ZnS starts out with about

5% hexagonal phase). However, dynamic recrystallization and extensive cubic phase twinning took place at this temperature as flow stress from deformation reached a peak value. For the 975 °C samples, true stresses were calculated to be in the range of 90 –

63

140 MPa (13 – 20 ksi) depending on the strain rate used. Note this temperature and pressure condition which produced recrystallization in hot pressing is very similar to the conditions used commercially in hot isostatic pressing of CVD ZnS.

Recent sintering studies of monodispersed submicron ZnS powders also provide information on the effect of temperature on phase transformation in ZnS [110]. Sintering temperatures from 900 to 1250 °C were investigated and x-ray diffraction measurements were made of the sintered bodies. There was no evidence of hexagonal phase peaks in material sintered at 900 °C, and the wurtzite (10.0), (10.1), and (10.3) reflections only appear in material sintered at 1000 °C or higher. ZnS powder sintered at 1250 °C is almost entirely hexagonal phase, showing only a very small diffraction peak of the sphalerite (111) plane. These results suggest that heat treatment temperatures below

1000 °C without pressure should not create additional hexagonal phase in ZnS.

3.3.2.2 Hot Isostatic Pressing of CVD ZnS

The application of hot isostatic pressing (HIPing) to CVD ZnS has been found to increase the visible and near-infrared transmission to levels identical to that achievable from the best single crystal ZnS. Typical temperatures and pressures used are 990 °C at

15 – 30 ksi. As shown above analogous to powder processing, this temperature should prevent formation of hexagonal phase and this pressure should induce dynamic recrystallization and extensive twinning, which is exactly what is observed. Hot isostatic pressed ZnS has been shown to have slightly different property values for Young’s modulus, Poisson’s ratio, and thermal conductivity. The microstructure of HIPd ZnS is significantly different than that of the as-deposited CVD material.

64

Hot isostatic pressing has traditionally been applied in materials processing to remove porosity and increase density of metals and ceramics [111]. Densification can typically be achieved by temperature alone (sintering) or by pressure alone (cold isostatic pressing), but the application of heat and pressure together reduces the necessary level of either by itself. The temperature used for a HIP process is generally at least half the melting point of the material, in order to lower the yield strength and raise the diffusivity of the material thus allowing pore closure and densification. The pressure is generally chosen such that it is greater than the reduced yield point of the material at the process temperature. This allows plastic flow due to high temperature creep processes such as dislocation creep (slip, climb), lattice diffusion (Nabarro-Herring) creep, and grain boundary (Coble) creep.

HIPing causes grain growth and can influence phase transformations due to the high pressure [111]. By itself, the pressures in HIPing, 100 – 200 MPa (~15-30 ksi), are generally not enough to cause phase transformations. However, for the ZnS system as an example, the sphalerite cubic structure is denser than the hexagonal wurtzite structure and so is more thermodynamically favorable at moderately high pressures. The transformation of the cubic to hexagonal phase ZnS near the transformation temperature is impeded by high pressure, since the volume must increase to convert to the hexagonal phase [55].

HIPing is a thermo-mechanical process with both consolidation and heat transfer occurring simultaneously [111]. Constitutive equations, generally applicable to powder consolidation, have been developed which describe the various microscopic mechanisms

65 contributing to densification rate, including power-law creep, boundary diffusion, volume diffusion, and diffusional flow. Microscopic models depend on three variables, pressure, temperature, and relative density, which can be plotted in various ways on densification maps. Macroscopic models, more empirical in nature, better describe the relationship of strain and strain rate to densification and densification rate and so better account for shape change.

Additionally, grain growth may occur faster with HIPing, as impurity diffusion is increased at higher temperatures and grain boundaries may no longer be pinned by precipitates but may migrate. Specific HIPing schedules must consider the densification mechanism and the thermal conductivity so as to prevent thermal cracking, especially in ceramics.

The impact of HIPing on diffusion is difficult to discern, since high temperatures generally increase diffusion, while high pressures generally decrease diffusion. A number of studies have been conducted looking at “self-diffusion” of Zn and S in ZnS as well as a number of dopants and defects. These studies have consisted of calculations as well as measurements. See Appendix C for a review of these diffusion studies and data.

3.3.2.3 Creep Mechanisms

Different plastic deformation and densification mechanisms are operative in ceramic processing depending on temperature and applied stress. Generally speaking, high pressures and low temperatures favor power-law (dislocation) creep mechanisms, whereas low pressures and high temperatures favor diffusion-controlled mechanisms

[25]. On a creep deformation map of temperature versus stress, the temperature is plotted

66 as “homologous temperature,” which is the temperature normalized by the melting point

(T/T m

), and the stress is plotted normalized to the shear modulus of the material (τ/μ).

For a polycrystalline ceramic, high stresses result in dislocation glide. The

Peierls-Nabarro (frictional) stress is the force needed to move a dislocation along the slip plane [25]. Regardless of temperature, the deformation mechanism is dislocation glide above the Peierls’ stress and below the theoretical shear stress. The Peierls’ stress is given by:

τ

f

=

μ exp

2

π

w b

where μ is the shear modulus, w is the dislocation width, and b is the magnitude of the

Burgers vector of the dislocation. For covalent ceramics, w~b, and for cubic ZnS, μ=33

GPa [25]. Therefore for ZnS, the Peierls’ stress is 61.6 MPa (8.9 ksi) which normalized to the shear modulus is 0.0019 or ~10

-3

.

At low stresses, diffusional creep mechanisms are dominant. For polycrystalline ceramics, the mechanism of lattice (or volume) atomic diffusion (Nabarro-Herring creep) is favored at high temperature, while grain boundary diffusion (Coble creep) is favored at lower temperature. Although both mechanisms can be operative at the same time, when the temperature is higher than 40-50% of the melting point lattice diffusion is usually assumed to be dominant. The melting point of ZnS is 1830 °C [92], so the transition point for lattice diffusion in ZnS should be approximately 732 - 915 °C.

Karaksina et al. [112] did a study of recrystallization of CVD ZnS where samples were subjected to high pressure HIP at various temperatures between 810 to 1100 °C.

Grain size change due to recrystallization was modeled as a power-law function of time.

67

The power-law exponent for ZnS recrystallization was found to be 0.1, considered to be a low value for ceramics, presumably due to a high density of structural defects which slow the grain boundary motion. The power-law also contains a prefactor with a temperaturedependent activation energy term.

Grain size was found to increase rapidly after the first hour of HIPing (T ≥ 940

°C) then increase very slowly with time up to 25 hours. The effect of HIP pressure on grain size in the range of 90 – 200 MPa was very minimal, with grain sizes increasing with pressure only about 10% in this range for a 980 °C, 16 hour HIP. The change in grain size (from the original CVD material) was plotted versus temperature in an

Arhennius-type analysis for samples HIP’d at 180 MPa (26 ksi) for 3 hours. A change of slope was found at 985 °C, indicating a recrystallization mechanism with different activation energy. Between 810 and 985 °C, the activation energy for grain size change was 150 kJ/mol, and was attributed to grain boundary diffusion. Between 985 and 1100

°C, the activation energy was 550 kJ/mol, and was attributed to volume diffusion of Zn in

ZnS.

Activation energies for diffusion of Zn in ZnS in the literature vary widely (see

[113] and Appendix C), and both of the activation energies from the recrystallization study cited above are higher than the generally accepted values for Zn interstitial diffusion in ZnS. However, the larger activation energy at the higher temperatures suggests a smaller diffusion coefficient, and lattice diffusion coefficients are smaller than grain boundary diffusion coefficients. Thus the conclusions seem reasonable, but the

68 authors neglect the important role of pressure in plastic deformation, since diffusion will be reduced at high pressures.

The relative role of diffusion versus plastic deformation at these higher temperatures is unclear. Both mechanisms seem to be important, with some authors emphasizing diffusion [114] and others plastic deformation [112]. It still remains to be clarified how much of the recrystallization is due to the pressure (i.e. a deformation mechanism), and how much is due to the temperature alone (i.e. a diffusion mechanism).

Given the fact that the Peierls’ stress for ZnS is much lower than typical HIP pressures, it seems likely that dislocation motion would play a significant role at all temperatures.

3.3.2.4 Effects of HIP on Microstructure and Mechanical Properties

The structure in CVD ZnS material exhibits characteristic features which are typical of bulk materials grown with chemical vapor deposition. The yellow material in transmitted visible light often displays patchy dark regions which appear as spherical features on the order of a few millimeters in diameter. Original nucleation of these structures is thought to be from particles present on the mandrel surface, and growth proceeds from these heterogeneous nucleation sites. These areas protrude farther into the reactant gas stream and so become preferentially built up. At the limit of long deposition times, the growth surface of CVD deposited ZnS is bumpy and nodular [93] (see Figure

3-4). These structures have been variously referred to as “haystacks” [115], “pebbles”

[116], “botryoidal hillocks” [117], “spicular” growth [118], “globules” [118],

“cauliflower,” “rosettes,” “petals,” and even “alligator skin.”

69

Figure 3-4: Nodular growth on the surface of CVD ZnS away from the mandrel

Recently, it was found that this macrostructure influenced material removal rates in CVD ZnS (manufactured by II-VI Infrared) by magneto-rheological finishing (MRF)

[119]. Shape, areal density, and intensity of the “decorated” structure emphasized by

MRF polishing were different for mandrel-side and growth-size CVD ZnS surfaces.

Material on the mandrel side was yellowish and, with polishing, showed distinct circular

“craters” of <1mm diameter which were only visible using interferometry. Material on the growth side was orangish and, with polishing, showed large overlapping “petals” some >5mm diameter which were visible to the unassisted eye as well as to interferometry. These structures are explained as resulting from “cones” of material which increase in diameter from the mandrel side to the growth side, and are probably associated with the anomalous exaggerated grain growth visible in cross-sectioned material.

The surface morphology of vapor grown material has been modeled [120]. It was shown that a flat growth front is unstable in growth conditions that are not purely

70 kinetically-limited. Whenever there is a degree of diffusion-control, where the reactant must be transported to the surface, the concentration gradient in the growth front will result in “domes” at the growth front. A mathematical model was generated which welldescribes this surface behavior from supersaturated reactant gases such as used with CVD

ZnS and the resulting surface morphology described above.

Microstructural images and crystallographic considerations suggest that the preferred growth during the CVD deposition is on {110} and {311}“atomically rough”

(i.e. not close-packed) planes [121], which encourages formation of kinks and dislocations and results in increased rates of nucleation and growth [32]. The resulting

CVD microstructure is composed of a complex hierarchy, starting with nanometer scale lamellae which assemble into domains of which several comprise a grain. Each hierarchical level is separated by stacking faults or other nonequilibrium boundaries.

The microstructure of the HIP’d material typically does not contain all the original hierarchical structure, but rather consists of large perfect single crystallite grains separated by low angle twin boundaries (hexagonal stacking faults). The extent of the recrystallization depends on various factors including as-deposited chemical make-up of the CVD material along with temperature, gas atmosphere, pressure, and time of the HIP process. Grain growth during HIPing in argon is believed to occur via a widening of the columnar grains until grains are equiaxed, with twins appearing roughly parallel to the original columns [101]. Some intermediate microstructures have been observed with

HIP temperatures of 840 – 880 °C and pressures of 180 MPa, where coarse 30 μm grains coexist with fine 3 to 5 μm grains [112]. Fully recrystallized samples from higher

71 temperature HIPs (940 – 985 °C) consist of “mosaics” of twinned {111} textured crystallites. Higher temperature HIP runs of 1030 °C, above the transition temperature to the hexagonal phase, result in euhedral grains with 120 degree sharp tilt boundaries.

These euhedral grains, by definition, are evidence of crystal growth unconstrained by adjacent crystals with crystallites, and hence are bounded by characteristic facets.

The relationship between mechanical properties and microstructure in both CVD

ZnS and HIP CVD ZnS has been closely examined. Grains of as-deposited CVD ZnS are columnar, with the long axis of the grain oriented perpendicular to the mandrel[122].

Grain size generally scales with deposition temperature for un-HIP’d material. Average grain sizes (columnar cross-sections) ranging from 2μm in the lowest temperature deposits up to over 200μm for high temperature deposits [117].

HIPing of ZnS has been shown to dramatically affect mechanical properties.

HIPing reduces hardness and strength and increases fracture toughness relative to starting

CVD material. Direct comparison is not always straightforward, however, since commercial CVD ZnS has a 4 - 8 μm average grain size and commercial HIP CVD ZnS has a 75 – 150 μm average grain size. Compared to CVD ZnS of comparable grain size, hardness in HIP’d ZnS is unchanged, while indentation toughness is increased due to a higher density and elastic modulus in recrystallized material [123]. Young’s modulus increases about 15% after HIPing, from 75 GPa to 87 GPa, possibly due to the reduction of charged grain boundary defects which scatter phonons [124]. Density in CVD ZnS reportedly increases from 4.085±0.001 before HIPing to 4.095±0.001 g/cm

3 after HIPing

[121]. Room temperature thermal conductivity increases from 17 W/(m·K) for CVD ZnS

72 to 27 W/(m·K) for HIP’d ZnS, due primarily to removal of impurities and perfection of the crystallites and increased grain size[125, 126]. Thermal conductivity of ZnS produced by various methods has been reviewed and shown to follow this trend[127].

See Appendix D for further discussion on hardness, toughness, and elastic properties in

ZnS.

Strength is decreased by HIPing from about 83 MPa before HIPing to 53 MPa after HIPing (about 65% reduction) [123]. Published strength numbers for these materials vary somewhat due to surface conditions, residual stress, and test conditions.

Fractography performed recently on HIP’d ZnS indicate that failure still occurs predominantly from surface flaws [126], so ascribing lower strength to HIP’d ZnS simply due to grain size is not supportable. Increased strength of CVD ZnS over HIP’d ZnS may be due to residual stresses, present from the CVD process in the as-deposited material, which are released when the material is HIP’d [128]; (HIP temperatures are higher than original deposition temperatures). Additionally, there is evidence that hydrogen, present in CVD ZnS material and removed with HIPing, may strengthen grain boundaries [129].

Grain growth during HIPing has been ascribed to various processes. Savage et al.

[101] suggest that stacking faults become mobile above 500 °C and defect clusters diffuse towards the surface. Hydrogen and excess Zn leaving the lattice provide the driving force for grain growth. Yashina et al. [114] invoke vacancy-mediated pore coalescence as the main driving force for HIPing, with plastic deformation playing a secondary role.

73

3.3.2.5 Effects of HIP on Optical Properties

The most dramatic effect of HIPing CVD ZnS is the reduction in visible and infrared scatter and removal of the 6 μm absorption, resulting in transmission near if not identical to single crystal cubic ZnS. More subtle but more precisely telling is the effect of HIPing on the luminescence spectra of CVD ZnS. In Zn-rich as-deposited ZnS, changes in the cathodoluminescent spectrum after HIPing have been attributed to outdiffusion of interstitial zinc and diffusion of oxygen from optically inactive twin boundary sites to active sulfur lattice sites [55]. It is the interstitial zinc atoms that are thought to be the donor species that give rise to the n-type electrical behavior in CVD

ZnS, though the resistivity is still very high due to compensation by other intrinsic defects such as zinc vacancies. Annealing in sulfur vapor reduces the number of interstitial zinc atoms and creates zinc vacancies [55]. Changes in the luminescence spectrum, particularly in the oxygen-related SA and SAL bands (see Chapter 2) are provided as evidence for these defect structure changes.

Oxygen, thought to be present at stacking faults, becomes mobile during the recrystallization from the HIP process and fills lattice sulfur vacancies and in extreme cases precipitates out a separate phase of different composition. According to Morozova

[15], in melt grown crystals precipitate sizes are on the order of 1 to 10 μm and appear slightly dark in backscattered electron images, but in general are not visible in SEM unless cathodoluminescence is used since they are compositionally different from the background by only about 1%.

74

The rate of diffusion and oxygen solubility in ZnS is thought to be decreased by the high pressure during HIPing. Dissolved oxygen in ZnS is thought to be detrimental to infrared transmission [55]. HIP treatments at 980 °C at 21 ksi provide the highest transmissions, while HIP treatments at 1050 °C at 14.5 ksi (100 MPa) show the biggest changes in luminescence from the as-deposited material. At these conditions it is surmised that the system is in equilibrium, with the high temperature causing the defect migration to speed up, and the pressure decelerating the diffusion and suppressing the phase transformation, since the hexagonal phase occupies larger volume.

3.3.3 Summary

CVD ZnS is radically transformed by heat treatment with annealing or hot isostatic pressing. HIPing produces the most dramatic transformation in optical and mechanical properties. Multiple mechanisms are operating during the HIP to change the microstructure of ZnS. Grain boundary and lattice diffusion are important for redistribution and/or removal of oxygen and hydrogen impurities. Strain from high pressure in HIPing and hot pressing assists in the martensitic transformation from hexagonal to cubic packing in ZnS. Optical transmission in HIP’d ZnS is increased to single crystal levels, and electronic defect equilibria changes produce new luminescence behavior.

75

4 Research Approach

Having now summarized the available research on ZnS, I will briefly restate the research goals in the context of this background. To reiterate, the goal of this dissertation was to describe the relationships between processing and properties in chemical vapor deposited (CVD) ZnS both before and after heat treatment. Specifically, I aim to provide insight into the mechanisms of reduced optical transmission in the visible and nearinfrared in as-deposited CVD ZnS. Additionally, I offer an explanation of the transformation process of yellow opaque CVD ZnS into clear and colorless multispectral

ZnS which occurs after hot isostatic pressing (HIPing) in the presence of platinum foil. I also comment on the relative importance of various process parameters in the conversion of ZnS. These parameters include temperature, pressure, process time, and presence or absence of process metal (i.e. Pt, Ag, Co). To this end I have employed a wide range of spectroscopic, imaging, and mechanical techniques to elicit information about variation and change in physical, chemical, and optical properties of ZnS.

The exploration of transmission loss in ZnS must be separated into absorption and scattering phenomena. Absorption has been shown to be due to native defects and extrinsic impurities, whether hydrides particular to the CVD deposition process, oxygen isoelectronic defects, or metal impurities from source materials or deposition equipment.

Scattering has been suspected to be due to pores, distinct hexagonal polytypes, or stacking disorder. The electronic structure of ZnS and its associated defects has been probed for this dissertation by transmission (absorption) measurements in the ultraviolet and visible as well as with photoluminescence. The vibrational structure of intrinsic ZnS

76 and its key defects have been probed using molecular spectroscopy methods for assessing infrared absorption and Raman scattering. The crystallographic structure of ZnS, important for understanding mechanisms of light scattering due to polytypes or stacking disorder, has been investigated by x-ray diffraction and electron diffraction.

The physical structure of CVD ZnS has been studied at various scales by appropriate microscopic imaging methods including optical light microscopy and electron microscopy (SEM and TEM). The analysis of the structure at various hierarchical scales allows an understanding of microstructural change that occurs as a result of various heat treatments. The anisotropy in CVD ZnS was explored using spectroscopy, microscopy, and physical testing techniques.

The various factors at play in the hot isostatic press were investigated separately by producing samples using only temperature, only temperature and pressure, temperature plus various metals, and temperature plus pressure with various metals. The understanding of these material processing factors required an assessment of the formation of various metal sulfides and possible catalytic actions in influencing recrystallization. Microstructural changes that occurred with heat treatment required investigation of the hexagonal to cubic phase transition in ZnS, plastic flow and diffusion mechanisms, and the presence of impurity atoms (like oxygen and hydrogen) and their effect on processing.

In the process of researching this dissertation, I felt a bit like E. H. Nickel in his

1965 “Review of the Properties of Zinc Sulphide” [130] when he stated that “in undertaking this review, the author was confronted with an embarrassment of riches. The

77 vast amount of literature on zinc sulphide, particularly its luminescent properties, makes a complete review practically impossible.” Although hardly as important a semiconductor as silicon or gallium arsenide, zinc sulfide is probably the best studied of the II-VI semiconductors. Yet there are still so many aspects of it which are poorly understand.

For instance, its tendency to polytypism is unique among the II-VI materials and production of pure materials for study has itself been difficult. Some crystallographers have spent their entire careers just describing the various short and long period stacking sequences that can be produced in ZnS crystals. The mixed covalent and ionic character in ZnS, particularly wurtzite, makes it fall in between many of the theories used to understand the optical properties of crystalline ceramics[131].

Add to this the unique defects created by processes like chemical vapor deposition and the difficulties of assessing exactly how much of each reactant is used in an open system process like conventional CVD. The lack of fundamental understanding of how

ZnS is produced via CVD causes one to wonder how bulk CVD infrared windows have been so commercially successful for over thirty-five years. If it were not for the renewed interest in using ZnS windows near visible wavelengths, much of this current work might never have been done.

Finally, I am compelled to reiterate a quotation overheard at a recent technical conference which seems relevant.

“Crystals are like people. It’s their defects that make them interesting.”

If we believe this, ZnS is a truly interesting material to study.

78

5 Experimental Description

In this chapter I describe the starting materials, processing methods, and characterization of zinc sulfide. First I describe some of the motivation for the series of experiments that were planned. Then I discuss sample description, including samples which were characterized as-received from suppliers and also those which were further processed by annealing or hot-isostatic pressing in various environments. Next I describe the equipment and techniques used to characterize these materials. Finally, I comment on some of the data processing methods and software used to analyze the data presented in the following chapters.

5.1 Sample Description and Preparation

5.1.1 Samples characterized as-received

Before going into great detail on the production of the heat treated samples, I will describe the other samples that were characterized along with them. These are summarized in Table 5-1. In addition to various chemical vapor deposited samples

(described below), a source was located for Bridgman grown cubic zinc sulfide single crystals (MTI crystals, Richmond, CA, http://mtixtl.com/index.asp), and several were purchased for analysis. Some hydrothermally grown hexagonal zinc oxide single crystals were also purchased from MTI. CVD zinc selenide material was available from

Raytheon stock as it was also grown by the same group in the 1980’s. It was hoped that

ZnO and ZnSe could provide “out-group” comparisons for various properties and that the single crystal ZnS could provide a “pure” standard for comparison. At the same time property trends could be observed in the zinc chalcogenides. Additionally, a very old

79 sample (c. 1960’s) of Kodak hot-pressed zinc sulfide (IRTRAN2) was located and characterized as well.

CVD ZnS was procured from as many suppliers as possible to compare various properties and establish a database of variability. Several lots of commercial standard

CVD ZnS were obtained from Rohm and Haas Advanced Materials (Woburn, MA and

Weeks Island, LA, http://www.cvdmaterials.com) as ground surface coupons or plate stock. Material was also obtained from the other U. S. supplier, II-VI Infrared

(Saxonburg, PA, http://www.iiviinfrared.com/) as plate stock. CVD ZnS, multi-spectral

ZnS, and a special “half-HIP’d” grade (unspecified processing conditions) was also obtained from Princeton Scientific (Princeton, NJ, http://www.princetonscientific.com/) who operated as an importer for Impex High-Tech (Rheine, Germany, http://www.impexhightech.de/). CVD (“FLIR” – forward looking infrared) grade ZnS, multi-spectral ZnS

(Pt HIP), and a special clear grade ZnS (HIP no Pt) was generously donated by Vitron

Spezialwerkstoffe GmbH (Jena/Thuringia, Germany, http://www.vitron.de/). Samples of both CVD ZnS and multispectral ZnS commercial materials were donated by Rafael

(Haifa, Israel, http://www.rafael.co.il/). Commercial material from IKhVV RAN

(Institute of the Chemistry of High-Purity Substances, Russian Academy of Sciences,

Nizhnii Novogorod, Russia) [106] was unfortunately not available for testing.

Raytheon provided CVD ZnS grown in their Waltham, Massachusetts facility, along with commercially processed multi-spectral ZnS from their starting material.

Additionally, some legacy “red ZnS” material (low temperature deposited, high zinc to hydrogen sulfide ratio) was found as well as some “elemental ZnS” material (sulfur and

80 hydrogen used as starting materials rather than hydrogen sulfide). These experimental

CVD ZnS grades were described in some detail in Chapter 3. Finally, some material akin to elemental ZnS was located through a U. S. university colleague, but which was produced in Beijing, China [104] by a process very similar to the Raytheon elemental

ZnS process.

Finally, a set of materials from a single CVD run from Rohm and Haas were investigated to explore any orientational effects of properties. These experiments are referred to as the experiments with the “cores,” as the original samples consisted of several parts coredrilled from a thick CVD deposit all the way from the mandrel surface (first material deposited) to the growth surface (last material deposited). These samples consisted of 1” diameter and 1” thick cores representing approximately 500 hours (~21 days) of growth in a CVD chamber. Samples were cut from these cores in two orientations and at three points in the core. Samples cut so that the largest face was in the plane of the growth direction were denoted “S,” while those where the largest face was perpendicular to the growth direction (i.e. what is normally encountered) were denoted “P.” Additionally, both S and P samples were cut from the mandrel side (denoted C), the middle (denoted

B), and the growth side (denoted A). Thus there were six possible samples representing two orientations of three positions in the core. This nomenclature is illustrated in Figure

5-1, showing the position of the “columns” or preferred orientation that exists in the growth direction in conventional CVD standard ZnS.

81

Table 5-1: Materials characterized as-received without further processing

All are ZnS unless otherwise noted. Standard CVD refers to the process which has Zn and H

2

S precursors.

Nomenclature Material type Supplier Country of

Origin

Notes

Bridgman Melt grown crystal

ZnS grown crystal ZnO pressed powder ZnS

Kodak USA Polycrystalline

RH Standard CVD Rohm & Haas USA Polycrystalline, various material lots

’04, ’05, ’06, cores

II-VI Standard Infrared USA Polycrystalline

Scientific (US) importer

Germany

(manufacturer unknown)

Polycrystalline

PS msZnS CVD + (Pt?) HIP

PS HH

Vitron msZnS

CVD + (?) HIP

CVD + Pt HIP

Princeton

Scientific (US) importer

Princeton

Scientific (US) importer

Germany

(manufacturer unknown)

Germany

(manufacturer unknown)

Polycrystalline,

“multispectral ZnS”

Polycrystalline, special short HIP

(unknown details)

Polycrystalline,

“FLIR grade”

Polycrystalline,

“clear grade”

Vitron HH

Rafael msZnS

CVD + HIP (no

Pt)

CVD + (?) HIP Rafael Israel

“special clear grade”

Polycrystalline,

“FLIR grade”

Polycrystalline,

“multispectral ZnS”

Raytheon USA Polycrystalline Raytran, Ray, or

Raytheon

Ray msZnS

(or just msZnS)

Standard CVD

CVD + Pt HIP Raytheon USA Polycrystalline,

“multispectral ZnS”

Red ZnS CVD Raytheon USA ~600 °C deposition, very Zn rich reactant eZnS CVD

(H

2

+ S + Zn)

Raytheon USA Polycrystalline,

“elemental ZnS”

Chinese CVD

(H

2

+ S + Zn)

Research institute synthetic crystals

China Polycrystalline

82

Figure 5-1: Orientation nomenclature for samples taken for CVD ZnS “cores”

5.1.2 Annealed Samples

Samples of Rohm and Haas CVD ZnS were subjected to various temperatures from 750 to 850 °C for times ranging from 48 to 100 hours in a graphite vacuum furnace

(Seco/Warwick Horizontal Vacuum Furnace 12Wx12Lx12H, vacuum level approximately 10

-2 to 10

-3 torr) or a quartz tube furnace (~1 atm flowing argon). These early experiments will be referred to as “annealing” experiments.

Much later a few annealing experiments were repeated but at higher temperatures to test out some suspicions regarding the relative effects of band-edge absorption and scattering on the transmission loss. Both Rohm and Haas and II-VI starting materials were subjected to 900 °C for 32 hours in the Seco graphite furnace described above. A single experiment was done at 990 °C for 10 hours to compare directly to a similar

83 condition in the hot-isostatic press, with one sample having no metal and the other wrapped in platinum foil. Quite surprisingly, these latter samples were completely gone at the end of the process run; in other words, they had evaporated completely. I shall return to this result later. A single sample of Rohm and Haas ZnS was run in a different furnace, a Centorr (Refractory Metal Zone Heat Treating Furnace, Model 16-8x12-WA-

056-B-25) with 200 torr of hydrogen gas (H

2

). In future discussions this set of experiments will be called the “UV-edge” experiments.

Finally, a series of samples of red ZnS, which has the lowest energy (longest wavelength) ultraviolet cut-on, were annealed for various temperatures and times. A total of four different vacuum anneal runs were done on these samples to assess the effects of the UV edge transmission and the 6 μm absorption, which is more intense in red ZnS than any other form. Annealing conditions ranged from 650 to 800 °C for 2 to 24 hours.

In future discussions these experiments will be called the “red ZnS” experiments.

Table 5-2 summarizes the conditions and sample names for these experiments.

84

Table 5-2: Summary of the different annealing experiments on CVD ZnS

Experimental series

# Starting material

Furnace

(atmosphere)

Temperature

(°C)

Time

(hours)

Sample thickness

(mm)

0 4.09 Anneal Co36_1 N/A

Anneal Co20_1,

Co21_1

R&H(04) Graphite

(vacuum)

No treatment

(control)

R&H(04) Quartz 700 48 2.92

R&H(04) Seco 96 2.79

Anneal Co32_1,

Co32_2

R&H(05) Seco 750 100 4.21

Raytheon Seco 100 4.22

R&H(05) Seco 32 4.75

R&H(05) Quartz 800 32 4.78

Anneal Co33_1,

Co33_3

R&H(04) Seco 850 24 2.64

UV edge

UV edge

UV edge

UV edge

UV edge

Red ZnS

Red ZnS

Red ZnS

Red ZnS

V1

V2

V3

V4

H1

S2

S3

S4

S5

R&H(06) Seco 850 24 4.11

R&H(06)

II-VI

Seco (vacuum)

Seco (vacuum)

900

900

32

32

3.53

3.49

II-VI

II-VI

R&H(06)

Red ZnS

Red ZnS

Red ZnS

Red ZnS

Seco (vacuum)

Seco (vacuum)

990

990

10

10

Centorr (H

2

) 700 16

Seco (vacuum) 850 2

Seco (vacuum)

Seco (vacuum)

Seco (vacuum)

650

650

620

24

2

6

5.1

5.1

1.52

3.18

3.18

3.18

3.18

5.1.3 Hot-Isostatically Pressed (HIP’d) Samples

5.1.3.1 Selection of Treatment Types

In this group of samples there were several subgroups. Platinum foil has already been mentioned as the commercial process for making multi-spectral ZnS for which the typical HIP temperature is 990 °C. According to the inventor of the patented process, Dr.

Chuck Willingham (personal communication), this temperature was chosen to be slightly below the transformation temperature of sphalerite ZnS to wurtzite ZnS, which is normally listed as 1020 °C. A number of platinum foil experiments were conducted, in

85 an attempt to lower the temperature of the HIP to see if grain growth could be slowed and mechanical properties of the converted material could be improved.

Work started with HIPing of samples with sputter deposited cobalt on their surfaces, and later cobalt foils were used to compare more directly to the commercial Pt foil process. Subsequently, all metals that were selected were tried using foils and only some using sputtered layers. Sputtered cobalt after HIPing left a greenish layer which was not identified but suspected to be a cobalt sulfide. This reactivity of the sputtered metal layer with the ZnS was noticed in earlier annealing work as well. Temperatures at or below 700 °C were required to observe cobalt absorptions in the ZnS (see Chapter 7), while higher temperatures seemed to “poison” the process, presumably by reacting with sulfur from the ZnS and preventing cobalt from diffusing into the ZnS.

Early hypotheses about the importance of reactivity in the metal layer led to the selection of silver as a candidate material. Subsequently several sequences of sputtered

silver and silver foil experiments were conducted. However, due to the low melting point of silver (962 °C), HIP runs were limited to 850 °C as a maximum temperature. Copper is right above silver in the periodic table and right next to zinc and is known to form many sulfides as mentioned in Chapter 2. A few sputtered copper and copper foil HIP experiments were conducted for this reason. Additionally as was described in Chapter 2, both silver and copper are common additives to ZnS phosphors and have been shown to have some effects on phase transformation in ZnS[19].

Finally, a few experiments with iron foil were tried due to its very low cost and relation to cobalt, being just to the left of it on the periodic table. The initial Raytheon

86 work in the 1980’s that identified cobalt as a promising metal for use with ZnS was originally intended to show strengthening of ZnS by in-diffusion of transition metals. For this reason iron, cobalt, chromium, and nickel were diffused into ZnS. Although iron in

ZnS had a large absorption around 3 μm which was undesirable, it was unknown whether the foil would result in the in-diffusion of iron in the same way that sputtering did. As shown in Chapter 2, iron is the most important solid solution element in natural sphalerite

ZnS.

A few HIP experiments were also done using sputtered nickel and nickel foil.

Nickel is on the right side of cobalt and in the column with platinum on the periodic table and also forms a large number of sulfides. In previous Raytheon work it had been found that nickel gave rise to many sharp absorption lines in ZnS in the visible and nearinfrared. It was also shown to be very reactive at the surface, supposedly inducing surface crystallization and “diffusion induced grain boundary migration” (DIGM). It was not known if nickel foil would have these same effects that were observed for sputtered nickel.

The range of experimental conditions run for each of the above-mentioned metals is summarized in Table 5-3. More detailed descriptions of some of the experiments are provided later.

87

Table 5-3: Summary of the different metal treatments for HIP experiments

Experimental series

# Starting material

Furnace Temperature range (°C)

No foil HIP various R&H(06), II-VI

Co sputter anneal various Raytheon,

R&H(04),

R&&(05)

Co sputter HIP various R&H(06), II-VI

HIP

Graphite

(vacuum),

Quartz (Ar)

HIP

750, 900

(650) 700 –

850

750 – 990

Time

(hours)

16, 32

(15) 24 –

100

3 - 16

Sample thickness

(mm)

various various

Ag sputter HIP 25,26,34 R&H(06) HIP 600 –750 12 - 16

1.58 -

4.95

1.51 -

1.56

Cu sputter HIP

Ni sputter HIP

64

66,

67Ni_s

II-VI

II-VI

Co foil HIP Various R&H(06), II-VI HIP

Ag foil HIP Various R&H(06), II-VI

Cu foil HIP 63

Fe foil HIP 49,51,53

,60

Ni foil HIP

Pt foil HIP

67Ni_f

Many

R&H(06)

II-VI

II-VI

Ray, R&H

(various), II-VI

HIP

HIP

HIP

HIP

HIP

750 – 1020 1,12 - 16 1.58 -

4.95

700 – 850

750

750 – 950

900

750 - 990

3 - 72

1,16

1,9 - 32

1

1,3 - 10

1.46 -

4.93

1.5

3.96 -

4.75

5.1

1.37 -

7.60

5.1.3.2 Preparation of Samples for HIPing

Typical sputtering thicknesses were about 100-200nm of metal per side, and foil thicknesses were 0.001” for the platinum, 0.003 to 0.004” for the cobalt, 0.001” for the silver, 0.004” for the iron, 0.001” for the copper, and 0.001” for the nickel. For the foil materials, ZnS coupons were carefully covered on either side with a layer of foil cut to the shape of the large surface of the sample. In some cases ZnS surfaces were polished, but in most cases a ground surface was used. No significant differences in results were observed between ground and polished surfaces with foil or sputtering.

The hot isostatic press (Autoclave Engineers Model #30P) consists of a graphite inner fixture with a sample plate. On the sample plate are placed the parts to be HIP’d, in

88 this case coupons of ZnS some with sputtered or foil metal. First a sapphire isolator (1” diameter disk, 0.05” thick) is placed on the graphite plate, then any bottom side foil, then the sample and the top foil. An additional sapphire isolator is placed on top of the foil, then two 1” diameter x 1” tall alumina weights are placed on top to ensure good contact of the foil with the ZnS sample. When multiple samples were placed in a run, they were separated from each other by sapphire isolators. The samples and the base fixture are then mounted up inside the steel insulation hood. The assembly is lowered into the HIP by a lift, and the massive threaded cap designed to hold the high pressures is placed on the HIP and sealed. Figure 5-2 shows the sample plate and insulation hood.

Figure 5-2: HIP fixturing and insulation hood

Each HIP run consists of a simultaneous temperature and pressure ramp up then a temperature and pressure hold. Quoted experimental conditions include the hold time and temperature (e.g. 750 °C, 16 hours). Unless otherwise indicated, HIP pressures for the experiments were 29.5 ksi (203 MPa), which is the approximate limit of the HIP pressure at temperatures in the range of the experiments (600 – 1000 °C). To end the run,

89 there is a depressurization to a low but above ambient pressure with a relatively short temperature hold at a very slightly reduced temperature. Finally, the temperature is ramped down slowly to reduce stresses and during the final ramp down the pressure is relieved to ambient as the temperature reaches ambient. An example of such a profile is shown in Figure 5-3.

Upon removal, samples are sent to be optically polished for transmission measurements. In the case of sputtered samples, the excess sputtered metal and/or any reaction products must be removed as part of the optical polishing.

Figure 5-3: Typical HIP schedule showing temperature and pressure

90

5.1.3.3 Specific Experimental Series

Several experimental sub-sets deserve specific mention. In these, specific groupings of parameters were varied to test out individual hypotheses. The results are discussed more fully in the following chapters.

As controls, some samples were subjected to HIP conditions like some of those with the metals but without any metal treatments (either sputter deposition or foil). These

“no foil” experiments were conducted at both low temperature (750 °C) and high temperature (900 °C) at standard HIP pressure of 29.5 ksi with two starting material pedigrees. The effect of pressure can be ascertained by similar experiments in the annealing furnace (see the “Annealing” series described above). This would apply for the low temperature “no foil” HIP, but for the high temperature analogy in the annealing furnace it has already been indicated that the samples evaporated completely during the

10 hour 900 °C experiment. This would indicate the importance of the pressure in allowing high temperature treatment without evaporation of the ZnS.

Also, a single experiment was done using low pressure in the HIP. At the time the HIP was having maintenance problems and could not maintain pressure above 7 ksi at

990 °C (recall from Chapter 2 that this is below the calculated Peirels stress for dislocation motion in ZnS). This sample was HIP’d with platinum foil for 10 hours at

990 °C and 7 ksi. The transmission of the sample is indistinguishable from those HIP’d at higher pressures. However, weight loss of the ZnS and weight gain on the foil is greater than for any other experiment of all the HIP’d samples, lending credence to the hypothesis about overpressure preventing evaporation. Additionally, the resulting

91 recrystallized grain size of this sample was more similar to other samples HIP’d with platinum foil for only 3 hours at 990 °C and 29.5 ksi, whereas a 10 hour processed sample at 29.5 ksi showed significantly greater grain growth. This is discussed further in

Chapter 6 in discussions on grain structure. This experiment will be called the “low-P

HIP” experiment.

An experiment was also conducted to ascertain the effect of contact of the foil on the transformation of the ZnS. Obviously the samples with sputter deposited metal have

“intimate contact” with the ZnS, but this was more difficult to assess using the foils which had some stiffness to them even with weights holding them down to the ZnS. A sample was prepared with 0.5 mm sapphire ring spacers on either side of the ZnS holding pieces of platinum foil away from the ZnS and HIP’d at 990 °C for 10 hours. At this temperature and time there was no distinguishable difference between the “fully contacted” samples and this sample. This experiment is referred to as the “non-contact,

Pt” experiment. These experiments are shown in Table 5-4.

Table 5-4: Special experiments: no metal, low-P HIP, non-contact HIP

Special samples # Starting material

Furnace Temperature range (°C)

Time

(hours)

low-P, Pt

Non-contact, Pt

No foil

No foil

No foil

No foil

No foil

65

54

33none

53none

55Anone

55Bnone

57none

II-VI

II-VI

R&H

II-VI

R&H

II-VI

II-VI

HIP

HIP

HIP

HIP

HIP

HIP

HIP

990

990

750

900

900

900

750

10

10

16

32

32

32

16

Sample thickness

(mm)

~3mm

4.83

1.52

4.78

3.49

3.49

4.27

92

In an effort to directly compare different metal treatments as well as the effect of temperature and time, several series of experiments were performed. One low temperature condition (750 °C, 16 hours) was set as the “standard” for comparison, given that this temperature was the optimum one for silver foil (details to be explained in

Chapter 7). An effort was made to assess the temperature and time limits of the platinum foil HIP process as well. The high temperature limit in all but a very few of the experiments with any metal was 990 °C which is the practiced high temperature with the commercial platinum foil process.

A study was done to assess the efficacy of the platinum to transform the standard

CVD ZnS to multi-spectral ZnS over a larger temperature range than was previously documented. Temperature in the HIP was varied from 750 to 990 °C in approximately

50 °C increments, keeping the HIP time constant at 3 hours. This experimental dataset is known as the “Pt foil temperature limits” study. These experiments are shown in Table

5-5.

Table 5-5: Platinum HIP experiments

Experimental

Series

# Starting material

Furnace Temperature

(°C)

Time

(hours)

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

Pt foil T limit

54

38

II-VI

Raytheon

37 Raytheon

6, 7, 8 R&H(06)

10, 11 R&H(06)

22 R&H(06)

28

31

R&H(06)

R&H(06)

32 R&H(06)

39, 44 R&H(06)

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

990

990

990

990

990

950

900

850

800

750

3

3

3

3

10

10

6

3

3

3

Sample thickness

(mm)

4.83

7.60

7.60

1.55

1.63 - 1.65

1.38

1.37

1.47

1.52

1.54

93

Silver was selected as a candidate metal for ZnS conversion, for comparison with the platinum and cobalt studies. The effect of time at 750 °C (from 3 to 72 hours) was studied for silver foil in contact with ZnS under the “Ag foil time optimization” study.

This temperature was found by conducting HIPs at a series of temperatures from 750 to

850 °C for a set time of 16 hours under the “Ag foil temperature optimization” study.

The importance of the selection temperature for silver is discussed in Chapter 7 with the optical properties. A few samples were prepared with sputter deposited Ag on ZnS and

HIP’d at slightly lower temperatures than the foil samples due to the experience of

“poisoning” the cobalt in the cobalt anneal studies. Some blackish material was typically found on the outside of the silver sputtered samples after HIPing. This experimental set is known as the “Ag sputter” study. These experiments are shown in Table 5-6.

Table 5-6: Silver HIP experiments

Experimental

Series

Ag foil t optimiz.

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil t optimiz

Ag foil T optimiz

Ag foil T optimiz

Ag foil T optimiz

Ag sputter

Ag sputter

Ag sputter

# Starting material

Furnace Temperature

(°C)

Time

(hours)

27

25_f

13, 33,

41_Ag1

R&H(06)

R&H(06)

HIP

HIP

750

750

3

12

Sample thickness

(mm)

1.51

1.54

R&H(06) HIP 750 16 1.52-1.56

41_Ag2

57

R&H(06)

II-VI

HIP

HIP

750

750

16

16

4.19

4.24

30

42_Ag1

42_Ag2

50

R&H(06)

R&H(06)

R&H(06)

II-VI

HIP

HIP

HIP

HIP

750

750

750

750

24

32

32

72

1.46

4.93

1.55

4.71

56

13

9

5

34

26

25_s

R&H(06)

R&H(06)

R&H(06)

R&H(06)

R&H(06)

R&H(06)

R&H(06)

HIP

HIP

HIP

HIP

HIP

HIP

750

750

800

600

700

750

72

16

16

16

16

12

4.57

1.52

1.59

1.51

1.56

1.56

94

In addition to the “cobalt-doped” anneal study already mentioned, some HIP experiments were done with cobalt foil and cobalt sputter deposited coatings. These HIP experiments tended to be at higher temperatures than silver, because of the ability of cobalt to handle higher temperatures in line with the temperatures for the commercial platinum process. HIP temperatures for cobalt foil were varied from 900 to 1020 °C for

12 hours in the “Co foil temperature optimization.” Thicker samples and a second starting material supplier were assessed at 950 °C for 12 hours in the “Co foil thickness

and pedigree” study. It was found that for the same thickness and process conditions, the

II-VI material improved its transmission more readily, as shown by the extinction coefficient. The effect of HIP on sputtered cobalt was assessed at comparable temperatures to the annealing study (i.e. 750 - 850 °C) as well as at 990 °C where the platinum foil process is practiced. This “Co sputter” study allowed comparison of some annealed versus HIP’d data, as well as comparison with the commercial platinum process. These experiments are shown in Table 5-7.

As a means to down-select the most promising metal foils under consideration at the time, a single HIP run was done at the standard low temperature process condition

(750 °C, 16 hours) for which there was a large body of data for silver foil. This

“simultaneous foils” experiment pitted cobalt and iron foils against platinum to convert a fairly thick piece of ZnS, where each sample was isolated from the others with sapphire spacers. Under these conditions, platinum fully clarified the ZnS, while the others showed minimal if any improvement in transmission over the starting materials.

95

Table 5-7: Cobalt HIP experiments

Experimental

Series

Co foil T optimiz

Co foil T optimiz

Co foil T optimiz

Co foil T optimiz

Co foil, thickness

& pedigree

Co foil, thickness

& pedigree

Co foil, thickness

& pedigree

Co sputter

Co sputter

Co sputter

Co sputter

Co sputter

Co sputter

Co sputter

Co sputter

Co sputter

# Starting material

Furnace Temperature

(°C)

Time

(hours)

17

15

14

12

R&H(06)

R&H(06)

R&H(06)

R&H(06)

HIP

HIP

HIP

HIP

900

950

990

1020

12

12

12

12

Sample thickness

(mm)

1.58

1.63

1.57

1.57

43Co_1 R&H(06) HIP 950 12 1.57

43Co_2 R&H(06) HIP 950 12 4.95

46, 47 II-VI

25

29

23

48_Co2

R&H(06)

R&H(06)

R&H(06)

R&H(06)

48_Co1

52_Co1

R&H(06)

II-VI (pol)

52_Co2 II-VI (grnd)

20 R&H(06)

21 R&H(06)

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

HIP

950

750

800

850

850

850

850

850

990

990

12

12

12

12

3

12

12

16

12

12

4.06

1.56

1.41

1.49

1.57

3.94

4.59

4.59

1.38

1.47

One final experiment resulted in the complete abandonment of further studies with iron and cobalt. Since iron and cobalt were shown to be less effective than silver or platinum at low temperatures, it was decided to try one last experiment with them at high temperature. In this case times were chosen to be very short so as to hopefully observe a partial conversion to multi-spectral ZnS. This set of experiments is referred to as the

“interrupted HIP” series. A single HIP run (HIP67) was conducted at 900 °C for 1 hour with a number of different samples, each treated with metal on only a single side. This configuration was designed to observe any preferential change in microstructure from the metal side as opposed to the side without any treatment. Samples were prepared with foils of Pt, Cu, Ni, Co, and Fe, along with samples with sputter deposited Cu and Ni.

Copper and nickel had, at the time, been identified as other potentially reactive metals

96 with ZnS and were thus included in both sputtered and foil form. Silver could not be included in this series due to the high temperature. Processed samples were crosssectioned through the thickness to observe what was hypothesized as a “transformation front” which would have been heterogeneously nucleated from the surface in contact with the metal.

A second “interrupted HIP” series was conducted at a low temperature, 750 °C for

2 hours, to remove the possibility that the higher temperature promoted a different mechanism (e.g. homogeneous recrystallization), and also to allow direct comparison with silver. In this experiment, only Pt, Ag, Cu, and Ni were included in both foil and sputter deposited form. This latter experiment hoped to more clearly assess the different mechanisms or extents of foil versus sputtered metal effects on transformation, if any.

All these experiments are summarized in Table 5-8.

Some initial impressions about the results of these numerous experiments have been given in this section. More complete analysis as well as transmission data, micrographs, and other relevant data are provided for a subset of these samples in

Chapters 6 and 7. Next I will provide an overview of the characterization techniques used in this study.

97

Simultaneous foils (Pt)

Simultaneous foils (Co)

Simultaneous foils (Fe)

Cu foil HIP

Cu sputter HIP

Fe foil

Fe foil

Fe foil

Ni sputter HIP

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - HT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Interrupted HIP - LT

Table 5-8: Simultaneous foils, Interrupted HIP, Fe, Ni, and Cu HIP experiments

All metals are foils unless noted as “sputter” for sputter deposited

Experimental Series metal #

Pt

Co

Fe

Cu

Cu - sput

Fe

Fe

Fe

60_Pt

60_Co

60_Fe

63

64

49

53

51

Ni - sput

Fe

Co

66

67_Fe_f

67_Co_f

Pt

Cu

67_Pt_f

67_Cu_f

Ni 67_Ni_f

Cu - sput 67_Cu_s

Ni - sput 67_Ni_s

Ag

Pt

68_Ag_f

68_Pt_f

Cu

Ni

C

None

68_Cu_f

68_Ni_f

68_C_f

68_none

Ag - sput 68_Ag_s

Pt - sput 68_Pt_s

Cu - sput 68_Cu_s

Ni - sput 68_Ni_s

Starting material

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

R&H(06)

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

II-VI

Temperature

(°C)

Time

(hours)

Sample thickness

(mm)

750 16 4.5

750 16 4.5

750

750

16

1,16

4.5

1.5

750 1,16 4.4

850 9 3.96

900

950

32

16

4.75

4.70

750

750

750

750

750

750

750

900 1 5.1

900 1 5.1

900 1 5.1

900 1 5.1

900 1 5.1

900 1 5.1

900 1 5.1

750

750

750

2

2

2

2

2

2

2

2

2

2

5.1

5.1

5.1

5.1

5.1

5.1

5.1

5.1

5.1

5.1

5.2 Characterization Techniques

Numerous characterization efforts were undertaken during the course of this project. Categorically, they can be separated into techniques involving imaging microscopy, diffraction, and spectroscopic analysis. Techniques exploited electron beams, x-rays, ultraviolet light, visible light, and infrared light to probe ZnS at various structural levels.

98

5.2.1 Microscopy

Three different microscopy methods were used to characterize ZnS – optical microscopy, scanning electron microscopy (SEM), and transmission electron microscopy

(TEM). Specific samples investigated along with the results are described in Chapter 6.

First, standard optical microscopy was used with etched samples to observe grain structure and grain size. A standard optical microscope with Nomarski contrast was used. Samples were optically polished then etched with a K-etchant (5 wt% H

2

SO

4

- 0.3 wt% K

2

Cr

2

O

7

- H

2

O) solution. According to Lendvay [132], this K-etchant is a selective agent for ZnS, with close packed planes being attacked the most, most preferentially in the

[ 10 .

0 ]

hexagonal directions more than the

[ 1 2 .

0 ] directions. Samples were etched for

15 minutes at 55 - 60 °C by placing the edge to be etched in the solution in a borosilicate beaker. The beaker was placed on a hotplate with a thermometer in the solution to monitor temperature. Occasionally a film residue remained on the sample which could be removed with a light rubbing of calcium carbonate powder paste with water.

Photographs were taken at 200x or 500x, some using Nomarski contrast, and grain sizes, when assessed, were done so using the intercept method. On different occasions, either face (i.e. plane of image perpendicular to the CVD growth direction, “P”) or cross-section

(i.e. plane of image parallel to the CVD growth direction, “S”) samples were made to assess anisotropy and extent of conversion from standard ZnS to multi-spectral ZnS.

Second, a scanning electron microscope (SEM) was used to get finer resolution of microstructure and grain boundaries. Samples were typically etched then lightly gold coated. The SEM used was a LEO 1455VP (variable pressure) microscope, which was

99 typically operated in vacuum mode about 10

-3

torr at 20kV with a filament current of about 2.45 μA.

Third, a transmission electron microscope (TEM) was used to assess the nanostructure of various grades of CVD ZnS and HIP’d CVD ZnS. The TEM used was a

Hitachi H-8100 electron microscope with a thermionic LaB

6

electron source, operated at

200keV and 10

-7

torr. Sample preparation for ZnS for the TEM was very difficult. Thin,

1 mm slabs were mounted between pieces of silicon then diced, dimpled, ion milled using argon to create a reference hole, then ring mounted. Because of the nature of this preparation, most TEM images ended up showing the plane parallel to the growth direction (i.e. the cross-section), and only in specific cases was the other orientation imaged. Final thickness of the electron transparent samples was on the order of 50 nm.

ZnS posed a real challenge for making TEM samples, as has been reported by other researchers. Samples often broke in the ion mill without thinning, and in some cases the electron beam exacerbated existing cracks and small fragments flew off into the vacuum pump during imaging. Thus it was difficult to find sufficiently thin areas for TEM imaging in many samples. Other sample preparation techniques may have been better for

ZnS, but were not readily available. Due to the polycrystalline nature of the ZnS, it was not possible to perform atomic resolution imaging,

5.2.2 Diffraction

Diffraction was used to identify crystal structures, using both electron beams and x-rays. TEM selected area electron diffraction (SAED) was performed on several ZnS samples, in order to identify the local structure as consisting of cubic phase, hexagonal

100 phase, and twinning. For the SAED investigation, a 50 μm aperture was used. It was the smallest aperture available on this TEM, and though a smaller area could have been sampled in convergent beam electron diffraction (CBED) mode, it was felt that the results would not have improved much, being fundamentally limited by the thermionic gun electron source. Diffraction pattern indexing was done using standard techniques [133] and by comparison with published diffraction patterns for ZnS.

X-ray diffraction (XRD) was performed using two different instruments. A large dataset of diffraction data for many samples of polycrystalline ZnS samples were obtained using a Rigaku Geigerflex II X-ray diffractometer fitted with a CuK

α

(1.5406

Angstroms for CuK

α1

) x-ray source, a horizontal diffraction plane, a diffracted beam monochrometer to reduce the background, and a sodium iodide scintillation detector.

Preliminary analysis of the data showed that some preferred orientation (texture) in the samples which varied considerably, especially around the location of the wurtzite (10.0) and (10.1) peaks for those samples containing some hexagonal phase. Data was collected from 2θ angles of 20 to 100°.

In a few cases, lattice parameter measurements were taken, using a NIST traceable alumina standard internal to the instrument. The alumina (0.2.10) plane reflection, expected at 2θ = 88.995°, was measured at 89.02° and the deviation used as an angular correction. Diffracted intensity was measured from 87.7 to 89.3° for ZnS at

0.02° intervals with 5 second integration time. The sphalerite (422) reflection, measured at 2θ = 88.4 – 88.6°, was corrected by the standard. This angle was used to calculate the

101 interplanar spacing (d hkl

) using Bragg’s law, and the lattice parameter using the definition for cubic crystals[134]:

d hkl

=

λ

2sin

θ

hkl

=

a

0

h

2

+

k

2

+

l

2

=

a

0

24

The estimated accuracy of lattice constant measurements is better than ±0.001 Å and may be as good as ±0.0003 Å.

Powder diffraction was performed first on a different instrument, a Scintag XDS

2000 with a CuKα source and liquid nitrogen cooled lithium-drifted silicon detector with a beryllium window. ZnS samples were crushed in a clean agate bowl and lightly packed into a sample holder. The goniometer and vertical diffraction plane only allowed data collection from 2θ angles of 10 to 70°. Considerable background was present in the data from this instrument for the low angles since it did not have a monochrometer. As has been previously mentioned, there is evidence in the literature that crushing transforms some of the hexagonal phase material to cubic phase [19], though it was found that the powder diffraction results divided nicely into three groups despite these concerns (see

Chapter 6). Additionally, some of the foils from various HIP runs were analyzed using the Scintag (and a few with the Rigaku) to identify any materials deposited on the foils.

Later, a new series of powder diffraction data was taken on the Rigaku for better quantitative evaluation of hexagonality and lattice parameter. The Rigaku offered lower backgrounds. For these tests the samples were crushed with mortar and pestle and lightly packed in vacuum grease on glass slides. For hexagonality measurements, 2θ angles of

102

25 to 35° were sampled at 0.02° intervals with an integration time of 2 seconds. The same samples were measured for lattice parameter as described above.

Identification of various phases was performed using the powder diffraction file

(PDF) database. Various analyses were performed on the x-ray diffraction data, including peak by peak fitting (to determine integrated intensity and full-width half max) for further use in texture analysis [135]. These results are discussed further in Chapter 6.

5.2.3 Spectroscopy

Spectroscopic investigation of ZnS and materials associated with its processing was done using x-rays, ultraviolet light, visible light, and infrared light.

Elemental composition of various grades of ZnS was investigated using the energy dispersive x-ray spectrometer (EDS) attached to the aforementioned SEM. It was desired to have a quantitative estimate of the atomic fraction of zinc versus sulfur in these materials. Even though the EDS does not provide totally reliable quantitative measures of stoichiometry, atomic ratios of zinc to sulfur were assessed for various samples. A

Bridgman grown ZnS crystal, essentially stoichiometric according to the supplier, was used as a “standard” to assess the accuracy of the EDS measurement. Samples were carbon coated and grounded to the sample plate with copper tape. Carbon x-ray peaks were ignored when performing the quantitative analysis. Oxygen was noted on several samples, but in quantities of 5 to 12 atomic percent if oxygen was included in the quantitative analysis. It was therefore assumed that background oxygen was too high to accurately assess any oxygen content in ZnS via EDS.

103

The main desired result of the EDS measurements was to assess whether samples were mostly zinc-rich or mostly sulfur-rich. For this experiment the electron gun was set at 30keV, and x-rays were collected from 1 to 30 keV for 300 seconds. The integrated intensity under the zinc Kα (8.631 KeV) and the sulfur Kα (2.307 KeV) peaks were used to assess atomic fraction. The “Inca” software associated with this instrument performs the ZAF (atomic number Z, x-ray absorption A, x-ray fluorescence F) correction before reporting results[136]. After Zn/S ratios were obtained, they were normalized to the

Bridgman crystal results assuming the latter was a 1:1 ratio.

Additionally, the EDS was used to determine chemical composition on the foils after their use with ZnS in the HIP. The foils were imaged in the SEM and x-ray spectra were taken of notable features. The chemical information was used in association with the x-ray diffraction data to identify the species on the foils.

Since measurement of less than one atomic percent of oxygen was impossible using the EDS method, a more reliable and quantitative test for oxygen was sought.

Several samples were sent out to Evans Analytical Group (formerly SHIVA technologies) for Interstitial Gas Analysis (IGA), which can detect sub-ppm levels of H,

C, N, O, and S. The equipment used was a Horiba EMGA-820W Oxygen/Nitrogen analyzer which works by volatilizing a small sample, converting any oxygen to carbon monoxide, and analyzing the content using a non-dispersive infrared detector. Also, a single sample was sent for Glow Discharge Mass Spectrometry (GDMS) where samples are sputtered with Ar+ ions then subsequently ionized by plasma before being channeled into a mass spectrometer. The GDMS was used to see detect metal impurities that were

104 present. Even though GDMS also measures oxygen levels, it is not considered reliable for quantitative assessment.

Transmission (absorption) spectroscopy was performed in both the ultravioletvisible-near-infrared regions and the mid-wave and long-wave infrared. Ultraviolet through near-infrared measurements (175 – 3300 nm) were taken on a Varian Cary UV-

Vis-NIR spectrophotometer. Various sources, gratings, and detectors operate on this double monochrometer based on the wavelength range. Infrared transmission measurements were taken on two different Fourier Transform Infrared (FTIR) instruments. The two instruments used were a Nicolet Protégé 460 FTIR and a Bruker

Equinox 55 FTIR. These spectrometers both report data from 2 to 20 μm with standard operating procedure being spectral resolution of 4 cm

-1

and 32 scans. The Bruker is set up with a periodic nitrogen purge which reduces the effects of variable water background.

Both FTIRs use DTGS detectors.

Photoluminescence measurements were made using an optical bench-top setup.

Both room temperature (300 K) and liquid helium (sample temperature approximately 7 to 10 K) measurements were made. Samples consisted of small wafers approximately

12mm x 2mm x 1mm, optically polished on the large faces.

A single photoluminescence measurement was done using a 337.1 nm pulsed N

2 laser (5 Hz rep rate, approx 0.45 mJ/cm

2

/pulse, 800 ps length, 10 s integration time).

Since this energy was not high enough to excite the excitonic luminescence in ZnS, it was decided to go to a 248 nm pulsed KrF excimer laser for all future measurements.

Standard integration time for the KrF excited photoluminescence was 5 seconds with the

105 signal averaged over ten scans, but for the high resolution grating measurements

(described below) and for two samples (redZnS and Chinese ZnS) the integration time was 10 seconds. Pulse energies and repetition rates for the excimer were the same as for the N

2

laser.

The pulsed excimer laser beam is directed using several dielectric mirrors, then proceeds through a diverger lens and an iris to reduce the intensity of the beam and grow it to approximately 1 mm in diameter to minimize sample damage. The samples were mounted onto the cryocooler tip with vacuum grease at 45 degrees to the excitation beam with the long axis of the sample in the vertical direction. Due to the angle of incidence, the beam diameter on the sample is slightly elongated along the short axis of the sample.

Most of the excitation beam is then reflected and scattered away from the monochrometer due to the sample angle.

The luminescence from the sample goes through another iris then a long-pass filter for excitation laser light rejection before going into a 0.15 meter monochrometer with a back-thinned Peltier-cooled CCD Si array. Monochrometer grating efficiencies had been previously optimized for the ultraviolet and visible range on this instrument, and since the responsivity of the detector degrades above 900 nm, spectral collection was limited to the range of energies slightly less than the excitation wavelength (248 nm) to about 900 nm. A rough schematic is shown in Figure 5-4.

106

Figure 5-4: Schematic of set-up for photoluminescence measurements

Several variations were tried early on before settling on a standard practice. One grating was used to cover collection of the 260.45 - 741.86 nm (4.767 - 1.674 eV) spectral range. A different grating was needed to cover the longer wavelength range of

550.02 – 1019.46 nm (2.257 – 1.218 eV). No additional significant peaks were found at the longer wavelengths, so these additional runs were discontinued after the first few samples. A higher resolution grating (600 grooves/ mm) was not found to improve the short wavelength data resolution so was discontinued after the first few samples.

Estimated worst case resolution of the “survey” lower-resolution grating was ±2 nm, with the high-resolution grating being ±1 nm.

5.2.4 Mechanical and Physical Testing

Some amount of physical and mechanical property testing was deemed useful.

This consisted of measurements of density, Young’s modulus, and biaxial flexure strength.

Density measurements were performed by the Archimedes water immersion method. A consistency test resulted in an estimated standard deviation of ±0.001 g/cm

3

.

107

Young’s modulus measurements were performed using uniaxial compression of cylindrical samples cut from the “cores” previously mentioned. The goal was to ascertain any anisotropy in Young’s modulus in the growth and perpendicular to the growth directions. Samples were cut both parallel to the growth direction and perpendicular to the growth direction. Sample diameter was 0.2512” and was very consistent due to the core drill used in the optical shop. The length varied slightly from sample to sample, approximately 0.8 to 0.9”. Sides of the cylinder were left “as-ground” but the ends were polished. The ends of the three samples cut parallel to the growth direction had excellent parallelism on the order of ±0.00005” of wedge. The three samples cut perpendicular to the growth direction, however, had varying amounts of wedge between the ends, from

0.00125” to 0.00375.” It was found that this amount of wedge was too much to get good compression data. Pieces of the samples flaked off when even small loads were applied.

Load was applied at a 0.02” per minute cross-head speed with a MTS Sintech

5/GL mechanical tester. A 500 pound load cell was used for the load-unload tests, and three cycles were completed for each of the samples cut parallel to the growth direction.

Peak load was optimized at 250 pounds to avoid any yielding. After the load-unload cycles, the load cell was changed to a 12,500 pound cell and two samples were loaded to failure, one cut parallel to the growth direction and one cut perpendicular to it. The latter sample had some wedge but continued to take load up until catastrophic failure. The failure mechanisms and ultimate compression strengths were different as will be described in Chapter 6.

108

Calculations of Young’s modulus based on the linear region of the stress-strain curve resulted in values significantly less than that reported in the literature. At this time it is currently unknown as to why there is such a large discrepancy, but compliance of the load cell has been suggested as a possible explanation. However, even with corrections for this, calculated values were still only about 70% of the literature values obtained using strain gauges. Two alternate test methodologies have been reported in the literature as described in the previous chapter. The first is a static method involving a strain gauge on the tensile side of a four-point-bend flexure bar, while the second is the dynamic method involving the acoustic wave propagation. Unfortunately there was not time or budget to repeat these experiments using one of the other methodologies.

Biaxial flexure strength testing was performed in accordance with ASTM C1499

[137]. The load ring diameter was 10.67 mm, the support ring diameter was 20.32 mm, and the sample diameter was 25.4 mm. Sample thickness was 1.9 to 2.0 mm. Various samples were tested, with some datasets having more data points than others due to sample availability. All sample surfaces were polished using a controlled grind procedure to minimize the effects of the surface flaw population differences. The exact grind recipe is proprietary to Raytheon but involves successive steps of grinding and polishing with smaller grits such that the subsurface damage induced by the previous step is removed. A rule of thumb in this kind of operation is that approximately three times the diameter of the abrasive grit must be removed in the subsequent polishing step with smaller grit size [138]. Details of the analysis procedure for the fracture data are provided in Appendix D and results are summarized in Chapter 6.

109

5.3 Data manipulation, processing

Most of the data manipulation and plotting was done using Microsoft Excel. This includes transmission, photoluminescence, and x-ray diffraction. Some curve fitting was done using Thermo-GRAMS/AI 8.0 suite of spectroscopy software, particularly for the xray diffraction intensity integration.

110

6 Characterization of Microstructure and Physical Properties

In this chapter I give an account of the results of characterization of the structure of ZnS through microscopy methods, diffraction, physical property testing, and chemical composition analysis. In the following chapter (Chapter 7) I relate the results of the optical characterization.

6.1 Microscopic Investigation of ZnS

Before delving into the details of the physical structure of ZnS, it is important to make some statements about terminology. CVD ZnS has many hierarchies of observable structure and so careful use of nomenclature must be maintained in order to avoid confusion. Figure 6-1 shows these scales of structure and the techniques used to access information about the structure. This chapter will start at the macroscale and work down to the atomic scale.

Figure 6-1: Hierarchies of structure in CVD ZnS

111

6.1.1 CVD ZnS

At the macroscopic scale, CVD ZnS can be seen to have striations or bands as well as “globules” from anomalous growth at heterogeneities in the growth process.

These phenomena were described briefly in Chapter 3. Both banding and “globules” or botryoidal structure are most readily viewed in thin specimens in transmitted light (see

Figure 6-2).

Figure 6-2: Macroscopic features in polished CVD ZnS

(L) View of growth layers of CVD ZnS in the plane parallel to the growth direction. The divisions of the “bands” are evident in the vertical direction in this photograph. This sample was taken from one of the “cores” representing approximately 20 days of growth in the CVD chamber. (R) View of inhomogeneities in elemental ZnS hemisphere using transmitted light. They are difficult to photograph, but can be seen as dark specks or regions in the part (sometimes mistaken for water residue marks). These are believed to be due to the nodules or globules that grow anomalously during CVD.

With simple etching and a light microscope, CVD ZnS can be seen to be composed of 5 – 10 μm “grains” which are more or less equiaxed when viewed on the plane perpendicular to the growth direction in the CVD chamber (i.e. the “P” orientation, called by some authors the “transverse section” or the “plane section”) (see Figure 6-3a).

112

When viewed perpendicular to this (i.e. the “S” orientation, called by some authors the

“longitudinal section” or the “cross-section”), the “columns” show an aspect ratio of about 10 times the “grain size” (see Figure 6-3b). While many authors use the term

“columnar” for this structure, it is strictly not correct, as columnar structures are more typical of magnetron sputtered thin films and do not have the “fan-like” structures shown here but are very regular and encompass the entire thickness of the film[139]. It is likely that these elongated grains are nucleated by gas phase created ZnS or impurities which fall to the substrate. These protrude farther into the gas stream, exacerbating the nonuniform growth front visible on the surface and resulting in faster growth along these grains. Temperatures are not low enough for rapid nucleation all across the surface for standard ZnS (mandrel or deposition temperature ~670 - 720 °C).

Figure 6-3: Micrographs of CVD ZnS perpendicular and parallel to growth direction

Optical micrographs of standard CVD ZnS (a,L) plane perpendicular to the growth direction and (b,R) plane including the growth direction

113

Red ZnS is deposited about 600 °C, does not show this elongated structure in the growth direction, has smaller grains by a couple of microns, and has a different crystallographic texture than standard ZnS (see XRD results below). The grain structure looks identical in both the parallel and perpendicular planes. These results collectively suggest that nucleation is more continuous in this material resulting in fine grained isotropic materials. This has been reported for continuously nucleated CVD ZnSe as well[139].

For standard CVD ZnS there are finer scales of structure which are only visible using SEM or TEM. Looking perpendicular to the growth direction, within each “grain” there appear fine striations or layers (called “bands” by some authors but called

“lamellae” here to distinguish from mesoscopic bands described previously), and the grain boundaries can be seen to be very irregular (i.e. not in obvious crystallographic orientations or facets). Other authors have described the lamellae within grains as being the “growth layers” in ZnS on the order of 600 to 800 nm in thickness[121]. Close inspection shows that these layers are parallel within a grain but have different orientations between grains. This is also evident when looking at a highly etched grain boundary under high magnification SEM. These fine growth layers appear as individual fibrous lamellae which are locally oriented (i.e. like bundles of logs) (see Figure 6-4).

114

Figure 6-4: SEM micrographs of standard CVD ZnS showing nanosized lamellae

SEM micrographs of standard CVD ZnS perpendicular to the growth direction. (a,top) showing lamella within grains orientated uniformly within grains but differently between grains; (b,bottom) showing irregular boundary and very local orientation of lamella which change noticeably across the grain boundary. Small arrows have been added to help aid the eye to see the local orientation of lamellae inside each grain.

115

Transmission electron microscopy (TEM) gives additional insight into the lamellae observed in high magnification SEM. TEM images were taken at magnifications between 10,000 and 80,000x. All the as-deposited CVD samples investigated (including several suppliers of standard ZnS, elemental ZnS, six positions and orientations of the cores, and red ZnS) had essentially indistinguishable nanostructures. Unless otherwise noted, all the TEM images show the plane that includes the growth direction due to the sample preparation technique. Structure consisted of parallel layers (lamellae) with widths of 10 to 90 nm arranged in blocks which were further arranged into sub-grains or “domains.” Contrast between the layers is believed to be due to diffraction contrast [140]. No precipitates were observed on any boundaries and no pores were observed. The layers terminate at boundaries where a new set of parallel layers takes off at a different angle, often the mirror of the angle of the previous side of the boundary. At the highest magnifications, the curved incoherent boundaries at the tips of twins can be observed. As will be shown in the section on diffraction, these layers are believed to be crystallographic twins on {111} planes. Some of the micrographs, particularly those of red ZnS, illustrate how the growth layer sets are stacked on top of each other at different angles. Figure 6-5 through Figure 6-9 illustrate these features at magnifications of 10kx to 80kx for CVD ZnS samples.

116

Figure 6-5: TEM images of CVD ZnS at 80k

(a,L) Raytheon CVD ZnS; (b,R) II-VI produced CVD ZnS.

Figure 6-6: TEM images of CVD ZnS at 60kx

(a,L) II-VI produced CVD ZnS; (b,R) Raytheon red ZnS.

117

Figure 6-7: TEM images of CVD ZnS at 30kx

(a,L) Raytheon CVD ZnS (b,R) Rohm & Haas CVD ZnS cut from the core on the

“growth” side (denoted A); because this was cut as an “S” sample the plane of the TEM is perpendicular to the growth direction, unlike most of the rest of the TEM images.

Figure 6-8: TEM images of CVD ZnS at 20kx

(a,L) Raytheon red ZnS, with the section highlighted being approximately 1500nm; (b,R)

Raytheon red ZnS, showing many overlapping subgrains at different orientations in this thin sample.

118

Figure 6-9: TEM images of CVD ZnS at 10kx

(a,L) Raytheon CVD ZnS; (b,R) Raytheon elemental ZnS. Note the similarity of structure and contrast at all levels of magnification previous to and including this one.

6.1.2 Hot Isostatic Pressed CVD ZnS

Hot isostatic pressing of ZnS produces recrystallization and grain growth. The grain morphology is completely different, consisting of angular plates with no remaining exaggerated grain size in the CVD growth direction (see Figure 6-10). In the case of

HIP’d ZnS, the contrast is somewhat better using a light microscope than using SEM, since HIP’d ZnS is typically colorless and very transparent in the visible. Thus only a few SEM pictures were taken as sample preparation for optical microscopy was easier and large areas could be investigated quickly. First some images of HIP’d ZnS are presented, then the microstructures of CVD ZnS and HIP’d ZnS are juxtaposed.

Grain sizes of annealed and heat treated materials were equal to or larger than asdeposited materials. Typical grain size for HIP’d material was 25 to 100+ μm. Early on some quantitative grain sizes were determined. The large irregular shape of the

119 recrystallized grains from HIPing and some of the higher temperature anneals made grain size measurements unreliable and not very repeatable. Defining what constituted a

“grain” was particularly difficult. Most of the subsequent investigations on grain size, therefore, were done on a more qualitative basis by visual comparison of micrographs taken at the same magnification.

Figure 6-10: Optical and SEM micrographs of Hot Isostatic Pressed CVD ZnS

(a,L) plane perpendicular to the growth direction and (b,R) plane including the growth direction. Note the isotropic nature. (c, bottom) Zygo white light interferometry view

0.1 mm

2

area of etched HIP’d ZnS (courtesy of Aric Shorey, QED Technologies).

120

Transmission electron microscopy of heat treated CVD ZnS, either annealed or hot-isostatic pressed, showed a vastly different nanostructure than that of as-deposited

CVD material (see Figure 6-11). All or almost all of the fine layer structure has disappeared, and in its place are large areas of seemingly layer-free ZnS. Because larger defect-free areas were available for imaging, the thickness contrast (“bend contours”) from the uneven ion milling is more evident in these samples. There are twin boundaries, but crystallites are on average very large, 300 to 1000 nm or larger. Nanostructure for recrystallized annealed and hot-isostatic pressed ZnS seems to be very similar if not the same. Also, from the two samples investigated, the presence of the metal layer in heat treatment does not appear to have had a significant effect on this nanostructure.

Figure 6-11: TEM images of heat treated ZnS

(a,L) Rohm & Hass CVD ZnS, annealed at 850 °C for 24 hours in a vacuum furnaces,

10kx; (b,R) Princeton Scientific Pt Hot-Isostatic Pressed CVD ZnS, 20kx. Note in both cases the large regions without fine lamella. The discolorations without straight boundaries are “bend contours” due to differential thickness from the ion beam (denoted with single ended arrows). The twin boundaries are denoted with double ended arrows and dotted lines. In the right figure a grain boundary is also evident, shown by a dashed line offset from the boundary. The areas that are completely white are where the ion mill broke through the material.

121

6.1.3 Comparison of CVD ZnS and HIP’d CVD ZnS

A side-by-side comparison of the structures of CVD ZnS before and after HIPing shows the dramatic structure change. Figure 6-12 and Figure 6-13 below shows this juxtaposition at various length scales. On the left is CVD ZnS and on the right is HIP’d multispectral ZnS. The growth axis orientation is consistent in each row of images as indicated in the leftmost image of the row.

Figure 6-12: CVD ZnS (L) versus HIP’d CVD ZnS (R) at various length scales (1)

Orientations and magnifications are the same for both images in a single row. Row 1):

SEM, growth axis in plane of image, exactly as shown; note columnar structure in CVD

ZnS at left but not in HIP’d material at right. Row 2): TEM, growth axis in plane of image, not necessarily exactly as shown; note presence of lamellae in CVD ZnS but not in HIP’d ZnS.

122

Figure 6-13: CVD ZnS (L) versus HIP’d CVD ZnS (R) at various length scales (2)

Orientations and magnifications are the same for both images in a single row. All images in this figure have the CVD growth axis perpendicular to the plane of the image. Row 1):

Optical micrograph, low magnification. Row 2): (L) SEM of CVD ZnS, (R) optical micrograph of HIP’d CVD ZnS; the dark areas in the HIP’d microstructure are polishing defects on the surface. Row 3): SEMs of CVD ZnS and HIP’d ZnS; individual grains are visible in the CVD ZnS image, on the order of 10 μm; features shown in the HIP’d

CVD ZnS are features which are internal to grains; these large layers are the equivalent to the very small nano-lamellae in CVD ZnS but at a much larger scale.

123

6.1.4 Recrystallization

One of the most useful results of the investigation of the etched ZnS was the photographic capture of samples which had only partially recrystallized. As will be discussed further, these included lower temperature HIP experiments with no metal foils or sputtering and some low temperature silver sputtered experiments (see Figure 6-14).

To my knowledge this has never been observed before, or at least is not in the literature.

The effect of a minimum temperature necessary for recrystallization can readily be seen be comparing the silver sputtered HIP series at 600 versus 750 °C (see Figure 6-15). It should be noted that these micrographs are only evidence for surface recrystallization as will be shown shortly.

Figure 6-14: Etched HIP’d ZnS samples with partial recrystallization.

(a,L) HIP25Ag_sput 200x optical micrograph (750 °C, 12 h) side 1; oddly enough, the etch of the opposite side did not show any recrystallization. This sample was sputtered on both sides, but it is possible that different amounts of the sample were removed upon polishing and any recrystallized layer that proceeded from the surface may have been removed. (b,R) HIP57_none 200x optical micrograph (750 °C, 16 h) showing partial recrystallization.

124

Figure 6-15: Etched HIP’d ZnS samples showing threshold for recrystallization

For the Ag sputtered case. Two magnifications each of two samples are shown for the treatments at 600 °C, 16h, HIP34Ag-sput (left) and 750 °C, 12h, HIP25Ag-sput (right).

A comparison of recrystallization trends at 750 °C is instructive in comparing the effects of annealing, HIPing, and the presence of metals. At 750 °C, a 12-16 hour treatment produces some recrystallization when HIP’d with sputtered Ag or with no metal

(previously shown in Figure 6-14), but full recrystallization when HIP’d with Ag or Pt foil (see Figure 6-16a,b). Vacuum annealing with sputtered cobalt at 750 °C for 96 hours produced only slight coarsening of the grains without recrystallization (see Figure 6-16c).

125

Figure 6-16: Etched ZnS samples showing 750 °C treatments

(a,L) HIP with silver foil (HIP41_Ag2, 750 °C, 16h), (b,C) HIP with platinum foil

(HIP62_Pt, 750 °C, 16h), and (c,R) anneal with sputtered cobalt (Co22_2, 750 °C, 96h).

The latter sample has only coarsened its original structure and not recrystallized, despite the 96 hour anneal.

However, it seems that no metal is required to achieve recrystallization at 850 °C in the anneal furnace and 900 °C in the HIP, based on inspection the samples in Figure

6-17. There was not a concerted attempt to find the lowest temperature for full recrystallization without metal in the HIP, but it is somewhere between 750 and 900 °C.

850 °C or possibly lower seems likely, given the experiences in the anneal furnace and the additional pressure acting as a driving force for recrystallization in the HIP.

Figure 6-17: Etched heat treated ZnS samples recrystallized without metal

(a,L) vacuum anneal (Co33_1, 850 °C, 24h); (b,R) hot isostatic press (HIP55a_none, 900

°C, 32h). Note that the transmission of the HIP’s sample was superior, with an extinction coefficient at 1064nm of 0.17 cm

-1

versus 0.60 cm

-1

for the annealed sample, even though the HIP’d sample was ~1mm thicker (see Chapter 7).

126

Although only some of the samples were etched, others with similar conditions had high visible transmission, indicative of full recrystallization throughout the entire thickness of the part. As will be shown in the following chapter, the partially recrystallized samples have very poor visible and near-infrared transmission due to scattering. Once the full recrystallization had been identified as key, it was used as a diagnostic of the microstructure.

Microstructures which have recrystallized almost all appear similar, regardless of whether pressure was applied during heat treatment, what metal was used, the process of putting down the metal, or whether metal was used at all. Two notable exceptions are worth mentioning. The very high temperature long time experiment of HIP35_Pt

(990°C, 10 hours, 30 ksi) shows evidence of coarsening after recrystallization. A similar run using cobalt foil (HIP14_Co, 990°C, 12 hours, 30 ksi) did not show the same grain growth but rather had a microstructure akin to the other HIP runs. A sample HIP’d at lower pressure (HIP65_Pt, 990°C, 10 hours, 7 ksi) does not appear to have experienced much grain growth but appears more like that of the 3 hour sample under full pressure

(see Figure 6-18). It should be reiterated that the microstructure and grain size of Figure

6-18a was representative of nearly all of the microstructures investigated, including very long time (72 hours) at low temperatures.

127

Figure 6-18: Etched HIP’d ZnS samples

(a,L) the typical microstructure observed for most recrystallized samples regardless of temperature and time, (b,C) grain coarsening of recrystallized grains, and (c,R) the same conditions as the center except low HIP pressure (7ksi rather than the usual 30 ksi) .

The observation of the threshold temperature effect on recrystallization led to an interest in determining the mechanism of recrystallization and the role of the metal on the recrystallization. It was unknown whether recrystallization proceeded from the surfaces or whether it nucleated homogeneously in the bulk. As described in the previous chapter, samples were prepared to attempt to observe the results of an “interrupted HIP.” Rather than etching the plane on which the metal treatment is applied, in this case the samples were cross-sectioned and etched so that the through-thickness could be viewed. Metal treatment was applied to only one side, so as to also compare any recrystallization on the metal versus bare side. The first batch of samples was HIP’d at 900 °C for 1 hour. A second batch was done with a subset of materials at 750 °C for 2 hours.

The results of the high temperature interrupted HIP can be summarized as evidence for a homogeneous nucleation of recrystallization throughout the bulk combined with heterogeneous nucleation of recrystallization at the surface near certain metals but not others. The most dramatic example of this is the copper foil, which is shown in

Figure 6-19. In this figure the face in contact with the copper surface has been

128 recrystallized and grains have grown markedly. On the surface without metal some recrystallization has taken place. In the cross-section, it can be seen that the recrystallization is relatively complete through the bulk with the exception of a narrow region near the surface without metal. This suggests that the recrystallization propagates from the surface in contact with the metal.

Figure 6-19: Recrytallization of sample exposed to Cu foil in 900 °C interrupted HIP

(UL) face unexposed to metal; (UR); face exposed to copper foil; (B) cross section, showing the recrystallization through the bulk and possible evidence of a recrystallization

“front” (dotted line) which moved from the metal side to the nonmetal side. Regions of new grain recrystallization are evident even in advance of the “front” suggesting some component of homogeneous bulk recrystallization at this temperature. Regions which appear unrecrystallized are indicated by small arrows. The CVD growth axis for the upper two figures is perpendicular to the plane of the image and in the bottom figure is in the plane of the image.

129

Figure 6-20 and Figure 6-21 show comparisons of all the metals tried in the high temperature interrupted HIP. These figures show a portion of the cross-section (center) and the faces exposed to the metals foil (right) and the faces unexposed to metal (left).

The side without metal is seen to be mostly unrecrystallized with a few regions recrystallized suggesting some component of homogeneous nucleation of stress free grains in the bulk of the material without requiring any metal promoter. This is evident in especially in the samples exposed to Co and Fe which did not show preferential recrystallization on the side with the metal. The heterogeneous nucleation of new grains occurred for the samples exposed to Pt, Cu, and Ni. In the case of Cu foil, extreme grain growth on the surface exposed to the metal is also evident.

Due to the short duration of the heat treatment, recrystallization had not gone to completion (with a possibly exception of the Pt sample which appears nearly fully recrystallized including the side not exposed to the Pt). Surfaces in contact with Pt, Cu, and Ni were more fully recrystallized than the surfaces without the metal treatment. In the samples exposed to metals which were promoters of recrystallization, there is evidence of a recrystallization “front” which propagated through the material from the surface exposed to the metal to the surface unexposed to the metal.

The foils seemed to provide better recrystallization than sputtered metal, presumably because the metal is not in intimate contact to diffuse in. On the sputtered metal, there are still some areas that seem unrecrystallized though more areas are recrystallized than on the side with no metal. It may be the nucleation of a metal sulfide is required to start recrystallization, and foils provide better nucleation sites. Certainly

130 surface electronic reconstruction in foils will be different than in sputter deposited layers, so available sites for nucleation will have a considerably different character in foils versus metal coatings.

There was no apparent surface effect of the Fe and Co foils, and the surfaces exposed to these metals looked identical to the surfaces which were not exposed.

Apparently the early successes at improving transmission with these materials were due solely to the high temperatures used and not due to any active role of the metal.

Figure 6-20: Samples exposed to Pt & Ni in 900 °C interrupted HIP

(L) face unexposed to metal; (C) cross section; (R) face exposed to metal. Note different scale used for center column. Some unrecrystallized areas noted with small yellow arrows. Note complete recrystallization on metal side of Pt and Ni foils, whereas some areas of Ni sputter are still unrecrystallized. It appears the Pt sample has recrystallized even on the opposite side of the sample from the metal (the “no metal face”) whereas both the Ni foil and Ni sputtered samples show large unrecrystallized regions.

131

Figure 6-21: Samples exposed to Cu, Co, & Fe in 900 °C interrupted HIP

(L) face unexposed to metal; (C) cross section; (R) face exposed to metal. Note different scale used for center column, where CVD growth axis is in plane of image with side exposed to metal to the right of the image. Some unrecrystallized areas noted with small yellow arrows. On “metal face,” note extreme grain growth in Cu foil sample compared to Cu sputter sample; Co and Fe samples show unrecrystallized regions on metal side.

All samples show unrecrystallized regions on “no metal face.” Even Co and Fe show some recrystallization which seems the same in all parts of the sample, suggesting a homogeneous recrystallization of new grains throughout the bulk at this temperature.

Heterogeneous mechanisms are also important for the Cu samples, as shown by the nucleation and growth of new grains preferentially at the Cu surfaces.

132

The low temperature interrupted HIP provided additional insight. The intent was to eliminate the effects of the homogenous recrystallization that occurred in the bulk due to thermally induced recrystallization. Pt, Cu, Ni, and Ag foil and sputter deposited metals were used on one side, in addition to grafoil (carbon) on one side of a single sample. A control sample with no metals was HIP’d with the others at 750 °C for 2 hours. Figure 6-22 shows the face exposed to the metal foils of Ni, Pt, Ag, and Cu.

Figure 6-22: Low temperature interrupted HIP micrographs for metal foils

All micrographs show the face of the sample in contact with the metal foil. Nickel and silver samples show very small isolated areas of recrystallization indicated by dotted circles. Platinum sample shows more extensive recrystallization, similar to some of the high temperature no metal experiments discussed in previous sections. The copper sample is extensively recrystallized, with the black areas in the micrograph indicating poor polishing. The copper foil sample micrograph is repeated in a larger size below for clarification.

133

Figure 6-23 compares the high and low temperature interrupted HIPs for Pt foil and Cu foil, clearly shows that the homogeneous nucleation seen in the higher temperature experiment has been eliminated.

The Pt, Ni and Ag foil samples showed only a very few regions suggestive of recrystallization on the metal side and none elsewhere. The Cu foil sample showed a dramatic “front” that went about 800 μm into the polished surface. Some material had to be taken off the surface by polishing to get a good etch, but this was minimized, resulting in micrographs that showed more black defects from pitting of the polishing than previous micrographs.

The Cu foil sample (Figure 6-24) also had a brownish color through the bulk, indicating that some copper had likely diffused into the material. From this series and the previous high temperature one, it seems that Cu is the most reactive in recrystallizing

ZnS followed by Pt. Both Ni and Ag seem only mildly effective at recrystallization at

750 °C for the short processing time of 2 hours.

None of the sputtered samples showed significant recrystallization (i.e. more than a few identifiable new crystallized grains) either on the metal surface, the non-metal surface, or the bulk cross-section. Of the sputtered samples, the Pt showed the most recrystallization, Ni and Ag about the same minimal amount, and Cu none at all. In all cases the sputtered samples showed less recrystallization than their metal foil counterparts. The faces which were not coated with metal (foil or sputtering) showed no recrystallization and appeared similar to the control sample which had no metal on either side.

134

Figure 6-23: Comparison of Pt and Cu foil interrupted HIP at 2 temperatures

The layout of the figure is the same as the previous ones on interrupted HIP. The crosssection micrograph for the high temperature interrupted HIP has been scaled to be the same magnification as the low temperature interrupted HIP for visual comparison. The top two lines show the copper interrupted HIPs at the high temperature (row 1) and lower temperature (row 2). The cross-section of the low temperature interrupted HIP shows an obvious “front” (dotted yellow line) of recrystallized ZnS extending about 800 μm from the surface that was in contact with the Cu foil. The face of the ZnS shows recrystallization, but it is difficult to see due to the defects from polishing. The bottom two lines show the platinum interrupted HIPs at the high temperature (row 3) and lower temperature (row 4). The low temperature Pt foil HIP shows some signs of recrystallization on the metal face which is also visible in cross-section (circles).

135

Figure 6-24: Recrystallization by copper foil at low temperature

Significant grain growth has not taken place such as was seen in the high temperature copper foil experiment.

6.2 Assessment of Metal Foils

This different behavior of iron and cobalt was corroborated by the investigation of some of the different foils after use on ZnS samples in the HIP. Pieces of used foil of Pt, Co,

Fe, and Cu were placed in the scanning electron microscope (SEM) for imaging and chemical identification using x-ray microanalysis through energy dispersive spectrometry

(EDS). These same foils were also examined using x-ray diffraction (XRD) to identify the structure of any materials on the foil. The results were intriguing in light of the

136 results of the interrupted HIP experiment. Both iron and cobalt foils show crystals of

ZnS attached to the foil, which by their morphology are likely vapor phase deposited.

XRD showed the crystals on the cobalt foil to be cubic, possibly a solid solution of

(Zn,Co)S, and those on the iron foil to be hexagonal, possibly a solid solution of (Zn,Fe)S

(see Figure 6-25 and Figure 6-26).

Figure 6-25: SEM of Co foil after treatment of ZnS and corresponding XRD pattern

(HIP15_Co, 950° C, 12 hrs); (a,L) SEM of the edge of the foil showing area free of crystals and area with small crystals; EDS showed the crystals to be ZnS or (Zn,Co)S;

(b,R) XRD of this same foil and one from a different sample showing cubic and hexagonal cobalt and cubic ZnS.

Cubic ZnS cannot be distinguished from cubic (Zn,Co)S by XRD except for one unique peak in the cobalt containing species out at 2θ = 117° which was outside the range of either X-ray apparatus available. The reason for suspecting possible cobalt inclusion was that the EDS spectrum of the crystals showed cobalt peaks, but these could have come from the underlying cobalt metal. Relative intensities of the cobalt base metal peaks were variable due to texturing effects from rolling of the foils. Both cubic and hexagonal phases of cobalt were indicated. Several of the characteristic sphalerite ZnS peaks are very close to those of either cubic or hexagonal cobalt, notably the second

137 strongest sphalerite peak at 2θ =47.5°. Only a very weak peak indicative of sphalerite is evident at 2θ = 28.65°, which is normally the strongest peak for sphalerite

Figure 6-26: SEM of Fe foil after treatment of ZnS and corresponding XRD pattern

(HIP51_Fe, 950° C, 16 hrs); (a,UL) SEM of the center of the foil showing many small crystals; EDS showed the crystals to be ZnS or (Zn,Fe)S; (b,UR) close-up of one of the crystals, showing its morphology; (c,LC) XRD of this same foil showing cubic iron and hexagonal ZnS.

The iron foil was much easier to interpret. There are no iron peaks near the ZnS peaks, so despite the high x-ray background at the shallow angles the wurtzite ZnS peaks are clearly visible. This ZnS can be distinguished from the cubic phase due to the peaks

138 at the 2θ = 26.9° and 30.5° indicative of the wurtzite (10.0) and (10.1) planes, respectively.

The platinum, copper, and silver foils investigated were completely different from the cobalt and iron just mentioned. Platinum and copper foils clearly formed sulfides – cooperite (PtS) and an unidentified copper sulfide with structure similar to digenite.

Silver foil showed nothing on it and exhibited no trace of anything other than silver.

However, it is apparent that the silver had some effect as previously mentioned, since the same HIP conditions without silver foil do not produce recrystallization while with silver there is recrystallization.

Two different platinum foils were run on the two different XRD instruments as described in the previous chapter. Evidence of the mineral cooperite (PtS) was unmistakable in both cases, though the relative intensities were different, and there was a very slight shift in the Scintag, about 2θ = 0.15° to the longer angles, with respect to the

Rigaku (see Figure 6-27). The Rigaku, which is a newer and more sophisticated instrument, seems to be closer to the diffraction file angles. As previously mentioned, intensity differences can be explained as due to the texturing effect of rolling the foil, and it is likely that the two foils had slightly different rolling texture. When one of the foils was probed with the EDS, it was found to contain only Pt and S in the center of the foil and only Pt and Zn at the edge. It should be noted that XRD on the opposite side of the foil showed only platinum. The side of the foil in contact with ZnS is greenish, which is the color reported for PtS [141]. PtS is known to form on Pt electrodes in fuel cells in the presence of H

2

S, impeding the transport of reactant gases from contacting the reactive

139

Pt surface [142]. However, at equilibrium temperatures higher than 850 °C, PtS thermally decomposes into Pt and sulfur vapor more readily than new PtS forms [143].

Figure 6-27: SEM of Pt foil after treatment of ZnS and corresponding XRD pattern

(a,L) SEM of the center of the foil from HIP37_Pt (990° C, 6 hrs); EDS showed only Pt and S were present; (b,R) XRD of two different Pt foils on two different x-ray diffractometers. The other foil was from HIP28_Pt (900° C, 3 hrs).

The copper foil after use had a powdery black material on the side in contact with the ZnS which could easily be rubbed off with a finger. The EDS investigation showed that the black substance consisted of copper and sulfur. The SEM image showed that there was some carbon contamination on the foil, which on careful investigation was visible to the naked eye since where the carbon was present there was no black copper sulfide. As shown in Figure 6-28, the carbon appears to have prevented copper sulfide from forming in the centers of the “ribbons” of carbon, and only copper is found there with no sulfur. The XRD investigation of this foil showed very intense copper lines and some very weak other lines which could not be satisfactorily indexed, but seem to be close to some hexagonal Cu

2

S lines. None of the lines could be ascribed to graphite.

140

Figure 6-28: SEM of Cu foil after treatment of ZnS and corresponding XRD pattern

(a,L) SEM of the center of the foil from HIP63_Cu (750° C, 16 hrs); EDS showed Cu, S, and C was present; (b,R) XRD of this foil. Identification of the copper sulfide is tentative as the best fit found was not very satisfactory.

Finally, two different silver foils were investigated using two x-ray diffractometers. No evidence of any other phase than silver was found on either side of the foil (see Figure 6-29). Visually the foil looks pristine. The SEM image with EDS analysis showed nothing but silver. It is possible that any silver that may have formed had dissociated, as one reference indicates that Ag

2

S begins to dissociate at 200 °C[141,

142], and its melting point is about 842 °C while silver melts at 962 °C[144]. Close inspection of the thermal decomposition data for Ag

2

S [145] were inconclusive.

A summary of the investigation of the activity of the metal foils is presented in

Table 6-1. It is apparent that the metals which actively react with sulfur (Pt, Ag, Cu, and

Ni) have the most effect on the recrystallization behavior of CVD ZnS. The iron group materials, Fe and Co, only cause a precipitation of vapor phase ZnS crystals, with the cobalt foil favoring the creation of cubic ZnS and the iron foil favoring the creation of

141 hexagonal ZnS. At this time it is not clear why these two metals produced different ZnS structures since the treatment temperature, 950 °C, was the same in both cases.

Figure 6-29: SEM of Ag foil after treatment of ZnS and corresponding XRD pattern

(a,L) SEM of the center of the foil from HIP33_Ag (750° C, 16 hrs); EDS showed only

Ag was present; (b,R) XRD of two different Ag foils on two different x-ray diffractometers. The other foil was from HIP25_Ag_f (750° C, 12 hrs). Each shows different texturing due to rolling.

Table 6-1: Summary of the Effects of Different Metals on ZnS

ZnS Xsection, foil

side (900 C)

ZnS X-section, sputter side

(900 C)

Foil, visual &

SEM

Foil

EDS

(center)

Foil

EDS

(edge)

Foil XRD Notes

Pt Fully recrystallized

N/A Greenish Pt, S Pt, Zn PtS, Pt PtS breaks down above ~850 °C

Ag N/A N/A Nothing base metal

Ag Ag Ag Ag

2

S breaks down above

~200 °C

Cu Fully recrystallized but grains smaller than on foil

Ni Fully recrystallized

Some small unrecrystallized areas

Black fine powder

Cu, S, C Cu, S, C Cu, unidentified but likely

Cu

2

S

“bonded” to ZnS difference in foil & nonfoil sides difference in foil & nonfoil sides phase crystals (cubic)

Co, Zn,

S phase crystals

(hexagonal)

Co Co, cubic ZnS or (Zn,Co)S

Fe, Zn, S Fe, S Fe, hexagonal

ZnS or

(Zn,Fe)S

142

Some speculations can be made about the different effects on ZnS recrystallization. Electronically, Cu and Ag have a filled d-shell and a single unpaired s electron in their outer shell which they tend to lose and become monovalent cations. Ni has 3d

8

4s

2

outer electrons while Pt has 5d

9

6s

1

. It could be that the transfer of electrons from the s to the d shells of Ni and Pt is related to their ability to provide nucleation sites for recrystallization.

Fe has 3d

6

4s

2

and Co has 3d

7

4s

2

electronic structure. Co tends to become a divalent cation when associating with ZnS, with equal preference for tetrahedral over octahedral sites, and a very similar ionic radius to Zn, accounting for the range of solid solubility of Co in cubic ZnS[146]. Fe-Zn-S is probably one of the well-studied ternary systems in geology. Iron has been implicated in the formation of polytypes in ZnS, and a wurtzite form of Zn(Fe)S is stable down to 900 °C with as low as 8 atomic percent Fe incorporation[16]. This could account for the ZnS crystals on the Fe foil being hexagonal while the ZnS crystals on the Co foil were cubic.

6.3 Diffraction Investigation of ZnS

As illustrated in the previous section with the foils, diffraction methods are extremely useful to identify crystalline phases. In this section the crystallographic structure of ZnS is investigated using both x-ray diffraction and electron diffraction. Xray diffraction studies show the evidence of hexagonal wurtzite phase ZnS in many of the samples, disordering in the closest packed direction, and texturing. Electron diffraction

143 presents evidence of the nanoscale twinning of ZnS creating hexagonal layers (stacking faults) between regions of cubic sphalerite ZnS.

6.3.1 X-ray diffraction

X-ray diffraction was performed on numerous samples of as-deposited and heat treated ZnS to investigate any differences which might suggest mechanisms for visible scattering. Polycrystalline samples were investigated using a Rigaku diffractometer and powder samples using a Scintag or Rigaku diffractometer as described in the previous chapter. An additional use of x-ray diffraction for determining the material on the surface of the foils has already been presented.

6.3.1.1 Polycrystalline sample diffraction

For ease of understanding, the results of the polycrystalline samples will each be discussed very briefly, with comments as to how they relate to the other samples. Figure

6-30 shows all the polycrystalline XRD samples plotted together with circles to denote specific features which will be pointed out in the text. Graphs have been offset so as to more easily observe the features. Commonly observed planes of both sphalerite and wurtzite ZnS have been shown at the top of the figure for reference. The description starts with the sample at the bottom of the figure and works upwards, to ease in following the explanation. The full sample name is indicated, followed by the abbreviation used in the XRD figure and the color of the line used in the graph. Only the distinguishing results of each sample will be discussed. When referring to specific planes, “s” indicates cubic sphalerite and “w” indicates hexagonal wurtzite. In general all the samples can be considered predominantly cubic phase.

144

1. Rohm & Haas 2004 lot (RH-04, blue): very broad background around the s(111) close-packing direction; evidence of a broadened peak around the w(10.1) plane; a very weak unidentified peak around 2θ = 37.8° was observed in this and most every other ZnS sample which was not found in other standards tested in this diffractometer. The s(222) and s(331) peaks seem a bit less intense than average.

2. Rohm & Haas 2005 lot (RH-05, red): broad background around the s(111) direction but with a sharp peak in the w(10.0) side and a broader one in the w(10.1) side. Some evidence of the w(10.3) peak at around 2θ = 52° is observable as well in this sample.

3. Rohm & Haas 2006 lot (RH-06 (SS), black): very sharp peak at w(10.0) and only a weak background at w(10.1).

4. Raytheon CVD ZnS c1980s (Raytran, plum-brown): no evidence of a feature at w(10.0) but very broad background on the long-angle side of s(111) and noticeable peak at w(10.1). The s(222) peak seems a bit less intense than average and the s(400) is exceptionally strong.

5. II-VI-Infrared produced ZnS (II-VI, sky blue): a hint of a step at w(10.0) and a noticeable step at w(10.1) .

6. Princeton Scientific FLIR grade ZnS (PS, green): no peak at w(10.0) but broad background between s(111) and s(200) and sharp peak at w(10.1).

7. Raytheon Red ZnS (Red, lavender): Very slight step at w(10.0) which is accentuated when longer integration times (step scans) are used. No features at w(10.1) and very little broadening of s(111).

145

8. Raytheon Elemental ZnS, c1990s (eZnS, pink): no peaks at w(10.0) or w(10.1) and very little broadening of s(111). Very little s(222) peak.

9. Rohm & Haas 2006 ZnS, annealed in vacuum 850 °C for 24 hours (RHSS Anneal, purple): no peaks at w(10.0) or w(10.1). Slight broadening around s(111).

10. Raytheon ZnS, Pt HIP at 1045 °C (HIP 1045, grey): this sample was in a HIP run where the temperature controller failed and the experimental temperature got up to 1045

°C, above the transformation temperature for wurtzite. A small shoulder is evident in the w(10.0) position but not much broadening around the s(111) which is also the w(00.2).

There is also a considerable broad background between 2θ = 45.4 and 47.2°, just to the short angle side of the s(220) which is also the w(11.0) plane. The w(10.3) is now a very sharp peak, and the s(400) and s(511) have disappeared completely.

11. Raytheon Multispectral ZnS, Pt HIP at 990 °C (msZnS, orange): The most obvious change with this sample is the huge intensity increase in the s(111) and s(222) as compared to all the other peaks which is not very evident from this particular figure.

This is no doubt due to the recrystallization favoring large twinned crystallites. There remains a sharp peak at 2θ = 27.35 ° which could be the w(10.0) peak. There is a small unidentified blip at 2θ = 54.85° along with the one at 37.8° observed in most samples.

12. Rohm & Haas 2006 core, mandrel side (RHSS core bottom, black): Very broad background between s(111) and s(200) with a hint of a peak where w(10.1) should be above the background. The s(222) peak very weak.

13. Rohm & Haas 2006 core, middle (RHSS core middle bottom, red): Very similar to the previous, with a bit more sharpening of the w(10.1) area and of the s(222) peak.

146

14. Rohm & Haas 2006 core, growth side (RHSS core top, blue): Clear w(10.0) sharp peak along with less broadened but still evident w(10.1) and evident w(10.3). The s(220) peak is much stronger than in the previous two samples which grew first in the CVD chamber and thus were held at growth temperature for longer.

15. HIP33_Ag, Ag foil, 750°C, 16 hours, RH06 starting material (Ag foil, lavender): No peaks or broadening where w(10.0) and w(10.1) should be. Very low intensity s(200) and s(400) peak.

16. HIP33_none, no metal, 750°C, 16 hours, RH06 starting material (HIP no foil, peach): Identical sample to the one above except for treated with no metal. Very broad background between s(111) and s(200) where w(10.1) should be. The s(200) and s(400) peaks are very intense, especially compared to the previous sample. It is likely that this sample did not fully recrystallized, though it was destroyed to make a photoluminescence sample before this could be verified.

17. HIP15_Co, Co foil, 950°C, 12 hours, RH06 starting material (Co foil, blue): Small shoulder and peak where w(10.0) should be. Low intensity s(200) and s(400) similar to the silver foil sample.

18. Rafael msZnS (not shown in graph, powdered sample): s(220) is strongest intensity line. There an identified peaked at 2θ = 28.8° with intensity about half that of the s(111) peak. Sample appears textured despite using a powdered sample.

19. IRTRAN2, hot pressed ZnS (not shown in graph, powdered sample): s(111) is strongest intensity line. Sample appears similar to standard powder pattern.

147

Figure 6-30: X-ray diffraction of polycrystalline specimens of ZnS

Figure 6-31 compares the diffraction pattern of a typical as-deposited CVD sample versus a typical multispectral ZnS sample. When each is normalized to its peak intensity, the reduction in the{100} lane intensities (reflections s(200) and s(400)) and s(311) and s(511) reflections is striking. By inspecting the points for the powder diffraction file, it can be seen that the multispectral ZnS, which has recrystallized after

HIP, is strongly textured on{111}planes. Even when normalized, the other peaks are not as intense as they should be according to the powder pattern. The quantitative texture analysis is presented in Appendix E.

Figure 6-32 shows the difference the silver foil has in changing the diffraction pattern of samples HIP’d at 750 °C for 16 hours with silver foil (HIP33Ag) and without

148 any metal (HIP33none). The silver foil sample was recrystallized, and the sample HIP’d without metal was presumably not recrystallized due to the strong visible scattering (see

Chapter 7). The sample HIP’d with no metal at this low temperature appears crystallographically very similar to the un-HIP’d standard ZnS shown in the previous figure. The quantitative texture analysis confirmed that without the silver foil, there was a lack of conversion from the predominant { 100 } texture of as-deposited samples to the predominant

{ 111 } texture of HIP’d fully recrystallized samples. In the silver foil case, there is a residual strong (422) reflection, or

{211} texture, which is removed when the samples are fully HIP’d at high temperature (i.e. 990 °C) as in the previous figure.

Presumably the high temperature along with the pressure allows the dislocations to move and recrystallization to take place more completely in shorter times.

A significant amount of quantitative analysis was performed on the polycrystalline x-ray diffraction data. The details of this analysis are provided in

Appendix E, and only the results will be presented here. X-ray data was used to analyze the nonstoichiometry, crystallite size, close-packed direction disorder, percent hexagonality, crystallographic texturing, and lattice parameter. The nonstoichiometry factor was found to be inconclusive with the methodology as presented in the literature being highly suspect. Crystallite size, which is more aptly described as the coherently diffracting domain or crystallographically perfect subgrain size surrounded by defects and dislocations, was found to be on the order of 50 to 100nm regardless of sample type.

Disordering on the close-packed direction, defined as the integrated area of the (111)

149

Figure 6-31: XRD comparisons for CVD versus HIP CVD

Intensities for the polycrystalline samples are normalized to the strongest peak in each case. The intensities of the non-textured powder pattern are shown for reference. All indexed peaks are for sphalerite.

Figure 6-32: XRD comparisons for Silver versus No Metal HIP

Intensities for the polycrystalline samples are normalized to the strongest peak. These samples had the same HIP treatment at 750 °C for 16 hours. Note that the sample HIP’d with no metal still shows signs of wurtzite w(10.1) and texturing like the as-deposited

CVD material. The powder pattern is shown for reference.

150 sphalerite reflection divided by the fitted background area, was found to vary substantially among the as-deposited CVD materials. The disorder in the close-packed direction was essentially absent in the transparent as-deposited samples (i.e. elemental

ZnS and red ZnS) and the heat treated samples (with the exception of the no metal low temperature HIP33none which is considered to be unrecrystallized and more like standard CVD ZnS). The percent hexagonality analysis yielded similar categorical results to what was found in the powder x-ray diffraction investigation (see below). Asdeposited CVD ZnS was determined to have about 2 to 5 mole percent hexagonality, fully transformed HIP’d ZnS has less than 0.5 mole percent hexagonality, and red and elemental ZnS are in the middle.

The texturing analysis was more fruitful and leant insight into the recrystallization process. It was found that as-deposited samples, with the exception of redZnS, were predominantly

{ 100 } textured (i.e. the intensity of the (200) and (400) x-ray reflections were significantly greater than that from the standard powder pattern). By contrast, most of the heat treated samples were predominantly

{ 111 } textured. A discussion of the crystallographic significance and insight into HIP recrystallization mechanisms is provided in Appendix E. These calculated textures agree with those stated in the literature for HIP’d and CVD ZnS [121, 147].

Lattice parameter measurements were conducted on a limited number of samples using a high angle alumina peak as a standard to assess angular error. Details of the interpretations are provided in Appendix E. Fundamentally, the lattice parameter

151 measurements lend support to the notion that CVD ZnS in all its forms has considerable amounts of substitutional oxygen in sulfur sites which reduces the lattice parameter.

6.3.1.2 Powder diffraction

The considerable variability found in these polycrystal diffraction experiments, especially around the areas where wurtzite reflections can be identified, led to the investigation of ZnS using powder diffraction. The desire was to eliminate any affects of preferred orientation and allow the x-ray to have access to all the possible planes. There was a concern that grinding would convert hexagonal phase to cubic to some extent.

Even if this was the case, there was enough difference in the hexagonality of the measured samples that the powder diffraction was useful. The results of this series of experiments can be summed up very simply.

All the powdered samples measured fell nicely into three groups considering the diffraction around the closest packed planes. Intensities of the other peaks were very similar if not identical as expected from powder diffraction, though some evidence of minor powder orientation in the sample holder resulted in relative peak intensities being somewhat different than the powder diffraction file data for sphalerite ZnS. Some small spurious peaks were evident at 2θ = 25.66°, 42.52°, and 50.32° due to the silicon detector in the Scintag. These peaks were verified as being due to silicon by running a silicon standard. The major silicon peaks are very close to the major sphalerite ZnS peaks, since silicon has the diamond structure which is just the zincblende structure where all the atoms are identical.

152

The first classification group consisted of all the heat treated samples – annealed

Rohm & Haas (same sample as the TEM, 850 °C, 24 hours in vacuum furnace), multispectral ZnS (990 °C Pt HIP), and two Vitron HIP’d samples, one their commercial multispectral material (VIT-CLR) and the other their special no Pt HIP material (VIT-

HH). None of these samples showed evidence of a w(10.0) peak or any broadening around s(111) or near w(10.1) (see Figure 6-33). Note that in this group of samples, all had fully recrystallized. It is likely that heat treated samples which had not fully recrystallized would still show some hexagonal peaks and would not be classed in this group. The appearance of a very weak shoulder in the annealed sample places it on the borderline with the next group.

The second group consisted of as-deposited samples which are intrinsically highly transparent without heat treatment, including red ZnS, elemental ZnS, Chinese ZnS,

Raytheon standard ZnS (which was very transparent compared to other “standard ZnS” samples) and Bridgman single crystal. These samples all had a small shoulder in the w(10.0) position (see Figure 6-34). It is interesting that the single crystal would be in this group, but the literature reports that fully cubic single crystals are almost impossible to make and invariably contain some amount of stacking faults and disordered hexagonal regions. The red ZnS powder sample had one unidentified peak at 2θ = 40.57°.

Finally, the last group included all the CVD materials that appear opaque in the visible. This included the Vitron FLIR grade (i.e. yellow standard ZnS), Rohm & Haas

2004 and 2005 lots (which appeared different when investigated as polycrystals), and six samples from the Rohm & Haas 2006 core (2 orientations each from the mandrel, middle,

153 and growth side of the core). These samples all appear nearly identical, showing a sharp peak in the w(10.0) position with no other broadening in other wurtzite positions (see

Figure 6-35).

The core sample comparison is particularly interesting since core sample diffraction measurements were taken using polycrystals as well (see Figure 6-36). From the powder diffraction, it can be seen that the amount of the hexagonal phase is essentially identical among all the parts of the core. From the polycrystalline diffraction, it can be seen that hexagonal wurtzite has a preferred orientation which is different in different parts of the core since the relative intensities of the different wurtzite peaks and disordered background varies with the different positions in the core. Texture data presented in Appendix E supports the crystallographic differences in core position, but the overall phase composition seems to be the same according to this powder diffraction.

Figure 6-33: X-ray powder diffraction of the heat treated group.

Note absence of distinct wurtzite peaks as shown by arrow. All curves are on top of each other so are not differentiated.

154

Figure 6-34: X-ray powder diffraction of the transparent as-deposited group.

Note small wurtzite peaks as shown by arrow. All curves are on top of each other so are not differentiated. Unidentified peak is from red ZnS

.

Figure 6-35: X-ray powder diffraction of the opaque group.

Note sharp wurtzite peak as shown by arrow. All curves are on top of each other so are not differentiated

155

Figure 6-36: Powder diffraction versus polycrystalline diffraction for the cores

Intensities for each dataset are normalized to their most intense peak, here the s(111) in all cases. Note that in the powder diffraction (Scintag), all the samples are essentially identical, while in the polycrystalline diffraction (Rigaku) there is evidence for texturing of the hexagonal phase. The high background at the short angles in the Scintag is independent of the sample for this instrument. Sample names are as described in Chapter

5, with S and P being orientations in the core.

6.3.2 Electron diffraction

A number of diffraction patterns were taken using the TEM. Some were indexed per the usual methods [133] and by comparison to other published electron diffraction patterns of ZnS. The most obvious feature of the diffraction patterns was their evidence for cubic sphalerite twinning. This was evident at many scales in the as-deposited CVD samples and at a single (larger) scale for the HIP’d samples. Also possibly evident in the as-deposited samples was the simultaneous presence of wurtzite phase or at least some

156 hexagonal stacking not associated with twins. The exact nature of the atomic arrangement is still not fully understood.

Figure 6-37 shows a selected-area electron diffraction (SAED) pattern. Only a few diffraction patterns were taken, but the general results are consistent within the asdeposited samples. In this pattern the splitting of spots from twinning is apparent. The pattern is very similar to those patterns shown by Ma et al. [148] from their nanobelts which had coexisting twinned cubic and hexagonal ZnS phases, and Hao et al. [149] from their nanowires periodically twinned every 4 nm. The one difference between this pattern and the aforementioned one is that only some of the hexagonal spots seem to be present in this pattern. In the figure below, the blue and green hexagons represent the two cubic twinned lattices and the orange rectangle the proposed hexagonal phase. The spots in the horizontal row of the beam axis should be hexagonal structure spots, but the ones that should be between the twins are not evident. This may be indicative of stacking disorder and imperfect hexagonal phase.

Figure 6-38 shows another diffraction pattern of a twinned region. It can be described as cubic with twins, similar to the x-ray precession diffraction pattern from

Fleet[35]. TEM diffractions from Holt and Culpan [150] also show similar patterns, but theirs are very streaked along the [111]* reciprocal lattice row, and they indicate that the

6H polytype would produce nearly the same pattern since it is a regularly twinned structure. Streaking is usually caused by disordering along that particular reciprocal lattice vector. Figure 6-39 is a similar pattern but from a redZnS sample that shows more spot streaking.

157

The two diffraction patterns in Figure 6-40 suggest hexagonal orientation with some cubic spots. Somewhat similar patterns were presented for wurtzite single crystals by Blank et al. [140], and for wurtzite nanobelts by Wang et al. [37]. Blank has indexed the pattern in a way that appears to represent the

[ 2 1 1 0 ] beam direction. In the pattern shown here, the spots that should be absent for pure hexagonal but which may appear due to double diffraction (see Edington [133]) are present. This may be indicative again of disordering, with twinned cubic and hexagonal occurring together. The exact beam direction could not be determined, as the length ratios did not follow any of the standard crystallographic directions for hexagonal crystals.

Figure 6-37: Raytheon ZnS 60kx TEM and SAED of highly twinned region

Subscript “T” indicates twin. The [111] reciprocal lattice row is indicated as [111]*.

There are two sets of twins diffracting here as shown by the two hexagons in the indexed pattern. Some of the other spots cannot be assigned to the cubic lattice, but the whole of the hexagonal lattice is not diffracted either. This suggests simultaneous presence of cubic and hexagonal layers with some disorder in addition to the cubic twinning which is a single hexagonal sequence.

158

Figure 6-38: Raytheon ZnS 40kx TEM and SAED of highly twinned region

Beam direction is [011]. Subscript “T” indicates twin. The [111] reciprocal lattice row is indicated as [111]*. The row of single dots is common to both twins and to 2H. Other than the twin row, no evidence of other hexagonal phase is shown here.

Figure 6-39: Red ZnS 10kx TEM and SAED of highly twinned region

Beam direction is [011]. Subscript “T” indicates twin. The [111] reciprocal lattice row is indicated as [111]*. Here more streaked spots indicating disorder are evident.

159

Figure 6-40: Raytheon ZnS 40kx TEM and SAED of two similar regions

Patterns could not be indexed as hexagonal or cubic.

Figure 6-41 shows a red ZnS diffraction which shows the small grains in almost a hexagonal pattern anticipated from cubic material.

Figure 6-41: Red ZnS 20kx TEM and SAED showing small variously oriented grains

160

Figure 6-42 shows the effect of hot isostatic pressing, with large fault free regions with some vague double diffraction spots and thin twin bands remaining. Note the complete absence of fine twinned lamella and disordered regions which are typical of the as-deposited CVD materials. In general, the diffraction patterns of the HIP’d or annealed samples were very uninteresting compared to the standard CVD ZnS samples. The regions all appear cubic, but the beam direction was not determined and patterns were not indexed. A few twins were evident, as shown in the previous sections as well, but in general the nanostructure was uniform and cubic.

Figure 6-42: Princeton Scientific ZnS 40kx TEM and SAED of two regions

Region (1) is a bulk crystal region and Region (2) is on a thin twin. Both areas appear cubic.

Together with the microstructural images, these diffraction results suggest that

CVD ZnS is composed of fine nanosized twins arranged in domains then grains. The stacking is predominantly cubic, heavily twinned, and contains some other hexagonal stacking disorder. Heat treated ZnS which is HIP’d above ~750 °C or annealed above

~850 °C is at least somewhat recrystallized. This recrystallization eliminates most if not all of the nanostructured twins, leaving large fault-free cubic regions. The grains are twinned on a larger scale, and fully recrystallized HIP’d ZnS shows strong

161

{111}texturing from these mesoscale cubic twins. As there are fewer twins in HIP’d

ZnS, there are fewer hexagonal layers (twin boundaries) and so the overall hexagonality of the recrystallized HIP’d ZnS is lower than for CVD ZnS.

6.4 Physical Properties

In this section I summarize the results for density, fracture strength, and chemical composition.

6.4.1 Density

The x-ray density of cubic ZnS is 4.123 g/cm

3

and for hexagonal ZnS is 4.103 g/cm

3

based on the perfect unit cell [15]. The actual density of materials is always lower than the x-ray density due to grain boundaries and defects [151]. In practice wurtzite is usually quoted as 4.08 and sphalerite as 4.09 g/cm

3

. Early assessments of CVD ZnS density using toluene immersion reported an average density of 4.08 g/cm

3

[92]. Later assessments of Raytheon CVD ZnS state that run-to-run density fluctuations were

4.08335±0.00020 g/cm

3

or Δρ/ρ≈ 5 x 10

-5

[59]. Russian researchers report an increase in density with hot isostatic pressing of CVD material, raising the measured density from

4.085±0.001 to 4.095±0.001 g/cm

3

[121]. The effect of dissolved oxygen at the level of

10

18

/cm

3

(0.004 atomic percent) on density is deemed negligible, but zinc vacancy concentration in sulfur annealed single crystals has been estimated from Δρ of 0.0004 g/cm

3

[15].

Twenty-two measurements were made of various ZnS samples. Figure 6-43 below shows the density measured and frequency a given value was recorded. Each different sample type contributed one to the frequency count, so the frequency histogram

162 represents the distribution of densities among the sample types. Most of the samples measured 4.089±0.001 g/cm

3

with a few slightly more or less. The large differences reported by the Russians between CVD and HIP CVD ZnS were not observed. The

HIP’d samples were 0.002 and 0.003 g/cm

3

denser that the average, with an estimated error on all measurements of ±0.001. Based on the limited number of measurements, it seemed that there could be a correlation between high optical transparency in the visible and density (i.e. redZnS, eZnS, and HIPZnS all had densities on the high side).

Figure 6-43: Histogram of measured densities of ZnS samples

Each sample type measured contributes one to the frequency, so the histogram can be seen as representing the distribution of measured densities. For example, in the measured

CVD ZnS type, multiple suppliers and multiple lots of CVD ZnS were measured.

6.4.2 Fracture Strength

Fracture strength of ZnS was assessed on various sample types. Most of the testing performed was ring-on-ring biaxial flexure. One experiment was performed in compression to assess the relative strengths and Young’s moduli of the two directions in the CVD growth process.

163

For biaxial testing, nine sets of samples were tested. Details on the test setup were provided in Chapter 5 and details on the analysis method are in Appendix D. A simple T-test was used to see if the mean fracture strengths were alike or different. The results are shown in Table 6-2 where the samples, their designations, and the number of samples is as follows: Rohm & Haas CVD ZnS (RH, 20); RedZnS (Red, 23); elemental

ZnS (E, 26); Raytheon multispectral HIP’d ZnS (MS, 7); Princeton Scientific multispectral HIP’d ZnS (PS HH, 6); Princeton Scientific CVD FLIR grade (PS, 6); II-VI

CVD ZnS (II-VI, 26), Rohm & Haas CVD ZnS HIP’d with Ag foil 750 °C 16 hours

(AgHIP61, 20); and Rohm & Haas CVD ZnS HIP’d with Pt foil 750 °C 16 hours

(PtHIP62, 17). Boxes with a red “x” indicate that the row versus column T-test (twotailed distribution, two-sample unequal variance) shows these are statistically different.

One outlier sample was removed from the PtHIP62 dataset which resulted in the table as shown below.

This analysis basically shows that the samples can be grouped into four categories based on the mean strength (values shown in Table 6-3). The strongest samples were the

II-VI standard CVD ZnS samples (group 1). Statistically lower strength as-deposited

CVD samples (group 2) included the yellow standard commercial CVD grades produced by Rohm & Haas (RH) and Princeton Scientific (PS), as well as the special CVD grades of elemental ZnS (E) and Red ZnS (Red). Next lower in strength were the samples

(group 3) which saw a 750 °C HIP treatment, AgHIP61 and PtHIP62. These HIP’d samples HIP’d at low temperature had slightly higher average strength than those HIP’d at high temperature (990 °C). This latter dataset (group 4) included the HIP’d samples

164 processed using the normal high temperature HIP process of commercial multispectral

ZnS (MS and PSHH datasets).

Table 6-2: T-test on biaxial flexure strength of ZnS samples

A Weibull analysis was performed using the maximum likelihood estimate (MLE) method as described in Appendix D. A summary of the parameters determined from this analysis are shown in Table 6-3. The strengths were normalized to 1 cm

2

stressed area based on the effective tensile stressed area (ASTM C1499 eq. X1.3). For this analysis, the Poisson’s ratio for the HIP’d samples was assumed to be 0.32 and that of the CVD samples was assumed to be 0.29 per Klein and Willingham[152]. The biased Weibull modulus does not take into account the number of specimens while the unbiased one does and is therefore more conservative. These values agree well with previous published

Weibull numbers[153], when all the data is normalized correctly (see Appendix D).

A probability of failure plot based on these parameters is shown in Figure 6-44.

This is based on using the unbiased Weibull modulus to account for small sample sizes.

It can immediately be seen that the high temperature HIP’d samples have a relatively high Weibull modulus (slope of the probability of failure curve) with a lower scale factor

(strength of the 62% point in the distribution). The lower temperature HIP’d samples,

165 though they have a higher average strength and scale factor, have a low Weibull modulus indicating a broad distribution of strength (large standard deviation). The CVD materials all have a moderate Weibull modulus and higher scale factor than the heat treated samples. All tests were done on the same apparatus with the same surface preparation by the same optical shop, so surface flaws should be similar for all samples. However, the different Young’s modulus between as-deposited and HIP’d samples (as described in

Chapter 2), could influence the behavior of abrasives interacting with the sample and result in a different flaw size in the CVD versus HIP’d materials.

Table 6-3: ZnS biaxial flexure results

Mean and standard deviation is shown for actual samples. Weibull scale factor is normalized to 1cm

2

stressed areas.

The shape of the curve for the low temperature HIP’d samples suggests that they may have a fracture character intermediate between the as-deposited and high temperature HIP’d samples. It should be recalled that these low temperature Pt and Ag

HIP’d samples appear to be fully recrystallized with a grain size comparable to that of the high temperature HIP’d samples. However, the texture analysis on similar Ag foil samples showed a somewhat intermediate crystallographic texture between CVD ZnS and high temperature HIP’d ZnS.

166

Figure 6-44: ZnS biaxial flexure probability of failure versus applied stress

The literature shows that the transition between the lower single crystal fracture energy to the larger polycrystalline fracture energy is best understood in terms of the ratio of the flaw size to the grain size [154]. Ratios of about one or higher generally indicate polycrystalline fracture mechanics should be applied. Flaw sizes in HIP’d ZnS have been measured at 100 to 200 μm [126], suggesting that even the relatively large-grained HIP’d material is controlled by polycrystalline fracture mechanics. However, the caveat is that crystallographic texturing and preferred cleavage planes can greatly influence the transition from single crystal to polycrystalline fracture energy [154]. CVD ZnS, like the studied CVD ZnSe in the previous reference, has preferred texturing and a single preferred cleavage along the [110] direction. The difference in crystallographic texturing

167 between the low temperature and high temperature HIP’d samples can probably explain the different fracture behavior.

Finally, a set of compression tests were done on cylinders to assess the different behavior when stress was applied in the axis of the growth direction (i.e. in the direction of the “columns”) and perpendicular to the axis of the growth. Initially these tests were designed to measure the Young’s modulus, since even in compression the limit of small strain should give the same elastic modulus, thus avoiding the associated problems of tensile testing brittle ceramics. However, a number of problems were encountered that rendered this test more qualitative than quantitative. Firstly, the samples cut perpendicular to the growth direction ended up having excessive wedge as described in

Chapter 5. Hence there was some anomalous behavior as the samples seated under load.

Secondly, the mechanical tester had a significant amount of compliance due to the threaded features and the load frame. Even when calibration runs were done and the instrument function was subtracted out from the displacement, calculated values of

Young’s modulus were too low.

The failure mode of the two different types of samples in compression is unique and has not been reported previously. The corrected stress versus strain curve is shown in Figure 6-45 along with photographs of the failed samples. The stronger sample with the columnar exaggerated grain grown in the axis of the stress failed at 377 MPa at

1.86% strain. Before failure the center of the cylinder got very white, indicative of a porous zone such as that which develops under the compression of Vickers’ diamond indenters [155]. The porous zone that formed under Vickers’ indenters was shown to

168 form as a result of plastic shear along flow lines and nucleation of voids where flow lines crossed. The authors reported that all cxracks nucleated from the boundary of the plastic porous zone to the large elastic field or at the crossing of shear flow lines. so it may be that there are local stresses and strains that are much higher than that of the overall sample. These researchers found that the porous zone formed at 2.5% strain and 1.5 GPa.

The samples tested here fractured in the middle of the sample length at considerably lower measured stresses and somewhat lower strains than reported for the indentation work. It may be that the localized stresses in the compression sample were much higher than the maximum calculated stress for the overall sample, and that the nucleation of the porous region occurred under similar conditions to that reported for the indenation tests.

SEM micrographs of one of the fractured pieces are shown in Figure 6-46.

The sample loaded perpendicular to the columns showed a much lower failure stress and a different failure mode. As mentioned there was some wedge in this sample which affected the transient regime. This sample held a maximum stress of 237 MPa at

1.6% strain. As can be seen from the graph, there were events where the stress jumped down then went up again. This happened when sections of the cylinder were sheared off longitudinally. The stress plotted is calculated from the original cross-sectional area and so the actual stress is higher when whole longitudinal sections fell off as they did at these points. There was no way to normalize the instantaneous cross-sectional area, but likely the stress would show a more usual shape if this were done. The white zone never developed in these samples as they did with the other. Rather, whole sections of the sample along its entire length broke off. The lower compressive strength and unique

169 failure mode is indicative of a shearing of cleavage of weakly bonded planes. Since these samples were known to be significantly {100}textured along the CVD growth direction

(i.e. in the axis perpendicular to the long axis of this cylindrical sample, see Appendix E), it appears that the crystallographic texture plays a large role in the compressive strength as it does the biaxial strength. SEM micrographs of one of the fractured pieces are shown in Figure 6-47.

Figure 6-45: Stress-strain curve of fractured samples in compression test

170

Figure 6-46: Images of fractures by compression parallel to CVD ZnS columns

Note the smooth face on the left image which is the outside of the cylindrical sample.

Right image shows the cavities in the porous zone which initiated fracture. The fracture areas shown in these images were white to the eye.

Figure 6-47: Images of fractures by compression perpendicular to CVD ZnS columns

Note the very smooth fracture surfaces (left) which are much different in character to those shown above. No porous zones were evident. Sample fractured into long pieces with surfaces 90 degrees apart (i.e. into rectangular prisms which often were as long as the entire specimen. One of these 90 degree surfaces can be seen in the righthand image, where the light area meets the darker area.

6.4.3 Chemical Composition

The results from Interstitial Gas Analysis (IGA) and x-ray microanalysis via energy dispersive spectroscopy (EDS) are discussed here. A report on the chemical composition from Glow Discharge Mass Spectrometry (GDMS) is in Appendix D.

171

6.4.3.1 X-ray Microanalysis

X-ray microanalysis was performed using the EDS on an SEM. Five minute counts were performed on carbon-coated and grounded parts excited with a 30kV electron beam at 250 μA in a 4x10

-3

torr vacuum. The analysis was done in two set-ups.

The same Bridgman ZnS single crystal to be used as a stoichiometric standard was measured in both set-ups. X-rays were collected from approximately 400 x 400 μm areas, and three areas were sampled and the results averaged. Table 6-4 shows the raw atomic percent of Zn and S reported from integrated intensities and the Zn to S ratio. The data for each of the three areas is shown along with the mean and standard deviation for the ratio. HIP5-Ag was tested to see if any silver showed up in EDS and it did not.

As can be seen from the data, the raw data shows most samples were very close to a 1:1 zinc to sulfur ratio with a small standard deviation. The deviation however in most cases puts the material on the Zn-rich or S-rich side depending on whether the deviation is added or subtracted. Assuming that the Bridgman crystal is stoichiometric, however, changes this assessment. Two different normalization cases were considered. First, the average of the two Bridgman runs (6 measurements total) gave a zinc to sulfur ratio of

0.83. If this value is used to normalize all the other values, only one sample remains sulfur-rich, the Raytheon multispectral ZnS. The measurement on this sample is somewhat suspect, however, since the standard deviation on the raw ratio is 0.14 and the other multispectral ZnS sample by Rafael shows a much higher Zn/S ratio. To be more conservative, the ratios of all the samples were also normalized to the highest value reported for any of the 6 Bridgman measurements, which was 0.97. In this case all

172 samples except the Raytheon multispectral still showed up zinc-rich but much less so than in the former case. These normalized ratios are shown in Table 6-5. Any other choices for the normalization constant from measured Bridgman crystal data would have resulted in materials which were more zinc-rich or between these two values.

Table 6-4: X-ray spectrometry: EDS, 300 seconds, raw data

Sample Atomic% Zn Atomic% S Zn/S ratio Zn/S average Zn/S stdev

Bridgman (1 st

) 42.57 57.43 0.74

Red ZnS

Raytheon msZnS

50.88

39.59

49.12

60.41

1.04

0.66

50.45

Bridgman (2 nd

) 49.22 50.78 0.97

Rafael msZnS 50.64 49.36 1.03

Vitron stdZnS

Rafael stdZnS

51.31

50.77

48.69

49.23

1.05

1.03

173

Table 6-5: X-ray spectrometry: EDS, data normalized to Bridgman crystal

*note that Bridgman was always considered 1.00 after normalization

Bridgman (1

Red ZnS

average normalization)

Zn/S average

(normalized to 0.83)

st

& 2 nd

) 0.83 1.00

1.02 1.23

Zn/S average

(normalized to 0.97)

1.00*

1.05

Rafael stdZnS

Vitron stdZnS

Raytheon msZnS

Rafael msZnS

1.00

1.03

0.78

1.02

1.21

1.24

0.94

1.23

1.03

1.06

0.81

1.05

From these analyses it is tentatively suggested that all of the CVD ZnS and hotisostatically pressed CVD ZnS (“multispectral” or Raytheon msZnS) is slightly zinc-rich as expected from thermodynamic considerations for charged defect equilibria and oxygen reduction of the sulfur. The exact off-stoichiometry is probably small and somewhat variable, but does not seem to be correlated to whether or not the samples were hot isostatic pressed. Also red ZnS does not appear to be any more Zn-rich than some other materials, despite the fact that it is grown with a higher zinc to sulfur reactant ratio.

Finally, the heavily doped silver sample does not register any silver in the EDS though it shows a large absorption which may be due to silver particles.

The question has to be raised as to whether these results were sufficiently reliable to conclude that CVD ZnS in all its forms is Zn-rich. The only better spectroscopic test would be one with an independently confirmed standard, such as using Rutherford

Backscattering Spectrometry (RBS) to assess a stoichiometry which could then be used as a standard for EDS. Secondary Ion Mass Spectrometry (SIMS), while more accurate at quantification, would only provide a very local assessment of chemistry and so would

174 require many iterations for reasonable statistics. Alternatively, many more samples and areas could be tested in EDS leading to better statistics. Overall it was felt, however, that none of these additional tests would lead to any more fundamental understanding of CVD

ZnS so they were not pursued.

6.4.3.2 Interstitial Gas Analysis (IGA)

Since background oxygen levels were too high in the EDS to assess the suspected quantities in ZnS, some samples were sent out for Interstitial Gas Analysis (IGA) to quantify the oxygen levels. Results were reported from the lab in weight percent oxygen which was then converted to various quantities as shown in Table 6-6. The oxygen levels should be compared with Appendix B. IGA has a detection limit of 10 ppm weight for oxygen, with an uncertainty of ± 5 to 10% for levels in the range determined here [156].

To test the assessment of the uncertainty, a second round of pieces from the same samples of each ZnS type was sent off again for IGA of oxygen and also hydrogen. IGA has a detection limit of <3 ppm weight for hydrogen, with a stated uncertainty of ± 5 %

[156]. The results of this second analysis and comparison to the first are shown in Table

6-7.

Table 6-6: Interstitial Gas Analysis first run results for oxygen redZnS

wt %

O ppmwt

O at %

O ppma

O

# density

(#/cm3)

0.20 2000 0.60 6000 1.5 x 10

20

0.13 1300 0.40 4000 1.0 x 10

20 eZnS

RHZnS msZnS

0.13 1300 0.40 4000 1.0 x 10

20

0.13 1300 0.40 4000 1.0 x 10

20

Bridgman 0.11 1100 0.33 3300 8.25 x 10

19

175

Table 6-7: Interstitial Gas Analysis second run results for oxygen and hydrogen

ppmwt O

(2 nd

run) ppmwt O

(2 nd

run) (1

Delta st

– 2 nd

) ppmwt

H

redZnS 1000 2000 1000 <3

RHZnS 880 1300 420 <3 msZnS 1000 1300 300 <3

Bridgman 1400 1100 -300 <3

RayZnS 1200 N/A N/A <3

Obviously these results vary by a factor of two for different pieces of the same

ZnS sample. However, given these larger-than-expected differences, it can still be stated that ZnS in all these forms contains 10

19

to 10

20

/ cm

3

oxygen (about 0.2 to 0.6 atomic percent). The amount of hydrogen is 10

18

/ cm

3

(~0.01 atomic percent) or less (see below).

It is important to assess whether these large oxygen concentrations could be from adsorbed oxygen on the surface of the samples. The size of the samples submitted for

IGA was approximately 1 to 2 cm

3

in volume. Taking ~10

22 atoms/cm

3 of ZnS[15], this gives ~10

22

atoms in each sample, the equivalent of about 2 grams of ZnS. The number of sites on the surface layer versus in the bulk was calculated assuming cases of “hard” spheres or cubes. Table 6-8 summarizes these calculations. Thus it can be seen that the ratio of volume sites to surface sites in these samples was at least 10

6

, so at most about

1ppm atomic of the signal could have been from surface oxygen. This confirms that the

IGA measurement, while crude, does identify oxygen as a significant bulk impurity in all forms of ZnS tested, including single crystal Bridgman.

176

Table 6-8: Volume versus surface sites for oxygen in IGA samples

Sample shape

# atoms Surface area

(atoms)

Volume of all (atoms)

Volume of interior

Cube x 6x

2.2 x 10

7

2

(x)

2.9 x 10

15

3

(x-2)

1 x 10

22

(atoms)

1 x 10

3

22

(per edge)

Sphere x 4π(x/2)

2

2.7 x 10

7

2.3 x 10

15

(4/3)π(x/2)

3

1 x 10

22

(diameter)

(4/3)π[(x-2)/2]

1 x 10

22

3

Ratio of volume to surface atoms

3.7 x 10

6

4.5 x 10

6

Estimates of the zinc-hydride concentration can be taken from Lewis[52], which estimates up to 420 ppm atomic (ppma) based on absorption coefficients, or from

Drezner [100] which estimates about an order of magnitude more (equivalent of 500 –

1300 ppma) based on mass spectrometer measurements. According to Drezner et al.

[100], hydrogen content, as measured by mass spectroscopy, decreased from 0.0027 weight% to 0.0013 weight% as the deposition temperature is increased from 650 to 770

°C. Hydrogen content was highest in ZnS material deposited near the mandrel (0.0013 weight%) and decreased toward the growth surface (0.0011 weight%). These latter differences are slight and may not indicate real differences. (Note that 6 μm absorption measurements for this current work suggested lower hydride levels at the mandrel as described in the following Chapter.) Interstitial Gas Analysis (IGA) measurements presented here suggest that hydrogen in a variety of ZnS samples contained levels <3 ppmwt (i.e. below the detectability limit). This is equivalent to saying the samples had

<150 ppma, <0.015 atomic%, or ~4 x 10

18

/ cm

3

hydrogen. These levels of hydrogen are better measured by another technique such as secondary ion mass spectroscopy (SIMS).

177

7

Characterization of Optical Properties

In this chapter I present the results of the characterization of the optical properties, which was of primary concern to this development program. First I briefly show the results of the photoluminescence measurements before launching into a long discussion of transmission measurements.

7.1 Photoluminescence

Due to the great expense of performing photoluminescence measurements, only a limited number of samples were tested. Some judicious choices were made to attempt to sample the range of possible material types. Both as-deposited CVD samples as well as heat-treated ones were tested. As mentioned in Chapter 5, all samples were tested at 300

K and 10 K with KrF excimer excitation (248 nm) and a single one with N

2

laser excitation (337 nm). A few of the early samples were tested with two gratings, one of which picked up the longer wavelength luminescence out to about 1000 nm. Peak wavelengths when indicated are listed to the first decimal place if they are sharp and without the decimal if the peak is more difficult to ascertain due to noise or broadness.

7.1.1 Spurious artifacts

Spectral artifacts due to the test environment arose in a few samples. A couple samples show emissions at very high energies above the bandgap (277 and 287 nm).

These are spurious emissions resulting from the filter cut-off around the excitation energy of the KrF laser. Many of the samples had a sharp emission peak of varying intensity around 496 – 497nm. This is identified as the second order diffracted KrF excitation

178 beam (n=2, excitation wavelength is 248nm, 2 x 248=496nm). This resulted from the limitation of the filter on the collection line to block all of the scattered excitation light.

A few samples showed a weak shoulder around 387 nm which is believed to be a result of scattered light bouncing around and hitting sample mounting materials which then emit, including a silicon jig, thermally conductive grease to hold sample in place during cryo-cooler piston vibration, and the stainless steel radiation shield package.

Finally, a second-order diffracted beam of the sharp exciton lines are visible for

ZnS samples at 2 times the wavelength of the original emission. This resulted from the limitations in the free spectral range of the grating used in the monochrometer. In one case, even the third-order diffraction of the exciton is evident.

7.1.2 As-deposited CVD ZnS

Figure 7-1 shows luminescence spectra of Raytheon produced CVD ZnS

(“rayZnS”). This was the only sample excited with the longer wavelength laser that had energy lower than the bandgap, and the data was only recorded for low temperature. The intensity recorded for this N

2

laser excitation was about four times that of the KrF excitation, but both peaked very close to each other, 450 nm (“blue”) for the N

2 excitation and 447.7 nm for the KrF excitation. The N

2

excitation shows a second wellresolved peak at 506 nm (green) which is poorly-resolved and weakly evident in the KrF spectrum at a similar position. The N

2

excitation also has a broad red/infrared luminescence with a peak about 754 nm which is much weaker and red-shifted slightly in the low-temperature KrF spectrum. A cathodoluminescence band at 850 nm has been

179 reported and assigned to sulfur vacancies [56], but the band in this sample seems to be at shorter wavelengths.

Figure 7-1: Photoluminescence in Raytheon CVD ZnS

Turning now to exclusively the KrF excitations, the most obvious difference is the inclusion of the exciton related peaks near the band edge. Low temperature spectra indicate a sharp peak at 327.2 nm which following Morozova et al. [56] is assigned to the

SA donor-bound exciton I

2

(Zn i

). The peak at 447.7 nm can therefore be assigned to the

SA(I) center or

{

O

S

* ⋅

Zn i

V

Zn

''

}

’ [54] present in materials with an excess of metal.

In the room temperature spectrum, the main band has red shifted to 495.4 nm.

The exciton-related peak at 339 nm is close to the I

1

(SAL) at 338 nm, which would

180 indicate stoichiometric composition since the prevalence of the SAL band is linked to the charge state of the oxygen complex dominating in stoichiometric compositions,

{

O

S

* ⋅

Zn i

• •

V

Zn

''

}

x

[56]. However, this spectrum lacks the related SAL band at 355 – 370 nm in the low temperature spectrum, and the binding energy for the SAL center is listed as

24.5 meV which is very close to room temperature (~25 meV). These indications suggest that the observed exciton is not related to the SAL center. Rather, it could be a red shift of the free exciton band from its usual position at 336 nm [56]. Red shifts of the room temperature exciton band up to 342 nm have been reported due to the presence of dissolved oxygen and the associated change in the band gap [55]. The longer wavelength hump at 350 nm is likely a phonon replica of this red shifted free exciton. The overall conclusions from this sample are that oxygen is present in a luminescent center in a material with excess metal.

Figure 7-2 shows the photoluminescence spectra of Raytheon produced elemental

ZnS subjected to the KrF laser excitation. The visible wavelength portion of the lowtemperature emission has collapsed into a single asymmetric peak with a shoulder. The main part of the peak is getting close to the stated SA(I) region which is listed as 445 nm at 80 K[54]. The exciton region shows two distinct sharp peaks, one of which can be assigned to the free exciton (326.3 nm) and the other one to one of the components of the acceptor-bound exciton I

1

(SA) (331.1 nm)[56]. The presence of this acceptor-bound exciton assigned to SA lends credence to the identification of the visible luminescence with the SA defect complex

{

O

S

*

Zn i

• ⋅

V

Zn

''

}

′. The room temperature spectrum shows an exciton at 337.8 nm which is red shifted due to ZnS-O contributions to the band gap. The

181 visible band at 446.8 is again asymmetric and probably related to the SA band which is still stable at room temperature (see Chapter 2).

Figure 7-2: Photoluminescence in Raytheon Elemental ZnS

7.1.3 CVD ZnS Core Sections

Next, the variation of the photoluminescence through the cores of Rohm and Haas

ZnS was investigated. The orientation of the various samples was described in Chapter 5.

Figure 7-3a shows the photoluminescence spectra of the mandrel side sample with the excitation beam oriented in the same way as all the samples described previously in this section, in the nomenclature of this project, sample “C-P.” Figure 7-3b shows the mandrel side with the samples cut so that the long axis of the “columnar” grains in the growth direction would be excited by the laser, sample “C-S.” Figure 7-4 compares the

182 low temperature spectra of the two, showing that the spectral features are similar but the intensities in the P orientation are much greater. Likewise, the room temperature spectrum for the P orientation is more intense and the peaks are more well-defined.

Figure 7-3: Photoluminescence in R&H ZnS core, mandrel side

(a, top) column tops; (b, bottom) column sides

183

Figure 7-4: Photoluminescence in R&H ZnS core, mandrel side, comparison

Luminescence spectra from the samples from the growth side of the core (i.e. the last material to grow) are shown in Figure 7-5a and Figure 7-5b. These samples are denoted “A-S” and “A-P” analogous to those previously described. Within each sample, the room temperature and low temperature emissions are very similar spectrally and in intensity. None of the spectra show exciton peaks. When comparing the two orientations, it can be seen that the “S” orientation is almost two orders of magnitude more intense. In fact, this intensity is by far the greatest of any of the samples tested.

Comparing the two samples excited in the “S” orientation, C-S and A-S, the latter shows a similar comparison in intensity to the A-P sample. Comparing the two samples excited in the “P” orientation, C-P and A-P, the peak visible intensity had red-shifted in the latter

184 and the exciton has been quenched. Figure 7-6 shows the large difference in luminescence intensity observed for the two orientations.

Figure 7-5: Photoluminescence in R&H ZnS core, growth side

(a, top) column tops; (b, bottom) column sides

185

Figure 7-6: Photoluminescence in R&H ZnS core, growth side, comparison

Luminescence spectra from the center of the core are shown in Figure 7-7.

Because the shapes and intensities were very similar for the “B-S” and “B-P” spectra they are plotted together on the same graph. In the low temperature spectrum, the S orientation is slightly less intense, but intensities are almost the same for the room temperature spectrum. Excitons are evident in all spectra, unlike the growth side mentioned previously.

A comparison of all the low temperature spectra for the mandrel, middle, and growth sides in both orientations is shown in

Figure 7-8

. The spectrum for the A-S is plotted as well, but on a different axis since it is so much more intense than the other spectra. The spectral positions for the mandrel side and middle seem to be the same, with two lobes in the visible, while the growth side seems to have the lobe at 500 nm dominating the emission.

186

Figure 7-7: Photoluminescence in R&H ZnS core, middle

Figure 7-8: Photoluminescence comparison of all the cores at 10 K

187

It was hypothesized that these samples would show somewhat different luminescence behavior, but the magnitude of the difference was much greater than expected. To further examine this phenomenon, another core from the same run was sliced longitudinally and polished on both sides. Immediately evident was the visible growth layers which appeared to the eye as bands of different color and scattering intensity. The bands started out wide and very clear and colorless at the mandrel end and became narrower and more yellow at the growth end. The visible scattering differed from band-to-band with no clear trend. A photograph taken in transmitted light is shown in Figure 7-9 along with the positions of the samples cut from a companion core from the same run for the luminescence testing. This observation lends support to the idea that the luminescence behavior is strongly influenced by layering, whether by the layers themselves or by particular optically active centers segregating to these layers or even causing the layering behavior. Internal electric fields are known from ZnS which can be very strong in areas with many faults [157]. These fields could influence the photoluminescent behavior, as they are known to play a major part in the electroluminescent behavior.

188

Figure 7-9: Layering in the cores and luminescence difference along growth direction

The picture on the left indicates the positions of the “S” samples cut from a similar core slab.

7.1.4 Hot isostatic pressed ZnS

Figure 7-10 shows the spectra collected for multispectral ZnS. The free exciton shows up at low temperatures (326.8 nm) and at room temperature (336.8 nm) without any spectral shift due to band-gap changes. There is some weak emission at 359 nm in the low temperature spectrum which may be indicative of the SAL(II) complex indicating stoichiometry [54]. Some weak features are evident in the visible portion of the luminescence which may be related to some portion of the other oxygen-related centers of SA(I) (blue) and III (green). The second order diffracted exciton is observable at

654nm (=2x327) and the third order at 986nm (close to 3x327=981nm).

189

Figure 7-10: Photoluminescence in Pt HIP Raytheon multispectral ZnS

Overall, the results of the photoluminescence measurements can be summarized as follows. The Raytheon CVD ZnS and elemental results support the idea that the oxygen defect is important in ZnS. Both samples show a red-shifted free exciton due to the change in the electronic band gap due to oxygen incorporation. The shift in this exciton position is a way to estimate the oxygen concentration. The dominance of the visible spectra of the blue SA band suggests that the important defect is

{

O

S

* ⋅

Zn i

V

Zn

''

}

′ and the materials are Zn-rich.

There is considerable variation in the luminescence observed in different positions and orientations along the CVD growth direction. This may be due to the concentration

190 of quenching impurities (by nonradiative mechanisms or charge transfer) which increases at the growth side. As shown in the following section, the concentration of hydrogen

(assumed by the absorption due to zinc hydride at 6 μm) is smaller on the mandrel side.

It is apparent from this short discussion that photoluminescence provides an extremely complicated picture of the electronic defects in ZnS. Whether or not the PL signatures can be seen as “diagnostic” or “characteristic” of a particular sample type cannot be determined without many more measurements. However, it can be seen that even within a single CVD growth run the material at various positions has substantially different electronic character, and thus defect structure. The detailed investigation of these defect structures and their significance on processing is beyond the scope of this dissertation.

7.2 Transmission Measurements

The primary assessment of usability of ZnS for multispectral infrared window applications is its transmission of light in the visible and near-infrared portion of the electromagnetic spectrum (0.4 to 2 μm wavelength). For this reason, measurements were routinely and frequently taken from the band edge in the ultraviolet to the limit of standard laboratory monochrometer equipment (in this case 2.5 μm). Infrared transmission (on FTIR equipment) was also taken in a large number of cases due to the presence of extinction in transmission in standard as-deposited CVD ZnS out to at least 7

μm wavelength. Additionally, the association of the 6 μm absorption with visible color had been repeatedly stated in the literature (e.g. [52]) and this feature is the most prominent one in the infrared spectrum of red ZnS and elemental ZnS particularly.

191

It should be noted that transmission measurements which are to be interpreted quantitatively take careful preparation and measurement protocol. It was found in the course of these studies that surface roughness and wedge in the samples had to be carefully controlled to get reproducible results especially in the visible and near-infrared.

Wedge in the part being measured was routinely kept below five minutes of arc once these difficulties were identified. Additionally, direct comparison of two samples where several percent transmission differences was of interest required polishing the samples all on the same block throughout each step of the process. The effect of a poor surface polish was shown to reduce the transmission at 1 μm wavelength as much as 10% in

0.14” thick samples. Precise surface roughness was never measured, but repolished parts had significantly improved transmittance without having appreciably changed thickness.

Finally, tilt in the measurement caused repeated problems and special fixturing had to be devised especially for the thin single crystal samples. In the long-wave infrared, the effect of unintentional tilt (i.e. the effect of slight differences in how the sample was held in the fixture) was shown to vary the transmission measurement by 4% and sometimes more across the entire spectrum. In the visible, tilt was readily identified by the periodic oscillations in the transmission due to the etalon effect.

One quick way of assessing the quality of the measurement was to plot the measured data versus the maximum transmittance based on the index of refraction [125]:

T

max

=

n

2

2

n

+

1

192 where T is the transmittance (a number between zero and one) and n is the index of refraction which is a function of wavelength. Tabulated index data or those based on fitted Sellmeier dispersions were used. In this prediction the transmittance is only based on the reflection from the two surfaces and extinction by absorption and scatter are neglected. Frequently some of the FTIR measurements indicated transmittances higher than the theoretical maximum, indicating a measurement with some tilt.

Finally, the spectral artifacts due to the measurement were frequent and had to be identified. In the UV-VIS measurement, there was a detector and grating change-over at

800 nm which gave a sometimes large discontinuity in the transmittance graph despite internal instrument corrections and calibrations. The spectrometer source change-over at

350 nm was rarely seen due the opacity of most ZnS at this wavelength. In the infrared, the strong atmospheric water absorption bands from 2 – 3 μm and from 5 – 7 μm limit the spectral resolution that can be used. By trial and error it was found that spectral resolutions of as low as 1 cm

-1

in the Thermo-Nicolet FTIR were tolerable with some

“bleed through” of the water bands that were not corrected by the background correction.

The default setting is a spectral resolution of 4 cm

-1

and was generally adequate.

Additionally, a strong CO

2

band at 3.4 μm wavelength occasionally does not get corrected or gets overcorrected by the software. This happens due to differences in CO

2 levels from when the background and sample measurements were taken. Finally, since the UV-VIS-NIR and FTIR spectrometers have an overlap in spectral range in the nearinfrared from 2 – 2.5 μm wavelength, in more cases than not there is a disagreement by a few percent to as much as ten percent in transmission between the two types of

193 instruments. Apparently this is common and spectrometer manufacturers blame this on different electronics processing and in some cases optimization of gratings and other components for certain spectral ranges. There is also no doubt a component of tilt in this difference which was difficult to control when moving samples between instruments.

As much as possible samples were polished to similar thicknesses so direct comparisons could be made using transmission graphs. Since the samples were never exactly the same thickness, and since sometimes it was desirable to assess process progression as a function of thickness, extinction coefficients were calculated and used as a metric. The wavelength of interest in this case was 1.064 μm, the Nd:YAG solid state laser line. The extinction coefficient was calculated as follows:

α

=

l

L

* ln

( 1

2 (

T

R

)

2

/ 100 )

+

R

2 +

( 1

R

)

4

4 (

T

/ 100 )

2

⎦ where α is the extinction coefficient (in cm

-1

), L is the sample thickness, T is the measured transmittance in percent, and R is the single surface reflectivity in air calculated from the index of refraction (real part n and imaginary part κ which can be neglected in this low-absorbing region):

R

=

(

n

1 )

2

(

n

+

1 )

2

+

κ

2

+

κ

2

(

n

1 )

2

(

n

+

1 )

2

.

The same equation rearranged is the familiar one for calculating the transmission of an absorbing window [125] as:

T

(%)

=

100 *

( 1

1

R

)

R

2

2

e

− α

L e

2

α

L

194

It should be reiterated that alpha is the extinction coefficient and thus includes the effects of both absorption and scattering. It turns out that this is a critical distinction in ZnS, as the effects of absorption at the ultraviolet edge can be eliminated by heat treatment while at the same time accentuating the effects of scattering.

7.2.1 Single Crystal ZnS

Transmission of single crystal ZnS (Bridgman melt grown) was tested alongside single crystal ZnO (hydrothermally grown) and CVD ZnSe. The ZnS crystal (110) plane sample was largely cubic, but as shown in the x-ray diffraction section had some hexagonal component to it. The ZnO crystal was a (00.1) c-axis plane, and no difference was seen in transmission coming through the zinc terminated surface first versus the oxygen terminated surface first. The ZnSe sample was polycrystalline chemically vapor deposited material. Single crystal ZnSe was not readily available, and this particular sample was thick and somewhat scratched, so the infrared transmission is lower than expected due to surface scattering. Other than this, the average transmission in the transparency region can be seen to be due mainly to reflection, since the refractive index is highest in ZnSe, followed by ZnS and ZnO.

Comparison of the transmission of zinc with increasingly heavier anions is illustrative of trends in II-VI materials. Figure 7-11 is a composite of UV-VIS-NIR and

FTIR spectrometer measurements. Some of the measurement issues, described in the previous section, are clearly evident and illustrated. Also in this figure are labels for some of the characteristic multiphonon absorptions and their assignments for ZnO [57,

158], ZnS [66], and ZnSe [159]. Note that considerable disagreement among various

195 sources was found on multiphonon assignment, particularly for ZnO [160], and preference was given to more recent articles. The effect of the heavier anion moving the infrared transmission cutoff to longer wavelengths is clearly evident.

Figure 7-11: Transmission of ZnO, ZnS, and ZnSe

Figure 7-12 shows the effect of the electronic bandgap on the ultraviolet transmission cut-on wavelength. Of the three materials, ZnS has the largest electronic band gap, followed by ZnO and ZnSe. The measured room temperature ultraviolet (UV) cut-off, defined as the last spectrometer datapoint before a negative transmission value is recorded, was 340 nm (3.65 eV) for ZnS, 387 nm (3.20) for ZnO, and 473 nm (2.62) for

ZnSe. These band gap values are slightly smaller than the usually accepted numbers (e.g.

196

[64]) but these samples were fairly thick and the ZnSe material was not the purest available.

Figure 7-12: Band edge transmission of ZnO, ZnS, and ZnSe

7.2.2 Polycrystalline samples, no heat treatment

It will be shown that CVD ZnS materials without heat treatment vary widely in their visible and near-infrared transmission and their presentation of the 6 μm absorption band. Many researchers have shown the effects of deposition temperature on the presence of the 6 μm absorption band and associated scattering, but definitive conclusions regarding the interplay of reactant stoichiometry and deposition temperature on the transmission are lacking [100, 118, 161-163].

Samples from several suppliers of standard CVD ZnS were measured for transmission, including legacy Raytheon (Raytran), Vitron, Rohm and Haas, II-VI,

197

Rafael, and Princeton Scientific. Additionally, legacy Raytheon CVD materials red ZnS and elemental ZnS were compared to current commercially available standard ZnS.

Finally, a single sample each was obtained of Chinese ZnS made by a process similar to elemental ZnS [104] and powder hot-pressed ZnS from the 1950’s made by Eastman

Kodak and marketed as IRTRAN2 [3].

It was expected that, among the various materials, there would be variation in the amount of transmission loss in the visible and near infrared (due to scattering), the visible coloration (due to position of the band edge), and intensity of the 6 μm absorption band.

These variations might be expected due to differences in impurity levels of hydrogen and oxygen, different manufacturing processes (e.g. CVD with H

2

S, CVD with H

2

, hot pressing from powders), and different heat treatments.

The main classes of CVD ZnS are compared in Figure 7-13. Materials were of comparable thickness so transmission graphs can be compared directly. The curve labeled “1” is multispectral ZnS which is CVD ZnS which has undergone a high temperature (~990 °C) HIP step with platinum foil. Note that the transmission is very high through the visible and there is no absorption around 6 μm. The curve labeled “2” is red ZnS, CVD ZnS deposited at a very low temperature (~640 °C) with a very high excess of zinc. Note the low extinction in the visible region, though the position of the band edge at ~400-415 nm accounts for the color (see Discussion). In red ZnS the 6 μm absorption band is very wide and deep and is resolved into two sharp minima. The curve labeled “3” is elemental ZnS, CVD ZnS made from reacting hydrogen and sulfur together directly before mixing with zinc vapor. Note that there is some extinction in the visible

198 and the 6 μm absorption band is present and weakly resolved into three lines (see

Discussion for details). Finally, the curve labeled “4” is standard CVD ZnS produced from commercial hydrogen sulfide gas and zinc vapor. Note the large transmission reduction that continues through to at least 10 μm and the higher-order absorption at 6

μm superimposed on the scattering losses. It has often been claimed that this is due to

Rayleigh scattering, but this is not so, as will be shown.

Figure 7-13: CVD ZnS Transmission: redZnS, msZnS, eZnS, stdZnS

Figure 7-14 illustrates the different wavelength dependencies of transmission loss encountered. The curve labeled “4” is the same standard CVD ZnS sample as in the previous figure. Curve “5” is a slightly thicker hot-pressed ZnS from powder precursors.

199

Note the different wavelength dependency of transmission loss as well as a strong extrinsic absorption at 8.9 μm. Curve “6” is the sample of Chinese ZnS which in many respects is similar to elemental ZnS (curve “3” in the previous example) with slightly better visible transmittance for its thickness but unlike eZnS has no 6 μm absorption.

Finally curve “7” is the reflection only transmission (T max

, defined earlier) which would result if losses were only due to the refractive index. It is apparent from this comparison that all three of these materials have extrinsic losses from scattering and/or absorption in the visible and near-infrared.

Figure 7-14: ZnS Transmission: stdZnS, hot-pressed powder ZnS, Chinese CVD ZnS

200

7.2.2.1 Cores

Transmission measurements were taken at various positions in the CVD growth cores described in Chapter 5 to see if there were any noticeable spectral differences from the mandrel to the growth side. The reader should note that x-ray diffraction and photoluminescence were taken at various comparable positions along the core. A slice was made along the longitudinal axis of the core (i.e. parallel to the growth direction).

The sample was optically polished to less than five minutes of wedge. A photograph of the sample is shown in Figure 7-15. What appears to be a chamfer is the result of the slab being cut from a cylinder which was core-drilled from a large plate of Rohm and Haas

CVD ZnS. The sample measured about 23 mm long by 21.5 mm wide and was 2 mm thick. Assuming an average deposition rate of 0.002” per hour as stated by Rohm &

Haas patents (USPTO #6,083,561), this core represents about 450 hours of ZnS deposition.

A series of bands is evident in the short axis direction which corresponds to sections of growth layers. Also evident on careful inspection (difficult to see in the photograph) are features which are parallel to the growth direction. These striations in both directions have been observed in single crystal vapor grown ZnS (e.g. Holt and

Culpan [150]). On the left of Figure 7-15 is a back-lit photograph and on the right a similar photograph with lines added to aid the eye in identifying the bands. Assuming the deposition rate above, the three bands are estimated to be composed of about 96 hours of deposition, with the next two being about 40 hours each. Also shown is a representation of the size of the aperture and indication of the various points used to first investigate the

201 transmission of these various sections (~4.5 mm diameter circle). Visual inspection showed that the first section next to the mandrel (labeled 1) appeared colorless with low scatter while the section right next to it (labeled 2) was yellower and more scattering.

This sample was analyzed in transmission using both UV-VIS (175 – 3300 nm) and FTIR (2 – 20 μm) spectrometers. In the first round of measurements, a maximum deviation between the highest and lowest transmission in the UV-VIS was 5% (between regions 1 and 2), and the largest deviation measured in the FTIR was 3% (again, between regions 1 and 2). These deviations were small and likely to be confounded by problems with tilt, especially since the part was difficult to fixture such that it was held flat and the beam was sampling the correct portion. In fact, a quick check using calculated reflection loss only (T max

) indicated much lower long-wave infrared transmission than expected.

Therefore a new set of corroborative measurements on a different set of spectrometers was commissioned to independently investigate this phenomenon.

Figure 7-15: Slab of ZnS cut from the core along the growth direction

202

The second set of measurements sampled the same bands as before but with a smaller aperture (1.7 mm diameter). In the UV-VIS the maximum deviation was about

5% (this time between regions 1 and 5). The absolute values of the transmission in the two iterations were comparable though not identical. The second set of FTIR measurements, however, were much different and on average 15% higher than the first set. This is more in-line with what was expected, so the sample was likely tilted during the first set of FTIR measurements. The transmission spread in the second set of FTIR measurements averaged 1% among different regions on the ZnS sample except for the 6

μm absorption region where transmission differences were up to 3%. Transmission was calculated for a 2 mm slab using recent absorption coefficients for 8 – 14 μm derived from NIST data [164]. Even assuming a reasonable error in the thickness measurements, transmission at 8 to 9 μm where the multiphonon absorption is not yet strong was about

1% below where it should have been assuming no scattering. This could be residual tilt in the measurement or could be due to real scattering effects even at these long wavelengths.

It was hoped that by comparing the two measurement data sets that some repeatability and confidence in the interpretations could be made. The ordering (highest to lowest transmission) was not comparable between the first and second set of UV-VIS measurements so nothing definitive can be said about the relative visible transmission in the various bands. Comparing the two sets of FTIR measurements did prove useful, however. In both cases the area closest to the mandrel (labeled 1) had the smallest 6 μm absorption and had 3% higher transmission than the region with the largest 6 μm

203 absorption. In the first set of FTIR measurements, the other 6 μm absorptions were all clustered together, with about 3% lower transmission than region 1. In the second set of

FTIR measurements, the spread was more even.

Thus careful FTIR measurements corroborated by this author and a research partner on two different instruments showed that the absorption around 6 μm was lowest in the colorless striation by the mandrel and increased in intensity toward the growth side.

Differences between other bands or other general trends could not be ascertained as the error in the measurement was on the order of the differences in the bands. Additionally, if losses were due to scattering, differences in lateral position on the sample may have influenced results. Despite these difficulties, it can be definitively stated that at least with this sample, there was less hydride present in the area directly in contact with the mandrel.

The reasons for this difference on the mandrel side are not totally clear. Rohm and Haas has stated in their patents (USPTO 6,083,561) that the initial mandrel temperature is higher for up to the first 20 hours and that the initial zinc retort temperature (and thus zinc concentration) is lower for up to the first 90 hours. It is conceivable given that the first band is estimated to be about 96 hours of deposition that the change at this point is due to a change in the stoichiometry of the vapor at this point to one which is more zinc-rich. A more zinc-rich vapor would favor more zinc hydride formation, as seen in the example of red ZnS. The cause of termination of the second and subsequent bands can only be speculated. It is possible that gas phase contaminants (e.g. oxygen or carbon dioxide levels changing to do pumps going on or off, leaks, etc.) from

204 the process have caused local chemical differences. The narrow bands during the final part of the deposition could be interpreted as being caused by the final bit of zinc raw material being evaporated which would be enriched in contaminants. Of course these are only speculations at this point and further investigation is required.

It had been reported that the transmission of CVD ZnS was slightly better measured perpendicular to the growth direction than measured parallel to the growth direction (what is normally measured) [165]. So the slab measurements discussed just previously were compared to the CVD ZnS from the same supplier (Rohm and Haas) measured in the other orientation. Because samples were not the same thickness, the extinction coefficient was calculated as described previously. Only one section of the slab was chosen for comparison, that of region three which appeared to have the most consistent data between the UV-VIS and FTIR wavelength ranges.

Data shown in Figure 7-16 is limited to the wavelength range of 0.5 – 14 μm where the Sellmeier equation for refractive index, given in Harris[125], is valid. It can readily be seen that the extinction measured in the direction perpendicular to the growth direction (i.e. perpendicular to the “columns”) seems to be lower. It is currently not understood why this is the case or if it is just an indication of the variability in the material. One could speculate that the crystallographic texturing (i.e. of the hexagonal stacking) or preferential segregation of impurities causes this difference, but a definitive understanding of this effect awaits further research.

205

Figure 7-16: Extinction parallel and perpendicular to the CVD growth direction

7.2.3 Heat treated samples

In this section transmission curves are presented for ZnS materials which have undergone various heat treatments as described in Chapter 5. First the samples heat treated without metal are described. Here I have lumped those merely annealed with those hot isostatic pressed. The annealed samples include those annealed in vacuum in a graphite furnace, those annealed in argon in a quartz tube furnace, and a single sample annealed in hydrogen. All of the hot isostatic pressed samples were processed in argon at

30ksi unless otherwise noted.

The effect of heat treatments on transmission of ZnS can be separated into two effects: changes in absorption and changes in scattering. Changes in absorption can be identified by a change in the ultraviolet cut-on wavelength near the band gap. Scattering

206 can be seen as an overall reduction in transmission over a broad wavelength range. More quantitative discussion of scattering will be presented later.

7.2.3.1 Heat treated without metal

In order to compare the effects of temperature on the change in transmission of

ZnS, transmission curves of various samples were plotted together as a function of processing temperature. Since not all samples were the same thickness, the extinction coefficient at 1064 nm was also calculated and shown on the graph for some examples.

7.2.3.1.1 Red ZnS

A series of samples of red ZnS were subjected to vacuum anneal treatments from

650 to 850 °C as described in Chapter 5. The purpose of these experiments was two-fold.

First, there was a desire to confirm some hypotheses about the 6 μm absorption that were suggested by the ab initio calculation data. It was thought that the 6 μm absorption due to zinc hydride would change its spectral position and number of sharp absorptions based on its level of hydrogen saturation. It was assumed that annealing would enable hydrogen release and a change in this characteristic absorption. Since red ZnS had the strongest and most well-defined absorption here, it seemed the right candidate to explore.

Second, there was some anecdotal evidence that red ZnS visible transmission changed differently under annealing conditions than CVD ZnS did. There was a desire to explore this aspect as well, since red ZnS had the longest wavelength ultraviolet edge of any of the samples tested.

207

Figure 7-17 shows a summary of the transmission measurements for these annealing experiments along with photographs of some of the samples before and after the heat treatment. All three samples had similar starting transmissions so only one

(Red1) is presented. All three samples were approximately the same thickness.

Red1 was subjected to a high temperature for a short time, 850 °C for 2 hours

(S2), resulting in a complete elimination of the 6 μm absorption, a large scattering from the visible through the infrared, and a band edge blue-shift despite strong scattering.

Red2 was subjected to a low temperature for a long time, 650 °C for 24 hours (S3), resulting in a some blue-shift of the band edge, similar scattering in the visible, an unidentified broad absorption centered at 4.7 μm, and some sharp fine structure from 5 to

8 μm but with elimination again of the hydride absorption. Red3 was subjected to the same low temperature, 650 °C, but for successive small amounts of time, first 2 hours

(S4), and then an additional 4 hours for a total of 6 hours (S5). After the first two hours

(S4) very little if any change was evident in the infrared but the UV band edge blueshifted slightly. The slightly higher long-wave transmission in this sample is due to slight tilt or wedge. After the next four hours (S5) again there was very little change in the infrared with the hydride absorption remaining deep, sharp, and two pronged, with no spectral or intensity change. Visible transmission degraded again and the band edge further blue-shifted.

208

Figure 7-17: Transmission and photographs of annealed Red ZnS samples

The most striking change in these samples was their change in visual appearance.

In the case of both Red1→S2 and Red2→S3 the reddish-yellow color was eliminated and the samples became whitish yellow and highly scattering. In the case of Red3, the first 2 hour anneal (S4) changed the sample from reddish to orangish, with the edges of the disk being more effected than the bulk. The subsequent 4 hour anneal (S5) changed the sample to a colorless material which was still fairly low scatter despite the reduced transmission from the band edge to about 3 μm. This is interesting since the color in the sample is gone but the 6 μm absorption generally associated with zinc hydride remains.

Most authors have claimed an intimate relationship between the hydride absorption and

209 the color of ZnS (e.g.[52]). An etched microstructure was taken of S5 which revealed no recrystallization and minimal if any coarsening.

A closer look at the band edge transmission in these samples provides additional information. Figure 7-18 shows the band edge (< 500 nm wavelength) in these samples at two scales, the first from 0 – 40% transmission and the other from 0 – 1% transmission. From the larger scale graph it is apparent that there are two phenomena taking place from the heat treatment. Firstly, in every heat treating case the ultraviolet band edge (defined as the last transmission measurement >0 on the short wavelength side) is blue-shifted. Additionally, some scattering is beginning to take place, most evident in the change from Red1 to S2 where the overall transmission is reduced in the visible even though the band edge is blue-shifted. The close-up of the band edge shows the reason for the coloration in red ZnS. Samples with a reddish hue have band edges around 400 to 420 nm, absorbing the violet most strongly and hence appearing red. By way of comparison, standard ZnS which is yellowish typically has a band edge of around

375 nm which transmits all the visible to some extent but transmission in the violet is low.

210

Figure 7-18: Transmission at band edge of annealed Red ZnS samples

(a, top) showing the change in transmission as a function of a wide wavelength range in the in UV and visible; (b, bottom) showing only the band edge

211

Finally, it was considered whether these samples (specifically S3) exhibited what had been reported as evidence for ZnO precipitates. Since these samples were initially deposited at low temperature (~640 °C) and at high zinc excesses, it was suspected that there could be more oxygen incorporation. Such was found to be the case in IGA measurements (Chapter 6) where the red ZnS sample had about 30% more oxygen than the standard ZnS. Morozova et al. [55] recently reported that the 6 μm absorption in

CVD ZnS had not been assigned (presumably because they missed the work of Lewis et

al. [52] or it had never been translated into Russian). These authors then assigned the absorption to “imperfect ZnO” and claimed that fine structure arising from 5 to 8 μm after annealing CVD ZnS [55] or CVD ZnSe [57] was due to multiphonon absorption of

ZnO precipitates in these materials.

After careful consideration, measurements, and analyses I have to reject these assertions. Morozova et al. [57] measured single crystal ZnO (details about thickness and manufacturing method not given) and assigned several sharp absorptions in the 6.2 to

7.1 μm spectral range to multiphonon processes in ZnO. In the ZnO crystals obtained and measured for this work, none of these absorptions were observable at room temperature.

Furthermore, these putative ZnO absorptions lie right on top of the strong water lines from 5 to 8 μm that are a constant problem in FTIR spectrometers even with purging. The assignment of this fine structure to ZnO at these wavelengths in ZnSe [57] and ZnS [55] I find dubious. The transmission data presented by these authors for this spectral region after annealing resemble closely the measurements taken for this current

212 work. However, this fine structure was found in nearly all the FTIR measurements of

ZnS, with a few exceptions. In these exceptions the sample was highly scattering in the spectral region where the fine water lines would have been, and so any fine structure was overwhelmed by scattering losses. Even samples with strong 6 μm absorptions showed the “fine structure” supposedly due to ZnO, appearing on both the “blue wing” and the

“red wing” of the broad absorption from the hydride. It is therefore my assertion that these reported features are measurement artifacts and not evidence for ZnO precipitates in

ZnS and ZnSe. It is possible that these precipitates exist, but I contest the identification of these lines as being evidence for ZnO until more convincing data is presented.

Figure 7-19a shows an example of two samples (there were many others) which show the structure in question. Figure 7-19b shows one of these superimposed on a typical background spectra of sharp atmospheric water lines. If in fact there is evidence of weak ZnO multiphonon absorptions in these spectra and those in the literature, it is insufficiently resolved from the background corrections accomplished by typical spectrometers. Further evidence that this is a background correction artifact is given by samples which show the inverse of the “absorptions” (i.e. in the opposite direction).

Again, it is possible that some of these lines are not artifacts, but even those presented in the literature are surrounded by measurement artifacts and thus much more careful measurements (i.e. low temperature, nitrogen purged for a long time) would be required to fully ascertain their importance.

213

Figure 7-19: Putative evidence for ZnO precipitates and comparison with water lines

(a,top) Samples with the strongest structure in the 5 – 8 μm region. (b,bottom) two samples which show influence of water lines on the transmission especially around 6.2 –

6.4 μm.

214

7.2.3.1.2 Standard ZnS

Figure 7-20a shows the effects of heat treatment of CVD ZnS at 700 °C on UV-

VIS transmission. A representative untreated starting material has an extinction of 2.45 cm

-1

at 1064 nm and an absorption edge of 374 nm. The same sample annealed in H

2

has increased its extinction coefficient at 1064 nm to 4.64 cm

-1

but moved its absorption edge to 340 nm. The other four samples, annealed for 48 to 96 hours in vacuum, show slightly worse extinctions than the starting material (~3.4 to 3.5 cm

-1

), but the UV edge has moved to 359 – 367 nm. In the case of the H

2

anneal, the shape of the curve is altered dramatically, whereas in the vacuum anneals the effect is more subtle.

Figure 7-20b shows the heat treatments at 750 °C. One sample shows significantly lower extinction than the baseline, Co32_5 which was an annealed legacy

Raytheon CVD sample. Though the extinction coefficient for this sample is 0.91 cm

-1

, some of the Raytran starting materials tended to be very clear to begin with, and so this result should not be taken as representative of the typical effects with other starting materials. However, some Raytheon CVD material has measured starting coefficients of up to 3.11 cm

-1

which is higher than the Rohm & Haas baseline material. The “before treatment” transmission of this particular sample was not measured but there is some indication that the Raytheon starting material may “clarify” better than some of the other starting materials, which may be indicative of the nature of the starting defect population.

The slope of the scattering in the as-annealed sample is very similar to the H

2

anneal sample in the previous figure.

215

Figure 7-20: Transmission of ZnS heat treated without metal: 700 °C & 750 °C

(a,top) 700 °C, (b,bottom) 750 °C

More typical 750 °C annealing results are Co32_1 and Co32_2. These samples were both annealed for 100 hours but show vastly different amounts of scattering for reasons still not clear. The hot isostatic pressed samples HIP33none and HIP57none show a very different shape to the scattering which more closely resembles the original untreated curve. As described in Chapter 6, HIP57none was shown to have partially recrystallized. The 1064 nm extinction coefficient and UV edge for these is similar, with

216

HIP33none at 2.56 cm

-1 and 372 nm and HIP57none at 2.82 cm

-1

and 379 nm. Here the extinction coefficient is useful since inspection of only the transmission indicates that the latter sample is much lower in transmission but was also much thicker. No attempt has been made to correct the UV edge for thickness but there is no doubt that thinner samples would have a smaller wavelength edge by a few nanometers.

Figure 7-21a shows the annealing treatments at 800 °C. The sample annealed in the graphite vacuum furnace shows very large extinction of 10.74 cm

-1

but the UV edge has blue-shifted to 355 nm. Two samples of the same starting material and thickness annealed for the same time in the quartz furnace have improved their transmissions, however. Co34_2 has an extinction of 1.61 cm

-1

and a UV edge of 346 nm while Co34_1 has an extinction of 1.86 cm

-1

and a UV edge of 334 nm. It is currently unknown why the quartz tube furnace produces better parts, but could be due to an increased evaporation and loss of sulfur in the vacuum furnace.

Figure 7-21b shows the annealing treatments at 850 °C. At this temperature, standard ZnS is converting well and transmission is high. Microstructures of this type of sample were shown in Chapter 6 as having complete recrystallization. Co33_3 has an extinction of 0.53 cm

-1

and Co33_1 is 0.60 cm

-1

. The only sample which did not fare well at this temperature was the red ZnS which significantly degraded its transmittance resulting in an extinction coefficient of 9.85 cm

-1

. However, time for this red ZnS treatment was only two hours, and longer times may have brought further improvement.

The UV edge on the red ZnS changed significantly, from a starting value of 400 nm to a post-treatment value of 345 nm.

217

Figure 7-21: Transmission of ZnS heat treated without metal: 800 °C & 850 °C

(a,top) 800 °C, (b,bottom) 850 °C

Figure 7-22 shows the heat treatments at 900 °C, both vacuum anneal runs and hot isostatic press runs. The annealed samples have a higher extinction than those treated at 850 °C, but the UV edge has moved to 343 – 344 nm. These samples are thicker than the ones treated at 850 °C and they were treated for 32 hours rather than 24. It is possible that the conversion has not completed throughout the thickness or that the higher temperature has lead to more evaporation and sulfur loss. The HIP’d parts are as thick or

218 thicker than the annealed ones at this temperature, but in contrast all show very high transmissions with extinctions 0.14 – 0.17 cm

-1

and UV edges at 345 nm. HIP55A was etched and shown in Chapter 6 to be fully recrystallized and microstructurally indistinguishable from the best samples HIP’d with metal. From these results it can be concluded that the HIP provides additional mechanisms for transmission improvement, either by aiding recrystallization, coalescing any pores, or allowing the microstructure to evolve without significant evaporation.

Figure 7-22: Transmission of ZnS heat treated without metal at 900 °C.

These results of heat treating without metal can be summarized as follows, with general reference to the 1064nm extinction coefficient changes. At 700 °C (anneal), extinction coefficient increased for all samples with treatments up to 96 hours. The shape of the transmission curves did not change with the exception of the hydrogen annealed one which had a sharp transmission improvement at the band edge and then recovered the slope of the other samples indicative of scattering. At 750 °C (anneal) up to 100 hours,

219 all extinctions increased over as-deposited material, except the one sample of Raytheon

CVD ZnS which may or may not have had better transmission to begin with

(unfortunately it was not measured prior to heat treatment). The shape of these curves did change with transmission being improved at the shortest wavelengths near the band edge but made worse at longer wavelengths. Hot isostatic pressing at 750 °C essentially did not change the extinction or the shape of the curve. At 800 °C (anneal) up to 32 hours, extinction increased over the as-deposited material for the sample in the graphite vacuum furnace but improved for the two samples in the quartz tube furnace with flowing argon. All samples showed an increase in transmission at the band edge as at the lower temperatures. At 850 °C (anneal), the two samples of standard ZnS annealed in the vacuum furnace for 24 hours showed substantial improvement in transmission at all wavelengths measured (i.e. up to 3 μm). The red ZnS sample which was annealed for only 2 hours showed a substantial decrease in transmission. Finally, the samples annealed at 900 °C for 32 hours again showed band edge improvements but scattered like samples annealed at 800 °C. Hot isostatic pressing at 900 °C for 32, however, produced superior results for the same thickness or thicker samples.

All these results suggest that temperatures of at least 850 °C are required for transmission improvement, which is hypothesized to require complete recrystallization

(see following sections). The samples which did not improve on annealing at 900 °C were about 1 mm thicker than the samples which did improve on annealing at 850 °C

(~3.5 mm versus ~2.5 mm). It could be, then, that the recrystallization had not completed in the allotted time for the 900 °C anneal. It has been shown that HIPing, on the other

220 hand, results in more repeatable improvement in transmission, presumably due to more rapid recrystallization due to the additional driving force of pressure to move dislocations and create the martensitic transformation from hexagonal to cubic stacking. This suggests that the activation energy for dislocation motion favors temperatures at least 850

°C. There may be an upper limit to the usable temperature in the vacuum anneal furnace due to the high vapor pressure of ZnS, since it should be recalled that 990 °C in the vacuum furnace for 10 hours completely evaporated the ZnS samples. Thus the HIP also allows processing at higher temperatures than would be possible in the vacuum furnace, due to the overpressure confining the ZnS from evaporating. Additionally, the thickness dependence observed for transmission improvement supports a diffusion-limited mechanism where a species, such as sulfur, has to diffuse to the surface. However, the diffusion would be impeded in the high-pressure HIP, so there may be a trade-off between the additional driving forces of pressure and the limitations imposed of outdiffusion.

7.2.3.2 Heat treated with metal

In this section I describe the results of the heat treatments in the presence of metal. Annealing was conducted with sputter deposited cobalt on one or two sides. Hot isostatic pressing (HIPing) was conducted with sputter deposited cobalt, silver, copper, or nickel on the surface of CVD ZnS. HIPing was also performed with the ZnS in contact with foils of platinum, cobalt, silver, iron, copper, or nickel as described in Chapter 5.

221

7.2.3.2.1 Pt series

Platinum foil HIPing is the baseline process for making multispectral ZnS. In practice, there is a certain amount of optimization in treatment time that is necessary to adjust for sample thickness. This series of experiments explored the sensitivity of thickness and starting material supplier to time and temperature. As a reminder, commercially HIPing of ZnS is performed around 990 °C, slightly below the temperature for transformation to wurtzite phase (~1020 °C).

There is considerable insensitivity of the platinum foil process to many of the variables considered here. It was found during the course of this study that even moderately thick samples of almost 5 mm could be successfully HIP’d at 750 °C for 16 hours and with much shorter times for thinner parts. “Successfully” here is defined as an extinction coefficient at 1064 nm of 0.1 cm

-1

or better. Transmission of a number of these samples along with a few of their microstructures is shown in Figure 7-23.

Microstructures look similar from 750 °C on up to 990 °C with the possible exception of the 990 °C, 10 hour microstructure which appears to have slightly larger grain size. All the samples etched from the platinum foil series (six total samples etched) had very similar microstructures and appeared fully recrystallized, as described in Chapter 6.

The transmission curves were all very similar at 1064 nanometers. Small deviations in either transmission or thickness measurement can have a large effect on calculated extinction. Therefore another extinction calculation was made at 700 nm where there was more difference between the curves. The results of these calculations are shown in Table 7-1. Also included is the UV edge as ascertained by the last positive

222 transmission measurement in the ultraviolet. In general, all the UV edge measurements were similar and blue-shifted from the average UV edge before treatment.

Figure 7-23: Transmission curves for CVD ZnS HIP’d with platinum foil

Table 7-1: Extinction coefficients and UV edge of Pt foil HIP’d samples

Sample

(HIP- # -Pt)

Starting

Material

Temperature

(°C), time (hrs)

44 R&H(06)

α @ 700 nm

(cm

-1

)

α @ 1064 nm

(cm

-1

)

UV edge

(nm)

223

Platinum thus has the ability to promote recrystallization in CVD ZnS at temperatures as low as 750 °C. As shown in Chapter 6 in the discussion on recrystallization, at higher temperatures (900 °C), platinum appears to have induced recrystallization throughout the entire thickness of the part (~5 mm) even when present only on one side. Also shown in Chapter 6, the Pt foils have converted to PtS (cooperite) on the surface. This suggests that the platinum may be acting as a preferential sink for sulfur, unlike the cobalt foil which induced vapor phase re-deposition of ZnS crystals.

Thus it seems that the formation of the platinum sulfide is an important component in the promotion of recrystallization and hence improved transmission. The exact mechanisms responsible for this dramatic effect are not yet clearly understood.

7.2.3.2.2 Co series

As mentioned in Chapter 5, the investigation of cobalt was initially spurred from earlier work on the sputter deposition of cobalt followed by annealing. In general it was found that higher temperatures (800 to 850 °C) tended to more reliably produce transmission improvements but that lower temperatures (≤ 650 °C) were required to get characteristic cobalt absorptions (see, for example [166, 167] for the cobalt absorptions in

ZnS). The main absorptions of interest are the 700 nm absorption (

4

A

2

(4F) →

4

T

1

(4P)) and the 1500 nm absorption (

4

A

2

(4F) →

4

T

1

(4F))[168]. Eventually this route was abandoned when it was discovered that 850 °C annealing for 24 hours resulted in almost identical transmission improvement regardless of whether cobalt was present or not.

Much later, it was confirmed with the “interrupted HIP” experiment described in Chapter

6 that cobalt does not play an active role in recrystallization.

224

When the same types of samples were hot isostatic pressed, however, a more repeatable trend was established. When the temperature was 850 - 990 °C and 250 nm of

Co was sputtered on each side (e.g. HIP23-Co, 850 °C, 12 hours), characteristic cobalt absorptions appeared which became stronger with higher temperatures or longer times

(see Figure 7-24). At 850 °C with only 200 nm of sputtered Co (HIP48-Co1, HIP48-

Co2, HIP52-Co1, HIP52-Co2), no absorptions were evident despite the otherwise identical 12 hour HIP to the previously mentioned one (Figure 7-25). This suggests that there was insufficient cobalt to drive in to the material to create strong absorptions when only 200 nm was present on the surface to start.

The starting material was not the factor here as both HIP23-Co (cobalt absorptions) and HIP48-Co (no cobalt absorptions) were Rohm & Haas and HIP52-Co

(no cobalt absorptions) was II-VI starting material. Material HIP’d at temperatures lower than 850 °C did not show any cobalt absorptions even after 16 hours. Another good comparison is the 12 hour HIP series with 250 nm of sputtered cobalt, where 750 °C

(HIP25-Co) shows no cobalt absorptions while 850 °C (HIP23-Co) shows mild absorptions and 990 °C (HIP21-Co) shows strong absorptions. Note that cobalt absorption was produced with lower temperature annealing heat treatments (700 °C), but these samples were heated for at least 48 hours, as opposed to the much shorter runs in the HIP (see Figure 7-24).

225

Figure 7-24: Transmission of ZnS HIP’d with sputtered cobalt showing absorption

Note the lower transmission for the low temperature cobalt annealed (“Co-dope”) samples which still scatter strongly but nonetheless show cobalt absorptions.

Figure 7-25: Transmission of ZnS HIP’d with sputtered cobalt showing no absorption

Cobalt foil was next tried with hot isostatic pressing of ZnS (see Figure 7-26a). In this case, no cobalt absorptions were observed even at the highest temperatures. Good transmission improvement was achieved with temperatures of 950 °C, but in light of the other evidence it seems likely that this was due solely to the temperature effect and not to

226 the cobalt. Occasionally the parts appeared somewhat greenish afterheat treatment. In one test it was shown that a ground sample (Figure 7-26a, curve 4) clarified slightly better than an identical sample which was polished (Figure 7-26a, curve 5). Figure 7-26b shows that for identical HIP conditions, the sample with sputter deposited cobalt exhibited strong cobalt absorptions while the one exposed to cobalt foil showed none of these absorptions.

Figure 7-26: Transmission of ZnS HIP’d with cobalt foil

(a, top) Cobalt foil HIP experiments; (b, bottom) Cobalt foil versus sputtering for same heat treatment

227

7.2.3.2.3 Ag series

Since silver is very active with sulfur (i.e. tarnishing of silver), it was surmised that silver might have an effect on ZnS during hot isostatic pressing (HIPing). Both sputter deposited silver and silver foil were put in contact with CVD ZnS during HIPing.

Because the melting point of silver is 962 °C, HIPing temperatures were kept at or below

850 °C. For comparison, silver sulfide melts at 842 °C at standard pressure.

Thin (~1.5 mm) samples of ZnS were HIP’d in contact with silver foil for 16 hours at temperatures between 700 and 850 °C. The ultraviolet through near-infrared transmission of representative parts are shown in Figure 7-27 along with transmission of an untreated part for comparison. Curve “1” is an unHIP’d sample with increasing numbers representing increasing temperature treatments for 16 hours. Curve “2” is at

700 °C and shows slightly increased transmission (this was verified with extinction coefficient since the heat treated sample is slightly thicker). Curve “3” at 750 °C shows excellent transmission and was selected for further refinement. Curve “4” at 800 °C begins to show some decrease in transmission from the optimum, and curve “5” at 850

°C shows substantial degradation and the beginnings of a new loss process.

Extinction in the visible and near-infrared for samples HIP’d at 750 °C for various times is shown in Table 7-2. Qualitatively, etched surface microstructures did not seem to change from HIP times of 16 hours to 72 hours (see Figure 7-28), but extinction continued to decrease slightly for thicker samples with longer HIP times (i.e. 0.9 → 0.8

→ 0.7 cm

-1

at 700 nm for similar thicknesses at 16h (#41_2), 32h (#42_1), 72h (#56)).

Though surface etches seemed similar in all these samples, it is likely that

228 recrystallization had not completed in the interior of the thicker (>4 mm) samples with extinction above 0.1 cm

-1

. The ability to “clarify” ZnS with low temperature silver foil

HIP was notably different between ZnS from Rohm and Haas versus II-VI. Rohm and

Haas material clarified well for thin samples but poorly for thick samples, while II-VI material clarified well for thick samples. Rohm and Haas thick samples still showed evidence of 6 μm infrared absorption even after HIPs of 72 hours. While II-VI material

HIP’d for 72 hours did not show 6 μm absorption, it did show signs of the “silver absorption” in a decrease in transmission around 3.5 μm.

Figure 7-27: Transmission of ZnS (visible wavelengths) HIP’d with silver foil

229

Table 7-2: Extinction coefficients and UV edge of 750 °C Ag foil HIP’d samples

Errors (± cm

-1

) on the extinction coefficients are 0.08 to 0.10 for thin samples and 0.03 to

0.05 for thick samples based on assuming ± 1% on the transmission measurement. No assumption was made on the accuracy of the thickness measurement. High extinction numbers in some of the earlier experiments may have been due to wedge in the parts prior to specification of <5 minutes wedge.

Sample

(HIP- # -Ag)

Starting

Material

HIP time

(hrs)

Thickness

(mm)

α @ 700 nm (cm

-1

)

α @ 1064 nm (cm

-1

)

UV edge

(nm)

41_1 R&H(06) 16 1.56 0.26 0.14 342

41_2 R&H(06) 16 4.19 0.92 0.59 347

42_1 R&H(06) 32 4.93 0.84 0.54 348

42_2 R&H(06) 32 1.55 0.42 0.29 343

Figure 7-28: Transmission and micrographs of silver foil HIP experiment samples

230

Thin (~1.5 mm) samples of ZnS were HIP’d with sputter deposited silver for up to 16 hours at temperatures from 600 to 750 °C. Figure 7-29 shows the visible wavelength portion of the transmission results. Curve “A” is the theoretical reflection only, maximum transmission (Tmax), curve “F” is the best Ag foil transmission (same sample as curve 3 on the previous graphs), and curve “asD” is the as-deposited no-HIP sample (same as curve 1 on the previous graphs). Curve “NF” represents the same HIP conditions as the best foil same (curve “F”) but without any silver foil being used. Curve

“6” is a 700 °C sample with sputter deposited silver, and curve “7” is a 600 °C sample with sputter deposited silver. As can be seen, the visible transmittance of curves “asD,”

“NF,” “6,” and “7” are virtually indistinguishable. Curve “8” HIP’d at 750 °C shows a marked decrease in transmittance. As these samples are all similar thicknesses, the transmission curves can be compared directly.

Figure 7-29: Transmission of ZnS (visible wavelengths) HIP’d sputtered silver

231

A composite transmission graph is shown in Figure 7-30 including visible through the infrared and a few key micrographs. It can readily be seen that curves for “as-D,”

“6,” and “7” are so similar that only “7” is labeled for clarity. The micrograph of “7” shows that at 600 °C with sputtered silver there is essentially no microstructural change and no recrystallization from the as-deposited material. Also the 6 μm absorption is still present in both the 600 °C sputter (curve “7”) and the 700 °C sputter (curve “8”). The latter was not etched but judging from the transmittance it is suspected to be essentially identical to that shown for the 600 °C sputtered silver (HIP34-Ag). Further, the 700 °C

Ag foil curve (curve “2” in Figure 7-28) shows very little change from the no-HIP sample as well, indicating that it too is likely unrecrystallized. Curve “8” shows that not only is the visible transmittance decreased as shown in the previous figure, but also the infrared transmittance is decreased by 5 to 10% from the other curves, becoming less different at longer wavelengths. The microstructure of this sample revealed an excellent example of a partially recrystallized microstructure as shown also in Chapter 6.

Several preliminary observations can be made comparing the silver foil to the sputtered silver transmission results. First, it seems that temperatures of at least 750 °C are required to initiate recrystallization in ZnS with Ag. For the Ag foil at 750 °C (curve

“F”), recrystallization is complete after 16 hours while for the sputter deposited Ag at 750

°C (curve “8”) only partial recrystallization has taken place. It is likely that 750 °C without any foil (HIP33-none for a similar thin sample, curve “NF”) is insufficient to cause any recrystallization since the transmission has not degraded in the visible.

232

To summarize, at 750 °C without any metal, no recrystallization will take place after 16 hours. With sputtered silver, recrystallization will begin but not finish after 16 hours, presumably because of some interference of in-diffusing silver with the recrystallization process. With silver foil, recrystallization will start and finish in 16 hours resulting in very high transmission though the visible and infrared. Partial recrystallization results in increased scatter in the visible and infrared. This can be seen in the 750 °C sputtered silver run (HIP25-Ag, curve “8”) where the loss mechanism is large particle scattering (i.e. wavelength independent scattering treated by geometrical optics[169, 170]) due to the different composition of recrystallized grains until the intrinsic multiphonon processes set in around 10 microns.

Figure 7-30: Transmission of ZnS (infrared wavelengths) HIP’d sputtered silver

Note the different magnifications for the 2 micrographs on the left versus on the right.

233

7.2.3.2.4 Others (Fe, Ni, Cu)

A limited number of HIP experiments were done with other metals including Fe,

Cu, and Ni. The results of the recrystallization experiments at high temperature and the analysis of the iron and copper foils were presented in Chapter 6. This section shows the spectral changes of ZnS when hot isostatic pressed with these other metals.

Figure 7-31 shows the results of several experiments done with iron foil. As can be seen, the 750 °C 16 hour HIP (curve “4”) has essentially no better transmission than a similar thickness sample before HIP. 850 °C (curve “3”) begins to show improvement and 900 °C (curve “2”) and 950 °C (curve “1”) show considerable improvement.

Considering the results discussed in Chapter 6 regarding the “interrupted HIP” microstructural cross sections and the lack of an iron sulfide on the foils, it is believed that these transmission improvements in curves 1, 2, and 3 were a result of temperature alone and not any action on the part of the iron foil. As was shown in the previous case for silver, 750 °C without an active recrystallization promoter (like silver) does not produce significant recrystallization (and hence transmission improvement) while temperatures of 850 – 900 °C do produce recrystallization.

These iron foil experiments can essentially be considered “no metal” experiments.

In this case proper comparison is with the “no metal” anneal and HIP experiments at 850

°C and above. Also shown on Figure 7-31 is a sample HIP’d with identical conditions of curve “2” except with no metal. It is apparent that this curve of the sample HIP’d with no metal (HIP53none, curve “NF”) is indistinguishable from the one HIP’d with the iron foil

(HIP53-Fe, curve “2”). The sample HIP’d at 850 °C has extinction at 1064nm of 0.87

234 cm

-1

, which is comparable to results from annealing runs with no metal (Co33_3 had 0.53 cm

-1

and Co33_1 had 0.60 cm

-1

, and both were thinner parts). This reinforces the notion that it is temperature and not any action of the iron which is causing transmission improvements here. These improvements are believed to be related to the recrystallization which, without an active metal, begins at 850 °C and substantially complete at 900 °C with either annealing or HIPing.

Figure 7-31: Transmission of ZnS HIP’d with iron foil

In contrast to the above example of iron, copper and nickel are believed to be

“active” metals in that they participate in starting the recrystallization of ZnS during heat treatment. A few experiments were conducted with sputter deposited copper and copper foil and a single experiment was conducted with sputtered nickel. As mentioned in

Chapter 6, nickel foil was also used in the “interrupted HIP” experiment and shown to be active.

235

Figure 7-32 compares sputter deposited nickel HIP transmission to sputter deposited cobalt HIP transmission. The absorption bands of nickel are strong, as they are for cobalt, and some of them are in similar spectral locations. These nickel absorption transitions in ZnS have been reviewed in the literature[166, 171]. The UV edge in this nickel doped ZnS is 364 nm, but the transmission is essentially zero below about 440 nm with a long exponential tail.

Figure 7-32: Transmission of ZnS HIP’d with sputtered nickel

(a, top) comparison of absorption bands from cobalt and nickel in ZnS; (b, bottomw) comparison of UV edge for as-deposited ZnS and nickel-doped ZnS

236

Only a few copper experiments were conducted and definitive conclusions await more experimentation. Figure 7-33 shows a couple of the copper experiments. Curve

“1” is a thin sample which was HIP’d at the low temperature conditions (750 °C, 16 hours) with some copper foil that was found as excess in the laboratory. Excellent transmission was achieved and the foil turned black with copper sulfide as described in

Chapter 6. When new foil was obtained and an identical experiment was performed on a thicker part, the part came out black even after polishing. The companion sample with it which had sputter deposited copper (curve “2”) showed some slight degradation in transmission.

Some deviation from the expected absorption edge slope can be seen in the sputtered sample. This is probably due to incorporation of some copper, and the UV edge is at 399 nm. A long exponential tail is evident in the sputtered copper sample similar to that shown for the nickel sputtered, with less than 1% transmission occurring below 460 nm. This is not the case for the copper foil sample, in which the transmission comes sharply down at 341 nm.

Figure 7-34 shows five different metals compared at the same low temperature condition, 750 °C for 16 hours. All samples were fairly thick (~4.5 mm). Three of the samples were processed in the same HIP run (HIP60) while the other two are comparable conditions. It is apparent that the platinum foil (curve “1”) is the most active and provides the most transmission improvement in the visible and infrared. Next best is silver (curve “2”) which is not quite as good as platinum. Note that the infrared Ag measurement seems low and dips in the aforementioned “silver absorption.” The next

237 three curves— cobalt foil (curve “3”), iron foil (curve “4”), and as-deposited CVD ZnS with no heat treatment (curve “asD”)—are essentially identical in transmission. The last curve, sputter deposited copper (curve “5”) shows lower transmission. This curve is similar to HIP25-Ag which is partially recrystallized. As copper is known to be active at these temperatures, it is likely that this thick sample is partially recrystallized and hence has reduced visible and infrared transmission that would improve on full recrystallization.

Figure 7-33: Transmission of ZnS HIP’d with sputtered copper and copper foil

(a, top) comparison of transmission; (b, bottom) comparison of UV edge

238

Figure 7-34: Transmission curves compared for 750 °C, 16 hours with various metals

7.2.4 Discussion

It is clear from the aforementioned data that the changes in transmission of ZnS with heat treatments are accompanied by many effects. Without exposure to metals,

CVD ZnS can undergo recrystallization beginning at 850 °C and above. With platinum or silver foil recrystallization promoters, this recrystallization temperature can be lowered to 750 °C with complete recrystallization achievable for thick (~5 mm) samples within 16 hours. No attempt was made to find a lower limit for the time of recrystallization or for a lower limit in temperature. It is suspected that temperatures above the original deposition temperature (in this case probably 670 to 720 °C) are necessary to move “frozen in”

239 defects. It was shown that 600 °C was inadequate for sputtered silver to initiate any recrystallization.

For pairs of materials in which both sputter deposition and foils were used, foils were more effective. For sputtered silver, 750 °C was only just enough to start recrystallization as may be the case for sputtered copper as well. This may be due to indiffusing metal interfering with the ZnS recrystallization process. Sputtered materials were shown to be more likely to incorporate into the material and cause sharp absorptions

(e.g. Co, Ni) or broad absorption (e.g. Ag).

Transmission losses in the visible, other than sharp absorption lines like in Co and

Ni, can be ascribed mainly to scattering. However, there is a discernable shift in the UV edge after heat treatment that can be ascribed to an electronic defect state in the band gap of as-deposited CVD ZnS which is removed after heat treatment. It is often noted that heat treated samples (even without any metal) change from the yellowish color to white or clear after annealing or HIPing.

Broad band extrinsic transmission loss in the infrared can be due to either large particle scattering, free carrier absorption, or both. Particular extrinsic impurities with vibrations in the infrared, such as zinc hydride stretching or even zinc oxide lattice modes, can produce localized impurity absorptions. Even when these combine with phonons they do not produce as broad extinction as some of the losses noted previously.

Now I explore these loss mechanisms in some more detail.

240

7.2.4.1 Extrinsic impurity absorptions

The most prominent absorption feature in as-deposited CVD ZnS materials is, of course, the 6 μm absorption which has historically been assigned to zinc hydride. It has already been discussed how the evidence for ZnO around this spectral area is dubious and confounded with atmospheric water lines. Known absorptions from incorporation of metal ions such as nickel and cobalt show up in samples where the metal was sputtered on the surface and diffuses into the lattice.

In CVD ZnS the main infrared feature is the broad absorption with one to three resolved minima at 5.9, 6.0, and 6.1 μm. This absorption was investigated closely. It was found to consist of a doublet in red ZnS and a triplet which was often blurred out into just a series of weak shoulders in CVD ZnS and elemental ZnS. Ab initio vibration calculations assign these absorptions to Zn-H-Zn stretch modes [172]. The doublet should be from a ZnH

2

in a sulfur site and the triplet from a ZnH

3

in a sulfur site. Figure

7-35 shows the measurements of several samples and the spectral position in wavelength and wavenumber for these vibrations.

A few other absorptions have been identified in this work in a few samples, but not yet definitively assigned. Figure 7-36 shows the transmission spectra of successively thicker samples of ZnS, the first two of CVD ZnS and the third of hot-pressed ZnS.

Plotting the three samples of increasing thickness allows the appreciation for the change of the multiphonon edge (wavelengths longer than 10.1 μm which is assigned to 3LO after[59]). This also helps to distinguish between those absorptions extrinsic and those intrinsic in ZnS. Some of these absorptions are labeled with their wavelength in microns.

241

Figure 7-35: Transmission of ZnS in the vicinity of the hydride absorption

The hot-pressed ZnS of course does not have the hydride absorption because of how it is made, but does have a number of absorptions which have not yet been identified. These impurity absorptions remain unidentified, as this sample of hot-pressed

ZnS was of unknown origin and pedigree. In the hot-pressed sample, the absorption at

4.3 μm can readily be assigned to the spectrometer artifact of CO

2

, and the fine structure from about 6 to 8 μm (as well as 2.6 to 2.9 μm in the thin CVD sample) can be assigned to water lines in the spectrometer cavity (or zinc oxide according to others). The absorptions at 8.6, 8.9, 9.6, and 10.0 μm were not easily assignable. They are not known

ZnS 2 or 3 phonon sum or difference multiphonon absorptions [66] and also not obviously assignable to ZnO multiphonon spectra [57, 158].

242

Figure 7-36: Extrinsic infrared absorptions in ZnS samples

7.2.4.2 Scattering

For many years, scattering has been blamed for the low visible and near-infrared transmission of CVD ZnS. For the purposes of CVD ZnS, scattering could be due to (1) pores, (2) hexagonal phase ZnS in a cubic matrix, (3) metal impurities, or (4) refractive index inhomogeneity due to differential incorporation of impurities such as oxygen. The driving force for scattering is ultimately the difference in refractive index between the matrix (cubic ZnS) and the scatterer. The largest index difference from the ZnS matrix would be air (pores) and the smallest difference would be local density fluctuations or birefringence due to stacking differences (i.e. hexagonal layers) or chemistry differences

(i.e. oxygen in sulfur sites).

243

The size scale of scatterers is generally the most important factor to consider when assessing elastic scattering where no energy is transferred [170, 173]. Scatterers which are very, very small with respect to the wavelength of light can be treated by volumetric mixing models such as Bruggeman or Maxwell-Garnett and act as “effective media.” Effects of scatterers only somewhat small with respect to the wavelength of light can be computed using Rayleigh scattering, which has a λ

-4

dependency. The application of Rayleigh scattering requires that the system meet a criterion based on the “size factor” which relates the relative refractive indices of the matrix and scatterer, and the size of the scatterer, to the wavelength of interest. Finally, larger particle scattering must be computed using the much more complex Mie scattering theory which uses spherical harmonic expansions to describe scattering, but which is valid only for spherical particles. Mie theory collapses to Rayleigh scattering when proper combinations of sizes and refractive index differences are small enough. At the limit of large particles with respect to the wavelength, Mie theory becomes geometrical optics or “large particle scattering” where scattering is independent of wavelength. One other scattering phenomenon of note is surface scattering which depends on interfacial roughness and has a λ

-2

dependency.

7.2.4.2.1 Rayleigh Scattering

Rayleigh scattering depends on concentration, size, and refractive index of scatterers. The extinction due to scattering by Rayleigh mechanisms can be shown to be

[174]:

244

γ

Sca

=

N

A

C

Sca

=

32

9

π

4

a

3

λ

0

4

(

ε

A

ε

B

)

2

f

A

( 1

f

A

)

2 where γ is the extinction coefficient (taken to be only from scattering, units cm

-1

), N

A

is the number of scatterers per volume (units cm

-3

), C

Sca

is the scattering cross section (i.e. the strength of the scatterers, not necessarily their geometrical cross section, units cm

2

), a is the radius of the scatterer, λ

0

is the wavelength, ε is the permittivity (A being the scatterer and B the matrix), and f

A

is the volume fraction of the scattering phase.

Scattering in hot-pressed ZnS has been analyzed using Mie theory [175]. To match their experimental transmission data, the authors had to include a bimodal lognormal distribution of pore sizes. Small pores (radius of 80 – 100 nm) were required to model the Rayleigh scattering in the visible and large pores (radius 2 – 10 μm) were required to model the wavelength independent (large particle) scattering through the long-wave infrared.

The sample of hot-pressed ZnS obtained for this present analysis did not have any appreciable large particle scattering, so it was modeled using the Rayleigh equation above using data from the band edge out to 8μm. At longer wavelengths, extrinsic absorptions and intrinsic multiphonon absorptions become important which are not accounted for in the scattering model. The results are shown in Figure 7-37a. As the equation above does not provide a unique solution, but rather a family of solutions, three of which are shown along with the actual measured data. For the transmission calculation, the extinction coefficient due to scattering γ is determined as above, R is the

245 reflectance determined from the refractive index, L is the sample thickness (6.39mm) and transmission is

T

=

( 1

R

)

2 exp(

γ

L

)

Absorption is not explicitly accounted for here, but since the scattering is a much stronger extinction than the absorption in this wavelength range, it can be neglected. Also the denominator sometimes included in the transmission equation relating to multiple reflections (1-R

2 exp(-2γL)) is ignored in this case. The inclusion of this term only changed the predicted transmission by about 1% at the most and so was eliminated for simplicity.

Generally good agreement is seen between the models and the measured data from the 6.39mm thick sample without resorting to distributions of scatterers. The two models are essentially identical beyond about 3 μm. Model 1 assumed pores with nondispersive refractive index = 1.003 with 0.1% volume fraction of 130 nm diameter pores. Model 2 assumed pores with 1% volume fraction and 30 nm diameter. Model 3 assume pores with 10% volume fraction and 20 nm diameter. Obviously the latter case is not likely and could easily be verified by a density measurement, but it illustrates the interplay of volume fraction with scattering size. Decreasing either the size or the volume fraction of the scattering phase will increase transmission, but the shape of the curve will remain essentially unchanged. According to Palik[176], IRTRAN ZnS is composed of 5% hexagonal phase and 95% cubic phase. Since the refractive index of air

(n~1) is so much different than ZnS (n~2.2), pores will scatter much more than any

246 birefringent phase and will dominate the scattering behavior if present in sufficient size or concentration.

Xue and Raj [177] have studied the effect of ZnS hot-pressed near and above the transition temperature, creating samples with volume fractions of hexagonal phase from 4 to 26% (determined by XRD) with diameters 2 to 8 μm. They found that at the largest volume fractions and sizes the scattering approached a λ

- 4

dependency. However, they admit that their grain sizes are on the order of the wavelength where scattering is evident

(i.e. the long-wave infrared) which is not the regime where the Rayleigh scattering equation is most valid. Xue and Raj’s data shows what appears to be wavelength independent scattering for low volume fractions and small sizes of hexagonal phase, whereas larger sizes and fractions transmit essentially nothing until about 4 μm wavelengths.

When such a model is applied to standard CVD ZnS, it fails due to the large deviation from a λ

- 4 for scattering in this material. Figure 7-37b shows this discrepancy.

No combination of volume fractions and pore sizes will fit this CVD ZnS curve even reasonably closely. Many in the literature have claimed that the low transmission in standard CVD ZnS could be explained by Rayleigh scattering of pores, but this is clearly not the case as can be seen. Any smaller difference in refractive index, such as that due to birefringence or oxygen inclusion would have a different size and volume fraction dependency but the same wavelength dependency. Thus some other scattering model must account for this relatively shallow slope with wavelength.

247

Figure 7-37: Rayleigh scattering models for hot-pressed ZnS and CVD ZnS

(a,L) hot-pressed IRTRAN ZnS assuming scattering pores (b,R) CVD ZnS assuming various volume fractions and sizes of pores

248

7.2.4.2.2 Internal Surface Scattering

One scattering model with different wavelength dependence is the surface scattering model. Surface scattering phenomena can also be described as “diffraction phenomena resulting from phase variations induced upon the reflected wavefront by random nanostructure surface features” [178]. Consider for a moment that ZnS scatters due to some very local refractive index changes, but that the geometry of the scattering is two-dimensional, with planes along the stacking direction with surfaces in between them.

In essence I am proposing a surface scattering that occurs internal to the CVD ZnS, not just at the external surfaces open to air.

It has been observed that even a single layer of hexagonal phase such as occurs between cubic rotational twins in ZnS behaves electrically like a quantum well due to the difference in electronic band gap between the structures. Such is also the case with polytypes. This phenomenon is related to the anomalous photovoltaic effect observed in

ZnS, and is believed to be due to spontaneous polarization in cubic and hexagonal regions creating internal fields in ZnS [179]. Thus faulted crystals of ZnS should be expected to exhibit unusual electromagnetic behavior. Anisotropic electrical conductance has also been observed in ZnS [180].

While purely cubic ZnS has zero birefringence, hexagonal ZnS has a theoretical maximum birefringence (defined as Δn = n e

– n o

) of 0.024 at 546 nm [181]. All polytypes with less hexagonality should have smaller birefringence values. Anomalously high birefringence values up to 0.05 have been observed in crystals with severe onedimensional disorder [182]. Even twinned cubic crystals have shown birefringence

249 values of 0.004 and 0.028 [181]. Difference in refractive index of wurtzite and sphalerite varies from about 0.03 at 1 μm to zero at 7 μm [176]. Over the range from 350 nm to 1.5

μm the absolute difference in index between the 3C and 2H phases fluctuates substantially going as low as 0.002 and as high as 0.03. Much larger differences in index can be expected in regions with substantial oxygen content, as the refractive index of

ZnO is about 0.3 lower than ZnS.

The scattering of successive lamellae with slightly different refractive indices was modeled as a series of internal surface scatterers. A scattering coefficient is derived by computing the surface scattering of one internal surface (unitless) and multiplying this by the number of such surfaces per centimeter (units of cm

-1

). This scatter loss coefficient is then added to the absorption coefficient inside the exponential when transmission is calculated.

T

= −

R

γ

abs

+

γ

sca

Since absorption coefficient data in the 0.3 – 7 μm range is scanty, a constant value of k=10

-7

(imaginary part of the refractive index used to calculate absorption coefficient) was assumed. Since the modeled region is not near the band edge or the multiphonon region, the choice of absorption should not affect the results much. See Figure 7-38 for a graphical representation of this model.

The scatter coefficient, then, is derived of two parts. The scatter per interface part

(TIS perInt

) is based on the calculation for total integrated scatter (See Appendix F for derivation of this equation).

250

TIS

=

2

πσ

Δ

n

λ

n av

⎟⎟

2 where TIS is total integrated scatter (a number between zero and one), σ is the root-meansquared surface roughness (in units of length), Δn is the difference in refractive index between the material on either side of the internal surface, n av

is the average refractive index between the two lamella, and λ is the wavelength.

The number of interfaces per unit length, N

IntperL

, is the inverse of the lamella thickness, l lamella

. The scatter coefficient γ sca

can then be shown to be:

γ

sca

=

TIS N perInt IntperL

=

TIS perInt l lamella

=

4

π σ

2

( )

2

λ

2 2

n l

=

4

π

2

1

2

n av

λ

2

σ

2

( )

2

l lamella

Therefore, the critical parameters in determining the scatter are the difference in index between the lamellae (Δn), the interface roughness (σ), and the thickness of the lamella

(l lamella

). For a given value of γ sca as determined by transmission from experiment, there are an infinite number of solutions that satisfy the scatter coefficient value. Some of these will be unphysical, but a few possible ones will be considered below for standard

CVD ZnS, elemental ZnS, and multispectral ZnS. The important result is not the actual values of these parameters, but the wavelength dependence which for the first time can reasonably reproduce the transmission curves for standard CVD ZnS.

251

Figure 7-38: ZnS internal surface scattering model schematic

The best and most reasonable fit to measured standard ZnS data was achieved by assuming a roughness of 50 nm, Δn of 0.050, and a number of internal surfaces of 10

5

per cm (i.e. one surface per every 100 nm). This Δn value seems conservative, given that values close to this have been observed in ZnS without any oxygen. The TEM data indicates thin layers on the order of 50 nm, and roughness could be seen as the steps in an interface which would create differential optical path lengths.

Alternatively, if roughness is assumed to be about a tenth of the lamellar thickness, another solution is possible. In this case, standard ZnS is well-fit by assuming lamellar spacing of 100 nm, as before, roughness of 10 nm, requiring a Δn of 0.3. This

252 index difference seems high, except when considering the index difference between ZnO

(n

1064nm

≈ 1.95) and ZnS (n

1064nm

≈ 2.3) is about this value.

The attractiveness of this model is that it can match the transmission of multispectral ZnS, elemental ZnS, and standard CVD ZnS just by changing the amount of birefringence and leaving the other parameters constant. This seems reasonable given that the powder x-ray diffraction data noted differing strengths of the hexagonal peaks

(and hence different atomic fractions of hexagonality) in these three samples. The hexagonal layers would be expected to be enriched in oxygen giving lamella with lower refractive index.

Figure 7-39 shows the computed results of the model along with measured curves of the three types of samples as above. Model parameters for all samples were 50 nm roughness and 10

5

internal surfaces per centimeter. Measured samples were 4.65 mm

(0.183”) thick. Effective Δn for CVD ZnS was 0.040, for eZnS was 0.010, and for msZnS was 0.002.

Recall that the microstructure is very different for multispectral ZnS and that the fine nanoscale lamellae are not present in this material. The assumption, then, that there are 10

5

per cm (i.e. one internal surface per every 100 nm) surfaces to scatter is not valid.

It is easy to see, however, that choosing a more reasonable value for this parameter, such as 10

3

per cm (i.e. one internal surface per every 10 μm) results in only slightly better transmission right along the band edge. Thus at the low levels of birefringence assumed for fully recrystallized multispectral ZnS, Δn=0.002, the number of internal surfaces is not nearly as important in determining transmission. This is an important result, as it

253 implies that if refractive indices of neighboring regions are very similar, the lamella can be thin and closely spaced, and the structure will still have low scatter even in the ultraviolet.

Figure 7-39: Internal surface scattering model of msZnS, eZnS, & standard ZnS

The weakness of the model is that it that it does not reproduce band-edge results such as seen in vacuum annealed red ZnS and hydrogen annealed standard ZnS. More complex scattering theories including effects of coherent and incoherent multiple scattering, hierarchies of size scales, or distributions of sizes may have to be employed to improve the predictions for these types of samples. Similar effects have been noted in the calculations for transparency in glass ceramics [169].

254

8 Discussion and Conclusions

In this final chapter I pull together all of the aforementioned investigation and attempt to present a coherent picture of the properties and processing relationships in

CVD ZnS. This is broadly divided into two sections: the nature of CVD ZnS and the nature of its transformation during heat treatments. The bulk of this chapter is my own assessment of the literature and the results of this investigation, so copious citations and references to previous data presented will be minimized.

First I talk about the nature of the CVD ZnS material chemically, crystallographically, structurally, and optically. I review the differences encountered along the different positions of the growth cores from the first to last material to deposit.

I then speculate on the nature of red ZnS and elemental ZnS and their differences from standard yellow CVD ZnS.

Then I switch to the heat treatment of the above material. I focus on the hot isostatic press and elicit the role of temperature and pressure. I relate a proposed mechanism of recrystallization, and a role of the metals in inducing it.

Finally, I offer some overall observations and conclusions. Much work has been done, but I also relate some specific examples of work yet to be done to test out the conclusions and extrapolations from this work.

8.1 What is the nature of standard CVD ZnS?

CVD ZnS is a polycrystalline material with multiple levels of hierarchical structure. Its typical impurities are hydrogen from the hydrogen sulfide precursor and

255 oxygen from process gases and the low vacuum deposition environment. CVD ZnS consists of nano-scale twins of cubic sphalerite phase on the order of 10 to 100 nm separated by a hexagonal packing layer. These twins are associated into domains and finally grains, where each higher order boundary involves disordered hexagonal and cubic packing. Strictly speaking polytypes probably do not exist within individual crystallite lamella. Of course, the degree of ordering or disordering totally depends on the scale being considered.

CVD ZnS exhibits about 5 to 10 atomic percent of hexagonality. The hexagonal phase measured by x-ray diffraction should not be considered as a “phase” in the sense that there is a clearly identifiable region in the material with the 2H or 4H structure. It has been shown that hexagonal reflections in polycrystalline XRD and selected area electron diffraction indicate that hexagonal packing is present. The hexagonal disorder is in a single dimension, the close-packed stacking direction. In the polycrystalline solid, the close-packing direction is not quite randomly oriented, since preferred orientation is evident in XRD, in some cases emphasizing the wurtzite (10.0) and in others the wurtzite

(10.1) reflection. It is possible that some needle-like hexagonal ZnS crystal are occasionally produced in the vapor phase and then get trapped in the growing ZnS deposit or are flushed away into the exhaust of the CVD chamber [103, 165]

The “grain” size of CVD ZnS as commercially deposited around 670 °C is 4 to 8

μm in the direction perpendicular to the growth direction, and about ten times that in the growth direction, giving CVD ZnS a “columnar” habit. The compressive strength of

CVD ZnS is considerably higher when the load is along these “columns” than when it is

256 perpendicular to them. CVD ZnS has a preferred texture (25 – 60 %) along the

{ 100 } planes (i.e. the (200) and (400) x-ray reflections) and to a much lesser extent

{ 311 } and

{ 511 } planes which are at small angles to the

{ 100 }

planes. In some cases the

{ 111 } texture, including both (111) and (222) reflections, is more important than the

{ 511 }

.

Macroscopically, CVD ZnS and even vapor grown single crystal ZnS shows many “growth bands.” In CVD ZnS these are perpendicular to the growth direction.

They probably represent slightly different chemical compositions, such as impurity content of hydrogen and oxygen. Previous researchers have suggested that banding is due to gas phase (homogeneous) reactions, non-uniform zinc vapor usage, or formation and dissociation of zinc hydride [76]. It was confirmed that the hydride absorption was lowest in the colorless band near the mandrel. This “growth banding” has been observed in CVD zinc sulfo-selenide where the bands were clearly different sulfur to selenium ratios and had different refractive indices. In single crystals of ZnS, these bands are not always associated with different crystal structures or polytypes, but nonetheless exhibit measurably different birefringence. The exact nature of this banding is still poorly understood.

Data presented here on the variation of properties of CVD ZnS along the growth direction (i.e. the cores) are the first such data in the literature. It was shown by powder x-ray diffraction that the absolute amount of hexagonality remains unchanged from the first deposited ZnS at the mandrel to the last deposited ZnS at the growth surface.

257

However, the texture of the material changes, as the relative intensities of the different hexagonal ZnS peaks change with position in the core.

Electronic defects were shown by photoluminescence to be greatly different in material on the mandrel side versus the growth side and in the growth direction versus perpendicular to it. The exciton luminescence was prominent on the mandrel side but completely absent on the growth side. The visible luminescence goes from emphasizing blue at the mandrel side to green at the growth side. On the mandrel side, the luminescence intensity is slightly higher for the excitations of the “tops” of the columns.

On the growth side, the luminescence intensity is an order of magnitude higher for excitations along the sides of the columns. From these observations I conclude that different optically active defects dominate on the mandrel and growth sides. There is an apparent increase in macroscopic growth banding on the growth side, and the texture of the hexagonality is different there. This leads to the possibility that the observed difference in luminescence is due to the effect of different internal electrical fields in ZnS caused by differential stacking disorder. These internal fields have been observed and analyzed in single crystals (e.g. [157, 183]) but never in CVD ZnS. Additionally, it has been shown that motion of dislocations in ZnS induces electric fields which can modify the photoluminescence signature[184].

Chemically, ZnS in general tends to be a Zn-rich n-type semiconductor due to the relative magnitude of the reaction constants of zinc and sulfur condensation and evaporation and the presence of oxygen as a reducer for sulfur. CVD ZnS has been verified by EDS to be Zn-rich as well, assuming a Bridgman melt-grown crystal is

258 nominally stoichiometric. CVD ZnS has the additional driving force for Zn-richness since the growth process is typically carried out at zinc to hydrogen sulfide reactant ratios greater than one. Zn-rich materials tend to favor the introduction of oxygen since the material is already sulfur deficient so sulfur vacancy sites exist, and interstitial zinc needs volume compensation by nearby sulfur vacancies or substitutional oxygen.

Competition exists for these sulfur sites during the CVD process. Hydrogen sulfide is thought to adsorb onto the growing solid body and interact with a sulfur vacancy, filling the sulfur site and giving off hydrogen. Oxygen, if available, will preferentially bond to sulfur sites before hydrogen sulfide. The evolved hydrogen from the split hydrogen sulfide may react exothermically with zinc vapor in the gas phase near the growth surface to form ZnH

2

, often getting incorporated into the growing solid.

When deposited at temperatures up to about 850 °C, CVD ZnS shows a characteristic zinc hydride absorption around 6 μm which can be resolved into a triplet of roughly equally spaced spectral lines. This defect is believed to have the structure of (ZnH

3

)

S based on ab inito assessments. Some CVD ZnS (e.g. that produced by Rafael) has no 6

μm absorption. Such material can be produced by growing at higher temperatures, though it is not known if this is the case with the Rafael material.

The measurable oxygen content in CVD ZnS, about 0.2 - 0.4 atomic percent, gives rise to some of the characteristic behavior during hot isostatic pressing to be discussed shortly. The oxygen is believed to be located both in the lattice on sulfur substitutional sites and along disordered boundaries. Oxygen preferentially inhabits sulfur sites along twin boundaries (hexagonal packing sequences), which have been

259 shown in other studies to be sulfur deficient and oxygen rich. This preferred location is due to the smaller ion size of oxygen and the favored hexagonal structure of ZnO. Since substitutional oxygen induces lattice compression, it tends to associate spatially with other defects which stretch the lattice such as zinc interstitials or monovalent cations like silver. This phenomenon is referred to as the compensation of volume.

8.1.1 What is red ZnS?

What is referred to here as Red ZnS is ZnS grown by CVD at low mandrel temperatures, ~640 °C at very high zinc excesses. The resulting material when polished varies from deep red to brownish in color. It is characteristically very low scattering in the visible and infrared but has very strong hydride absorption and a notably longer wavelength ultraviolet edge. The “ultraviolet” edge measures at 400 to 415 nm, compared to 361 to 386 nm for standard CVD ZnS. The 6 μm absorption band is resolved into two distinct minima at different spectral positions than in standard ZnS. It has been said that the absorption edge (and hence coloration) is tied to the hydride absorption. In these experiments this was not confirmed. The S5 sample having been annealed at 650 °C for a total of 6 hours showed an ultraviolet edge shift to 376 nm, creating a white, scattering sample which still had an unchanged 6 μm absorption. This leads me to conclude that the color in red ZnS, at least, is due to the position of the band edge and the relative transmission in the various visible wavelengths.

Microstructurally, red ZnS is similar to other standard ZnS but with a slightly smaller grain size. This smaller grain size does not contribute to stronger ZnS, and biaxial flexure strength is essentially the same for red ZnS and standard CVD ZnS. The

260 grains do not have columnar habit, however, unlike standard ZnS. The nanostructure is similar, with small twins collectively organized into larger domains and then grains. The nanostructure does seem to be more randomly oriented compared to other samples investigated. Selected area electron diffraction showed more “spotty rings” indicative of small structures. It is conceivable that because of the low deposition temperature, red

ZnS actually forms ZnS in the gas phase by homogeneous nucleation via a transition state complex then deposits on the substrate more as a particle[98]. Alternatively, the dense isotropic polycrystalline form could be due to a kinetically limited process, where the rate of the reaction to form ZnS is the limiting factor, and increased nucleation rate inhibits anomalous grain growth seen in higher temperature grown ZnS.

Red ZnS has about 2 atomic percent hexagonality, which is considerably less than in standard ZnS which is more scattering. As with elemental ZnS, the slow growth rate

(<0.001”/hour) probably contributed to the lower incidence of hexagonal stacking disorder. The crystallographic texture of red ZnS is substantially different from standard

ZnS, with most of the texture being in the

{ 111 } planes and much less in the

{ 100 } planes.

Chemically, the main difference in red ZnS seems to be its oxygen content. Red

ZnS is not any more Zn-rich than standard ZnS, at least within the limits of EDS assessment. It does contain more oxygen, as measured by Interstitial Gas Analysis

(IGA), at 0.3 - 0.6 atomic percent, as expected for a lower temperature deposition.

Lattice parameter measurements confirm the high presence of oxygen (as ZnO) in the concentration range measured by IGA. It is possible that the red-shifted ultraviolet edge is due to the high dissolved oxygen content as ZnS(O). Electronic defects in red ZnS are

261 sufficient to quench the exciton in photoluminescence, although additional structure in the ultraviolet is present in red ZnS that is not present in the other samples.

It is possible, but not proven, that the scattering induced by annealing red ZnS is due at least in part to precipitation of zinc oxide particles. Some annealed red ZnS (S3) in which the 6 μm absorption has been removed shows structure that some in the literature have claimed is due to ZnO. Due to its close proximity to water bands that are difficult to remove even in a careful spectroscopic measurement, this identification is still uncertain .

8.1.2 What is elemental ZnS?

Elemental ZnS as it is defined here is any CVD ZnS formed from hydrogen and sulfur reacted just before injection into the CVD chamber to combine with Zn vapor.

This is distinguished from standard (or FLIR grade) CVD ZnS which uses a commercial hydrogen sulfide gas precursor. The two known versions of elemental ZnS are on average superior in visible and near-infrared transmittance to standard CVD ZnS.

It has been shown by x-ray diffraction that both Chinese ZnS (grown from the elements) and Raytheon elemental ZnS have measurably less hexagonality (estimated at

2 atomic percent). As a result, both show improved transmission (lower scattering) in the visible and near-infrared. In the internal surface scatter model, this is reflected by a lower “birefringence” for elemental ZnS than for standard CVD ZnS.

Structurally, elemental ZnS should be like other CVD ZnS materials grown at similar temperatures. Recall that Raytheon commercially produced standard CVD ZnS and elemental ZnS were both grown with a 670 °C mandrel temperature. Grain

262 diameters for Raytheon elemental ZnS are 4 to 8 μm, and the material has a “columnar” grain which is about ten times as large in the growth direction. At the nanoscale, lamellar twin structures are observed with diffraction contrast in the TEM. The x-ray texture is nearly identical to standard CVD ZnS except that there is a component of

{ 210 } texture which is not significant in standard ZnS. Elemental ZnS may be slightly denser than standard ZnS, but barely within the resolution of an immersion-type measurement.

Biaxial flexure strength is essentially identical to CVD ZnS.

The reason for the reduction in hexagonality in these materials can be postulated as due to several factors. Firstly, it is highly likely that there is some unreacted hydrogen gas which is introduced into the reactor with the hydrogen sulfide. Thermodynamically, this will make the hydrogen sulfide more stable in the gas phase and prevent the back reaction to the elements (the low chamber pressures prevent the need to consider activity coefficients). This allows the splitting of the hydrogen sulfide to occur at the surface of the growing solid body rather than in the gas phase. Also, Raytheon elemental ZnS was historically produced at much lower growth rates than commercial CVD ZnS

(<0.001”/hour versus 0.002”/hour). This lower growth rate should assist in preventing hexagonal stacking disorder.

Chemically, elemental ZnS is very similar if not identical to standard CVD ZnS

(i.e. made from bottled H

2

S). It is zinc-rich and its measured oxygen content is the same at 0.2 - 0.4 atomic percent. The hydrogen content of Raytheon elemental ZnS is probably similar to (most) standard CVD ZnS based on the absolute depth of the 6 μm absorption.

The Chinese and Raytheon elemental ZnS materials are different, however, since the

263

Chinese ZnS shows no 6 μm absorption while the Raytheon elemental ZnS shows the characteristic CVD ZnS hydride absorption with three identifiable minima.

In photoluminescence, Raytheon elemental ZnS is very similar to standard ZnS complete with exciton luminescence. A minor difference was noted from standard ZnS, however, in that elemental ZnS showed a strong peak assigned to the acceptor bound exciton related to oxygen. Both Raytheon CVD ZnS and elemental ZnS have a redshifted room temperature free exciton due to bandgap changes due to oxygen. Rohm and

Haas CVD ZnS and multispectral ZnS, however, show no free exciton red shifting despite having measurably identical oxygen levels. This result could be explained by having oxygen located differently in the lattice or boundaries in the two groups of materials. Chinese ZnS, on the other hand, has very different luminescence, characterized by a broad featureless visible luminescence without the exciton, suggesting additional defects.

8.2 What is the nature of transformation to Multispectral ZnS?

Multispectral ZnS is a desirable optical material due to its broadband transparency from the ultraviolet through the long-wave infrared. The ability to make very thick parts of conformal shapes sets chemical vapor deposition apart from other techniques for making ZnS. Therefore, the transformation from as-deposited scattering standard CVD

ZnS is critical for applications requiring a window with visible through longwave infrared spectral bandwidth.

264

Many aspects of the changes induced by hot isostatic pressing (HIPing) on the microstructure and properties of CVD ZnS have been reviewed and assessed here.

Chemically, the zinc to sulfur ratio does not appear to change, at least within the resolution of the EDS measurement. Hydrogen is removed by HIPing or annealing for sufficient times, as evidenced by the reduction and eventual removal of the 6 μm absorption. In the samples assessed, the overall oxygen concentration does not appear to change, measuring 0.3 – 0.4 atomic percent.

The most obvious change, other than the removal of scattering, is the complete recrystallization of the CVD ZnS after HIPing, resulting in large single crystallite grains without nano-lamellar twins. This dissertation has shown that this process can be observed before it is complete, with the intermediate material being more scattering than the original CVD ZnS. Seemingly similar multispectral ZnS can be distinguished by its crystallographic texture. Upon complete recrystallization, the HIP’d ZnS is strongly textured in the

{ 111 } planes, close to 50 percent.

In samples HIP’d at low temperature without any metal present (HIP33none), texturing remained as if it were still CVD ZnS (i.e. predominantly

{ 100 } texture), recrystallization did not take place or at least did not complete, and visible transmission was low and scattering. Samples may have improved transmission without “full” recrystallization (defined here as having the predominant crystallographic texturing in the

{ 111 } planes), such as those annealed at 850 °C (AnRH) and those HIP’d with silver at

750 °C (HIP33Ag) where the material is predominantly

{ 211 } textured with

{ 111 } being the next biggest fraction.

265

Thus it can be said that true “multispectral ZnS” is by definition fully recrystallized to strong

{ 111 } texture, so a certain minimum process involving a combination of temperature, pressure, and/or metal promoter is required. Biaxial flexure testing of

{ 111 } textured multispectral ZnS has a lower average strength and higher

Weibull modulus than material HIP’d at lower temperature with silver or platinum. The differential crystallographic texture of the low temperature HIP’d samples (HIP61Ag,

HIP62Pt) and the removal of internal stresses induced during CVD deposition may be responsible for the different biaxial flexure behavior. The flaw size to grain size ratio is large enough in both CVD ZnS and HIP’d ZnS that fracture should be governed by the polycrystalline fracture energy in both cases, and there should not be a significant grain size effect[154].

8.2.1 What is the HIP doing?

The hot isostatic press combines two driving forces for recrystallization – temperature and pressure. By performing annealing experiments, the effect of pressure has been independently assessed. It has already been shown that the pressures typically used in HIPing ZnS (15 – 30 ksi) are well above the Peierls’ stress of 9 ksi necessary to move dislocations in ZnS. Thus it is expected that HIP’d microstructures will show more signs of plastic deformation than annealed structures. An additional surprising effect of the high pressure was that it allowed higher processing temperatures to be usable without evaporation of the ZnS. It was shown that a 990 °C anneal for 10 hours completely evaporated 5 mm thick samples of CVD ZnS, whereas samples held at 30 ksi for this time and temperature fully recrystallized with some subsequent grain growth.

266

Unlike other authors who claim to have been unable to recrystallize without applying pressure, this work observed some microstructures that did recrystallize with annealing only when temperatures were ≥850 °C (e.g. anneal Co33_1a). This temperature would seem to be the threshold for thermal activation of dislocation motion involving diffusion, grain boundary sliding, and nucleation of stress free grains[33]. It should be recalled that lower temperatures are effective at recrystallization in the HIP when appropriate metal recrystallization promoters are used.

Fully recrystallized multispectral ZnS has a very low fraction of hexagonal phase, less than about a half an atomic percent. Since the starting material has 5 to 10 atomic percent hexagonality, the hexagonality is reduced by HIPing. Since the phase transformation is martensitic, it is expected that pressure would help induce the transformation. Since cubic sphalerite is slightly denser than hexagonal wurtzite, it is also expected that the denser phase would be slightly favored at higher pressure.

Additionally, lattice diffusion will likely be important at temperatures ≥850 °C, so zinc diffusion and rearrangement of the zinc sublattice may also assist in transformation to the cubic phase [108].

Recrystallization due to the HIP can be seen as a combination of diffusion and plastic deformation. The mechanism proposed here implicates oxygen as the important species which is normally at the twin boundaries, inducing hexagonal layers at the twin boundaries and contributing to nonuniform chemistry and hence nanoscale birefringence.

Careful measurements of natural sphalerites have indicated that oxygen concentrates at twin boundaries, substituting for sulfur for less than two monolayers[185]. In the model

267 proposed here, at elevated temperatures these oxygen atoms at the boundaries can diffuse into the lattice into sulfur sites, thus unpinning the twin boundaries to move to lower energy stacking configurations. The total energy of the system is lowered by removing the internal hexagonal surfaces and recrystallizing into large (30 to 100+ μm) grains which are relatively fault free. These large “perfect” crystallites are readily evident in

TEM in materials HIP’d at high temperature. Grains of HIP’d ZnS are irregular in shape and can consist of lathes of material with aspect ratios of up to 10.

This phenomena could fall under the heading of “diffusion-induced grain boundary migration” (DIGM) or “diffusion-induced recrystallization.” Diffusion induced grain boundary migration (DIGM) has been modeled at an atomistic level as dislocation climb in the grain boundary as a result of interdiffusion (Kirkendall effect) [186]. In classic DIGM, the flux of solute atoms to the boundary is not equal to the flux of matrix atoms out of the boundary, and the result is a “wavy” grain boundary due to atoms diffusing across the boundary rather than along it as in normal grain boundary diffusion[187]. DIGM and diffusion induced recrystallization (DIR) has been observed mostly in precipitates in metal alloys, but with very few exceptions not at all in ceramic systems presumably due to the requirement for charge neutrality and thus the simultaneous movement of anions and cations[188].

Diffusion induced recrystallization (DIR) occurs when a misfitting solute, such as oxygen in ZnS in this hypothetical example, diffuses through a lattice producing strain and thus nucleating new grains[189]. Solute segregation at interfaces results from elastic forces from size misfit, electrostatic interactions between charged solutes and interfaces,

268 and dipole interactions involving electric fields at interfaces[190]. The elastic energy change that results from the movement of the solute provides the driving force for the boundary motion in the new grain. “Coherency strains” can be thought of as variable lattice parameter in nearby regions of the crystals which can be caused by various things including diffusion, phase transformation, and nucleation. Carter and Handwerker [189] have shown that morphologies of recrystallized grains depend only on elastic anisotropy and linear compressibility, each defined by a relation of the elastic tensor components. In the ideal case of DIR, for materials with elastic anisotropy of greater than one, like ZnS, the fastest growth in strain free crystals will be parallel to the <111> directions and the slowest growth parallel to the <100> directions and the resulting recrystallized grains will be cuboidal.

Fully recrystallized ZnS does have a strong

{ 111 } texture, suggesting that a mechanism like diffusion induced recrystallization could be taking place. However, the resulting structure is not cuboidal, but rather elongated tabular grains. A more likely candidate for a detailed mechanism for producing the microstructures observed in recrystallized ZnS is polytype induced exaggerated grain growth [191]. In this model, an impurity or dopant with low solubility forms prototypic growth faults at a twin or other special boundary rather than forming a separate phase within the host. Because the symmetry of the system is altered, these polytypic sequences then grow in the direction of the fault plane (i.e. along the

{ 111 } planes) at the expense of fault free regions in order to minimize the total energy of the system. The result is anisotropic grain growth, which

269 has been observed in the cubic to hexagonal phase transition in barium titanate and has been implicated in twining behavior in natural ZnS sphalerites.

Exaggerated or discontinuous grain growth (also called secondary recrystallization), or the coarsening of the average grain size due to preferential consumption of some grains in favor of others, has been observed in many ceramics and has traditionally been ascribed to a bimodal size distribution of grains or non-uniform pore pinning [32]. It has been shown that there is another mechanism for exaggerated grain growth which can be explained by the presence of nanostructural polytypism in these systems that imparts anisotropy to an otherwise isotropic grain system. This phenomenon may be able to explain the microstructure of recrystallized CVD ZnS.

Nanostructural polytypism is favored in systems where there is a very low solubility of the additive in the host material and rather than form an isolated phase, there is a tendency to form faults which become enriched with the additive [191]. These faults tend to be low index planes such as twin, antiphase, or inversion boundaries and are produced (nucleated) during the initial growth and never by deformation. Examples of this phenomenon include the hexagonal to cubic transformation in BaTiO

3

, where the dopants of calcium tend to segregate to polytypic faults. Polytypism in SiC has been shown to be enhanced by the addition of aluminum, and additions of SnO

2

to ZnO produce inversion boundaries. In all cases, when these systems are subsequently reheated they exhibit anisotropic grain growth along the polytypic faults (i.e. twin faults or inversion boundaries). Resulting recrystallized microstructures can be seen to be preferentially elongated along the planes of the polytypic faults at the expense of fault

270 free regions. Sometimes secondary faults are produced as a result of impingement of oppositely growing domains.

No porosity was identified in any of CVD materials used for this dissertation.

Based on this lack of evidence for porosity, a pore-closure mechanism cannot be invoked as a driving for HIPing. Rather, it is suggested here that oxygen normally pins grain

(twin) boundaries. At elevated temperatures the oxygen can diffuse, and boundaries can move and grains recrystallize. The exact threshold temperature for this unpinning mechanism cannot be separated from the threshold temperature for thermal activation of dislocation motion mentioned above. Other impurities like hydrogen leave the lattice, and zinc diffusion will promote the phase transformation from hexagonal to cubic stacking by the crystallographic rotation of zinc tetrahedra. Plastic deformation (strain) will also promote the martensitic phase transformation and provides the main basis for strong{111}texturing through slip and twinning of the cubic grains.

8.2.2 What is the metal doing?

One important factor not yet discussed is the role of the metal present during the hot isostatic pressing. It has been shown by the “interrupted HIP” experiments that certain metals (Pt, Ag, Ni, and Cu) accelerate recrystallization from the contacted surface.

In the case of platinum and copper, the formation of a sulfide on the metal foil has been verified. Those metals which did not assist recrystallization (Fe and Co) form only hexagonal or cubic ZnS on their surfaces after HIPing. Sputter deposited metals on the

ZnS surface do not seem to be as effective as foils in inducing the recrystallization, presumably because the indiffusion of the metal is facilitated which complicates the

271 lattice restructuring and recrystallization. This could be due to the need to nucleate a metal sulfide, and the nature of nulcation sites on a foil surface and a metal coating are very different, the former having electronic surface reconstructions.

A detailed understanding of why certain metals, like Cu, Ag, Ni, and Pt, are effective at recrystallization of ZnS while other metals like Co and Fe are not effective is the subject of future work.

What is proposed here is that the so-called “active” metal recrystallization promoters, typified here by platinum, react with sulfur coming out of the ZnS during

HIPing. Removal of sulfur from ZnS creates open lattice sulfur sites into which oxygen can diffuse from the twin boundaries. This dissolving of the oxygen, initially lying at the nanopolytype twin boundaries, back into the cubic lattice allows recrystallization to take place. This creates large new crystallites with relatively uniform oxygen concentration separated by twin boundaries enriched in oxygen. These large recrystallized grains grow to impingement along the twin boundaries then thicken to impingement in the other direction in an anisotropic exaggerated grain growth mode.

HIPing or even annealing at high temperature (>850 °C) alone without a recrystallization promoter can allow the oxygen diffusion and dislocation motion necessary for recrystallization, but the promoter metal foils allow recrystallization to occur down to 750 °C or lower in the HIP. An absolute lower limit for recrystallization with the foils was not determined, but it is likely that it must be greater than the deposition temperature of the CVD ZnS (~670 °C). The time required at temperature is

272 more for thicker samples to achieve the same effect of transmission improvement via recrystallization.

The addition of an active foil lowers the temperature required for full recrystallization as shown by a couple of examples. A sample (HIP15-Co) HIP’d with an inactive metal, cobalt, at 950 °C shows texturing nearly identical to that of high temperature platinum HIP’d samples and high visible transmission. A cobalt sample

HIP’d at 750 °C (HIP60-Co) shows transmission unchanged from the original CVD ZnS, thus indicating no recrystallization. Thus it can be said that cobalt is inactive at recrystallizing ZnS.

The same low temperature 750 °C HIP conditions with silver (HIP33-Ag) produce high visible transmission. The texture has strongly changed in favor of

{ 111 }

and

{ 211 }

, those planes most associated with material fully recrystallized with platinum at high temperature. Material HIP’d at 750 °C without metal (HIP33none) has remained in the

{ 100 }

texture associated with the as-deposited material. Though the microstructure of this particular sample was not investigated, a similar sample (HIP57none) showed only the first signs of recrystallization, and the low transmission of both samples suggests their microstructures are similar and only beginning recrystallization. Microstructures of samples HIP’d at this temperature and time with active metals (e.g. HIP62-Pt, HIP33-

Agf) are fully recrystallized and have high visible transmission.

Additionally, both samples HIPd without foil at low temperature (HIP33none,

HIP57none) still show signs of the hydride absorption. Since the samples of the same thickness HIP’d at this temperature with Ag or Pt do not have any remnants of this 6 μm

273 absorption, this suggests that one of the other roles of the metal foil is to facilitate removal of the hydride. Zinc hydrides present in sulfur sites will break apart, releasing the hydrogen which diffuses out of the ZnS. Zinc stays in sulfur sites as an antisite point defect, diffuses to a zinc vacancy.

On additional possible mechanism exists for the efficacy of the metal in recrystallizing the ZnS and improving its transmission. It has been shown in the literature that small amounts of metals such as copper or silver in ZnS phosphors can form sulfides during heat treating of the powders. The energy evolved from the formation of these sulfides nucleates the hexagonal to cubic ZnS phase transition. This is more likely to be important with the sputtered metals where intimate contact between the

ZnS and metal is assured. As has been mentioned, however, these sputtered metals have a greater tendency to diffuse into the ZnS. This may create small metal particles which can have surface resonance effects that absorb in the infrared [173]. Though this was not shown in detail in this dissertation, this surface plasmon resonance may be responsible for the infrared absorption in ZnS HIP’d with silver at higher temperatures.

8.3 Conclusions

The goals of this dissertation were to understand the transmission loss behavior of various forms of ZnS and to elucidate the mechanism of transmission improvement through hot isostatic pressing in the presence of platinum.

Degradation of transmission in the visible and near-infrared can be correlated to hexagonal phase content. However, scattering cannot be modeled by a simple Rayleigh

274 model assuming a difference in refractive index between the cubic and hexagonal phases, as the slope of the scattering is much smaller than the wavelength to the inverse fourth power characteristic of Rayleigh scattering. A pseudo-empirical model was generated which relies on scattering at surfaces between nanosized lamella with slightly different refractive indices. This model estimates the transmission of standard CVD ZnS, elemental ZnS, and multispectral ZnS very closely. No additional complexities were added to the model to account for internal electric fields, size distributions, or other features known to be present in ZnS. Adding these additional complexities to the model could allow better predictions assuming effects of coherent and incoherent multiple scattering.

The ZnS investigated in this study shows a large range of optical extinction behaviors. Hot-pressed powder ZnS (IRTRAN) showed localized absorptions from impurities and Rayleigh scattering due to pores. CVD ZnS showed the unconventional scattering behavior modeled above, as well as hydride absorptions described as ZnH

2

and

ZnH

3

species on sulfur sites. Samples which were stopped in the midst of recrystallization showed large particle scattering behavior independent of wavelength.

ZnS HIP’d with silver at high temperature showed broad-band extinction probably due to resonance of indiffused silver particles reflecting and absorbing the infrared light. This last phenomenon deserves future inquiry.

Recrystallization of the CVD ZnS, whether by annealing or hot isostatic pressing, has emerged as the most important function of the heat treatment. HIPing has the added benefits over annealing of allowing processing at temperatures that normally evaporate

275

ZnS, as well as providing the additional driving force of pressure for dislocation motion and martensitic phase transformation. The platinum foil was found to be far from inert, as has traditionally been assumed. Though recrystallization is possible without an active metal promoter, the metal allows it to occur at lower temperatures. Other active metals aiding recrystallization have been identified. Silver was the one most explored while copper and nickel were discovered later and have yet to be optimized. Platinum still remains the most effective for recrystallizing thick ZnS at low temperatures.

This dissertation offers a large body of data and new insights to the literature.

A large number of commercial grades of CVD ZnS have been compared using various techniques and shown to be distinctly different in terms of their defects, particularly with respect to the hydride absorption and the defects responsible for photoluminescence behavior. Photoluminescence of CVD ZnS is reported here for the first time including excitonic effects, as previous luminescence studies of CVD ZnS have not included band edge effects or were done using cathodoluminescence.

Additionally for the first time, the crystallographic and electronic defects of CVD

ZnS has been characterized as a function of its growth position with respect to the mandrel, and shown to be highly variable. Compression data in the anisotropic orientations of ZnS are shown here for the first time.

The new model of internal surface scattering and the importance of recrystallization for transmission improvement have already been described. The structure of CVD ZnS and recrystallized ZnS has been explained by its tendency toward

276 nanostructural polytypism and polytype induced exaggerated grain growth, both of which are related to impurity segregation of oxygen at twin boundaries.

Insights into the HIP process of ZnS have been offered as well. The band edge of

CVD ZnS has been investigated in detail here, showing that heat treating moves the edge further into the ultraviolet indicating an absorption phenomenon separable from visible scattering. The role of the metal had previously been described as “inert” or “protection” from the graphite reducing environment of the HIP chamber. To the contrary, platinum and other metals have been shown to be active in the recrystallization process.

Identification of the materials on the surface of the metal foils is presented for the first time, lending credence to the importance of the metal in reacting with sulfur.

8.4 Suggestions for Future Work

Despite being known as a phosphor for over 100 years, the luminescence behavior of ZnS in its various forms still defies coherent and consistent explanation. The presence of blue, green, and combination bands in various amounts in the samples studied in this work cannot be easily explained by electronic defect models available in the literature.

Better structural characterization of point defects could be gained by using photoluminescence excitation (PLE) to vary the input excitation energy for luminescence and time resolved methods for viewing spectral shift and decay. More complicated and expensive techniques like electron paramagnetic (or spin) resonance (EPR/ESR), positron annihilation spectroscopy (PAS), or capacitance methods like deep level transient spectroscopy (DLTS) provide information about defect charges due to the inclusion of

277 electric and magnetic fields[43]. Even better are methods that combine structural and electronic knowledge of defects such as scanning cathodoluminescence spectroscopy

(SEM-CL) and imaging combinations with EPR such as optically detected magnetic resonance (ODMR)[146].

Absorption and scattering measurements could also aid in understanding electronic and microstructural defects. Absorption or emissivity measurements at low temperatures on very thin samples would help determine the presence of point defects and impurities, especially those present in low concentrations. This work could further probe the possibility that zinc oxide is responsible for the 5 – 7 μm bands in ZnS. Further work is certainly needed to understand the intense infrared absorption presented by the high temperature silver HIP samples, possibly due to silver particle incorporation and resulting absorption resonance and large particle scattering. Bidirectional transmittance distribution function (BTDF) and integrating sphere scattering measurements may also be able to identify these as metal particles.

In microscopy, more detailed investigation of the microstructure and nanostructure of ZnS will aid in producing better scattering and diffraction models. The advent of electron back scattered diffraction (EBSD) in scanning electron microscopes allows identification of crystal orientations at the same time they are imaged. This technique could lead to new insight about the texturing of CVD ZnS and the changes it undergoes during recrystallization. Careful chemical analysis of twin boundaries in CVD and HIP CVD ZnS should be carried out using transmission electron microscopy (TEM) combined with x-ray energy dispersive spectroscopy (EDS) and TEM electron energy

278 loss spectroscopy (EELS) similar to what has been done with natural sphalerites.

Convergent beam electron diffraction (CBED) in TEM may allow individual nanotwin lamella to be investigated and stacking sequences determined.

On a more macroscopic scale, there is still work to be done in understanding the banding and growth layers in CVD ZnS. Cores of Red ZnS and elemental ZnS were not available, and when they are these should be similarly characterized. Also, variation in these layers could be assessed as a function of position in the CVD deposition chamber.

If possible, micro-x-ray diffraction should be used to identify any hexagonality, texture, or lattice parameter differences between the growth bands that could help explain their origin. The formation, structure, and optical properties of the macroscopic “hillocks,” visible mostly prominently in elemental ZnS (and in standard ZnS as well, just more difficult to see due to its visible opacity), deserves more attention. It may be that these features are, at their root, associated with gas phase created hexagonal crystallites which have embedded in the growing ZnS body.

In the areas of recrystallization promotion, there are still a number of unknowns.

Given more time, I would have liked to investigate the phase transformation using only pressure, such as with a cold isostatic press (CIP), to see if recrystallization could be induced without elevated temperature. The exact mechanism for the promotion of recrystallization by certain metals and not others needs some dedicated work. This current dissertation has only opened the door to the possibility of understanding the roles that these and other metals have on inducing recrystallization and grain growth in CVD

ZnS.

279

Finally, a better understanding of the chemistry of the CVD ZnS deposition chamber is sorely needed. The lack of information about the actual mechanism for ZnS formation and deposition during bulk chemical vapor deposition heeds further optimization of the process. The installation of a mass spectrometer such as a residual gas analyzer (RGA) somewhere in the gas stream would give some insight into the species and levels of contaminants in the process. Detailed investigation of purity of the process gases such as argon, hydrogen sulfide, and hydrogen is an easy first step at understanding the full complexity of gas species available for reaction in the CVD chamber.

8.5 Final thoughts

I am definitely indebted to the serious work on ZnS carried out in previous generations which established the basis of understanding of the physics and chemistry of this material. In the 1960s many corporate labs such as Westinghouse, Bell, GE, IBM,

RCA, and Phillips had well-funded programs to explore ZnS as a semiconductor material and phosphor. The Raytheon Research Division developed CVD ZnS in the 1970s, but since then dedicated funding towards the understanding of ZnS has been spotty at best. I was fortunate to discover the vast Russian literature on ZnS, some previously untranslated. It is my hope that the heyday of basic science and research and development funded and published by big corporations is not over but only suffering a hiatus. Clearly much of the work described in this dissertation would not have been done without the continued interest by Raytheon in CVD ZnS.

280

At the close of this dissertation it is worth reconsidering the mineral name for cubic ZnS—sphalerite from the Greek σφαλερός (sphaleros) meaning treacherous. The

Ancients demonstrated their particular wisdom in their naming of this material, which has required careful and diligent study I order to pry loose any of its secrets. Though known as an infrared window material for now over 50 years and as a phosphor for twice that,

ZnS continues to offer new surprises. With the advent of new characterization techniques to probe the chemistry and physics of its structure, ZnS is sure to be a fruitful material for study for decades to come.

281

9 APPENDIX A: Detailed Crystallography

9.1 Structures of ZnS

The designations alpha, beta, and gamma ZnS are not used consistently in the literature so will not be used here. In mineralogy, α is wurtzite and β is sphalerite, while in phase diagrams, α is usually the low temperature form (sphalerite) β the high temperature form

(wurtzite)[30]. Both the rhombohedral 3R form and the rocksalt form have been called γ-

ZnS. Details on crystal system notation after Allen and Thomas[192]; Strukturbericht: B refers to AB compounds; Pearson: crystal system (c=cubic, h=hexagonal), Bravais lattice symbol (F =face-centered cubic, P = primitive), and number of atoms in unit cell;

Schoenflies: T= tetrahedral, C=cyclic monaxial; d =diagonal mirror plane, v=vertical mirror plane; international notation also known as Herrmann-Mauguin and has abbreviated and full forms.

Sphalerite/

Zincblende

Face-centered-cubic

(FCC)

Wurtzite Hexagonal-close-

Distorted packed (HCP)

(FCC)

3C cF8

43m

T d

2H hP4 6mm

C

6v

C

4h

F

(no. 216)

T d

2

P6

C

43

3

6v

Fm3m

O h

5

4

m mc

B3 Stable

T<1020 C

B4 Stable

T>1020 C

B1 P=14-65

Orthorhombic Cmcm

GPa orthorhombic

Rhombohedral

Rhombohedral 3R* 3m

C

3v

R3m

C

3v

5

Rhombohedral

Rhombohedral 8L* 3m

C

3v

*and related polytpes

P3m1

C

3v

1

[27]

[27]

[10]

282

9.2 Polytypes of ZnS

Reasons for formation of these periodic structures of one-dimensional disorder arise from consideration of close-packing of equal spheres and the three crystal systems that accommodate this. Briefly, close-packing can result in cubic, hexagonal, or trigonal crystal systems[192]. Cubic systems have 4-fold plane symmetry and are usually indexed to a cubic lattice but can also be transformed to a hexagonal lattice. Hexagonal systems have 6-fold plane symmetry and can only be described by a hexagonal lattice.

Trigonal systems have 3-fold plane symmetry and can be described by a rhombohedral lattice or a hexagonal lattice. It is obvious, then, that the common lattice is hexagonal and it is often convenient to describe all the structures with a hexagonal basis, as is often done for studying the x-ray reflections of polytypes.

Layer arrangement and periodicity of polytypism has typically been explained by either thermodynamic theories or dislocation theories [27, 39, 193, 194].

Thermodynamic theories attempt to explain stable polytypes by their free energy and, for

ZnS polytypes, the energies for a large range of stacking sequences are found to be very similar[195]. Thermodynamic theories alone, then, do not explain why one polytype and not another is formed. Dislocation theories, on the other hand, easily explain the periodic sequences of polytypes. Movement of partial dislocations and presence of a screw dislocation can model the periodicity observed in polytypes. These structural theories, however, cannot easily explain why only a few layer sequences are commonly observed.

An explanation combining elements of both these approaches has been offered by

Mardix[29], who has worked extensively with ZnS polytypes. In this model, Shockley

283 partial dislocations glide across the basal plane and provide the mechanism by which solid state transformations occur between the high temperature hexagonal (2H) phase and the low temperature cubic (3C) phase. The presence of a screw dislocation imposes the periodicity on the structures which gives rise to repeating as long as hundreds of angstroms. The driving force for moving partial dislocations is the free energy difference between parent and transformed structure. Stability of the hexagonal phase decreases as temperature is reduced, and hexagonal stacking sequences are replaced with cubic ones at lower temperatures. This reduction in hexagonality has been observed to occur spontaneously at room temperature [196]. The polytype structure results from local minimization of free energy in the absence of sufficient activation energy to move to still lower energy states. Similarly, Engel [197] has shown, using a Monte Carlo approach, that the screw dislocation is performing the ordering during transformation, and that without it a metastable disordered cubic phase results.

Small amounts of hexagonal phase are typical in chemical vapor deposition zinc sulfide, and have been characterized as distorted layer stacking composed of stacking faults, twin boundaries, and dislocation pileups [102]. When reheated above 500 °C, the stacking faults are mobile, and tend to realign towards the more stable cubic phase [81,

198]. The main polytype reported in Russian grown CVD ZnS is the 8H polytype, which is most prevalent on the mandrel side of the chemical vapor deposited material [147]. It has been said that twinned cubic structure in polycrystalline zinc sulfide is favored by near stoichiometric ratios of zinc to sulfur [199].

284

9.2.1 Notations for stacking

Various notations have been put forward and are in use in the literature to describe periodic stacking sequences and these are summarized in the book by Verma and Krishna[30]. One of the most common is the ABC notation. In this notation, each letter corresponds to the position of a plane of equally-sized spheres relative to another .

A sequence of ABAB is the simplest hexagonal close-packed lattice, while the sequence

ABCABC is the simplest cubic close-packed lattice. In structures where more than one atom type is involved per layer, sometimes the notation is extended to Greek letters for the anion, for example. Cubic zinc sulfide, then, would be represented by AαBβCγ with

Roman letters for Zn atoms and Greek letters for S atoms. For wurtzite, the close packed plane is (0001) in the stacking direction [0001], and for sphalerite, the (111) plane is close packed in the [111] stacking direction.

Another useful construct is that sometimes referred to as the Hägg notation. In this construct, a “clockwise” sequence A→B→C→A is denoted as a “+”, 1, or ↑, and a

“counterclockwise” sequence of C→B→A→C is denoted as a “–“, 0, or ↓. Sometimes this is referred to as a “spin.” If the number of positives is denoted p and the number of negatives denoted n giving the total number of layers N=p+n in a repeat unit, a lattice can be shown to be hexagonal if p-n≡0(mod3) and rhombohedral if p-n≡±1(mod3) (see below).

Zhadnov numbers, similarly, give the widths of bands of parallel spins [200]. In other words, each pair of numbers reflects the number of layers of hexagonal then cubic types, so there must be an even number of these in a Zhadnov sequence to describe a

285 periodic layer. The order is not important, as long as the pairs stay together. Another way to look at this is that each component of the Zhdanov number is the sum of the number of layers in a positive sense in the Hägg notation followed by the sum of the number of layers in the negative sense in the Hagg notation. Almost 200 ZnS polytypes have been identified by their Zhadnov numbers and tabularized by Mardix [29].

Ramsdell notation is the most compact, and consists of a number for the period of the stacking and a letter denoting the lattice type, C for cubic, H for hexagonal, and R for rhombohedral. In ZnS polytypes, a third letter L is used to denote polytypes of the space group P3m1 (International notation) or C

3v

1

(Schoenflies) [29]Ramsdell notation is most frequently encountered, and the other notations are useful when describing faulting and specific stacking sequences, since a given Ramsdell number can have several different unique stacking sequences, as described by the Zhdanov symbols [201].

Finally, the Jagodzinski, or hc notation, simply indicates whether the neighboring layers are alike (h) for hexagonal, or different (c) or (k) for cubic[194]. A polytype can be seen as a periodic arrangement of alternating hexagonal and cubic domains, which can be illustrated by Jagodzinski’s notation. Stability of these domains is a function of temperature and the domain size. Domains need not be polytypes, but can be incoherent sequences which ,when viewed by x-ray diffraction, exhibit broad peaks [201] sometimes termed “disordered structure” or DS [201].

It is obvious that all of these notations are related, and this is most easily seen by comparing them to the ABC sequence and to each other. Such a comparison of these

286 various notations for the same structures is provided in Figure 9-1 for an arbitrary stacking sequence and Table 9-1 for some common polytypes.

Table 9-1: Some of the more important polytypes of ZnS.

In the Ramsdell notation, the number refers to the number of layers in the repeat units in cubic (C) and hexagonal (H) lattices, and three times the number for rhombohedral (R) structures, since these are referred to the hexagonal lattice. This data is compiled from

Roth[27], Verma and Krishna[30], Smith [7] and Mardix [29].

Stacking sequence &

Hägg notation

AB (wurtzite)

+ -

Ramsdell notation

Zhdanov notation

Jagodzinski notation

Percent hexagonality

100

9R 21

3

hhc, , etc 66.7

Powder Diffraction file

36-1450, 75-1534,

75-1547, 79-2204,

80-0007

ABCBCACAB

+ + - + + - + + -

ABCB

+ + - -

ACABCBCABA

BCAC

-+++-+++-+++

ABACBCACBA

BCBAC

ABCACB

+++- - -

ABCACBACAB

CBACBCABAC

B

ABACBABC

+- - - - + + +

ABACBACABC

+- - - - - + + + +

ABCABCACBA

CB

++++++- - - - - -

ABCABCABAC

BACBABCABC

ABACB

(+++++++- - - - -

- -+++++++ - - -

)

3

ABC (sphalerite)

++++ or - - - -

4H 22 hchc 50 1-089-7334

12R (31)

3

6H

21R (3112)

3

(hcchccc)

3

28.6

2191, 1-089-2156

72R (7773)

3 though other sequences exist in ZnS

50

(hcchc)

3

40 1-089-7386

12H 66 hccccc 16.7 1-089-2155

3C ±∞ or 30

(hcch)

3

(cccccchccccc chcccccchccc)

3 c

12.5 1-089-2208

0 05-0566, 77-2100,

80-0020

287

3 6 5 4

+ + + - - - - - - + + + + + - - - -

ABCACBACBABCABCBACB

c c h c c c c c h c c c c h c c c

Zhadnov

Hägg

ABC

Jagodzinski

Figure 9-1: Correspondence of various notations for stacking sequence after Gosk [202] and Verma[30]

9.2.2 X-ray methods and fault modeling

Single crystal polytypes of ZnS have typically been investigated by oscillating crystal x-ray diffraction methods (e.g. work by Sebastian and Krishna [193, 203, 204]).

Intensities are plotted versus L or L/m, where L is the coordinate on the reciprocal lattice vector for the characteristic 10.L (hexagonal notation HK.L where H, K, and L are the

Miller indices) reflection and m is the number of repeat layers in a given polytype (i.e. 4 for 4H). Due to the symmetry of close-packing, determining the reflections along this reciprocal lattice vector is sufficient to describe the periodicity of the crystal. Reciprocal lattice rows HK.L are either the type H-K=0mod3 (i.e. H-K=3m) or type H-K=±1mod3

(i.e. H-K=3m±1). Smith [7] has tabularized the relationship between the cubic notation and hexagonal notation and indicated which reflections are to be expected for 3C, 2H, and the polytypes 4H, 6H, 9R, 15R, and 21R.

The crystal is oscillated around the c-axis between 10.5-20.5 degree angle between the a-axis and the x-ray beam, giving the 10.L row for -0.8 ≤ L/m ≤ 0.8 which corresponds to approximately 26º < 2θ < 34º in a powder diffraction pattern. For polycrystalline samples, a powder pattern is more useful, and an invaluable tabulated cross-reference of reflections and 2θ angles is given in Palosz and Przedmojski [201].

288

Almost every other reference on measured or simulated polytype x-ray spectra plots intensity versus L/m from reflections of single crystals.

In close packed structures, four kinds of stacking faults can be present [205]. A

growth fault (sometimes called a twin fault) results from the addition of an extra layer while subsequent layers follow the proper stacking sequence. A deformation fault occurs when parts of the crystal slip past one another on the basal plane, where the slip vector

(or stacking offset vector) can be

( / 3)[1 100]

,

( / 3)[01 10]

, or

( / 3)[1010]

in hexagonal notation. In this example, is easy to see how the use of the Miller-Bravais (hkil) indices enables recognition of equivalent directions. Thirdly, a layer displacement fault displaces only one or two layers, and nucleates by the accumulation of vacancies then expands by diffusion to neighboring regions. Finally, an extrinsic fault involves the insertion or deletion of an entire layer of a close-packed structure and thus requires much higher energy, making its occurrence infrequent. These various faults are shown in Table

9-2, with fault planes indicated by underlining, a vertical line, or a box following convention.

These various fault types are important because their effects on the x-ray diffraction pattern can be calculated and compared to measured data. This gives insight into the mechanisms of faulting, phase transformation, and structure. Various computational methods have been used to predict the scattering intensities of zinc sulfide and other close-packed structures in the presence of stacking faults or long period polytypes. The most straight-forward approaches utilize the structure factor and the probabilities of growth faults [193] and deformation faults [206] to develop analytical

289 relations for the scattered x-ray intensity. Consequently, the probability of a growth fault or deformation fault can be determined using measured x-ray diffraction data. The implicit assumption here is that faults are distributed at random.

Table 9-2: Examples of stacking fault types

After Sebastian[205]. In some cases the faults are difficult to visualize without looking at the stacking sequence of multiple layers. An example of the latter is presented in

Sebastian and Krishna [193]. Detailed compact notation for various fault configurations is described in the International Tables for Crystallography [200]. It can be appreciated here that a growth fault results in one layer change from c↔h whereas a deformation fault results in two consecutive layers changing from c↔h.

ABCABCABCABC

ABCABCBACBA

c c c c h c c c c growth fault in 3C growth fault in 3C

ABCABC│BCABCA deformation fault in 3C

c c c c h h c c c c deformation fault in 3C

ABCABACBCABC

ABCABABCABC layer displacement in 3C extrinsic insertion in 3C deletion

ABABCBCB growth fault in 2H

h h c h h h growth fault in 2H

ABAB│CACA deformation fault in 2H

h h c c h h deformation fault in 2H

ABABCBAB layer displacement in 2H

ABABCAB extrinsic insertion in 2H cannot occur because AA or BB not allowed

Similar models using three parameters (fault probability for random insertion of deformation fault and fault probabilities for deformation faults at 3 and 2 layer separations) have been applied to the phase transformations of ZnS and ZnS with Mn and

Cd and have been compared with measured data with reasonable agreement. Early models by Sebastian and Krishna found the disordered 3C structure resulted from the non-random distribution of growth faults [193], but that the 2H-3C phase transformation resulted from the non-random nucleation of deformation faults from thermal stresses during cooldown after annealing [206].

Other models are necessary when it is assumed that faults are not distributed at random, such as might be the case with a dislocation mechanism for faulting. Reverse

290

Monte Carlo methods have been developed by Gosk for faulting in 3C [202], 2H [207], and 4H [208] structures. In this model, disordered structure (DS), growth stacking faults, deformation stacking faults, and extrinsic stacking faults are examined with the intensities plotted versus the 10.L/m reciprocal lattice row as described previously. It is clear that different stacking fault mechanisms and model parameters radically change the output diffraction intensity and position as predicted by the model. These Monte Carlo type models incorporate the previous generation of models while providing more possibilities for exploring the diffraction effect of complicated non-random stacking sequences.

One final set of models for ZnS polytypes is worthy of mention. Varn and colleagues [209-211] have used computational mechanics to create recursive genetic algorithms that fully describe the stacking sequences of many polytypes using simple stochastic processes. Unlike previous “fault models”, those espoused by this group to not presume a mechanism for faulting, but only seek first to adequately describe the stacking sequence in terms of a set of spin-flip operations. In principle, this method does not distinguish between a starting structure and a fault, only the relative arrangement of atoms. All the information from a diffraction spectrum is taken into account in assessing the stacking, including the diffuse background, not just the Bragg peaks as most other models use to match to measured data. An exact stacking sequence giving rise to the diffraction is not found, rather a statistical description of the disorder. Using this method, they have shown that a sequence of hexagonal layers transforms into a set of cubic

291 twinned domains when undergoing a simulated deformation fault process. This group’s work is ongoing and available in the Sante Fe Institute working papers.

It was considered whether or not CVD ZnS contains some polytypic forms intermediate between hexagonal and cubic. Figure 9-2 shows some possible evidence for this. However, given that the noise in the measurement is high compared with the peaks, in addition to the broad background on either side of the sphalerite (111) peak in most untreated samples, it seems unlikely that true polytypes exist. It may be more correct to say that at least some of the crystallites at the lowest level of structure in ZnS exhibit a disordered stacking sequence that is mostly cubic with some hexagonal elements. In some very local regions these might be called short period polytypes, but there are probably only a few repeat sequences.

Figure 9-2: XRD of polycrystals and possible evidence for short period polytypes

Several different samples of standard CVD ZnS are shown. Planes are indexed in hexagonal notation except where noted, with HK.L/M reflection plane shown, as described in Chapter 2.

292

10 APPENDIX B: Electronic Structure

10.1 Electronic band gap and absorption edge

Sphalerite

Wurtzite

Direct energy gap (eV)

@ 300K

3.4 (E therm

3.54 – (c)

) –(e)

3.6 – (a)

3.7 (E ) – (e) opt

3.7 – (b,j)

3.68 – (g)

3.5 (E therm

3.67 –(c)

) – (e)

3.8 – (a, b)

3.8 (E opt

) – (e)

3.78 (E perp c) – (j)

3.74 (E para c) – (j)

3.8 (E perp c, theor) – (j)

3.92 (E para c, theor) – (j)

Direct energy gap (eV) @ T K

3.84 (4 K) – (b)

3.833 (4 K) – (e)

3.824 (77 K) –(e)

3.882 (77 K) –(e)

3.910 (4 K) – (e)

3.911 (4 K) – (b)

3.913 (4 K) – (e)

3.901 (77 K) –(e) dE g

/dT

(meV/K)

-0.53 – (a, c)

-0.47 – (g)

dE g

/dP

(meV/GPa)

57 – (a, c, g)

-0.38 – (a) 90 – (a)

Electron mass m e

*/m

0

Hole effective mass m h

*/m

0

0.58 – (d)

Electron mobility: μ

(cm

2

/(Vs)) e

Hole mobility:

μ h

(cm

2

/(Vs))

10 – (d, h)

Sphalerite 0.39 – (a, b, h)

0.34 – (d)

0.34±0.02 – (f)

200 – (e)

180 (r.t.) – (c)

160 – (d)

230 (theor max, r. t.) – (i)

140 – (e)

Wurtzite

0.28 – (a, b, k)

0.28±0.03 – (f)

0.27±0.03 perp – (h)

0.39 – (h)

>1 para, 0.5 perp – (a)

1.4 para, 0.49 perp – (b)

1.4 (heavy hole), 0.49 (Light hole) – (k)

>1 para, 0.51±0.05 perp – (h)

0.67 perp – (h)

5 (673 K)

– (c)

(a) – Pankove [212]

(b) – Yen [46]

(c) – Lide [213]

(e) – Morozova [15]

(f) – Kukimoto [215]

(g) – Yu [64]

(i) – Ruda [217]

(j) – Adachi [218]

(k) – Thomas and Timofeev [219]

(d) – Park [214] (h) – Segall [216]

Recommended values in bold and green. “Perp” and “para” are in reference to the c-axis of wurtzite. For polytypes, band gap varies linearly with degree of hexagonality, becoming smaller with more hexagonal phase [220]. The band gap changes with temperature and pressure per the equation [221].

E g

=

E

0

g

+

⎜⎜

E

P g

⎟⎟

T

Δ

P

+

⎜⎜

E g

T

⎟⎟

P

Δ

T

293

Table 10-1: Room temperature ultraviolet cut-on edge of ZnS

Various materials as determined from transmission measurements or from the literature.

Material

ZnS, hexagonal

ZnS, “pure” (hexagonal)

ZnS, (hexagonal)

ZnS, polytypes

ZnS, hexagonal with stacking faults

ZnS, cubic 0.1mol% O

ZnS, cubic

ZnS, cubic, Bridgman

ZnS, “intrinsic” cubic

ZnS, N anneal

2

ZnS, CVD, H

2

anneal

ZnS, cubic crystals

ZnS, CVD any, anneal or HIP

ZnS, cubic 0.6mol% O

ZnS, CVD, yellow

ZnS + 1mol% O

ZnS, CVD (H

2

S or S process)

ZnO, hydrothermal

ZnS + 2mol% O

ZnS with ZnO·S precipitates

ZnO, “pure” hexagonal

ZnS, CVD (red)

ZnS, CVD, orange-yellow

UV edge (nm) Notes

~329 (para)

335 Must be c-axis (i.e.

E field perp. to caxis)

334.5 (perp)

331.5 (para)

335 - 339 More hexagonality, shorter UV edge

330 -> 340

(para)

335

~339

Introduction of stacking faults into

2H

Isolated O

S

340 MTI crystals

340

Reduced, Zn rich

Reference

Shachar [222]

Kroeger [23]

Brafman [181]

Shachar [222]

Shachar [222]

This work

Kroeger [23]

340

341.0

342-346

350 Widely separated

670 °C deposit,

H

2

S/Zn = 0.7

This work

Brafman [181]

This work

Morozova [15]

350

350

361-386

387

390

O

S

-O

S pairs

950 °C deposit,

H

2

S/Zn = 0.8

MTI crystals, c-axis

385-400

Collins [53]

Kroeger [23]

This work

This work

Kroeger [23]

385, 400 Due to random orientation of c-axis

Kroeger [23]

400-415

450

This work

Collins [53]

294

10.2 Band Structure of Sphalerite and Wurtzite at Zone Center

Below is shown the crystal field splitting and spin-orbit splitting of the bands in wurtzite and sphalerite, after Segall [216] and Birman[223]. Note that J refers to the total angular momentum of the spin-orbit coupling. Values for the valence band splitting and band gap are from Morozova [15] and are stated at 77 K. Letters A, B, C correspond to the absorptions from the named valence band to the n=1 exciton just below the conduction band.

E g

=

3 .

824

eV

Δ

SO

=

0 .

064

eV

Γ

6

Γ

8

Γ

7

Γ

1

Γ

7

Γ

8

Γ

6

Γ

6

Γ

5

(

x,

y

)

Δ

CR

Γ

1

( )

Δ

SO

=

0

Γ

Γ

4

1

Δ

SO

= Δ

CR

=

0

J

J

=

=

3

2

,

m

J

3

2

,

m

J

= ±

3

2

= ±

1

2

J

=

1

2

,

m

J

= ±

1

2

Γ

6

Γ

8

(

A

)

Δ

SO

Γ

7

(

C

)

Δ

CR

=

0

Γ

7

Γ

9

Γ

7

Γ

7

Γ

7

Γ

9

(

A

)

Γ

7

(

B

)

Γ

7

(

C

)

Γ

7

Γ

9

Γ

7

Γ

7

E g

=

3 .

901

eV

Δ

CR

=

0 .

028

eV

=

0 .

116

eV

295

10.3 Point Defects in ZnS

10.3.1 Kröger Vink and charged defects

Standard notation used is due to Kröger and Vink. Charged defect is indexed with respect to the host lattice charge. Defect becomes a donor or acceptor as it is ionized.

Defect

Sulfur vacancy

(F

2+

center)

Sulfur vacancy + 1 electron

(F

+

center)

Sulfur vacancy + 2 electrons

Zinc vacancy

Normal lattice site

charge

2+

Zinc vacancy + 1 hole

Zinc vacancy + 2 holes

Zinc interstitial

Zinc interstitial +

1 electron

Zinc interstitial +

2 electrons

Isoelectronic oxygen

Monovalent cation

(Ag, Au, Cu)

Trivalent cation

(Al, In, Ga)

Monovalent anion

(Cl, I, Br)

Zn at S site

ZnH at S site

0

2-

2-

Charge on defect

Relative

charge

Notation Function

0

2+

2+ ˙˙ Donor

V

S

˙

2-

S x

Donor

(fully ionized)

V

S x

= V

S

˙+ e’

V

Zn

''

Acceptor

V

Zn

' Acceptor V

Zn

' = V

Zn

'' + h˙

2+

Ionization Reaction

Donor V

S

˙= V

S

˙˙+ e’

Zn x

Acceptor

(fully ionized)

V

Zn x

= V

Zn

' + h˙

Zn i

˙˙ Donor

Zn i

˙

Donor Zn i

˙ = Zn i

˙˙+ e’

i x

Donor

(fully ionized)

Zn i x

= Zn i

˙ + e’

O

S x

Ag

Zn

' Acceptor

Al

Zn

˙

Donor

Cl

S

˙

Donor

ZnH

2

at S site

ZnH

3

at S site

ZnH

4

at S site

“A” Center

2-

2-

2-

0

2+

1+

0

1-

2-

4+

Zn

S

˙˙˙˙

Donor Zn

S

˙˙˙˙ = Zn

S x

+4 e’

3+

(ZnH)

S

˙˙˙ Donor (ZnH)

S e’

˙˙˙ = (ZnH)

2+

(ZnH

2

)

S

S x

+3

˙˙ Donor (ZnH

2 e’

)

S

˙˙ = (ZnH

2

)

S x

+2

1+

0

(ZnH

3

)

S

˙ Donor (ZnH

3 e’

)

S

˙ = (ZnH

(ZnH

4

)

S x

Fully ionized

3

)

S x

+1

1+

' V

S

et al. [224]

1- 1-

(V

Zn

'' Cl

S

˙)' Accepter

296

10.3.2 Defect Equilibria

Approximate band gap levels of intrinsic defects are as shown in Figure 10-1.

Equilibrium defect concentration of native defects can be shown schematically on a

Brouwer diagram, as a function of partial pressure of the species. Below the defect reactions including the Schottky and Frenkel defects as well as their subsequent ionization energies are shown. Albers [225] describes the important of defect equilibria for II-VI compounds in determining expected conductivity as due to electrons or holes.

The following treatment is similar to that of Schmidt-Mende [226] for the ionization reactions and Wiedemeier [47] for the Schottky defect and intrinsic carrier reactions. (Note that 1 kJ/mol = 0.0104 eV)

Intrinsic charge carriers

K

][

h

+

]

i

=

[

e

=

[

n

][

p

]

K i

=

4 { 2

π

m

* 1 /

e

2

m

* 1 /

h

2

kT

) /

h

2

}

3 exp(

E

G

/

kT

) k and h are Planck’s and Boltzmann’s constants, and values for the density of states’ effective masses (m*) and energy gap (E

G

) can be found in Figure 10-1.

Formation of Zn

Zn

Zn x i

and V

↔ Zn i x

Zn

by Frenkel reaction:

+ V

Zn x

Enthalphy of formation of Zn Frenkel defect in ZnS: 5.3eV [48]; 3.32eV (ZB), 5.8eV (WZ) [113]

Enthalpy of formation for S Frenkel defect in ZnS: 6.8eV (ZB), 6.4eV (WZ) [113]

No information was found for the Anti-Frenkel (anion interstitial) defect in ZnS.

Subsequent ionization (ionization energies can be found in Figure 10-1):

Zn

Zn i x i

˙

↔ Zn

↔ Zn

Formation of V

S i

˙ i

˙˙

+ e’

+ e’

and V nul ↔ V

Zn x

Zn

+ V

S x

K = 2N

C exp(-E

K = (1/2)N

by Schottky reaction:

C

1

/kT) exp(-E

2

/kT)

Enthalphy of formation of Schottky defect in ZnS: 5.1eV [48]

Δ

H

°

S

, 298

=

110 .

2 kcal/mol[47], 3.53eV (ZB), 4.4eV (WZ) [113]

K

S

K

S

=

=

[

V

Zn

''

][

V

S

• •

]

since II-VI vacancies are thought to be doubly ionized. exp(

− Δ

G

°

S

, 298

/

RT

)

= exp(

− Δ

H

°

S

,

T

/

RT

) exp(

Δ

S

°

S

,

T

/

R

)

297

Where R is the gas constant,

Δ

H

°

S

,

T

is the Schottky enthalphy defined as the energy required to move a Zn

2+

and S

2-

pair to the surface, and

Δ

S

S

°

,

T

is the entropy which is about 10R or 10 to 20

(cal/deg*mol) [47].

Subsequent ionization (ionization energies can be found in Figure 10-1):

V

V

V

V

S x

S

˙

↔ V

↔ V

Zn x

Zn

↔ V

˙

S

S

˙˙

↔ V

+ e’

+ e’

Zn

Zn

’’

+ h

+ h

˙

˙

K = 2N

C exp(-E

3

K = (1/2)N

C exp(-E

K = 2N

V exp(-E

5

K = (1/2)N

V

/kT)

4

/kT)

/kT) exp(-E

6

/kT)

E1 E2 E3 E4

Zn i

˙˙/Zn i

˙ Zn i

˙/Zn i x

V

S

˙/V

S x

V

S

˙˙/V

S

˙ V

Zn

E5

’/V

E6

Eg,therm

(77K)

Reference

Zn x

V

Zn

’’/V

Zn

0.2 eV 0.1 eV 1.56 eV 2.0 eV

0.2 eV 0.1 eV 1.7 eV 2.17 eV

0.52 eV

0.35 eV

1.9 eV 3.52 eV Morozova [15] fig

1.6-IV

1.9 eV 3.54 eV Morozova [48]

2.8 eV 1.1 eV Neumark [43]

Donor Donor Donor Donor Acceptor Acceptor

Figure 10-1: Approximate intrinsic defect locations in the bandgap of cubic ZnS.

Energies correspond to the equilibrium reactions as shown on the next page. These energy levels correspond to the equilibrium states. The absorption and emission energies of the defect level are different in ZnS (Stokes shift). The electrons interact with the lattice atoms and are slightly displaced by a polaron (interaction between phonon and electron[15]). Absorption energies from a particular defect level are larger energies than emission energies, so the defect level can be thought of as a “band” that extends to either side of the equilibrium defect energy listed above.

298

Brouwer diagrams for defect equilibria have been presented in various reports

(e.g. Gurvich [50], Albers [225]), suggesting that vacancies of Zn and S are the dominant defects. Many of the defect equilibria in the literature do not include interstitials in the calculation. A more comprehensive assessment was done by Morozova [48] which indicated that interstitial zinc and zinc vacancies are the defects with the highest concentrations, depending on zinc partial pressure. Charged Zn defects (vacancies and interstitials) will be present in much higher concentrations than sulfur defects over a wide range of zinc (sulfur) pressures, assuming only Frenkel disorder on the Zn sublattice [22] or Schottky disorder plus Frenkel disorder on the Zn sublattice [48].

A generic diagram for metal chalcogenide systems is given in Stevenson [22], which assumed the Henry’s law line for activity, since defects are in low concentrations.

Examples are given for singly ionized Frenkel defects on the Zn sublattice, as well as the addition of a shallow donor. Another insight gained from the Brouwer type diagram is the change-over from n- to p-type conductivity versus partial pressure of the component.

However, a system such as ZnS is not easy to make a p-type conductor, since the required partial pressures of sulfur determined are higher than the saturation pressure of sulfur[47]. Ion implantation is required for p-type doping of ZnS with sulfur. For comparison, sulfur saturation pressure curves can be found in Vaughan and Craig [79].

Some assessments of defect concentration in ZnS are shown in Table 10-2.

299

Table 10-2: Defect concentrations of oxygen and hydrogen in ZnS

Various literature sources often report number density and atomic (mole) percent. Others, particularly mass spectrometer measurements, report weight percent or parts per million weight (ppmwt). Since atomic percent (or parts per million atoms) is more meaningful in terms of lattice occupation, these units have been preferred here. Also shown are some comments regarding concentrations of oxygen and hydrogen defects in ZnS at certain levels, as well as published general sensitivity limits for particular characterization methods which might be used to identify oxygen defects (and hydrogen for IGA & SIMS).

Parts per Notes density

(#/cm

3

)

Atomic %

(for ZnS)

[Zn]+[S]

[Zn] or [S]

[O]

[O]

[O]

2.5 x 10

1.25 x 10

22

22

Atomic density in 2H or 3C ZnS [15]

# of Zn or S atoms in perfect crystal

3 x 10

20

1.2 12,000 [15]

2 x 10

1.5 x 10

20

20

100

50

0.8

0.6

8,000

6,000

Lower solubility limit of O in ZnS [15]

1.0 wt% ZnO (=0.6 at% O) in ZnS stated at solubility limit in wurtzite [24]

[O] 1 x 10

20

0.4

million atoms

(ppma)

(for ZnS)

4,000 Minimum [O] to see changes in E

G

; decomposition of ZnS·O at this level of [O] and greater leads to ZnO precipitates [15];

0.7 wt% ZnO (=0.4 at% O) in ZnS stated as

[O]

General

[H]

[O]

7.5 x 10

19

3,000 Minimum O concentration to see UV edge shift to from 335 to 350 nm (at 80 K) due to

O

S

-O

S

pairs [15]

5 x 10

19

3.25 x 10

0.2 2,000

19

0.13 1,300 Maximum level of hydrogen stated in CVD

ZnS [100]

2.5 x 10

19

0.3

0.1 1,000 solubility limit for sphalerite [24]

UV edge in sphalerite moves to 335nm (at

80 K) at this [O] due to isolated O

S

[15];

best

[H]

General

[H]

[O]

[O]

[O]

[H]

2.5 x 10

18

detection limit for SEM-EDS (could be order of magnitude worse)

3.75 x 10

18

0.015 150

Detection limit [H] in IGA, <3ppmwt [156]

1.25 x 10

0.01 100

Detection limit for FE-Auger or XPS

18

0.005 50 Mimimum level of hydrogen stated in CVD

ZnS [52]

1 x 10

18

0.004 40 Minimum [O] in ZnS when no special purity precautions are taken;

at this [O] and

1 x 10

14

smaller, O is undetectable by x-ray methods

or activation analysis[15]

; approximate [O] in Czochralski Si

7.5 x 10

2.5 x 10

17

17

0.003 30

Detection limit [O] in IGA, ~10ppmwt [156]

0.001 10 Controlled low level O doping in ZnS [15]

Approximate detection limit for Secondary

Ion Mass Spectroscopy (SIMS)

; values cited vary between 10

12 and 10

16

/cm

3

300

11 APPENDIX C: Thermodynamics and Kinetics

11.1 Sulfur Activity and Gibbs Energy of Creation of ZnS

The chemical potential is defined as:

μ

i

=

⎜⎜

G

n i

⎟⎟

T

,

P

,

n i

j

=

μ

i

* +

RT

ln

a i

=

μ

i

* +

RT

ln

p i p i

*

Where G is the Gibbs free energy, n is the number density, the “*” symbol represents the state at the pressure of interest, and p is the partial pressure of species i.

The reaction

2H

2

S = 2H

2

+ S

2

(reaction I) determines the activity of sulfur. The free energy of the reaction is defined as the difference between the free energies of the products and the reactants, and at equilibrium is equal to zero. Here G is the Gibbs free energy, the “º” symbol represents the standard state value, a is the activity, K is the reaction equilibrium constant, R is the gas constant, and T is the temperature in Kelvin.

Δ

G rxn

,

I

= Δ

G prod

− Δ

G reac

= Δ

G

°

rxn

,

I

+

RT

ln(

K

I

)

=

(

Δ

G

°

S

2

+

2

Δ

G

°

H

2

2

Δ

G

°

H

2

S

)

+

RT

[ln(

a

S

2

)

Then, at equilibrium where ΔG rxn,I

equals zero:

+

2 ln(

a

H

2

)

2 ln(

a

H

2

S

)]

Δ

G

°

rxn

,

I

=

(

Δ

G

S

°

2

+

2

Δ

G

°

H

2

2

Δ

G

°

H

2

S

)

= −

RT

ln(

K

I

)

= −

RT

ln

(

a

S

2

)(

a

(

a

H

2

S

H

)

2

2

)

2

The equilibrium constant for reaction I as a function of T is shown in Sack [78] to be log

10

K

I

= log

10

(

a

S

2

)

+

2 log

10

(

a

H

2

)

2 log

10

(

a

H

2

S

)

=

9480

T

(

Kelvin

)

+

5 .

16

301

The activity coefficients for hydrogen and hydrogen sulfide can be assumed to be ≈ 1 since the CVD chamber is operated at low pressures (< 100 torr). Therefore the activities can be approximated as the concentrations or partial pressures.

a a

H

2

=

[

[ ]

]

=

H S p p

H

2

The activity of sulfur can then be calculated based on the temperature dependent equilibrium constant and the relative concentrations of H

2

and H

2

S (see Figure 11-1).

Increasing the [H

2

]/[H

2

S] ratio in the gas phase then must lower the activity of sulfur for a given temperature.

Consider now the reaction:

2Zn + S

2

= 2ZnS (reaction II)

Compiled reaction data assumes reactants and products except S

2

are in their standard solid state, so for

Δ

G

°

rxn

,

II

Therefore

=

( 2

Δ

G

°

ZnS

2

Δ

G

°

Zn

− Δ

G

°

S

2

)

=

2 ( 1 )

2 ( 1 )

− Δ

G

°

S

2

= − Δ

G

°

S

2

= −

RT

ln

K

II

log

10

K

II

=

2 log

10

(

a

ZnS

)

2 log

10

(

a

Zn

)

− log

10

(

a

S

2

)

= − log

10

(

a

S

2

) and log

10

(

a

S

2

)

=

Δ

G

°

rxn

,

II

=

Δ

G

°

rxn

,

II

(

cal

)

(

Kelvin

)

=

Δ

G

°

rxn

,

II

(

J

)

19 .

148

T

(

Kelvin

)

.

2 .

303

RT

4 .

573

T

At sulfur saturation, the activity of S

2

has been shown by Vaughan [79] to be log

10

(

a

S

2

)

=

6 .

8052

6032 .

3

T

(

Kelvin

)

.

302

Figure 11-1: Activity of sulfur in the hydrogen sulfide reaction

Figure 11-2: Gibbs energy of formation of CVD ZnS from H

2

S.

303

11.2 Oxygen in ZnS: Thermodynamics and Kinetics

It is important to consider the conditions leading to the availability of oxygen for incorporation into ZnS despite the presence of gaseous H

2

byproducts of the CVD ZnS process. Thermodynamically, it can be shown that oxygen has a preference for zinc over hydrogen. This is illustrated by plotting the standard Gibbs free energy change versus temperature (Ellingham diagram) for the two reactions [227, 228]. The change in standard state Gibbs free energy (ΔG°) is lower for the reaction of 2Zn(s) + O

2

(g) =

2ZnO(s) than for 2H

2

(g) + O

2

(g) = 2H

2

O(g) until about 1200 °C (see Figure 11-3 generated from HSC [80]). CVD ZnS deposition temperatures are typically around 700

°C, well within the regime where ZnO is the thermodynamically favored oxide.

Figure 11-3: Standard state free energy change per mole for ZnO and H

2

O

The equilibrium pressure (dissociation pressure) is the pressure where the driving forces for the forward and backward reactions are equal. For an oxide reaction, partial pressures higher than the equilibrium pressure result in oxidation (i.e. a tendency towards

304 the products of the reaction which are zinc oxide and water), while those pressures lower than this value dissociate the products. This pressure can also be seen as the threshold oxygen partial pressure for formation of the oxide. It can be calculated from the standard state Gibbs free energy as:

p

O

2

eq

,

T

= exp

Δ

G

RT

°

The results for H

2

O and ZnO are given in Table 11-1 for 650 °C. Numbers for standard state Gibbs energy are taken from the HSC thermodynamics database [80].

It can be seen that very small partial pressures of oxygen are required to produce zinc oxide, seven orders of magnitude less than to make water from hydrogen. The pressure of a typical

CVD ZnS run is 40 torr (0.053 atm) and of this about 90% is argon. It is almost certain that the partial pressure of oxygen in the CVD chamber is much higher than 10

-20

torr, so both ZnO and H

2

O are favored to be produced if Zn and H

2

are present.

Table 11-1: Dissociation pressures of ZnO and H

2

O(g) at 650 °C

Calculation for 2 moles allows comparison with Figure 11-2 and Figure 3-1.

2 moles 2 moles 2 moles

T P

2Zn + O

2H

2

2

(g) = 2ZnO

(g) + O

2

(g) = 2H

923

2

O(g) 923

-255.98

-196.94

O2,eq

(atm) P

-511.96 1.1 x 10

-29

-393.88 5.1 x 10

-23

O2,eq

(torr)

8.1 x 10

3.9 x 10

-27

-20

The use of the Ellingham diagram reaction for ZnO is not quite right, however, since the activity of oxygen in ZnS is really the activity (i.e. mole fraction) of oxygen in

ZnS

1-x

O x

which is very much less than one. Assume that the partial pressure of oxygen in the gas phase is controlled by the ratio H

2

/H

2

O (it could be controlled by S

2

/SO

2

or some other ratio). The H

2

comes from the decomposition of H

2

S and the O

2

from leaks in

305 the CVD chamber or contamination of the process gases (H

2

S or Ar). New ΔG° curves with negative slope can be derived for the Ellingham diagram based on the lower assumed activity of oxygen. These imply that 1) if H

2

/H

2

O ~ 10

3

, ZnO forms a stable distinct phase; 2) if H

2

/H

2

O ~ 10

6

, ZnS picks up about 10

-4

atomic fraction O; and 3) if

H

2

/H

2

O ~ 10

9

, ZnS picks up ~10

-8

atomic fraction O. In other words, as H

2

/H

2

O decreases, [O] increases in ZnS. H

2

O must build up to a certain concentration in the gas phase, either by reacting hydrogen and oxygen or by residual water vapor. Said another way, the less H

2

is around, the higher the concentration of oxygen in ZnS. The driving force for the first bit of oxygen to be incorporated is high according to Le Chatlier’s principle.

It is likely that there is some hydrogen present in the ZnS CVD chamber, either residual from the elemental ZnS in-situ production of H

2

S, or produced as the H

2

S is broken up when forming ZnS. Some of the free hydrogen, especially that near the surface of the ZnS, will likely form ZnH or ZnH

2

(depending on whether the hydrogen is present in atomic or molecular form) in the gas phase and be trapped in the ZnS body.

Unfortunately, thermodynamic data does not exist for the zinc hydride species, by ab

initio calculations suggest that they may form in the gas phase with only minimal activation energy (~16 kJ/mol electronic energy, calculated for T ~ 0 K) [172]. This energy could easily be provided by the strong exothermic reaction of hydrogen sulfide with gaseous zinc (~ -233 kJ/mol electronic energy, calculated for T ~ 0 K). Database values for the reaction of gaseous Zn with H

2

S show Gibbs energy change of -150 kJ per mole ZnS at 700 °C [80].

306

Some hydrogen will still be in the gas phase, and the possibility for its violent reaction with oxygen must be considered. However, it has been shown that there is a lower critical pressure (P

L

) below which hydrogen and oxygen will not react explosively[229]. The mechanism for avoiding a chain reaction explosion at very low pressures is believed to be deactivation of the chains by collision with other gas molecules. At these low pressures (P<P

L

) the reaction is bimolecular with hydrogen peroxide playing an important interim role (i.e. H

2

+ O

2

= H

2

O

2

then H

2

O

2

+ H

2

= 2H

2

O).

At pressures higher than this, a chain reaction takes place and there is an explosion.

Above a certain upper critical pressure (P

H

), the reaction rate slows again due to chain breaking at the walls of the vessel. At these pressures the reaction is believed to be termolecular (2H

2

+ O

2

= 2H

2

O). This pressure regime is illustrated in Figure 11-4, after

Thompson [229].

Figure 11-4: Reaction rate of hydrogen with oxygen as function of pressure

307

Between the lower and upper critical pressures (P

L

<P<P

H

) reaction rates are extremely high and explosion occurs. At 550 °C, the upper critical pressure (P

H

) for [2H

2

+ O

2

+ 2Ar] was found to be 137 torr (55 torr Ar and 82 torr [2H

2

+O

2

]) [229]. The determination of the lower limit is much more difficult. It has been determined by finding the temperature where there is no ignition at all (i.e. where the lower and upper pressure limits coincide). In one particular vessel, P

L

was found to be 450 °C which corresponded to 8 torr (2H

2

+O

2

only). Below the lower critical pressure (P

L

), the reaction rate is very slow but nonzero [229]. At atmospheric temperatures and pressures, hydrogen and oxygen can remain in contact “indefinitely” without any sign of chemical reaction [230].

Oxygen is more easily ignited by hydrogen when the H

2

:O

2

ratio is 1:4 (i.e. much more oxygen), at which ratio widely diverging ignition temperatures have been reported from 514 to 845 °C [230]. The lower critical pressure for ignition has been determined for a vessel at 500 °C as a function of O

2

and H

2

partial pressures [231]. The curve defining the lower critical pressure is more tolerant of higher partial pressures of hydrogen than of oxygen. For example, 1 torr each of H

2

and O

2

at 500 °C would not ignite. About 6 torr of H

2

and 1 torr of O

2

would still be below the threshold. By 2.5 torr each of H

2

and O

2

the gas mixture will ignite at 500 °C.

Even though example scenario is for a temperature lower than the CVD process temperature in question (650 to 700 °C), it gives a sense of the order of magnitude of H

2 and O

2

pressures required to induce ignition. Based on these examples (and on practical experience) it must be that the combined levels of hydrogen and oxygen in the CVD

308 chamber are below the lower critical pressure for ignition of H

2

with O

2

. Therefore some water is produced at very low rates, but the chain reaction is broken by the other gas molecules in the chamber (H

2

S, Ar) since the pressure is below the lower critical limit.

Since (1) kinetics for reaction of hydrogen with oxygen will be slow, (2) oxygen has a thermodynamic preference for zinc over hydrogen, and (3) zinc has thermodynamic preference for oxygen over sulfur, ZnO will be present in ZnS. The source of low levels of oxygen is likely leaks, residual atmospheric oxygen, low level contamination of process gases (argon, hydrogen sulfide) and zinc raw material [76]. Of these, the ones which will be important through the whole deposition run are leaks and oxygen in argon and hydrogen sulfide. When these gases are specified for purity by gas manufacturers, their total purity levels typically do not include oxygen. Since there is evidence in the literature for oxygen in ZnS produced by CVD (e.g. photoluminescence), it seems likely that the mechanisms responsible for the incorporation of oxygen into ZnS are as described above.

The situation in reality is more complicated than just described, since hydrogen sulfide will react with oxygen and zinc will react with hydrogen. This gas phase equilibrium has not been worked out, owing to the complexity and the lack of thermodynamic data for the formation of important species like zinc hydride. Some possible reaction paths in this system are shown in Table 11-2 . This list is hardly comprehensive, but included to give an appreciation of the types of species that could be

309 present. For example, it has already been mentioned (Chapter 2) that oxygen is a reducing agent in ZnS producing Zn-rich ZnS and SO

2

[23].

Table 11-2: Some possible reactions in the Zn-O-S-H complex equilibrium

Reaction Notes

2H

2

S + 3O

2

= 2H

2

O + 2SO

O + 2S

2

2H

2

S + O

2

= 2H

2

2ZnS + 3O

2

= 2ZnO + 2SO

2

Explodes on ignition

Roasting in air requires T>1000 K; this is the commercial process for making zinc; sulfur dioxide byproduct is used to make sulfuric acid according to

Porter [232] At lower temperatures and reduced oxygen contents this produces Zn-rich (reduced)

ZnS[23].

Reference

Mellor [85]

[5]

Mellor[5], kinetics in

Prabhu [233]

2SO

2

+ O

2

2ZnO + SO

ZnS + H

2

= 2SO

3

3

= ZnSO

4

O = ZnO + H

2

S

ZnS + 3H

SO

2

+ 3H

2

2

O = ZnO + SO

2

+ 3H

= 2H

2

O + H

2

S

Mellor [85]

Mellor [85]

2

Takes place at a reasonable rate at 750 °C in steam Sohn[234]

Unlikely to occur compared to previous

Happens spontaneously if reactant gases present

Sohn[234]

Sohn[234]

11.3 Vapor Pressures

Figure 11-5: Mole fraction and partial pressures of sulfur allotropes in sulfur vapor

Sulfur gas species as a function of temperature, after Rau [90]. Curves on left were digitized from the original so show discontinuities not present in the original data.

Curves are right are plotted from tabulated data in the paper.

310

Figure 11-6: Vapor pressures of various materials important in the CVD ZnS process

Compiled from various source: oxygen from[235], hydrogen sulfide from [92] and NIST web thermo tables (WTT), water from Hyperphysics ( http://hyperphysics.phyastr.gsu.edu/hbase/kinetic/watvap.html

), sulfur from[79, 92, 235, 236], zinc from[92,

237]. There is some disagreement as to the vapor pressure of ZnS, with values from

[238] (triangles) being somewhat lower than from [92] (squares and circles).

11.4 Diffusion

Interatomic potentials in the ZnS system have been determined by fitting experimental data models, and have allowed the calculation of several defect parameters of interest in ZnS [113]. Binding and defect formation energies were calculated for

Schottky defects and Zn and S Frenkel defects for both sphalerite and wurtzite structures, and suggest that S Frenkel defects are uncommon compared to the other two. Arhennius ion migration energies were calculated as well. When migration energies are combined

311 with the defect formation energies, the activation energy for intrinsic diffusion is obtained. Calculated activation energies for diffusion are higher for wurtzite than sphalerite in the intrinsic regime, but nearly equal in the extrinsic regime where defects are created by impurities. Activation energies for Zn diffusion are within the range of values calculated by various experiments, though the range of these reported values is very large. Activation energies for S diffusion are more difficult to match since experimental data is even more sparse and inconsistent. The overall picture of the analysis suggested that Zn in ZnS diffuses by way of an interstitial mechanism, while S diffuses by a vacancy mechanism or some more complicated mechanism involving more than one defect.

Sulfur self-diffusion is similar in all II-VI compounds [22]. Normally, the dominant self-diffusion defect is an electrically neutral chalcogen interstitial (S i x

), but for metal-rich materials, sulfur self-diffusion is thought to occur via a vacancy or vacancyinterstitial mechanism [239]. This results in higher sulfur self-diffusion than normally expected, probably due to doubly ionized sulfur vacancies (V

S

••

) [22]. Nonetheless, the sulfur diffusion coefficient is generally smaller than the zinc self-diffusion coefficient in

ZnS. Diffusivity of sulfur is reportedly increased with increasing sulfur pressure.

Zinc self-diffusion increases with Zn partial pressure. The diffusion coefficient of zinc in ZnS at 1025 °C increases as

p

3/2

in Zn atmospheres between 0.25 and 2 atm [22,

Zn

240]. Diffusion below 1025 °C is believed to be via interstitials but via vacancies at higher temperatures. Vacancy-controlled diffusion may also be preferred under extreme

312 chalcogen pressures or certain doping conditions, though the usual mechanism is interstitial [239].

ZnS saturated with excess Zn is a p-type conductor, and monitoring the electrical conductivity is one way to determine diffusion parameters. However, it has been found that no measurable changes in conductivity are seen with out-diffusion of excess zinc, presumably due to the fact that only a small part of the excess zinc is electrically active

[241]. This is in marked contrast to experiments involving in-diffusion of Zn, where changes in conductivity are measurable.

No published studies were found which accounted for ambipolar diffusion in ZnS, which would help to explain any differential diffusion due to cations versus anions [242].

Oxygen diffuses in ZnS via sulfur vacancies and accumulates at stacking faults

(twins) if it is not able to leave the crystal [15]. Thus oxygen concentrations at stacking faults are always higher relative to the matrix. Oxygen-rich and sulfur-deficient stacking faults have been shown with a special TEM-EDS technique in natural sphalerite crystals

[42]. The diffusion coefficient of oxygen in ZnS is thought to be one or two orders of magnitude smaller than the diffusion coefficient of interstitial zinc in ZnS [55].

Self diffusion (isoconcentration diffusion) is related to interdiffusion (chemical diffusion) through the Darken equation, depending also on the mole fractions of the diffusing species and the activity of the metal atom [22]. This self-diffusion coefficient is really the sum of the individual diffusion coefficients of each charge state of the defect times its concentration, though in real systems typically one species and hence one term dominates. The self-diffusion coefficient is also related to the intrinsic diffusion

313 coefficient, which describes the diffusion of parts of a binary alloy relative to the lattice planes [243]. Finally, it is important to consider the difference between atomic disorder and electronic disorder. Self-diffusion behavior assessed by high temperature electrical conductivity methods (i.e. Hall coefficient) generally do not give the same results as defect models or as self-diffusion coefficients measured by radioactive tracer atoms [22].

Lott [244] has described chemical self-diffusion for Zn in ZnS. Chemical-self diffusion (D(Δ)) is distinguished from self-diffusion (D* or D e

) as representing the change in the level of stoichiometry (Δ) in a crystal, or the level of intermixing of the two atom types. At high zinc partial pressures, the high temperature electric conductivity is a function of stoichiometry. Chemical self-diffusion coefficients were determined for

Bridgman grown single crystal ZnS between 750 and 850 °C and between 1050 and 1150

°C. In the intermediate region, two different mechanisms seem to be operative, one between 850 and 950 °C and the other between 950 and 1050 °C. Earlier self-diffusion experiments, which give contradictory results, are explained as due to the change of mechanism around the phase transformation temperature (~1020 °C). Metal interstitials are more mobile than vacancies, and doubly ionized zinc interstitials (Zn i

••

) are assumed to be the dominant charge carrier, as Frenkel disorder dominates in the cubic phase under zinc saturation[245].

In the following data compilation (Table 11-3), the diffusion coefficient can be calculated by:

D

=

D

0 exp

Q

RT

314 where D is the diffusion coefficient in cm

2

/s, D

0

is the prefactor in cm

2

/s, Q is the activation energy in kJ/mol, R is the gas constant (=8.314 kJ/mol), and T is the temperature in Kelvin.

Table 11-3: Diffusion data for ZnS

Note that 1 kJ/mol = 0.0104 eV= 0.239 kcal/mol, and Boltzmann constant k

J/K = 8.62x10

-5

B

= 1.38x10

-23

eV/K. All diffusion listed is in ZnS unless otherwise noted. Diffusion data for Al, Ga, In, Cl, and Mn dopants in ZnS is reviewed in Dutt [240] and for Al, Ga, and In in Lott [246].

D

0

(cm

Zn 3 x 10

2

-4

1 x 10

4

1 x 10

16

/s) Q (eV)

1.5

3.26

6.5

3.2 x 10

7

4.2

T (K)

1198-1213

1213-1303

1303-1348

1198-1348

4.5 2.7

P (atm) Notes

Radiotracer hexagonal single crystals

Same as above, data described as single equation method; hexagonal single XL

Reference

Secco in

Sharma

(compilation)

[239]

Stevenson

(compilation)

[22]

Stevenson

(compilation)

[22]

Bansagi

[108]

0.1 2.9 ± 0.2 Preannealed in vacuum,

Kinetics of transformation; single XL, 20% sphalerite,

Zn: 0.1 atm, or S atm

2

:0.1

80% wurtzite

10

-7

0.75±0.15 Zn Supports idea of interstitial mechanism of diffusion;

4.5 x 10

-3

0.69 vacancy mechanism would require Q=6.6 eV at high temperature

1023-1123 Zn: 0.1 atm Chemical self-diffusion

D(Δ); crystals sphalerite at this temperature

1.2 x 10

-4

0.43 1323-1423 Zn: 0.1 atm Chemical self-diffusion

D(Δ); crystals wurtzite at this temperature

1.2 x 10

-2

1.39 Interdiffusion coefficient, rate of film growth using zinc single crystal and zinc liquid

Morozova

[15]

Lott [244]

Lott [244]

Stevenson

(compilation)

[22]

315

Table 11-3 (continued)

D

0

(cm

2

/s)

S 8 x 10

-5

Q (eV) T (K) P (atm) Notes

2.2 1013-1373 : 0.5 atm Radiotracer method; diffusion along (111) in sphalerite XL; increasing D

2.16 x 10

4

3.15 with increasing P

S2

773-1073 Ar: 3 atm Radiotracer method; in sphalerite XL known to

Reference

Gobrecht

[247]

Williams

[248]

Cu 4.3 x 10

-4

Ag

7 x 10

5

3.4 973-1164

9.75 x 10

3

2.6 x 10

-

-3

2 contain faults and polytypes

: 0.5 atm Radiotracer method; hexagonal single crystals preannealed in saturated Zn

0.645 523-1473

2.08 673-1073

0.79 743-1023

Electroluminescence measurements; ionization

Dutt

(compilation)

[240]

Sharma

(compilation)

[239] energy is 1.2 eV from top of valence band

Luminescence measurements

In CdS, Ag diffuses interstitially at dilute concentrations but complexes of >1Ag atom at high concentrations; ionization energy 0.55 eV from top of valence band

Dutt

(compilation)

[240]

Radiotracer Dutt

(compilation)

[240]

Sharma

(compilation)

[239]

Co Solubility is 2.4 x 10

19

/cm3 at 800 °C

Sharma

(compilation)

[239]

316

12 APPENDIX D: Engineering Properties and Miscellaneous Data

12.1 Elastic Properties

Calculations of the elastic properties of polycrystalline ZnS have been conducted based on the single crystal tensor properties of cubic ZnS [152]. This study showed that the calculated values for Young’s modulus and Poisson’s ratio of polycrystalline ZnS agree with the measured values for HIPd CVD ZnS, but are about 17% higher than measured values for as-deposited CVD ZnS. The differences are explained as being due to the presence of hexagonal phase in CVD ZnS and the considerable stacking disorder.

The variational principles of Hashin and Shtrickman [249], were used in this analysis, which give much better bounds than Reuss and Voigt methods for calculating bulk and shear moduli. For example, the calculated Voigt shear modulus for ZnS is 19% higher than the Reuss shear modulus, while the Hashin shear modulus is only 3% larger than the

Shtrickman shear modulus. There is some indication of anisotropy in the elastic constants of CVD ZnS, particularly the Poisson’s ratio, due to the columnar preferential growth perpendicular to the mandrel [250]. A comparison of these values is in Table

12-1.

The increase in Young’s modulus after HIPing has recently been proposed as being due to a decrease in dangling bonds in nonequilibrium grain boundaries [121, 124].

These authors corroborate this hypothesis with low frequency dielectric constant measurements and electron paramagnetic resonance. Note, however, that these dielectric results have not been reproducible by US researchers. Elastic energy losses at the 1 – 2 kHz range are said to be due to the excitation of dislocations in grain boundaries.

317

Table 12-1: Published values for Young’s modulus (E) and Poisson’s ratio (ν) for CVD ZnS and HIP’d CVD ZnS. Static tests involve strain gauges on the tensile side of a flexural bend test. Dynamic methods are those such as ASTM E1259 which involve sound wave propagation with non-contact transducers.

CVD ZnS HIP’d CVD ZnS Manufact. Orientation (CVD) Method Ref

E=10.7±0.5 Msi

(74.1±3.6 GPa)

E=10.8±0.5 Msi

(74.5±3.5 GPa)

E=12.7±0.1 Msi

(87.6±0.7 GPa)

E=12.4 Msi (85.5 GPa) - Shtrikman limit

E=12.7 Msi (87.7 GPa) - Hashin limit

E=11.69±0.26 Msi

(80.6±1.8 GPa)

E= 11.3 – 11.7 Msi

(77.9 – 80.9 GPa) independent orientation

E=12.74±0.06 Msi

(87.8±0.4 GPa)

E= 12.5 – 13.3 Msi

(86.2 – 91.5 GPa)

Single XL tensor data from [251]

Rohm & Haas

IKhVV RAN,

Russia

Assumed isotropic Calculated Klein [152]

Not indicated

Not indicated

Dynamic Henneman

Not indicated

[126] and Pers.

Comm..

Shchurov [121]

ν=0.30±0.003 N/A normal to growth direction

ν=0.35±0.017 N/A to growth direction

ν=0.29±0.1

ν = 0.318 - Shtrikman limit

ν = 0.314 - Hashin limit

ν=0.30±0.008 ν=0.33±0.007

Single XL tensor Assumed isotropic Calculated Klein [152] data from [251]

Rohm & Haas Not indicated Dynamic Henneman

[126] and Pers.

Comm..

12.2 Hardness

Hardness relates to the resistance of a material to indenter penetration. Minimum

Vickers’ hardness in CVD ZnS for 10 Newton loads was ~1.6 GPa for 15 - 25 μm average grain size material [252]. Hardness increased with decreasing grain size, grain following a Petch mechanism due to dislocation pile-ups at grain boundaries, reaching a maximum value of 2.2 GPa for 2.5 μm grain sizes [252] and 4.0 GPa for 0.5 μm grain sizes [117]. Hardness measurements for grain sizes > 17 μm varied widely due to

318 anisotropic hardness of individual grains. Though hardness appeared to increase with increasing grain size for 10 N loads, measurements using 100 N loads showed more constant hardness at large sizes, ~1.4 GPa for grain sizes > 30 μm, showing that this value probably represents the hardness of single crystals (grains). Standard commercial

CVD ZnS at 4 – 8 μm has a Vickers’ hardness around 2.0 GPa for both 10 N and 100 N loads. Hardness of HIP’d material is indistinguishable from hardness of unHIP’d material at the same grain size, about 1.8 GPa for 150 μm grain sizes in the recrystallized

HIP’d material [123].

12.3 Toughness

Toughness was measured for CVD ZnS by fracturing notched rings with hydrostatic expansion [253]. Samples had average grain diameters of 8 μm and aspect ratios of about 8 for the columnar structure. Expansion of the cylinders was in the direction perpendicular to the growth direction (column direction). Measured toughness was 0.75 ± 0.01 MNm

-1.5

and crack propagation was by mixed intergranular and transgranular mode. Toughness measured from double-cantilever beam experiments was determined to be 0.67 ± 0.04 and 0.69 ± 0.07 MNm

-1.5

for two different lots of CVD ZnS, with cracks propagating parallel to the growth direction [250].

Indentation fracture toughness increased with HIPing, however, for comparable grain sizes. Toughness values for CVD ZnS peak at 0.8 MNm

-1.5

for 8 μm grain size material [252]. Materials with 2.5 μm grain size exhibit toughness of 0.6 MNm

-1.5

, while large grain size material (500 μm) exhibits toughness of 0.3 MNm

-1.5

[252]. Large grain size toughness values represent single crystals, and maximum toughness corresponds to

319 an indenter size effect. No nanoindentation studies have been published for CVD ZnS.

HIPing increases the toughness in CVD ZnS for comparable grains size material [123].

Toughness for HIP’d ZnS was 0.8 to 1.2 MNm

-1.5

for 100 N loads compared to about 0.5

MNm

-1.5

for comparable grain sizes (~100 μm diameter) for CVD ZnS. The recrystallization of HIP’d ZnS results in more pristine grain boundaries and a denser material which tends to improve material toughness [128].

12.4 Strength and Weibull Analysis

12.4.1 Fracture testing of ceramics by ring on ring testing

In looking up strength values in any handbook or paper it is important to realize that those values are characteristic of 1) the particular stress state in that particular test and 2) the specific material being tested, including the surface and subsurface state which typically control the fracture behavior of a ceramic.

There are various standard tests which stress the ceramic in different ways. For the purposes of testing infrared window materials, usually the biaxial flexure test is preferred. In some cases, as with anisotropic single crystals such as sapphire and single crystal MgF

2

or CaF

2

, four-point bend testing of various orientations may be preferred.

Table 12-2 is a list of ASTM standards relating to strength testing and data analysis of fracture data.

320

Table 12-2: ASTM standards relating to ceramic strength testing

Number Name

ASTM F394 Biaxial flexure strength (modulus of rupture) of ceramic substrates

ASTM C1499 Monotonic equibiaxial flexural strength of advanced ceramics at ambient temperature

ASTM C1161 Flexural strength of advanced ceramics at ambient temperature

ASTM C1211 Flexural strength of advanced ceramics at elevated temperature

ASTM C1239 Reporting uniaxial strength data and estimating Weibull distribution parameters for advanced ceramics

Comments

Older standard, different test fixture; not recommended

Load ring and support ring; very detailed requirements for sample size, loading rate, maximum deflection

3 point bend and 4 point bend

3 point bend and 4 point bend

Great detail on maximum likelihood method; unbiasing factors for small sample sizes; multiple flaw populations, etc.

The ring-on-ring test is described best in ASTM C1499 [137]. In this test the failure stress is calculated from the load at breaking, various material parameters, and geometric test parameters as (ASTM C1499, eq. 7):

σ

f

=

3

F

2

π

h

2

( 1

ν

)

D

S

2

2

D

2

D

L

2

+

( 1

+

ν

) ln

D

S

D

L

⎟ where σ f

is the fracture stress, F is the load at fracture, h is the thickness of the sample, ν is Poisson’s ratio for the ceramic, D

S

is the diameter of the support ring, D

L

is the diameter of the load ring, and D is the diameter of the sample.

12.4.2 Weibull Analysis procedures

In the most basic sense, a Weibull distribution when it relates to ceramic strengths is a probability of failure given a particular stress. The distribution itself has a particular

“s-like” shape to it, with few samples breaking at very low or very high strengths. The

Weibull modulus “m” is sometimes called the shape parameter and denotes the spread of fracture data; in other words, a large m is a tight distribution and a small m is widely spread distribution. Ceramic materials typically exhibit values of m between 2 and 20, while most infrared window materials are between 5 and 10.

321

The scale parameter (called variously the characteristic strength (which according to ASTM includes effects of the specimen), or the scale factor (which according to ASTM takes the Weibull modulus and specimen geometry out of the picture, but which ends up with really funny units which I like to avoid in favor of normalizing everything to a 1cm

2

stressed area) gives the value where 63.21% of the points lie below this value of the stress.

There are two main procedures used for calculating Weibull parameters. The simplest procedure which is introduced in most texts is the simple rank method (SRM)

(also known as the linear regression method) which involves a ranking and linear fit to find the Weibull modulus. Alternatively, there is a more sophisticated statistical method known as the maximum likelihood estimate method (MLE) which is espoused by

ASTM, and the details are in ASTM C1239[254]. In general this method allows for the removal of tests which do not conform to the expected flaw population (i.e. a volume flaw failure when a surface flaw failure is the most common) but retaining them in the overall statistics since they might have broken at a different stress had they fractured from the surface flaw population. This is important when generating statistics where two flaw populations (e.g. surface and bulk pore) are involved in fracture.

In some cases, it is desirable to do an initial estimate of the Weibull modulus and scale parameter using the SRM and use the beginning values to quickly converge on the

“real” parameters using the MLE method [255]. Numbers achieved using simple rank method and maximum likelihood are slightly different, and there is no clear trend which

322 is larger or smaller. Both methods are essentially interchangeable except the MLE method is slightly more rigorous statistically (but is also more difficult to calculate).

The ASTM C1239 [254] also provides details and tables of factors to apply for

“unbiasing” the shape and scale parameters to account for the number of specimens tested. Also confidence bounds can be established to get error bars on the values.

12.4.3 Definitions of strength

Probability of failure calculations using Weibull statistics show up in a dizzying number of forms, making it difficult to compare across various published datasets. The relationships between them can be calculated with some effort (e.g. McCloy in [256]).

See Table 12-3 below.

A lot of different terminologies and notations are used, and it becomes confusing if one does not keep close track of units and test setups. Many authors are sloppy in this regard and terminology is not yet standardized. According to ASTM, “characteristic strength” (σ

θ

) is the strength characteristic to a particular test, in that it has area effects embedded in it. In some papers, “characteristic strength” (σ c

) refers to the strength normalized to 1 cm

2 stressed area multiplied by the gamma function (see

Table 12-4 below).

In this paper, “unit strength” will be used to refer to the strength normalized to 1 cm

2

stressed area. Also, in general terms, the 63% point of the Weibull distribution is called the “scale parameter” and is usually designated σ

0

[257].

However, “scale factor”

0

) has a very specific meaning in ASTM and refers to the value of the 63% point

323 convolved with the area. It is very important to check exactly how the analysis was done!

Table 12-3: Various ways of calculating the probability of failure

Cumulative distribution function

Probability of failure

Notes Reference

P f

=

1

− exp

⎜⎜

σ

σ

θ

⎟⎟

m

σ

θ must have units of stress

ASTM 1239

[254]

P f

P f

P f

=

=

=

1

1

1

− exp

⎪⎩

− exp exp

⎜⎜

A

A

0

A

⎜⎜

⎜⎜

σ

⎟⎟

⎜⎜

σ

0

σ

⎟⎟

σ

⎟⎟

θ

m

⎟⎟

m

A

σ

σ

0

dA m

⎪⎭

σ

θ must have units of stress; here A

0

is the effective area for which obtained, and A is the area of interest for determining probability of failure

σ

0 must have units of stress*length^(2/m)

σ

0 must have units of stress*length^(2/m); usually A is the area used to calculate

σ

0

σ

θ

was

Harris [125]

Harris et al.

[258]

P f

=

1

− exp

⎪⎩

⎜⎜

A

A

0

⎟⎟

Γ

1

1

m

m

⎜⎜

⎛ σ

σ

c

⎟⎟

m

⎪⎭

σ

is the unit area stress scaling factor

c

with A

σ

c

=

0

Γ

being the unit area. Therefore

1

1

m

σ

θ

when A

0

is the same unit area for both cases

Klein and

Miller [255]

Table 12-4: What is meant by “strength”

Klein and Miller [255] ASTM C1239

[254]

σ

Effective strength

Units Area dependent?

σ

Mean strength Pa Yes

σ

N

Nominal strength

N/A

σ c

Characteristic strength

σ

θ

Characteristic strength

σ

0

Scale factor

N/A

Pa

Pa*meters

(2/m)

, where m is the

Weibull modulus

Pa

Yes

No

No

In other texts, often represented by σ

Awkward due to

0 units; not used consistently

Area scaled to unit area; expression includes the gamma function

324

One of the main problems in doing comparative Weibull analysis (or metaanalysis) on published data is that not all authors present enough information to do the analysis. Most commonly, a mean and standard deviation is all that is supplied. If the

Weibull modulus is also supplied, this is sufficient information to extract the other parameters necessary for a probability of failure analysis for the particular geometry

tested. As shown below, there is a relationship between the mean of the population and the “Weibull scaling factor” (here not properly in the ASTM sense) of the 63% point in the distribution (σ

θ in ASTM [254] and σ

N in Klein and Miller [255]).

σ

=

0

σ

dP d

σ

d

σ

=

σ

θ

Γ

1

1

m

=

(

σ

kA

)

0

1

m

Γ

1

1

m

It is my opinion that this is the best way to analyze other researchers’ data, since oftentimes it is difficult to extract the stressed areas under which others apply the “scale factor” (see more on this below). It turns out that when actual simple rank method

(SRM) Weibull statistics are generated, the value calculated for the “scaling factor” is very close to that obtained from the mean and Weibull modulus (see McCloy in [256]).

Because the raw data is not always available but the mean is almost always reported, the addition of the Weibull modulus allows the scale factor to be calculated.

The problem with taking the scale factors at face value from published literature is that they nearly always have an area effect compounded into them. The true “scale factor” in the ASTM sense accounts for this by multiplying the scaling factor (in units of stress) by the area factor (in units of meters

(2/m)

, where m is the Weibull modulus. A is the area and kA is the “effective area” with k being a dimensionless number which accounts for geometry and stress gradients in the individual test sample. This matter can

325 be confused even more when a paper publishes a “scale factor” without the units to deconvolve the area considered.

1 1

σ

0

=

σ

θ

A m

=

σ

θ

(

kA

)

m

Therefore, it is recommended that the value that is reported be the one with units of stress (i.e.

σ

θ

) in addition to the geometry of the load ring, support ring, etc. The geometry will aid future researchers in calculating the stressed area of the test, since stressed area determination depends on the geometry of the test.

12.4.4 Area effect

When dealing with stressed areas, there is a characteristic relationship between measured stresses, stressed areas, and the Weibull modulus [259]:

σ

σ

θ

θ

1

2

⎜⎜

A e

2

A e

1

1

=

⎟⎟

m

The question becomes, then, what to use for the area in this equation? Two main possibilities have been proposed which have consequences on the relative ranking of datasets and on the scaling in finite element models. One area to consider is simply the load ring area which is

A load

=

π

r load

2

Alternative, ASTM has proposed the use of the “effective tensile stressed area” which for ring-on-ring testing is estimated to be (ASTM C1499, eq. X1.3 [137]).

A e

=

π

2

D

L

2

1

+

44(1

+

ν

) 5

+

3(1

+

m

) 2

+

m D m

S

D

D D

L

2

D

2

(1

ν

(3

) (

+

ν

D

S

D

L

ν

) (1

ν

)

This “effective area” is a close approximation to the result of solving the following integral, with σ

1

being the principal tangential stress, σ

2

the principal radial stress, and

326

σ max

the maximum tensile stress (stress at failure) as defined by σ f

. This accounts for stresses out to the load ring [259, 260]

A e

=

A

σ

σ

1 max

m

+

σ

σ

2 max

m

dA

Oftentimes parameters are missing from published papers that preclude ability to analyze the effective area. If the load ring geometry is known, then a “unit strength” can be calculated where the strength is normalized to a unit area where A load

and A

0

are the same units (i.e. 1cm

2

or 1in

2

).

σ

unit

=

σ

θ

A load

A

0

The method above is also useful for converting the scale parameters of tested samples to stressed areas in finite element models. In other words, one would not want to use the scale parameter directly from a test coupon to apply to a different load case in finite element analysis (FEA). At the very minimum, one should scale the stressed area for the coupon case to the stressed area for the FEA case and thereby get the correct

Weibull scale factor (using the Weibull modulus which does not change with stressed area) for the probability of survival calculations.

This can be understood intuitively by considering that the flaw population of the ceramic which is sampled in a test using a coupon with a very small stressed area (i.e. the inner ring of a ring-on-ring setup) is much smaller than the flaw population sampled by a uniform pressure load on the interior of a hemispherical dome. In the latter case the flaw population over the whole inside of the dome must be considered, and since it is a much

327 larger area than the coupon, one would expect the strength to be lower. This can be accounted for using the area scaling equation above.

The recommended standard test and analysis procedure for ceramic fracture testing is:

1. Use the ASTM stress equations

2. Use the maximum likelihood method (however, analyses have been performed both ways and there is very little difference in most cases).

3. Normalize to a ring-on-ring test of 1 cm

2

stressed area (choose the load ring area or the effective area and state as such, knowing that these give very different results).

4. Unbias the shape and scale parameters on the normalized basis

12.5 Impurity content of CVD ZnS by GDMS

Glow Discharge Mass Spectrometry (GDMS) data on ZnS was compiled from various sources. Data on samples from 1971 are taken from diBenedetto et al. [102].

Data on 2006 sample is from Shiva Technologies (Evans Analytical Group). Data on

2007 sample of Bridgman ZnS was supplied by the manufacturer (MTI crystals). Note that the GDMS value for oxygen is much lower (4 orders of magnitude) than what was determined using Instrumental Gas analysis (IGA). GDMS results for oxygen should only be taken as a “very rough estimate,” as IGA is a much more reliable measure of low oxygen contents [156].

328

329

13 APPENDIX E: Details of XRD analysis

13.1 A Note about method

Strictly speaking, all x-ray intensities should be analyzed using integrated intensity. This was the procedure employed first, using the GRAMS spectroscopic data analysis software tool described in Chapter 5. However, especially for the texture analysis, this procedure proved cumbersome and prone to inaccuracies. First of all, in any integrated peak intensity analysis, a lineshape must be chosen. GRAMS offers various lineshapes, and based on visual inspection of the fitted data, different lineshapes gave better fits depending on the sample. For samples with very small peaks, the fitting algorithms had to be told where peaks were located and when optimization runs were performed, often these small peaks (which could be seen by the human) were blurred out into the background. Also, for the texture analysis where multiple peaks were fit on a peak-by-peak basis, often the best lineshape did not come close to capturing the peak intensity value. This problem was particularly severe for the CVD ZnS samples which were very broadened. A small difference in the integrated area in some cases made large differences in the texture ranking of families of planes, especially for weak peaks.

Therefore it was decided that a more consistent and repeatable method was to choose peak heights. In general this gave good results, and allowed a coherent picture to be presented for the texture analysis. For the hexagonality analysis, the method of peak heights was questioned since the calculation for hexagonality requires the intensity at a specific angle, which would change slightly depending on lattice parameter. For these

330 samples both peak heights and integrated intensities were calculated and were found to be consistent. For small peaks the integrated intensity gave values of zero while the peak height method gave small but nonzero values. The concern about lattice parameter shifting the peak of interest was alleviated when comparing the XRD spectra and calculated hexagonality of samples with the same measured lattice parameter. It was shown that hexagonality as calculated by peak height was not very sensitive to the lattice parameter differences within the range of the materials studied.

13.2 Texture analysis

A texture analysis was performed on the x-ray data from the polycrystalline ZnS samples to see if any information could be gleaned about preferred orientation and possible mechanisms for crystallographic transformation during annealing and hot isostatic pressing. The use of the texture or orientation index has been described for

CVD ZnS [121] and CVD ZnSe [139].

The analysis consists of taking the ratio between the fraction of the intensity of an observed peak to the sum of all the observed peaks and the fraction of the intensity of the peak in an ideally isotropic sample to the sum of all ideal peak intensities.

f hkl

=

txt

I

hkl

I txt hkl

/

iso

I

hkl

I iso hkl

In an isotropic polycrystalline sample f hkl

=1 for all (hkl) planes. The larger the value, the larger the volume fraction of

{hkl } planes orientated in the Bragg diffraction position. The powder pattern for cubic ZnS #05-0566 was used as the “isotropic” sample

331 since it had the data for the (511) peak unlike the diffraction files #77-2100 and #80-

0020. The intensity of each peak in #05-0566 powder pattern was ratioed with the sum of the intensities of all the peaks from 2θ = 20 to 100° to calculate the denominator of f hkl

.

Seventeen samples were analyzed this way. At first the GRAMS spectroscopy software suite was used to calculate the integrated intensities of the following cubic ZnS peaks (111), (200), (220), (311), (222), (400), (311), (420), (422), and (511). Data was taken from the Rigaku diffractometer as described in Chapter 5. There were issues with analyzing the data this way. Firstly, a functional form had to be chosen for fitting, and

Voigt lineshapes were used. The fitting algorithm did not always find a good fit, especially for very small peaks. Also, sometimes the peak intensity was not captured by the resulting fit. Fitting algorithms had trouble with higher angle peaks, since the Cu

K α 2 x-ray wavelength starts to give a lower intensity side peak to the primary Cu

K

α

1 peak for angles greater than about 2θ=45°, which included the (220) and higher index reflections.

When fitting difficulties were found using this method it was decided to use instead the peak intensity value from the raw x-ray data. Since the Rigaku data had very little background to begin with, this method proved reasonable. The peak intensities were then ratioed with the sum of all the intensities then divided by the normalized isotropic numbers shown below. The results were simplified by identifying only the planes, hkl, with the four highest values of f hkl

and assigning the texture by these four planes.

In each sample individually, the f hkl

value for each plane was divided by the total of the f hkl

values for that sample to give the percent of texturing for each family of

332 crystallographic planes. Here the equivalent planes were reduced to their lowest indices and combined where appropriate (i.e. (200) and (400) reflections were combined as

{ 100 }

and (422) is shown as

{ 211 }

, etc). Following is an example of this procedure for calculating the

{111} texturing of Raytheon multispectral ZnS (msZnS). Note that this value shows up as 49% in the table due to different rounding than shown here.

f

111,

msZnS

=

I msZnS

111

I msZnS hkl

⎞ ⎛

/

I

111

I hkl

⎞ ⎛

82,386

/

100

115,314 224

=

1.6

Where

I hkl

=

I

111

+

I

200

+

I

220

+

I

311

+

I

222

+

I

400

+

I

331

+

I

420

+

I

422

+

I

511

%{111}

msZnS

=

f

111,

msZnS

f

+

f

222,

msZnS

=

6.18

=

0.5 50%

Texture index for each reflection by sample

Percent texturing of families of planes by sample

333

It can be seen that a large number of the as-deposited CVD materials are textured very similarly. For example, the order of texturing is the same for Core-mandrel, Coremiddle, RH05, PSZnS, Raytran, and RH04. Only slightly different are RH06 (where

(311) and (400) are switched), II-VI (where (111) is significantly textured), and Coregrowth (where (200) and (400) are switched). EZnS is very similar but has (420) as important, and HIP33none which has the same planes as the first group with (200) and

(400) switched. Thus of the 11 samples without heat treatment, 10 of them have (200) and (400) as highest volume fractions of planes in the Bragg position. The other one,

RH06, is very close. Additionally, one heat treated sample which did not have any metal layer, HIP33none, has similar texture as well. RedZnS is anomalous. It shows very strong (111) and (222) texture, though not as strong as Raytheon’s multispectral ZnS

(msZnS). It is very possible that this is due to the different growth regime (i.e. low temperature, very kinetically controlled) of redZnS where continuous nucleation results in a lack of anomalous grain growth (i.e. “columnar cross-section”) as occurs in most

CVD ZnS.

Heat treated samples, with the exception of HIP33none already mentioned, show very different texturing than the as-deposited samples. HIP1045 is another anomaly, since it was held above the wurtzite transformation temperature and is beginning to convert to wurtzite. It should be noted that none of the wurtzite peaks (weak or strong) were included in this texture analysis. The remaining heat treated samples (msZnS,

HIP15Co, HIP33Ag, and AnRH) all have (422) as an important textured plane. If we take these four samples as steps in the recrystallization process it is instructive. At the

334 fully recrystallized end we have msZnS (HIP’d at 990 °C) with strong (111) and (222) texture, followed by (422) and (220). The importance of (422) and (220) can be appreciated when considering that the preferred slip system is along <110> directions on

(111) planes[261], while partial dislocations have Burgers vector 1/6[112] on (111) planes[34]. HIP15Co (HIP’d at 950 °C) has an intermediate texture of (422) as the most preferred as well as the (331) which is remnant of as-deposited CVD material. Similarly,

HIP33Ag (HIP’d at 750 °C) also has the intermediate (422) texture along with the residual CVD (511) texture. Finally, AnRH (vacuum annealed 850 °C) does not show the (111) and (222) texture characteristic of the HIP plastic deformation, but does show the intermediate (422) and the residual (311).

Another explanation can be offered by comparison to CVD ZnSe. According to

Hartmann et al. [139], CVD ZnSe preferred reflections are (111), (422), and (331). The first two of these are preferred orientations in heat treated ZnS. These authors’ explanation for this combination of directions being included together has to due with their angular deviation from the [111] preferred texture direction for ZnSe. [211] and

[331] directions are at small angles from [111], while [311], [110], and [100] directions are all at larger deviations. In the case of HIP’d ZnS, the planes mostly found together are (111), (222), and (422), similar to the situation described above. For the as-deposited materials, the most common planes are (200), (400), (311), and (511). Both [511] and

[311] have small angular deviations from [100]. In the cases of samples which do not show these preferred groupings, one could argue that the transformation is not totally completed, such as in the annealed sample, and the alternate HIP samples (Ag and Co) as

335 well as the HIP1045 which is beginning a different transformation to the hexagonal phase.

Angles between planes/ directions

From this discussion we can postulate that the isostatic pressure does play a large part in texturing the sample (i.e. producing deformation twins) since the HIP pressure is above the Peierls’ stress to move the dislocations. As pointed out by Shchurov et al.

[121], nucleation and growth rates are highest on “atomically rough” planes which have higher densities of kinks and steps, like { 100 } planes (shown by (200) and (400) reflections in Face-centered cubic materials like ZnS) and (311). These, as has been shown, are the most common textures for the as-deposited ZnS.

A comparison of the calculated { 100 } versus {111} texture is shown graphically in Figure 13-1 below. Note that the HIP samples for the most part cluster to the top left of the figure (large {111}/ { 100 } ratios) while the CVD ZnS cluster to the bottom right and center.

336

Figure 13-1: Comparison of predominant crystallographic texture by sample

13.3 Nonstoichiometry Factor

Calculations were also made for the “nonstoichiometry factors” as described in

Shchurov et al. [121] estimated by I

200

/I

400 and relating to the atomic scattering factors and angular intensity factor. Results here were very different than reported by these authors. Values calculated from raw polycrystalline diffraction data ranged from 1.34 to

4.37 whereas these authors report 0.84 to 1.02 with 1.06 indicating ideally stoichiometric material and deviations from it resulting in smaller numbers. Values calculated from integrated areas of these peaks gave 0.17-3.99 (though most were close to 1). When normalized powder diffraction data was used with integrated intensities, values were calculated to be between 0.61 and 1.11, which is much more in line with that reported in

337 the literature. For reference, the calculated I

200

/I

400 for powder diffraction files are 1.67

(05-0566), 1.56 (77-2100) and 1.6 (80-0020). This author is skeptical about the utility and validity of the “nonstoichiometry factor” as described by the cited reference.

13.4 Disordered Fraction and I

111

/I

200

Ratio

The fraction of disordered material in the closest packing direction was assessed as suggested by Shchurov et al. [121] by diving the integrated intensity of the (111) peak by the integrated intensity of the background around this peak. The larger this number, the less disordered. This analysis was done in GRAMS using integrated intensities, where the two theta data was fitted from 20 to 37.6 degrees. The (200) peak was included in this fitting procedure so the contributions of its tails were not added to the background of the (111). The curve-fitting algorithm found the peaks, then the fit was allowed to iterate until a converged solution was found. Several iterations were made to get a sense of the error in this method. Lorentzian or mixed Gaussian-Lorentzian profiles were chosen depending on which gave a better fit with no baseline function assumed since the instrumental background was small from the Rigaku diffractometer. Sometimes the algorithm had to be told there was a peak at the w(10.0) and w(10.1) position, and in the case of some HIP’d samples even where the (200) peak was located. This was required when peaks were very broad or very small.

The results are shown in the table below. In a number of cases, there was insufficient background to fit any peaks other than the characteristic (111) and (200) peaks. In this case the value of I hkl

/I bkgrnd

was arbitrarily set to 999. It can be quickly seen that these samples correspond to the “heat treated” and “inherently transparent”

338 groups as defined by the powder process analysis (see Chapter 5). The one heat treated sample which did not make it in this group, HIP33none, did not have a metal present during its heat treatment. Otherwise HIP33none had the same processing conditions as

HIP33Ag also shown below. Again it is shown that the metal is required for low temperature conversion of the disordered material.

For disordered packing fraction I

111

/I bkgrnd

, really only two groups can be assessed. (1) those samples with a finite value <3, (2) the HIP 1045 sample which is crystallographically different due to significant wurtzite formation, and (3) the high values which could not measure significant disordered background. Within group (1) the differences are probably negligible

For the relative recrystallization I

111

/I

200

, it can be seen that the as deposited materials all have values below 5 (with the exception of red which has been shown to have more (111) texture). Heat treated samples (ignoring HIP 1045 which is substantially different), have values in excess of 10 and for very recrystallized samples (msZnS) the number is an order of magnitude larger

Also listed is the value of I

111

/I

200

, determined from integrated intensities. This can be seen as a measure of recrystallization from the texture of the as-deposited material to the texture of the fully HIP’d material. By comparing the ranking of these two values it is obvious that elemental ZnS (eZnS) sticks out, having not recrystallized by having

I

111

/I

200

=1.49 in the middle of the range of other as-deposited materials (0.97- 4.64, not including redZnS). RedZnS has a relatively high value of I

111

/I

200

= 9.44 which is similar to HIP1045 and the annealed sample AnRH. HIP’d and fully recrystallized samples have much larger values of I

111

/I

200

from 47.5 (HIP33Ag , 750 °C) to 80.6 (HIP15Co , 950 °C) to 813.7 (msZnS, 990 °C). Note that the higher temperature increases this transformation

(750 vs. 950°C) even when an active metal is not present (cobalt is not active). The highest temperature has the best conversion, presumably due to the mobility of

339 dislocations and ability for lattice diffusion at these temperatures. It can be seen that there is a limit to the increase of temperature, as the HIP1045 shows that higher temperatures with platinum actually reduce the desirable conversion and lead to different transformations toward the hexagonal structure.

It has been observed in the literature that the I

111

/I

200

ratio increases with increasing deposition temperature in the CVD chamber, at least from 560 to 800 °C[262].

Not knowing what the deposition temperatures are for the various starting materials, this is difficult to corroborate. However, it can easily be said that the ratios vary even within a single deposition run, as shown by the increasing ratio from the mandrel (1.08) to the middle (1.25) to the growth side (2.95) of a single core. It may be that these values are not significantly different, since no comparative quantitative data is available. Also, since it is known that redZnS was deposited at a fairly low temperature (~640 °C) and that commercial material is deposited at 670-720 °C, one would expect the ratio to be lower in redZnS which it is not. However, judging the graphs presented by Chang et al.

[262], obvious differences in I

111

/I

200

ratio are not very evident until 550 °C on the low end and 760 °C on the high end, while the temperatures in between show very similar ratios. Therefore, it is concluded that there are substantially more complicated factors influencing the crystallographic texture of the deposited material than simply the deposition temperature. One possible explanation for the enhanced (111) texture in the higher temperature deposits is the onset of boundary layer diffusion control in the deposition process rather than kinetic control, as has been described for CVD ZnSe[139].

However, this still does not explain the redZnS texture which is deposited at low enough

340 temperature to be kinetically controlled. It is likely that some other mechanisms are operable in making redZnS more highly textured on (111), such as its higher oxygen content which leads to stacking fault layers on the (111) plane.

13.5 Crystallite Size and Coherently Diffracting Domains

The size of the hexagonal phase “particles” (if they exist) was estimated using xray data to provide inputs to potential scattering models. The procedure followed was similar to the other GRAMS analyses described previously. In this case one representative sample was selected from each of the three groups in the powder diffraction data (i.e. the heat treated samples with no w(10.0), the transparent CVD samples with a shoulder at w(10.0), and the opaque CVD samples with a peak at w(10.0).

A Voigt lineshape was fit to the sphalerite (111) and to the w(10.0) peak in the case of the opaque samples. The linewidth at full-width half-maximum (FWHM) was determined from the Voigt fit, and designated Δ2θ meas,ZnS

. (Gaussian profiles were also tried but they resulted in particle sizes about half of that assessed by the Voigt fits. This illustrates a weakness in the integrated intensity method since it can be very sensitive to the lineshape chosen.) The linewidth of the Scintag x-ray instrumental function was subtracted from the linewidth of ZnS by using a large grain powdered silicon standard, which gave Δ2θ instr,Si

= 0.181 degrees for the instrument function.

For the opaque samples, the core sample (Rohm & Haas ‘06 A-S) was chosen.

For the transparent sample, elemental ZnS (eZnS) was chosen, and for the heat treated samples, 990 °C Pt HIP multispectral ZnS (msZnS) was chosen. The half-width as determined from the GRAMS peak fitting was corrected for the instrument function in

341 the opaque and transparent samples. The heat treated sample had a narrow enough linewidth that it did not need to be corrected. Corrected linewidth was determined by:

Δ

2

θ

corr

=

(

Δ

2

θ

meas

,

ZnS

2

) (

Δ

2

θ

instr

,

Si

)

2

The Scherrer formula (e.g Cullity[134]) was then used to determine the particle diameter or coherently diffracting (scattering) domain:

d

=

0 .

9

λ

Δ

2

θ

corr

cos

θ where d is the diameter, λ is the x-ray wavelength (1.54439 Å for Cu

), Δ2θ is the corrected linewidth as above, and cos θ refers to the angle at the peak of interest, in this case the (111) where 2θ=28.559° or the w(10.0) where 2θ=26.915°. The results are shown below. Note that in this analysis the disordered background was not fit but just the peak of the w(10.0), since the powder diffraction samples did not show any evidence of w(10.1).

It can be seen that the values of the “particle size” or coherent diffracting domain do not vary much among the samples. This is probably due to the phenomena in crystals formerly known as the “mosaic structure” but now more properly the crystallographically perfect sub-grains or cells separated by “walls” of dislocations[134]. This is the case even within relatively fault-free large crystallites such as occur in msZnS as shown by the

TEM images. The quoted normal size for coherently diffracting domains even in

“perfect” crystals is 100nm [134], which is very close to the values obtained here. The stated values for coherently diffracting domain in the literature are 60-80nm for CVD

ZnS and >100nm for HIP’d ZnS [121], which is well within the errors of this analysis technique.

342

Figure 13-2: Peak broadening fitting for phase size analysis using Scherrer formula

It was concluded that the investigation of lineshape could not really give useful information on the actual size of the hexagonal phase (if it exists as a discrete phase and not just random stacking). The size of the coherently diffracting domain (i.e. 70 – 110 nm) is on the order of the size of the smallest hierarchy of lamella imaged TEM for standard CVD ZnS and elemental ZnS. However, it should be recalled that the multispectral ZnS did not have these fine features, and yet the coherently scattering domain size is about the same. This suggests that this analysis is sampling small blocks of defect free crystal and not real feature sizes in the nanostructure.

13.6 Lattice parameter analysis

Considerable variation is found in lattice constant values in the literature.

Probably many of the samples had unknown amounts of oxygen content, which

343 considerably shrinks the lattice parameter in a roughly linear fashion according to

Vegard’s law, according to Skinner and Barton [24]. However, Chechetkina et al. [263] found a considerable deviation from Vegard’s law in the solubility of oxygen in wurtzite

ZnS.

Impurity additions of Fe, Cd, and Mn, which are commonly present in mineral specimens, result in additional uncertainty in measured lattice constants. A relation for estimating the lattice parameter of ZnS as a function of impurity content of FeS, CdS,

MnS, CoS, ZnSe, and ZnO mole percent is shown in Barton [14]. A relation for estimating the lattice constant of ZnS based only on ZnO is shown in Morozova [15].

The lattice parameters of various natural sphalerites are reviewed in Smith [7].

It should be recalled that Skinner and Barton [24] state the lattice parameter of cubic ZnS free of oxygen to be 5.4093 Å. Various other values are presented in the literature (see Table 13-1), which can probably be accounted for by undetermined concentrations of oxygen and other impurities. Barton and [14] list the lattice parameter as a function of mole percent impurity concentration as:

a

=

5 .

4093

+

0 .

000700 * [

0 .

000456 *

CoS

]

+

[

FeS

]

0 .

002592

+

* [

0 .

00424

ZnSe

]

* [

CdS

]

+

0 .

00202 * [

MnS

]

0 .

003 * [

ZnO

]

Morozova et al. [15] prefer the value of 5.4092 Å for the unaltered lattice parameter, and similarly write:

a

=

5 .

4092

0 .

008584 * [

ZnO

]

344

Table 13-1: Literature ZnS lattice parameters

Sphalerite

α = β = γ = 90º

a (Angstroms)

5.42, 5.406,

5.4109

c (Angstroms) c/a

N/A N/A

Reference, Notes

Sharma

5.41 [113]

Wurtzite

α = β = 90º; γ =

120º calc.)

5.4093 N/A N/A

5.4050 ± 0.0030 [121]

= layer repeat* √2/√3

= 2*.8166 = 1.6332; i.e. c-axis for 4H is

4*1.6332 = 6.53 Å

Skinner [24]

Ideal spherical close-packing of anions [200]

1.6379, 1.638, 1.6378 Sharma [6]

3.8226 6.2605

(ab initio calc.)

6.05 (ab initio calc.)

Lattice parameter measurements were taken on powdered samples, using a NIST traceable alumina standard internal to the Rigaku diffractometer. The alumina (0.2.10) plane reflection, expected at 2θ = 88.995°, was measured at 89.02° and the deviation used as an angular correction. Diffracted intensity was measured from 87.7 to 89.3° for ZnS at

0.02° intervals with 5 second integration time. The sphalerite (422) reflection, measured at 2θ = 88.4 – 88.6°, was corrected by the standard. This angle was used to calculate the interplanar spacing (d hkl

) using Bragg’s law, and the lattice parameter using the definition for cubic crystals [134]:

d hkl

=

λ

2sin

θ

hkl

=

a

0

h

2

+

k

2

+

l

2

=

a

0

24

The estimated accuracy of lattice constant measurements is better than ±0.001 Å and may be as good as ±0.0003 Å. The data is shown in Table 13-2 below.

345

Note that in two cases (Chinese ZnS and IRTRAN ZnS), the lattice parameter is smaller than that of the “oxygen-free” number quoted above which is 5.409 Å (to four significant figures), suggesting some lattice shrinkage due to oxygen substitution and/or sulfur vacancies. All the other samples show lattice parameter in excess of the oxygenfree lattice parameter.

Table 13-2: Measured ZnS Lattice Parameters

Sample 2θ meas

Chinese ZnS

2θ adj d hkl a

0

88.6 88.575 1.103143 5.404

88.455 5.410

Rafael msZnS 88.48 88.455 1.10433 5.410

Vitron no Pt HIP 88.48 88.455 1.10433 5.410

Ray msZnS

Rafael CVD

88.435 5.411

88.46 88.435 1.104528 5.411

88.44 88.415 1.104726 5.412

Co33_1a 88.4 5.414

For samples where measured oxygen concentration was available from Interstitial

Gas Analysis (IGA), lattice parameters were calculated based the equations above and compared to the measured data (see Table 13-3). Obviously, the calculated values given that oxygen is known to be present are too low, so some other factor must be increasing the lattice constant. Current and historical results from glow discharge mass spectrometry (GDMS) on similar samples suggest that iron was an impurity in some early runs in 1971, but cadmium, manganese, cobalt, and selenium were not detected. None of these was detected in the most recent GDMS of Rohm and Haas CVD ZnS. Thus the

346 impurities indicated in available formulas for ZnS lattice parameter probably do not account for the defects responsible for measured deviations in lattice parameter in CVD

ZnS. Point defects and substitutional impurities like zinc hydrides, zinc antisites, interstitial atoms (likely zinc) increase the lattice parameter are all candidates. These are

not included in the above models, however.

Table 13-3: Lattice parameters correlated to oxygen concentration only

Raytheon msZnS

Elemental ZnS

Bridgman ZnS

Red ZnS

XRD Lattice

Parameter a

0

(Å) IGA mol% O

5.410 0.3 – 0.4

5.411

5.411

5.414

0.22 – 0.4

0.33 – 0.43

0.3 – 0.6

Calculated after

Morozova (Å)

5.407 – 5.406

5.407 – 5.406

5.406

5.407 – 5.404

Calculated after

Skinner (Å)

5.408

5.409 – 5.408

5.408

5.408

Existing literature models underpredict the lattice parameter of real ZnS materials.

Therefore, a substantial concentration of point defects which increase the lattice parameter must be present. Since the Bragg refraction is measuring coherently scattering domains, only point defects and not grain boundaries should be considered. Since the hydrogen content is known to be at least two orders of magnitude lower than the oxygen content, the zinc hydrides in sulfur sites must not contribute significantly to the lattice parameter. The main point defects in these materials, therefore, must be substitutional oxygen (and possibly oxygen vacancies) along with zinc interstitials to compensate the volume change and grow the lattice parameter. If interstitials are charged, charge compensation could be accomplished with some charged zinc vacancies.

It is also possible that the degree of hexagonality could influence the measured lattice parameter in CVD ZnS. Currently this possibility has not been considered. The

347 analysis is complicated by the possibility of orientation of the hexagonal phase, since the lattice parameter of the a and c axis in wurtzite differ by about a factor of two.

Ab initio calculations have allowed estimation of the effects of individual point defects on the lattice parameter of ZnS (described in upcoming article to be submitted to

Journal of Applied Physics). Given the measured oxygen concentrations of 0.3 to 0.6 atomic percent in these samples, and the calculated lattice parameter changes for point defects, it is estimated that the samples described above contain between 0.4 and 1.0 atomic percent interstitial zinc, with red ZnS containing the most interstitial zinc. These estimates agree with the EDS results presented which imply ZnS is zinc rich on the order of a few percent at most.

13.7 Hexagonal fraction

Bansagi et al. [108] and Xue and Raj [109] have suggested a simple measure of the degree of hexagonality in polycrystalline ZnS powders and hot-pressed compacts.

Essentially the procedure involves taking ratios of characteristic peaks in the wurtzite and sphalerite XRD spectra. The relative molar fraction (γ) of the hexagonal (α) phase to the cubic (β) phase is described by

γ

=

I

{

α

(

28 .

53

I

002

{

α

( 002 )

+

28 .

)

+

53

β

( 111 )

β

}

( 111 )

}

+

I

α

( 100 )

26 .

91 where the subscripts are 2θ angles and the superscripts are the characteristic planes, since the α(20.0) and the β(111) overlap at 2θ=28.53°. From this it can be shown [177] that the mole fraction of the hexagonal phase is given by

α

=

1 .

84 ( 1

γ

)

348 where the constant is determined for the value of γ when there is no cubic phase (β=0) from standard powder diffraction patterns.

This procedure was used to quantitatively analyze a new set of powder data collected on the Rigaku. The original intent was to compare the calculated values to the rough “categories” described in Chapter 6 which were determined based on visual comparison of powder patterns from the Scintag. The peak height intensity values were used rather than integrated intensities for simplicity and avoidance of fitting issues. The powder XRD data from the Scintag diffractometer proved unusable for the quantitative analysis since x-ray backgrounds were so high and values obtained for percentage of hexagonal phase ranged from 24 to 43% which cannot be correct. The Rigaku has a diffracted beam monochrometer, so backgrounds are low enough to enable use of peak heights.

The polycrystalline diffraction data from the Rigaku were examined, using 2θ angles of 28.55 and 26.9° as these were the closest datapoints. Calculated values of hexagonality (α) were very reasonable and consistent with the powder data visual inspection with only a couple minor exceptions. Not all the same samples were tested in the Scintag and the Rigaku, so some additional information was gained by analyzing the

Rigaku data this way. The data is listed in the following table.

Hexagonality in polycrystalline samples as determined from comparing sphalerite and wurtzite peaks

349

Most of the as-deposited CVD materials have hexagonal phase content

(determined only from the w(10.0) peak) of 2 to 5+ %. In the visual analysis of the

Scintag powder data, no tested heat treated samples fell into the “opaque” category with a strong w(10.0) peak. In the present analysis, two samples that were not tested on the

Scintag did show this type of behavior. HIP1045 sample also has hexagonality in the high range, but this sample was quite different due to its being processed above the transformation temperature, so was expected to have more hexagonal phase. The

HIP33none sample was a low temperature (750 °C), no metal experiment which came out visibly scattering, likely due to the incomplete transformation of hexagonal phase as seen here.

Two samples which were expected to show higher degrees of hexagonality

(RH06-Core middle and II-VI) showed hexagonality below 2%, which is in the range of the low scattering materials. However, on inspection of the polycrystalline XRD pattern of these two samples, much of the hexagonality is reflected more in the w(10.1) peak which is not accounted for in this analysis. Some of the other samples have hexagonality in the w(10.1) region (e.g. PS and RH04) but still registered high levels of hexagonality by this calculation, presumably due to the height of the s(111)/w(00.2) peak. It should be recalled that when powder diffraction was done on these samples, all the hexagonality if present was strongly shown in the w(10.0) peak and only a small amount of broadening showed up in the w(10.1) position.

The “transparent” sample category from the x-ray powder analysis (see Chapter

6) consisted of Raytran, eZnS, and redZnS. In the present hexagonality analysis, Raytran

350 had to be classed in the opaque category with hexagonality >2%. Again this could be due to the variability of preferred orientation of hexagonality or to the variability of the material itself, since it is known to be in some cases highly scattering and in other cases virtually indistinguishable from elemental ZnS. The “transparent” category here had hexagonality values between 0.5 and 2%. The arbitrary setting of 0.5% as being included in the “transparent” category left the high temperature annealed (850 °C) sample in this category, but it could have just as easily been put in the “heat treated category” by arbitrarily choosing the hexagonality class boundary at 0.6%.

Finally, the “heat treated” category again includes multispectral ZnS, but here also samples untested by the Scintag, including HIP33-Ag (same conditions as HIP33none, 750 °C, except with Ag foil) and HIP15-Co (950 °C). The low values of hexagonality here can be explained by a full recrystallization of the ZnS by high temperature and metal promoter (msZnS – Pt foil), high temperature only (HIP15-Co, since Co foil has been shown to be inactive), and by metal promoter only (HIP33-Ag, since temperature was too low to cause full recrystallization without recrystallization promoter).

This exercise has been useful to put more quantitative estimates on the hexagonality of various grades of ZnS, but should be seen as an indication only. For comparison, commercial hot-pressed ZnS sold by Kodak was advertised to contain about

5% hexagonal phase [176], very similar to what we see here in the CVD ZnS before any treatments. Lastly, it should be remarked that we are not dealing with specific “grains”

351 of hexagonal phase, per se, but more likely have broadly disordered areas such as at the boundaries of domains and grains, which manifest their hexagonal nature in XRD.

To make a final assessment of these issues, one more set of data was taken for hexagonality assessment. Some sample types were deliberately chosen from the previous analysis and others were chosen as representing possible borderline cases or as-yetuntested cases. Samples were ground and put in vacuum grease on glass slides and the diffraction was recorded at 2θ angles of 25 to 35° only (the region around the stacking direction) at intervals of 0.02° for 5 second integration times.

Both intensities at particular angles (26.91 and 28.53° as above, calculated as average of bounding datapoints) and integrated intensities were calculated, since there was concern that the lattice parameter would influence the position of the peak. It was found that the integrated intensity method could not detect hexagonality below 2%, but that in general calculated values bounded or were slightly larger than those calculated by raw intensity data. It should be noted that in the integrated intensity calculation, no baselining was done since the background of the raw data was very low. Ranges are given for the hexagonality calculated by integrated intensity since the area under the peaks was sensitive to the fitting parameters.

Example XRD spectra for calculated hexagonality values of ~1%, ~2%, ~3%, and

~10% are shown in

Figure 13-3

below. All powder spectra are normalized to their largest peak, the s(111) reflection. In the inset is a blow-up of the w(10.0) region. Also listed are the calculated values for hexagonality by raw intensity and by integrated intensity.

352

Figure 13-3: Example of hexagonality determination and corresponding XRD

In Figure 13-4 the powder pattern of the single crystal Bridgman ZnS is shown. It is evident that there is much more structure in the region around the stacking direction, indicating polytype structures. The exact polytype was not identified, but it appears to be a long-period rhombohedral with possible contributions from hexagonal and major cubic regions.

353

Figure 13-4: Polytype stacking in single crystal ZnS

13.8 Summary of XRD quantitative analysis

The results of the various hexagonality and texture analyses are now presented in a comparative and coherent form. A summary of this data is shown in Table 13-4 and

Table 13-5 below. In a few key cases, hexagonality was determined from polycrystalline and powder peak height calculations as well as powder integrated intensity and found to be relatively consistent. Additionally, texture percentages (determined from peak intensities) were compared with the I

111

/I

200

ratios (determined from integrated intensities). It can be seen that the texture analysis and I

111

/I

200

ratio give similar data on the relative recrystallization. Together these tell a consistent story and suggest that there are really four categories of CVD ZnS with a few outlier samples.

354

Group 1 consisted of all high temperature HIP’d samples (HIP at ≥ 900 °C with no metal, Pt, Ag, or Co) and low temperature HIP’d sample (750 °C) with Pt and Ag.

Samples in this category were HIP57Ag, HIP33Ag, HIP60Pt, Ray msZnS, HIP55a-none,

Vitron no Pt HIP, and HIP15Co. These parts have <1% hexagonality (defined here and in the following paragraphs as the raw intensity determined hexagonality) and have large ratios of {111}to {100}texture. These samples all had very good visible transmission and were fully recrystallized.

Group 1A consisted of high temperature annealed samples (850 °C) and low temperature HIP’d samples (750 °C) with no metal present. Samples here included

HIP57none, HIP33none, Co33_1a, and Co42_2a (AnRH). Hexagonality was slightly higher, at 1-2%, and texture was intermediate and variable, with {211} predominant texture or with {100}texture. These samples had good visible transmission but not quite as good as group 1, and had recrystallized at least partially.

Group 2 consisted of as-deposited CVD samples with high intrinsic visible transmission. Samples in this category were red ZnS (600 °C deposition temperature,

H

2

S process), elemental ZnS (H

2 process) and Chinese ZnS (H

2 process). Hexagonality was 2 – 5%, and texturing generally favored the {100} planes, but red ZnS had an unusually high component of {111}texture. These samples had moderately good visible transmission, but were colored. Since they had not been heat treated, there was no recrystallization.

Group 3 consisted of as-deposited CVD samples with poor intrinsic visible transmission. Samples in this category were all CVD ZnS by the standard H

2

S process

355

(most likely 670 – 730 °C deposition temperature) from suppliers Rohm & Haas (3 different lot years, 3 different core orientations), II-VI, Rafael, Raytheon, and Princeton

Scientific. Hexagonality was 5-10%, and texture was dominated by{100}planes. These samples were all opaque in the visible and were yellow to orange except Rafael CVD which was pale yellow. Again, since they had not been heat treated, they were not recrystallized.

Table 13-4: CVD ZnS groupings by hexagonality

Table 13-5: Hexagonality and texture of all x-rayed samples

This analysis gives added credence to the idea that hexagonality, however it is measured, is directly correlated to visible transmission and crystallographic texture. The actual calculated atomic percent of the hexagonality is probably accurate to within a factor of two, given the stacking irregularity and broadening which is not taken into account when calculating integrated intensity of the w(10.0) peak.

356

14 APPENDIX F: Derivation of the Surface Scattering Equation

Below the relation used in the internal surface scattering model is derived.

TIS

=

scattered

_

power incident

_

power

=

Φ

diffuse

Φ

i

Where TIS is total integrated scatter, Φ diffuse

is the scattered power and Φ i

is the incident power. Note that this is different than the way some define TIS as scattered power divided by specular reflected power [264], the difference being that in the latter case the denominator becomes Q*Φ i where Q is the Mueller matrix (see below). Other sources define TIS as diffuse reflectance divided by the sum of diffuse and specular reflectance, which can be shown to be the same as Φ diffuse

/ Q*Φ i

[265].

Rayleigh-Rice or micro-roughness theory allows calculation of the Mueller matrix

Bidirectional Reflectance Distribution Function (BRDF) [266].

f r

=

4

π

λ

2 cos

θ

i

cos

θ

r

Q

(

θ

i

,

θ

r

,

φ

r

)

S

2

(

f x

,

f y

)

f x

=

(sin

θ

r

cos

φ

r

− sin

θ

i

) /

λ

f

=

y

(sin

θ

r

sin

φ

r

) /

λ

Where f r

is the BRDF, Q is the Mueller matrix (reflection matrix which includes incident to reflected/ scattered polarization terms s→s, s→p, p→p, and p→s), S

2

is the power spectral density (Fourier transform of the autocorrelation function) of a two-dimensional profile propagating in x and y, and angles with subscript i are incident and r are reflected.

The Q matrix has the optical properties of the two media within it.

357

TIS

=

2

πθ

0

∫ ∫ max

θ min

(

d

Φ

/

Φ

i d

Ω

)

R

d

Ω

r

=

f f f

max

∫ ∫ max

f

min min

f r df x df y

Where Ω is the solid angle, and the variable exchange is made to allow integration over the power spectrum [264]. Substituting in for f r

gives

TIS

=

f f

max max

∫ ∫

f

min

f

min

f r df x df y

=

4

π

λ

2 f

f f

max max

∫ ∫

f

min min

[ cos

θ

i

cos

θ

r

Q

(

θ

i

,

θ

r

,

φ

r

)

S

2

(

f x

,

f y

)

]

df x df y

The RMS (root mean squared) roughness σ is related to power spectral density by [266]

σ

2 =

f f f

max

∫ ∫

f

max min min

S

2

(

f x

,

f y

)

df x df y

And the small scatter angle approximation gives θ i

= θ r

so

f

max

f

∫ ∫

max

[ cos

θ

f

min

f

min

i

cos

θ

r

S

2

(

f x

,

f y

)

]

df x df y

=

(

σ cos

θ

i

)

2

Substituting this in to TIS gives

TIS

=

⎛ 4

πσ cos

θ

i

λ

2

Q

(

θ

i

,

θ

r

,

φ

r

)

Now the Mueller matrix Q is the same as the specular reflectance R when polarization is neglected and at normal incidence, giving

Q

=

R

=

⎜⎜

n n

1

1

n

2

+

n

2

⎟⎟

2

=

⎛ Δ

⎜⎜

2

n n av

⎟⎟

2

, since

n av

=

n

1

+

n

2

2

and

Δ

n

=

n

1

n

2

Now substituting Q into the TIS equation,

TIS

=

2

πσ

Δ

n

cos

λ

n av

θ

i

⎟⎟

2

, which for normal incidence where θ i

=0° is just

TIS

=

2

πσ

Δ

n

λ

n av

⎟⎟

2

358

1.

15 REFERENCES

Harris, D.C. "History of the development of hot-pressed and chemical-vapordeposited zinc sulfide and zinc selenide in the United States," in Window and

Dome Technologies and Materials X, SPIE 6545, 654502 (2007).

2. McCloy, J. "International development of chemical vapor deposited zinc sulfide," in Window and Dome Technologies and Materials X, SPIE 6545, 654503 (2007). and Boston: Adam Hilger, Ltd. (1985).

4. Hilton, R., Sr. "Development of Chalcogenide Glasses as Optical Materials for

Infrared Systems," in Windows and Domes Technologies and Materials IX, SPIE

5786

, 258-261 (2005).

5. Mellor, J.W., "Zinc and Cadmium Sulphides," in A Comprehensive Treatise on

Inorganic and Theoretical Chemistry. London: Longmans, Green and Co. Pp.

586-612 (1956).

6. Sharma, R.C. and Y.A. Chang, "The S-Zn (Sulfur-Zinc) System," Journal of

Phase Equilibria 17(3): 261-266 (1996).

7. Smith, F.G., "Structure of zinc sulfide minerals," American Mineralogist 40: 658-

675 (1955).

8. Buck, D.C. and L.W. Strock, "Trimorphism in Zinc Sulfide," American

Mineralogist 40: 192-2000 (1955).

9. Trigunayat, G.C., "A survey of the phenomenon of polytypism in crystals," Solid

State Ionics 48(1-2): 3-70 (1991).

10. Ves, S., et al., "Cubic ZnS under pressure: optical-absorption edge, phase transition, and calculated equation of state," Physical Review B 42(14): 9113-

9118 (1990).

359

11. Jaffe, J.E., R. Pandey, and M.J. Seel, "Ab initio high-pressure structural and electronic properties of ZnS," Physical Review B 47(11): 6299-6303 (1993).

12. Desgreniers, S., L. Beaulieu, and I. Lepage, "Pressure-induced structural changes in ZnS," Physical Review B 61(13): 8726-8733 (2000).

13. Krongauz, V.G., et al., "A Polymorphic Transition in Crystalline ZnS," Inorganic

Materials 13(8): 1122-1124 (1977).

14. Barton, P.B., Jr. and B.J. Skinner, "Sulfide Mineral Stabilities," in Geochemistry

of Hydrothermal Ore Deposits, H.L. Barnes, Editor. New York: Hold, Rinehart and Winston, Inc. Pp. 236-333 (1967).

15. Morozova, N.K., V.A. Kuznetsov, and M.V. Fok, Sul’fid tsinka. Poluchenie i

opticheskie svoistva [Zinc Sulfide: Preparation and Optical Properties]. Moscow:

Nauka (1987).

16. Knitter, S. and M. Binnewies, "Der Chemische Transport von Mischkristallen in den Systemen MnS/ZnS,FeS/ZnS und FeS/MnS," Zeitschrift für anorganische

und allgemeine Chemie 625: 1582-1588 (1999).

17. Balabin, A.I. and R.O. Sack, "Thermodynamics of (Zn,Fe)S sphalerite: A CVM approach with large basis clusters," Mineralogical Magazine 64(5): 923-943

(2000).

18. Dicarlo, J., et al., "Preparation and properties of iron-doped II-VI chalcogenides,"

Journal of Solid State Chemistry 87(2): 443-448 (1990).

19. Aven, M. and J.A. Parodi, "Study of the Crystalline Transformations in ZnS:Cu,

ZnS:Ag, and ZnS:Cu,Al," Journal of the Physics and Chemistry of Solids 13: 56-

64 (1960).

20. Nickerson, J., P. Goldberg, and D.H. Baird, "The Influence of Copper on

Structural Transformations in ZnS:Cu,Cl," Journal of The Electrochemical

Society 110(12): 1228-1230 (1963).

360

21. Scott, S.D. and H.L. Barnes, "Sphalerite-wurtzite equilibria and stoichiometry,"

Geochimica and Cosmochimica Acta 36: 1275-1295 (1972).

22. Stevenson, D.A., "Diffusion in the Chalcogenides of Zn, Cd, and Pb," in Atomic

Diffusion in Semiconductors, D. Shaw, Editor. London: Plenum Press. Pp. 449-

541 (1973).

23. Kroeger, F.A. and J.A.M. Dikhoff, "The Function of Oxygen in Zinc Sulfide

Phosphors," Journal of the Electrochemical Society 99(4): 144-154 (1952).

24. Skinner, B.J. and P.B. Barton, Jr., "The Substitution of Oxygen for Sulfur in

Wurtzite and Sphalerite," American Mineralogist 45: 612-625 (1960).

25. Carter, C.B. and M.G. Norton, Ceramic Materials: Science and Engineering:

Springer (2007).

(1996).

27. Roth, W.L., "Crystallography," in Physics and Chemistry of II-VI Compounds, M.

Aven and J.S. Prener, Editors. Amsterdam: North-Holland. Pp. 117-164 (1967).

28. Yacobi, B.G. and Y. Brada, "The nature of the pyroelectric effect of faulted ZnS,"

Journal of Applied Physics 47(4): 1243-1247 (1976).

29. Mardix, S., "Polytypism: A controlled thermodynamic phenomenon," Physical

Review B 33(12): 8677 (1986).

30. Verma, A.R. and P. Krishna, Polymorphism and polytypism in crystals. New

York: John Wiley & Sons (1966).

31. Chiang, Y.-M., D.P. Birnie, and W.D. Kingery, Physical Ceramics: Principles

for Ceramic Science and Engineering: Wiley (1996).

32. Kingery, W.D., H.K. Bowen, and D.R. Uhlmann, Introduction to Ceramics, 2nd ed: Wiley-Interscience (1976).

361

33. Balluffi, R.W., S.M. Allen, and W.C. Carter, Kinetics of Materials. Hoboken, NJ:

Wiley Interscience (2005).

34. Kulakov, M.P. and S.Z. Shmurak, "Structural changes in ZnS crystals on account of partial dislocation movement," Physica Status Solidi (a) 59: 147-153 (1980).

35. Fleet, M.E., "Structural transformations in natural ZnS," American Mineralogist

62

: 540-546 (1977).

36. Azikuzi, M., "Investigation of phase transition of natural ZnS minerals by high resolution electron microscopy," American Mineralogist 66: 1006-1012 (1981).

37. Wang, Z., et al., "Morphology-tuned wurtzite-type ZnS nanobelts," Nature

Materials 4: 922-927 (2005).

38. Churchman, A.T., G.A. Geach, and J. Winton, "Deformation Twinning in

Materials of the A4 (Diamond) Crystal Structure," Proceedings of the Royal

Society of London. Series A, Mathematical and Physical Sciences 238(1213): 194-

203 (1956).

39. D'Aragona, F.S., P. Delavignette, and S. Amelinckx, "Direct evidence for the mechanism of the phase transition wurtzite - sphalerite," Physica Status Solidi (a)

14

: K115-K118 (1966).

40. Omel’chenko, S.A. and M.F. Bulanyi, "Reversible changes in the structure of zinc sulfide crystals during elastic deformation," Physics of the Solid State 39(7): 1091

(1997).

41. Takeuchi, S., K. Suzuki, and K. Maeda, "Stacking-fault energy of II-VI compounds," Philosophical Magazine A 50(2): 171-178 (1984).

42. Srot, V., et al., "Stacking faults and twin boundaries in sphalerite crystals from the

Trepca mines in Kosovo," American Mineralogist 88: 1809-1816 (2003).

43. Neumark, G.F., "Defects in wide band gap II-VI crystals," Materials Science and

Engineering: R: Reports 21(1): 1-46 (1997).

362

44. Suzuki, S., et al., "Structure and resistivity of heat-treated polycrystalline zinc sulfide prepared by chemical vapor deposition," Journal of Crystal Growth 134:

67-74 (1993).

45. Thomas, A.E., G.J. Russell, and J. Woods, "Preparation of Low Resistivity Zinc

Sulphide," Journal of Crystal Growth 63: 265-275 (1983).

46. Yen, W.M., S. Shionoya, and H. Yamamoto, eds. Phosphor Handbook. 2nd ed.

Laser and Optical Science and Technology Series, ed. M.J. Weber, Boca Raton:

CRC Press. (2007).

47. Wiedemeier, H., "Thermodynamic Estimations of Defect Equilibria and Vacancy

Concentrations in ZnS," Zeitschrift für anorganische und allgemeine Chemie

632

(10-11): 1717-1727 (2006).

48. Morozova, N.K. and O.N. Morozova, "Phase Diagram of Equilibrium Point

Defects and Deviations from Stoichiometry in Zinc Sulfide," Inorganic Materials

17

(8): 989-993 (1981).

49. Schneider, J. and A. Rauber, "Electron Spin Resonance of the F-Centre in ZnS,"

Solid State Communications 5: 779-781 (1967).

50. Gurvich, A.M. and R.V. Katomina, "Influence of the positions of levels of intrinsic defects on the deviations from stoichiometry and the electrical conductivity of zinc and cadmium sulfides," Soviet Physics - Semiconductors

5

(7): 1189-1195 (1971).

51. Matsuura, K., I. Tsurumi, and F. Takeda, "Optical absorption and lattice defects induced by neutron irradiation and heat treatment in ZnS crystals," physica status

solidi (a) 28(1): 379-385 (1975).

52. Lewis, K.L., G.S. Arthur, and S.A. Banyard, "Hydrogen-related Defects in

Vapour-Deposited Zinc Sulphide," Journal of Crystal Growth 66: 125-136

(1984).

363

53. Collins, A.K., et al., "Optical Characterization of Polycrystalline ZnS Produced

Via Chemical Vapor Deposition," 11th International Conference on Chemical

Vapor Deposition: 626-633 (1990).

54. Morozova, N.K., et al., "A Study of Luminescence Centers Related to Copper and

Oxygen in ZnSe," Semiconductors 35(1): 24-32 (2001).

55. Morozova, N.K., et al., "Pressure and Temperature Effects on Point-Defect

Equilibria and Band Gap of ZnS," Inorganic Materials 40(11): 1138-1145 (2004).

56. Morozova, N.K., et al., "Transformation of Luminescence Centers in CVD ZnS

Films subject to a high hydrostatic pressure," Semiconductors 38(1): 36-41

(2004).

57. Morozova, N.K., et al., "Absorption Spectrum of ZnO Precipitates in ZnSe,"

Inorganic Materials 39(8): 783-787 (2003).

58. Vagelatos, N., D. Wehe, and J.S. King, "Phonon dispersion and phonon densities of states for ZnS and ZnTe," The Journal of Chemical Physics 60(9): 3613-3618

(1974).

59. Klein, C.A., et al., "Lattice Absorption, Phonon Assignment, and Image-Spoiling

Properties of CVD ZnS in the Infrared," Laser Induced Damage in Optical

Materials (NBS Special Publication 541, Washington): 86-98 (1978).

60. Blodgett, D.W., et al. "Longwave infrared absorption properties of ZnS and

ZnSe," in Window and Dome Technologies VIII, SPIE 5078, 137-147 (2003).

61. Hahn, D.V., M.E. Thomas, and D.W. Blodgett, "Modeling of the frequency- and temperature-dependent absorption coefficient of long-wave-infrared (2-25 um) transmitting materials," Applied Optics 44(32): 6913-6920 (2005).

Y.S.,

Thermophysical Properties of Matter-Thermal Radiative

Properties - Nonmetallic Solids. John Wiley & Sons. (1972).

364

63. Birman, J.L., "Theory of Infrared and Raman Processes in Crystals: Selection

Rules in Diamond and Zincblende," Physical Review 131(4): 1489-1496 (1963).

64. Yu, P.Y. and M. Cardona, Fundamentals of Semiconductors: Physics and

Materials Properties, 3rd ed. Berlin: Springer (2005).

65. Bendow, B., H.G. Lipson, and S.P. Yukon, "Multiphonon absorption in highly transparent semiconducting crystals," Physical Review B 16(6): 2684 (1977).

66. Klein, C.A. and R.N. Donadio, "Infrared-active phonons in cubic zinc sulfide,"

Journal of Applied Physics 51(1): 797-800 (1980).

67. Marshall, R. and S.S. Mitra, "Optically active phonon processes in CdS and ZnS,"

Physical Review 134(4A): A1019-A1025 (1964).

68. Arguello, C.A., D.L. Rousseau, and S.P.S. Porto, "First-Order Raman Effect in

Wurtzite-Type Crystals," Physical Review 181(3): 1351 (1969).

69. Brafman, O. and S.S. Mitra, "Raman Effect in Wurtzite- and Zinc-Blende-Type

ZnS Single Crystals," Physical Review 171(3): 931 (1968).

70. Nilsen, W.G., "Raman spectrum of cubic ZnS," Physical Review 182(3): 838-850

(1969).

71. Serrano, J., et al., "Raman scattering in beta-ZnS," Physical Review B 69: 014301-

1 to 04301-11 (2004).

72. Xu, J., et al., "Surface vibrational mode of ZnS nanoparticles," Journal of Vacuum

Science and Technology B 15(4): 1465-1467 (1997).

73. Myers, S.M., et al., "Hydrogen interactions with defects in crystalline solids,"

Reviews of Modern Physics 64(2): 559-617 (1992).

74. Van de Walle, C.G. and J. Neugebauer, "Hydrogen in Semiconductors," Annual

Review of Materials Research 36(1): 179-198 (2006).

365

75. Khan, M.A., "New Band Systems of ZnH and ZnD in the Far Ultra-violet

Region," Proceedings of the Physical Society (London) 80: 599-607 (1962).

76. diBenedetto, B.A., J. Pappis, and A.J. Capriulo, "Chemical Vapor Deposition of

Multispectral Windows," in Technical Report AFAL-TR-73-252. (1973).

77. Iwata, H., S. Suzuki, and Y. Sasaki, "Iodine related grain growth in vapour deposited ZnS," Journal of Crystal Growth 125: 425-430 (1992).

78. Sack, R.O. and D.S. Ebel, "Thermochemistry of Sulfide Mineral Solutions,"

Reviews in Mineralogy & Geochemistry 61: 265-364 (2006).

79. Vaughan, D.J. and J.R. Craig, Mineral chemistry of metal sulfides: Cambridge

University Press (1978).

80. Roine, A., "HSC Chemistry 5.11 for Windows (computer program)," Pori,

Finland: Outokumpu Research Oy (2002).

81. Lewis, K.L. and J. Hill, "The Effect of Impurities on the Optical Properties of

CVD ZnSe and ZnS," 7th International Conference on Chemical Vapor

Deposition: 629-631 (1979).

82. Hertl, W., "Surface chemical properties of zinc sulfide," Langmuir 4: 594-598

(1988).

83. Karan, K., A.K. Mehrotra, and L.A. Behie, "On Reaction Kinetics for the Thermal

Decomposition of Hydrogen Sulfide," AIChE Journal 45(2): 383-389 (1999).

84. Higashihara, T., K. Saito, and H. Yamamura, "S

2

Formation during the Pyrolysis of H

2

S in Shock Waves," Bulletin of the Chemical Society of Japan 49(4): 965-

968 (1976).

85. Mellor, J.W., "Hydrogen Sulphide," in A Comprehensive Treatise on Inorganic

and Theoretical Chemistry. London: Longmans, Green and Co. Pp. 114-154

(1956).

366

86. Tsuchiya, K., K. Kamiya, and H. Matsui, "Studies on the oxidation mechanism of

H

2

S based on direct examination of the key reactions," International Journal of

Chemical Kinetics 29(1): 57-66 (1997).

87. Frenklach, M., et al., "Oxidation of hydrogen sulfide," Combustion and Flame 41:

1-16 (1981).

88. Greenwood, N.N. and A. Earnshaw, "Sulfur," in Chemistry of the Elements.

Oxford: Pergamon Press (1984).

89. Steudel, R., Y. Steudel, and M.W. Wong, "Speciation and Thermodynamics of

Sulfur Vapor," Topics in Current Chemistry 230: 117-134 (2003).

90. Rau, H., T.R.N. Kutty, and J.R.F. Guedes De Carvalho, "Thermodynamics of sulphur vapour," The Journal of Chemical Thermodynamics 5(6): 833-844 (1973).

91. Wiedemeier, H., "Computational analysis of solid-vapor equilibria for ZnS and

SrS phosphor synthesis conditions," Journal of Vacuum Science & Technology A:

Vacuum, Surfaces, and Films 24(3): 450-458 (2006).

92. Steele, S.R. and J. Pappis, "Chemical Vapor Deposition of IR Materials," in

Technical Report AFAL-TR-71-200 (1971).

93. Goela, J. and R.L. Taylor, "Monolithic material fabrication by chemical vapor deposition," Journal of Materials Science 23: 4331-4339 (1988).

94. De, A.K., et al., "Modeling of transport phenomena in a low-pressure CVD reactor," Journal of Crystal Growth 267: 598–612 (2004).

95. Sharifi, Y. and L.E.K. Achenie, "Effect of substrate geometry on the deposition rate in chemical vapor deposition," Journal of Crystal Growth 304(2): 520-525

(2007).

96. Bryant, W.A., "Review: The fundamentals of chemical vapour deposition,"

Journal of Materials Science 12: 1285-1306 (1977).

367

97. Sladek, K.J., "The Role of Homogeneous Reactions in Chemical Vapor

Deposition," Journal of the Electrochemical Society 118(4): 654-657 (1971).

98. Sharifi, Y. and L. E. Achenie, "Using density functional theory to postulate a mechanism for zinc sulfide formation in a CVD reactor," Journal of Crystal

Growth 307(2): 440-447 (2007).

99. Willingham, C.B., Personal Communication, "Hot isostatic pressing of ZnS,"

(2007).

100. Drezner, Y., S. Berger, and M. Hefetz, "A correlation between microstructure, composition and optical transparency of CVD-ZnS," Materials Science and

Engineering B 87: 59-65 (2001).

101. Savage, J.A., et al. "The role of a CVD research reactor in studies of the growth and physical properties of ZnS infrared optical material," in Advances in Optical

Materials, SPIE 505, 47-51 (1984).

102. diBenedetto, B.A., J. Pappis, and A.J. Capriulo, "Chemical Vapor Deposition of

IR Materials for Reconnaissance," in Technical Report AFAL-TR-72-309 (1972).

103. diBenedetto, B.A. and J. Pappis, "Chemical Vapor Deposition of Multispectral

Domes," in Technical Report AFML-TR-75-27. (1975).

104. Fang, Z., et al., "CVD growth of bulk polycrystalline ZnS and its optical properties," Journal of Crystal Growth 237-239: 1707-1710 (2002).

105. Gavrishchuk, E.M. and E.V. Yashina, "The Growth Mechanism of Zinc Sulfide from the Gas Phase," Vyskochistye Veshchestva 328(5): 36-39 (1994).

106. Gavrishchuk, E.M. and E.V. Yashina, "Zinc sulfide and zinc selenide optical elements for IR engineering," Journal of Optical Technology 71(12): 822-827

(2004).

107. Devyatykh, G.G., et al., "The Effect of CVD Conditions on the Porosity of Zinc sulfide," Inorganic Materials 31(8): 936-938 (1995).

368

108. Bansagi, T., et al., "Kinetics of hexagonal-cubic phase transformation of zinc sulfide in vacuo, in zinc vapor, and in sulfur vapor," Canadian Journal of

Chemistry 46(18): 2881-2886 (1968).

109. Xue, L.A. and R. Raj, "Superplastic deformation of zinc sulfide near its transformation temperature (1020 °C)," Journal of the American Ceramic Society

72

(10): 1792-1796 (1989).

110. Kim, Y.-D., et al., "Sintering behaviour of monodispersed ZnS powders," Journal

of Materials Science 32: 5101-5106 (1997).

111. Atkinson, H.V. and S. Davies, "Fundamental Aspects of Hot Isostatic Pressing:

An Overview," Metallurgical and Materials Transactions A 31A: 2981-3000

(2000).

112. Karaksina, E., V. Ikonnikov, and E. Gavrishchuk, "Recrystallization behavior of

ZnS during hot isostatic pressing," Inorganic Materials 43(5): 452-454 (2007).

113. Wright, K. and R.A. Jackson, "Computer simulation of the structure and defect properties of zinc sulfide," Journal of Materials Chemistry 5(11): 2037-2040

(1995).

114. Yashina, E.V., E.M. Gavrishchuk, and V.B. Ikonnikov, "Mechanisms of

Polycrystalline CVD ZnS Densification during Hot Isostatic Pressing," Inorganic

Materials 40(9): 901-904 (2004).

115. Adler, W.F., J.C. Botke, and T.W. James, "Response of Infrared-Transparent

Materials to Raindrop Impacts," in AFML-TR-79-4151 (1979).

116. Zwaag, S.v.d. and J.E. Field, "Liquid jet impact damage on zinc sulphide,"

Journal of Materials Science 17(9): 2625-2636 (1982).

117. Lewis, K.L., et al., "The Mechanical Properties of CVD-Grown Zinc Sulphide and their Dependence on the Conditions of Growth," 9th International

Conference on Chemical Vapor Deposition 84-6: 530-545 (1984).

369

118. Campbell, A. and C. Hayman. "Manufacturing Aspects of Zinc Sulphide," in

Recent Developments in Infrared Components and Subsystems, SPIE 0915, 79-83

(1988).

119. Kozhinova, I.A., et al., "Minimizing artifact formation in magnetorheological finishing of chemical vapor deposition ZnS flats," Applied Optics 44(22): 4671-

4677 (2005).

120. Gavrishchuk, E., P.L. Krupkin, and L.A. Kuznetsov, "Surface morphology of vapor-grown layers," Inorganic Materials 33(11): 1117-1120 (1997).

121. Shchurov, A.F., et al., "Structure and Mechanical Properties of Polycrystalline

Zinc Sulfide," Inorganic Materials 40(2): 96-101 (2004).

122. Devyatykh, G.G., E.M. Gavrishchuk, and E.V. Yashina, "Effect of Deposition

Conditions on the Microstructure of CVD ZnS," Inorganic Materials 32(6): 667-

669 (1996).

123. Lewis, K.L., et al. "Toughening Effects Induced in Zinc Sulfide by Hot Isostatic

Pressing," in Infrared and Optical Transmitting Materials, SPIE 683, 64-71

(1986).

124. Shchurov, A.F., et al., "Effect of Hot Isostatic Pressing on the Elastic and Optical

Properties of Polycrystalline CVD ZnS," Inorganic Materials 40(4): 336-339

(2004).

Performance. Bellingham, WA: SPIE Press (1999).

126. Henneman, L., et al. "Thermal, Structural and Optical Properties of Multispectral

Zinc Sulfide," in 11th DoD Electromagnetic Windows Symposium (2006).

127. Lugueva, N.V. and S.M. Luguev, "The Influence of Structural Features on the

Thermal Conductivity of Polycrystalline Zinc Sulfide," Physics of the Solid State

44

(2): 260-265 (2002).

370

128. Mecholsky, J.J., Jr., Personal Communication, "Grain size effect and toughness in

CVD ZnS and HIP'd CVD ZnS," (2008).

129. Wimmer, E., "Atomstic Modeling of Zinc Sulfide," Materials Design, Inc., prepared for Raytheon (2007).

130. Nickel, E.H., "A Review on the Properties of Zinc Sulphide," in Department of

Mines and Technical Surveys, Ottowa, Mines Branch Information Circular IC

170, Ottowa, Canada (1965).

131. Wemple, S.H. and M. DiDomenico, "Behavior of the Electronic Dielectric

Constant in Covalent and Ionic Materials," Physical Review B 3(4): 1338 (1971).

132. Lendvay, E., "Revealing of dislocations in ZnS by chemical etching," Journal of

Crystal Growth 59: 375-383 (1982).

Macmillan, Phillips (1975).

134. Cullity, B.D. and S.R. Stock, Elements of X-ray Diffraction, 3rd ed: Prentice Hall

(2001).

(2007).

136. Goldstein, J.I., et al., "Quantitative X-ray Analysis: The Basics," in Scanning

Electron Microscopy and X-ray Microanalysis. New York: Kluwer Academic &

Plenum. Pp. 391-451 (2003).

137. ASTM, "Standard Test Method for Monotonic Equibiaxial Flexural Strength of

Advanced Ceramics at Ambient Temperature," in ASTM-C1499-05: ASTM

Committee C28 on Advanced Ceramics (2005).

138. Pepi, J.W., Personal Communication, "Stress corrosion in ZnS and allowable stresses in glasses and ceramics," (2006).

371

139. Hartmann, H., et al., "Morphological stability and crystal structure of CVD-grown zinc selenide," Journal of Materials Science 26: 4917-4923 (1991).

140. Blank, H., et al., "Fault Structures in Wurtzite," physica status solidi (b) 7(3):

747-764 (1964).

141. Macintyre, J.E., ed. Dictionary of Inorganic Compounds. London: Chapman &

Hall. (1992).

142. Liu, M., et al., "Use of Metal Sulfides as Anode Catalysts in H

2

S-Air SOFCs,"

Journal of the Electrochemical Society 150(8): A1025-A1029 (2003).

143. He, P., et al., "Stabilization of Platinum Andode Catalyst in a H

2

S-O

2

Solid Oxide

Fuel Cell with an Intermediate TiO

2

Layer," Journal of the Electrochemical

Society 149(7): A808-A814 (2002).

144. Sharma, R.C. and Y.A. Chang, "The Ag-S (Silver-Sulfur) System," Bulletin of

Alloy Phase Diagrams 7(3): 263-268, 289-290 (1986).

145. Piacente, V., P. Scardala, and D. Ferro, "Study on the thermal decomposition of

Ag

2

S," Journal of Materials Science Letters 9(3): 365-367 (1990).

146. Becker, W. and H.D. Lutz, "Phase Studies in the Systems CoS-MnS, CoS-ZnS, and CoS-CdS," Materials Research Bulletin 13: 907-911 (1978).

147. Mironov, I.A., et al., "Physico-Mechanical and Structural Properties of

Polycrystalline Zinc Sulfide Obtained by Different Methods," Soviet Journal of

Optical Technology 59(7): 417-421 (1992).

148. Ma, C., et al., "Nanobelts, nanocombs, and nanowindmills of wurtzite ZnS,"

Advanced Materials 15(3): 228-231 (2003).

149. Hao, Y., et al., "Periodically Twinned Nanowires and Polytypic Nanobelts of

ZnS," Nano Letters 6(8): 1650-1655 (2006).

372

150. Holt, D.B. and M. Culpan, "Scanning electron microscope studies of striations in

ZnS," Journal of Materials Science 5(7): 546-556 (1970).

151. Smith, J.F., "Crystallography and Phase Equilibria: A Review: Part II - Space

Groups and Structure," Journal of Phase Equilibria and Diffusion 25(6): 497-506

(2004).

152. Klein, C.A. and C.B. Willingham, "Elastic properties of chemically vapordeposited ZnS and ZnSe," in Basic Properties of Optical Materials. Washington,

DC: NBS Special Publication Pp. 137-140 (1985).

153. Klein, C.A., R.P. Miller, and R.L. Gentilman, "Characteristic strength and

Weibull modulus of selected infrared-transmitting materials," Optical

Engineering 41(2): 3151-3160 (2002).

154. Rice, R.W., S.W. Freiman, and J.J. Mecholsky, "The Dependence of Strength-

Controlling Fracture Energy on the Flaw-Size to Grain-Size Ratio," Journal of the

American Ceramic Society 63(3-4): 129-136 (1980).

155. Zwaag, S., J.T. Hagan, and J.E. Field, "Studies of contact damage in polycrystalline zinc sulphide," Journal of Materials Science 15(12): 2965-2972

(1980).

156. Nam, F., Personal Communication, "Evans Analytical Group communication regarding IGA, GDMS, and SIMS detectability and uncertainty limits," (2008).

157. Yacobi, B.G. and Y. Brada, "Franz-Keldysh effect and internal fields in ZnS crystals," Physical Review B 10(2): 665 (1974).

158. Emelie, P.Y., et al., "Free carrier absorption and lattice vibrational modes in bulk

ZnO," Journal of Electronic Materials 35(4): 525-529 (2006).

159. Klein, C.A., R.P. Miller, and D.L. Stierwalt, "Surface and bulk absorption characteristics of chemically vapor deposited zinc selenide in the infrared,"

Applied Optics 33: 4304 (1994).

373

160. Mitra, S.S. and R. Marshall. "Multiphonon infrared absorption in II-VI semiconductors," in 7th International Conference on Physics of Semiconductors,

Academic Press, 1085-1090 (1964).

161. diBenedetto, B.A., J. Pappis, and A.J. Capriulo, "Improved Large ZnS Windows," in Technical Report AFAL-TR-73-176. (1973).

162. Pickering, M.A., et al. "Effect of Process Conditions and Chemical Composition on the Microstructure and Properties of Chemically Vapor Deposited SiC, Si,

ZnSe, ZnS and ZnS x

Se

1-x

," in Chemical Vapor Deposition of Refractory Metals

and Ceramics II, Materials Research Society 250, 145-160 (1992).

163. Sato, T., Furukawa, and H. Kashi, "CVD-ZnS IR Dome," Journal of the Japan

Society of Infrared Science and Technology 11: 44-49 (1986).

164. Harris, D.C., Personal Communication, "Cleartran calculated n & k values in longwave infrared from NIST data," (2008).

165. diBenedetto, B., Personal Communication, "On hexagonal needles in CVD ZnS & anisotropy in transmission in growth direction and perpendicular to growth direction," (2008).

166. Weakliem, H.A., "Optical Spectra of Ni

2+

, Co

2+

, and Cu

2+

in Tetrahedral Sites in

Crystals," Journal of Chemical Physics 36(8): 2117-2140 (1962).

167. Gumlich, H.-E. and H.-J. Schulz, "Optical Transitions in ZnS Type Crystals

Containing Cobalt," Journal of Physics and Chemistry of Solids 27: 187-195

(1966).

168. Tsai, T.-Y. and M. Brinbaum, "Co

2+

:ZnS and Co

2+

:ZnSe saturable absorber Q switches," Journal of Applied Physics 87(1): 25-29 (2000).

169. Trager, F., "Interaction of light with optical materials," in Springer Handbook of

Lasers and Optics, F. Trager, Editor. New York: Springer. Pp. 250-281 (2007).

374

170. Van de Hulst, H.C., Light Scattering by Small Particles, originally published 1957 by John Wiley & Sons ed. New York: Dover (1981).

171. Muller, B., G. Roussos, and H.-J. Schulz, "Photoluminescence and Excitation

Spectroscopy of Ni

2+

in ZnS Crystals," Journal of Crystal Growth 72:

360-363 (1985).

and Ni

+

172. Wimmer, E., "Atomistic Modeling of Zinc Sulfide: Project 2: Interpretation of

IR Absorption," Materials Design, Inc., prepared for Raytheon (2007).

173. Bohren, C.F. and D.R. Huffman, Absorption and Scattering of Light by Small

Particles. New York: John Wiley & Sons (1983).

174. Zelinski, B., et al., "Unpublished MS on Effective media and scattering,"

Raytheon (2008).

175. Chen, W.W. and B. Dunn, "Characterization of pore size distribution by infrared scattering in highly dense ZnS," Journal of the American Ceramic Society 76(8):

2086-2092 (1993).

176. Palik, E.D., ed. Handbook of Optical Constants of Solids, Volume 1. New York:

Academic Press. (1980).

177. Xue, L.A. and R. Raj, "Effect of hot-pressing temperature on the optical transmission of zinc sulfide," Applied Physics Letters 58(5): 441-43 (1991).

178. Harvey, J.E. and A. Krywonos. "Understanding diffraction effects in novel systems containing nanostructures," in Novel optical systems design and

optimization IX, SPIE 6289, 6289M1-12 (2006).

179. Neumark, G.F., "Theory of the Anomalous Photovoltaic Effect in ZnS," Physical

Review 125(3): 838-845 (1962).

180. Brada, Y., "Electrical conductivity of ZnS crystals: ZnS crystals as quasi-onedimensional conductors," Physical Review B 39(11): 7645 (1989).

375

181. Brafman, O. and I.T. Steinberger, "Optical Band Gap and Birefringence of ZnS

Polytypes," Physical Review 143(2): 501 (1966).

182. Singer, J., "Study of the bands of uniform birefringence in faulted ZnS," Journal

of Physics and Chemistry of Solids 24(12): 1645-1646 (1963).

183. Yacobi, B.G. and Y. Brada, "Defect-induced electric fields in ZnS," Solid State

Communications 18(1): 135-138 (1976).

184. Omel’chenko, S.A., M.F. Bulanyi, and O.V. Khmelenko, "Effect of the electric fields of immobile dislocations on photoluminescence and EPR in deformed ZnS crystals," Physics of the Solid State 45(9): 1608-1613 (2003).

185. Recnik, A., et al., "Structure and Chemistry of Basal-Plane Inversion Boundaries in Antimony Oxide-Doped Zinc Oxide," Journal of the American Ceramic

Society 84(11): 2657-2668 (2001).

186. Flewitt, P.E.J. and R.K. Wild, Grain Boundaries: Their Microstructure and

Chemistry. Chichester: John Wiley & Sons (2001).

187. Butrymowicz, D.B., et al., "Diffusion-induced grain-boundary migration," in

Character of Grain Boundaries, M.F. Yan and A.H. Heuer, Editors. Columbus,

Ohio: American Ceramic Society. Pp. 202-212 (1983).

188. Handwerker, C.A., R.L. Coble, and J.E. Blendell, "Diffusion-induced grainboundary migration and discontinuous precipitation experiments in ceramic systems," in Character of Grain Boundaries, M.F. Yan and A.H. Heuer, Editors.

Columbus, Ohio: American Ceramic Society. Pp. 213-223 (1982).

189. Carter, W.C. and C.A. Handwerker, "Morphology of Grain Growth in Response to Diffusion Induced Elastic Stresses: Cubic Systems," Acta Metallurgica et

Materialia 41(5): 1633-1642 (1993).

190. Yan, M.F., R.M. Cannon, and H.K. Bowen, "Solute Segregation at Ceramic

Interfaces," in Character of Grain Boundaries, M.F. Yan and A.H. Heuer,

Editors. Columbus, Ohio: American Ceramic Society. Pp. 255-273 (1982).

376

191. Recnik, A., M. Ceh, and D. Kolar, "Polytype induced exaggerated grain growth in ceramics," Journal of the European Ceramic Society 21: 2117-2121 (2001).

192. Allen, S.M. and E.L. Thomas, The Structure of Materials. New York: John Wiley

& Sons (1999).

193. Sebastian, M.T. and P. Krishna, "An X-ray diffraction study of faulting in single crystals of cubic ZnS grown from the vapour phase," Philosophical Magazine A

49

(6): 809-821 (1984).

194. Geilikman, M.B., "Mechanisms of Polytype Stabilization During Wurtzite-

Sphalerite Transition," Physics and Chemistry of Minerals 8: 2-7 (1982).

195. Engel, G.E. and R.J. Needs, "Total energy calculations on zinc sulphide polytypes," Journal of Physics: Condensed Matter 2(2): 367-376 (1990).

196. Kiflawi, I., Z.H. Kalman, and Y. Sonnenblick, "Direct observation of polytype transformations in a vapour-phase grown ZnS crystal," Journal of Crystal Growth

34

: 145-148 (1976).

197. Engel, G.E., "Simulation of polytype formation in zinc sulphide," Journal of

Physics: Condensed Matter 2(33): 6905-6919 (1990).

198. Hill, J., K.L. Lewis, and A. Cullis. "The Defect and Band Structure of CVD-

Grown Zinc Sulphide," in 6th International Conference on Chemical Vapor

Deposition, Electrochemical Society, 276-282 (1977).

199. Yu, H., et al. "The study of the relationship between inside defects and optical properties of CVD ZnS," in Advanced Optical Manufacturing and Testing

Technology 2000, SPIE 4231, 224-230 (2000).

200. Durovic, S., P. Krishna, and R. Pandey, "Layer Stacking," in International Tables

for Crystallography. Pp. 744-765 (1999).

377

201. Palosz, B. and J. Przedmojski, "Analysis of Polytypic Structure with Stacking

Faults on the Basis of X-ray Powder Patterns," Acta Crystallographica A34: 165-

170 (1978).

202. Gosk, J.B., "Investigation of One-Dimensionally Disordered Structures of AII-

BVI Crystals by Monte Carlo Technique: I. The 3C Disordered Structure and the

3C Structure with Different Kinds of Stacking Faults," Crystal Research

Technology 35(1): 101-116 (2000).

203. Sebastian, M.T. and P. Krishna, "X-ray investigation of the mechanism of phase transformation in single crystals of ZnS, Zn x

Cd

1-x

S and Zn x

Mn

1-x

S (I):

Calculation of diffraction effects by a three parameter model," Crystal Research

Technology 22(7): 929-941 (1987).

204. Sebastian, M.T. and P. Krishna, "X-ray investigation of the mechanism of phase transformation in single crystals of ZnS, Zn x

Cd

1-x

S and Zn x

Mn

1-x

S (II):

Comparison of observed and calculated diffraction effects," Crystal Research

Technology 22(8): 1063-1072 (1987).

205. Sebastian, M.T., "X-ray diffraction from faulted zinc sulphide crystals undergoing phase transitions," Journal of Materials Science 23(6): 2014-2020 (1988).

206. Sebastian, M.T., R. Pandey, and P. Krishna, "X-ray diffraction study of the 2H to

3C solid state transformation in vapour grown single crystals of ZnS," Physica

Status Solidi (a) 71: 633-640 (1982).

207. Gosk, J.B., "Investigation of One-Dimensionally Disordered Structures of AII-

BVI Crystals by Monte Carlo Technique: II. 2H Structure with Different Kinds of Stacking Faults," Crystal Research Technology 36(2): 197-213 (2001).

208. Gosk, J.B., "Investigation of One-Dimensionally Disordered Structures of AII-

BVI Crystals by Monte Carlo Technique: III. 4H Structure with Different Kinds of Stacking Faults," Crystal Research Technology 38(2): 160-173 (2003).

209. Varn, D.P., "Language Extraction from ZnS," Doctor of Philosophy, University of

Tennessee, Physics (2001).

378

210. Varn, D.P., G.S. Canright, and J.P. Crutchfield, "Discovering planar disorder in close-packed structures from x-ray diffraction: Beyond the fault model," Physical

Review B 66(17): 174110 (2002).

211. Varn, D.P. and J.P. Crutchfield, "From finite to infinite range order via annealing: the causal architecture of deformation faulting in annealed close-packed crystals,"

Physics Letters A 324(4): 299-307 (2004).

J.I.,

Publications (1971).

213. Lide, D.R., ed. CRC Handbook of Chemistry and Physics. 87th ed., Boca Raton:

Taylor and Francis. (2006).

214. Park, Y.S. and B.K. Shin, "Recent Advances in Injection Luminescence," in

Topics in Applied Physics: Electroluminescence, J.I. Pankove, Editor. Berlin:

Springer-Verlag. Pp. 133-170 (1977).

215. Kukimoto, H., et al., "Infrared absorption due to donor states in ZnS crystals,"

Journal of the Physics and Chemistry of Solids 29: 935-944 (1968).

216. Segall, B., "Band Structure," in Physics and Chemistry of II-VI Compounds, M.

Aven and J.S. Prener, Editors. Amsterdam: North-Holland. Pp. 3-72 (1967).

217. Ruda, H.E. and B. Lai, "Electron transport in ZnS," Journal of Applied Physics

68

(4): 1714 (1990).

Numerical and Graphical Information. Boston: Kluwer Academic Publishers

(1999).

219. Thomas, G.A. and V.B. Timofeev, "A Review of N=1 to infinity Particle

Complexes in Semiconductors," in Handbook on Semiconductors, M. Balkanski,

Editor. Amsterday: North-Holland. Pp. 45-63 (1980).

379

220. Morozova, N.K., L.V. Gamosov, and E.E. Lakin, "Ultraviolet spectra of some zinc sulfide polytypes and crystals with stacking faults," Optics and Spectroscopy

37

(6): 639-642 (1974).

221. Peyghambarian, N., S.W. Koch, and A. Mysyrowicz, Introduction to

Semiconductor Optics. Englewood Cliffs, NJ: Prentice-Hall (1993).

222. Shachar, G., et al., "Electroabsorption and Internal Fields in ZnS Crystals,"

Journal of Applied Physics 41(2): 723-728 (1976).

223. Birman, J.L., "Polarization of Fluorescence in CdS and ZnS Single Crystals,"

Physical Review Letters 2(4): 157 (1959).

224. Dielman, J., et al. "Reversible light-induced blackening by charge transfer in zinc sulphide single crystals," in 7th International Conference on Physics of

Semiconductors, Academic Press, 941-944 (1964).

225. Albers, W., "Physical Chemistry of Defects," in Physics and Chemistry of II-VI

Compounds, M. Aven and J.S. Prener, Editors. Amsterdam: North-Holland. Pp.

165-222 (1967).

226. Schmidt-Mende, L. and J.L. MacManus-Driscoll, "ZnO - nanostructures, defects, and devices," Materials Today 10(5): 40-48 (2007).

Press, Taylor and Francis (2006).

York: Taylor and Francis (2003).

229. Thompson, H.W. and C.N. Hinshelwood, "The Mechanism of the Homogeneous

Combination of Hydrogen and Oxygen," Proceedings of the Royal Society of

London. Series A, Containing Papers of a Mathematical and Physical Character

122

(790): 610-621 (1929).

380

230. Mellor, J.W., "The Chemical Properties of Water," in A Comprehensive Treatise

on Inorganic and Theoretical Chemistry. London: Longmans, Green and Co. Pp.

483-493 (1956).

231. Egerton, A. and D.R. Warren, "Kinetics of the Hydrogen/Oxygen Reaction. I. The

Explosion Region in Boric Acid-Coated Vessels " Proceedings of the Royal

Society of London. Series A, Mathematical and Physical Sciences 204(1079): 465-

476 (1951).

York: Marcel Dekker, Inc. (1991).

233. Prabhu, G.M., D.L. Ulrichson, and A.H. Pulsifer, "Kinetics of the Oxidation of

Zinc Sulfide," Industrial & Engineering Chemistry Fundamentals 23: 271-273

(1984).

234. Sohn, H.Y. and D. Kim, "Intrinsic Kinetics of the Reaction between Zinc Sulfide and Water Vapor," Metallurgical Transactions B 18B: 451-457 (1987).

235. Perry, R.H. and D.W. Green, eds. Perry's Chemical Engineers' Handbook. 7th ed.

McGraw-Hill Professional. (1997).

236. Ritter, E. and R. Hoffmann, "Influence of substrate temperature on the condensation of vacuum evaporated films of MgF2 and ZnS," Journal of Vacuum

Science and Technology 6: 733-736 (1969).

237. Stolen, S. and T. Grande, Chemical Thermodynamics of Materials: Macroscopic

and Microscopic Aspects: John Wiley & Sons (2004).

238. Goldfinger, P. and M. Jeunehomme, "Mass Spectrometric Studies of the

Thermodynamic Properties of Group III-V and II-VI Compounds and the Group

V and Group VI Elements," in Advances in Mass Spectrometry, J.D. Waldron,

Editor. London: Pergamon. Pp. 534-546 (1959).

(1970).

381

240. Dutt, M.B. and B.L. Sharma, "Diffusion in Compound Semiconductors," in

Diffusion in Semiconductors: Springer-Verlag. Pp. 1-87 (1998).

241. Baba, H., "Some Effects of Zinc Atmospheres on Zinc Sulfide," Journal of the

Electrochemical Society 110(1): 79-81 (1963).

242. Langdon, T.C. and F. Mohamed, A., "The incorporation of ambipolar diffusion in deformation maps for ceramics," Journal of Materials Science 13: 473-482

(1978).

controlled Processes. Berlin: Springer (2007).

244. Lott, K., T. Nirk, and O. Volobujeva, "Chemical self-diffusion in undoped ZnS and in undoped CdSe," Crystal Engineering 5: 147-153 (2002).

245. Lott, K., et al., "High Temperature Electrical Conductivity in Cu-Doped ZnS,"

physica status solidi (b) 229(1): 361-364 (2002).

246. Lott, K., et al., "High temperature electrical conductivity in donor-doped II–VI compounds," physica status solidi (c) 0(2): 618-621 (2003).

247. Gobrecht, H., et al., "Self diffusion of sulphur in zinc sulphide," Solid State

Communications 5: 777-778 (1967).

248. Williams, V.A., "Diffusion of some impurities in zinc sulphide single crystals,"

Journal of Materials Science 7: 810-812 (1972).

249. Hashin, Z. and S. Shtrickman, "A Variational Approach to the Theory of the

Elastic Behaviour of Polycrystals," Journal of the Mechanics and Physics of

Solids 10: 343-352 (1962).

250. Wimmer, J.M., "Mechanical and Physical Properties of Chemically Vapor-

Deposited (CVD) Zinc Sulfide," in AFML-TR-79-4013 (1979).

382

251. Berlincourt, D., H. Jaffe, and L.R. Shiozawa, "Electroelastic Properties of the

Sulfides, Selenides, and Tellurides of Zinc and Cadmium," Physical Review

129

(3): 1009 (1963).

252. Townsend, D. and J.E. Field, "Fracture toughness and hardness of zinc sulphide as a function of grain size," Journal of Materials Science 25: 1347-1352 (1990).

253. Shockey, D.A., D.J. Rowcliffe, and K.C. Dao, "Fracture Toughness of CVD

ZnS," in Topical Report from Stanford Research Institute for Office of Naval

Research (1977).

254. ASTM, "Standard Practice for Reporting Uniaxial Strength Data and Estimating

Weibull Distribution Parameters for Advanced Ceramics," in ASTM-C1239-06:

ASTM Committee C28 on Advanced Ceramics (2006).

255. Klein, C.A. and R.P. Miller. "How to do a Weibull Statistical Analysis of Flexural

Strength Data," in Window and Dome Technologies and Materials VII, SPIE

4375

, 241-257 (2001).

256. Huie, J., C. Dudding, and J. McCloy. "Polycrystalline Yttrium Aluminum Garnet

(YAG) for IR Transparent Missile Domes and Windows," in Window and Dome

Technologies and Materials X, SPIE 6545(2007).

257. Wu, D., J. Zhou, and Y. Li, "Unbiased estimation of Weibull parameters with the linear regression method," Journal of the European Ceramic Society 26: 1099-

1105 (2006).

258. Harris, D.C., et al. "Mechanical Evaluation of Yttria," in Window and Dome

Technologies and Materials II, SPIE 1326, 23-32 (1990).

259. Salem, J.A. and L. Powers, "Guidelines for the Testing of Plates," 27th

International Cocoa Beach Conference on Advanced Ceramics and Composites:

B 24(4): 357-367 (2003).

260. Salem, J., Personal Communication, "ASTM standards on fracture testing effective area calculation and conversion of data between test types," (2007).

261. Holt, D.B., "Defects in the Sphalerite Structure," Journal of Physics and

Chemistry of Solids 23: 1353-1362 (1962).

262. Chang, C.S., J.L. He, and Z.P. Lin, "The grain size effect on the empirically determined erosion resistance of CVD-ZnS," Wear 225: 115-120 (2003).

263. Chechetkina, E.A., et al., "Solubility of ZnO in ZnS (Wurtzite)," Inorganic

Materials 14(8): 1086-1088 (1978).

J.C.,

WA: SPIE Optical Engineering Press (1995).

265. Elson, J.M., J.P. Rahn, and J.M. Bennett, "Relationship of the total integrated scattering from multilayer-coated optics to angle of incidence, polarization, correlation length, and roughness cross-correlation properties," Applied Optics

22

(20): 3207 (1983).

266. Germer, T.A., "Predicting, Modeling, and Interpreting Light Scattered by

Surfaces," in SPIE Short Course SC492, course notes (n.d.).

383

Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement

Table of contents