The Pennsylvania State University

Home , Dealkalization, Sodium silicate

The Graduate School

College of Engineering

UNDERSTANDING SHEAR-INDUCED HYDROLYSIS REACTIONS ON SODA LIME

SILICA GLASS SURFACE

A Dissertation in

Chemical Engineering

Jiawei Luo

 2018 Jiawei Luo

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

May 2018

The dissertation of Jiawei Luo was reviewed and approved* by the following:

Seong H. Kim Professor of Chemical Engineering and Materials Science Dissertation Advisor Chair of Committee

Carlo G. Pantano Distinguished Professor of Materials Science and Engineering Dissertation Coadvisor

Manish Kumar Associate Professor of Chemical Engineering Associate Professor of Biomedical Engineering Associate Professor of Civil and Environmental Engineering

Kristen Fichthorn Merrell Fenske Professor of Chemical Engineering Professor of Physics Head of the Department Graduate Program

*Signatures are on file in the Graduate School

iii

ABSTRACT

Soda lime silica (SLS) glass is the most widely used glass materials in terms of mass in windows, bottles & containers, etc. One of the biggest challenge for SLS glass is its brittleness or propensity to be damaged at ambient conditions. One important reason is that cracks propagation on the SLS glass surface can be significantly accelerated by tensile stress induced corrosion reactions of amorphous silicate network, which breaks the Si-O-Si bonds to form two silanol groups, known as stress corrosion effect. It has also been found that stress corrosion reactions are faster with the increase of relative humidity (RH). The propagation of cracks will eventually lead to the failure of SLS glass. However, very little is known on the reactions between glass and water in the environment induced by interfacial shear, which results in the removal of SLS glass.

Recently, it has been observed that SLS glass shows unique response to interfacial shear in humid conditions through a ball-on-flat wear test. For most of the commercial flat glasses, including silica, borosilicate glasses, boroaluminosilicate glasses and aluminosilicate glasses, the amount of wear will increase as the humidity level increases. Only SLS glass shows wear resistance at high relative humidity. It is found that shear-induced hydrolysis reactions take place on SLS glass surface for medium and low RH conditions. Then this shear-induced hydrolysis reaction must be “suppressed” or have similar reaction rates of its reverse reactions (condensation

+ + of silanols) at high RH. One hypothesis is that Na /H + H2O exchange could take place at high

RH, which creates surface residual compressive stress at the meantime. Compressive stress which could lower the effective tensile stress on glass surface, is believed to stabilize or hinder the propagation of cracks. Then the amount of wear could also be lowered if the applied stress is not large enough to induce hydrolysis reactions.

To test this hypothesis and explore the factors that govern the shear-induced reactions on

SLS glass surface, effects of different structural units, surface mechanical properties and

+ + compressive stress, Na /H + H2O exchange are investigated in this dissertation. It should be noted that these factors cannot be easily isolated. Therefore, multiple different treatments need to be performed to modify the surface chemical structures of SLS glass. These surface treatments include hydrothermal reactions, leaching in acid solution, thermal poling in controlled environment, ion-exchange with KNO3. The governing factors are determined through correlating the changes in surface chemical structure, mechanical properties with the wear behavior. It is found that surface mechanical properties, compressive stress created by ion-exchange process and

+ + alteration layers created by Na /H + H2O exchange are not responsible for SLS glass’s unique wear behavior at high RH. These conclusions has been demonstrated by analyzing the chemical structure and mechanical properties of modified SLS surface by hydrothermal treatment,

Na+/K+& Na+/Ag+ exchange, thermal poling treatment and leaching in acid solutions. While no specific mechanism has been identified, mobile Na+ cations and silicate network structure are suggested to be the dominating factors.

Another contribution of this dissertation is the improved understanding of amorphous silicate network of glass surfaces which plays vital roles in shear induced reactions. Several novel approaches based on non-destructive spectroscopy have been developed to describe the wide distribution of structural parameters of silicate network. With the assist of molecular dynamics

(MD) simulation, the weighted mean of Si-O bond length is in linear relationship with the weighted mean of Si-O-Si asymmetric stretch absorbance. Then a new mathematical algorithm has been developed to extract absorbance of Si-O-Si asymmetric stretch from specular reflectance infrared spectra. By utilizing the selection rule of sum frequency generation (SFG) spectroscopy,

O-HO distribution in the subsurface of SLS glass has been roughly identified, which is in good agreement with theoretical prediction. Speciation of hydrous species (ratio of SiOH/H2O) can be determined through a new algorithm based on attenuated total reflectance infrared (ATR-IR) spectroscopy and H depth profile.

A new hypothesis on the unique wear behavior based on the current findings is proposed: shear induced hydrolysis and condensation reactions are in dynamic equilibrium in the presence of Na+. Existence of Na+ can lower the activation energy barrier, especially for condensation reactions. At high RH, Na+ can migrate through the adsorbed water layers and moved to the silanol sites to “catalyze” condensation reactions. To prove this hypothesis, MD simulation with reactive force fields as well as experimental work are required.

TABLE OF CONTENTS

List of Figures ...... x

List of Tables ...... xix

Publications ...... xx

Preface… ...... xxii

Acknowledgements ...... xxiii

Chapter 1 Introduction ...... 1

Motivation ...... 1 Understand shear-induced hydrolysis reactions of SLS glass at high RH by controlling and analyzing the surface chemical structure and mechanical properties ...... 3 Background and literature review ...... 5 Unique surface chemical structure of SLS glass ...... 5 Stress-induced reactions between SLS glass surface and H2O (g) in the environment ...... 10

Chapter 2 Summary of employing specular reflectance infrared (SR-IR) spectroscopy, attenuated total reflectance infrared (ATR-IR) spectroscopy, environmental-control indentation and wear tests to analyze the surface structure and mechanical properties of SLS glass ...... 20

Overview ...... 20 Applying SR-IR and ATR-IR to analyze the structure of Si-O-Si network and hydrous species in the surface region of SLS glass ...... 21 Differences of SR-IR and ATR-IR in analyzing glass materials ...... 21 Obtaining surface Si-O-Si structure with SR-IR and hydrous species with ATR- IR ...... 24 Friction and wear test for glasses under applied tensile stress ...... 31 Cleaning procedure prior friction and wear test ...... 31 Characterization of wear track with optical profilometry ...... 32 Holder design for wear test of glass under applied tensile stress ...... 33 Labview software development for crack initiation load analysis of glass materials with Hertzian indentation ...... 34 Environmental controlled Vickers indentation ...... 37

Chapter 3 Hydrothermal reactions of soda lime silica glass – revealing subsurface damage and alteration of mechanical properties and chemical structure of glass surfaces ...... 40

Overview ...... 40 Introduction ...... 40 Experimental methods ...... 43

vii

Results and Discussion ...... 46 Selective etching of regions with subsurface damage or residual stress via vapor-phase hydrothermal reaction ...... 46 Mechanical properties of hydrothermally-treated SLS glass surface ...... 48 Mechanochemical wear of hydrothermally-treated SLS glass surface ...... 52 Subsurface structural changes upon vapor-phase hydrothermal reactions ...... 53 Conclusion ...... 60

Chapter 4 Thermal poling of soda lime silica glass with nonblocking electrodes –: Effects of sodium ion migration and water ingress on glass surface structure and wear behavior ...... 62

Overview ...... 62 Introduction ...... 63 Experimental methods ...... 65 Results & Discussions ...... 69 Chemical composition and thickness of the thermally-poled surface layers ...... 69 Silicate network structure in thermally-poled surfaces ...... 74 Hydrous species (Si-OH, H2O) in poled surfaces ...... 78 Effects of thermal poling on mechanochemical wear in humid ambient ...... 85 Conclusion ...... 89 Supporting information ...... 89

Chapter 5 Chemical structure and mechanical properties of soda lime silica glass surfaces treated by thermal poling in inert and reactive ambient gases ...... 91

Overview ...... 91 Introduction ...... 92 Experimental methods ...... 94 Results and discussion ...... 98 Chemical composition and thickness of the thermally-poled surface layers ...... 98 Changes in silicate network in the sodium-depleted surface after thermal poling ... 101 Molecular species trapped in the sodium-depleted surface ...... 103 Probing hydrogen bonding interactions among hydrous species with SFG ...... 108 Surface mechanical properties after thermal poling reactions in different environments ...... 111 Conclusion ...... 115

Chapter 6 Effects of Na+/K+-ion exchange on mechanical and mechanochemical properties of soda lime silica glass ...... 117

Overview ...... 117 Introduction ...... 118 Experimental details ...... 119 Results ...... 122 Concentration profile of the ion-exchanged SLS glass surface ...... 122 Effects of ion exchange on mechanical responses under normal indentation ...... 123 Effects of ion exchange on mechanochemical wear under tangential shear ...... 127 Effects of ion exchange on subsurface silicate network and hydrous species ...... 128

viii

Discussion ...... 130 Conclusion ...... 132 Supporting Information ...... 133

Chapter 7 Complex refractive index of silica, silicate, borosilicate, and boroaluminosilicate glasses-Analysis of glass network vibration modes with specular-reflectance IR spectroscopy ...... 135

Overview ...... 135 Introduction ...... 135 Methods ...... 138 Experimental analysis ...... 138 Mathematical algorithm to obtain 풏() + 풊풌() from two-angle SR-IR data ...... 140 Results ...... 143 Validation of the mathematical algorithm using known 풏() + 풊풌() ...... 143 Comparison with spectroscopic ellipsometry (SE) analysis ...... 146 Discussions...... 150 Comparison of 풌() spectra of various types of glass ...... 150 Silica and Silicate glass ...... 153 Borosilicate glass...... 156 Boroaluminosilicate glass ...... 157 Challenges and Opportunities in analysis of 풌() spectrum of glass ...... 158 Conclusions ...... 160 Supporting Information ...... 160 Alignment of a SeagullTM variable angle reflection accessory ...... 160 Comparison of the SR-IR spectra of fused quartz used in this study and those calculated from the refractive index reported in the literature ...... 162 Comparison of 풏() and 풌() spectra of ISG obtained from TASR-IR and IR- VASE methods ...... 164 Kramers-Kronig consistency of the results obtained with the TASR-IR and IR- VASE methods ...... 164 Raw SR-IR spectra of the flat glass surfaces analyzed in this study ...... 166 IR penetration depth in SR-IR analysis ...... 167

Chapter 8 Molecular dynamics study of correlations between IR peak position and bond parameters of silica and silicate glasses: Effects of temperature and stress ...... 168

Overview ...... 168 Introduction ...... 168 Experimental and simulation methods ...... 172 Experimental details ...... 172 MD simulation details ...... 173 Results and Discussions ...... 176 Comparison of refractive index calculated from three different potentials ...... 176 Validation of MD simulations with experimentally-observed temperature dependence ...... 177 MD simulations of mechanical stress effect on refractive index and IR reflectance ...... 178 Correlations between the simulated IR spectral features and bond parameters ...... 179

Simplified vibrational normal mode analysis ...... 183 Implication for chemically-strengthened and thermally-tempered SLS glass ...... 184 Conclusion ...... 187 Supporting information ...... 188 SR-IR analysis of a SLS float glass (microscope slide glass) under tensile stresses and a SLS glass (5 mm thick glass) with surface compressive stress prepared by thermal tempering...... 188 Potentials used in MD simulation ...... 189 Correlation between 풌(흎) with bond parameters ...... 191 Nanoindentation analysis of the convex surface of a SLS glass under three- point bending ...... 192

Chapter 9 Vibrational Sum Frequency Generation (SFG) Spectroscopy Study of Hydrous Species in Soda Lime Silica (SLS) Float Glass ...... 193

Overview ...... 193 Introduction ...... 194 Experimental details ...... 199 Results and discussion ...... 201 Conclusion ...... 215 Supporting Information ...... 216 Appendix A Wear behavior of SLS glass surfaces after leaching in acid solution ...... 220 Preparation of leached sample with ~100 nm layer leached layer ...... 220 Silicate network modification and introduction of hydrous species ...... 221 Modification of wear behavior of SLS glass with 100 nm thick surface layer ...... 222 Appendix B Determining the relative abundance of SiOH/H2O in monolithic flat glass surfaces using ATR mid-IR and hydrogen depth profiles ...... 224 Introduction ...... 224 Experimental details ...... 226 Results and discussion ...... 227 Conclusion ...... 236 Reference ...... 237

LIST OF FIGURES

Figure 1-1. Schematics of modified surface chemical composition of SLS float glass. Na+ and H will form gradient from surface to the bulk. Depending on the commercial delakalization treatment, the thickness of this modified layer could be between 50 and 150 nm...... 8

Figure 1-2. Schematic of chemical environment of Na+ and hydrous species in the surface region of SLS glass...... 9

Figure 1-3. Schematics of the relationship between crack propagation velocity and applied stress from Ref. 20...... 13

Figure 1-4. Line profiles of the wear tracks on different glasses under different humidity conditions: (a)fused quartz; (b) soda-lime silica glass; (c) BF33; (d) AF45; (e) sodium alumino-silicate; (f) K-exchanged alumino-silicate.36...... 17

Figure 2-1. Schematic illustration of transmission, specular reflection (SR), and attenuated total reflection (ATR) IR spectroscopy of a flat glass sample...... 22

Figure 2-2. Real (n) and imaginary (k) components of refractive index of soda lime glass.46 ...... 23

Figure 2-3. (a) IR penetration depth, dp, inside soda lime glass calculated with k(λ) shown in Figure 2-2. (b) SR-IR spectrum calculated for a 700 μm thick soda lime glass using equations (1) – (5) at an incidence angle of 40. The dotted lines are the components calculated for the reflection from the front and back surfaces. (c) Experimentally obtained SR-IR spectra of soda lime glass with different backside reflection conditions...... 27

Figure 2-4. (a) Diamond ATR-IR, and (b) Ge ATR-IR spectra of soda lime glass calculated with eqs. (1) – (3) using the refractive index shown in Figure 2 at different incidence angles (θi). Note that the calculated spectra for i = 1 and 20 correspond to the SR-IR data since the incidence angle is lower than the critical angle (c)...... 29

Figure 2-5. (a) Comparison of ATR-IR spectra of SLS glass from Asahi Co. (0.7 mm) collected by diamond and Ge ATR crystal; (b) Calculated information depth of SLS glass with diamond crystal (45 incidence angle) and Ge crystal (60 incidence angle)...... 30

Figure 2-6. Schematic of Friction and wear test in controlled environment...... 32

Figure 2-7. Typical optical profilometry image of substrate and ball after friction and wear test in 40% RH and 90% RH conditions...... 33

Figure 2-8. Drawings of sample holder to perform applied tensile stress on glass substrate...... 34

Figure 2-9. Schematic of crack initiation load detection with a Hertzian indentater...... 35

Figure 2-10. Labview software interface to perform crack initiation load analysis on glass surface...... 37

Figure 2-11. Schematic of Vickers indentation in humidity controlled environment...... 38

Figure 2-12. Vickers imprint of SLS glass in 90% and 0% relative humidity. The load is 300gf for both cases...... 39

Figure 3-1. Schematic representation of vapor-phase hydrothermal treatment of SLS glass at temperatures higher than 100 oC...... 44

Figure 3-2. Optical profilometry of SLS glass surface slide tracks produced in n-pentanol VPL environments (a) before and (b) after hydrothermal treatment at 150 °C for 24 hours; (c) cross-section line profiles of the wear tracks marked with dashed lines in (a) and (b)...... 47

Figure 3-3. (a) Vickers indent prepared with 500 gf load in 40% relative humidity; (b) indent from (a) after hydrothermal treatment at 150 °C for 24 hours...... 48

Figure 3-4. (a) Hardness and (b) reduced modulus of the SLS glass surfaces before and after hydrothermal treatment determined from nanoindentaiton...... 49

Figure 3-5. Vickers indentation before and after hydrothermal treatment at 150 C for 72 hours under 300 gf load conditions at 40% relative humidity. (a) Summary of Vickers hardness; (b) Optical image of Vickers indentation of SLS glass before hydrothermal treatment; (c) Optical image of Vickers indentation of SLS glass after hydrothermal treatment...... 50

Figure 3-6. Comparison of wear tracks formed after 400 sliding cycles in (a) 40% relative humidity (b) 90% relative humidity on the pristine sample, the sample treated at 150 C for 24 hours in dry air, and the sample treated at 150 C for 24 hours in steam. (c) Comparison of the wear volume. The p-values in the figures are from comparison between two cases: pristine vs. heat-only and heat-only vs. steam-treated for the 40% humidity case and pristine vs. heat-only for the 90% humidity case...... 52

Figure 3-7. (a) SR-IR spectra of SLS glass before and after hydrothermal treatment. The treatment conditions is at 150 C for 24 hours, 48 hours and 72 hours; (b) ATR-IR spectra of SLS glass before and after hydrothermal treatment for 12 hours at various temperatures; (c) Raw data and fitted curve for determining the activation energy of diffusion of water and formation of silanol groups into the glass network...... 54

Figure 3-8. SFG spectra of SLS glass before and after hydrothermal treatment at 150 C for 24 hours, 48 hours and 72 hours respectively...... 58

Figure 4-1. (a) Schematic experimental set-up of thermal poling experiment. (b) Graphical illustration of typical poling conditions with respect to the temporal profiles of temperature (T), bias voltage (V), and measured current (A)...... 66

xii

Figure 4-2. SR-IR spectra of the cathode-side glass surface before and after cleaning with a cotton tip wet with water. Inset shows an AFM image of sodium carbonate (image size = 25m  25m; height full scale = 650nm; poling time = 20 min)...... 67

Figure 4-3. (a), (b), (c): EDX line profiles of cross-sections of pristine and poled surfaces. Poling time = 40 min...... 70

Figure 4-4. Ellipsometric spectra (in  and ) and model fits for (a) pristine, (b) anode- side, and (c) cathode-side surfaces. The poling time was 40 min and the incidence angle of SE analysis was 70o. The lines are least squares fit results. All un-weighted error functions were below 2 x 10-3, indicating good agreement between experimental data and the least squares model fit. (d) Refractive index of bulk glass, pristine surface layer, and Na+-depleted layer as a function of photon energy. The refractive index of the cathode-side surface was almost identical to the bulk value...... 71

Figure 4-5. Schematics of ellipsometric model for (a) anode (sodium-depleted surface); (b) pristine surface; (c) cathode (sodium-gradient surface)...... 72

Figure 4-6. Topography (a,c,e) and adhesion force (b,d,f) maps obtained with peak-force tapping AFM imaging of (a,b) pristine, (c,d) anode-side, and (e,f) cathode-side surfaces. Poling time =10 min. The sampling size is 600nm  600nm...... 73

Figure 4-7. SR-IR spectra of thermally-poled soda lime glass surfaces. For comparison purpose, spectra of cathode-side, pristine, and anode-side surfaces as well as fused quartz are stacked with offsets...... 75

Figure 4-8. SR-IR spectra of the poled (solid lines) and then reverse-poled (dashed lines) glass surfaces. Both poling and reverse-poling times were 10 min...... 77

Figure 4-9. ATR-IR spectra of (a) Na+-depleted and (b) Na+-gradient surfaces of thermally-poled soda lime glass...... 78

Figure 4-10. SFG spectra of (a) Na+-depleted and (b) Na+-gradient surfaces of thermally- poled soda lime glass...... 80

Figure 4-11. (a,b) ATR-IR and (c,d) SFG spectra of Na+-depleted (a,c) and Na+-gradient surfaces of thermally-poled soda lime glass before and after post-annealing at 200oC for 2 hours in ambient air...... 82

Figure 4-12. Characteristic line profiles of wear track of Na+-depleted surface, Na+- gradient surface and pristine SLS glass substrate when rubbed with a pyrex ball for 400 cycles under an applied load of 0.2 N in (a) 40% RH (b) 90% RH conditions. (c) The overall wear volume of treated SLS glass substrate in humidity conditions...... 87

Figure 5-1. Schematics of thermal poling of SLS glass with non-blocking electrodes in controlled gas environment...... 95

Figure 5-2. EDX cross-section profiles, from the external surface, of the anode surfaces treated by thermal poling in (a) Ar, (b) N2, (c) N2 + 20% O2, (d) N2 + 1.3% H2O and

xiii

(e) N2 + 2.9% H2O. The dash lines in the graphs represent the atomic ratios in the bulk of pristine glass. (f) Comparison of the thickness of sodium-depleted layer determined by EDX and impedance spectroscopy...... 99

Figure 5-3. Comparison of impedance spectra of SLS glass before and after thermal poling treatment in Ar, N2 + 20% O2, N2, N2 + 1.3% H2O and N2 + 2.9% H2O. The inset magnifies the high frequency region. The AC frequency is the highest near the origin and decreases as the data move away from the origin...... 100

Figure 5-4. SR-IR spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O...... 102

Figure 5-5. (a) Three-dimensional representation of the SLS glass surfaces thermally poled in Ar and N2. (b) Statistics on the number of craters and depth of those craters. ... 103

Figure 5-6. Raman spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O...... 104

Figure 5-7. (a) ATR-IR spectra (measured with Ge crystal) spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O. (b) ATR-IR spectra (measured with diamond crystal) of the sodium-depleted surfaces formed by poling in N2 and Ar. In each gas condition, one sample was retrieved directly to ambient air o after thermal poling and the other was exposed to D2O vapor at 25 C after the thermal poling treatment and then retrieved to ambient air. In (b), the diamond ATR was used since it makes easier to identify the HOH band due to a flatter baseline than the Ge ATR...... 106

Figure 5-8. (a) SFG spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O. Note that the SFG spectra of pristine, N2, and N2+20%O2 cases are magnified by the factors marked in the figure for comparison. The sharp peaks between 2800 cm-1 and 2950 cm-1 are from adventitious organic contaminants. (b) SFG spectra of sodium-depleted surfaces poled in dry Ar. The red data is the sample retrieved to ambient air right after poling and the blue one is the sample exposed to saturated D2O vapor spectra at room temperature without electric field after poling and then retrieved to ambient air. The dotted line is the simulated OD spectrum assuming all SFG-active OH species in the red data are fully deuteriated and become SFG-active OD species.218 ...... 109

Figure 5-9. Hardness (a) and reduced elastic modulus (b) of sodium-depleted surfaces compared with pristine SLS glass with nanoindentation analysis...... 112

Figure 5-10. Optical images of Vickers indent of (a) pristine SLS glass surface; sodium- depleted surfaces prepared in (b) Ar, (c) N2, (d) N2 + 20% O2, (e) N2 + 1.3% H2O, (f) N2 + 2.9% H2O. The applied normal load is 300gf. The relative humidity is about 40% during the indentation analysis...... 114

xiv

Figure 6-1. EDX cross section profile of Na/Si, K/Si, Mg/Si and Ca/Si ratios of (a) Na+/K+-exchanged and (b) reverse-exchanged SLS surface; (c) Na/Si, K/Si, and Ca/Si ratios determined from XPS analysis of the pristine, Na+/K+-exchanged and reverse-exchanged SLS surfaces...... 123

Figure 6-2. (a) Hardness and (b) reduced modulus, measured at indentation depths of 100 nm, 150 nm and 200 nm, of the pristine, Na+/K+-exchanged and reverse-exchanged SLS surfaces...... 124

Figure 6-3. Optical images of Vickers indentation imprint at a 500 gf normal load at 0%, 40%, and 90% RH conditions on (a) pristine, (b) Na+/K+-exchanged, and (c) reverse- exchanged SLS surfaces. The dwell time at the maximum load was 15 seconds...... 125

Figure 6-4. Weibull plot of crack initiation loads measured with Hertzian indentation on pristine (black), Na+/K+-exchanged (red), and reverse-exchanged (blue) SLS surfaces in (a) 0%, (b) 40%, and (c) 90% RH conditions...... 126

Figure 6-5. Cross-section line profiles of wear tracks produced on the pristine, Na+/K+- exchanged SLS glass, and reverse-exchanged SLS surfaces after tribo-testing in (a) 40% RH and (b) 90% RH condition...... 128

Figure 6-6. (a) SR-IR, (b) ATR-IR, and (c) SFG spectra of the pristine (black), Na+/K+- exchanged (red), and reverse-exchanged SLS surfaces. In (c), the SFG spectrum of the pristine glass after heating to 400 oC in air is also shown for comparison...... 130

Figure 6-S1. (a) Depth profile of modifier ions, (b) hardness and modulus measured with nanoindentation, (c) humidity dependence of mechanochemical wear of the pristine and Na+/Ag+-exchanged SLS glass...... 134

Figure 7-1. (a) Schematic representation of SR-IR analysis of glass. (b) SR-IR of soda lime silica (SLS) glass spectra collected at an incidence angle of 10o and 45o. (c) Complex refractive index (n and k) values of SLS glass obtained using the Kramers- Kronig transformation algorithm of the built-in software of a FTIR instrument...... 138

Figure 7-2. Flow diagram of the mathematical algorithm to obtain n + ik from SR-IR spectra collected at two incidence angles (10o and 45o). The 3D plot in the left is the correction factor, C = R(i = 0o) / R(i = 10o), for all possible combinations of n and k values. Note that this value is close to the unity...... 141

Figure 7-3. (a) Theoretical SR-IR spectra at an incidence angle of 10o and 45o calculated using the literature value of refractive index of SLS glass in Rubin’s work 46 (shown in the inset). (b) Comparison of the experimental 푅(45o) value with the calculated reflectance, 푔(푛), in equation (11). The cross-point is the solution for 푛. (c) Errors for the (n1, k1) and (n3, k3) values calculated from the two-angle SR-IR algorithm...... 144

incidence angle to 45 and with a reduced beam size (aperture = 50%, instead of 100% full opening) to reduce the divergence of incidence angle...... 145

Figure 7-5. (a)  and  spectra at 푖 = 55o simulated using the known n + ik values for ISG glass from Ref. 16. (b) point-by-point calculations and (c) fitting of the simulated  and  spectra (dashed lines). The symbols are the values calculated from the simulated spectra and the lines are the original data used for the simulated spectra...... 147

Figure 7-6. Comparison of the two-angle SR-IR method (a, b) and the SE method (c, d). The raw reflectance spectra of the ISG glass are shown in (a). The  and  spectra of the ISG glass are displayed in (c). The n() and k() spectra of the ISG glass calculated from the two-angle SR-IR and SE methods are shown in (b) and (d), respectively...... 149

Figure 7-7. Comparison of the n()[black] and k()[red] spectra of (a) fused quartz, (b) annealed 700 m thick SLS float glass (from Asahi), (c) annealed 4mm thick SLS float glass (from PPG), (d) thermally-tempered 4mm thick SLS float glass (from PPG), (e) borosilicate glass (BOROFLOAT®33, from Schott), (f) barium boroaluminosilicate glass (AF45, from Schott), (g) alkaline-earth boroaluminosilicate glass (Willow®, from Corning), (h) sodium calcium boroaluminosilicate glass with a trace of Zr (ISG, from MoSCI). The 푅() spectra collected at 10o and 45o incidence angles are shown in the Supporting Information. The data files of all spectra shown in this figure are available in the Excel format in the Supporting Information...... 152

Figure 7-8. Four stretch modes of an ideal SiO4 unit with the tetrahedral (Td) symmetry. ... 154

Figure 7-S1. (a) Schematic diagram of the variable-angle reflection accessory. (b) Picture of the optics inside the SeagullTM unit. The sample mounting unit is removed to show the hole that aligns the incidence angle to 45o. (c) Picture of top and back side of the SeagullTM unit...... 161

Figure 7-S2. Comparison of the experimentally-obtained SR-IR spectra of fused quartz at 10o and 45o incidence angles and the simulated spectra using the 푛() and 푘() value from ref. 283The small difference is likely due to the difference in samples and experimental systems used in this study and ref. 283 Errors due to misalignment and divergence of the incidence IR beam...... 162

Figure 7-S3. SR-IR spectra collected (a) at 푖=10o and 42o with the 100% opening of the IR aperture and (b) at 푖=10o and 45o with the 50% opening of the IR aperture. The 푛() and 푘() spectra are calculated using the spectra in (a) and the TASR-IR algorithm shown in Figure 7-2; these spectra are different from the ones calculated with the spectra in (b). The spectra in (b) and (d) are the same ones in Figure 7-4(a) and 7-4(c)...... 163

Figure 7-S4. Comparison of 푛() and 푘() spectra of ISG obtained from TASR-IR and IR-VASE methods...... 164

xvi

Figure 7-S5. Comparison of the experimentally-obtained 푅(30표) spectrum with the cimulated spectrum using the 푛() and 푘() spectra obtained with (a) TASR-IR and (b) IR-VASE method. The data used in this calculation are shown in Figure 7- S4...... 165

Figure 7-S6. Raw SR-IR spectra of the glasses used for calculation of 푛() and 푘() spectra shown in Figure 7-7...... 166

Figure 7-S7. IR penetration depth (dp) calculated using the 푛() and 푘() spectra shown in Figure 7 for an incidence angle of 10o...... 167

Figure 8-1. Compilation of stress effects on the maximum intensity position of Si-O-Si,as stretch band of silica and soda lime silica glass. The silica glass data (open symbols) were taken from ref. 15 ( ), 16 ( ) where uniaxial stress was applied along a fiber or bending a fiber. The compressive stress data of SLS glass surface was obtained by SR-IR analysis of ion-exchanged glasses and thermally-tempered glass. The stress of SLS was estimated based on the literature: ref. 24, 26 for Na+/Ag+ exchange and ref. 23, 25, 27 for Na+/K+ exchange.9,294–297 The tensile stress data of SLS were measured on the convex surface of a slide glass under three-point bending...... 171

Figure 8-2. Comparison between experimentally measured values and calculated values based on simulated glass (a) real part of the refractive index, n; (b) imaginary part of the refractive index, k; (c) reflectance spectra. The experimental data were reproduced from Ref. 261...... 176

Figure 8-3. (a) Experimentally measured SR-IR spectra of fused quartz at 300K, 473K and 673K; (b) Calculated SR-IR spectra based on the simulated 푛(휔) and 푘(휔) at 300K, 500K and 700K; (c) Simulated real part of the refractive index (푛(휔)) at 300K, 500K, 700K; (d) Simulated imaginary part of the refractive index (푘(휔)) at 300K, 500K, 700K. Note that in (a), the experimental setup could not detect the spectrum below 600 cm-1...... 178

Figure 8-4. Stress effect on (a) SR-IR spectral features calculated from the simulated (b) 푛(휔) and (c) 푘(휔) of the refractive index. The calculations were done for the silica system at 300 K with a 5% uniaxial strain along the applied stress axis which was achieved at 3.2 GPa and 3.6 GPa under compressive and tensile stress, respectively. Strains normal to the directions normal to the stress axis varied following the simulated Poisson’s ratio...... 179

Figure 8-5. (a) Si-O bond length distribution at 300K, 500K and 700K; (b) Si-O-Si and O-Si-O bond angle distribution at 300K, 500K, 700K; (c) Stress effect on Si-O bond length distribution; (d) Stress effect on Si-O-Si and O-Si-O bond angle distribution. .... 180

Figure 8-6. Plots of the weighted mean of the calculated Si-O-Si,as stretch band in k as a function of (a) weighted mean of Si-O bond length, (b) weighted mean of Si-O-Si bond angle, (c) weighted mean of O-Si-O bond angle, and (d) density...... 182

xvii

Figure 8-7. Plots of the weighted mean of the calculated Si-O-Si,as stretch band in k as a function of (a) weighted mean of Si-O bond length, (b) weighted mean of Si-O-Si bond angle, (c) weighted mean of O-Si-O bond angle, and (d) density...... 185

Figure 8-S1. SR-IR spectra of a pristine SLS glass and a convex surface of slide glass under three-point bending. The tensile stress of the convex surface was estimated using the beam bending theory from the deformation, the sample dimension, and the modulus of the slide glass. The lower intensity for sample under tensile stress was due to the light scattering from the slightly curved surface...... 188

Figure 8-S2. SR-IR spectra of a tempered glass (quenched in air) and the same type of glass that has been annealed (slow cool in air)...... 189

Figure 8-S3. (a) Number of defects as a function of temperature during quenching from 6000K; (b) density changes during quenching of glass with the BKS force field; (c) density as a function of temperature simulated with the FFSiOH force field; (d) thermal expansion coefficient as a function of temperature calculated with the BKS force field; Poisson’s ratio and elastic modulus during (e) compression and (f) tension at 300K calculated with FFSiOH force field...... 190

Figure 8-S4. (a) Maximum intensity position of imaginary part (k) of the simulated refractive index versus weighted mean of bond parameters; (b) maximum intensity position of simulated reflectance versus weighted mean of bond parameters; (c) weighted mean of simulated reflectance versus weighted mean of bond parameters...... 191

Figure 8-S5. (a) Elastic modulus and (b) hardness of the SLS glass with or without applied tensile stress. The p-values shown in the figure are from student’s t-test with N = 30 data set...... 192

Figure 9-1. Schematic illustration of depth profiles of H and modifier ions in the surface region of SLS float glass and the probe depth of SFG and ATR-IR. Note that the schematic is not drawn in scale...... 195

Figure 9-2. (a) Schematics of ssp-polarization SFG analysis of a glass surface as a function of temperature in a controlled vapor environment; (b) Schematics of monitoring the intensity of two SFG peak intensities at the same time during the temperature ramping at a constant rate ()...... 201

Figure 9-3. SFG spectra taken at room temperature, at 200C, and after cooling back to room temperature for (a) 0.7 mm thick SLS float, (b) 1mm thick SLS float, (c) fused quartz, and (d) BF33 glasses...... 203

Figure 9-4. TP-SFG analysis for (a) 3700 cm-1 peak of fused quartz (b) 3544 cm-1 peak of 0.7 mm thick SLS float glass. The symbols are experimental data and the red lines are the non-linear fit with equation (7)...... 207

Figure 9-5. Summary of the desroption energy of hydrous species responsible for different OH SFG peaks detected for (a) 0.7 mm thick SLS float, (b) 1 mm thick SLS float, (c) fused quartz, and (d) BF33 glasses...... 209

xviii

Figure 9-6. (a) SFG spectra of 1mm thick SLS float glass exposed to D2O vapor (ii) and comparison with the SFG spectra before (i) and after (ii) D2O exposure. The relative partial pressure of D2O in (ii) was 90 % in N2 and that of H2O in (i) and (iii) was 30 % in air. The bottom spectrum is the subtraction of (i) from (ii); the dotted line represents the sharp SFG spectral features obtained by shifting the OH features using theoretical OH/OD peak positions. (b) SFG spectra of 1mm thick SLS float o glass heated to 200 C and then cooled back to room temperature in D2O vapor; the spectra were collected in ambient conditions. (c) ATR-IR spectra of 1mm thick SLS float glass collected before heating to 200 oC, at 200 oC, and after cooling back from 200 oC to room temperature...... 211

Figure 9-7. Effect of heating above Tg on the surface structure of SLS float glass. (a) SFG, (b) ATR-IR, and (c) SR-IR spectra collected before and after heat treatment at 600 oC. (d) Modifier ion (Na+, Ca2+, Mg2+) concentrations measured with XPS before and after heat treatment at 600 oC. (e, f) Changes in O1s XPS peak shape before (e) and after (f) heat treatment at 600 oC...... 212

Figure 9-8. (a) Relation between vibrational wavenumber and O-HO distance in hydrogen bonding regenerated from literature (b) & (c) Square root of SFG signal for 0.7 mm SLS float glass and for 1mm SLS float glass respectively as a function of O-HO distance...... 214

Figure 9-S1. SFG spectra of pristine 1mm SLS float glass with and without CCl4 on the back side...... 217

Figure 9-S2. Anti-Stokes Raman signal of 1mm SLS float glass before and after heating above Tg. This signal is detected only in the 532 nm reflection direction. It does not overlap with the SFG signal direction when the IR incident angle is different from the 532 nm incident angle...... 218

Figure 9-S3. SR-IR spectra of 1mm thick fused quartz before and after heating at 600 oC. ... 219

xix

LIST OF TABLES

Table 4-1. Position, width, and height of OH stretch peaks observed in SFG spectra ...... 90

Table 5-1 Summary of chemical, structural, and mechanical changes of the sodium- depleted surface formed by thermal poling (200 oC, 2 kV, 20 minutes) of SLS float glass (thickness = 1 mm) in various gas environments...... 116

Table 7-1. Deconvoluted peaks of the 푘() spectrum of silica and soda lime silicate glasses. The data shown in the order of peak position (cm-1)/peak intensity/ width (cm-1) of Gaussian fit function. Note that the Gaussian function was used in the peak deconvolution only because of its simplicity...... 153

Table 7-2. Deconvoluted peaks of the 푘() spectrum of borosilicate and boroaluminosilicate glasses. The data shown in the order of peak position (cm- 1)/peak intensity/ width (cm-1) of Gaussian fit function. Note that the Gaussian function was used in the peak deconvolution because of its simplicity...... 153

Table 8-S1. Physical properties of fused quartz at 300 K. These values are for the fused quartz substrate from Technical Glass Products used in the experimental part of this study ...... 191

Publications

(1) Luo, J.; Smith, N.; Pantano, C. G.; and Kim, S. H., Complex refractive index of silica, silicate, borosilicate, and boroalumunosilicate glasses -Analysis of glass network vibration modes with specular-reflection IR spectroscopy. (submitted) (2) Luo, J.; Zhou, Y.; Pantano, C. G.; Kim, S. H., Fictive temperature effect on the specular reflectance infrared spectra- a molecular dynamics study. (submitted) (3) Luo, J.; Amma, S.; Chen, L.; Ngo, D.; Pantano, C. G.; Kim, S. H., Determining the relative amount of SiOH/H2O in soda lime silica glass surface by using ATR-IR and hydrogen depth profiles. (in progress) (4) Luo, J.; Grisales, W.; Rabii, M.; Pantano, C. G.; Kim, S. H., Effects of Na+/K+-ion exchange on mechanical and mechanochemical properties of soda lime silica glass. (in progress) (5) Sheth, N.; Luo, J.; Pantano, C. G.; Kim, S. H., Surface structural, mechanical and mechanochemical behavior of leached soda lime silica glass. (in progress) (6) Xiao, C.; Yao, Y.; Luo, J.; Ngo, D.; Kim, S. H.; Qian, L.; Chen, L., Evolution of surface wettability of monocrystalline silicon exposed in liquid alcohol, humid air and water. (in progress) (7) Luo, J.; Bae, S.; Yuan, M.; Schneider, E.; Lanagan, M. T.; Pantano, C. G.; Kim, S. H. Chemical Structure and Mechanical Properties of Soda Lime Silica Glass Surfaces Treated by Thermal Poling in Inert and Reactive Ambient Gases. J. Am. Ceram. Soc. 2018. doi: 10.1111/jace.15476 (8) Luo, J.; Zhou, Y.; Milner, S. T.; Pantano, C. G.; Kim, S. H. Molecular Dynamics Study of Correlations between IR Peak Position and Bond Parameters of Silica and Silicate Glasses: Effects of Temperature and Stress. J. Am. Ceram. Soc. 2018, 101 (1), 178–188. (9) Amma, S.; Luo, J.; Kim, S. H.; Pantano, C. G. Effect of Glass Composition on the Hardness of Surface Layers on Aluminosilicate Glasses Formed through Reaction with Strong Acid. J. Am. Ceram. Soc. 2018, 101 (2), 657–665. (10) Sheth, N.; Luo, J.; Banerjee, J.; Pantano, C. G.; Kim, S. H. Characterization of Surface Structures of Dealkalized Soda Lime Silica Glass Using X-Ray Photoelectron, Specular Reflection Infrared, Attenuated Total Reflection Infrared and Sum Frequency Generation Spectroscopies. J. Non. Cryst. Solids 2017, 474 (August), 24–31. (11) Zhang, Y.; Vulfson, Y.; Zheng, Q.; Luo, J.; Kim, S. H.; Yue, Y. Impact of Fiberizing Method on Physical Properties of Glass Wool Fibers. J. Non. Cryst. Solids 2017, 476 (September), 122–127. (12) Amma, S.; Luo, J.; Kim, S. H.; Pantano, C. G. Effects of Fictive Temperature on the Leaching of Soda Lime Silica Glass Surfaces. J. Am. Ceram. Soc. 2017, 100 (4), 1424– 1431. (13) Chae, I.; Ahmed, S.; Atitallah, H. Ben; Luo, J.; Wang, Q.; Ounaies, Z.; Kim, S. H. Vibrational Sum Frequency Generation (SFG) Analysis of Ferroelectric Response of PVDF-Based Copolymer and Terpolymer. Macromolecules 2017, 50 (7), 2838–2844. (14) Luo, J.; Huynh, H.; Pantano, C. G.; Kim, S. H. Hydrothermal Reactions of Soda Lime Silica Glass – Revealing Subsurface Damage and Alteration of Mechanical Properties and Chemical Structure of Glass Surfaces. J. Non. Cryst. Solids 2016, 452, 93–101. (15) Luo, J.; Banerjee, J.; Pantano, C. G.; Kim, S. H. Vibrational Sum Frequency Generation Spectroscopy Study of Hydrous Species in Soda Lime Silica Float Glass. Langmuir 2016, 32 (24), 6035–6045.

xxi

(16) Luo, J.; He, H.; Podraza, N. J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 1: Effects of Sodium Ion Migration and Water Ingress on Glass Surface Structure. J. Am. Ceram. Soc. 2016, 99 (4), 1221–1230. (17) He, H.; Luo, J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 2: Effects on Mechanical and Mechanochemical Properties. J. Am. Ceram. Soc. 2016, 99 (4), 1231–1238. (18) Barthel, A. J.; Luo, J.; Hwang, K. S.; Lee, J.-Y.; Kim, S. H. Boundary Lubrication Effect of Organic Residue Left on Surface after Evaporation of Organic Cleaning Solvent. Wear 2016, 350–351, 21–26. (19) Amma, S.; Luo, J.; Pantano, C. G.; Kim, S. H. Specular Reflectance (SR) and Attenuated Total Reflectance (ATR) Infrared (IR) Spectroscopy of Transparent Flat Glass Surfaces: A Case Study for Soda Lime Float Glass. J. Non. Cryst. Solids 2015, 428, 189–196. (20) Alazizi, A.; Barthel, A. J.; Surdyka, N. D.; Luo, J.; Kim, S. H. Vapors in the ambient—A Complication in Tribological Studies or an Engineering Solution of Tribological Problems? Friction 2015, 3 (2), 85–114. (21) Barthel, A. J.; Luo, J.; Kim, S. H. Origin of Ultra-Low Friction of Boric Acid: Role of Vapor Adsorption. Tribol. Lett. 2015, 58 (40).

xxii

Preface

Portions of this dissertation are adapted from a collection of previously published work used with the permissions from each of the manuscript holders. The author’s contributions to multiple-authored work are summarized as follows. Chapter 2: calculation of SR-IR and ATR-IR spectra and comparison with experimental results. Chapter 3-6: all the surface structure and mechanical properties analysis. Chapter 7: methodology development and SR-IR. Chapter 8: summary of experimental SR-IR results in comparison with MD simulation. Chapter 9: SFG and

IR analysis. Chapter 1 and 6 are original and not taken from previously published work.

xxiii

ACKNOWLEDGEMENTS

I am deeply grateful for all of my fellow advisers, collaborators, mentors, friends and family who helped me during the journey towards Ph.D. since Aug. of 2012. I would be lying if I say this journey is easy and without sacrifice. But I would never regret every decision I made during this journey. They help me to be honest with myself, understand myself and prepare myself for the upcoming journeys in life. I am confident that I have become a better man.

First, I will give my special thanks to my advisors. It is my great honor to have two excellent advisors: Dr. Seong Kim and Dr. Carlo Pantano during my PhD. I am sure that one would learn enough if you could have one of them as your advisor. I am fortunate enough to have both of them at the same time. Their knowledge, guidance, life lessons during the meetings are the treasures of my life. Dr. Kim is always frank with me and help me like a father. Dr. Pantno is so insightful and a model person that many people want to become. They show me how to be successful not only in their careers, but also in their family.

Second, I would like to thank my collaborators and labmates during my study. It has been a pleasure to work with so many wonderful minds. For my last couple years, the weekly Friday night gathering with Dr. Yuxing Zhou, Dr. Wenlin Zhang, Xin Qi and Xin He has been a combination of pleasure and sparkles of ideas with the help of a magic chemical. The collaboration between Dr. Zhou and I starts from one of the conversation in this gathering. We are working on our third collaborative publication as of now. Dr. Hongtu He, Dr. Shin-ichi

Amma and Dr. Lei Chen are three brilliant visiting scholars in our group. I am inspired during the productive collaboration with them. Dr. Christopher Lee and Dr. Anthony Barthel have been model labmates for me. They are both persons with character and did wonderful work for their

xxiv

PhD. I am also very lucky that all my labmates are nice people. Every one helps each other and learns from each other. I guess this is also the secret that the group has been so successful.

Third, my special thanks to the undergraduate students who worked with me on different projects during my PhD. They are Brennan Heiser, David Belias, William Grisales, Erik

Schneider, Connor Maust, Hoang Huynh, Trisca Ng, Matthew Rabii, Mohaammed Alshehri and

Maujouri Sau. I really enjoy the experience and opportunities with these inspired young people.

Their curiosity, grit and desire to learn have given me the motivation to work hard. In my mind, they are not undergraduate researchers but teachers to show me the unknown world.

Last but not least, my family members’ support is the harbor of my heart. Their sacrifice and efforts give me the strength to finish the PhD. I miss them so much and wish I could do something more for them. This dissertation is for my grandma who just went to another world to enjoy the life she should have three weeks ago.

Chapter 1

Introduction

Motivation

Soda lime silica (SLS) glass has been widely applied in construction, automobile and containers. One unsatisfactory feature of SLS glass is its brittleness. In other words, SLS glass is easy to break due to applied stress in ambient conditions. Previous studies suggest that this brittleness is related to the stress-induced reactions between glass and water molecules in the environment, or stress-corrosion reactions. These studies are typically carried out to monitor the velocity of crack propagation of the crack tips under applied tensile stress. However, very little is known on how the glass material are removed upon shear stress in humid environment, or wear behavior of glass materials. It should be noted that the applied normal stress (~300 MPa) alone in this study will not cause plastic deformation to SLS glass. Effect of shear and its induced reactions are the major reasons for materials removal. Under these conditions, the wear behavior of glass materials are found to be not only affected by the applied normal and shear stress, but also highly affected by the shear-induced hydrolysis reactions on glass surfaces.

Recently, SLS glass has been found to behave uniquely upon shear stress at high relative humidity (RH) conditions -- little wear are observed on the SLS glass substrate while significant amount of material is removed on the counter surface (borosilicate glass, stainless steel, silicon nitride, sapphire,etc). For most of the other commercial available flat glasses including fused quartz, AF45, BF33, sodium-aluminosilicate and potassium-exchanged aluminosilicate glasses, more glass materials will be removed as the amount of water in the gas phase increases given the same applied normal stress. Two possible mechanisms could contribute to this unique wear

2 behavior for SLS glass: 1. the shear-induced hydrolysis reactions are “suppressed” at high RH condition; 2. Additional chemical reactions take place with the assistance of interfacial shear, which compete with the hydrolysis reactions. For both cases, surface mechanical properties which affect the stress upon shear, the surface chemistry of SLS glass and dynamic interaction among glass surface and water molecules at high humidity conditions all play vital roles.

In this dissertation, factors that could govern the shear-induced hydrolysis reactions on

SLS glass surface region at high RH conditions have been investigated. These factors including

+ + ion-exchange process between Na and H + H2O, modifications of mechanical properties like hardness and elastic modulus, Na+ depletion/accumulation in the surface region, are studied by modifying the chemical structure of SLS glass in the surface region. These modifications are achieved by performing hydrothermal treatment, leaching in acid solutions, thermal poling treatment, ion-exchange treatment in molten salt (KNO3, AgNO3). The major accomplishments of this dissertation are as follows: 1. The unique wear behavior is not due to the ion-exchange

+ + process between Na and H + H2O in the surface region (Chapter 3, 4, 5) and not due to the formation of surface compressive stress from ion-exchange process or modification of surface elastic modulus and hardness (Chapter 5, 6); 2. The unique wear behavior is best correlated with the high surface concentration of mobile Na+ and unique surface silicate network of SLS glass

(Chapter 4, 6, Appendix A). It is speculated that Na+ can lower the activation energy of condensation of silanol groups to form BO which is the reverse reaction of hydrolysis; 3. To better understand the unique silicate network structure of SLS glass, several novel approaches have been developed. These novel approaches could experimentally reveal non-equilibrium glass structure features, including distribution of bond Si-O bond length from infrared spectroscopy and molecular dynamics (MD) simulation (Chapter 7&8), distribution of O-HO distance from sum frequency generation (SFG) spectroscopy (Chapter 9) and SiOH/H2O ratio from attenuated total reflectance infrared (ATR-IR) spectroscopy (Appendix B) in the surface region of glass

3 materials. Theoretical description of infrared spectroscopy applied on SLS glass surface has also been summarized for appropriate data interpretation (Chapter 2).

Understand shear-induced hydrolysis reactions of SLS glass at high RH by controlling and analyzing the surface chemical structure and mechanical properties

Glass materials can be removed by “softer” materials through interfacial shear in humid environment. For example, the Gorilla glass used on cellphones which has higher hardness than most of daily objects, can still be “scratched” by the coins or keys in our pockets. It takes hundreds of cycles before these scratches are observable to our eyes, just like the wear test in this study.

Shear-induced hydrolysis reactions of SLS glass studied in this dissertation could be governed by the following factors: surface composition, silicate network structure, surface mechanical properties (hardness and elastic modulus), surface residual stress, etc. It should be noted that surface mechanical properties and residual stress could be very sensitive to the surface structural modification. For example, surface composition and structural modification through ion-exchange

+ between Na and other cations with different ionic radii below Tg could change the surface mechanical properties of glasses. To identify the real governing factors, surface chemical structure and mechanical properties need to be characterized for controlled surface treatments.

There are several important features that makes SLS glass surface unique among silica and silicate glasses: 1. Relative high leachable Na+ concentration in the surface region; 2. Amorphous silicate network with high concentration of non-bridging oxygens (NBO) which are associated with

+ 2+ Na and Ca ; 3. Relative high concentration of hydrous species (SiOH & H2O) in the surface region. There are always concentration gradients for these three structural features from surface to the bulk. What makes this study even more challenging is that these three features cannot be isolated easily. This is because introduction of hydrous species are always accompanied with the

4 partial depletion of Na+ and modification of silicate network. For these reasons, various surface sensitive techniques need to be used to probe these structural features and their impact on the shear- induced hydrolysis reactions. Researchers have attempted to relate properties of glass materials including refractive index, water diffusion into the glass, hardness, fictive temperature, bond parameters, etc with vibrational features of silica and silicate network.1,2 For convenience, peak position shifts of spectra features in infrared spectroscopy such as Si-O-Si asymmetric stretch, OH stretch, are commonly used to correlate with the changes of these properties. However, these correlations fail to describe the non-equilibrium nature of glass inevitably since they did not provide the distribution of those structural features. Therefore, efforts are also needed to develop new methodology and new interpretation of these vibrational features. In this dissertation, new methodologies based on SR-IR and SFG have been developed to describe the amorphous structure of SLS glass surface.

The role of surface mechanical properties need to be investigated since they directly affect the surface stress upon shear, which is the driving force for hydrolysis reactions. Resistance to crack formation and plastic deformation, which are best described by surface compressive stress and hardness, also play important roles in the material removal process. Nanoindentation has been developed as a powerful tool to probe the surface hardness and elastic modulus of glass materials from Oliver and Pharr’s method.3 Surface compressive stress, on the other hand, is still hard to measure directly on a small sample. Indirect measurement like Vickers indentation, can provide some qualitative information on the surface residual stress. It should also be noted that the crack formation upon indentation like Vickers are also affected by the humidity levels in the environment.

Therefore, environmental control is critical during the indentation measurements in order to have better comparison among samples. Environmentally controlled indentation analysis has been developed in this dissertation.

5 After combining and summarizing the changes in surface chemical structure and mechanical properties from various controlled surface treatments, the dominated factors (surface mobile Na+ and silicate network) are then identified. This finding could also provide new insights in designing glass materials with desired surface properties that hinder the materials’ removal and failure.

Background and literature review

Unique surface chemical structure of SLS glass

Before describing the surface chemical structure of SLS glass, it is necessary to address the bulk structure of silica and silicate glasses briefly. Unlike crystalline materials, the distribution of bond parameters (Si-O bond length, Si-O-Si bond angle, etc) is much broader in glass materials.

For example, neutron scattering shows that crystalline SiO2 and a SiO2 glass’s Si-O bond length are both centered ~1.6 Å. However, the distribution of crystalline SiO2 is extremely sharp (close to a step function ~1.61 Å) while the Si-O bond length in silica glass (centered ~1.62 Å) can be as low as ~1.0 Å and as high as ~2.3 Å.4 The bond angles (Si-O-Si, O-Si-O) distribution for silica glass is also very broad accompanied with wide distribution of ring sizes. For a simple sodium silicate glass with similar modifier contents as SLS glass, the Si-O bond length is centered ~1.63

Å.5 While no experimental EXAFS and neutron scattering data is available for SLS glass, the bond parameters of silicate network are believed to be similar to sodium silicate glasses which have similar modifier ion content. The introduction of network modifiers such as Na+, Ca2+ will lower the connectivity of the glass network. The most accepted model for silicate glass like SLS glass is the “modified random network” (MRN) model proposed by Greaves.6 This model regard Si-O-Si network as the backbone of the glass while the modifiers are associated with the non-bridging

6 oxygen (NBO). It is also believed that channels exist for modifiers like Na+ to migrate within the glass network, which is related to its electronic conductivity (Na+ migration within the glass network) and leaching of Na+ out of the glass when exposed to water.

The surface chemical structure of SLS glass is in general different from the bulk structure and can vary significantly depending on the history of the surface treatment. There are two major reasons: 1. the bridging oxygen at the glass/air interface will react and form SiOH groups which become active sites for adsorbed water molecules; 2. Na+/NBO sites have higher reactivity than the other structural units which result in preferential leaching of Na+ out of glass. When the most reactive species in the environment are water molecules, this leaching process is usually accompanied with the formation of SiOH and ingression of H2O into the surface region of SLS

glass. In ambient air, Na2CO3 and NaHCO3 can be formed on top of SLS glass through further

+ reactions between Na /NBO sites with H2O and CO2, which will make the transparent glass “hazy”.

To mitigate this effect and increase the durability of SLS glass, dealkalization treatment prior to distributing the SLS glass is normally performed. For example, SO2 dealkalization treatment around Tg is carried out for the SLS float glasses used in this dissertation. While the details of SO2 treatment cannot be obtained from the provider (Asahi Co.), some key structural features can be

+ found in the literature. The SO2 treated surfaces can have a Na gradient layer with the thickness of 50-150 nm where Na+ has lowest but non-zero concentration at the glass/air interface.7 The formation of the concentration gradient of alkali and alkaline earth ions can be attributed to the fact that rate of removing Na+ out of the glass is higher than the diffusion of Na+ from the bulk to the surface. Ca2+ and Mg2+ are also slight lower in the top surface. The NBO sites where Na+ are removed, are converted to SiOH. At the same time, H2O also migrate into and “trapped” in the silicate network. In addition, condensation of two SiOH groups could occur, which will increase the network connectivity, relax the silicate network and change the amount of H2O in the dealkalized layer. The variation of the layer thickness as well as the structure of modified structure is

7 related to the treatment time as well as the temperature profiles (near Tg) during the commercial

SO2 treatment. Na2SO3/Na2SO4 can also be formed as products which is washed away during the manufacturing process. Then the surface chemical structure of SLS glass can be considered as the reaction product of a series of non-equilibrium reactions in the surface treatment history.

An inter-diffusion model can be used to describe the concentration gradient profiles of H or Na+ in some conditions. The inter-diffusion was first developed by Doremus for describing the depth profile of H for leached SLS glass with nuclear reaction analysis (NRA).8 The core assumption for this model is defining an inter-diffusion coefficient combining the diffusion coefficient of both Na+ and counter ion (H+). The physical meaning of this term is that the diffusion of two Na+ and counter ions are dependent on each other. This model has been successful in describing the H depth profile in leached glasses as well as the K+ profiles after ion- exchange process.2,9 One indication of this model is that the diffusion of Na+ out of the glass will depend on the counter ions diffusivity into the glass network. While a modified version of this model can account for the structure changes as a function of Na+ concentration in the alteration layer of SLS glass, the dynamic change of glass structure during this reaction can sometimes be subtle to describe. A schematic describing the changes of the composition due to commercial treatment is presented in Figure 1-1.

While the profile of H can be described by model, the speciation and chemical environment of Si-OH and H2O in the surface region is not well resolved yet. This is mainly due to the complicated hydrogen-bonding interactions among SiOH and H2O. For non-destructive

10,11 method like IR, the near-IR (NIR) have been utilized to distinguish Si-OH and H2O. However, the interpretation of NIR is not as straightforward as is discussed in the literature, which can cause some disagreement in the interpretation of results. In mid-IR region, the molecular H2O can be identified by the bending vibration peak, however, Si-OH groups cannot be directly identified in the OH stretching region due to the complex hydrogen-bonding interaction. OH stretch vibration in solid materials is also found to be a function of OH-O distance. However, it is possible to determine the Si-OH/H2O ratio with proper deconvolution of OH stretching peaks. For destructive method like temperature program desorption (TPD) analysis, the amount of “weakly bonded” H2O, “strongly-bonded” H2O, and Si-OH can be distinguished based on their temperature of desorption.12–14 However, the dynamic transformation in the silicate network

9 during the heating process is not well accounted for. It can be expected that the vacancy will be created when H2O molecules are removed, the silicate network might relax and even reorganize.

The chemical environment of SiOH and H2O is also a reflection of glass network. Based on MRN model, local structural heterogeneity might exist in SLS glass. Na+ might accumulate locally and

+ form “clusters”. When these Na ions are replaced with hydrous species far below Tg (the Si-O-Si network will not change much), the hydrogen bonding interactions will be largely depending on the initial configuration of Na+ in the glass network. However, detecting such structural heterogeneities is not easy experimentally. A schematic describing the chemical environment of hydrous species in the surface region of SLS glass is shown in Figure 1-2.

Figure 1-2. Schematic of chemical environment of Na+ and hydrous species in the surface region of SLS glass.

More importantly, the alteration of silicate network structure in the surface region due to the reactions between SLS glass and H2O is not very well understood yet. The differences in the ionic radii between Na+ and hydrous species can introduce “stress” on the Si-O-Si backbone

+ assuming Si-O-Si backbone is “rigid” at temperature way below Tg. Assuming that the size of Na ,

H2O, O-H bond length determined from crystalline materials can be applied to these species in glass, the speciation of hydrous species and amount of hydrous species will have a significant impact on the sign of the stress (tensile or compressive) as well as the degree of stress. For example, surface silicate network should be under tensile strain is one Na+ exchanges with one H+ and one

H2O equivalently; but if multiple H2O molecules are introduced for one sodium cation, compressive strain for silicate network is expected.15–18 This “strain” on silicate network can be reflected in vibrational band of Si-O-Si network.19 SR-IR analysis shows that the Si-O-Si asymmetric stretching

+ + vibration shows blue shift or become more “silica like” when Na is replaced by H and H2O.

However, the interpretation of this peak shift from molecular level is still lacking. Proper interpretation of IR spectra could lead to better understanding of the surface silicate network structure, which can be correlated with the surface mechanical properties and shear-induced reactions.

In summary, the surface chemical structure of SLS glass can be considered as a product of SLS melt surface reacting with H2O and other species in the environment. It is featured with relative large amount of mobile Na+ while it is always partially depleted due to reactions, abundance of hydrous species (Si-OH & H2O) and altered amorphous silicate network. It should be noted that the “surface” described in this dissertation is the alteration layer created by various surface treatments, not just the interface between SLS glass and air.

Stress-induced reactions between SLS glass surface and H2O (g) in the environment

Stress-induced reactions is normally described as the reactions that are triggered and accelerated by stress at ambient conditions while these reactions usually require much higher thermal energy. For SLS glass surface, both applied stress like indentation, and surface residual tensile stress like strained silicate network can induce the breakage of Si-O-Si bonds through

11 reactions with water. Since H2O(g) needs to adsorb on the SLS glass or be absorbed by SLS glass to participate in this reaction, the amount of water in the environment or relative humidity (RH) also plays critical role in stress-induced reactions. The structure of adsorb/absorbed water layer at different RH levels are affected by the adsorption isotherm of water on these surfaces. At high

RH, multiple layers of adsorbed water or absorbed water molecules can form on glass surfaces. It is speculated that adsorbed water layers at high RH also have reactivity close to liquid water.

Here, the reactions between SLS and H2O(g) in the environment that are induced by normal and shear stress will be described in details.

Stress corrosion of SLS glass at ambient conditions

Stress corrosion is defined as the hydrolysis reactions between bridging oxygen (BO) and water molecules in the air, which result in the formation of silanol groups and breakage of the silicate network at ambient conditions or low temperature conditions (<200 C). Ciccotti made an excellent summary on stress corrosion mechanisms in silicate glasses. In brief, hydrolysis reactions of BO can be initiated and accelerated by the applied tensile stress on the glass network.

For SLS glass, three major reactions take place under applied tensile stress20: (1) penetration or diffusion of molecular water into the glass matrix; (2) hydrolysis reactions of BO to

+ + form two silanol groups; (3) ion-exchange between Na in NBO sites with H + H2O, which can result in the formation of Si-OH and introduction of molecular water. It should be noted that these reactions are considered to be reversible. The relationship between nature log of crack propagation velocity and stress intensity factor (KIC) can be divided into 3 regions by the magnitude of KIC as is shown in Figure 1-3.

Region I is considered to be “stress corrosion regime” where the relationship is almost linear. Several models have been proposed to describe this region. Wiederhorn’s model (shown

12 below in equation (1)) is most widely accepted for silicate glasses since it describes that the thermal activation energy will be lower linearly with the applied tensile stress (K is the stress intensity factor). Besides, this model also describes the linear dependence of log[velocity] with the water partial pressure/RH at given stress. The validity of this model has been demonstrated for different glass compositions as well as stress corrosion experiments at different humidity and temperature

(up to 200 C).21

퐼 푝퐻2푂 푚 ∆퐸푎−푏 퐾 푣 = 푣0 exp[훼 퐾] = 퐴 ( ) exp (− ) (1) 푝0 푅푇

Region II is normally described as the “transport-limited” region. The crack propagation velocity hinders for the medium range of applied tensile stress. This claim can be supported by the shortened stress range as humidity increases or in liquid water conditions.22 The transport of water molecules is not only affected by the total pressure, RH and temperature, but also highly affected by the crack tip geometry in the confined space.

Region III also refers to as inert propagation where environment plays little role on the crack propagation velocity. This is because stress is high enough to cause fracture without the assistance of water. The crack propagation velocity is significantly higher and even cause catastrophic failure on the glass materials. This region can be regarded as “mechanical-driven” regime.

I Figure 1-3. Schematics of the relationship between crack propagation velocity and applied stress from Ref. 20.

The amplitude of the stress applied in the wear behavior falls in the Region I or stress- corrosion regime. Noted that surface compressive stress from ion-exchange or other treatment can increase the threshold for crack to initiate and propagate in this regime. While shear during wear test is quite different from applied tensile stress in classical stress corrosion experiments, the breakage of BO due to tensile stress during wear test can still be analogous to stress corrosion phenomena.

Stress-induced Na+ migration

The concentration of Na+ on SLS glass surface upon shear could affect the rate as well as the type of the shear-induced reactions. Mobile Na+ can migrate within the glass as well as diffuse out of SLS glass in the presence of stress. Weber and Goldstein first observe the migration of sodium ions by monitoring the current changes when mechanical stress is applied through four

14 point bending.23 Na+ migrated from compressive stress end to tensile stress end in those cases.

Langford et al observed the emission of Na+ from fracture surface in vacuum.24 Celarie et al also observe the accumulation of Na+ at the vicinity of the crack tip where the glass network is under tensile stress.25,26 In this case, Na+ are in the form of hydrated salt and carbonate particulates by atomic force microscope (AFM) and time-of-flight secondary ionic mass spectroscopy (TOF-

SIMS) analysis, which suggests these Na+ are actually leached out of the glass and react with the water molecules in the environment.25

Another important factor in these previous studies is the water molecules in the environment. While the controlled humidity was not carried out in previous studies, the ambient humidity condition cannot be ignored. H2O molecules in the environment can react with the excess

Na+ at the surface and forms hydroxide or carbonate outside SLS glass, which will creates a local

+ + Na concentration gradient. The presence of H2O molecules not only provides H to charge balance the Na+, but also allows more Na+ to migrate out of the glass by maintaining the stress gradient.

Similar to dealkalization process, this migration is expected to cause surface silicate network to relax through reactions with H2O.

The migration of Na+ induced by stress gradient and further reactions with chemical species in the gas phase indicates that the surface structure of SLS glass will change dynamically upon applied stress. Na+ will migrate favorably towards the region where high tensile stress and high

+ concentration of H2O molecules are present. As RH increases, the stress-induced Na migration might be enhanced, which can accelerate surface reactions upon interfacial shear.

Stress induced reactions between glass and water around Tg

The existence of water molecules at temperature around Tg of SLS glass can help heal the surface flaws and cracks, which is quite opposite to what is observed for water at low

15 temperature.27,28 This “healing effect” is not observed if very little water exist in the environment

(e. g. Ar atmosphere) during the heat treatment. As the amount of water in the environment increases, the “healing” rate is also enhanced. Even though no solid scientific explanation has been provided to this observation, the phenomena suggest that condensation reactions of Si-OH groups to form Si-O-Si bond can be favorable over hydrolysis reactions on glass surfaces even tensile stress is present.

It is speculated that the healing of surface flaws is due to the rearrangement of SiOH due to the viscous flow of silicate network. Since the temperature is really high, both physisorbed and chemisorbed water should desorb from the glass surface region.29 Another important factor in this process is the change of viscosity of SLS glass when H2O is present in the silicate network.

Existence of small amount of water in SLS glass can significantly reduce the viscosity of glass around Tg up to 3 orders of magnitude.30 At the surface crack tip or surface flaws, it is possible that condensation of H2O (g) occur, which allow the silicate network to flow and rearrange locally.

The healing behavior of SLS glass in the presence of large amount of H2O(g) suggests that the hydrolysis of Si-O-Si on SLS glass is not always favorable as water concentration increases. A Vickers imprint could be totally healed by exposing to 22 torr water vapor around

Tg. In contrast, the cracks along the diagonal of Vickers imprint will propagate at temperatures below 200 C.31,32 These results suggest that condensation of SiOH groups could also become favorable at certain conditions.27 The key factors to trigger the condensation reactions are believed to be the high water content in the environment and the high mobility of glass network and mobile cations like Na+. It remains to be investigated if condensation reactions of SiOH groups can be the favorable reactions on SLS glass surface upon shear in humid environment.

16 Shear induced reactions on glass surfaces

When SLS glass is under shear stress in a reactive environment (For example: humid environment), chemical reactions mentioned in 1.1 could also take place. However, the mechanisms for these reactions to take place is still not clear. The study of wear behavior of SLS glass was initiated from the experimental observations by Bradley et al.33 It was carried out with a borosilicate ball on flat glass surface with ~300MPa normal stress in controlled humidity environment. It was found that very little wear was observed on SLS glass at high relative humidity

(RH) compared with medium humidity (~40%) while fused quartz (pure silica glass) showed more wear as RH increases. Further analysis on the wear behavior of SLS glass shows that this unique behavior of SLS glass only takes place when humidity level is close to saturation (RH90%) or in liquid water. It should be noted that wear behavior in humid environment and dry environment are due to different mechanisms. For wear behavior in dry/inert environment, glass materials including

SLS glass are all damaged due to mechanical scratching effect and should be governed by

Archard’s law. For wear behavior in humid environment, the debris generated are all hydrated glass, which indicate that hydrolysis reactions take place during the wear process.34 In other words, wear behavior in humid environment/reactive environment are sometimes denoted as

“mechanochemical wear” in contrast with “mechanical wear” which is used to describe the wear behavior in dry/inert conditions.

This unique behavior was not due to the borosilicate ball used in this study. He et al found that even materials with very high hardness and inertness like Si3N4, stainless steel and sapphire, will still be eroded at high RH conditions while very little damage is observed on SLS glass counter surface.35 Studies on the different glass substrate also shows that in addition to pure silica glass,

AF45, BF33, Na-aluminosilicate and K-exchanged aluminosilicate glass all show similar behavior for RH dependence wear behavior, which makes SLS glass unique among commercial available

17 flat glasses. The summary of the wear behavior of commercial flat glasses are shown in Figure 1-

Among these different glasses, the comparison between Na-aluminosilicate and K- exchanged alumiosilicate glasses is of interest. K-exchanged aluminosilicate glass has higher bulk hardness (up to 50 µm) than Na-aluminosilicate glass. For the surface hardness, this difference is believed to be even greater. However, the wear behavior of these two glasses are quite similar, which suggests that the surface compressive stress or mechanical properties might not be an important factor for wear behavior of Na-aluminosilicate glass. Both Na-aluminoslicate glass and

SLS float glass have ~ 3% Na+ atom% on the glass surface. However, the chemical environment

+ + - of Na for both of these glasses are quite different: Na is associated with AlO4 tetrahedral in Na-

18 aluminosilicate glass while Na+ is associated with NBO in SLS glass. The drastic differences in the wear behavior of these two Na+-rich glass surfaces suggest that glass network structure is also a very important factor. Another recent study on the wear behavior of Na+-rich surface after thermal poling of AF45 glasses also confirms the effect of Na+ associated with NBOs on the unique wear resistance at high RH.37 AF45 is an “alkali free” glass with ~400 ppm of Na+ in the bulk composition. However, after thermal poling, the Na+ can accumulate on the cathode side of the glass and formed a Na+-rich layer. This surface shows similar wear resistance behavior at high RH as the SLS glass which is quite different from AF45 glasses before this treatment.

Surface residual stress could also play a pivotal role in the wear behavior. As is discussed previously, the introduction of surface compressive stress makes the glass surface more resistant to crack initiation and propagation. By analyzing the failure strength of SLS glass before and after reactions with water15, surface compressive stress is expected to be generated. At high RH condition, such reactions are also expected to take place. Even though the K+/Na+ ion-exchanged aluminosilicate glasses do not have significant different wear behaviors than the aluminosilicate glass before ion-exchange, the effect of surface compressive stress cannot be ruled out easily because it has a different glass network from SLS glass.

Another factor that remains unclear is the role of counter surface during the wear test. It has been demonstrated that borosilicate ball, which is widely used in this study, is not the only material that results in the unique wear behavior of SLS glass at high RH. However, the behavior of the counter surface varies significantly at different RH levels. When RH level is below 90% but greater than 0%, the wear debris of the substrate adhere to the borosilicate ball or silica ball. At

90% RH, the counter surface will be polished and no debris from the substrate will adhere to the counter surface. The threshold for this transition of adhesion behavior of wear debris is lower in stainless steel and Si3N4 surfaces. While probing the adhesion of the wear debris in situ is extremely difficult, first-principle simulation that can describe the chemical reactions could reveal potential

19 insights on this matter. Some molecular insights on the wear behavior of silica materials have been proposed based on the molecular dynamics simulation with reactive force field (ReaxFF).38 It is suggested that the existence of H2O during sliding will help form a Si-O-Si bridge between substrate and counter surface.38

Therefore, the shear-induced reactions on SLS glass can be treated as a function of surface mechanical properties, structure of glass substrate and counter surface. Modification of these factors in controlled experiments would bring more physical insights on the nature of shear- induced reactions.

Chapter 2

Summary of employing specular reflectance infrared (SR-IR) spectroscopy, attenuated total reflectance infrared (ATR-IR) spectroscopy, environmental- control indentation and wear tests to analyze the surface structure and mechanical properties of SLS glass

Part of this chapter is reproduced with permission from Elsevier: Luo, J.; Huynh, H.; Pantano, C. G.; Kim, S. H. Hydrothermal Reactions of Soda Lime Silica Glass – Revealing Subsurface Damage and Alteration of Mechanical Properties and Chemical Structure of Glass Surfaces. J. Non. Cryst. Solids 2016, 452, 93–101.

Overview

This chapter summarizes the principles and experimental set-up of several techniques that are applied extensively to analyze the glass surface structure, mechanical properties and mechanochemical properties in this dissertation. These techniques include specular reflection

(SR) and attenuated total reflection (ATR) - infrared (IR) spectroscopy for glass surface analyses, friction and wear test for samples under tensile and compressive stress, environmental control crack initiation load analysis with Hertzian indentation and humidity control Vickers indentation.

The proper methodology and interpretation of SR-IR and ATR-IR techniques can provide surface-sensitive structure information, which is crucial in understanding the effect of surface treatment on the Si-O-Si network and hydrous species (SiOH & H2O) in glass surfaces.

Understanding the principles of SFG can ensure the proper interpretation of SFG spectra for different glasses. The setup of Hertzian indentation to characterize the crack initiation load of glass surfaces allow further investigation on the stress induced reactions on glass surfaces.

Combination of these principles and techniques allow one to explore the origin of wear behavior of SLS glass at high RH.

21 Applying SR-IR and ATR-IR to analyze the structure of Si-O-Si network and hydrous species in the surface region of SLS glass

Differences of SR-IR and ATR-IR in analyzing glass materials

For silicate glasses, mechanical strength and chemical reactivity are known to be affected by its surface condition, especially the speciation and concentration of hydrous species (Si-OH or

- H2O) and the network structure (Si-O-Si bridging oxygen (BO) and Si-O non-bridging oxygen

(NBO)) 33,39. For example, distributions of these groups can affect surface corrosion 40,41, adhesion of coatings 42,43, and fictive temperature 44 of silicate glasses. Therefore, full characterization of the chemical structure of the surface and subsurface region of silicate glass is very important.

Among various surface-sensitive analytical techniques, infrared (IR) spectroscopy has been widely used for this purpose. IR can probe BO and NBO groups, as well as Si-OH and H2O species inside the glass 40–44.

In order to obtain surface information, IR analysis of glass must be carried out in a specular reflection (SR) or attenuated total reflection (ATR) configuration, rather than typical transmission mode 45. Figure 2-1 schematically compares these three methods. In the transmission mode, the signal intensity is often expressed as absorbance (A) which is defined as –log(It/Io) where Io and It are the intensities of the incident and transmitted IR beams, respectively. The absorbance follows the Beer-Lambert law, 퐴 = 푎푏푐 where a is specific absorptivity, b is sample thickness, and c is concentration of the species of interest. Since the thickness of the surface region affected by glass manufacturing processes or environmental corrosion is much thinner than the total thickness of the sample (bsurface << bbulk), the transmission signal is always dominated by bulk species and their structure.

Figure 2-1. Schematic illustration of transmission, specular reflection (SR), and attenuated total reflection (ATR) IR spectroscopy of a flat glass sample.

In SR-IR, the intensity of the reflected beam (Ir) is measured and expressed as a reflectance (R=Ir/Io), which is dimensionally equivalent to the transmittance (It/Io) in the transmission mode experiment. When the IR frequency resonates with the absorption band of the sample, the reflected beam intensity is enhanced in SR-IR. Thus, the peak in the R vs. wavenumber (cm-1) plot of SR-IR is positive, while the peak in transmission IR is negative when plotted in the percent transmission (It/Io100%) scale.

In the case of ATR-IR, a crystal with a refractive index (n1) higher than that of sample of interest (n2) is in intimate contact with the sample surface and the IR probe beam is irradiated through this crystal. When the IR incidence angle (i) is higher than the critical angle

(c=arcsin(n2/n1)), then the IR beam is totally reflected from the crystal/sample interface and only an evanescent wave penetrates into the sample. The interaction of this evanescent wave and the absorption band of the sample attenuates the total reflection. When the measured reflectance (R)

23 is plotted as log(1/R), then it is dimensionally equivalent to the absorbance; thus, peaks in ATR-

IR spectra are often interpreted as absorption bands as in the case of transmission spectra.

However, it should be noted that peaks in the transmission IR spectra are governed by the absorptivity (a) which is a function of the imaginary (k) part of the complex refractive index of glass (n+ik) [i.e., 푎 = 4휋푘⁄ where  is the IR wavelength], while those in the SR-IR and ATR-

IR spectra are governed by both the real (n) and imaginary (k) parts that vary over a large range

46. As shown in Figure 2-2 46, the real part of the glass refractive index is not constant; due to the

Kronig-Kramers relationship, 47,48 n varies significantly along with k especially near the absorption band region. Thus, the SR-IR and ATR-IR spectra of glass cannot be interpreted in the same manner as the transmission IR peaks. The peak position, shape, and relative intensity can be drastically different in the SR-IR and ATR-IR spectra of glass.

Figure 2-2. Real (n) and imaginary (k) components of refractive index of soda lime glass.46

24 Obtaining surface Si-O-Si structure with SR-IR and hydrous species with ATR-IR

There are two major factors that need to be considered when applying IR technique with certain configuration on analyzing glass surfaces: 1. Can this IR configuration reveal the structure features of glass materials without optical artifact; 2. The information depth of this IR configuration. Here these factors will be demonstrated by simulation results which is based on the

Fresnel equations and refractive index shown in Figure 2-2. The details of the simulation is illustrated in the following paragraphs.

The reflectance of the IR beam was calculated using the Fresnel equations 49 and the refractive index of soda lime glass was taken from a literature (Figure 2-2) 46. The reflection coefficients of p-polarized and s-polarized beams (rp and rs, respectively) at an interface between media 1 and 2 are expressed as:

푛2푐표푠휃1−푛1푐표푠휃2 푛1푐표푠휃1−푛2푐표푠휃2 푟푝,12 = and 푟푠,12 = (2-1) 푛2푐표푠휃1+푛1푐표푠휃2 푛1푐표푠휃1+푛2푐표푠휃2

where n1 is the refractive index of the media 1, n2 is the complex refractive index of

46 media 2 (푛2 = 푛() + 푖푘() from Figure 2 ), θ1 is the incident angle (i in Figure 1) and θ2 is the transmission angle (t in Figure 1) calculated from Snell’s equation: 푛1푠푖푛휃1 = 푛2푠푖푛휃1. In the case of SR-IR, n1 = 1 (air); in the case of ATR-IR, n1 = ~2.4 diamond and ~4 for germanium.

50 51 The exact n1 value as a function of IR wavenumber can be found at for diamond and for germanium. Then, the reflectance of p-polarized (electric vector polarized parallel to the incidence plane) and s-polarized (perpendicular to the incidence plane) component of the IR beam are calculated as:

∗ ∗ 푅푝 = 푟푝,12푟푝,12 and 푅푠 = 푟푠,12푟푠,12 (2-2) and the total reflectance of the unpolarized beam is the average of these two values:

푅 = (푅푝 + 푅푠)/2 (2-3)

25 Within the glass sample, the electric filed intensity of IR decreases with an exponential function of the distance from the surface. The characteristic attenuation or penetration distance is expressed as the distance required for the electric field amplitude (transmitted beam in SR-IR or evanescent wave in ATR-IR) to fall to 36.8% (e-1) of its value at the surface. The penetration of

40 depth of SR-IR is defined as 푑푝,푆푅 = ⁄4휋푘() . The penetration depth in ATR-IR is expressed

2 2 2 as 푑푝,퐴푇푅 = ⁄(2휋Im [√푛2 − 푛1푠푖푛 휃1]) where Im[] means the magnitude of the imaginary

52 part. Note that when 푘() ≈ 0, then 푑푝,퐴푇푅 can be calculated using only the real part of n2. The information depth from which most of IR signal comes from is three times the characteristic penetration depth which is wavelength dependent.

When the glass sample is thin, a small portion of the transmitted IR beam can be reflected from the backside of the sample and can be detected along with the beam reflected from the front surface (Figure 1). In that case, the total reflectivity of SR-IR can be calculated as:

2푖훽 2푖훽 푟푝,121 = 푟푝,12 + 푡푝,12푟푝,21푡푝,21푒 and 푟푠,121 = 푟푠,12 + 푡푠,12푟푠,21푡푠,21푒 (2-4)

where 푡푝,12 and 푡푠,12 is the transmission coefficients at the front surface for the beam entering from air, 푟푝,21 and 푟푠,21 is the reflection coefficient at the back surface, 푡푝,21 and 푡푠,21 is the transmission coefficient at the front surface for the beam reflected from the back surface, and

훽 = 2휋푏푛1푐표푠휃1⁄ (where b = sample thickness; Figure 1) is the phase of the IR beam propagated through the glass. The transmission coefficients are calculated as:

2푛1푐표푠휃1 2푛1푐표푠휃1 푡푝,12 = and 푡푠,12 = (2-5) 푛2푐표푠휃1+푛1푐표푠휃2 푛2푐표푠휃2+푛1푐표푠휃1

The 푟푝,21, 푟푠,21, 푡푝,21, and 푡푠,21 terms can be calculated by changing the subscript order in the 푟푝,12, 푟푠,12, 푡푝,12, and 푡푠,12 equations, respectively. Then, the total reflectance (R121) from both front and back surfaces is calculated using the same equations (2-2) and (2-3) where 푟푝,121 and

푟푠,121, are used, instead of 푟푝,12 and 푟푠,12.

26 Experimental SR-IR spectra of both as-produced float glass with 0.7 mm thickness and polished and leached samples with 5.0 mm thickness were obtained by three instruments: (i) 20o incidence angle from the surface normal direction using a Bruker Hyperion 3000 micro-FT-IR system equipped with a 15x infrared microscope objective lens (Bruker Optics Inc.), (ii) ~45o incidence angle using a Thermo-Nicholet 670 FTIR system equipped with a custom-arranged optics, and (iii) 43o, 53o, 58o, 63o, and 68o incidence angles using a Bruker Hyperion 3000 system equipped with a Pike VeeMAX II ATR accessory. SR-IR spectra were obtained in the range of

4000 – 500 cm-1. Spectra were acquired for three spots per sample in 400 scan passes at a 4 cm-1 resolution. A gold mirror was used as a standard reference for all measurements.

A Bruker Vertex70 FT-IR system was used for ATR-IR analysis of both float (0.7mm thickness) and leached (5.0mm thickness) soda-lime glass samples using diamond and Ge ATR crystals. For the diamond ATR (MVP-Pro, Harrick Scientific Products), the IR beam incident angle was 45°. The sample was contacted against the ATR diamond crystal with a force of pushed by 420 N over a 1.5 mm2 sampling area. For the Ge ATR (VariGATR, Harrick Scientific

Products), the IR incident angle was set at 60o. The sample was contacted against the ATR germanium crystal with a force of pushed by 600 N over 1 cm2 sampling area. Spectra were collected for 100 scans with a spectral resolution of 6 cm–1 from 4000 to 400 cm–1 for diamond

ATR and from 4000 to 800 cm–1 in Ge ATR. The diamond ATR crystal absorbs IR in the 2300 and 2000 cm-1 region; so, this region cannot be probed. For the same reason, the Ge ATR-IR spectrum cuts off at 800 cm-1. For quantitative comparison of water related species as a function of acid treatment time, the peak intensity at 3400cm-1 was plotted after background correction.

In SR-IR analysis, the contribution from the backside reflection is sometimes ignored, which could lead to artifacts in the spectra. Figure 2-6(a) plots the IR penetration depth, dp, calculated using k(λ) in Figure 2-2. The penetration depth varies from ~0.65 m at the BO absorption band position to ~2 cm in the >3700 cm-1 region. It is noted that when the sample thickness is 700 m, then the >2200 cm-1 region of the IR beam can be reflected from the backside of the glass sample. This backside reflection can contribute to the signal intensity in the

o SR-IR experiment. When a microscope objective lens is used (i=20 data shown in Figure 2-

3(a)), the backside reflection is negligible since the focal depth of the objective lens is only on the order of several microns. However, in the typical SR-IR experiment with an elliptical mirror with

o a long focal distance (i=45 data shown in Figure 2-6(a)), the backside reflection should be taken into account.

Figure 2-3(b) shows how the backside reflection alters the SR-IR spectrum of a 700 μm thick float glass sample. When the IR beam is reflected from the front surface only, the SR-IR spectrum has no features at λ > 2000 cm-1. The reflection from the backside has zero signal at λ <

28 ~2240 cm-1, but has non-zero signal at λ > ~2240 cm-1. Since the front and backside reflected beams are spatially separated, the interference effect is negligible and the signals from these surfaces can be added. When these two components are added, the sum spectrum (solid line in

Figure 2-6(b)) shows negative peaks at ~2240 cm-1, ~2800 cm-1, and ~3400 cm-1, which is in good agreement with the experimentally observed spectrum (Figure 2-3(a), inset). This simulation clearly explains that the apparent peak at ~2240 cm-1 should not be interpreted as the silane (Si-H) species in the soda lime glass; it is simply the onset of the backside reflection contribution in SR-IR of a thin optically-flat glass sample. Note that the Si-H peaks are normally very sharp. The negative peaks at ~2800 cm-1 and 3400 cm-1 in the SR-IR spectrum are due to the absorption by hydrous species in the bulk. In this sense, the region above ~2240 cm-1 in the SR-

IR spectrum is similar to the transmission absorption spectrum.

In order to further understand the backside reflection contribution in SR-IR, we conducted a few control experiments (Figure 2-3(c)). When the sample thickness is changed from

700 m to 5mm, then the onset of the backside contribution is shifted from ~2240 cm-1 to ~3600 cm-1. When the backside of the 5mm thick sample is roughened, then this contribution is further suppressed. When CCl4 is placed on the back of the 700 m thick sample, then the backside reflection is completely suppressed since the refractive index of CCl4 is very close to that of soda lime glass.

In ATR-IR, the IR incident beam travels through the high refraction index crystal and is reflected from the crystal/glass interface; thus, the reflection behavior can be divided into two regimes: (i) θi < θc (SR-IR region) and (ii) θi > θc (ATR-IR region). The critical angle (θc) is

o o ~38.7 for the diamond crystal and ~22.0 for the germanium crystal. When θi < θc, the peak shapes in the reflection spectra shown in Figures 2-7(a) and 2-7(b) appear to be governed mostly by the real part, n(λ), of the glass refractive index (Figure 2-2). When θi > θc, the signal in the non-absorbing region (>1200 cm-1) is almost zero since the IR beam undergoes total internal

-1 reflection (i.e., R = Ir/Io = 1; log(1/R) ≈ 0). In the absorbing region (<1200 cm ), k(λ) has non- zero value and the total internal reflection is attenuated (i.e., R < 1; log(1/R) > 0). It is noted that the Si-O-Si peak position of the maximum intensity is not the same as the absorption band position in k(λ); it is significantly red-shifted to ~910 cm-1 for the diamond-ATR data in Figure 2-

4(a) and ~990 cm-1 for the Ge-ATR data in Figure 2-4(b). Such a red-shift in the absorption peak position is insignificant for organic materials since k(λ) is very small and thus variance in n(λ) is negligible. However, in the case of soda lime glass, both n(λ) and k(λ) vary over a large range, causing distortion of the peak shape in the ATR-IR spectrum. Thus, the difference in the peak positions of the SR-IR and ATR-IR spectra should not be interpreted as structural modification in the surface region. These differences are simply due to the anomaly of the IR beam reflection resulting from changes in n(λ).

Figure 2-5 (a) compares the ATR-IR spectra of a SLS glass collected by two different

ATR crystal. The origin of the different spectra shape is related to the refractive index of the ATR crystals as well as the different information depth of these two crystals. The information depth is estimated with the refractive index reported in Figure 2-2. As can be seen in Figure 2-5(b), Ge crystal has a much shallower information depth than diamond crystal due to its larger refractive index. It should also be noted that the baseline of ATR spectra collected with two different crystal can be quite different due to the force applied during the measurement and crystal geometry. Thus caution must be taken when comparing the absolute intensity of ATR-IR spectra collected with different crystal and experimental setup.

To summarize, SR-IR can be very useful in analyzing the surface Si-O-Si network structure (1300 – 600 cm-1) while ATR-IR will provide excellent surface sensitivity in the surface structure of hydrous species (4000 – 1500 cm-1). For SR-IR analysis, the information depth is less than 2 microns in the Si-O-Si vibration region and true absorption bands can be obtained with further analysis (discussed in details in Chapter 7). However, the much larger information depth

31 in OH stretch region makes SR-IR complicated in interpreting the OH stretch in the glass surface.

ATR-IR analysis can provide good surface sensitivity in OH stretch and H2O bending vibration region. However, the analysis of Si-O-Si network using ATR-IR is affected by the optical defects due to the smaller refractive index differences between glass and ATR-IR crystal in this region.

Friction and wear test for glasses under applied tensile stress

Cleaning procedure prior friction and wear test

Schematic of friction and wear test carried out in this study is shown in Figure 2-3. A pyrex ball from McMasterr Carr with 2.4 mm diameter is used as counter surface. For a typical wear test, 0.2 N normal load is used. The environment is controlled by purging a small chamber

(volume ~25 cm3) with desired gas. The number of cycles is 400 cycles to compare with the previous studies.

Cleaning protocol is essential in performing friction and wear test, especially the cleaning procedure for the substrate. As received SLS float glass will be contaminated by both the manufacturing process and handling process during the transport. As a result, the SLS glass is mostly contaminated with sodium containing mineral compound including sodium carbonate, sodium sulfate and organic contaminant like oil and volatile organic compound (VOC) from the atmosphere. Typically the inorganic salt can be rinsed off with DI water and organic contaminants can be rinsed off by sonication in pure ethanol or acetone. In our recent study, it is found that the organic residual from rinsing in pure ethanol can leave a flim of organic residual on the glass surface.53 This thin layer of organic residual can have lubrication effect during the friction and wear test. This thin film of orgnic residual can be removed by UV-ozone treatment or rinsing in DI water again. Therefore, the cleaning protocol of friction and wear test on glass

32 surface is set as follows: 1. Rinsing SLS glass air side with DI water 3-5 times; 2. Rinsing SLS glass air side with pure ethanol; 3. UV-ozone treatment on the air side of SLS glass for 20 mins.

The cleaning protocol for the counter surface (pyrex ball) is the same as the substrate.

Interestingly, UV ozone treatment has not been found to be necessary for cleaning the counter surface. Both substrate and counter surface will be checked with an optical microscope equipped with a 20x objective lens prior to friction and wear test.

Figure 2-6. Schematic of Friction and wear test in controlled environment.

Characterization of wear track with optical profilometry

After the friction and wear test is performed, the topography of ball and substrate will be characterized with optical profilometry. The principle of optical profilometry is based on the interference of the light. From the topography image, the cross section line profile can be obtained from the image with height information. The wear depth and wear volume can then be calculated based on the optical profilometry images.

Figure 2-7. Typical optical profilometry image of substrate and ball after friction and wear test in 40% RH and 90% RH conditions.

Holder design for wear test of glass under applied tensile stress

To explore the surface stress effect on the friction and wear behavior of SLS glass at high relative humidities, a three-point bending holder was built as is shown in Figure 2-5. The slide glass will be placed in the slot. A polished aluminum rod is located in the middle of the holder and served as the middle point in three-point bending. The two side of the glass will be fixed by screws and polished aluminum rod. The displacement can be adjusted by changing the height of the aluminum rod above the slot in the middle. Based on the mechanical properties and practical strength of SLS glass, the maximum tensile stress of ~100 MPa tensile stress can be applied. This holder can also be used in other characterization techniques like SR-IR, ATR-IR analysis.

Figure 2-8. Drawings of sample holder to perform applied tensile stress on glass substrate.

Labview software development for crack initiation load analysis of glass materials with Hertzian indentation

A Hertzian indenter was modified to perform crack initiation load analysis on flat glasses.

A schematic of this experimental setup is shown in Figure 2-9. A 1mm WC ball is used as the indenter material. The load is applied through a stepping motor and recorded by a load cell attached to the Hertzian indenter. The initiation of the crack is detected by an acoustic sensor

(Model No.: PICO HF1.2, from Mistras Group, Inc.). To detect the crack initiation load, the stepping motor and acoustic sensor need to be synchronized. Once the sound of the crack is detected by the acoustic sensor, the Hertzian indenter needs to be retracted immediately to

35 prevent further damage of the glass. Here an AD converter from National Instrument is used to convert and synchronize the signal from stepping motor controller and acoustic sensor. In a typical test, the normal load continuously increase under certain humidity conditions until the crack is formed and detected. A Labview software is developed with the help of Mohammed

Alshehri (an undergraduate researcher in Electrical Engineering Department of Pennsylvania

State University).

Figure 2-9. Schematic of crack initiation load detection with a Hertzian indentater.

The details of the software interface is shown as follows. There are six major sections in this software marked as A-F. In Section A, the physical address of the motor and data file path will be shown. The time stamp will be automatically added to the file name to keep track of the data file. Section B allows a quick calculation of Hertzian contact stress based on the Hertzian

36 contact mechanics when radius of the indenter, the elastic modulus, Poisson’s ratio of indenter and the substrate are typed in. Sigma will display the Hertzian contact stress in MPa. Section C is the control section of stepping motor. “AUTO (SLOW)” is the most common tab to use. It allows the user to set the ramping rate of stepping motor (steps/s) to control the loading rate of Hertzian contact normal load (N/s). Typically 5 steps/s correspond to 1N/s. Total number of steps can be set at the beginning. The number of actual steps will be recorded when crack is initiated. Section

D is showing the parameters of acoustic emission (AE) sensor before and after picking up the sound from cracks. The details of the meaning can be found in the manual of AE sensor from

Mistras. Section E will display the acoustic wave and Fourier Transform (FFT) of the acoustic wave. The acoustic wave can be related to the patterns of the crack. Section F is to display and record the force applied before and after crack is initiated. For safety concerns, the maximum force need to be set as 1000 N prior to the test. After one indentation test is completed, the user has the capability of save all the data displayed here and write short note on the experiment. The information directly related to the analysis including crack initiation load, crack initiation stress, energy of the sound wave and the type of the crack (ring, cone or radial crack) based on the observation from optical microscope, is summarized in a excel file that can be appended to.

Figure 2-10. Labview software interface to perform crack initiation load analysis on glass surface.

Environmental controlled Vickers indentation

The Vickers indentation was very useful in characterizing the materials’ mechanical properties like hardness and fracture toughness. However, it has been found that the H2O molecules in the gas phase can have an impact on the Vickers indentation results. The formation of cracks from the diagonal of Vickers indent seem to be related to the stress corrosion phenomena that is widely observed in the silica and silicate glasses.

Figure 2-11 shows the schematic of environmentally controlled Vickers indentation analysis on glass materials. The design is very similar to that of Hertzian indentation with humidity control capabilities. The load and loading rate is normally fixed.

38 Figure 2-12 shows the Vickers indentation on a SLS glass in different humidity conditions. For the same load and same piece of SLS glass, the cracks will pop from the diagonal of the Vickers imprint when glass is exposed to 90% relative humidity gas. However, when very little water exist (~0%) in the environment, the crack does not form. It has been recorded that the cracks will immediately form when dry N2 stops flowing (system try to recover to room humidity,

~40% RH). Similar phenomena is observed when Vickers indentation is performed close to liquid

54 N2 temperature (partial pressure of gas phase water would be very low) . The nature of such behavior remains to be studied.

Figure 2-11. Schematic of Vickers indentation in humidity controlled environment.

Figure 2-12. Vickers imprint of SLS glass in 90% and 0% relative humidity. The load is 300gf for both cases.

Chapter 3

Hydrothermal reactions of soda lime silica glass – revealing subsurface damage and alteration of mechanical properties and chemical structure of glass surfaces

Reproduced with permission from Elsevier: Luo, J.; Huynh, H.; Pantano, C. G.; Kim, S. H. Hydrothermal Reactions of Soda Lime Silica Glass – Revealing Subsurface Damage and Alteration of Mechanical Properties and Chemical Structure of Glass Surfaces. J. Non. Cryst. Solids 2016, 452, 93–101.

Overview

Hydrothermal treatment provides a unique way to reveal subsurface damage of soda lime silica (SLS) float glass. During the hydrothermal treatment, the penetration of water, ion- exchange, and hydrolysis reaction can take place. These reactions are accelerated at locations where subsurface damage exists. Interestingly, the hydrothermally-treated glass surface exhibits higher fracture toughness, lower hardness, and less resistance to mechanochemical wear at high humidity compared to the pristine SLS float glass. These property changes can be explained by the leaching and polymerization of the silicate network and the chemical environment of hydrous species in the surface region of SLS glass.

Introduction

The surface and subsurface damage of soda lime silica (SLS) glass can affect the strength and durability of the glass.55–58 Such damage can be created when contact stresses are applied to the glass surface during the manufacturing, handling, and storage, but the severity of damage can

41 be mediated through environmental effects. For example, adsorption of a monolayer of lubrication molecules on the surface can prevent physical wear at the sliding interface whereas interfacial shear in the presence of reactive molecules will damage and degrade the surface.36,59 The history of normal indentation or interfacial shear at the surface can lead to a distribution of local chemical structures and residual stresses.35,60–62 In the case of sharp cracks, tensile residual stress is believed to be present and, in humid environments, the crack will propagate in the direction where this residual stress is largest.20 Changes in the stress profile in the surface region at the vicinity of the crack tip can be accompanied by changes in the chemical structure.25 While it is generally accepted that these phenomena can be explained by stress corrosion mechanism,21,63,64 revealing the subsurface damage and understanding their effect on mechanical properties and mechanochemical behavior is still challenging.

This study describes a method to reveal subsurface damage without using strong acid or base etching. Wong et al. showed that HF can be used to reveal subsurface damage on fused silica by controlling the etching rate precisely.65 In this case, HF etching treatment reveals the subsurface damage at the cost of destroying the entire glass surface.66 Since the subsurface damage is accompanied by residual stress which strains the silicate network, it is hypothesized that the associated strained bonds will preferentially react with water molecules. Here, it is shown that if reactions between the SLS glass surface and water are accelerated through hydrothermal treatment, the residual stress and related chemical structures can be etched selectively to map the damage on the surface.

Another outcome of this study was the observed alteration of mechanical properties and the mechanochemical response due to the hydrothermal leaching and hydration of the surface. For example, if the sodium ion leaching creates a softer layer or region on top of the bulk glass, indentation loading and unloading along the surface normal direction will be altered since the surface layer will be compacted and absorb more mechanical energy.67 Recently, SLS glass has

42 been found to show very unique wear resistance in high relative humidity environment.36 While the mechanism of such mechanochemical wear behavior is not fully understood yet, the wear resistance was found to be related to the presence of sodium ions associated with the silicate network of SLS glass. The glasses without leachable sodium ions such as fused quartz, alkali-free display glass, and borofloat do not show the same wear resistant behavior at high humidity.33,39 In aluminosicliate glasses where sodium is associated with aluminum tetrahedral sites instead of non-bridging oxygen in silicate network, the wear resistance at high humidity is not observed.58 While removing the sodium ions from the surface region of SLS glass through thermal poling also removes the wear resistance effect of SLS glass, the silicate network is also altered drastically.67 Therefore, a means to leach sodium out of SLS glass with no or minimal change in the subsurface network is needed to investigate the effect of sodium ions in mechanochemical wear resistance of the SLS glass.

In order to achieve these goals, hydrothermal reactions of SLS glasses were studied.

Hydrothermal reactions are typically carried out in a sealed reactor at temperatures higher than

100 oC.68 During the hydrothermal treatment at 200 C and 250 C, water can diffuse into alkali- free glasses and react forming Si-OH groups.16,69 Such hydrothermal treatments in liquid water have been reported to strengthen a vitreous silica glass.16,18 In the case of SLS glass, hydrothermal treatment in liquid water was reported to create a porous surface layer.70 Our study investigated the effect of hydrothermal treatment in the vapor phase around 150 C with the focus on the aforementioned goals – revealing the subsurface damage; studying mechanical properties of the modified surface layer; and investigating how the wear behavior is affected by sodium ion leaching. It should be noted that liquid water and water vapor have the same reactivity in hydrothermal reactions since their chemical potentials are the same at this saturated condition.

However, transport properties in the vapor phase - especially the removal of reaction products from the glass - would be different than for the liquid phase.

43 Experimental methods

SLS float glass with a 1mm thickness from Asahi Co. (Asahi Co., Tokyo, Japan) was used in this study. The bulk composition of SLS (weight %) was found to be 72.3% SiO2, 13.3% Na2O,

7.7% CaO, 1.9% Al2O3, 4.4% MgO, 0.3% K2O, and 0.1% Fe2O3 from X-ray fluorescence (XRF).

During the manufacturing process, SO2 treatment was applied after the glass was lifted from the tin bath onto the rollers, and this dealkalized the surface. This study investigated the air side of the float glass only because it is more pristine and defect free initially.71

The vapor-phase hydrothermal treatment system is schematically shown in Figure 3-1. It was performed in a sealed stainless steel vessel with a volume of 100 cm3. 10 mL of miliQ water was placed in the bottom of the vessel that was rinsed with ethanol and water before each treatment.

The amount of water was enough to create saturated vapor between 100 C and 200 C in the sealed vessel. At 150C, the saturated water vapor pressure is 0.48 MPa. The actual temperature of the vessel was monitored with a thermocouple. All samples were cut in 2 cm  2 cm and placed on an elevated sample stage in the vessel so that the glass would interact with saturated steam only. The sealed vessel was then put in an oven set to a desired temperature. After the hydrothermal treatment, the sealed vessel was rapidly cooled with running water. The treated samples were cleaned with miliQ water, pure ethanol and UV-ozone before any mechanical tests and surface characterizations.53

Figure 3-1. Schematic representation of vapor-phase hydrothermal treatment of SLS glass at temperatures higher than 100 oC.

Specular reflectance infrared (SR-IR) spectroscopy was carried out with a Bruker Hyperion

3000 Microscope (Bruker, Co.) with a 15 objective lens. A gold mirror was used as a reference background. Attenuated total reflectance infrared (ATR-IR) spectroscopy was performed with the same IR microscope equipped with a Ge ATR crystal accessory with 60 incident angle. Sum frequency generation (SFG) vibrational spectroscopy was used to study the chemical environment of hydrous species in the surface region of SLS glass. The detailed description of the SFG system can be found elsewhere.72 In brief, visible laser pulses (532nm) from a 27 ps Nd:YAG laser and tunable IR pulses (2.5-10 m) from an optical parameter generator and amplifier were spatially and temporally overlapped on a glass surface of interest. The incident angles of visible and IR pulses were 60 and 56 with respect to surface normal, respectively. The SFG signal was collected in a reflection geometry at the angle determined by the phase matching condition of the SFG process.

The polarization combination used in this study was s for SFG signal, s for 532nm laser pulses, and p for IR laser pulses (ssp).

45 The surface composition of SLS before and after hydrothermal treatment was analyzed with X-ray photoelectron spectroscopy (XPS). A Kratos Analytical Axis Ultra spectrometer

(Chestnut, NY) fitted with a monochromatic AlK (1486.6 eV) X-ray source was used. Survey scans of O 1s, Na KLL, Ca 2p, Mg KLL, C 1s, Si 2p peaks and high-resolution narrow binding energy of O 1s and C 1s peaks, were conducted at 80 and 20 eV pass energies, respectively. The binding energies of all elements were corrected with the adventitious alkyl peak at 285 eV. The surface composition was determined after removing the adventitious carbon and carbonate species on the glass surface.73

Nanomechanical properties including elastic modulus and hardness of the hydrothermal treated surface were obtained using a nanoindenter (Hysitron TI 950, Minneapolis, MN) equipped with a Berkovich tip. The nanoindenation measurements was performed with displacement control. The maximum penetration depth was held at 50 nm, 100nm, 150nm and 200nm for 2 seconds. The loading and unloading rate were both 20 nm/s. The results are averaged from more than 40 indentations for each indentation depth. Vickers indentation was performed with a microindenter (MHT Series 200; Leco Corporation, St. Joseph, MI). The duration time at maximum load was 15 seconds. The Vickers hardness was averaged from more than 15 measurements for each sample. The duration at maximum load for both nanoindentation and

Vickers indentation are small enough that influence from indentation creep effect will not be significant.74 Wear test was done using a custom-designed ball-on-flat tribometer with an environment control capability. All wear tests were conducted with 0.2 N normal force and 400 reciprocating sliding cycles. The ball used in this test was a borosilicate ball with 2.4 mm diameter (McMaster-Carr Products Inc., Elmhurst, IL). The contact pressure was calculated to be

350 MPa based on Hertzian contact mechanics. “Invisible” wear tracks, tracks that did not undergo plastic deformation during the reciprocating shear loading, were created by performing

46 the wear test in n-pentanol vapor environment.36,75 The wear tracks before and after hydrothermal treatment were analyzed with an optical profilometer (Zygo NV7300, Middlefield, CT).

Results and Discussion

Selective etching of regions with subsurface damage or residual stress via vapor-phase hydrothermal reaction

Stress corrosion is a phenomenon that accelerates hydrolysis reactions of glass upon application of mechanical stress to the glass;63,76,77 a similar effect could be expected when a glass with residual stress is exposed to hydrolysis reaction conditions. The subsurface damage could be associated with subsurface defects such as non-bridging oxygen (NBO), three-coordinated silicon, and strained silica tetrahedron.78 It was hypothesized the hydrothermal treatment can reveal the residual stress in the surface region since the defect sites or areas would react more readily with water at high temperature and pressure. To test this hypothesis, we ran two control tests: (i) hydrothermal reactions of the surface subjected to interfacial shear and (ii) hydrothermal reactions of an indentation crack. The diffusion and chemical reactions of water can lead to selective etching of these defective regions without using any chemical etchants such as strong alkali or hydrofluoric acid.

Figure 3-2. Optical profilometry of SLS glass surface slide tracks produced in n-pentanol VPL environments (a) before and (b) after hydrothermal treatment at 150 °C for 24 hours; (c) cross- section line profiles of the wear tracks marked with dashed lines in (a) and (b).

It is known that as long as the bulk can support the applied load, the ‘visible’ wear of the glass surface can be suppressed by continuous adsorption and replenishment of alcohol molecules from the vapor phase.36,39,75 This is called vapor phase lubrication (VPL). The key for the success of VPL is to avoid detrimental reactions at the sliding interface that can induce chemical wear of the surface. Although the formation of visible wear tracks can be suppressed in VPL, the interfacial shear under a high normal load (~350 MPa in the test condition used in this study) could leave some damage in the subsurface glass network. Figure 3-2 demonstrates how this

“invisible” wear track with subsurface damages can be revealed by hydrothermal treatments.

Figure 3-2a and 3-2b compare the optical profilometry images of the same wear track before and after hydrothermal treatment. Figure 3-2c compares the cross section of wear tracks before and after hydrothermal treatment. The wear track became approximately 50 nm lower than the outside surface after hydrothermal treatment. This result clearly supports the hypothesis that the region

48 with subsurface damage is more reactive than the pristine surface and can be preferentially etched by water under hydrothermal conditions.

Figure 3-3. (a) Vickers indent prepared with 500 gf load in 40% relative humidity; (b) indent from (a) after hydrothermal treatment at 150 °C for 24 hours.

Vickers indentation can be used to create regions with residual stress; when a radial crack is formed, the region in front of the crack tip is under residual stress.57,79,80 As shown in Figure 3-

3a, radial cracks were normally formed on the SLS float glass after indenting with a 500 gf normal load in 40% relative humidity air. After the sample was hydrothermal treated at 150 °C for 24 hours, the length of the cracks was increased by 22%. For some indents, new cracks were also formed as shown in Figure 3-3b. Again, this result supports that the glass surface with residual stress around crack tips can be selectively etched by hydrothermal treatment.

Mechanical properties of hydrothermally-treated SLS glass surface

The elastic modulus and hardness of the glass surface before and after hydrothermal treatment were measured with nanoindentation and are summarized in Figure 3-4. The elastic modulus and hardness were calculated from force-displacement curves using the Oliver-Pharr

49 model.3 The analysis was performed with four different indentation depths: 50nm, 100nm, 150nm, and 200nm. Since a surface layer with a modified chemical structure is expected to form during the hydrothermal treatment, the Poisson’s ratio of the modified layer cannot be determined accurately.

Therefore, the reduced modulus is reported rather than elastic modulus. It should be noted that the modified layer formed by hydrothermal treatment is not like an external film with an abrupt interface with the bulk substrate, but rather is a gradient layer that gradually changes in chemical and physical properties from the surface to the bulk. For these reasons, the accurate determination of mechanical properties for the modified layer was difficult; however, the qualitative trend of modulus and hardness for the modified layer can still be analyzed.

Figure 3-4. (a) Hardness and (b) reduced modulus of the SLS glass surfaces before and after hydrothermal treatment determined from nanoindentaiton.

Figure 3-4a shows that the overall hardness decreases after the hydrothermal treatment.

This indicates that the silicate network is altered due to the reaction with water in the surface region, which could result in a hydrated layer. The longer treatment makes the glass surface softer, which can be associated with the thickening of a modified layer. As shown in Figure 3-4b, the reduced modulus of the surface treated for 24 and 48 hours is slightly higher than the pristine glass surface, and the surface hydrothermal treated for 72 hours is similar to the pristine surface. This non-

50 monotonic trend must be related to an artifact associated with the modulus calculation in nanoindentation. When a soft layer is present on top of the glass, the softer layer could pile up around the probe tip during the indentation. Then the actual contact area is hard to determine from the indentation curves and could be larger than the area estimated from contact depth based on

Oliver-Pharr method. As a result, the contact area estimated by the Oliver-Pharr method could be smaller than the actual contact area, leading to an overestimation of the reduced modulus.81 In such a case, as the ratio of the indentation depth over film thickness increases, the reduced modulus will increase, reach a maximum, and then decrease. The reduced modulus estimated by the Oliver–Pharr method is smaller for the 72 hour sample than for the samples treated for 24 hours and 48 hours since a thicker modified layer is expected for SLS glass that has been hydrothermally treated for

72 hours and the ratio of the indentation depth over film thickness.81

Resistance to the nucleation of a crack in glass is a desirable property. Vickers indentation was performed with 300 gf on the SLS glass surface before and after hydrothermal treatment. Figure 3-5a summarizes the Vickers hardness before and after hydrothermal treatment.

51 It should be noted that the hardness obtained from Vickers indentation is smaller than that from nanoindentation. This is because the Berkvich tip used in nanoindentation has a smaller size and a different shape than the Vickers tip used in microindentation.82 The student t-test suggests that the difference in Vickers hardness is not statistically significant with 95% confidence (p= 0.39>0.05), which indicates that the bulk mechanical properties do no change by the hydrothermal treatment.

No cracks are observed on the samples that are hydrothermal treated at 150 C for 72 hours

(Figure 3-5c) while the pristine SLS glass (Figure 3-5b) shows radial cracks from the Vickers indenter corners. Based on the nanomechanical properties of the hydrothermally treated surface determined in Figure 4, the surface layer formed by hydrothermal treatment is softer than the bulk. Thus, the resistance to crack formation after hydrothermal treatment cannot not be due to the compressive stress as is observed in chemically strengthened glass surface through ion- exchange process.83 Instead, it must be due to the formation of a softer layer with modified chemical structure on top of the glass. The effect of such a modified layer formed from the bulk glass itself on fracture toughness of the surface has not been fully understood. One possible explanation is that the formation of the silica-like surface layer can prevent or suppress crack opening or initiation. This speculation is based on the fact that silica glass has higher fracture toughness than SLS glass.67 Another possible explanation could be the changes in the network connectivity in the modified surface layer based on the topological constraint theory.84 It is conceivable that the overall topological constraint in hydrothermally-modified surface region of

SLS glass is altered since the number of Na-NBO sites decreases and hydrous species are introduced.85

52 Mechanochemical wear of hydrothermally-treated SLS glass surface

Unlike the alcohol VPL which prevents wear of the surface, water adsorption from the surrounding gas phase can induce mechanochemical wear due to hydrolysis reactions facilitated by interfacial shear.34,38,86The mechanochemical wear behavior of SLS glass shows a very unique humidity dependence.33,39 Unlike other glass surfaces that show more wear as humidity in the test environment increases, SLS glass shows a decrease in wear as humidity increases, especially in near-saturation humidity.34 While exact mechanisms of such resistance to mechanochemical wear at high humidity are not fully understood yet, it is believed to come from dynamic interactions between leachable Na+ ions and the silicate network in the presence of water molecules.67 XPS analysis indicates that sodium ions are almost completely depleted in the hydrothermal-treated SLS glass surface (see Supporting Information). Thus, if the hypothesis involving the leachable Na+ ions is correct, it is expected that the resistance to mechanochemical wear at high humidity is reduced or lost for the hydrothermal-treated SLS glass surface.

Figure 3-6. Comparison of wear tracks formed after 400 sliding cycles in (a) 40% relative humidity (b) 90% relative humidity on the pristine sample, the sample treated at 150 C for 24 hours in dry air, and the sample treated at 150 C for 24 hours in steam. (c) Comparison of the wear volume. The p-values in the figures are from comparison between two cases: pristine vs. heat-only and heat- only vs. steam-treated for the 40% humidity case and pristine vs. heat-only for the 90% humidity case.

53 The cross sections of the wear tracks produced at 40% and 90% relative humidity conditions for SLS glass surfaces before hydrothermal treatment and after heat treated at 150 C for 24 hours in dry air and in steam are compared in Figure 3-6. The heat treatment in dry air was conducted to rule out the effect of thermal annealing on the wear behavior of SLS glass. As shown in Figures 3-6a, the wear behavior of SLS glass at 40% relative humidity is indistinguishable for the pristine, heat-treated, and hydrothermal-treated surfaces (the p-values from t-test are provided in Figure 3-6c). However, the wear depth and wear volume at 90% relative humidity are significantly increased after hydrothermal treatment, while both pristine and heat-only treated SLS glasses show very little wear (Figures 3-6b and 3-6c). In Figure 3-6b, small positive spikes on the substrate surfaces are physical wear debris of the counter surface deposited on the SLS surface.33 The increase in wear volume at 90% relative humidity compared to the 40% relative humidity case is similar to the mechanochemical wear behaviors of other glasses without leachable sodium ions.39,58 The results shown in Figure 3-6 clearly show that the resistance to mechanochemical wear at high humidity is a unique property for the SLS glass network containing leachable sodium ions; once sodium ions are depleted, the mechanochemical wear resistance effect is lost.

Subsurface structural changes upon vapor-phase hydrothermal reactions

It is generally known that during the hydrothermal treatment, molecular water can diffuse into the SLS glass and induce the hydrolysis of the silicate network at the bonding oxygen sites:

Si-O-Si+H2O ⟺Si-OH + HO-Si (1)

or the ion-exchange at the non-bonding oxygen (NBO) sites:20,87,88

Si-O-Na++ 퐻+ ⟺ Si-OH + Na+ (2)

- + + + Si-O Na + 퐻3푂 ⟺ Si-OH + Na + H2O. (3)

54 While these reactions are slow at ambient conditions, they can be significantly accelerated at a temperature higher than 100 oC.21,89–91 First, the diffusion of water penetrating into the glass network is accelerated at elevated temperatures.92 Second, the reactivity of water molecules is also higher. For example, the ionization constant of water (Kw) at 150 C is approximately 2 orders of magnitude higher than the value at room temperature.93,94 In other words, the equilibrium concentration of protons and hydroxides are 10 times higher than normal water at room temperature. And, their thermal energy is higher compared to the ambient condition. Thus, hydroxide ions can catalyze the hydrolysis reaction (Eq.1) and the protons and hydronium ions can facilitate the sodium leaching reaction (Eq. 2 and 3). These reactions can occur not only when glass is in contact with liquid water, but also when the glass surface is exposed to steam at high temperature and pressure. It is conceivable that such reactions would occur more readily at the defect sites with strained chemical bonds, which can explain selective etching of the ‘invisible’ wear track formed in the n-pentanol VPL condition (Figure 3-2) and the region with residual tensile stress in front of a crack tip (Figure 3-3).

In order to understand the mechanical and mechanochemical property changes of SLS glass after hydrothermal treatment, analyses of the silicate network structure as well as the amount of and chemical environment of hydrous species (silanol groups and molecular water) were carried out. Figure 3-7a shows the SR-IR analysis on the changes of silicate network before and after hydrothermal treatment. As the reaction time between glass surface and water increases, the peak centered around 1056 cm-1 gradually shifts to 1064 cm-1. In previous studies, this shift was normally observed at the surfaces where the density decreased due to the sodium depletion and the silicate network became a more silica-like structure.42,95 These findings indicate that substantial reactions occur between water and glass surface during the hydrothermal treatment, modifying the subsurface silicate network structure.

The reactions between glass surface and water, and the diffusion of water molecules into the glass network, can be studied by examining hydrous species in the surface region before and after hydrothermal treatment. ATR-IR analysis was carried out to monitor the changes in the amount of these hydrous species in the subsurface of SLS glass. The information depth of ATR-IR is about 0.5-0.7 m for the stretching vibration region of hydroxyl groups and ~1.1 m for the bending vibration region of molecular water.96,97 As shown in Figure 3-7b, the molecular water peak at ~1620 cm-1 grows as the hydrothermal treatment temperature increases. This provides direct evidence that water molecules diffuse into the glass network. The activation energy (퐸푎 ) for molecular water diffusion into the SLS glass can be obtained by assuming the Arrhenius relationship between the rate of diffusion (k) and temperature (T):

푘 = 퐴 퐸푥푝[−퐸푎/푅푇] Eq. 4 where 퐴 is a pre-exponential factor. Since k is in a linear relationship with the amount of water molecules in the glass which is linearly proportional to the peak area (M) of the water bending

56 vibration, k can be substituted with b M where b is a constant. Then, equation (4) can be expressed as follows:

퐸 1 푏 퐿푛(푀) = − 푎 − 퐿푛( ) Eq. 5 푅 푇 퐴

By fitting the experimental data (Figure 7c) with a linear relationship in equation (4), 퐸푎 for diffusion of molecular water is calculated to be 40  2.8 kJ/mol.

In the OH stretch region of mid-IR spectra of glasses, both silanol (Si-OH) and molecular

-1 water (H2O) are detected. The small peak at ~3740 cm is assigned to free OH of molecular water without hydrogen bonding with surrounding molecules.98,99 Although not well-resolved, the steep rise and shoulder in the ~3650 cm-1 region suggests the presence of a distinctive OH component; when it is deconvoluted through peak fitting, the ~3650 cm-1 component is generally assigned to free Si-OH without hydrogen-bonding with surrounding oxygen atoms.100 All other components contributing to the broad feature from 3600 cm-1 and 2600 cm-1 can be assigned to OH groups (of

101,102 both Si-OH and H2O) with varying degrees of hydrogen bonding interactions. In previous literature, the broad OH stretch peaks were fitted with an arbitrary number of components and varying peak widths and the component near 3400 cm-1 was attributed to the molecular water in the glass network.100,103,104 But, it should be noted that the deconvolution of the hydrogen-bonded

OH stretch peaks cannot be done through such a simple curve fitting procedure. The OH peak position and width are complicated functions of hydrogen bonding strengths and dynamics.105 For that reason, we did not attempt to deconvolute the broad OH stretch peak (Figure 3-7b).

It is noted that the Arrhenius plot of the OH stretching peak from 2600 cm-1 to 3800 cm-1 show 40.6  1.2 kJ/mol (Figure 3-7c), which is the same as the molecular water diffusion. This implies that the production of the Si-OH species is either limited or controlled by the diffusion of

+ + molecular water (H2O) or the transport rate of proton or hydronium ions (H /H3O ) is similar to the

57 diffusion of molecular water. The former would accelerate the hydrolysis reaction (Eq.1) and the latter would accelerate the reaction of ion-exchange with sodium ions (Eqs. 2 and 3).

The 퐸푎 value determined for diffusion in SLS glass under vapor phase hydrothermal conditions is lower than the activation energy determined for pure silica glass. The 퐸푎 for water diffusion in the pure silica glass was reported to be ~70 kJ/mol by Wiederhorn et al and Zouine et al in the same temperature range and saturated vapor pressure.16,106 For reactions between water and silica glass at 400 C - 1000 C, the activation energy for water penetration into the glass and formation of Si-OH was found to be ~83 kJ/mol or higher.107 It should be noted that during the reaction between water and silica glass in their studies, Si-OH is formed as a reaction product from the hydrolysis reaction.

The lower activation energy determined for SLS glass must be related to its alkali content and silicate network. Leaching of Na+ in SLS glass has been proven to assist the water diffusion into its silica network.97 The ion-exchange between Na+ and hydronium ions results in the formation of Si-OH from NBO sites and molecular water in the glass network. The silicate network of SLS glass is also quite different from the silica network of pure silica glass. Based on molecular dynamics (MD) simulations, Tilocca et al suggested that the addition of modifiers like Na+ can change the ring size distribution in the Si-O-Si network, which can create more open pathways for water molecules to move inside the glass.108 MD simulations also suggest that water will preferentially interact with the Na+-NBO site over the BO site at the surface to allow water

109 molecules to penetrate into the glass. Thus, the low 퐸푎 value implies a reaction-assisted penetration of water molecules into the SLS glass network and formation of Si-OH. Pure silica glass does not have leachable monovalent ions like Na+, which makes it thermodynamically difficult for molecular water to diffuse into the glass network and form Si-OH groups.

The changes in the silicate network and increase in the amount of hydrous species in the surface region of SLS glass after hydrothermal treatment can explain the lower hardness of the

58 surface layer (Figure 3-4). The hydrated silica-like layer with the lower hardness could suppress crack opening during the Vickers indentation test (Figure 3-5b). Based on the hypothesis involving the leachable sodium ions, the loss of mechanochemial wear resistance at high relative humidity can be explained by the depletion of sodium ions and the modification of the subsurface silicate network after hydrothermal treatments (Figure 3-6).

Figure 3-8. SFG spectra of SLS glass before and after hydrothermal treatment at 150 C for 24 hours, 48 hours and 72 hours respectively.

SFG is very sensitive to chemical environments of the hydrous species (Si-OH and H2O) arranged noncentrasymmetrically at the surface and in the subsurface region of SLS glass.95,110 The

SO2 dealkalization process of the manufactured float glass produces the subsurface hydrous species arranged or oriented noncentrosymmtrically which can be selectively detected with SFG as multiple sharp peaks in the OH stretch region.110 Figure 3-8 shows the effect of hydrothermal treatment on these hydrous species in a 1mm thick SLS float glass. Before hydrothermal treatment, this SLS float glass shows five OH peaks at ~3180 cm-1, ~3392 cm-1, ~3552 cm-1, ~3728 cm-1, and

59 ~3920 cm-1. These SFG peaks represent hydrous species in the subsurface of SLS glass rather than physisorbed water molecules on top of the glass.98,110,111 SFG intensities depend on the noncentrosymmetrically-ordered species in a complex manner; so, the accurate quantification of the detected species is difficult. Instead, the most useful information can be deduced from the peak positions and relative intensities of the detected OH vibrations.95,110

The peak at ~3728 cm-1 must be the free OH groups without hydrogen bonding interactions with surrounding oxygen species.99 The peak above 3800 cm-1 is proposed to originate from the hydrous species (probably OH group) within the strained Si-O-Si network.110,112–115 These strained sites in the silicate network would be more reactive than normal sites with bond distances and angles close to the equilibrium values; thus, they are expected to react more readily with water during the hydrothermal treatment, which would result in the selective disappearance of this peak after the hydrothermal treatment.

The peaks at <3650 cm-1 are the hydrous groups with varying degrees of hydrogen bonding interactions.101,116,117 In Figure 3-8, the SFG peak at ~3392 cm-1 is increasing compared to other peaks; similarly, the ATR-IR spectra in Figure 3-7b show the largest increase at ~3400 cm-1. The

OH stretch peak position can be correlated with the O-HO distance using the empirical relationship found from minerals.101,110 Using the empirical relationship found in the literature,101,110 the component near 3400 cm-1 could be attributed to the hydrous species with hydrogen bonding interactions with the O-HO distance of ~0.28 nm. This value is very close to the equilibrium distance between two oxygen atoms connected via hydrogen bonds in liquid water;118,119 but this does not assure that the only species contributing to the 3400 cm-1 peak is molecular water. Not only the OH stretch peak position of molecular water,98,120 but also the OH stretch peak of the silanol group, vary over the same spectral range depending on the hydrogen bonding strength or distance.

60 With this spectral interpretation, we note that the distribution of the SFG OH peak intensities (Figure 8) is shifted to the lower wavenumber side, implying the growth of populations of hydrous species with shorter O-HO distances (3400 cm-1 = 0.28 nm and 3250 cm-1 = 0.27 nm).101,110 This finding might provide an explanation for the loss of mechanochemical wear resistance. Density functional theory (DFT) calculations reported that at the transition state of the proton-catalyzed hydrolysis of the Si-O-Si linkage, the distance between the oxygen of a hydronium ion and the BO atom is about 0.25nm.121

One potential effect of the shear force exerted by the counter-surface during the wear test could be the distortion of the silicate network of SLS glass at the sliding interface, so that the distance between the hydrous species and the BO is reduced further toward 0.25 nm. If so, the hydrolysis reaction could be facilitated during the interfacial shear. If the hydrous species with short O-HO distances are highly populated at the sliding interface, the transition state for hydrolysis could be reached with a small distortion of the silicate network. This hypothesis could be one possible explanation for the increased wear of the hydrothermal-treated surface at high humidity (Figure 3-6). Further details of this mechanochemical reaction mechanism could be revealed through computation studies such as molecular dynamics simulations with reactive force fields.

Conclusion

The surface of SLS float glass was probed and modified by hydrothermal treatment in steam which involved penetration and diffusion of water molecules, hydrolysis reaction of bridging oxygen, and ion-exchange between Na+ and hydronium ions/protons. It was found that hydrothermal treatment at >150oC can be used to reveal the subsurface damage of SLS glass since the area with subsurface damage or residual stress is preferentially hydrolyzed during the

61 hydrothermal treatment. Hydrothermal treatment also creates a highly-hydrated silica-like layer with very little sodium. While the hardness of this modified surface layer is smaller than that of the bulk, the overall indentation fracture toughness of the glass surface increases. The hydrothermally-treated surfaces, unlike pristine SLS glass, become more susceptible to mechanochemical wear at high humidity. A hypothesis explaining the mechanochemical reactions leading to wear is proposed.

Chapter 4

Thermal poling of soda lime silica glass with nonblocking electrodes –: Effects of sodium ion migration and water ingress on glass surface structure and wear behavior

Reproduced with permission from Wiley: Luo, J.; He, H.; Podraza, N. J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 1: Effects of Sodium Ion Migration and Water Ingress on Glass Surface Structure. J. Am. Ceram. Soc. 2016, 99 (4), 1221–1230. He, H.; Luo, J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 2: Effects on Mechanical and Mechanochemical Properties. J. Am. Ceram. Soc. 2016, 99 (4), 1231–1238.

Overview

In this study, thermal poling is used to modify the sodium concentration and its in-depth profile at the surface of commercial soda-lime-silica (SLS) float glass. It is well accepted that sodium is largely responsible for the surface reactivity of glass, in general; not only the sodium itself, but also the non-bridging oxygen sites sodium creates in the glass network structure.

Although high temperature dealkalization and low-temperature acid leaching processes are often used to reduce sodium surface reactivity, the thermal poling process is unique in its ability to modify the distribution of sodium in both surfaces without the need for external chemical reactants. Here, a commercial SLS float glass was thermally poled using non-blocking electrodes in air. The Na+-depleted anode surface and the Na+-gradient cathode surface were characterized using a variety of methods to find the compositional, structural and morphological effects of poling. Of particular significance is the use of non-destructive vibrational spectroscopy methods, which can lead to new and improved understanding of water interactions with sodium and its sites

63 in the glass. It is shown that after poling, the Na+-depleted glass network on the anode side undergoes a condensation reaction to eliminate non-bridging oxygen sites accompanied by some

SiOH formation. More interesting was the relatively unreactive, water-free Na+-gradient cathode side surface where water and hydroxyl species did not increase. This process could offer new ways to modify glass surface chemistry and structure, or at least to better understand sodium effects on surface properties through ion-transport control of Na surface concentration, creation of unique sodium in-depth gradients and modification of the network structure. In a Part 2 companion paper, the mechanical properties of these thermally-poled surfaces are reported.

Introduction

Controlling alkali ion concentrations in the surface region of silicate glasses is of great interest and importance for many engineering applications since it plays pivotal roles in chemical and mechanical properties of glass materials.39,122 In the case of soda lime glass which is widely used for windows and containers, high-temperature SO2 and fluorocarbon treatments are commonly used during the manufacturing process to reduce the sodium concentration in the surface region.123–

126 Acid treatments and organic coatings can also be applied to further deplete leachable sodium ions or prevent sodium reactions at the glass surface.42,127–129 These treatments alter not only the sodium ion concentration but also the chemical structure of the silicate network in the glass surface, which greatly affects how the glass surface reacts with the environment, especially water in ambient air.33,78,130 Thus, fundamental surface studies of sodium-containing glasses are needed to fully understand and control the surface properties of soda lime glass.131,132

Recently, thermal poling has been explored as a means to both study and alter the mobile ion distribution near the glass surface.133–135 In this process, a glass sample is heated to a temperature

64 at which network modifier ions have sufficient ionic mobility in the glass when a high voltage direct current (DC) bias is applied to move cations from the anode side to the cathode side. When the temperature is lowered while maintaining the DC bias, the migrated cations are trapped, forming an ion depletion layer in the anode-side glass surface and an ion accumulation layer on the cathode-side glass surface. The electric field gradient in the glass surface subsequently formed by the ion migration induces a second-order optical nonlinearity (SON) effect such as second harmonic generation (SHG) which is otherwise not expected for amorphous glass materials.136–138

Although the charge distribution and SON properties of thermally-poled glasses are well documented in the literature, the chemical structure of the poled glass surface is not fully understood.134,139,140 During the poling process, the electric-field induced migration of sodium ions can result in various structural changes in the glass network. In the anode-side glass surface, the sodium ion depletion would leave negatively-charged non-bridging oxygen (NBO) groups. These negatively charged NBO groups could lose electrons which could be ejected to the anode, leaving neutral NBO groups. If two NBO groups are in proximity, they could react to form Si-O-Si linkages and release oxygen to the environment. When a non-blocking electrode is used,133 water could ingress into the sodium depleted surface and react with NBO groups forming Si-OH. Then, hydroxide anions would be produced which could further react with the Si-O-Si network. Other cationic species could be formed at the anode surface by ionization of gas phase molecules;141,142 these cations could enter the glass surface to compensate the negatively-charged groups produced upon sodium ion depletion. The chemical environment during the poling also affects the cationic species involved in charge compensation inside the glass poled at the anode side.143 On the cathode side, sodium ions accumulate in the glass if a blocking electrode is used.134,144 If a non-blocking electrode is used, sodium ions can exit the glass and be reduced to metallic sodium by electrons at the cathode, and further react with molecules from the gas phase.145 These poling-induced

65 compositional and structural changes are expected to impact the chemical and mechanical properties of the poled glass surfaces.146

In this paper, the silicate network structure and hydrous species (Si-OH and H2O) in the soda lime silica (SLS) float glass, thermally-poled using non-blocking electrodes in air, were studied. The non-blocking electrode was used so it could be easily removed for spectroscopic

(Part-1) and mechanical (Part-2) studies of the poled glass surface. The chemical composition, thickness, and surface roughness of the poled layers were characterized with cross-sectional energy-dispersive x-ray (EDX) spectroscopy, spectroscopic ellipsometry (SE), and atomic force microscopy (AFM). The Si-O-Si network and amount of hydrous species were studied with specular reflectance (SR) and attenuated total reflectance (ATR) infrared (IR) spectroscopy. The

OH groups (Si-OH and H2O) with distinct hydrogen bonding interactions within the poled layer were studied with vibrational sum frequency generation (SFG) spectroscopy. SFG is another type of SON effect which requires noncentrosymmetry like SHG.147,148 While SHG is responsive to the overall field gradient without any molecular specificity, SFG provides molecule-specific information since it probes vibrational modes of functional groups.148,149 Based on all results of these analyses, a comprehensive model is proposed for the structural modifications which occur inside soda lime glass upon sodium migration via thermal poling.

Experimental methods

Thermal poling experiments were performed on commercial soda lime silica (SLS) float glass (Asahi Glass Co, Japan) with a thickness of 0.7mm. The nominal bulk composition in weight percentage is 74% SiO2, 13.5% Na2O, 10.5% CaO, 1.3% Al2O3, 0.2% MgO, 0.3% K2O, and 0.2%

SO3. Only the air-side of the float glass was analyzed. A schematic diagram of the thermal poling

66 setup is shown in Figure 4-1. All samples were cleaned by first rinsing with water and ethanol, and then by UV-ozone to remove organic contaminants on the glass surface before thermal poling. A stainless steel plate and a highly oriented pyrolytic graphite (HOPG) disc were used as an anode and a cathode, respectively. The electrodes were in physical contact with the glass surface by its own weight; thus, there were physical gaps between the electrode and glass surfaces, which allowed the environmental gas species to interact with the surface during the thermal poling process. The thermal poling temperature was held constant at 200 oC,141,145 well below the strain point for this glass (~500 oC). The temperature of the sample was measured by a thermocouple which was in intimate contact with the cathode (ground electrode). After the sample was equilibrated at 200 oC in ambient air, a +2 kV DC bias voltage was applied to the anode. After a given poling time, the furnace temperature was lowered until the cathodic current decreased below the detection limit of the picoammeter while the DC bias was held constant at +2 kV. Figure 1b shows an example of the temporal profiles of temperature (T), bias voltage (V), and measured current (A).

After thermal poling, white particulates were observed on the glass surface inside the area contacting the cathode. An AFM topography image of the particulates on the glass surface is shown in the inset of Figure 4-2. In SR-IR analysis of this surface, a sharp peak was observed at ~1430

67 cm-1 corresponding to the asymmetric vibrational mode of carbonate species (Figure 4-2).150 Since the non-blocking electrode geometry was used, sodium at the cathode side reacted with CO2 forming sodium carbonate (Figure 4-1a).151 The white particulates could be removed by simple rinsing with a cotton tip wet with water. All cathode-side surfaces described in this paper were cleaned to remove the sodium carbonate deposits prior to analysis.

The energy-dispersive x-ray spectroscopy (EDX) analysis was performed using a FEI

Quanta 200 Environmental SEM system to measure the chemical composition changes in the soda- lime silica glass due to thermal poling. Samples were prepared by fracturing through the poled area.

The samples were analyzed without a conducting layer deposition and the water vapor pressure was kept at 100 Pa to mitigate surface charging problems.

Spectroscopic ellipsometry (SE) measurements were performed using a single rotating compensator multichannel ellipsometer (Alpha-SE, J. A. Woollam Co., Inc.).152,153 Ellipsometric spectra were collected at a 70o angle of incidence and a piece of frosted tape was attached on the

68 backside of the measured surface to avoid interference by incoherent backside reflections. The ellipsometric spectra were fitted using a least squares regression with an unweighted error function154 to a multiple layer structural model with fit parameters defining layer thicknesses, material volume fractions, and analytic expressions for the index of refraction (n). A Bruggeman effective medium approximation is used to represent n for composite and interfacial layers.155 For the interface layers, material fractions were fixed at 0.5 of the overlying and underlying layers.

Spectra in n for bulk glass, as well as the untreated and anode side surface layers, were represented using Sellmeier oscillator parameterizations.156 After n for the bulk glass was extracted, it was fixed in the structural model of the cathode and anode side samples. A sensitivity limit to differences in n of ~0.002 between adjacent layers was deduced from simulations of ellipsometric spectra.

The surface topography and adhesion force map were analyzed through peak-force tapping mode imaging with a Bruker Icon atomic force microscope (AFM). The deflection sensitivity of the silicon tip and the spring constant of the cantilever were calibrated before the test. The surface was pre-cleaned by 200-proof ethanol and UV-ozone.

Specular reflectance IR (SR-IR) spectroscopy was performed with a ThermoNicolet 670

FTIR spectrometer. All IR spectra were taken at a 40o incidence angle. A gold mirror was used as a reference background. Note that the shoulder peak at ~1250 cm-1 is due to the anomalous dispersion effect of the complex refractive index in the reflection spectrum (Figure 4-2), not an absorption band.96

Attenuated total reflectance infrared (ATR-IR) spectroscopy was carried-out with a Bruker

Hyperion 3000 Microscope equipped with a Ge ATR crystal accessory with 45° incidence angle.

The glass surface was pressed to the Ge crystal with a 600 N force. The information depth of ATR-

IR was calculated to be 0.5-0.75 μm in the OH stretch region.52

Vibrational sum frequency generation (SFG) spectroscopy was employed to identify hydroxyl groups with different hydrogen-bonding interactions. The detailed set-up of the SFG

69 spectroscopy system is described elsewhere.72 Visible pulses (532 nm) and tunable IR pulses (2.5

– 10m) generated with an EKSPLA laser system were spatially and temporally overlapped at the glass surface. The incident angles of visible and IR pulses were 60o and 56o with respect to the surface normal, respectively. The SFG signal intensity was normalized with the intensities of input visible and IR beams. The polarization combination for the spectra collected was s for SFG signal, s for visible beam, and p for IR beam (ssp).

Results & Discussions

Chemical composition and thickness of the thermally-poled surface layers

The atomic concentrations of modifier elements (Na, Ca, Mg and Al) as well as network atoms (Si and O) were measured in the outermost 20 μm of the anode- and cathode-side surfaces after 40 minutes of poling; these profiles, along with the pristine glass reference profile, are presented in Figure 4-3. The pristine surface showed no concentration gradient near the surface for all elements of the glass within the spatial resolution of EDX. The charging-induced signal decay of mobile sodium ions157,158 was not observed at the edge of the glass cross-section in our EDX measurement condition (water partial pressure = 100 Pa).

Figure 4-3. (a), (b), (c): EDX line profiles of cross-sections of pristine and poled surfaces. Poling time = 40 min.

On the anode side, the formation of a Na+-depleted layer was clearly observed (Figure 4-

3b). For 40 min poling, the thickness of this layer was about 3 m. The concentration of other multivariate cations did not change substantially from the bulk values, indicating that the monovalent sodium (Na+) ions are the primary mobile ions responding to the poling condition used in this study. Beneath this ‘completely’ Na+-depleted layer, a ‘marginally’ Na+-depleted region

(about 10 m thick) could be noted.

On the cathode side, the sodium ion concentration showed a gradient profile (Figure 4-3c).

The sodium concentration increased slightly in the sub-surface reaching a maximum value at 3-4

m from the surface and then gradually decreased at the surface to a value lower than the bulk concentration. This gradient profile was formed because a non-blocking electrode was used as the cathode.159 When the glass surface was subjected to the negative electrical potential, Na+ ions in the glass were extracted from the glass surface and then reduced to the metallic state by reaction with electrons supplied from the cathode (as recorded as cathodic current in Figure 4-1b). The sodium metal on the glass surface further reacted with CO2 from the environment, forming sodium carbonate particulates (Figure 4-2). At the same time, sodium ions were supplied from the bulk via

71 electrical field-induced migration. The balance between these two processes (sodium ion removal out of the surface and supply from the bulk) would determine the final concentration profile of the

Na+ ion seen in Figure 4-3c.

Figure 4-4. Ellipsometric spectra (in  and ) and model fits for (a) pristine, (b) anode-side, and (c) cathode-side surfaces. The poling time was 40 min and the incidence angle of SE analysis was 70o. The lines are least squares fit results. All un-weighted error functions were below 2 x 10-3, indicating good agreement between experimental data and the least squares model fit. (d) Refractive index of bulk glass, pristine surface layer, and Na+-depleted layer as a function of photon energy. The refractive index of the cathode-side surface was almost identical to the bulk value.

The thickness and refractive index (n) of the Na+-depleted layer at the anode side and the

Na+-gradient layer at the cathode side were analyzed with SE (Figures 4-4 and 4-5). For the untreated glass (Figure 4-4a), the structural model consisted of a semi-infinite bulk glass substrate as presented in Figure 5b: 261  4 nm thick untreated surface layer covered with 46  2 nm thick

72 composite layer of bulk glass and 1.93  0.04 % void fraction at the sample / air interface. The origin of voids at the sample / air interface is attributed to contributions from surface roughness or other subtle differences in the optical response arising from a density deficit or strain in that region.

The untreated surface layer might be due to the SO2 dealkalization process during the float glass manufacturing.7

Figure 4-5. Schematics of ellipsometric model for (a) anode (sodium-depleted surface); (b) pristine surface; (c) cathode (sodium-gradient surface).

The sodium-depleted surface at the anode side showed distinct oscillating fringe patterns

(Figure 4-4b). The structural model for the anode side is schematically shown in Figure 4-5a: (1) semi-infinite bulk glass substrate, (2) 90  2 nm thick interface layer, (3) 2417  5 nm thick Na+- depleted layer, (4) 581  6 nm thick interface layer, (5) 1117  3 nm thick composite of the Na+- depleted layer and 0.85  0.07 % void fraction, and (6) 10  1 nm thick bulk glass at the sample / air interface. The regression result showed that the total thickness of the Na+-depleted layer (from

2 to 5) was about 4.2 m, which was in agreement with the EDX measurement (Figure 4-3b), and n was 0.01-0.02 lower than the pristine bulk (Figure 4-4d).

The sodium-gradient within the surface of the cathode side did not show any undulating fringes (Figure 4-4c). The overall refractive index from the model fit was almost identical to the bulk in the visible wavelength region (Figure 4d). The fit result shown in Figure 4c was based on the structural model shown in Figure 5c which consisted of a semi-infinite bulk glass substrate and a 25  1 nm thick composite layer of bulk glass and 94.5  0.2 % void fraction. The large void fraction in the model could be interpreted as the roughness of this surface. Since the Na+ concentration varied gradually (Figure 4-3c), and the average composition of the gradient region was very close to the bulk value, the refractive index change was negligible within the resolution of our experiment and fit method.

74 The surface topography and adhesion force maps of the pristine, Na+-depleted (anode side), and Na+-gradient (cathode side) surfaces are shown in Figure 4-6. The prisitine sample had ring- like features (Figure 4-6a). It was speculated that they might originate from the SO2 sodium dealkalization treatment used in the commercial float glass process.160 The topographic images of the Na+-depleted surface were completely featureless other than a few random bumps (Figure 4-

6c). The Na+-gradient surface showed numerous small bumps (~50 nm wide, ~3 nm tall; Figure 4-

6e). Note that similar bumps were also seen, although less frequently, on the pristine surface upon

+ the SO2 dealkalization process of the float glass. Thus, the topographic features seen in the Na - gradient surface seem related to the removal of sodium from the surface (as seen in Figure 4-2).

The RMS roughness over 600nm  600nm was 0.3, 0.3, and 1.0 nm for the prisine, Na+-depleted, and Na+-gradient surface, respsectively. The significantly larger roughness for the Na+-gradient surface was also consistent with the SE analysis result (Figure 4-5c).

The adhesion force measured at 30% relative humidity condition during the peak-force tapping imaging implied chemical differences among the three surfaces. The average adhesion force was 6.8  0.3 nN for the pristine surface (Figure 4-6b). The Na+-depleted surface showed significantly higher adhesion, 11.1  0.6 nN (Figure 4-6d). The Na+-gradient surface showed a bimodal adhesion force distribution: ~5 nN for the the protruded bump region and ~8.6 nN for the depressed region (Figure 6f; in average, 7.2  1.7 nN).

Silicate network structure in thermally-poled surfaces

In order to understand how the electric field-induced migration of sodium ions altered the silicate network of glass,108 the poled glass surfaces were analyzed with SR-IR. In SR-IR spectra, structural information about the Si-O-Si (BO) network can be deduced from the peak position in

75 the 1060-1120 cm-1 region and from the Si-O stretching vibration of NBO or Si-OH groups which appear as a shoulder at ~940 cm-1.103

The SR-IR spectra of the poled surfaces are compared with those of the pristine surface in

Figure 4-7. For the Na+-depleted anode side, the Si-O-Si peak was blue-shifted from 1070 cm-1 for

-1 the pristine surface to 1096 cm (approaching toward the peak position of pure SiO2, fused quartz, at 1122 cm-1) as the poling time increased. Note that in the SE analysis, n of the Na+-depleted region decreased toward the silica value (1.47). These results implied that the Si-O-Si network becomes more “silica-like” as sodium ions migrated from the surface into the bulk.161 A small growth of the shoulder peak at 940 cm-1 was also observed as the poling time increased. This growth cannot be attributed the increase in the NBO population since Na+ ions were depleted. Instead, this growth must be related to an increase in the Si-OH groups, which was confirmed by ATR-IR (Section

3.3).103,124 Thus, it can be concluded that the Na+-depletion is accompanied by changes in the silicate

76 glass network structure and its terminal sites, most likely converting NBO groups to BO and Si-

OH groups.

In contrast, the SR-IR spectra of the Na+-gradient cathode side surfaces were almost identical to the pristine surface spectrum, regardless of the poling time (Figure 4-7). The only noticeable change was a slight red-shift, for example, ~2 cm-1 decrease after 20min and 40min poling. Since the thermal poling was carried 200 oC (much lower than the glass transition temperature), the Si-O-Si network would not have sufficient thermal energy to fully relax after the field-induced migration of Na+ ions. The local excess or deficiency of Na+ ions could impose a compressive or tensile stress to the surrounding Si-O-Si network, which could alter the Si-O-Si bond angle from its initial state.162 There was no discernable change in the shoulder peak intensity at 940 cm-1.

In order to test if these structural changes are reversible, re-poling under a reverse bias was performed. As shown in Figure 4-8a, the structure of the initially Na+-depleted surface did not change significantly upon applying the reverse poling. This suggested that the “silica-like” network formation upon the Na+-depletion is irreversible. The newly-formed Si-O-Si (BO) network did not dissociate and return to the original NBO state during the reverse poling at 200oC.

Figure 4-8. SR-IR spectra of the poled (solid lines) and then reverse-poled (dashed lines) glass surfaces. Both poling and reverse-poling times were 10 min.

In the case of the Na+-gradient cathode side (Figure 4-8b), the Si-O-Si peak position was shifted from 1068 cm-1 to 1090 cm-1 after reverse-poling for 10 min. This result suggested that the

Na+ ions in the cathode-side surface are still highly mobile and readily migrate back to the bulk when a positive bias voltage is applied. Since the NBO shoulder peak at ~940 cm-1 (as well as the

OH intensity as shown in Section 3.3) did not change in the Na+-gradient surface, the concentration gradient of sodium ions observed EDX (Figure 4-3c) does not seem to be accompanied by the restructuring of the NBO groups.

Hydrous species (Si-OH, H2O) in poled surfaces

Analysis with ATR-IR

ATR-IR was employed to monitor the concentration change of hydroxyl groups and water molecules in the glass surface. The thickness of the poled layer was on the order of microns based on the EDX and SE analyses. The sampling depth of ATR-IR with a Ge crystal is between 1.5 to 3

m in the 4000 – 1500 cm-1 region and the ATR-IR peak intensity is proportional to the concentration of functional groups within this sampling depth.96 Thus, ATR-IR is a good technique to assess changes in the amount of hydroxyl and water in the glass.

Figure 4-9. ATR-IR spectra of (a) Na+-depleted and (b) Na+-gradient surfaces of thermally-poled soda lime glass.

The ATR-IR spectra of the Na+-depleted and Na+-gradient surfaces are presented in

Figure 4-9. While very minimal spectral changes were observed for the Na+-gradient surface

(Figure 4-9b), the Na+-depleted surface spectra showed the significant growth of the broad peak spanning from 3700 to 2600 cm-1 with increasing poling time (Figure 4-9a). This is direct

79 evidence that the OH concentration in the Na+-depleted layer is significantly increased, which is consistent with the growth of the 940 cm-1 shoulder peak in SR-IR spectra of anode-side surfaces shown in Figure 4-7. For a poling time longer than 20 min, even the bending vibrational mode of molecular water at 1640 cm-1 was detected. These results indicate that proton and water enter into the Na+-depleted region to compensate the field-induced migration of cations into the bulk. In the cathode side, however, the ingress of water species into the Na+-gradient surface was negligible.

Speciation with SFG

Vibrational SFG spectroscopy is well known for its ability to detect molecular species at interfaces.163 In SFG, two photons irradiating the surface are combined and emitted as a single photon whose frequency is the sum of two input photons.163 When one of the irradiating photons is in the mid-IR range, then SFG can probe vibrational modes of species at the interface. While SHG provides information on the strength of the frozen dc electric field in the poled layer, SFG provides vibrational spectroscopic information of chemical species in this frozen electric field region.164,165

Note that the interface does not necessarily mean a sharp boundary of two phases, which would be mathematically described as a step function. If a compositional profile of a constituent element varies gradually across this boundary, the entire gradient region can be the interface responsible for the SFG response.

Figure 4-10. SFG spectra of (a) Na+-depleted and (b) Na+-gradient surfaces of thermally-poled soda lime glass.

For the poled glass surface, SFG signals (ISFG) can originate from two sources – the noncentrosymmetrically arranged functional groups and the electric field effects:166,167

2 (2) (3) 2 퐼푆퐹퐺 ∝ 퐸휔푉퐼푆+휔퐼푅 ∝ [( +  퐸퐷퐶)퐸휔푉퐼푆퐸휔퐼푅] (1) where 휔푉퐼푆 and 휔퐼푅 are the frequencies of visible and IR photons irradiating to the surface,

respectively, and 휔푉퐼푆 + 휔퐼푅 is the frequency of the emitted photon. 퐸휔푉퐼푆+휔퐼푅, 퐸휔푉퐼푆, and 퐸휔퐼푅 are the electric fields of sum-frequency, visible, and IR photons, 퐸퐷퐶 is the electric field produced by the concentration gradient of ions inside the surface, and (2) and (3) are the second-order and third-order nonlinear susceptabilities of vibrational modes in resonance with the incoming IR frequency. The (2) term can have non-zero values only for the noncentrosymmetrically-arranged vibrational modes. Typically, molecules at the surface are responsible for the (2) response.168,169

The molecules in the subsurface region (which are responsible for (3) term) of a bulk material with random structures are normally SFG-inactive because (3) alone does not meet the noncentrosymmetry requirement. However, if the subsurface is poled and a net EDC is formed, then

(3) 170–174 the  퐸퐷퐶 term will be SFG-active since EDC is noncentrosymmetric. The orders of magnitude for molecular hyperpolarizability for (2) and (3) terms are ~10-32 cm4V-1s-1 and ~10-39

5 -2 -1 175 8 143,171 cm V s . Since typical EDC of thermally-poled surfaces is on the order of ~10 V/m, the

(3) (2) hyperpolarizability of the  퐸퐷퐶 term becomes comparable to that of the  term. Thus, the

(3)  퐸퐷퐶 term can provide structural information about the subsurface OH groups present in the

33,143 region where 퐸퐷퐶 is generated due to the ion concentration gradient. For a glass surface with

33 negligible amount of ions (such as fused silica), EDC is negligible before thermal poling.

The SFG spectra of the Na+-depleted and Na+-gradient are shown in Figure 4-10. The insets show the peak deconvolution with Lorentzian functions176 for the surfaces poled for 40 min. A total of 5 peaks were used for the OH stretch region and three peaks for the CH stretch region. The CH stretch peaks come from adventitious hydrocarbon species adsorbed on the glass surface from airborne organic contaminants. The small negative doublet peaks near ~2350 cm-1 are simply because the IR power delivered to the sample is attenuated through the absorption by CO2 in air; thus, those two peaks were ignored in peak fitting. The same peak fitting method was applied to all spectra and the fit results are summarized in Table 4-1 in the Supporting Information.

Figure 4-11. (a,b) ATR-IR and (c,d) SFG spectra of Na+-depleted (a,c) and Na+-gradient surfaces of thermally-poled soda lime glass before and after post-annealing at 200oC for 2 hours in ambient air.

The multiple SFG peaks in the 3000-3700 cm-1 region for the pristine glass surface are consistent with the previously reported SFG spectra of soda lime float glass.33 These peaks are relatively sharper than typical SFG signals observed for the water molecules at the air/liquid interface or adsorbed on the fused quartz surface.117,177 The narrow peak width can be interpreted as the lack of long-range hydrogen bonding interactions.99,105,178 Thus, they could not be attributed to the water layer adsorbed on the glass surface ((2) responses). Instead, the sharp OH peaks could be attributed to the Si-OH groups converted from NBO upon leaching of Na+ and neutralization by

+ (3) H , i.e.,  퐸퐷퐶 responses. Depending on the size of the original NBO site, they can have hydrogen-bonding interactions with additional water molecules diffused from the gas phase into

83 the glass. Thus, the compensation of the Na+ ions separated from the NBO sites could be considered

+ + 8,179 to be stoichiometrically equivalent to either proton or hydrated protons (H3O , H5O2 , etc.).

Their relative abundance will depend on the size of the original NBO site and the activity (or concentration) of water in the glass. The SFG peak positions of these species are governed by the strength of hydrogen bonding interactions.102,180,181 As the hydrogen bonding interaction increases, the red-shift of the OH peak position becomes larger.102,182 Thus, the small variance in the O-HO distance will result in a larger broadening in the peak width for the lower frequency peaks, compared to the higher frequency ones with weaker interactions.

An extremely broad peak centered at 2650-2850 cm-1 species was detected for the poled surfaces, which was not noticed in the previous report for the pristine soda lime float glass.33 This peak must be the strongly hydrogen-bonded Si-OH species in the bulk. This component is observed as big as the >3000 cm-1 peaks in transmission IR spectra of soda lime glass.97 It is also detected as a minor component in ATR-IR spectra (for example, see Figure 4-9a). It appears that the 퐸퐷퐶 field formed inside the glass upon thermal poling makes this species SFG-active.

The weak peak at 3750-3800 cm-1 was observed for the pristine glass surface. Its intensity did not grow with thermal poling. This peak could be ascribed to the silanol groups at the glass surface, although its peak position is slightly higher than typical values for the free OH group (3700

– 3740 cm-1) without hydrogen bonding interactions with neighboring molecules. The origin of this slight blue-shift is not well understood at this moment.

Figure 4-10 and Table S1 clearly show that four peaks at 2650-2850 cm-1, 3200-3350 cm-1,

3500-3570 cm-1, and 3650-3700 cm-1 are growing with thermal poling. The ATR-IR analysis results

(Figure 4-9) revealed that the concentration of the OH species increases in the Na+-depleted surface

(anode side), while that does not change in the Na+-gradient surface (cathode side). Thus, the

+ (3) increase in the SFG signal for the Na -gradient surface must be mostly due to the  퐸퐷퐶 effect and the SFG change for the Na+-depleted surface is the combination of both OH concentration

(3) -1 change and the  퐸퐷퐶 effect. It is noted that the relative intensity of the 3500-3570 cm , compared to other peaks, is higher for the Na+-depleted surface (anode side) than the Na+-gradient surface (cathode side). This difference might imply that the 3500-3570 cm-1 species are the newly

+ + formed OH groups by the reaction of NBO with H or H2O upon the migration of Na ions into the bulk.

Another important observation is that the SFG peaks of the adventitious hydrocarbon contaminants on the surface appeared as positive peaks above the OH peaks in the Na+-depleted surface spectra and negative peaks in the Na+-gradient surface spectra (Figure 4-10 inset). This difference supported that the enhanced SFG signals for the thermally-poled surfaces are largely due

(3) (2) (3) 2 to the  퐸퐷퐶 effect. As shown in eq. (1), the SFG signal is proportional to ( +  퐸퐷퐶) .

Here the sign of the adsorbed hydrocarbon signal does not change, but the sign of the 퐸퐷퐶 field changes depending on the poling direction. Thus, the amplitude of the total signal depends on the relative phase relation between these two components.183 The fact that Na+-depleted and Na+- gradient surfaces showed hydrocarbon peaks in the opposite directions with respect to the OH signal implied that the 퐸퐷퐶 fields formed inside the glass surface upon thermal poling are opposite for the anode and cathode sides.

The thermal stability of the Na+-depleted and Na+-gradient layers was tested by annealing the poled glass at 200oC, the same temperature used for thermal poling, without applied electrical bias. At this temperature, sodium ions have enough mobility, so they can diffuse to reduce the concentration gradient. Figure 4-11 presents the ATR-IR and SFG spectra of the poled samples before and after the 200oC heat treatment in ambient air. In the ATR-IR spectra of the Na+-depleted surface, the 200oC heating caused only a minor decrease in the OH concentration (Figure 4-11a).

In other words, 200oC was not high enough to induce dehydration reactions of the OH groups inside the Na+-depleted layer. The SFG intensities showed a much larger decrease after heating at 200oC

85 (Figure 4-11c). Two possible reasons can be considered -- the slight reduction of the OH

+ concentration or the reduction of the 퐸퐷퐶 field due to repopulation of Na ions via diffusion from the bulk. Although it is difficult to determine which factor is more dominant, one can speculate that the latter is the case since the relative intensities of different OH species did not change substantially.

For the Na+-gradient surface, no additional OH groups were formed during the poling process (Figure 4-9b). So, there was no change in the OH intensity in ATR-IR upon heating to

200oC (Figure 4-11b). However, the OH SFG intensities dropped significantly after heating at

200oC (Figure 4-11d). This must be solely due to the repopulation of Na+ ions through diffusion,

171 reducing the 퐸퐷퐶 field in the surface. The structural distribution of the OH groups in the glass surface did not seem to vary since the relative intensities of the OH SFG signals did not change.

Effects of thermal poling on mechanochemical wear in humid ambient

In order to evaluate the mechanochemical wear properties of thermally-poled SLS glasses, the wear of glass surfaces is tested in various RH conditions. At an applied load of 0.2 N, the nominal Hertzian contact pressure is ~300 MPa; in our previous study, it was shown that the glass surface was not visually damaged within the spatial resolution of optical profilometry as long as the surface was lubricated by a monolayer of adsorbed alcohol molecules at this contact load.Error! Bookmark not defined. In a dry N2 environment, the SLS glass surface was scratched adly and the wear depth reached up to ~10 μm, regardless of thermal poling conditions (data not shown). The wear depth was so large and chaotic, no discernable difference could be found before and after thermal poling. When water vapor is introduced, the wear pattern changes from the purely mechanical process producing rough and deep scratches to a mechanochemical process leaving much smoother and shallower (less than a few hundred nanometers) wear tracks. This transition is

86 similar to the behavior observed for pristine glass surfaces with different compositions.Error! Bookmark ot defined.,Error! Bookmark not defined.,Error! Bookmark not defined. However, the RH dependence was different for the anode and cathode side surfaces.

Figure 4-12a and 4-12b compare the line profiles of wear tracks produced on the pristine and poled SLS glass surfaces at 40% and 90% RH, respectively. The wear track of the Na+-depleted anode-side glass surface is increased from ~100 μm wide at 40% RH to ~300 μm wide at 90% RH.

The wear depth at 90% RH is deeper for the Na+-depleted surface than the pristine surface. On the other hand, both wear depth and width of the Na+-gradient cathode-side glass surface are decreased as RH increases from 40% to 90%. Figure 4-12c compares the wear volumes of the pristine and thermally-poled glasses measured at 40% and 90% RH conditions. The data clearly show that as

RH increases, the wear volume of the Na+-depleted surface increases while the that of the Na+- gradient surface decreases to a value even smaller than that of the pristine surface.

Figure 4-12. Characteristic line profiles of wear track of Na+-depleted surface, Na+-gradient surface and pristine SLS glass substrate when rubbed with a pyrex ball for 400 cycles under an applied load of 0.2 N in (a) 40% RH (b) 90% RH conditions. (c) The overall wear volume of treated SLS glass substrate in humidity conditions.

Since these experiments are conducted with the same ball materials at the same applied load in the same RH conditions, the observed difference in RH dependence must originate from the alteration of the Na+ concentration profile by thermal poling. In our previously publications, it was hypothesized that the presence of hydronium ions interacting with Si-O- groups at the Na+-leached sites, forming Si-OHOH2, could be responsible for the enhanced wear resistance of SLS float glasses at high RH conditions.Error! Bookmark not defined.,Error! Bookmark not defined. This hypothesis was ggested based on the positive correlation in RH dependences of wear resistance and multiple OH

88 SFG peaks observed for the pristine SLS float glass surface.Error! Bookmark not defined. However, the data hown in Figure 4-12 could not be fully explained with this hypothesis.

From attenuated total reflectance infrared (ATR-IR) spectroscopy analyses of the thermally-poled SLS glass surfaces in the Part-1 of this series, it was found that the Na+-depleted

+ surface contains more hydrous species (Si-OH and H2O) compared to the pristine and Na -gradient surfaces.24 However, its wear resistance at 90% RH is worse than the pristine and Na+-gradient surfaces (Figure 4-12c). In Part-1, the specular-reflection infrared (SR-IR) spectroscopy analysis also revealed that the Na+ depletion during the thermal poling is accompanied by the restructuring of the Si-O-Si network toward a ‘silica-like’ structure.24 The network resembling the pure silica network would be subject to a stress corrosion effect.Error! Bookmark not defined. Thus, the increase of ear at high RH for the Na+-depleted surface appear to be consistent with the RH dependence of the silica glass containing no Na+ ions.Error! Bookmark not defined. Based on these analyses, it could be concluded that for the Na+-depleted anode-side surface, the restructuring of the glass network to the silica-like structure is the main cause for the reduction of wear resistance at high humidity.

In the case of the cathode-side surface, the cross-sectional energy dispersive x-ray (EDX) spectroscopy analysis in Figure 4-3 showed that the Na+ ion concentration gradually changed from its bulk value to a slightly higher value in the subsurface region and then a slightly lower value at the surface.24,Error! Bookmark not defined. The SR-IR analysis of this surface showed that the a+-gradient profile was produced without substantial restructuring of the Si-O-Si network.24 As shown in Figure 4-12, this surface exhibits superior wear resistance in the high RH condition, while the Na+-depleted silica-like surface of the anode side shows poor resistance against wear.

Note that the same phenomenon was not observed for Na+-containing aluminosilicate glass,

Na+/K+-exchanged chemically-strengthened aluminosilicate glass, and borosilicate glass containing sodium-rich phases.Error! Bookmark not defined. Also, this enhanced damage resistance at igh RH was observed for the lateral shearing of the glass surface, but not for the normal

89 indentation test.Error! Bookmark not defined. Thus, the good wear resistance at high RH must be a onsequence of dynamic interactions of the Na+ ions in the SLS network upon adsorption and absorption of water from the environment. The exact nature of these dynamic interactions is not fully understood at this time.

Conclusion

Through the use of non-blocking electrodes, we studied the effects of thermal poling with on the surface composition, morphology, and structure of commercial soda-lime silicate float glass.

On the anode side, a Na+-depletion layer was formed within the glass surface. On the cathode side, a Na+-gradient layer was formed. The Na+-depletion was accompanied by irreversible restructuring

+ of the silicate network and an increase in the Si-OH and H2O concentrations. The Na -gradient layer on the cathode side did not exhibit substantial restructuring of the network nor the uptake of water from the environment. SFG analyses of these surfaces showed at least three distinct hydrous species (3200-3270 cm-1, 3500-3550 cm-1, and 3650 – 3700 cm-1) in the glass surface, in addition to the strongly hydrogen bonded bulk species (<2800 cm-1). The relative abundances of these species were different for Na+-depletion and Na+-gradient layers.

Supporting information

The peak fitting results of SFG spectra in Figure 4-10 is summarized in Table 4-1. Five peaks that represent OH groups in different chemical environment, are used to fit each curve. The

Lorentzian line shape is applied in all the curve fitting since it works better to fit the SFG signal

90 that come from multiple domains176 (glass-air interface and subsurface region with concentration

gradient of sodium ions).

Table 4-1. Position, width, and height of OH stretch peaks observed in SFG spectra Poling time Peak position (cm-1) / FWHM (cm-1) / peak height (a.u.) for Na+-depleted surface 0 min 3810 / 32 / 0.27 3650 / 45 / 0.16 3550 / 60 / 0.58 3270 / 43 / 0.21 2650 / 170 / 0.46 5 min 3780 / 17 / 0.30 3710 / 32 / 0.69 3530 / 47 / 2.2 3240 / 73 / 1.4 2680 / 160 / 0.29 10 min -- 3700 / 38 / 8.0 3510 / 82 / 9.9 3220 / 150 / 7.3 2680 / 200 / 1.8 20 min -- 3660 / 37 / 14 3500 / 96 / 17 3200 / 180 / 12 2690 / 220 / 2.7

Poling time Peak position (cm-1) / FWHM (cm-1) / peak height (a.u.) for Na+-gradient surface 0 min 3810 / 32 / 0.27 3650 / 45 / 0.16 3550 / 60 / 0.58 3270 / 43 / 0.21 2650 / 170 / 0.46 5 min -- 3650 / 35 / 1.5 3578 / 56 / 1.9 3360 / 120 / 2.7 2830 / 280 / 0.75 10 min -- 3660 / 39 / 4.6 3550 / 71 / 2.4 3310 / 150 / 3.2 2780 / 270 / 1.5 20 min -- 3660 / 38 / 13 3550 / 77 / 7.4 3280 / 190 / 7.7 2850 / 260 / 3.3

Chapter 5

Chemical structure and mechanical properties of soda lime silica glass surfaces treated by thermal poling in inert and reactive ambient gases

Reproduced with permission from Wiley: Luo, J.; Bae, S.; Yuan, M.; Schneider, E.; Lanagan, M. T.; Pantano, C. G.; Kim, S. H. Chemical Structure and Mechanical Properties of Soda Lime Silica Glass Surfaces Treated by Thermal Poling in Inert and Reactive Ambient Gases. J. Am. Ceram. Soc. 2018. doi: 10.1111/jace.15476

Overview

This study employed thermal poling at 200 C as a means to modify the surface mechanical properties of soda lime silica (SLS) glass. SLS float glass panels were allowed to react with molecules constituting ambient air (H2O, O2, N2) while sodium ions are depleted from the surface region through diffusion into the bulk under an anodic potential. A sample poled in inert gas (Ar) was used for comparison. Systematic analyses of the chemical composition, thickness, silicate network, trapped molecular species and hydrous species in the sodium-depleted layers revealed correlations between subsurface structural changes and mechanical properties such as hardness, elastic modulus and fracture toughness. A silica-like structure was created in the inert gas environment through restructuring of Si-O-Si bonds at 200 oC in the Na-depleted zone; this occurred far below Tg. This silica-like surface also showed enhancement of hardness comparable to that of pure silica glass. The anodic thermal poling condition was found so reactive that O2 and N2 species can be incorporated into the glass, which also alters the glass structure and mechanical properties. In the case of the anodic surfaces prepared in a humid environment, the glass showed

92 an improved resistance against crack formation, which implies that abundant hydrous species incorporated during thermal poling could be beneficial to improve the toughness.

Introduction

The mechanical properties of soda lime silica (SLS) glass such as hardness, elastic modulus and fracture toughness are important parameters determining the durability in practical applications. To improve surface properties of commercial SLS glasses, the sodium ion concentration profile near the surface region is often modified through dealkalization or ion exchange processes.122,184–186 For example, upon depletion of the subsurface Na+ ions via exposure to SO2 or difluoroethane at high temperature (near glass transition temperature) or exposed to steam below 200 C, SLS glass becomes more resistant to crack formation.31,187,188 When subsurface Na+ ions are replaced with K+ ions by immersing the SLS glass in the molten potassium nitrate bath,

SLS glass becomes more resilient to mechanical damage.9,83 This paper explores the use of thermal poling as an alternative means for modification of surface mechanical properties at relatively low temperature (200 oC).

Thermal poling was originally studied as a process to induce non-linear optical properties and electro-optic effects in isotropic glass materials;189,190 it was recently reported that thermal poling can also be used as an efficient way to increase the indentation fracture toughness of SLS glass.67 Thermal poling is typically done by applying a high electric field (on the order of MV/m) across the glass at an elevated temperature, which facilitates diffusion of Na+ ions from the surface in contact with the anode to the surface in contact with the cathode, and then cooling the sample while holding the electric field to freeze any non-equilibrium distribution or polarization of ions in the glass matrix.67,95,191 In SLS glass, the Na+ ions are associated with non-bridging oxygen (NBO) atoms; upon their diffusion into the bulk, the residual negative charges of NBOs in the anode side

93 will be annihilated or compensated through electron ejection or various chemical reactions inside the glass..141,192,193 When thermal poling is performed with a non-blocking electrode, environmental gases can also be involved in such chemical reactions. Thus, the gas condition during thermal poling can have profound impacts on chemical structures and mechanical properties of the treated glass surface.

Generally, H2O and O2 are considered to be reactive molecules while N2 and noble gas molecules are inert in ambient conditions. Indeed, water can diffuse and react with the non-bridging oxygen (NBO) sites to form Si-OH in the sodium-depleted region.95,141,191 Such reactions can also be accompanied by or compete with alteration of the silicate network structure.95,145 When the amount of reactive species entering the glass is not sufficient enough to compensate or consume all

NBO sites left, emission of electrons and extinction of anions take place.193,194 One such process

141,195 could be the condensation of NBOs forming molecular O2 and a new bridging oxygen (BO).

These reactions could lead to restructuring of the silicate network even though thermal poling is carried out far below the glass transition temperature (Tg). It was also reported that NO2 and N2O4 species can be formed in a borosilicate glass during thermal poling in air.142 It is then critical to understand what chemical reactions occur between the glass and environmental gases during thermal poling and how the surface mechanical properties are altered after thermal poling in different gas environments.

In this paper, a comprehensive and systematic characterization with spectroscopic analyses and mechanical tests was carried out to understand structural changes in the SLS glass surface induced by thermal poling with the nonblocking electrode in inert and reactive gas environments and their effects on mechanical properties. The inert gas environment was modeled with Ar; reactive gas components tested were N2, O2, and H2O (consitituents of ambient air).

Note that even N2 (which was originally expected to be inert) is in the reactive gas category. The depletion layer thickness was measured with cross-sectional energy-dispersive x-ray (EDX) and

94 impedence spectroscopies. The silicate network structure was probed with specular-reflection infrared (SR-IR) spectroscopy. The molecular species formed and trapped in the sodium-depleted surface were identified with Raman, attenuated total internal reflection IR (ATR-IR), and vibrational sum frequency generation (SFG) spectroscopies. These spectroscopic analyses provided deeper insights into the chemical reactions and structural modifications taking place in

SLS glass upon depletion of sodium ions and reactions with surrounding gases (N2, O2, and H2O).

These chemical and structural alterations result in substantial changes in hardness, modulus, and indentation fracture toughness measured with nanoindentation and Vickers indentation.

Experimental methods

Commercial soda lime silica (SLS) float glass (Asahi Glass Co, Japan) with a thickness of

1mm was used in this study. The bulk composition, in weight percentage determined from x-ray fluorescence, is 72.3% SiO2, 13.3% Na2O, 7.7% CaO, 1.9% Al2O3, 4.4% MgO, 0.3% K2O, and

0.1% Fe2O3. All samples were exposed to SO2 during the float glass manufacturing process; otherwise, no chemical treatment or coating was applied after manufacturing. The samples were cut into roughly 3 cm × 3 cm size and cleaned by first rinsing with ethanol and DI water, and then by UV-ozone to remove organic contaminants on the glass surface before thermal poling. The poled samples were cleaned once again before surface characterization.53

A schematic diagram of the thermal poling setup with environmental control is shown in

Figure 5-1. Stainless steel plates were used for both anode and cathode. The electrodes were in physical contact with the glass surfaces up to their own weights. The poling area under the electrode was 1.6 cm2. The thermal poling temperature was held constant at 200C which is far below the strain point for this glass. The temperature of the sample was measured by a thermocouple that was in direct contact with the cathode. All samples were prepared by the same electrical bias and

95 thermal profile conditions except the gas atmosphere in the chamber. The gas environment during the thermal poling was controlled by a continuous flow of vapor with the flow rate of 5 L/min at atmospheric pressure. In this study, high purity Ar, dry N2, N2 + 20% O2 (dry air), N2 + 1.3% H2O

(low humidity) and N2 + 2.9% H2O (high humidity) were applied. The dew points of Ar, N2 and O2 gases used were below -50 C.

Figure 5-1. Schematics of thermal poling of SLS glass with non-blocking electrodes in controlled gas environment.

When the temperature and chemical environments were stable, +2kV of DC bias voltage was applied to the anode. The air-side of the SLS glass was in contact with the anode. After keeping the temperature at 200C for 20 minutes, the furnace was lowered to room temperature while the

DC bias was held constant. Around 50 oC, the cathode current decreased below the detection limit of the picoammeter. After thermal poling, a sodium-depleted surface on the air side of the SLS glass was created. The reference sample, which is labelled as “pristine” in this study, was outside

96 the poling area of the same glass. Thus, the pristine reference surface went through the same thermal history in the same gas environment as the treated surfaces except for the high DC electric field.

The energy-dispersive x-ray (EDX) analysis was carried out using an FEI Quanta 200 (FEI

Co., Hillsboro, OR) Environmental SEM system to measure the chemical composition changes in the sodium-depleted surface of the poled SLS glass. The samples were fractured through the poled area for cross-sectional analysis. An electron beam with a 20 keV energy was used to line scan over the cross-sectional surface. No conductive layer deposition was applied on the sample surface.

During the EDX analysis, a water vapor pressure was kept at 100 Pa to mitigate surface charging and minimize the migration of Na+.157,158

Impedance spectroscopy was used to analyze the ionic conductivity of mobile cations in the sodium-depleted layer. The analysis was performed with a commercial impedance analyzer

(Modulab, Solartron Analytical, Hampshire, UK). A platinum coating was deposited by sputtering at room temperature on both surfaces of the glass to serve as electrodes. The samples were placed in an oven with electrode attached and kept at 100C during the measurement. The temperature was monitored using a thermocouple close to the glass sample. The frequency was swept from 100 kHz to 0.01 Hz with the oscillation voltage of 0.5 V during impedance measurement.

Raman spectroscopy was performed with a Horiba LabRam system. A 458 nm laser equipped with a 50x objective lens was used in this study. The input laser power was 4.2 mW during the analysis. Surface topography of glass surface before and after thermal poling treatment was analyzed with an optical profilometer (Zygo NV7300, Middlefield, CT).

Specular reflectance IR (SR-IR) spectroscopy was carried out with a ThermoNicolet 670

FTIR spectrometer equipped with an MCT detector (Waltham, MA) at an incident angle of 45. A gold mirror was used as a reference background. The approximate information depth from SR-IR analysis is 0.52 m in the Si-OH and Si-O-Si asymmetric stretching vibration region (900-1100 cm-1).96 Attenuated total reflectance IR (ATR-IR) spectroscopy was performed with a Bruker

97 Hyperion 3000 Microscope (Bruker, Co.) equipped with a Ge ATR crystal with 60 incidence angle and a Bruker V70 (Bruker, Co) equipped with a diamond ATR crystal with 45 incidence angle.

Based on the refractive index of the pristine SLS glass, the information depth of ATR-IR in the spectral range between 4000 cm-1 and 1200 cm-1 is estimated to be 0.51.75 m when the Ge crystal is used and 1.52.5 m when the diamond crystal is used.52,96

Vibrational sum frequency generation (SFG) spectroscopy was employed to study the hydrogen bonding interactions of hydrous species (silanol group and molecular water) in the surface region of SLS glass. The detailed experimental configuration of SFG spectroscopy can be found elsewhere.72 Visible laser pulses (532nm) and tunable IR laser pulses (2.510 m) generated from an EKSPLA laser system (EKSPLA Co., Vilnius, Lithuania) were spatially and temporally overlapped at the glass surface. The incidence angle for IR and visible pulses were 56 and 60, respectively. The polarization combination in this study was s for SFG signal, s for visible laser pulses, and p for IR laser pulses (ssp). The energies of visible and IR pulses before reaching the glass surface were 174 J and 136220 J, respectively. The SFG signal intensity was normalized with the incident visible and IR laser intensities. The SFG spectra were plotted as a function of the wavenumber of the input IR beam.

Hardness and elastic modulus of the sodium depleted surface after thermal poling were measured with a nanoindenter (Hysitron TI 950, Minneapolis, MN) with a Berkovich tip. The tip area function was calibrated with a standard quartz single crystal. The indentation depth was controlled from 50 to 200 nm. The hardness and reduced modulus were calculated using the Oliver-

Pharr model.3 The average value and standard deviation of hardness and reduced modulus were obtained from more than 50 measurements for each sample.

Vickers indentation analysis was conducted to evaluate the fracture toughness of the sodium-depleted surface formed by thermal poling using a micro indenter (MHT Series 200; Leco

98 Corporation, St. Joseph, MI) equipped with a Vickers tip.58 During the indentation, a 300 gf normal load was applied to the Vickers tip in contact with the glass surface, held for 15 seconds and withdrawn from the glass. The optical images of the Vickers indent were obtained immediately after the indentation was completed. The relative humidity (RH) of the system was kept at 40% during the test.

Results and discussion

Chemical composition and thickness of the thermally-poled surface layers

The atomic concentration ratios of Na/Si, Ca/Si, Mg/Si and O/Si in the outermost 20 m of the sodium-depleted layer formed by thermal poling in different chemical environments are presented in Figures 5-2a to 5-2e. Overall, Ca/Si and Mg/Si do not change substantially while Na/Si varies significantly, which is due to a much greater mobility of Na+ over Ca2+ and Mg2+ in the SLS glass.196–199 The thickness of sodium-depleted layers (Figure 5-2f) is highly influenced by the gas environment during thermal poling. The sample poled in Ar shows a steep Na+ concentration gradient in the outermost 1.3 m and a gradual change over ~15 m until Na/Si reaches the bulk value. The sodium-depletion layer thickness is roughly ~2.6 m for poling in N2, ~3.3 m in N2 +

20% O2, ~3.6 m in N2 + 1.3% H2O, and ~4.1 m in N2 + 2.9% H2O.

Figure 5-2. EDX cross-section profiles, from the external surface, of the anode surfaces treated by thermal poling in (a) Ar, (b) N2, (c) N2 + 20% O2, (d) N2 + 1.3% H2O and (e) N2 + 2.9% H2O. The dash lines in the graphs represent the atomic ratios in the bulk of pristine glass. (f) Comparison of the thickness of sodium-depleted layer determined by EDX and impedance spectroscopy.

In Figures 5-2a – 5-2e, it should be noted that O/Si is also altered after thermal poling. For the sample poled in Ar (inert gas), the O/Si ratio is higher than the bulk value in the steep Na/Si gradient region (<1.3 m) and slightly lower in the gradual Na/Si gradient region (1.3 – 15 m). In contrast, the samples poled in N2 and N2 + H2O vapors (both 1.3% and 2.9 %) show a slight increase in the steep Na/Si gradient regions. The sample poled in the presence of O2 exhibits a slight excess in O/Si up to 10-12 m from the surface. These results indicate that the concentration and distribution of OH, H2O, O2, or other oxygen-containing species are also altered in the sodium- depleted region during or after thermal poling.195,200 More details will be discussed in Sections 3.3-

3.4 with Raman, IR, and SFG analyses.

Impedance spectroscopy was used as a complementary technique to verify the thickness of the sodium-depleted subsurface layer. In the Nyquist plot (Figure 5-3), the poled samples show two

100 semicircles while the pristine glass has only one semicircle. The change in the lower frequency semicircle is associated with the formation of a sodium-depleted layer with a resistivity higher than the bulk.134,201 The samples poled in different gas environments show quite different radii for the second semicircle in the high frequency region. The high frequency region is governed by the ion transport in the cathode side and the bulk of the glass, which is beyond the scope of this paper.

Fitting the semicircles in the Nyquist plot with an analytical model provided an estimation of the thickness and resistivity of the sodium-depleted layer (see the Supporting Information). In general, sodium-depleted layers formed in humid environments gave lower resistivity than those formed in dry conditions. The depletion layer thicknesses calculated from the impedance data fitting are plotted in Figure 5-2f. The thicknesses calculated from the impedance data cannot be

101 accurate since the exact value of relative permittivity (r) of the sodium-depleted layer is not known and that there is a concentration gradient of sodium in the depletion region. Although the thickness estimated using the r value of the bulk glass show less variance than the one determined from the

EDX analysis, their magnitudes are in reasonable agreement to each other. The inset in Figure 5-

3 shows that the Ar-poled sample has an equivalent bulk impedance value to the pristine sample, which is attributed to a very thin depletion layer (see Supporting Information Table S1). All other poling environments generate a thick depletion region that penetrates the bulk layer and increases the apparent bulk resistance.

Changes in silicate network in the sodium-depleted surface after thermal poling

The silicate network structure of the pristine and sodium-depleted surfaces of SLS glass was probed from the peak position of the Si-O-Si asymmetric stretching (Si-O-Si) vibration mode

-1 using SR-IR (Figure 5-4). The Si-O-Si band of the pristine surface is peaked at ~1070 cm ; compared to this, all sodium-depleted surfaces from thermal poling show a blue shift. Since the pure silica

-1 network has the Si-O-Si band centered at ~1100 cm , the blue shift of this band could mean that the silicate network in the poled surface is restructured upon sodium depletion and becomes more

“silica-like”.31,42,95 Based on a recent simulation study, this blue shift could also indicate a decreased weighted mean of the Si-O bond length,202 which suggests the formation of a strained silicate network compared to pristine SLS glass.

Changes in the intensity of the Si-O- (NBO) and Si-OH stretch modes at ~940 cm-1 are also evident in Figure 5-4. The intensity of this 940 cm-1 band decreases for the sodium-depleted layers formed by poling in Ar, N2, and N2 + 20% O2. These samples show a larger blue shift in the Si-O-Si band than the ones poled in the presence of H2O in the gas phase but with no significant increase

102 in the OH content during the poling (see Section 3.3). These results imply that a large fraction of

NBOs are converted to BOs upon sodium depletion by thermal poling in these gas environments.

In contrast, the intensity of the 940 cm-1 band is slightly increased upon thermal poling in humid environments. This must be due to the conversion of NBOs to Si-OH groups, which is consistent with the increase in the hydrous species peak in ATR-IR (see Section 3.3).

Figure 5-4. SR-IR spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O.

The surface topography of the sodium-depleted surfaces were analyzed with optical profilometry (Figure 5-5). In the area poled in Ar and N2, many craters with depths ranging from

~20 nm to ~80 nm can be seen frequently. The sodium depletion which accompanied the conversion of most NBOs to BOs must cause so much local deformation or aggregation of void space in the silicate network,195 that some of them are observed as craters in high-resolution optical

103 profilometry. Similar craters were also observed on the surface poled in N2 + 20% O2, although their abundance and depth were less than those found on the surface poled in Ar and N2. It is interesting to note that when water is present in the environment gas during thermal poling (N2 +

1.3% H2O and N2 + 2.9% H2O), no craters or collapse regions were observed. Along with the

-1 increase in the 940 cm band, these results imply that H2O molecules readily ingress into the glass and react with NBO sites left behind during the sodium depletion via thermal poling in the gas containing H2O vapor.

Figure 5-5. (a) Three-dimensional representation of the SLS glass surfaces thermally poled in Ar and N2. (b) Statistics on the number of craters and depth of those craters.

Molecular species trapped in the sodium-depleted surface

Detection of molecular oxygen trapped in subsurface with Raman

One of the charge compensation reactions for NBOs upon sodium depletion by thermal poling is the condensation of NBOs into molecular O2 and ejection of electrons to the anode.

Depending on the transport property (which would be dependent of the resulting network structure),

104

141,203,204 some O2(g) molecules produced can be trapped inside the subsurface region. Then, they can be detected with Raman spectroscopy. The characteristic peak of molecular O2 appears at

~1550 cm-1.

In Figure 5-6, the pristine surface shows no peak at ~1550 cm-1, as expected. The samples

-1 poled in Ar, N2, and N2 + 20% O2 show a clearly discernable O2 vibration peak at ~1550 cm . Note

-1 that these are the samples exhibited a larger blue shift in the Si-O-Si band (from 1070 cm to1094 cm-1) and a larger decrease in the NBO band at ~940 cm-1 in SR-IR (Figure 5-4) and more collapsed pits in the surface topographay (Figure 5-5). The trapped molecular O2 has also been observed in the cases where thermal poling was conducted with a blocking electrode configuration.141,195 When the gas environment during thermal poling contains 1.3% H2O, then the molecular O2 peak decreases substantially; when the H2O concentration is increased to 2.9%, the molecular O2 peak is completely suppressed.

Figure 5-6. Raman spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O.

105

The results shown in Figure 5-6, along with the SR-IR data in Figure 5-4 and the topography change shown in Figure 5-5, imply that the production of molecular O2 from NBOs upon depletion of Na+ ions is coupled with the formation of more BOs from NBOs (leading to restructuring of the silicate network to a “silica-like” structure). The amorphous silica has a density lower than the soda lime silica glass; the density difference between the surface layer and the bulk and the loss of gaseous O2 products (which are not detected in post-treatment Raman analysis) might be the causes for local collapses in surface topography. These repolymerization reactions between NBOs appear to be greatly suppressed when H2O is present during thermal poling. Stoichiometric relationship between these reactions could not be obtained in this study since the fraction of trapped O2 among the total produced amount is unknown.

Identification of subsurface chemical species using ATR-IR

ATR-IR was employed to detect other chemical species newly formed or located in the sodium-depleted subsurface region of the SLS glass during or after thermal poling treatments.

The probe depth of ATR-IR is comparable with the thickness of sodium-depleted layers determined from EDX and impedance spectroscopy analyses. The ATR-IR spectra of the sodium- depleted surfaces formed by thermal poling in different gas environments are summarized in

Figure 5-7.

When the SLS glass surface is thermally poled in Ar (inert gas), the ATR-IR analysis

-1 shows an increase in the OH stretch region (OH; 2700 – 3700 cm ) region and a small increase in

-1 the H2O bending region (HOH; 1630~1640 cm ). Initially, it was a puzzling observation since there was no source for H2O in the gas phase during the thermal poling process. It was hypothesized that the surface poled in Ar may be porous, allowing the ingress of water molecules

106 from ambient air after the poling. To test this hypothesis, the surface poled in Ar was exposed to

D2O vapor at room temperature for 30 min and then retrieved to ambient air. In Figure 5-7b, the

-1 sample poled in Ar and then exposed to D2O clearly shows the OD stretch (OD; 2200-2750 cm )

-1 205,206 and DOH bending (DOH; ~1440 cm ) modes, verifying the hypothesis about the post-poling ingress of water from the ambient air into the surface poled in Ar. Recall that the SLS glass surface poled in Ar shows numerous craters (Figure 5-5) and a significant restructuring in BO

(Figure 5-4); it is possible that the regions that are not collapsed might be porous enough to allow diffusion of water molecules from the ambient air. This might also allow further reactions with residual NBO sites that are not fully compensated after sodium depletion. This might explain the small increase of the O/Si ratio above the bulk value near the surface (<2 m from the surface) in the EDX cross-section profile of the surface poled in Ar (see Figure 5-2a).

Figure 5-7. (a) ATR-IR spectra (measured with Ge crystal) spectra of the pristine SLS glass surface and the sodium-depleted surfaces produced by thermal poling in Ar, N2, N2 + 20% O2, N2 + 1.3% H2O, and N2 + 2.9% H2O. (b) ATR-IR spectra (measured with diamond crystal) of the sodium- depleted surfaces formed by poling in N2 and Ar. In each gas condition, one sample was retrieved o directly to ambient air after thermal poling and the other was exposed to D2O vapor at 25 C after the thermal poling treatment and then retrieved to ambient air. In (b), the diamond ATR was used since it makes easier to identify the HOH band due to a flatter baseline than the Ge ATR.

107

-1 The SLS surface poled in dry N2 shows a sharp peak at ~1650 cm and a weak peak at

-1 -1 ~1300 cm (Figure 5-7a). The intense peak at ~1650 cm cannot be the HOH band of water since

-1 there is no peak in the OH region (2800 – 3700 cm ). Unlike the surface poled in Ar, the surface poled in N2 does not show discernable OD peak upon post-poling exposure to D2O (Figure 5-7b), indicating that it does not react with or uptake water from the ambient air. Surprisingly, the

- 207 positions of these two new peaks fall in the typical range of bidentate nitrate (NO3 ) species.

Such anions can contribute to the negative electric field trapped in the anode-poled surface.12 Or, the 1650 cm-1 peak could be attributed to –ONO species.208–210 Cremoux et al. previously reported

+ the detection of various nitrogen oxide species (including NO , NO2, N2O4, N2O3, etc) trapped in the subsurface of borosilicate glass after thermal poling in air.142

In any case, these results indicate that N2 is involved in compensation or annihilation reactions of NBOs during the sodium depletion by thermal poling and that reaction products can

211 be trapped in the sodium-depleted layer. In other words, N2 is not inert in the thermal poling condition. Again, stoichiometry of this reaction cannot be checked since there is no way to determine what fraction of the produced nitrogen oxide species is trapped in the poled surface.

Also, it is difficult to determine the exact speciation of reaction products because the peak position and shape of nitrogen oxide species vary with the electronic structure and binding geometry of the adsorption sties.211–216 Due to the nonequilibrium and amorphous nature of the glass, the structure of reactive sites could have a broad distribution.

The ATR-IR spectrum of the surface poled in N2 + 20% O2 (dry air) shows the enhancement of the ~1650 cm-1 and ~1300 cm-1 as well as a sharp peak at ~2225 cm-1 and a doublet peak at 1480 – 1550 cm-1 (Figure 5-7a). The ~2225 cm-1 peak can be assigned to nitrous

+ 217 -1 oxide (N2O) or nitrosonium ion (NO ). The peaks at 1480 – 1550 cm could be attributed to

207 the unidentate nitrate species. The supply of O2 from the gas phase appears to enhance or facilitate these additional reactions.

108

In contrast, the presence of H2O in the gas during the thermal poling appears to suppress the chemical reactions involving nitrogen species in the sodium depletion layer. The presence of

-1 1.3% H2O (low humidity) in the gas phase completely quenches the 2225 cm species and greatly suppressed the 1550 – 1480 cm-1 species. Instead, a small growth of the 1630 cm-1 shoulder

-1 (HOH) and a noticeable increase in the 2800 – 3700 cm (OH) region. This implies the formation of Si-OH species from NBOs as well as ingress of molecular water. When the gas-phase H2O concentration during the thermal poling is increased to 2.9%, then almost all nitrogen oxide species are suppressed in ATR-IR and the OH band of hydrous species (OH and H2O) and the

HOH peak of H2O are dominant (Figure 5-7a).

Probing hydrogen bonding interactions among hydrous species with SFG

The apparent simplicity of the ATR-IR band in the OH stretch region masks the underlying complexity associated with the hydrogen-bond interactions of hydrous species with the silicate network. Thus, vibrational SFG spectroscopy has been utilized to probe the chemical environment of hydrous species in the sodium-depleted layer formed by thermal poling in

33,95,110 different gas environments. For the poled glass surface, the SFG signal (퐼푆퐹퐺) can originate from two possible sources – the noncentrosymmetrically-arranged functional groups ((2) effect)

(3) 110,164,166 and the electric field-induced contributions ( 퐸퐷퐶 effect). The latter term is possible due to the internal electric field (퐸퐷퐶) accumulated in the sodium-depleted layer where excess electrons or ionic species are trapped in the amorphous glass network.95,143 The surface poled with the non-blocking electrode is expected to have much lower 퐸퐷퐶 than the one poled with the blocking electrode since a number of charge compensating reactions can occur during or after thermal poling. Even though the magnitude of 퐸퐷퐶 in the poled surface cannot be determined and

109

(2) (3) it is not possible to distinguish the relative contributions from the  and  퐸퐷퐶 terms, drastic differences in the SFG spectra of the sodium-depleted layers produced by poling in different gas environments (Figure 5-8a) provide useful information about the distribution of hydrous species and their hydrogen bonding interactions.110

First, it should be noted that the multiple OH peaks observed for the pristine SLS glass

(black line in Figure 5-8a; ~3160, ~3390, ~3570, ~3740, ~3900 cm-1) are not due to the physisorbed water on top of the glass surface. The SFG features of the physisorbed water layer on a melt-derived SLS glass is much broader and weaker than the spectral features (black color) shown in Figure 5-8a.110 These multiple peaks originate from the hydrous species trapped inside

110 the SO2-treated float glass and the separation of peaks is due to differences in the O-HO hydrogen-bond interaction.110 These subsurface hydrous species are produced by the high- temperature SO2 dealkalization process of the float glass manufacturing. This process is known to deplete sodium ions in the <100 nm region from the surface via chemical reactions.97

The surface poled in inert Ar shows about a 20 times increase in the SFG intensities, which is believed to originate from the field-induced enhancement of SFG signals. The enhancement might be in part due to the uptake of water species from the ambient air after thermal poling (as evidenced by the OD peaks for the post-poling D2O exposure sample in Figure

5-8b). It is noticed that the hydrogen-bonded OH bands (3100  3700 cm-1) are red-shifted, which indicates stronger hydrogen bonding interactions or shorter O-HO distances compared to the pristine surface.110 This might be the consequence of the large restructuring of the silicate network (as evidenced by the large blue shift in the Si-O-Si stretch peak) and the collapse of surface topography (Figure 5-5). In the previous study, the ~3900 cm-1 OH peak is speculated to originate from a strained silicate network site far from the equilibrium structure.110 This peak intensity is also enhanced upon poling in Ar.

For the samples poled in dry N2 and N2 + 20% O2 (dry air), the multiple SFG peak features are significantly enhanced (red and cyan colors in Figure 5-8a). Note that ATR-IR analysis shows no significant uptake of hydrous species during or after thermal poling in these conditions (Figure 5-7). So, the enhancement of the SFG signals must be due to the field-induced ordering of hydrous species in the sodium-depleted layer, not due to the increase in the subsurface water content. Similar to the case of poling in Ar, the SFG peak at ~3900cm-1 is significantly enhanced and the peaks of hydrogen-bonded OH groups (<3700 cm-1) are substantially red- shfted. These changes are additional evidence for the restructuring of the silicate network through

111 condentation reactions of NBOs resulting in trapped O2 (Figure 5-6), the increase in BO (Figure

5-4), and local collapse of surface topography (Figure 5-5).

The SFG spectra of the samples poled in the presence of H2O in the gas phase (1.3% H2O and 2.9% H2O in N2) are drastically different from those of the samples poled in dry conditions

- (Ar, N2, and N2 + 20% O2). The SFG intensities of the hydrogen-bonded OH species at <3700 cm

1 are substantially enhanced; the largest increase is in the weakly hydrogen bonded region (3300 –

3700 cm-1). These increases must be associated with the large uptake or formation of hydrous species as evidenced from ATR-IR analysis (Figure 5-7a), in addition to the field-induced ordering of hydrous species, inside the sodium-depleted layer.

Surface mechanical properties after thermal poling reactions in different environments

Indentation hardness (H) and modulus (E’) of sodium-depleted surfaces formed by thermal poling in different gas environments were analyzed with nanoindentation. Since the

Poisson’s ratio of the sodium depleted layer is hard to determine, reduced elastic modulus (E’) is reported rather than elastic modulus (E).3 The indentation depth was varied within 10% of the thickness of the sodium-depleted layer determined from EDX and impedance analyses (Figure 5-

2f);3,81,219 thus, the mechanical properties probed with nanoindentation pertain to the sodium- depleted layer with little contribution from the bulk (except the 200 nm data points of the Ar and

N2 cases). Figure 5-9 summarizes the H and E’ values measured at 4 different indentation depths.

The pristine glass has very little indentation depth dependence; the H value from nanoindentation of the pristine glass is slightly higher than the bulk value (~6.5 GPa), which must be due to the

SO2 dealkalization treatment of the float glass manufacturing process. In contrast to the pristine glass, the poled glass surfaces show indentation depth dependences. This is because the chemical composition and structure of the sodium-depleted layer change gradually along the depth from

112 the surface (Figure 5-2). Note that the chemical composition dependences of hardness and modulus are not a simple linear function.220,221

Figure 5-9. Hardness (a) and reduced elastic modulus (b) of sodium-depleted surfaces compared with pristine SLS glass with nanoindentation analysis.

It is well known that amorphous silica (a-silica) has a hardness larger than SLS glass; thus, the E/H ratio of a-silica is only ~8, while that of SLS is ~11.5.222 In Figure 5-9, it is noted that the E’/H ratio at a indentation depth of 150 nm for the sodium-depleted surface formed by poling in Ar and N2+20% O2 is 8.4-8.6, which is remarkably similar to the E/H ratio of a-silica.

The E’/H ratio of the surface poled in N2 is slightly higher (9.5), but still significantly lower than that of the pristine surface (10.4). The “silica-like” mechanical behavior of the poled surfaces in dry conditions are consistent with the structural rearrangement revealed from spectroscopic analyses Upon the depletion of Na+ ions, NBOs undergo condensation reactions forming more

BOs (Figure 5-4) accompanied by collapse of surface topography (Figure 5-5), some trapped O2

(Figure 5-6), and shorter O-HO distances (Figure 5-8).

If the increase of the Si-O-Si network connectivity is perturbed by formation of nitrogen oxide (NxOy) species during thermal poling or further reactions in ambient air after thermal poling, it could alter H and E’ (as can be seen in comparison of Ar vs. N2). Some differences

113 between N2 and N2 + 20% O2 (dry air) are noticed; however, detailed explanations for such differences cannot be discussed in this paper due to the lack of reaction stoichiometry information. The lower values of H and E’ at 50 nm indentation for the Ar data set, compared to those obtained at larger indentation depths, can be explained with the fact that the surface layer is porous (as evidenced from the uptake of water from ambient air, Figures 5-7b and 5-8b).

It appears that the presence of H2O in the gas phase during thermal poling makes a drastic

+ difference in the mechanical response of the Na -depleted layer. The surfaces poled in N2 + 1.3%

H2O and N2 + 2.9% H2O (humid nitrogen) show the lowest hardness among the ones tested.

These surfaces uptake a significant amount of hydrous species (Si-OH and H2O) during the poling (Figures 5-7 and 5-8). The smaller blue shift in the Si-O-Si band position (Figure 5-4) and the lesser amount of trapped O2 (Figure 5-6) suggest that the NBO condensation reaction is suppressed. Instead, most NBOs (Si-O-) seem converted to silanol groups (Si-OH; as evidenced by the 940 cm-1 peak in Figure 5-4). The large amount of hydrous species in the glass network can lower the hardness.31,67,223 It is also important to note that unlike the surfaces poled in the dry condition, the E’/H ratio (10.5-10.6) of the surface poled in the humid condition is still similar to the prisitine surface (10.4). This imples that the silicate network structure of the surface poled in humid conditions is far from the “silica-like” structure; in other words, the Si-O-Si connectivity is much lower than the surfaces poled in dry conditions.

114

Figure 5-10. Optical images of Vickers indent of (a) pristine SLS glass surface; sodium-depleted surfaces prepared in (b) Ar, (c) N2, (d) N2 + 20% O2, (e) N2 + 1.3% H2O, (f) N2 + 2.9% H2O. The applied normal load is 300gf. The relative humidity is about 40% during the indentation analysis.

Fracture toughness, probing the resistance to crack formation and propagation, is another mechanical property that is sensitive to differences in the silicate network composition and structure.224,225 The Vickers indentation test results in Figure 5-10 show a remarkable difference between the surfaces poled in dry (Ar, N2, and N2 + 20% O2) versus humid (N2 + 1.3% H2O and

N2 + 2.9% H2O) conditions. The apparent fracture toughness of the sodium-depleted surfaces formed by thermal poling in humid conditions is much higher than the rest of the samples tested.

The formation of cracks during the indentation can be viewed as a way to release the strain energy applied to the glass network by breaking the chemical bonds; however, some degree of plasticity can also play a role.225 It is speculated that polymerization between abundant hydroxyl groups under compressive stress might be responsible for the increased resistance to crack formation.

115 Conclusion

In this study, surface mechanical properties of SLS glass are altered by chemical reactions in inert and reactive gas environments during thermal poling and reactions with amibient air after thermal poling. Comprehensive analysis on the composition, thickness, silicate network changes, chemical species, chemical environment of hydrous species and topography of sodium-depleted surfaces are carried out and the results are summarized in Table 5-1. It is found that the sodium-depleted surface prepared in an inert gas environment has a “silica-like” network strucutre, while sodium-depleted surfaces prepared in reactive gas environments have substantially-modified networks with trapped molecular species. The enhancement of hardness is observed for the sodium-depleted surface prepared by poling in inert Ar gas and dry air. The sodium-depleted surfaces prepared by poling in humid environment become more resistant to crack formation.

116 Table 5-1 Summary of chemical, structural, and mechanical changes of the sodium-depleted surface formed by thermal poling (200 oC, 2 kV, 20 minutes) of SLS float glass (thickness = 1 mm) in various gas environments.

* (g) = molecular species trapped in the subsurface

-1 # SiOSi = 1070 cm outside the thermal poling area, directly read from the reflectance spectra

117

Chapter 6

Effects of Na+/K+-ion exchange on mechanical and mechanochemical properties of soda lime silica glass

Submitted to Journal of NonCrytalline Solids

Overview

The effect of exchanging Na+ in soda lime silica (SLS) glass with K+ on mechanical and mechanochemical properties were investigated. It is known that replacing smaller modifier ions with bigger ions in the silicate glass network at temperatures below glass transition (Tg) produces a compressive stress in the subsurface region, which enhances resistance to mechanical damages.

This study found that when Na+ ions in SLS are exchanged with K+ ions at 400 oC, the hardness, indentation fracture toughness, and crack initiation load of the surface are increased, which is consistent with the chemical strengthening effect. However, the resistance to mechanochemical wear in a near-saturation humidity condition (relative humidity RH = 90%) is deteriorated. When

K+ ions are exchanged back with Na+ ions at 350oC, the wear resistance in high humidity conditions is recovered. By analyzing the surface chemical composition, silicate network and hydrogen-bonding interactions of hydrous species in the subsurface region, it is suggested that the leachable Na+ associated and subsurface hydrous species in the silicate network play more important roles in the mechanochemical wear of SLS at high RH and the improvement of mechanical properties under indentation normal to the surface is irrelevant with the resistance to mechanochemical wear under tangential shear at the surface.

118 Introduction

Surface damages on flat glass panels produced by contact with foreign objects could have detrimental impacts on the practical strength of glass materials.32,77,226,227 Physical contacts causing surface damages can be divided into two processes – indentation normal to the surface and shear along the surface. There are numerous studies on damage mechanisms under normal indentation conditions.62,222,228–231 In the tangential shear process, mechanical scratch at high loads with sharp indenter tips are relatively well documented in the literature;232–234 but, the surface wear at light loads with blunt objects are not well studied. Recently, we have studied the effect of relative humidity (RH) on wear of bare glass surfaces. At the low load condition where mechanical damage could be prevented if the surface is lubricated a monolayer of alcohol molecules, surface wear occurs via mechanochemical processes.35,39 In a series of studies under the mechanochemical wear conditions, it was found that sodium calcium silicate (called more frequently as soda lime silica, SLS) exhibits a peculiar RH dependence of mechanochemical wear.33 While pure silica (fused quartz), alkali borosilicate (Schott Borofloat®33), barium boroaluminosilicate (Schott AF45), and sodium aluminosilicate (base glass of Corning Gorilla® 2

), chemically-strengthened aluminosilicate (Corning Gorilla® Glass 2) shows an increase in wear volume as RH increases to >80%, SLS shows a substantial reduction of wear volume at RH

>80% compared to the wear in lower RHs.33,34,36,37,39,58,146 Understanding of the origin of good wear resistance of SLS at RH>80% may provide critical insights needed to design and produce

SLS glass panels with a superior practical strength.

Based on the network former and modifier compositions of the glasses tested so far,

33,34,36,37,39,58,146 it is noted that only SLS has leachable Na+ ions associated with the non-bridging oxygen (Si-O-; NBO) atoms in the glass network. Then, it could be questioned whether its presence is critical for enhanced wear resistance at high RH conditions (which would produce

119 thick layers of adsorbed water on the glass surface). This hypothesis has been corroborated in several control experiments. When the SLS glass surface is thermally poled, the Na+-depleted surface (anode-side) loses the wear resistance at 90% RH, while the Na+-accumulated surface

(cathode-side) exhibits an enhanced resistance.146 When the subsurface Na+ ions are depleted via reaction with steam at 150-200oC, the SLS glass loses the wear resistance at 90% RH.31 The

AF45 glass contains a trace amount of Na+ ions in the bulk; those Na+ ions can be pushed to the surface via thermal poling. The accumulation of subsurface Na+ ions would be accompanied by production of NBO sites for charge compensations. The Na+-enriched AF45 glass surface also exhibits a good wear resistance at 90% RH.37

In this paper, the effect of exchanging Na+ ions in SLS with K+ ion was studied. It is well known that the replacement of smaller modifier ions in the subsurface region with larger ions creates a compressive stress in the surface region and increases the mechanical strength of glass.83,235,236 This is the chemical strengthening effect. In the case of SLS, the exchanged K+

(larger) ions will be associated with the NBO sites to which Na+ (smaller) ions are originally bound, creating a compressive stress in the surface.237–239 This system provides a unique testbed case to study (i) if the compressive stress in SLS can enhance the resistance to mechanochemical wear at RH > 80% and (ii) if the K+/NBO pair can provide the same wear resistance effect as the

Na+/NBO pair in SLS.

Experimental details

SLS float glass panels (thickness = 1 mm) from Asahi Co. were used in this study. The glass samples were cut into a dimension of ~2.5 cm  ~4 cm. A home-built stainless (SS) bath was used to hold the molten salt and SLS glass during the ion-exchange process. 20 g of KNO3 was pre-melted at 400 oC in the SS bath and kept covered. Prior to ion-exchange, the SLS

120 samples were cleaned by rinsing with DI water and pure ethanol, then UV-ozone exposure for 20 minutes.53 The glass samples were preheated to 400 oC in the same oven and then immersed into the molten KNO3 bath. The ion exchange time in the molten KNO3 bath was 48 hours. Then, the

K+-exchanged sample was taken out of the oven and cooled in ambient air. Hereafter, this K+- exchanged sample is referred to as “Na+K+” in the figures. Some of the Na+K+ samples had gone through a reverse-exchange process where the exchanged K+ is replaced back with Na+ ions.

+ The reverse-exchange was done by immersing the K -exchanged SLS glass in a molten NaNO3 bath at 350 oC for 8 hours. When the reverse-exchange time was less than 5 hours, there was no significant change between the Na+K+ and Na+K+Na+ samples in wear tests. The reverse- exchanged sample is marked as “Na+K+Na+” in the figures. Although ion exchange occurs in both air- and tin-sides of SLS, only the air-side was used in this study.

Surface mechanical properties of ion-exchanged SLS glasses were studied with a nanoindenter (Hysitron TI 950, Minneapolis, MN) with a Berkovich tip. Results from more than

60 indents were obtained and averaged for each indentation depth. Vickers indentation was performed with a microindenter (MHT Series 200; Leco Corporation, St. Joseph, MI) equipped with a Vickers tip to investigate the changes of fracture toughness. A home-built humidity control chamber around the indenter was utilized to carry out humidity dependent fracture toughness tests. A 1 mm WC ball (TRD Specialties, Inc., Pine Meadow, CT) was used as the tip of the indenter which is made of stainless steel. The load was controlled by a step-motor (Klinger

Scientific, Garden City, NY) and monitored with a load cell. An acoustic sensor (PICO sensor,

Mistras, Princeton, NJ) was physically attached to the indenter to detect the sound waves during the initiation of cracks. The acoustic sensor and Hertzian indenter were integrated and controlled by a home-written Labview program. At the time of crack initiation, the indenter was immediately withdrawn from the sample based on the acoustic detection, which allowed the crack initiation load to be recorded. In this study, a linear loading rate of 1N/s was used. The humidity

121 was controlled by purging the test chamber with N2 (g) and H2O (g) mixture. Crack initiation loads for glasses before and after ion-exchange at 0%, 40% and 90% were recorded. At least 30 data points was recorded at each conditions. The crack initiation stress was then calculated based on the Hertzian contact mechanics. Weibull statistics was then generated based on the distribution of crack initiation stress. The probability in Weibull plot was calculated based on the Blom methodology.

Mechanochemical wear tests were performed with a custom-built ball-on-flat tribometer with an environmental control capability. Details of this tribometer can be found elsewhere.240

Borosilicate (Pyrex) balls with a 2.38 mm diameter were used as a counter surface. All wear tests were done with 400 reciprocating cycles with an applied load of 0.2 N, which corresponds to a

Hertzian contact pressure of ~280 MPa (on the flat surface before wear occurs). The wear tracks were analyzed with an optical profilometer (Zygo, NV7300, Middlefield, CT).

Energy dispersive x-ray (EDX) analysis was performed with a FEI Qunata 200

Environmental SEM system (FEI Co., Hillsboro, OR). The samples were prepared by scoring the tin side of ion-exchanged SLS glass with a diamond cuter and then fracturing with a force. The samples were analyzed under a water vapor (pressure = 100 Pa) condition to mitigate surface charging. No conductive coating was deposited on the sample. X-ray photoelectron spectroscopy

(XPS) was utilized to probe the surface composition of SLS glasses before and after ion-exchange treatments. XPS analysis was done with a PHI VersaProbe system. Survey scans over narrow binding energy (BE) ranges of O 1s, Na KLL, Ca 2p, Mg KLL, K 2p, C 1s, Si 2p and Al 2p as well as high resolution scans of C 1s and O 1s were collected. The quantification was done using the method described earlier with home-calibrated relative sensitivity factors (RSFs).241

Specular reflectance infrared (SR-IR) spectra of SLS glass before and after ion-exchange treatment were obtained with a Bruker FTIR microscope spectrometer (Waltham, MA). All SR-

IR spectra were collected at a 20o incidence angle. A gold mirror was used as a reference

122 background. The spectrum was averaged over 100 scans at each spot. More than 5 spots were randomly selected on each sample and averaged. Very little variance was observed from spots to spots.

Vibrational sum frequency generation (SFG) spectroscopy was employed to probe hydrogen bonding interactions of hydrous species in the glass surface. The detailed system information can be found elsewhere.72 In brief, 532 nm pulses from a 27 ps Nd:YAG laser and tunable mid-IR pulses generated from an optical parametric generator/amplifier were spatially and temporally overlapped on the glass surface. The incidence angle for visible and IR laser pulses was 56o and 60o, respectively, and the SFG signal detection was made at the phase matching angle in the reflection direction. The polarization combination of SFG in this study is s for SFG signal, s for visible pulses, and p for IR pulses (ssp). The SFG signal was collected at an

8 cm-1 interval and averaged over 100 laser pulses at each data point.

Results

Concentration profile of the ion-exchanged SLS glass surface

Figures 6-1a and 6-1b shows the modifier ion depth profiles from cross-sectional EDX analysis of the Na+/K+-exchanged and reverse-exchanged SLS surfaces. The K/Si and Na/Si ratio profiles evolve oppositely from the surface, mirroring each other. The concentration profiles of divalent alkanine-earth ions (Mg/Si and Ca/Si) does not vary with the depth from the surface.

This data confirms that the ion-exchange occurs only between monovalent alkali ions (Na+ and

K+).242 Because the treatment time for reverse-exchange (K+Na) was shorter than that of initial exchange (Na+K+), the K+ ions in the shallower region are exchanged with Na+ and the K+ ions in the deeper remain in the sample.

123

Figure 6-1. EDX cross section profile of Na/Si, K/Si, Mg/Si and Ca/Si ratios of (a) Na+/K+- exchanged and (b) reverse-exchanged SLS surface; (c) Na/Si, K/Si, and Ca/Si ratios determined from XPS analysis of the pristine, Na+/K+-exchanged and reverse-exchanged SLS surfaces.

The sodium, potassium, and calcium concentrations detected with XPS are shown in

Figure 6-1c. The XPS probe depth is only about 10 nm from the surface. The Na+ concentration of the pristine sample is a bit lower than the bulk value due to the SO2 dealkalization process during the float glass manufacturing.7 After the Na+/K+-exchange at 400 oC for 48 hours, the Na+ concentration in the top 10 nm region is negligible; the K+ concentration increases to ~11% of the silicon network former concentration. Upon reverse-exchange of K+ with Na+ at 350 oC for 8 hours, the surface concentration of K+ decreases by a half and almost equal amount of Na+ is added to the surface.

Effects of ion exchange on mechanical responses under normal indentation

The chemical strengthening effect of Na+/K+-exchange for SLS was confirmed with nanoindentation measurements. Figure 6-2 compared the surface hardness and reduced modulus of the ion-exchanged SLS glasses. Since the Poisson’s ratio of the ion-exchanged region is not known, the exact elastic modulus of the surface could not be determined.243 For that reason, the

124 reduced modulus of the contact is plotted in Figure 6-2. The surface hardness and modulus are both significantly enhanced upon the Na+/K+ exchange. This significant increase confirms the formation of compressive stress after the Na+/K+-exchange process.57,60,244 Because the ion exchanged samples has the Na+ and K+ concentration gradients along the depth direction (so does the compressive stress), they exhibit more dependence on the indentation depth than the pristine glass. After K+ ions are exchanged again with Na+ ions, both hardness and modulus decrease towards the values of the pristine sample. During the reverse-exchange of K+ with Na+ at 350 oC, not only the alkali modifier composition is changed, but the surface compressive stress is also relaxed (as evidenced by the decrease in hardness decreases toward the value of the pristine surface).

Figure 6-2. (a) Hardness and (b) reduced modulus, measured at indentation depths of 100 nm, 150 nm and 200 nm, of the pristine, Na+/K+-exchanged and reverse-exchanged SLS surfaces.

125

Figure 6-3. Optical images of Vickers indentation imprint at a 500 gf normal load at 0%, 40%, and 90% RH conditions on (a) pristine, (b) Na+/K+-exchanged, and (c) reverse-exchanged SLS surfaces. The dwell time at the maximum load was 15 seconds.

Indentation fracture toughness is a good indicator of measuring the resistance to crack formation and propagation. Figure 6-3 compares the Vickers indentation images of the pristine,

Na+/K+-exchanged, and reverse-exchanged SLS glasses at 0%, 40%, and 90% RH conditions. The crack length from the corner of Vickers imprint is affected by the humidity in the gas phase of the test environment. In 0% RH, there is no visible radial crack after indentation with a 500 gf for 15 seconds for all pristine and ion-exchanged SLS surfaces. In 40% RH, the Na+/K+-exchanged SLS surface show no radial crack, while the pristine surface has a large crack (~86 m), confirming the chemical strengthening effect. The Na+/K+-exchanged SLS surface did not show any crack even at the highest load that we could apply with our system (1 kgf). Upon reverse-exchange of

126 K+ with Na+, the surface compressive stress decreases and thus the radial crack is formed, but its length (56 m) is shorter than the pristine surface case. In 90% RH, the trends are the same, except that the crack lengths are longer compared to the 40% RH data. It is understandable that the crack propagation will be affected by the stress corrosion in the presence water molecules at the surface. Therefore, the fracture toughness evaluation based on the Vickers indentation method remain to be revised.245,246

Since the crack initiates from the sharp corner in the Vickers indentation test, it may not fully reflect the distribution of surface detects. That problem can be circumvented by using a blunt probe such as a sphere. Figure 6-4 shows the Weibull plot of the crack initiation load distribution measured with the Hertzian indentation using a 1 mm diameter tungsten carbide ball.

In 0% RH (Figure 6-4a), there is no significant difference in critical load (around 2060 MPa) at

50% failure probability and the Weibull modulus (slope in the Weibull plot) is also similar, except a few occasions of very low load failures for the pristine glass. It is not clear if the absence

127 of these low probability outliers for the ion-exchanged surfaces are due to the consequence of ion exchange or just a thermal annealing effect during the ion exchange process.

In 40% RH (Figure 6-4b), the Weibull modulus is reduced, indicating a broader distribution of strength-affecting surface defects. The crack initiation load at 50% failure probability also decreases to ~1720 MPa for the pristine surface, ~1820 MPa for the Na+/K+- exchanged surface, and ~1760 MPa for the reverse-exchanged surface. In 90% RH (Figure 6-4c), the 50% failure load is even lower: ~1600 MPa for the pristine surface, ~1750 MPa for the

Na+/K+-exchanged surface, and ~1710 MPa for the reverse-exchanged surface. It is noted that the

Weibull plots of the 90% RH data appear to be bimodal – the modulus value is smaller for the data below 1600 MPa. The origin for this bimodal distribution is not known. In any case, the results shown in Figure 6-4 clearly demonstrate that the chemical strengthening effect of the

Na+/K+-exchange on mechanical strength of the SLS glass in humid conditions (40% and 90%

RH) is statistically significant.

Effects of ion exchange on mechanochemical wear under tangential shear

The mechanochemical wear behaviors of the pristine, Na+/K+-exchanged, and reverse- exchanged SLS surfaces at 40% and 90% RH are compared in Figure 6-5. In the 40% RH condition, the Na+/K+-exchanged and reverse-exchanged surfaces show slightly shallower wear track than the pristine surface (Figure 6-5a). Surprisingly, the Na+/K+-exchanged (chemically- strengthened) surface shows much larger wear depth than the pristine surface at 90% RH (Figure

6-5b), indicating that the wear resistance property of SLS in highly humid conditions is lost even though it has superior mechanical strength in normal indentation tests (Figures 6-2 – 6-4). When the subsurface K+ ions are exchanged back with Na+ ions (Figure 6-1b), the wear resistance at high humidity is regained (Figure 6-5b). Note that the reverse-exchange reduces the surface

128 compressive stress (as evidenced by the decrease in hardness from ~8.5 GPa to ~7.5 GPa at the

200 nm indentation test in Figure 6-2a). These results suggest that the resistance of SLS to the shear-induced mechanochemical reactions leading to wear in high RH conditions is a unique property or consequence of the presence of subsurface Na+ ions bound to NBOs in SLS and not related to the compressive stress.

Figure 6-5. Cross-section line profiles of wear tracks produced on the pristine, Na+/K+-exchanged SLS glass, and reverse-exchanged SLS surfaces after tribo-testing in (a) 40% RH and (b) 90% RH condition.

Effects of ion exchange on subsurface silicate network and hydrous species

The Si-O-Si asymmetric stretch peak at 1050-1100 cm-1 is sensitive to a minute change in the distribution of the Si-O bond lengths in the glass network.202 The SR-IR spectra of the pristine, Na+/K+-exchanged, and reverse-exchanged SLS surfaces are shown in Figure 6-6a. Upon the Na+/K+-exchange, the peak position of the Si-O-Si stretch band is red-shifted by ~3 cm-1, implying that the average Si-O bond length is slightly increased. Based on the bond length dependence of the peak position from previous molecular dynamics simulation study of silica,202

129 this red-shift may correspond to an increase in the bond length by 0.0006 Å. The accuracy of this estimation remains to be tested; but, its trend is congruent with the network dilation upon exchange of smaller Na+ ions with larger K+ ions.247,248 Upon the reverse-exchange, the peak position does not blue-shift back to the original position. The silicate network appears to remain as dilated as the ion-exchanged state.

The ATR-IR spectral features of the three surfaces are almost identical; differences are very minor or not obvious above the noise level (Figure 6-6b). This result suggests that the total amount of subsurface hydrous species does not vary significantly with these ion-exchange processes. However, this does not mean that the subsurface hydrous species are in the same chemical environments; it just means that the changes are too subtle to be detected with ATR-IR.

The hydrogen bonding interactions of hydrous species were analyzed with SFG (Figure

6c). From our previous study, the sharp OH features in the SFG spectra of SLS float glasses are found to vary with the glass thickness and are attributed to the subsurface OH species formed

110 during the SO2 dealkalization process during the float glass manufacturing. The peak near 3700

 3740 cm-1 is the free Si-OH group.104,249,250 The origin of the peaks at >3740 cm-1 are not fully understood, but they appear to be associated with the non-equilibrium nature of glass since they are not observed in the equilibrium systems (such as crystalline phase or bulk liquid).31,95,110,251,252

In Figure 6c, the spectral changes in the >3700 cm-1 region (highlighted with the light-yellow background) is negligible. The species responsible for the peaks at >3700cm-1 can be removed only after annealing the float glass above its glass transition temperature (the Tg of float glass tested here is 550 oC).110

In contrast, the spectral region of the stretch modes of hydrogen-bonded OH groups

(highlighted with the light-blue background in Figure 6-6c) shows some changes upon Na+/K+- exchange and reverse-exchange. First of all, it was confirmed that the hydrogen bonding interactions of subsurface hydrous species do not change by heating at 400 oC without ion

130 exchanges because the temperature was lower than the Tg of the sample. Note that the OH stretch peak position is a function of the hydrogen bond strength or the O-HO distance – the stronger the hydrogen bond is, the larger the OH stretch peak is red-shifted.101,102 After the Na+/K+- exchange at 400 oC, the ~3550 cm-1 peak of weakly-hydrogen bonded species disappears and the

~3160 cm-1 and ~3370 cm-1 are slightly blue-shifted to ~3200 cm-1 and ~3425 cm-1, respectively.

Upon the reverse-exchange of K+ with Na+ at 350 oC, it can be seen that these two peaks red- shifted back toward their original positions, although they become weaker and broader.

Figure 6-6. (a) SR-IR, (b) ATR-IR, and (c) SFG spectra of the pristine (black), Na+/K+-exchanged (red), and reverse-exchanged SLS surfaces. In (c), the SFG spectrum of the pristine glass after heating to 400 oC in air is also shown for comparison.

Discussion

It is well known that chemical strengthening originates from the compressive stress in the surface region upon exchange of smaller modifier ions in the glass network with larger ions. 83,238

The data presented in Figures 6-2 – 6-4 confirm the chemical strengthening effect due to compressive stress in the SLS surface after the Na+/K+-exchange. The chemically-strengthened surface has higher hardness (more difficult to make plastic deformation), higher fracture toughness

131 and higher crack initiation load under indentation applied normal to the surface. However, such mechanical property improvements observed in the indentation tests are found to be irrelevant with the resistance of SLS to mechanochemical wear in high humidity conditions (Figure 6-5). In order to confirm this surprising observation, we have carried out the Na+/Ag+-exchange for SLS glass and observed the same trend (see Figure S6-1 in the Supporting Information).

These results clearly indicate that the mechanisms responsible for the mechanochemical wear of glass under tangential shear are different from those improving mechanical responses under normal indentation.58 A similar observation was previously made from comparative studies of different types of glass. The current study on chemical strengthening effect of SLS glass further proves that the responses of glass network to normal and tangential stresses are quite different.

When this markedly superior resistance of SLS to mechanochemical wear in high RH conditions was observed for the first time,33,39 it was speculated if the leaching of Na+ ions from the

- NBO (SiO ) sites and the ingress of water to those sites forming Si-OH and H2O (so,

+ + stoichiometrically equivalent to replacing Na with H3O ) could induce a compressive stress in the

SLS surface and that might be responsible to the improved wear resistance. The data presented in this study clearly disprove that earlier hypothesis.

It seems that the property of alkali ions attached to NBOs is important in the mechanochemical wear of SLS. The K+ ions exchanged with the Na+ ions in SLS does not induce the wear resistance at high RH conditions. It might be due to a difference in chemical activity or leachability of Na+ versus K+ in the SLS network. Even if the glass network is dilated by the inclusion of larger ions, the molar volume of the ion-exchanged glass network is still smaller than that of the glass with the same composition produced from the melt.248 Thus, the K+ ions occupying the sites originally occupied by Na+ ions in SLS may not be as mobile or leachable as the original

Na+ ions. If those ions cannot come to the shear plane where mechanochemical reactions take place, then they may have a negligible impact on the wear behavior.

132 From the comparative studies of different types of glass,36 the interplay between the Si-O-

Na+ groups and the interfacial hydrous species are believed to be critical. The superior resistance to mechanochemical wear in high RH conditions is unique to SLS, not observed for pure silica

(fused quartz), alkali borosilicate (Schott Borofloat33), barium aluminosilicate (Schott AF45), sodium aluminosilicate (base glass of Corning Gorilla-2), and chemically-strengthened aluminosilicate (Corning Gorilla-2). Alkali ions in Borofloat33 are not leachable due to the presence of silica-rich domains formed through micro-phase separation in the glass.253 The trace amount of sodium ions in AF45 is not leachable since they are in the bulk; but, when they are pushed and accumulated as NBO/Na+ pairs near the surface via thermal poling, they can induce a

37 + - good wear resistance in high humidity. The Na ions associated with the AlO4 sites in the network are not as leachable as the ones associated with NBOs.254,255

Based on the SFG analysis results (Figure 6-6c), the free Si-OH groups responsible for the

~3700 cm-1 peak and the hydrous species responsible for the peaks above >3740 cm-1 are irrelevant to the mechanochemical reaction activities of the SLS surface. The hydrogen-bonded subsurface hydrous species ascribed to the ~3160 cm-1 and ~3370 cm-1 peaks might play critical roles. The loss of wear resistance at 90% RH (Figure 6-5b) appears to correlate (or coincide) with the blue-shifts of these peaks to higher wavenumbers (weaker hydrogen bond interactions; Figure

6-6c). Further details in the interplay between the hydrogen bonding interactions of these species and the wear resistance could not be obtained from the current experimental study and may require computational simulations.

Conclusion

The effects of ion exchange (chemical strengthening) for SLS glass on mechanical indentation strength and mechanochemical wear resistance were studied. The exchange of Na+ in

133 SLS with K+ produces a compressive stress in the surface region, increasing the resistance to crack opening upon normal indentation, which is congruent with chemical strengthening effects.

But, it deteriorates the resistance to mechanochemical wear upon tangential shear in high humidity conditions (90% RH). This result suggests that the mechanochemical wear behavior might be sensitive to the type of modifier ions associated with NBOs in SLS glass, but not the compressive stress in the surface. Comparison with structural characterization results suggests that subsurface hydrous species with certain hydrogen bonding interactions with the silicate network might be involved in mechanochemical reactions under tangential shear of the glass surface.

Supporting Information

For a comparison with the compressive stress effect produced upon exchange of Na+ with K+, a set of clean SLS samples were treated with the Na+/Ag+ exchange process. The clear SLS float glass

o samples were immersed in a AgNO3 molten bath at 250 C for 1 hour. The retrieved samples were air cooled and rinsed with DI water. The cross-sectional EDX analysis was performed to confirm the ion exchange. The hardness and modulus of the Na+/Ag+-exchanged SLS surface was measured with nanoindentation. The mechanochemical wear behavior was tested with the ball-on-flat tribometer at a 0.2 N load for 400 reciprocating cycles at RH = 40% and 90% conditions. These data are shown in Figure 6-S1.

134

135 Chapter 7

Complex refractive index of silica, silicate, borosilicate, and boroaluminosilicate glasses-Analysis of glass network vibration modes with specular-reflectance IR spectroscopy

Submitted to Journal of NonCrytalline Solids

Overview

A novel mathematical algorithm was developed to calculate refractive index (n + ik) from specular reflectance infrared (SR-IR) spectra in the strongly-absorbing glass network vibration region. The method is named as two-angle SR-IR (TASR-IR), since it is based on the

Fresnel equations of specular reflectance at two incidence angles (10 and 45). The results obtained from TASR-IR are comparable with the values obtained from spectroscopic ellipsometry. The TASR-IR method allows one to obtain the peak positions and intensities of fundamental network vibration modes of glass from the imaginary component (k) of complex refractive index without convolutions from the dispersion effect due to chains in the real component (n) of refractive index. The TASR-IR method is applied to silica, silicate, borosilicate, and boroalumunosilicate glasses; tentative peak assignments of glass network vibrations are proposed. The origin and concept of peaks in the vibrational spectra of glasses is discussed.

Introduction

The distribution of bond parameters such as Si-O bond length, Si-O-Si dihedral bond angle, and O-Si-O tetrahedral bond angle in silicate-based glass materials are believed to influence their chemical durability, mechanical property, and mechano-chemical reactivity (for example, stress

136 corrosion and wear in humid environments).1,20,31,64,77,252 The bond length and angle distributions could be obtained from neutron scattering or extended X-ray absorption fine-structure (EXAFS) analyses.4,5,256 However, such experimental techniques are not readily available; thus, a simple and routine experimental technique applicable to various compositions of glass would be practically useful. We recently reported that the peak position of the Si-O-Si asymmetric stretching vibration

-1 202 mode (Si-O-Si,as; typically 1000-1100 cm ) correlates with the average Si-O bond length (푑̅Si-O).

This correlation works best in the absorbance spectrum where the peak intensity, 퐴(휔), at a given wavenumber (휔) can be converted to the absorptivity, 푎(휔), using the Beer-lambert law: 퐴(휔) =

푎(휔) ∙ 푏 ∙ 푐 where 푏 is the length of beam path and c is the concentration of species of interest within the beam path. Then, 푎(휔) is directly related to the imaginary component of the complex refractive index, 푛(휔) + 푖푘(휔) , through the relationship 푎(휔) = 4휋푘(휔)/휆 where  is wavelength (= 1/휔).96

In transmission IR analysis, the intensity ratio of transmitted (퐼푡) and incident (퐼표) beam at the surface normal direction is defined as transmittance 푇() = 퐼푡/퐼표 . Absorbance is then calculated as: 퐴() = − log 푇(). Here, the dispersion effect arising from the real part of the complex refractive index, 푛(), is usually negligible in 푇() measurements at normal incidence

(휃푖 = 0 in Figure 7-1a); at least, this is the case for a weakly absorbing medium. Unfortunately, the transmission IR analysis of typical silicate glass networks is not possible unless the sample thickness is less than 1 m because of the extremely high absorptivity of glass in that spectral region.96,257 For this reason, IR analysis of glass is often performed by measuring the intensity of a reflected beam (퐼푟) at the specular reflection direction (휃푖 = 휃푟 in Figure 7-1a) and reflectance

푅() = 퐼푟/퐼푖 is plotted. Calculation of − log 푅() provides a spectrum that is dimensionally equivalent to 퐴(). However, it should be noted that 푅() is convoluted by the dispersion of 푛() and the absorption of 푘(); thus, it cannot be considered to truly reflect 푘().96,258

137 In theory, 푅() can be deconvoluted into 푛() and 푘() using the Kramers-Kronig (KK) transformation.259–261 However, in practice the KK conversion does not work well for evaluating the network stretching modes of silicate glass materials. One limitation arises from the fact that, in theory, the KK algorithm requires integration from - to + in the wavenumber () domain. The reality is that IR spectra obtained with typical mid-IR instruments are truncated on the low- wavenumber end near 200-400 cm-1 due to detector limitations, and problematically tends to overlap a region containing various low-frequency vibrational modes. Without definitive knowledge of the spectral features at  <400 cm-1, errors propagate in the KK transform into the network stretching mode region, and are often manifest in artifacts such as negative backgrounds and artificial peak shifts. Another issue is that the KK algorithm available in many commercial

표 instruments is derived for non-polarized reflection at the normal incidence angle (휃푖 = 휃푟 = 0 ).

Accordingly, errors will be inevitable if non-polarized specular reflectance (SR) IR spectra are

표 measured at an oblique angle (휃푖 = 휃푟 > 0 ). Figure 7-1b and 7-1c shows an example of converting the SR-IR spectra of a soda lime silica (SLS) float glass  collected at incidence angles of 10o and

45o  to the corresponding 푛() and 푘() spectra using the built-in KK conversion function typical of a commercial spectrometer. Note that 푘() values are negative at  < 1000 cm-1, which is physically not feasible.

In order to overcome these limitations, in this paper we describe a new algorithm that allows calculation of 푛() and 푘() from non-polarized SR-IR spectra collected at two incidence angles  10o and 45o; hereafter, this method will be referred to as the “two-angle SR (TASR)” IR method. The method was validated by processing a data set simulated with known 푛() and 푘() spectra, and compared with the data obtained from variable-angle spectroscopic ellipsometry

(VASE).262 The TASR-IR method was subsequently applied to flat glass samples with different types of glass networks, specifically silica, silicate, borosilicate, and boroaluminosilicate glasses.

138 The peaks observed in the 푘() spectra of these glasses were deconvoluted into multiple components, and the physical origins of individual components are discussed.

Figure 7-1. (a) Schematic representation of SR-IR analysis of glass. (b) SR-IR of soda lime silica (SLS) glass spectra collected at an incidence angle of 10o and 45o. (c) Complex refractive index (n and k) values of SLS glass obtained using the Kramers-Kronig transformation algorithm of the built-in software of a FTIR instrument.

Methods

Experimental analysis

SR-IR spectra were collected using a V70 spectrometer (Bruker) equipped with a DTGS detector and a SeagullTM variable-angle reflection accessory (Harrick Scientific). It should be noted that a small misalignment of the incidence angle in data collection can cause a large error in calculation; thus, it is important to accurately align the incidence angle to 45o (see Figure 7-S1 in

Supporting Information, SI). Also, the divergence of the incidence angle should be minimized; this can be done by reducing the IR beam diameter (see Figure 7-S1 in SI). A gold mirror was used as a reference for background. One hundred scans with a resolution of 6 cm-1 were collected and averaged. Spectra were collected under N2 environment. No atmospheric correction was used during the data collection. The incidence angle was calibrated under the manufacture’s instruction.

The aperture size was set to be 0.5 mm to minimize the divergence of IR beam.

139 For comparison, VASE measurements were performed using an IR-VASE instrument (J.A.

Woollam) equipped with a DTGS detector (working range 8000-250cm-1). Spectra were collected at a 16 cm-1 resolution at 50° and 55° incidence angles. The latter represents the Brewster angle for most silicate glasses in the mid-IR region where VASE data exhibits the highest signal-to-noise ratio. Although most data discussed in this paper refer to the strongly-absorbing IR region

<1200cm-1 wherein backside surface reflections are virtually eliminated, efforts were nevertheless made to evaluate any deleterious effects of backside reflections on the VASE data. Accordingly,

VASE data were collected using both a backside-index-matching method (i.e. using transparent tape263) and a more rigorous backside-roughening-plus-index-matching method. The latter was found to be identical with the former, verifying the negligible impact of backside reflections in this spectral region. IR-VASE data were processed using the WVASE® software platform (J.A.

Woollam). The same program was also used to model  and  for a SLS glass based on published

푛() and 푘() data.

Commercial flat glass samples were used in this study. The glass samples were cut into a square pieces of ~25 mm  ~25 mm. Fused quartz slides were obtained from Technical Glass

Products (GE124 grade), and used as a model system for “pure” silica glass. As a model system for sodium calcium silicate glass, two SLS float glass panels were analyzed: (1) 700 m thick panels provided by Asahi Glass Co. and (2) 4mm thick panels obtained from the PPG Industries

Works #6 plant (now Vitro Architectural Glass, Carlisle, PA). In order to minimize spectral artifacts due to a dealkalized surface layer, the SLS glass samples were annealed at 600 oC for 2 hours and slowly cooled down to room temperature over 10 hours.110 The 4 mm thick panel was thermally tempered at United Plate Glass (Butler, PA). A Borofloat®33 panel from Schott Inc. was used to represent a typical alkali borosilicate glass. Two types of alkaline earth boroaluminosilicate glass were analyzed: AF45 from Schott Inc. and Willow® from Corning

Incorporated. Unless specified, commercial glass samples were measured on air-side or melt

140 surfaces in their as-received condition. Another type of boroaluminosilicate glass is the

International Simple Glass (ISG), which was produced as a model nuclear waste glass by MoSCI

Corporation.264 A small disk of ISG glass was cut from a boule and polished to an optically clear finish with SiC slurry.

Mathematical algorithm to obtain 풏() + 풊풌() from two-angle SR-IR data

The SR-IR method utilizes a few special mathematical conditions within the polarization- dependent Fresnel equations, namely that (i) the polarization dependence of SR-IR vanishes at i

= 0o, and (ii) the s-polarized spectrum can be analytically calculated from the non-polarized

o o o spectrum at i = 45 . Subsequent conversion of the 푅(i = 0 ) and 푅(i = 45 ) spectra to the 푛() and 푘() spectra is thus equivalent to solving two algebraic equations for two unknowns; so, the degrees of freedom are zero and, in principle, an exact solution can be obtained. Practically, it is

o o not possible to collect the R() spectrum at i = 0 ; the lowest i is about 10 in typical

o o commercial FTIR instruments. Since the 푅(i = 10 ) value is very close to the 푅(i = 0 ) value (as

o o shown in the simulation plot in Figure 7-2), it can be assumed that 푅(i = 0 )  푅(i = 10 ) initially and the error introduced by this assumption can be corrected through iteration. The overall flow of the algorithm is shown in Figure 7-2.

141

Now, let’s discuss the mathematical algorithm of the TASR-IR method. At normal

o incidence (i = 0 ), there is no distinction between s- or p-polarization. Thus, non-polarized reflectance spectra can be expressed as:265

(푛−1)2+푘2 푅(0°) = (7-1) (푛+1)2+푘2

This allows expressing 푘 as a function of 푅(0°) and 푛. Since 푘 cannot be negative, only one solution can be obtained:

(푛+1)2∙푅(0°)−(푛−1)2 푘 = √ ≡ 푓(푛) (7-2) 1−푅(0°)

At an incidence angle of 45, the p-polarized reflectance (푅푝) is equal to the square of s-

266 polarized reflectance (푅푠), regardless of the refractive index. Since the non-polarized reflectance is the average of s- and p-polarized reflectance, 푅푠 can be obtained from the non-polarized reflectance at this specific angle:

−1 + √1 + 8 푅(45°) 푅 (45°) = > 0 (7-3) 푠 2

142 Use of both Equations (7-1) and (7-3) allows one to obtain both amplitude and phase information from non-polarized SR-IR spectra at a given wavenumber ().

The refraction angle (휃푡) is another important term that needs to be expressed as a function

2 of 푛 and 푘. The following expression of cos (휃푡) can be derived:

sin(휃) 2 sin2(휃)[푛2−푘2−푖(2푛푘)] cos2(휃 ) = 1 − sin2(휃 ) = 1 − ( ) = 1 − (7-4) 푡 푡 푛+푖푘 (푛2−푘2)2+(2푛푘)2

2 As shown here, cos (휃푡) is a complex number. To simplify the calculation process, we can

2 2 2 express cos(휃푡) = 푎 + 푖푏 and cos (휃푡) = 푎 − 푏 + 푖(2푎푏). By comparing this expression with equation (7-4), the following relationships can be derived:

(푛2−푘2) sin2(휃) 푎2 − 푏2 = 1 − ≡ 푦(푛) (7-5) (푛2−푘2)2+(2푛푘)2

2nk sin2(휃) 2푎푏 = ≡ 푧(푛) (7-6) (푛2−푘2)2+(2푛푘)2

Note that 푘 can be replaced with 푓(푛) using equation (7-2) in this expression. Thus, 푎 and 푏 terms in equations (7-5) and (7-6) can be expressed as a function of 푛 only:

√푦(푛)+√푦2(푛)+푧2(푛) 푎(푛) = (7-7) √2

−2푦(푛)√푦(푛)+√푦2(푛)+푧2(푛)+[푦(푛)+√푦2(푛)+푧2(푛)]3/2 푏(푛) = (7-8) √2푧(푛)

In the Fresnel equation, the reflection coefficient of the s-polarized light can be expressed as follows:

(푛+푖푘)∙(푎+푖푏)−cos(휃) [(푎푛−푏푘)−cos(휃)]+푖(푏푛+푎푘) 푟 = = (7-9) 푠 (푛+푖푘)∙(푎+푖푏)+cos(휃) [(푎푛−푏푘)+cos(휃)]+푖(푏푛+푎푘)

The equation can be rearranged into a complex number form:

(푎2+푏2)(푛2+푘2)−cos2(휃) 2 cos(휃)(푏푛+푎푘) 푟 = + 푖 (7-10) 푠 (푎2+푏2)(푛2+푘2)+cos2(휃)+2 cos(휃)(푎푛−푏푘) (푎2+푏2)(푛2+푘2)+cos2(휃)+2 cos(휃)(푎푛−푏푘)

143 Again, 푘 in this equation can be replaced with 푓(푛) using equation (7-2); thus, equation

∗ (7-10) is a function of 푛 only, at a given angle and wavenumber. Since 푅푠 = 푟푠푟푠 , equation (7-10) can be used to calculate the reflectance of s-polarized light at 45o, which can be compared with the non-polarized reflectance using equation (7-3):

(2 ∙ 푟 (45°) ∙ 푟∗(45°) + 1)2 − 1 푅(45°) = 푠 푠 ≡ 푔(푛) (7-11) 8

Since 푅(0°) value is not available, the 푘 = 푓(푛) in equation (7-2) can be initially approximated using 푅(10°). As shown in Figure 7-3b in Section 3.1, there is only one solution of

푛 for equation (7-11). Once 푛 is determined by solving equation (7-11), then 푘 is determined by equation (7-2). This calculation is repeated at every  in the spectrum.

o o To correct the errors generated by assuming 푅(i = 0 )  푅(i = 10 ), a short iteration

o o loop can be employed to estimate the real 푅(i = 0 ) from 푅(i = 10 ). Once the approximate values of 푛1() and 푘1() are found in the first cycle of the iteration, those values can be used to

o o predict 푅(i = 0 ) from 푅(i = 10 ) using the correction factor, 퐶 = 푅(푖 = 0°)/푅(푖 = 10°), shown in Figure 7-2. This iteration can be repeated until the 푛푗() and 푘푗() values converge to the final solution.

Results

Validation of the mathematical algorithm using known 풏() + 풊풌()

144

In order to verify the algorithm described previously, we generated theoretical reflectance spectra (Figure 7-3a) using known 푛() and 푘() values from the literature for an SLS glass

(shown in the inset of Figure 7-3a)46 at 10o and 45o incidence angles using the Fresnel equations.96

The theoretically-generated spectra were then treated as the inputs – 푅표(10°) and 푅표(45°) – in the algorithm described in Figure 7-2. As an example, Figure 7-3b compares the 푔(푛) value

-1 calculated from equation (11) and 푅표(45°) at  = 1097 cm . The cross-point is the solution for

푛1(). This calculation is repeated for all data points in the 푅표(10°) and 푅표(45°) spectra. The

푛1() and 푘1() values from the first cycle of iteration has some errors because the correction factor 퐶1 is assumed to be 1 (Figure 7-3c). This error is reduced in subsequent iterations by utilizing the correction factor 퐶푗() calculated from the 푛푗−1() and 푘푗−1() values from the previous iteration cycle. Once the difference between two subsequent cycles becomes negligible, then the converged 푛푗() and 푘푗() can be taken as the final solution (Figure 7-2). For the theoretical data set shown in Figure 7-3a, only three iteration cycles were needed to converge the calculated value to the true value: in Figure 7-3c, the difference of 푛3() and 푘3() from the initial input value is <10-2 % and <10-7 %, respectively.

145

Figure 7-4. Measured SR-IR spectra of (a) fused quartz and (b) SLS glass (700m thick) collected at incidence angles of 10o and 45o. Complex refractive index values (n, k) values are then calculated using the two-angle SR-IR method for (c) fused quartz and (d) SLS glass, respectively. The SR-IR data were collected after calibrating the incidence angle to 45 and with a reduced beam size (aperture = 50%, instead of 100% full opening) to reduce the divergence of incidence angle.

To further validate the TASR-IR method, experimental SR-IR spectra of fused quartz and

o o an annealed SLS glass (0.7 mm thick) were collected at 푖 = 10 and 45 (Figures 7-4a and 7-4b).

The experimental data were processed with the algorithm described in Section 2.2. The calculated

푛() and 푘() spectra are shown in Figures 7-4c and 7-4d. The calculated spectra exhibit

Kramers-Kronig (KK) consistency, and the overall spectral features agree reasonably well with the spectra previously reported in the literature.46,267 Note that small discrepancies with the literature data are attributable to differences in thermal history or surface preparation (Figure 7-S2 in SI). For the TASR-IR analysis, it is important to accurately set the incidence angle to 45o and reduce the IR

146 beam size on the focusing lens before the sample; an offset or divergence in the IR incidence angle of the FT-IR unit used will deteriorate the accuracy of the calculated data (Figure 7-S3 in SI).

In Figure 7-4, it should be noted that the “maximum” peak positions in 푅() and 푘() spectra are different. The peak shape in 푅() is convoluted with the dispersion of 푛(), so the peak position in 푅() cannot be taken as the resonance energy of the vibrational excitation upon

IR absorption. It is the peak position in the 푘() spectrum that truly reflects the IR absorption resonance energy.96,202

Comparison with spectroscopic ellipsometry (SE) analysis

Another technique commercially available to probe and analyze vibrational spectra in the mid-IR region is IR-VASE. Although commercial ellipsometry units typically come with powerful data fitting software, the accuracy of the fit result is often difficult to determine without independently obtained reference data. In that case, the 푛() and 푘() spectra calculated from the TASR-IR method can be used as a reference for the bulk. This section compares the 푛() and

푘() spectra obtained from IR-VASE and the TASR-IR method.

Similar to the TASR-IR method validation, we first used known 푛() and 푘() values to calculate () and () spectra (Figure 7-5a), and then fitted the () and () spectra with a general oscillator model to get new 푛() and 푘() spectra. When the 푛(푗) and 푘(푗) values are calculated at each data point,  and  values at 푗, the calculation result matches the original data perfectly (Figure 7-5b). This is the same as in the TASR-IR data processing (Figure 7-3).

However, when the overall () and () spectra are fitted with a model constructed with a combination of multiple oscillators, the fit result does not perfectly match (Figure 7-5c).

147

o Figure 7-5. (a)  and  spectra at 푖 = 55 simulated using the known n + ik values for ISG glass from Ref. 16. (b) point-by-point calculations and (c) fitting of the simulated  and  spectra (dashed lines). The symbols are the values calculated from the simulated spectra and the lines are the original data used for the simulated spectra.

In typical IR-VASE fitting, the model is constructed with multiple strong absorbers being represented by a series of Gaussian or Gaussian-Lorentzian peaks. The fitting procedure starts with initial guesses on the number and shape of possible peaks, and seeks to minimize the mean-squared- error over the entire spectral region. Although the overall shapes match reasonably well, the fit result shown in Figure 5c slightly deviates from the true value in the 1150 – 1300 cm-1 region. The misfit could arise from the fact that it is not known a priori how many peaks should be included in the fitting. In such cases, having independently-obtained reference data would be helpful to validate the IR-VASE fit results and/or improve the initial guess for oscillator fits. The TASR-IR method can be used for this purpose, at least for bulk samples without any surface layers or coatings. The current algorithm (Figure 7-2) cannot distinguish the surface versus bulk contributions; it will give a weighted-average value within the IR penetration depth for any surface layers present.

Figure 7-6 compares the TASR-IR and IR-VASE methods for the same sample. Here, the

ISG glass was used for comparison of two results. To the best of our knowledge, the refractive index in the silicate stretch vibration region of ISG has not been reported in the literature. For a single-side polished ISG glass, two SR-IR spectra at an incidence angle of 10o and 45o (Figure 7-

148 6a) and two SE spectra at an incidence angle of 50o and 55o (Figure 7-6c) were measured. The

SR-IR spectra were processed with the algorithm described in Figure 7-2 and the IR-VASE spectra were fitted with the WVASE® software. The processed 푛() and 푘() spectra are shown in Figures 6b and 6d. Both methods meet the KK consistency. Accordingly, when the 푛() and

o 푘() spectra shown in Figures 7-6b and 7-6d are used to simulate 푅(), for example at 푖 = 30 , both data reproduce the 푅() spectrum that matches well with the experimental spectrum (Figure

7-S4 in SI).

Note that the exact shape of the 푘() spectra in Figures 7-6b and 7-6d are slightly different (Figure 7-S5 in SI). This could be due to differences in how the sample surface roughness and the probe beam divergence manifest in two different techniques or the fitting model in IR-VASE. The IR lamp used in commercial FT-IR systems produces non-polarized light; but, reflections from several mirrors used to steer the IR beam inside the system could slightly enhance the s-polarization component over the p-polarization. Then, this could introduce an error in the 푅(45°) spectrum used in the TASR-IR algorithm. At this moment, it cannot be determined which is the main source of the small difference in TASR-IR and IR-VASE results.

149

150 Discussions

Comparison of 풌() spectra of various types of glass

With the validation of the TASR-IR method, it is now possible to convert 푅(휔) spectrum of glass to 푘(휔) spectrum, and use a peak deconvolution algorithm to determine peak positions

(oscillator energies) and peak intensities (oscillator strengths) of individual components in the network vibration modes of glass. The goal is to find commonalities and/or differences in network structures among different glasses or the same glass with different thermal histories. We applied the TASR-IR algorithm (Figure 7-2) to the reflectance spectra of various types of glasses

– fused quartz (representing pure silica glass); SLS float panels from Asahi and PPG (as a model for sodium calcium silicate glass); Schott Borofloat®33 (as an example of borosilicate glass);

Schott AF45 and Corning Willow® (as examples of alkali-free alkaline-earth boroaluminosilicate glasses); and ISG (a sodium calcium boroaluminosilicate glass with a trace of Zr). The SLS float glass samples were annealed at a temperature above Tg to avoid artifacts due to the dealkalized surface layer in the collected 푅() spectra.97,188 The thick tempered float glass was included to test the effect of compressive stress in the glass network vibration.161,202,268 The raw 푅() spectra collected at the incidence angle of 10o and 45o are shown in the Supporting Information (Figure

7-S6). The fully-processed 푛() and 푘() spectra are compared in Figure 7-7. Once the 푘() spectra are obtained, the spectral features can be deconvoluted into several components corresponding to specific vibrational modes. The deconvoluted peak position, intensity, and width are listed in Tables 7-1 and 7-2. In the peak deconvolution, Gaussian peak shapes were used just for convenience; the exact peak shape will vary depending on the relaxation and dephasing dynamics of the vibrational modes.269 The tentative interpretation of components commonly

151 present in all glass types and subtle differences among them will be given in the following sections.

152

153 Table 7-1. Deconvoluted peaks of the 푘() spectrum of silica and soda lime silicate glasses. The data shown in the order of peak position (cm-1)/peak intensity/ width (cm-1) of Gaussian fit function. Note that the Gaussian function was used in the peak deconvolution only because of its simplicity.

- O3Si-O-Si(O2)O Sample Si-O-Si,sym Si-O-Si,asym Si-O-Si,asym Si-O-Si,asym O3Si-O-Si(O2)OH 1034/1.13/44, Fused quartz 791/0.44/52 948/0.47/63 n/a 1195/0.79/49 1101/2.12/39 0.7mm thick 771/0.44/71 870/0.26/35 945/0.68/45 1030/1.29/62 1142/0.56/65 SLS 4mm thick SLS 769/0.43/63 864/0.26/24 942/0.83/48 1031/1.17/57 1134/0.60/65 Thermally- 770/0.42/65 862/0.19/20 933/0.60/53 1038/1.31/72 1160/0.41/53 tempered SLS

Table 7-2. Deconvoluted peaks of the 푘() spectrum of borosilicate and boroaluminosilicate glasses. The data shown in the order of peak position (cm-1)/peak intensity/ width (cm-1) of Gaussian fit function. Note that the Gaussian function was used in the peak deconvolution because of its simplicity.

- Si-O-Si,asym; O3Si-O-BO3 sample Si-O-Si,sym - - Si-O-Si,asym Si-O-Si,asym B-O O3Si-O-BO3 O3Si-O-AlO3 BF®33 792/0.43/85 913/0.36/40 987/0.48/39 1081/1.71/58 1188/0.62/56 1379/0.33/73 AF45 773/0.33/69 890/0.18/48 977/0.65/50 1074/1.08/64 1177/0.41/51 1384/0.37/89 Willow® 769/0.43/91 856/0.08/23 975/0.64/56 1084/1.15/52 1172/0.55/50 1375/0.33/105 ISG 768/0.37/45 858/0.35/41 949/0.46/43 1029/1.39/69 1150/0.48/56 1388/0.27/98

Silica and Silicate glass

Since the silica and silicate glass networks are mainly constructed with the SiO4 units

(often called Q4 species), let’s consider this constituent unit first. The isolated SiO4 unit will have the tetrahedral (Td) symmetry. In the Td geometry, the stretch modes are reduced into two groups: a totally-symmetric mode with A1 symmetry and triply-degenerate asymmetric modes with T2

270 symmetry (Figure 7-8). The energy of A1 mode is typically lower than that of T2 mode. In an isolated molecule such as CH4, the A1 mode is symmetry-forbidden in IR (i.e., IR-inactive); and three asymmetric modes are all symmetry-allowed (i.e. IR-active), but they cannot be separated because their energies are all identical.271

154

Figure 7-8. Four stretch modes of an ideal SiO4 unit with the tetrahedral (Td) symmetry.

However, the perfect Td symmetry cannot be strictly applied to the SiO4 unit in the glass because the bond parameters of individual SiO4 units in the amorphous network fluctuate and

202 deviate from the ideal Td geometry. Also, it should be noted that the Td symmetry of the individual SiO4 unit breaks even in crystalline solids. For example, in the olivine lattice (Mg2- xFexSiO4), the SiO4 vibrational modes are coupled and the symmetry is reduced to Cs(m), removing

272 the degeneracy of T2 modes and splitting them into three peaks. For the same reason, the A1 mode becomes detectable in IR of solid materials. It is important to note that the stretch modes of

SiO4 units in the silica and silicate glass are coupled; in other words, they are delocalized across multiple SiO4 units. In the glass network, it is not possible for the single SiO4 unit to vibrate independently without influencing surrounding units that are covalently connected. Even in the crystalline phase formed through hydrogen bond networks, which are much weaker connections than the covalent bond network in the glass, the delocalization of stretch modes is observed.273

Thus, it is obvious that the Si-O stretch peaks cannot be assigned to the symmetric A1 or asymmetric T2 modes of individual SiO4 units; they are the collective vibrations of inter-connected

-1 SiO4 units. Based on this argument, the 770800 cm component could be attributed to a collective vibration mode in which a large fraction of bridging oxygen (BO) groups of the SiO4 units vibrates with a relatively symmetric fashion (Si-O-Si,sym). Other higher energy components, except the 920 –

155

-1 940 cm component in SLS glass, would be collective vibration modes where most SiO4 units oscillate in a relatively asymmetric fashion (Si-O-Si,asym). Among these, one specific mode appears to be dominant, which appears at 1035 – 1090 cm-1. Although the exact description is different, this assignment is congruent with conventional peak assignments of the glass IR peaks.45,274–276

It is important to note that the peak position of the dominant asymmetric peak correlates well with the average Si-O bond length or its distribution in the glass network.202 For example, in the case of the thermally-tempered SLS glass, the silicate network within the IR penetration depth from the surface are under compressive stress; this glass shows the asymmetric stretch peak ~7 cm-

1 higher than the annealed glass (Table 7-1), which is consistent with the theoretical prediction.202

Being able to convert 푅(휔) to 푘(휔) opens up an unprecedented opportunity to use SR-IR to probe the distortion or variation in bond length distribution upon various thermal or mechanical treatments of silica and silicate glasses with high precision.

The shoulder peak near 930950 cm-1 in the 푘() spectra of SLS glasses (Figures 7-7b –

7d) is typically assigned to the Si-O stretch of the non-bridging oxygen (NBO, Si-O-) or silanol (Si-

OH) groups, which could be collectively called Q3 species. The intensity of this shoulder peak often correlates with the network modifier concentration in SLS glass, supporting its assignment to

NBO.277,278 However, in the context of the collective group vibration concept, this component cannot be assigned to the individual isolated Q3 species. It must be the collective vibrations of the

- network region where Q3 species are involved. In other words, it can be said that the 930950 cm

1 component appears when the network connectivity of Q4 species is interrupted by the presence of

Q3 species (Table 7-1).

In the case of pure silica glass (Figure 7-7a), the NBO component is absent; thus, the ~948

-1 cm component cannot be attributed to NBO (Table 7-1). It is conceivable that some Q4 species could be highly distorted locally when the amorphous network is entirely constructed with Q4 units only. Then, such locations could behave as a disruption point of the Q4 network connectivity in

156 delocalized vibrations, similar to the Q3 species in SLS glass. The highly distorted network region may cause a red-shift of the energy of asymmetric vibration modes. In SLS glass, the presence of

-1 Q3 species in the network may induce a larger red-shift to 860 – 870 cm . Further details cannot be determined in this study and may require computational simulations.

In some literature, the 11501200 cm-1 component is proposed to originate from the longitudinal optic (LO) mode, in analogy to the phonon bands observed for ionic crystals.279,280 If it is truly LO mode, then it should show up only in the p-polarized spectrum, but not in the s- polarization spectrum.279,280 We have shown that this band is observed in both p- and s- polarizations.96 So, it is unlikely to be a mode similar to the LO mode reported for ionic crystals.

Wilson et al. simulated IR spectrum of amorphous SiO2 using polarizable potentials and claimed

-1 270 that a totally-symmetric A1 mode may exist around 1250 cm . Or it could simply be due to the splitting of the asymmetric T2 mode in collective stretch modes of highly connected Q4 species in the amorphous network.270 For example, in the case of silica glass, the ~1034 cm-1 and ~1195 cm-

1 components could be viewed as splitting of the most prominent peak at ~1101 cm-1. In fact, the relative intensity of the ~1034 cm-1 and ~1195 cm-1 components are similar in the 푘(휔) spectrum

(Figure 7-7a); however, this cannot be seen in the 푅(휔) spectrum (Figure 7-4a) due to convolution with the dispersion of 푛(휔).

Borosilicate glass

The borosilicate glass analyzed in this study (Borofloat®33) contains a highly silica-rich phase; thus, it is understandable that the main peak near 1080 cm-1 is somewhat similar in the

푘() spectra of fused quartz (Figure 7-7a) and Borofloat®33 (Figure 7-7e). The borosilicate glass

-1 276 has an additional peak at ~1380 cm , which can be attributed to the B-O stretch (B-O) mode.

In the 푘() spectrum of Borofloat®33, the small peaks at 910~980 cm-1 cannot be attributed to

157 the NBO group. Based on the nominal composition of this glass,110 all sodium ions are likely to

- - 281,282 be associated with the BO4 (or trace amount of AlO4 ) units in the network; in other words, there is not enough sodium to create additional SiO- groups. The 910~980 cm-1 bands of the borosilicate network could be tentatively attributed to the disruption of the O3Si-O-SiO3

- connectivity due to the presence of BO4 groups. The vibrational energy of the delocalized

- stretches containing O3Si-O-BO3 units would be different from that of all O3Si-O-SiO3 units.

Boroaluminosilicate glass

The peak assignment of the 푘() spectrum of alkaline-earth boroaluminosilicate glass is even more challenging because the exact compositions of AF45 and Willow® are not publicly available (proprietary information of Schott and Corning). Nevertheless, the main peak at

-1 10501100 cm can be attributed to the Si-O-Si,asym mode of the glass network, based on the relative abundance of silica in these glasses; but, it is very likely to be modified due to the presence of B

-1 276 and Al elements in the network. The small peak at ~1380 cm is the B-O stretch (B-O). It is generally known that the NBO concentration is negligible in these glasses because of a rough

-1 balancing of alkaline-earth and Al2O3 concentrations. Thus, the shoulder peak at ~975 cm cannot be assigned to the NBO group, although its position is somewhat close to the NBO mode of the silicate glass. Similar to the borosilicate glass case, they could tentatively attributed to the

- disruption of the O3Si-O-SiO3 connectivity due to the presence of O3Si-O-AlO3 units in the glass network.

In the case of the ISG (a model glass for nuclear waste glass), the exact composition is known – 60.2% SiO2, 16.0% B2O3, 3.8% Al2O3, 12.6% Na2O, 5.7% CaO, and 1.7% ZrO2 (in

264 - - mole). Based on this composition, all cations are expected to be associated with BO4 and AlO4 anions in the glass network. This glass has the main peak at ~1029 cm-1, which is significantly

158 lower compared to other types of glasses. Based on the spectral shape in 푘(), it is clear that the boroaluminosilicate network in ISG is drastically different from those in AF45 and Willow®

(display glasses). But, further details could not be inferred without detailed knowledge of the network connectivity which might be obtainable in computer simulations.

Challenges and Opportunities in analysis of 풌() spectrum of glass

The TASR-IR algorithm works well in the region where the intensity in the 푘() spectrum is large. For the weakly-absorbing spectral region where 푘 values are close to zero (for example, 

>1300 cm-1 for silica, >1250 cm-1 for silicate, and >1500 cm-1 for borosilicate and boroaluminosilicate), the algorithm does not work well. In this spectral region, a small error in

o experimental intensity of 푅(푖=0 ) introduces a relatively large error in expressing 푘 in terms of

푓(푛) in equation (2).

The accuracy of the 푛() and 푘() spectra calculated from the TASR-IR method depends

o on the 푅(푖=45 ) spectrum, which in turn depends on how accurately the incidence beam is aligned at 45o, reducing the divergence of the incidence angle, and using fully un-polarized light. The first two conditions could be met by carefully tuning the variable-angle reflection unit and reducing the

IR beam aperture (Figure 7-S1). The IR beam is steered by a highly-reflected metallic mirror; when the numbers of mirrors reflecting the beam horizontally and vertically are not equal, then the IR beam arriving at the sample surface may be slightly polarized in one direction. For example, if IR is reflected 5 times with the gold mirrors in the horizontal direction only, then the s-polarization would be about 5% stronger than the p-polarization at a wavenumber near 1000 cm-1. It is difficult to mathematically correct the effects of this deviation. One way of circumventing such complications is to use a polarizer and make the beam purely s-polarized. Then, equation (3) can

159

∗ be skipped and the 푅푠 = 푟푠푟푠 calculated from equation (10) can be compared directly with the

o experimentally-obtained 푅푠(푖=45 ) spectrum.

The 푘() spectrum obtained from TASR-IR gives information of IR-active vibrational modes of glass; IR-inactive modes cannot be probed. In this context, it is important to note that the symmetry rule used to distinguish IR-active and Raman-active modes of isolated oscillators (such as small molecules) may not be strictly applicable to the vibrational modes of glass networks. The local bond parameters of glass network are not uniform and often deviate from the equilibrium

202 value. For example, the bond length and angle of the SiO4 unit in silica have broad distributions.

Thus, the symmetry selection rule of the ideal tetrahedral geometry may be substantially loosened in the glass network.

Also, the vibration modes are not fully localized to individual Si-O-Si bonds or tetrahedral units;270 the vibration modes of glass are delocalized over a large number of units which contains many non-ideal tetrahedral geometries. This is the main reason that the spectral features of glass are so broad. For this reason, the comparison of the 푘() spectrum with density functional theory

(DFT) calculations obtained for a small number of SiO4 clusters may not be straight forward, unless the peak broadening is externally accounted for. It is more desirable to compare the 푘() spectrum with the molecular dynamics (MD) simulation results which contains a large number of structural units to reflect fluctuations of the bond parameters.202

Our recent study of using MD simulations to reproduce the vibrational spectra of amorphous silica clearly showed that the peak shape and position in the Si-O-Si stretch region correlates reasonably well with the average Si-O bond length, contrary to the previous belief that the Si-O-Si stretch peak position is a function of the Si-O-Si bond angle.202 Based on MD simulations, it is predicted that the compressive stress resulting in a shrinkage of the average Si-O bond length will cause a blue shift of the Si-O-Si stretch mode.202 The data of the thermally- tempered glass shown in Figure 7-7d and Table 7-1 confirms this prediction. More studies on the

160 correlation between the 푘() spectral features and the bond parameter distribution need to be done for more accurate interpretation of the vibrational spectroscopic features to the glass network structure. The TASR-IR method described in this study will make such studies possible since the 푘() spectrum of glass can be obtained accurately.

Conclusions

The TASR-IR method is developed to calculate complex refractive index in network vibration region of silica and silicate glasses. Results from TASR-IR method is validated and compared with the results from spectroscopic ellipsometry. The true vibrational features based on the imaginary part of the refractive index (k) are obtained and compared for silica, soda lime silica, borosilicate, and boroaluminosilicate glasses. Tentative peak assignment and proper interpretation of the vibrational features of these glass are discussed.

Supporting Information

Alignment of a SeagullTM variable angle reflection accessory

Since the algorithm of the TASR-IR method is mathematically correct (Figure 7-3), the accuracy of the calculated 푛() and 푘() value depends on the accuracy of the experimental data. The 푅(0표) spectrum is relatively insensitive to a small error in the alignment since it is near surface normal. So, most error would come from the alignment of the incidence IR beam at 45o.

It should be noted that the angle indicator scale may not be accurate. In the case of SeagullTM accessory, there are two small holes in the back of the unit which was used during the

161 construction of the unit to set the system at 45o (see Figure 7-S1b). It is important to align these two holes to accurately align the system at 45o, instead of using the approximate angle indicator.

Also, the beam diameter should be reduced to the minimum size without sacrificing the signal-to- noise ratio to reduce the divergence of the incidence beam. This can be done by reducing the aperture opening at the IR beam source.

162 Comparison of the SR-IR spectra of fused quartz used in this study and those calculated from the refractive index reported in the literature

163

o o Figure 7-S3. SR-IR spectra collected (a) at 푖=10 and 42 with the 100% opening of the IR aperture o o and (b) at 푖=10 and 45 with the 50% opening of the IR aperture. The 푛() and 푘() spectra are calculated using the spectra in (a) and the TASR-IR algorithm shown in Figure 7-2; these spectra are different from the ones calculated with the spectra in (b). The spectra in (b) and (d) are the same ones in Figure 7-4(a) and 7-4(c).

164 Comparison of 풏() and 풌() spectra of ISG obtained from TASR-IR and IR-VASE methods

Figure 7-S4. Comparison of 푛() and 푘() spectra of ISG obtained from TASR-IR and IR-VASE methods.

Kramers-Kronig consistency of the results obtained with the TASR-IR and IR-VASE methods

165

166 Raw SR-IR spectra of the flat glass surfaces analyzed in this study

Figure 7-S6. Raw SR-IR spectra of the glasses used for calculation of 푛() and 푘() spectra shown in Figure 7-7.

167 IR penetration depth in SR-IR analysis

Figure 7-S7. IR penetration depth (dp) calculated using the 푛() and 푘() spectra shown in Figure 7 for an incidence angle of 10o.

168 Chapter 8

Molecular dynamics study of correlations between IR peak position and bond parameters of silica and silicate glasses: Effects of temperature and stress

Reproduced with permission from Wiley : Luo, J.; Zhou, Y.; Milner, S. T.; Pantano, C. G.; Kim, S. H. Molecular Dynamics Study of Correlations between IR Peak Position and Bond Parameters of Silica and Silicate Glasses: Effects of Temperature and Stress. J. Am. Ceram. Soc. 2017, No. August, 1–11.

Overview

In the IR spectra of silica and silicate glasses, the shifts of the maximum intensity position of the Si-O-Si,as band upon heating or applying mechanical stress could be attributed to changes in the distribution of bond parameters such as bond length and bond angle. Upon heating, isotropic expansion occurs and the density changes; upon applying mechanical stress, anisotropic strain is induced and a significant change in the Si-O-Si bond angle is observed. From molecular dynamics simulations of silica glasses, we show that the peak position shift correlates better with the asymmetric change in the Si-O bond length distribution, rather than the Si-O-Si bridge angle, the O-Si-O tetrahedral angle, or the density change. This new finding provides an insight into how and why the Si-O-Si,as IR peak of soda lime silica (SLS) glass shifts upon chemical strengthening via ion exchange and thermal tempering.

Introduction

Residual stress in sub-surface region of glass (or simply called surface stress) plays an important role in the mechanical and mechanochemical properties of silicate glass materials.31,33,67,132,236 In multicomponent silicate glasses, compressive surface stress can be introduced by exchanging smaller ions with larger ions through ion-exchange processes using

169 molten salts or through thermal tempering, which can improve modulus, hardness, and fracture toughness.234,284,285 Tensile stress, on the other hand, promotes crack nucleation and extension, in most cases facilitated by the stress corrosion process.20,64,77 The magnitude of compressive or tensile stress is typically determined by mechanical tests such as indentation analysis and bending tests.286–288 Optical characterizations such as polariscope and infrared (IR) spectroscopy have been used as a nondestructive and complimentary means to estimate the stress in the glass. Among these, IR is of special interest since it can provide structural information about the silicate glass network.

The maximum intensity position of the Si-O-Si asymmetric stretch (Si-O-Si,as) band in IR spectroscopy is reported to be sensitive to the stress in glass surfaces. An example of the pioneering work conducted by Tomozawa and his colleague for silica glass fibers is shown in

Figure 8-1.17,268 Mechanical strain of glass can induce changes in various bond parameters of the silicate network – Si-O bond length (dSi-O), Si-O-Si bridge angle (Si-O-Si), and O-Si-O tetrahedral angle (O-Si-O). It is conceivable that the largest change in bond parameters upon mechanical strain would occur in the Si-O-Si angle because it has the lowest force constant among three

289 parameters. Thus, it was speculated that the peak shift of the Si-O-Si,as band originates from the

1,268,290 changes in the Si-O-Si angle. However, it is not clear how blue or red shifts in the stretch mode are associated with changes in the bond angle. There are mathematical model explaining a

291,292 possible correlation between Si-O-Si and Si-O-Si using a noncentral force constant term; but, the origin of this additional term has not been elucidated from first-principles theories.

Also, it should be noted that the Si-O-Si,as band of glass is extremely broad (typically greater than 150 cm-1), while the same band of crystalline silica is very sharp.48 Unlike the crystalline materials which have atoms in highly-ordered lattice positions, the amorphous nature of the glass network inevitably gives broad distributions of bond parameters. This is the main

170 reason that the IR bands of glass are so broad. Thus, determining the magnitude of shift in the maximum intensity position (often called “peak position” for simplicity) of the Si-O-Si,as band with a FWHM of >150 cm-1 requires averaging from multiple measurements and statistical significance test.1,17,268,290

In contrast to pure silica glass, a multicomponent glass such as soda lime silica (SLS) shows a larger degree of shift in the Si-O-Si,as band upon build-up of compressive stress in the subsurface region. For comparison, the data of SLS glasses with surface compressive stress prepared by ion-exchange of Na+ with Ag+ and K+ are shown in Figure 8-1. Also shown in Figure

8-1 is the surface tensile stress data obtained on the convex surface of a slide glass mounted on a three-point bending rig. If we consider the direction of the change from these data sets, the trends in silica and SLS glasses appear to be similar, except the magnitude being much larger for the

SLS glass. However, the thermally-tempered glass with a compressive surface stress show an opposite trend from the ion-exchanged glass, i.e. a blue-shift in the Si-O-Si,as peak position by ~2

-1 cm with an estimated compressive stress of ~50 MPa. These results may indicate that the Si-O-

Si,as peak position cannot be directly associated with the sign or magnitude of stress.

In this paper, molecular dynamics (MD) simulations were employed to better understand the correlation between the IR spectral changes of glass and the bond parameters of the glass network. MD simulations were carried out with the van Beest-Kramer-van Santen (BKS) force field to create pure silica glass, which is the simplest and most studied glass although it is often considered an anomalous glass.293 The FFSiOH potential was used to calculate the dielectric constants of silica glass in the mid-IR frequency range. The simulation methodology was validated by comparing MD results with experimental data for peak shifts upon heating. In order to observe clear trends with less uncertainty in simulations, extremely large mechanical strains

(5% tensile and compressive strain) were imposed via uniaxial compression or tension. Assuming

171 that the cause for spectral shifts is the same for the thermally-induced isotropic deformation and mechanically-induced anisotropic deformation of the silica network, correlations between the

Si-O-Si,as peak position and three bond parameters (dSi-O, Si-O-Si, O-Si-O) as well as glass density

() were established with the simulation results. This revealed a physical insight into the structural origin for the blue- or red- shift in the Si-O-Si,as band for the pure silica glass. Based on this finding, new hypotheses for the Si-O-Si,as peak shift for SLS glass upon chemical strengthening, three-point bending, and thermal tempering were proposed.

172 Experimental and simulation methods

Experimental details

Fused quartz slides (Technical Glass Products, Inc. OH) were used for temperature effect study. This product is normally 100% SiO2 with ppm levels of impurities. Microscope slides

(VWR, Inc.) made of soda lime silica (SLS) glass was used in the three point bending test and chemical strengthening via ion exchange. The dimensions of the microscope slide were 75 mm 

25 mm  1 mm. A SLS glass with the dimensions of 5 cm  5 cm  5 mm was used for thermal tempering treatment in this study. The tempered SLS glass was prepared by heating the glass to

620 C and held for 30 mins before it was taken out to quench in air. The surface compressive stress was estimated to be approximately 50 MPa.298 An annealed SLS glass, which served as a sample without surface compressive stress, was slowly cooled from 620 C to room temperature at a 2 C/min cooling rate.

Specular reflectance infrared (SR-IR) spectra of fused quartz and SLS glass surfaces were measured with a Bruker Hyperion 3000 Microscope (Bruker, Co.) with a 15 objective lens. The incident angle of the IR beam is approximately 20. A gold mirror was used as a reference background. A Linkam Scientific heater (model No. TS1500, U. K.) was used to control the temperature of the fused quartz during SR-IR analysis. The top of the heater was covered with a small opening which allowed the infrared beam to interact with the glass surface directly.

A home-built three point bending apparatus was used to apply tensile stress on SLS glass.

In this apparatus, a polished stainless steel cylinder placed under the glass slide applied the stress uniformly at the contact location. The tensile stress in the opposing convex surface was calculated using a beam bending theory. The maximum tensile stress applied on the SLS glass surface was about 100 MPa (data shown in Figure 8-1 and Figure 8-S1 in Supporting Information); beyond

173 this point, measurements were not reliable since the glass slides broke easily. The aperture size of the SR-IR objective lens was set to be 125 m  125 m to minimize the curvature effect during the three-point bending analysis.

The compressive stress samples of SLS glass were prepared by ion exchange in a molten

+ + o 294 salt bath. The Na /K exchange was carried out in molten KNO3 at 425 C for 6 hours and at

o 9 + + o 375 C for 8 hours . The Na /Ag exchange was carried out in molten AgNO3 at 300 C for 1 hour295. After rinsing the ion-exchanged glass surface with water at room temperature, the SR-IR analysis was carried out. The data are shown in Figure 8-1. The compressive stress in the ion- exchanged SLS glassed were taken or estimated from the literature.296,297

MD simulation details

MD simulations of silica glass systems were performed using the Gromacs simulation package.299 For the sake of computational efficiency, the silica glass system was constructed using the BKS potential, which has been shown to successfully reproduce experimental structural properties (see Supporting Information).300,301 A pure silica system consisting of 3000 particles

(1000 Si atoms and 2000 O atoms) was first equilibrated in the NPT ensemble at 7000 K and 1 atm for 10 ns. The system reached equilibrium after 10 ns at which atoms moved more than 80 times the average Si-O bond length. The system was then quenched uniformly from 7000 K to

300 K over 100 ns, corresponding to a cooling rate of 67 K/ns, and finally annealed for another

10 ns at 300 K.

To calculate the dielectric constants of the silica glass, the rigid ionic BKS potential was replaced with the FFSiOH potential, which is a static core-shell model based on periodic B3LYP calculations.302 In the core-shell model, the oxygen ion is polarizable, represented by a core and a massless shell connected via a spring, allowing calculation of the infrared spectrum of the

174 amorphous silica.270 After changing to the FFSiOH potential, the system was further equilibrated for 10 ps to allow adjustments of structural properties, such as bond angle, bond length and density to the new potential, while the Si-O network remains unchanged.

To investigate the temperature and stress dependence of the IR spectra, isobaric simulations (P=1 atm) at 100K, 300K, 500K and 700K, and isothermal simulations (T=300K) under uniaxial compression or tension with a linear deformation of 3% and 5% in the z-direction were performed for each sample. We equilibrated the systems for 20 ps at each condition before collecting polarization data for a period of 80 ps with a time interval of 4 fs needed to cover the frequency range between 200 cm-1 and 1600 cm-1.

The generalized Kirkwood equation was used to extract the frequency-dependent

270,303–305 dielectric constant, (휖(휔) = 휖1(휔) + 푖휖2(휔)), from the total dipole moment M(t). This equation is derived by Neumann and Steinhauser specifically for a polarizable cubic system with the Ewald summation method:306

푑휙(푡) 휖(휔)−휖 ℱ [− ] = ∞ (1) 푑푡 휖(0)−휖∞ and

〈푀(푡)2〉 휖(0) − 휖∞ = (2) 3푉푘퐵푇휖0

[ ] ∞ −2휋푖휔푡 ( ) 〈 〉 〈 2〉 where ℱ 푓(푡) = ∫0 푓(푡)푒 푑푡 is the Fourier transform, 휙 푡 = 푀(0) ∙ 푀(푡) / 푀(푡) is the normalized autocorrelation function of the total dipole moment, and 휖(0) is the vacuum permittivity. The optical dielectric constant at the high-frequency limit, 휖∞, was obtained from the Clausius-Mossotti equation (휖∞ − 1)/(휖∞ + 2) = 4휋휌훼/3, where 휌 = 푁/푉 is the number density of oxygen ions and α is the polarizability of the oxygen ion, which is related to the core-

2 shell spring constant 푘푐푠 and the shell charge 푞푠 in the FFSiOH model by 푘푐푠 = 푞푠 /4휋휖0훼. Note that the average total dipole moment 〈푀(푡)〉 of the as-prepared silica system has a nonzero value because of the finite system size, and the time-independent average value was subtracted from the

175 total dipole moment before the autocorrelation calculation. Also, artificial jumps of ions across the periodic boundary in the MD trajectory were removed in this calculation. The real and imaginary components of complex refractive index (푛′(휔) = 푛(휔) + 푖푘(휔)) were calculated from the complex dielectric constant 휖(휔) and the reflectance spectrum (푅(휔) = 푛′(휔)푛′(휔)∗) was calculated from the complex refractive index.96

All simulations were performed in periodic boundary conditions with an integration time step of 2 fs. For both BKS and FFSiOH potentials, a short-range cutoff of 0.55 nm was used for the van de Waals interaction in order to match the experimental density,302 whereas the particle- mesh-Ewald (PME) method was used for the long-range Coulomb interaction with a real-space cutoff of 1.2 nm. To control the temperature and pressure in the NPT ensemble simulation, a stochastic velocity rescaling thermostat (τt = 0.1 ps) and the Berendsen barostat (τp = 1 ps) were used. Anisotropic pressure coupling was set for isobaric simulation, whereas a semi-isotropic coupling was used for uniaxial deformation, which fixed the box in z direction and only allowed fluctuation in x and y axes. To improve statistics, MD results over five samples generated independently were averaged.

In addition to BKS and FFSiOH potentials, another core-shell model proposed by

Sanders et al. was also tested.307 Unlike BKS and FFSiOH potentials, the Sanders model includes three-body interactions for O-Si-O bonds and uses full ionic charges. The same methodology was applied for these three force fields to calculate the dielectric constants in the mid-IR frequency range.

176 Results and Discussions

Comparison of refractive index calculated from three different potentials

Figure 8-2 compares the 푛(휔) and 푘(휔) of the refractive index calculated with three different potentials along with the experimentally-obtained values from the literature.261 The BKS potential cannot predict the optical constants properly. The FFSiOH predicts reasonably well the peak position and shape in 푛(휔) and 푘(휔) as well as the reflectance spectrum, although their magnitudes do not exactly match with the experimental values. The discrepancy in magnitude

3 appears because the FFSiOH model has a weaker polarizability of oxygen ion (αO = 0.603 Å )

3 308 than the experimental value (αO = 1.31 Å ). The Sanders method can reproduce the amplitude of 푛(휔) and 푘(휔), but not the peak position and shape. In the Sanders model, a larger core-shell

3 307 spring constant gives an oxygen polarizability (αO = 1.56 Å ) close to the experimental value, allowing better prediction of the intensity than the FFSiOH model. Since the peak position and shape are of primary concern in thermal and stress effects, the FFSiOH method was employed in

177 this study. MD simulations for a limited set of conditions were also run for a system of 24000 particles and were compared with those from 3000 particle systems. No significant size effect was observed.

Validation of MD simulations with experimentally-observed temperature dependence

In order to validate the theoretical refractive index calculated with the method described here, its temperature dependence was compared with experimental spectra. The SR-IR spectra of a fused quartz at 300 K, 473 K, and 673 K are displayed in Figure 8-3a. The experimental Si-O-

Si,as band shows a negligible change in the lower frequency region and a significant decrease in the higher frequency region. Overall, this results in a red shift of the maximum intensity position with a slight reduction in the peak width as the sample temperature increases. When the reflectance is calculated using the simulated 푛(휔) and 푘(휔), the theoretical spectra (Figure 8-3b) show the exact same trend observed in experiment (Figure 8-3a).

The overall changes in the peak position and shape of the reflectance spectra (Figures 8-

3a and 8-3b) are related to the changes in individual 푛(휔) and 푘(휔) terms shown in Figures 3c

- and 3d, respectively. In the 푘(휔) spectrum, the entire band of the Si-O-Si,as mode around 1100 cm

1 is shifted by -20 cm-1 as temperature increases from 300 K to 700 K. On the other hand, 푛(휔) shows a large decrease in the higher frequency part of the 1000 cm-1 band with increasing temperature while remains almost unchanged in the lower frequency part. Since the SR-IR spectrum is sensitive to both 푛(휔) and 푘(휔), the change in the lower wavelength side become relatively insignificant and the high wavenumber region decrease more significantly with the temperature increase. The red shift in the maximum intensity position in the simulated SR-IR spectrum is about -14 cm-1 (Figure 8-3b), which is in agreement with the experimental data (-11

178 cm-1 in Figure 8-3a). This validates that the simulation method described here can reproduce the experimentally observed trend.

MD simulations of mechanical stress effect on refractive index and IR reflectance

The effects of mechanical stress on 푛(휔), 푘(휔), and the SR-IR spectra, simulated with

MD, are summarized in Figure 8-4. The simulation results show a blue shift in the maximum

-1 -1 intensity position of the Si-O-Si,as band (from 1098 cm to 1102 cm ) under a compressive stress of

179 3.2 GPa and a large red shift (from 1098 cm-1 to 1086 cm-1) under a tensile stress of 3.6 GPa. Similar

-1 changes are also observed for the Si-O-Si symmetric stretch band (Si-O-Si,s) near 780 cm . The

-1 bending mode of the SiO4 tetrahedra near ~450 cm shows a slight blue shift in both compressive and tensile stress. Since the shift in this mode cannot distinguish compressive versus tensile stress and no experimental data for the stress dependence of this mode were reported, it will not be discussed in this paper.

Correlations between the simulated IR spectral features and bond parameters

In order to understand the dependence of the IR spectral features on temperature (Figure

8-3) and external mechanical stress (Figure 8-4), the structural changes in the glass network were analyzed. Note that unlike crystalline materials, the bond parameters (length and angle) have a broad distribution due to the amorphous nature of the glass. Figure 8-5 displays the distributions of the Si-O bond length and the O-Si-O and Si-O-Si angles of the silica glass network simulated at different temperatures under no stress and at different stresses at 300 K. In the case of temperature changes, expansion or shrinkage is isotropic in all three dimensions. Thermal excitation makes the bond length and angle distributions broader, which is the intrinsic property of the Boltzmann

180 distribution. On further inspection, it is noted that the broadening is not symmetric. As temperature increases, the longer length side of the dSi-O distribution increases more than the shorter length side

(Figure 8-5a), and the population of the smaller angle side become more abundant than that of the larger angle side in the O-Si-O angle distribution (Figure 8-5b). The change in the Si-O-Si distribution is relatively small; this is because its distribution is already broad at room temperature. The broadness of the Si-O-Si distribution reflects the softness of this bond angle.

When external mechanical stress is applied in a uniaxial direction to the silica glass at 300K, the deformation is not locally uniform. This non-uniform deformation is accommodated most easily in the softest bond parameter. In Figures 8-5c and 8-5d, it can be noticed that the Si-O-Si distribution

181 is much more susceptible to the applied stress than the dSi-O and O-Si-O distributions. The Si-O-Si distribution shifts to a smaller angle under compressive stress and a larger angle under tensile stress.

Although small, the dSi-O and O-Si-O distributions also show asymmetric changes upon applications of compressive or tensile stress.

The potential energy is a function of bond length and angle and the frequency of a vibrational mode is a function of the potential energy curvature. If the temperature or external stress alters the vibrational band of glass, then it must be through the changes in the bond length and angle distribution. Thus, it can be hypothesized that the bond parameter dependence of the IR vibrational peaks should be the same regardless of the source of change (internal thermal energy or external mechanical stress). With this hypothesis, one can test which bond parameter (dSi-O, O-Si-O, or Si-O-

Si) affects the Si-O-Si,as position in the same way upon changing temperature and applying mechanical stress. In order to represent the asymmetric change in the distribution, correlation plots can be made using the weighted mean value of each parameter (푑̅Si-O, ψ̅O-Si-O, and 휃̅Si-O-Si) calculated from the simulated distributions.

In representing the IR absorption band, reading the peak position from the 푘(휔) spectra is more accurate than reading from the reflectance (R) spectra. This is because the absorptivity is a function of 푘(휔) only, but R is convoluted with both 푛(휔) and 푘(휔). In principle, the Kramers-

Kronig (KK) algorithm can be used to convert the experimental reflectance spectrum to the 푘(휔) spectrum; but many commercial KK conversion does not work reliably for the silicate bands of glass. In such cases, the reflectance spectra could be used as an alternative (but, caution must be taken since the true peak position might be different).96 In most experimental works reported in the

95,161,162,309,310 literature (and Figure 8-1 as well), the maximum intensity position (푚푝) is normally used, which corresponds to the most probable value in the distribution. In order to reflect the peak shape of the absorption band, the weighted mean of the peak position (̅) could be used instead.

182

Figure 8-6 displays the correlation plots of the weighted mean position (̅) of theSi-O-Si,as band in the simulated 푘(휔) spectra versus the average Si-O length (푑̅Si-O), the average Si-O-Si angle (휃̅Si-O-Si), the average O-Si-O angle (ψ̅O-Si-O), and the density () of the simulated silica system at four temperatures (100K, 300K, 500K, and 700K under stress-free condition) and with two different compressive and tensile strains (3% and 5%) at 300 K. The density dependence is shown here since it was previously proposed as the cause for spectral changes in IR.311 The correlations with the 푚푝 of 푘(휔), ̅ of 푅(휔), and 푚푝 of 푅(휔) are also shown in the Supporting Information; they show the same overall trends.

183

Among all parameters considered, only 푑̅Si-O shows a quantitatively similar slope for both internal thermal energy and external mechanical stress effects (Figure 8-6a). The correlations between Si-O-Si,as peak position versus 휃̅Si-O-Si, ψ̅O-Si-O, and  are different for the thermal and mechanical cases (Figures 8-6b, 8-6c, and 8-6d). Thus, it can be concluded that the shift in the Si-

O-Si,as peak position upon heating or applying external stress must be associated with a small change in the Si-O bond length distribution, not with changes in the Si-O-Si or Si-O-Si bond angles or the density.

Simplified vibrational normal mode analysis

The correlation between the Si-O bond length and the peak position of the Si-O-Si,as band is further scrutinized using a simplified vibrational normal mode analysis for a single SiO4 cluster. The potential energy (U) for a perfect SiO4 tetrahedron, which consists of four Si-O pairs and six O-O pairs, are calculated first. Using the tetrahedral geometry, the potential energy associated with a single Si-O bond can be written as:

3 푈 (푑) = 푢 (푑) + 푢 (√8⁄3 푑) (3) SiO SiO 2 OO where 푑 is the bond length, 푢SiO(푟) and 푢OO(푟) are the pair interactions (including van der Waals and Coulomb forces) given by the FFSiOH force field. The 푈SiO(푑) is plotted in Figure 8-7a. The bond length at equilibrium is estimated to be 1.628 Å3, which is close to the average bond length obtained from MD simulations (1.636 Å3), suggesting this approximation is reasonable. This simple model of the local geometry of a SiO2 network can be used to predict how the Si-O-Si vibrational band will shift with small changes in bond lengths, with predictions that remain valid for silica glass networks. By taking the second derivative of the bond potential, the effective

2 2 spring constant 푘SiO = 휕 푈/휕푑 between Si and O atoms can be obtained. As shown in Figure 8-

184

7b, because the effective potential is highly anharmonic, the spring constant 푘SiO increases as bond length decreases. Thus, the conditions that make the Si-O bond longer – such as heating, or tensile stress – will shift the Si-O-Si vibrational band to lower frequencies.

To quantify the relation between 푑̅Si-O and the Si-O-Si,as position, normal mode analysis of an one-dimensional triatomic Si-O-Si model was performed. For different bond lengths, and thereby different force constants, the stiffness matrix was solved to obtain the eigenvalues and eigenvectors. For the simulated average bond length at 300K, the Si-O-Si,as peak is predicted to be at 1254 cm-1. Considering the oversimplification in this analytical calculation, the small discrepancy in the absolute peak position from MD simulation results is reasonable. Figure 8-7c compares the Si-O-Si,as peak shift as a function of 푑̅Si-O determined from MD simulations and analytical calculations. The excellent agreement between these two approaches confirms that the small change (less than 0.01 Å) in 푑̅Si-O can indeed induce a considerable shift in Si-O-Si,as peak position. This is because the potential associated with the Si-O-Si,as mode is highly anharmonic

(Figure 8-7a).

Implication for chemically-strengthened and thermally-tempered SLS glass

The physical insight obtained from MD simulations for pure silica glass can be extended to understand or explain the data of SLS glass shown in Figure 8-1. If the effect of modifier ions on SR-IR spectral features is assumed to be similar to the temperature and stress effects discussed in Section 3.4, then the Si-O-Si,as peak position of silicate glass can be correlated with 푑̅Si-O of the

Si-O-Si network. In the literature, the Si-O-Si,as peak position of sodium-silicate glasses (20%Na2O

-1 277,278,309 80%SiO2) is about 50 cm lower than pure silica glass. Experimental analyses of these glasses with extended x-ray absorption fine structure (EXAFS)256 and neutron scattering4,5 have

185 shown that the Si-O bond length is ~0.01 Å longer in sodium-silicate glasses (20%Na2O 80%SiO2) compared to pure silica glass. Then, using these literature values, the bond length dependence of

-1 the Si-O-Si,as peak position is predicted to be ~50 cm per a 0.01 Å change, which is in a reasonable agreement with the trend observed in Figure 8-7c (46 cm-1 / 0.01 Å). Better comparisons can be obtained if peak position as well as peak shape of 푘(휔) in the Si-O-Si vibrational region are obtained properly from spectra analysis.

In the thermal-tempering process which relies on the difference in cooling rate of the surface versus bulk regions, a compressive stress is developed in the faster-cooled surface region while the bulk region is shrinking at a much slower cooling rate. Thus, this situation is similar to the application of mechanical stress to the surface region by the bulk. Then, the average bond length in the compressed region could be smaller than the stress-free region, which can be manifested in the blue shift of the Si-O-Si,as band (as shown in Figure 8-1). This is consistent with the simulation result obtained with the pure silica network (Figure 8-4a). The magnitude of shift is much larger for the SLS glass than the silica case, which is also consistent with the trend observed in Figure 8-

186 In the case of chemically-strengthened SLS glass, the compressive stress is generated internally by the ion-exchange of small Na+ ions with larger K+ and Ag+ ions. The large modifier ions (M+) occupying the sites that were originally created for smaller Na+ ions can induce an increase in the bond length of the nearest surrounding Si-O bonds. This increase in dSi-O would decrease gradually as the distance from the exchanged ions increases. Then, this may induce a net increase in the average Si-O bond length, which could be manifested as a red-shift in the Si-O-Si,as band. In other words, the effect of the internal compressive stress induced by ion exchange on the the Si-O-Si,as band is not the same as the effect by the external compression. Based on the simulation result shown in Figure 8-7c, it can be deduced that the exchange of Na+ with a larger ion induces a slight increase in the Si-O bond length of the silicate network surrounding the exchanged modifier ion. This is consistent with the network dilation upon exchange of smaller modifier ions in the glass with larger ions.247,312 To test this hypothesis for the multicomponent silicate glass, force fields that can handle polarizability of modifier ions are necessary. The BKS force field cannot accurately reproduce the subtle change in the Si-O bond length upon addition of modifier ions.313 Such simulations will be possible in the future when more advanced force fields are developed.

The tensile stress data of the SLS glass in Figure 8-1 was obtained at the convex surface under three-point bending. In this case, the Si-O bond length might increase along the tensile stress direction which is parallel to the surface. However, the stress along the surface normal direction is compressive. This argument is further supported by the increase in the indentation modulus and hardness of the convex surface on the three-point bending rig compared to the pristine surface (see

Figure 8-S5 in Supporting Information). This could mean that the net change in the bond length distribution inside the convex surface of the plate under three-point bending is shortening, rather than lengthening expected in a simple one-dimensional analogue. This can explain the tensile stress data of SLS glass in Figure 8-1. In other words, the Si-O-Si,as band blue shift of the convex surface of the glass under three-point bending cannot be attributed to the tensile stress along the surface

187 tangential direction; it is governed by the compressive stress along the surface normal direction. A similar argument could be applied to the tensile data of the silica glass in Figure 8-1 which were obtained at the convex surface of the fiber under two-point bending.17,268

The correlation between density and Si-O-Si,as peak position obtained from MD simulations for low temperature thermal expansion and mechanical stress at room temperature may not be applied to the glasses with different fictive temperatures. For the thermally-grown silica on a silicon substrate, the density increases as the fictive temperature increases from 1000 oC and 1500 oC; this is known as an anomalous glass behavior. In this region, the Si-O-Si,as position decreases as the fictive temperature increases.1,309 This is opposite to the trend predicted for the low temperature thermal expansion (Figure 8-6d). Similarly, the fictive temperature effect of the SLS glass cannot be explained with the trend shown in Figure 8-6d. The SLS glass with a higher fictive temperature

44,45 shows a lower density but a higher Si-O-Si,as peak position.

The discrepancy in the Si-O-Si,as peak position versus density relation between the fictive temperature effects19,39,44,45 and the low temperature thermal expansion effect (Figure 8-6d) implies that the density alone cannot be used to justify or explain the dsi-o bond length change. It should be noted that fictive temperature can alter not only the bond length, but also the Si-O-Si connectivity.

For example, MD studies of silica showed that the density can be higher due to the smaller ring size distribution with a slightly longer bond length.301 Also, MD simulations of SLS glass found

314 that the density can be lower while the dSi-O bond length is shorter.

Conclusion

MD simulations were employed to elucidate the structural origin for the Si-O-Si asymmetric peak shift in SR-IR spectra of a silica glass upon heating or applying mechanical stress. The peak shift is correlated with the asymmetric change in the bond length distribution of

188 Si-O due to thermal excitation or mechanical stress. Other bond parameters factors such as the O-

Si-O tetrahedral bond angle and the Si-O-Si bridge bond angle as well as the density change cannot explain both temperature- and stress-induced changes of the peak position. The correlation found from the silica glass could explain the experimental trends observed for the soda lime silica glass upon ion exchange, bending, and thermal tempering.

Supporting information

SR-IR analysis of a SLS float glass (microscope slide glass) under tensile stresses and a SLS glass (5 mm thick glass) with surface compressive stress prepared by thermal tempering.

Figure 8-S1. SR-IR spectra of a pristine SLS glass and a convex surface of slide glass under three- point bending. The tensile stress of the convex surface was estimated using the beam bending theory from the deformation, the sample dimension, and the modulus of the slide glass. The lower intensity for sample under tensile stress was due to the light scattering from the slightly curved surface.

189

Figure 8-S2. SR-IR spectra of a tempered glass (quenched in air) and the same type of glass that has been annealed (slow cool in air).

Potentials used in MD simulation

The BKS potential is efficient for generating silica glasses with properties close to experiments. Since the BKS potential was used to prepare the glass network in this work, coordination numbers for Si and O should be low in the constructed silica model. The concentration of defects was defined as the number fraction of Si (O) atoms that do not have exactly four (two) neighboring O (Si) atoms within a cutoff distance of 2.2 Å. There were less than 0.1% defects in the resulting glass samples after cooling to 300 K (Figure 8-S3a). Then the thermal and mechanical properties of silica glasses were tested by simulating with the FFSiOH potential. The density of the silica model as a function of temperature for BKS and FFSiOH is shown in Figure 8-S3b and 8-

S3c. The density anomaly of liquid silica was observed in simulations with the BKS potential

(Figure 8-S3b), although it occurred at a temperature significantly higher than the experimental value (1820K).315 From the density-temperature curve, the volumetric thermal expansion coefficient at constant pressure p was obtained (Figure 8-S3d), which was close to the

-6 -1 -6 -1 experimental value (1.6×10 K ). The FFSiOH potential yields p = 11.8 ×10 K at 300K. This

190 overestimation of thermal expansion coefficient was also observed when other softer force fields, such as those proposed by Takada et al316 and Demicralp et al.317, were used. Although the BKS potential gives an excellent prediction of v, the BKS potential is not necessarily more accurate than the FFSiOH potentials. Since p decreases significantly when cooling rate decreases by several orders of magnitude, the BKS potential is expected to underestimate the p at experimental cooling rates, whereas the p from FFSiOH potential might become closer to the experimental value. Note

318 that for α-quartz, the p provided by both potentials agree with experiments. For mechanical properties, the Young’s modulus and Poisson’s ratio at 300K were calculated (Figure 8-S3e and 8-

S3f). The simulated mechanical properties of silica glass were close to experimental values that are summarized in Table 8-S1.

191 Table 8-S1. Physical properties of fused quartz at 300 K. These values are for the fused quartz substrate from Technical Glass Products used in the experimental part of this study Elastic Modulus Poisson Ratio Density Strain Point Thermal expansion 72 GPa 0.17 2.2 103 kg/m3 1120 C 0.55 ppm/C

Correlation between 풌(흎) with bond parameters

The trends observed in Figure 8-S4 are essentially the same as the one shown in Figure 8-

6 of the main text; the correlation with the peak position is much better for 푑̅Si-O, compared to 휃̅Si-

O-Si, ψ̅O-Si-O, and . It is interesting to note that, in the case of stress-induced changes (blue lines in the 휃̅Si-O-Si and ψ̅O-Si-O panels), the data appear to deviate from the linear fit. This might be due to the anisotropic nature of the strain. It is also noted that the correlation between peak position and

192

푑̅Si-O does not change noticeably with two different ways of representing the peak position  weighted mean position ̅ (Figure 8-6a) vs. most probably position 푚푝 (Figure 8-S4a)  in the

푘() spectrum; however, it shows some difference when the reflectance spectrum is used (Figure

8-S4b vs 8-S4c). This is believed to be due to artifacts caused by convolution of the real part,

푛(), of refractive index in the reflectance spectrum.

Nanoindentation analysis of the convex surface of a SLS glass under three-point bending

Nanoindentation analysis data of the SLS glass with or without applied stress are summarized in Figure 8-S5. Nanoindentation tests were carried out with a load control mode with a maximum force of 5mN. An increase in the modulus and hardness was observed for the convex surface of a SLS glass under tensile stress by three-point bending.

Figure 8-S5. (a) Elastic modulus and (b) hardness of the SLS glass with or without applied tensile stress. The p-values shown in the figure are from student’s t-test with N = 30 data set.

193 Chapter 9

Vibrational Sum Frequency Generation (SFG) Spectroscopy Study of Hydrous Species in Soda Lime Silica (SLS) Float Glass

Reproduced with permission from ACS : Luo, J.; Banerjee, J.; Pantano, C. G.; Kim, S. H. Vibrational Sum Frequency Generation Spectroscopy Study of Hydrous Species in Soda Lime Silica Float Glass. Langmuir 2016, 32 (24), 6035–6045.

Overview

It is generally accepted that the mechanical properties of soda lime silica (SLS) glass can be affected by the interaction between sodium ions and hydrous species (silanol groups and water molecules) in its surface region. While the amount of these hydrous species can be estimated from hydrogen profile and infrared spectroscopy, their chemical environment in the glass network is still not well understood. This work employed vibrational sum frequency generation (SFG) spectroscopy to investigate the chemical environment of hydrous species in the surface region of

SLS float glass. SLS float glass shows sharp peaks in the OH stretching vibration region in SFG spectra, while the OH stretch peaks of glasses that do not have leachable sodium ions and the OH peaks of water molecules in condensed phase are normally broad due to fast hydrogen bonding dynamics. The hydrous species responsible for the sharp SFG peaks for the SLS float glass were found to be thermodynamically more stable than physisorbed water molecules, did not exchange with D2O, and were associated with the sodium concentration gradient in the dealkalized subsurface region. These results suggested that they reside in static solvation shells with relatively slow hydrogen bonding dynamics, compared to physisorbed water layers on top of the glass surface. A putative radial distribution of the hydrous species within the SLS glass network was estimated based on the OH SFG spectral features, which could be compared with theoretical distributions calculated from computational simulations.

194 Introduction

Understanding the surface structure of soda lime silica (SLS) glass is of great importance to improve its chemical and mechanical properties, and to expand its applications beyond simple windows and bottles into high-tech industries such as electronic displays, photovoltaics, and so on.39,122,319,320 The surface properties of SLS glass are highly dependent on the manufacturing process, i.e., blowing, molding, polishing, or float processes.321 In the SLS float process, the SLS molten liquid is poured onto a molten tin bath; since the molten SLS and tin liquids are not miscible and tin has higher density than SLS, the SLS liquid floats and vitrifies on the liquid tin bath.322 This allows production of flat glass panels with a surface finish (roughness and flatness) as smooth as the molten liquid. In order to improve the durability of the produced glass panel, the commercial

122 float glass process exposes the glass to SO2 gas before the glass is fully cooled. During this exposure, sodium ions (Na+) associated with non-bridging oxygen (NBO; Si-O-) in SLS glass react

+ with SO2 and O2 and precipitate as Na2SO4 at the glass surface; the loss of Na ion is compensated

+ by proton (H ) from water or H2 in the environment which reacts with NBO, forming the silanol

+ 323 (Si-OH) group in the Na -leached site. During this SO2 dealkalization process and post- manufacturing steps (such as rinsing with water or storage in humid air), water molecules can alter the surface chemistry of SLS glass, producing concentration gradients of Na+ and hydrous species in the subsurface region.186 Here, the term “hydrous species” means both silanol and molecular water; the combination of these species could be stoichiometrically equivalent to hydronium ions

+ + 8,33 (H3O ) replacing sodium ions (Na ). The thickness of the dealkalization region varies with process conditions; it is typically about 60 – 100 nm.97,123 Thus, the hydrogen bonding interactions between silanol groups and water molecules as well as bridging oxygen (BO; Si-O-Si) groups become far more complicated in SLS float glass than pure silica and other silicate glasses without leachable ions.42,97,250 Figure 9-1 schematically illustrates the depth profiles of hydrogen and

195 modifier ions (Na+, Ca2+, Mg2+) in the surface region of SLS float glass based on the literature97,186 and the probe depth of SFG and ATR-IR on the hydrous species.52,96,324 The exact nature of these hydrous species is not well understood yet and remains as one of the grand challenges in fundamental surface science of silicate glasses.33

The formation and chemical environment of hydrous species in the surface region of glass have been investigated with theoretical calculations. Molecular dynamics (MD) simulations suggested that surface defect sites, including NBO, three-coordinated silicon, and two membered rings are prone to hydroxylation process when glass is exposed to water.78,325 Garofalini and colleagues have studied the interaction between silica surface and water using MD simulations with

196 a dissociative water potential. The results indicated that the water molecules can penetrate the silica network up to 0.7 nm during their simulation time frame, forming silanol groups and hydronium ions.326 This process can be enhanced in the presence of surface stress and excessive amount of water molecules.327,328 The bond angle distribution of Si-O-Si is also modified with the formation of new Si-OH on silica glass surface.329 Compared with silica glass, the chemical environment of hydrous species could be different in SLS glass since more NBO sites and Si-OH sites are available in the SLS surface region. Cormack and coworkers reported that sodium-rich and sodium-deficient sites can co-exist in MD simulations of silicate glass systems.313,330,331 When sodium ions are leached out of the surface region of SLS glass and hydrous species are introduced, it is likely that these hydrous species in the SLS glass network could be very different from those found in condensed liquid or ice phases or physisorbed states on solid surfaces.

It is well known that the OH stretching vibrational peak is very sensitive to the hydrogen bonding interactions. Thus, vibrational spectroscopy has been widely used to investigate the distribution of various hydrous species or hydrogen bonding interactions in the glass.42,95–97,124 For hydrous species in the surface region, attenuated total reflectance infrared (ATR-IR) spectroscopy is most extensively used.97,332 Since the information depth of ATR-IR for SLS glass is approximately one micron in the OH stretch vibration region (Figure 9-1) and its signal intensity is linearly proportional to the total concentration,52,96 it is suitable for quantification of the subsurface

OH and H2O species in glass. The overtone or combination bands in the near-IR region could distinguish molecular water and silanol groups;10,332 but subtle differences in hydrogen bonding interactions of water molecules and silanol groups in the glass matrix are difficult to interpret in near-IR analysis. The fundamental peaks in the mid-IR range are more sensitive to small changes in hydrogen bonding interactions; but the spectral overlap of various species makes deconvolution and interpretation of the data difficult.249 One of the reasons for this difficulty is that ATR-IR detects all hydrous species within the information depth (~1 µm), so the structural information of

197 the dealkalization region (within ~100 nm) is often masked by the spectral features of all hydrous species.

This difficulty could be circumvented by employing vibrational sum frequency generation

(SFG) spectroscopy.33 When a dielectric medium is irradiated with light, the medium is polarized in response to the electric field of the light (퐸⃗ ). The polarization (푃̃) of the medium can be expressed as follows:333

(1) (2) 2 (3) 3 푃̃ = 휀0(휒 퐸⃗ + 휒 퐸⃗ + 휒 퐸⃗ + ⋯ ) (1)

(1) (2) where 휀0 is the permittivity in the vacuum, 휒 is the first-order or linear susceptibility, 휒 and

휒(3) are the second and third-order nonlinear susceptibilities, respectively. Note that 휒(2) is about eight orders of magnitude smaller than 휒(1), and 휒(3) is even much smaller than 휒(2).175 When the glass is exposed to a relative weak 퐸⃗ field of light (i.e., the mid-IR beam in typical IR spectroscopy), only the contribution from 휒(1) in equation (1) is observed (Figure 9-1). The nonlinear responses from 휒(2) and 휒(3) terms could become significant and measurable only when high power pulsed laser beams are used. In the case of SFG spectroscopy, the ps-laser pulses with visible (휔푉퐼푆) and infrared frequencies (휔퐼푅) are spatially and temporally overlapped at the sample surface and the polarization response at the sum frequency (휔푉퐼푆 + 휔퐼푅) is detected. Since the SFG process requires phase matching of three photons with different wavelengths (휔푉퐼푆, 휔퐼푅, and 휔푉퐼푆

+ 휔퐼푅 ), it can be generated only within a coherence length defined by the phase mismatch condition. When the SFG experiment is carried out in a reflection geometry on an optically flat surface, then the SFG coherence length is less than 100 nm from the surface (Figure 9-1).324

One unique feature of SFG is that the second harmonic response of (2) vanishes in a random or centrosymmetric medium.333 Since the glass network is amorphous, it cannot meet the noncentrosymmetry requirement of (2). Thus, there will be no SFG signal from the bulk. However, the surface of glass breaks the randomness of two bulk phases (glass and surrounding gas). Thus,

198 the glass surface naturally provides the noncentrosymmetry, allowing detection of molecules at the surface.33,164 The third harmonic response of (3) in equation (1) does not require noncentrosymmetry; but when the molecules are under a strong electric field (퐸⃗ 퐷퐶 ), then the

(3) 휒 퐸⃗ 퐷퐶 term becomes non-centrosymmetric (because 퐸⃗ 퐷퐶 has polarity) and can be detected in

SFG.95,166 Such local electric field could originate from the subsurface concentration gradient of

Na+ ions in the glass.95 Therefore, both hydrous species at the glass/air interface (휒(2) response)

(3) and in the subsurface (휒 퐸⃗ 퐷퐶 response) can be probed using SFG spectroscopy (Figure 9-1). The

overall SFG intensity (퐼휔푉퐼푆+휔퐼푅) of the glass material could be expressed as follows:

2 ̃ (2) (3) ⃗ 퐼휔푉퐼푆+휔퐼푅 ∝ 푃푁퐿 ∝ (휒 + 휒 퐸퐷퐶) 퐼휔푉퐼푆 퐼휔퐼푅 (2)

̃ where 푃푁퐿 is the nonlinear part of the polarization from equation (1), 퐼휔푉퐼푆 and 퐼휔퐼푅 are the intensity of input visible and IR laser beams, respectively.

In this paper, the subsurface hydrous species in glass were analyzed with in-situ SFG measurements during heating to obtain thermodynamic information relevant to the interfacial structures. For glasses without leachable alkali ions, the main hydrous species detected with SFG are water molecules adsorbed on the glass surface. In the case of SLS float glass, the spectral features in the OH stretch region of the SFG spectrum are very different – multiple sharp peaks are observed which have not been reported in vibrational spectroscopy of other glass systems or condensed water phases.33,181 The hydrous species responsible for these sharp SFG peaks of the

SLS float glass are found to have higher stabilities than simple physisorbed molecules and do not readily exchange with the water molecules in the gas phase. Also, they disappear when the subsurface Na+ concentration gradient is suppressed by annealing at a temperature higher than glass transition temperature (Tg). These findings suggest that the multiple OH SFG features with narrow peak widths of SLS float glass originate from the structural differences of hydrous species within

+ the subsurface region where the Na ion profile is modified during the SO2 dealkalization process.

199 Using the known empirical relationship of the OH peak position and the hydrogen bond distance, a putative radial distribution of the OH···O distance for subsurface hydrous species in the SLS float glass is proposed, which could provide an experimental basis for verification of computational predictions.

Experimental details

The SLS float glasses were provided by Asahi Glass Co (Tokyo, Japan). The bulk composition of the glass (weight%) was measured by X-ray fluorescence and found to be 72.3%

SiO2, 13.3% Na2O, 7.7% CaO, 1.9% Al2O3, 4.4% MgO, 0.3% K2O, and 0.1% Fe2O3. Two different thicknesses (0.7mm and 1mm) of the SLS float glass were used in this study. In the float glass manufacturing, the glass with different thickness underwent different degree of stretching and

322 SO2 treatment time. The glass transition temperature of the SLS float glass used in this study was around 550C. During the float process, two sides of the glass were produced in different chemical conditions; one side was in contact with air and the other with the molten tin. Since the tin-side contains a small amount of tin which alters the glass structure, this study focused on the air-side only. For a comparative study, fused quartz and borosilicate glasses were also analyzed. Fused quartz (SiO2) slides with a thickness of 1 mm were obtained from Technical Glass Products.

Borosilicate (BOROFLOAT® 33; BF33) glass slides with a thickness of 1 mm were obtained from

Schott Glass. The mass composition of BF33 was 81% SiO2, 13%B2O3, 2%Al2O3 and 4%

Na2O/K2O. All glasses were cut with a diamond cutter to the size of 1 cm  1 cm. All samples were pre-cleaned by rinsing with MilliQ water and pure ethanol, then UV-ozone cleaning for 20 minutes.53

The surface composition of soda lime float glass in this paper was analyzed with x-ray photoelectron spectroscopy (XPS). A Kratos Analytical Axis Ultra spectrometer (Chestnut, NY)

200 fitted with a monochromatic Al Kα (1486.6 eV) source and a low-energy electron beam charge- neutralizing flood gun was used for elemental analysis of the top 10 nm region of the SLS float glass. In the survey spectra (80 eV pass energy and 0.3 eV step size), O 1s, Na KLL, Ca 2p,

Mg KLL, K 2p, C 1s, Si 2p and Al 2p peaks were used to calculate the atomic concentrations. The relative sensitivity factors (RSF) used for quantification were calibrated using a vacuum-fractured SLS float glass surface and the bulk composition acquired via inductively coupled plasma atomic emission spectroscopy (ICP-AES).13 In addition, O 1s and C 1s high-resolution, narrow energy peaks (20 eV pass energy and 0.1 eV step size) were captured for peak fitting. The binding energies of all elemental peaks were corrected with the adventitious alkyl peak (high-resolution) adjusted to 285.0 eV. The true glass composition was determined by a mathematical correction for the attenuation caused by adventitious hydrocarbon contamination on the glass surface.13,73 The total amount of hydrous species in the glass was analyzed with ATR-IR spectroscopy with diamond and Ge ATR crystals.96 The IR beam incident angles for Ge ATR (VariGATR; Harrick Scientific Products,

Inc. Pleasantville, NY) and diamond ATR (MVP-Pro; Harrick Scientific Prodcusts, Inc.,

Pleasantville, NY) were 60o and 45o, respectively. For ATR-IR analysis on the elevated temperatures, diamond ATR was used. Specular reflectance infrared (SR-IR) spectroscopy was carried out with a Bruker Hyperion 3000 Microscope (Bruker, Co.) with a 15x objective lens. A gold mirror was used as the reference background.

The detailed set-up of the vibrational SFG spectroscopy system was described elsewhere.72,334 In brief, frequency-doubled laser pulses (532 nm) from a 27 ps Nd:YAG laser and tunable IR pulses generated from an optical parameter amplifier and generator system were used as 휔푉퐼푆 and 휔퐼푅 incident beams, respectively. The incident angle of IR and visible pulses were 56

201 and 60, respectively. These two laser pulses were overlapped spatially and temporally on the glass surface to generate SFG signal. The SFG signal was detected at the phase matching direction. The polarization combination used in this study was s for SFG signal, s for 532nm laser pulses, and p for IR laser pulses (ssp). The visible and IR pulse energies before reaching the sample was 174 J and 136-220 J, respectively. The IR wavenumber was calibrated with SFG spectra of DMSO/air interface in ssp polarization.335 For thermodynamic analysis of hydrous species, the SFG signal intensities were recorded as a function of temperature while the glass sample was heated at a 2

K/min rate () using a Linkam Scientific heater (model No. TS1500, UK) in a controlled vapor gas environment, as shown in Figure 9-2. Similar set-up was applied in the ATR-IR analysis.

Results and discussion

Figure 9-3 compares the OH stretch region of the SFG spectra of 0.7 mm and 1 mm thick

SLS float glasses with those of fused quartz and BF33. For each sample, spectra were taken in air

(with 30% relative humidity) (i) at room temperature, (ii) at 200C, and (iii) after cooling back to room temperature for the same spot. These measurements were repeated 3 or more times at different

202 locations; the spectra shown in Figure 9-3 are the representative ones. In order to rule out the possibility of interferences from the signals reflected from the back surface, especially in the case of multiple peaks of the SLS float glass samples, SFG analysis was performed with a refractive index matching liquid (for example, CCl4) in the back surface; no differences were observed (see

Figure 9-S1 in the Supporting Information), indicating that the multiple peaks are not artifacts due to interferences of signals from two surfaces. The 0.7mm thick SLS float glass (Figure 9-3a) shows three sharp peaks at ~3272 cm-1, ~3544 cm-1, and ~3824 cm-1, while the 1mm thick float glass

(Figure 9-3b) shows five peaks at ~3180 cm-1, ~3392 cm-1, ~3552 cm-1, ~3728 cm-1, and ~3920 cm-

In contrast, the glasses without leachable sodium ions show broad features in SFG. In the

SFG spectrum of fused quartz (Figure 9-3c), a broad peak centered around 3300 cm-1 (spanning from 3000 to 3600 cm-1) and a sharp peak at 3760 cm-1 are observed, which is consistent with the previous report.33 The broad peak below 3600 cm-1 is generally assigned to the hydrogen-bonded

OH groups and the sharp peak at 3760 cm-1 is assigned to the free OH groups without hydrogen- bonding interactions.100,104,249,250 The BF33 glass also gives a broad SFG peak centered around 3500 cm-1 (spanning from 3000 to 3700 cm-1) and a relatively sharp peak centered around 3850 cm-1, as shown in Figure 9-3d. The broadness of the hydrogen-bonded OH peaks in the SFG spectra of fused quartz and BF33 is related to the dynamics of hydrogen bonding interactions. In the condensed liquid phase or adsorbate layers, the OH groups of water molecule undergoes very fast formation and dissociation processes of hydrogen bonds with surrounding water molecules, resulting in an ultra-short transient lifetime of the hydrogen bonds with solvating molecules.98,111

Thus, the vibrational relaxation time of the excited OH group is extremely short, which leads to the homogeneous line-broadening of the vibrational absorption band.105,117,336

203

Applying this relationship between the OH peak width and the hydrogen bonding dynamics, one could interpret the sharpness of the OH SFG peak (Fig. 9-3a and 9-3b), compared to the fused quartz and BF33 spectra (Fig. 9-3c and 9-3d), as a longer life-time or slower dynamics of hydrogen bonds of hydrous species in the SLS float glass. In other words, the hydrous species detected with SFG for SLS float glass are in a more static (or less dynamic) solvation shell. Such solvation shell could be formed by the solid silicate networks surrounding silanol groups or water molecules.337 Then, the distinct positions of the multiple peaks could indicate subtle differences in the hydrogen bond distance distribution among the species involved in hydrogen bonding interactions (for example, NBO, BO, Si-OH, and H2O) within the SLS float glass structure.

204

In Figure 9-3, it is noted that the SFG intensities of all glasses decrease to zero at 200 C.

As the samples are cooled back down to room temperature, all peaks fully recover to the same peak positions and intensities, implying the observed changes are reversible. Since 200C is well below the Tg of the glasses studied in this work, the silicate network does not change during the heating to 200 oC and cooling.321 Also, 200 C is not high enough for total dehydroxylation of silanol groups.338 Thus, the observed changes upon heating in Figure 9-3 must be due to physical changes in the state of molecular water at or within the glass.

(3) For fused quartz and BF33 which would not have the 휒 퐸⃗ 퐷퐶 contribution since there is no concentration gradient of leachable sodium ions, this reversible process must be solely due to desorption and adsorption of water molecules at the air/glass interface; i.e. 휒(2) responses (Figure

9-1).339 The broadness of the SFG features for these glasses is also consistent with the physisorbed water molecules which are subject to fast relaxation dynamics of the OH excitation through hydrogen bonding interactions with surrounding water molecules.116

(3) In the case of SLS float glass, 휒 퐸⃗ 퐷퐶 could contribute to SFG spectra because the

+ concentration gradient of Na in the dealkalization region could induce substantial 퐸⃗ 퐷퐶 (Figure 9-

340 1). In the dealkalization region, polar water molecules could be aligned if strong 퐸⃗ 퐷퐶 field exists.

In the thermal poling experiment, we have observed that the OH signals of water molecules have opposite phases for the anode-poled and cathode-poled surfaces.95 The phase of OH signals can be

341–343 related to the direction of OH bonds with respect to the surface. The subsurface 퐸⃗ 퐷퐶 field could also stabilize certain binding sites where the dipoles of water molecules align favorably with the electric field. As the temperature of the sample is increased, thermal energy would counteract this 퐸⃗ 퐷퐶-induced ordering and desorb water from the 퐸⃗ 퐷퐶-stabilized bindng site. If thermal energy of the system is high enough, then the degree of 퐸⃗ 퐷퐶-induced alignment of water molecules in the dealkalization region would decrease and so does the SFG intensity. This could cause the

205 disappearance of the sharp OH SFG peaks for the SLS float glass upon heating and reappearance upon cooling.

If these interpretations are correct, then we expect different thermodynamic properties for the water molecules adsorbed on fused quartz and BF33 glass surfaces and the hydrous species

(water molecules interacting with NBO, BO, and Si-OH) in the silicate network of SLS float glass.

In order to test this hypothesis, we measured the SFG signal intensity of each peak as a function of temperature (Figure 9-2b). This method can be called temperature-programmed SFG (TP-SFG) in analogy with temperature-programmed desorption (TPD) in typical surface science studies of adsorbed molecules.12,344,345 In a TPD experiment, the concentration of surface species is calculated by integrating the intensity of the desorbing species.345 In TP-SFG experiments, the SFG signal intensity is related to the concentration of surface species.165,333 In both TPD and TP-SFG, the rate of desorption from unit surface area could be modeled by the Polarnyi-Wagner equation:345

푑휎 퐸 − = 푣 휎푛퐸푥푝(− 푎 ) (3) 푑푡 푛 푅푇 where n is order of desorption process, 푣푛 is rate constant, 휎 is surface coverage, T is temperature

(K), R is gas constant (8.14 J/molK), and 퐸푎 (kJ/mol) is activation energy. The physical meaning of 퐸푎 will depend on what chemical species are actually measured in SFG. If SFG probes physisorbed water molecules on the glass surface, then 퐸푎 will be equivalent to the heat of desorption. In this experiment, the sample temperature (T) was ramped at a constant rate: T = T0 +

βt as shown in Figure 9-2b, where T0 is room the initial temperature (298 K), t is time, and β is 2

K/min. Thus, the time term in the left-hand side of equation (3) can be replaced with temperature: t = (T - T0)/β. Since SFG is insensitive to random molecules, it is insensitive to the zeroth-order bulk process. For simplicity, the first order relationship is assumed in equation (3). Then, equation

(3) can be converted to the following form:

푑휎(푇) 푣 퐸 − = 1 퐸푥푝 (− 푎 ) 푑푇 (4) 휎(푇) 훽 푅푇

206 Integrating equation (4), the analytical expression of desorption process could be obtained as follows:

휎 (푇) 푣 푇 퐸 − ln ( 푓 ) = 1 ∫ 퐸푥푝 (− 푎 ) 푑푇 (5) 휎0 훽 푇0 푅푇 where 휎푓(푇) and 휎0 are the concentration of species at T and T0, respectively.

The SFG intensity is proportional to the square of the number of molecules that are SFG active (as shown in equation (2)). It should be noted that the molecular orientation of OH group and 퐸⃗ 퐷퐶-induced alignment of OH can also be changed during the heating process. If the heating to 200 oC changes the orientation of OH groups drastically, it can be assumed that the square root of the SFG signal is proportional to the amount or concentration of the ordered species being probed:

( ) 푚√퐼휔푉퐼푆+휔퐼푅 푇 = 휎(푇) (6) where m is a proportionality constant relating the square root of the SFG signal to the analyte concentration, which depends on the incident and detection angles, refractive indices, and detector

335 sensitivity. Substituting equation (6) into the 휎푓(푇) term in equation (5), the following equation is obtained:

√퐼휔 +휔 (푇) 푉퐼푆 퐼푅 푣1 푇 퐸푎 = 퐸푥푝 (− ∫푇 퐸푥푝 (− ) 푑푇) (7) 퐼 (푇 ) 훽 0 푅푇 √ 휔푉퐼푆+휔퐼푅 0

Similar to TPD, equation (7) can be used to fit the experimental TP-SFG data to obtain Ea for hydrous species responsible for each OH SFG peak. In TPD, the chemical species desorbing from the surface are monitored; thus, if the same molecular species desorb from two different sites with similar binding energies, they may not be distinguished easily. In contrast, TP-SFG allows direct monitoring of the same molecules in different chemical environments based on their peak positions. Thus, the energetics of water molecules at different binding sites can be easily

207 distinguished. The dissociation or desorption energy of each binding site can be determined through nonlinear fitting of the TP-SFG data using equation (7). Representative fitting results are presented in Figure 9-4.

The desorption energies of all hydrous species detected in SFG are summarized in Figure

9-5. For fused quartz (Figure 9-5c), the desorption energy of water molecules is 422 kJ/mol for the ~3300cm-1 peak and 452 kJ/mol for the 3700 cm-1 peak. In the case of BF33 (Figure 9-5d), the desorption energy for water molecules detected at ~3500 cm-1 is 442 kJ/mol. These values are close to the values of heat of vaporization of water (40.7 kJ/mol) and heat of sublimation of ice

(46.7 kJ/mol), validating the TP-SFG analysis method and confirming the assignment of the broad

OH SFG peaks to the physisorbed water molecules on the glass surface. The origin of 3800 cm-1 peak in BF33 is not clear at this point since its desorption energy (523 kJ/mol) is higher than the physisorbed water molecules. One possible origin of this peak is the OH groups of B-OH or Al-

- - OH sites or water molecules interacting with BO4 /B-OH or AlO4 /Al-OH in the glass network of

BF33.346–348

208 The desorption energies of hydrous species detected for SLS float glass appear to be different from those of the water physisorbed on the fused quartz surface. For 0.7 mm thick SLS float glass (Figure 9-5a), the activation energies of three species detected at 3272 cm-1, 3544 cm-1, and 3824 cm-1 are 623 kJ/mol, 734 kJ/mol, and 422 kJ/mol, respectively. For 1 mm thick SLS float glass (Figure 9-5b), the activation energy of five species detected at 3180 cm-1, 3392 cm-1,

3552 cm-1, 3728 cm-1, 3920 cm-1 are 553 kJ/mol, 652 kJ/mol, 512 kJ/mol, 432 kJ/mol, and

313 kJ/mol, respectively. Again, the discrepancy of these values from the values for adsorbed water molecules (Figures 9-5c and 9-5d) supports that the hydrous species detected with SFG for

SLS float glass are not water molecules adsorbed on the glass surface. Especially, the fact that the desorption energies for hydrogen-bonded species detected at <3600 cm-1 are higher than those of condensed water phases (40.7  46.7 kJ/mol) supports the interpretation that they are mostly water molecules at 퐸⃗ 퐷퐶-stabilized bindng sites in the dealkalization layer, not the physisorbed molecules on the glass surface.

209

For SLS float glasses, the peaks centered around 3728 cm-1 (1 mm thick) and 3824 cm-1

(0.7 mm thick) have the desorption energies similar to the 3700 cm-1 peak of the fused quartz.

Thus, they could be assigned to the free hydroxyl groups without hydrogen bonding interactions with the surrounding hydroxyl or BO groups. The peaks above 3800 cm-1 are rarely reported in the vibrational spectroscopy literature of OH species.100,103 These peaks at high wavenumbers could be related to the water molecules that interact with the charged sites in the glass network. Similar to the 3800 cm-1 peak of BF33, it could come from the water molecules that interact with the Ca2+ or

- 347,349 -1 -1 AlO4 /Al-OH sites in the glass network. The 3920 cm peak position is about 200 cm higher than typical stretching vibrations of free OH groups. Such blue shifts in the OH stretching vibration

210 peak position has been reported for the OH groups attached the atoms involved in strained chemical bonds or hyper-conjugation.112,113 The low desorption energies of the 3920 cm-1 (313 kJ/mol) might imply this is a metastable site. The exact nature of this band remains unknown and requires further investigation.

Additional proof for the assignment that the sharp OH peaks are mostly due to subsurface hydrous species in the SLS float glass can be found in D2O adsorption experiments. If the sharp

OH SFG features are due to H2O molecules adsorbed on top of the SLS float glass surface, then they should be shifted by a factor of 1.34 to the lower frequency region (dotted lines in Figure 9-

205 6a) upon replacing H2O with D2O. Figure 9-6a compares the SFG spectra of 1mm thick SLS

-1 -1 float glass equilibrated in H2O and D2O vapors. Note that the sharp peaks at 2850 cm , 2870 cm , and 2920 cm-1 are due to organic species adsorbed from the ambient gas.53,350 The difference of the

SFG spectra collected in D2O and H2O vapors show broad peaks in the OD stretch region (between

-1 -1 2400 cm 2800 cm ), which can be attributed to the physisorbed D2O molecules. If a factor of 1.34 is multiplied to the difference spectrum, the broad OD peak shape is quite comparable with the broad OH peak of the physisorbed water on fused quartz (Figure 9-3c). Once again, these results are consistent with the assignment that the sharp SFG peaks in the OH stretch region are not due to the physisorbed water molecules on the float glass surface; based on these findings, we conclude that they originate mainly from the subsurface hydrous species.

211

Figure 9-6. (a) SFG spectra of 1mm thick SLS float glass exposed to D2O vapor (ii) and comparison with the SFG spectra before (i) and after (ii) D2O exposure. The relative partial pressure of D2O in (ii) was 90 % in N2 and that of H2O in (i) and (iii) was 30 % in air. The bottom spectrum is the subtraction of (i) from (ii); the dotted line represents the sharp SFG spectral features obtained by shifting the OH features using theoretical OH/OD peak positions. (b) SFG spectra of 1mm thick o SLS float glass heated to 200 C and then cooled back to room temperature in D2O vapor; the spectra were collected in ambient conditions. (c) ATR-IR spectra of 1mm thick SLS float glass collected before heating to 200 oC, at 200 oC, and after cooling back from 200 oC to room temperature.

Another important question is whether or not the subsurface hydrous species are able to desorb to the gas phase upon heating to 200 oC. If they desorb to the gas phase at 200 oC (Figure 9-

3a and 9-3b), their reappearance upon cooling back to room temperature must be re-absorption of water molecules from the gas phase. In order to answer this question, the 1 mm thick SLS float

o glass was heated to 200 C in D2O vapor (relative humidity 90% in N2) and cooled back to room temperature in D2O vapor. As shown in Figure 9-6b, the SFG spectrum of the glass heated and cooled in D2O vapor shows no change in the sharp OH features and no growth of the sharp OD features in the 2400-2900 cm-1 region (shown as the dotted line in Figure 9-6a). Also, the molecular

H-O-H bending vibration region of the ATR-IR spectra of SLS float glass collected at room temperature and 200 oC (Figure 9-6c) shows no significant decrease in the total amount of molecular water in the ~1 m region of the surface upon heating to 200 oC. These results indicate that the water molecules desorbed or dissociated from the 퐸⃗ 퐷퐶-stabilized binding sites (or oriented

212 sites) upon heating to 200 oC remain trapped inside the glass (probably near their original sites). As stated earlier, the silicate network of SLS glass is not modified since 200C is far below Tg. The sodium profile does not change substantially at temperatures far below 400C.351 Thus, upon cooling, the water molecules trapped in the glass network must bind back to the original sites, recovering the original ordering induced by the sodium depletion in the subsurface region.

Figure 9-7. Effect of heating above Tg on the surface structure of SLS float glass. (a) SFG, (b) ATR- IR, and (c) SR-IR spectra collected before and after heat treatment at 600 oC. (d) Modifier ion (Na+, Ca2+, Mg2+) concentrations measured with XPS before and after heat treatment at 600 oC. (e, f) Changes in O1s XPS peak shape before (e) and after (f) heat treatment at 600 oC.

To confirm the correlation between the multiple SFG peaks with narrow widths and the sodium concentration gradient caused by the SO2 treatment in the subsurface region, the 1 mm thick SLS float glass was heated above Tg. The annealing procedure was as follows: the sample was heated at 3C/min from room temperature to 600C, held at 600C for 30 minutes and cooled down to room temperature at a 3C/min cooling rate in ambient air. This heat treatment should be

213 sufficient to equilibrate the sodium ion concentration in the SO2-dealkalized region to the level close to the bulk concentration. Since the temperature was well below the softening point (~730 oC) of the SLS float glass, the glass surface remained optically flat for SFG and ATR-IR analyses.

As shown in Figure 9-7a, the heat-treated SLS float glass exhibits a broad SFG feature in the OH stretching region, which is drastically different from the pristine SLS float glass. In addition, the

SFG peak intensity decreases substantially after the heat treatment. Especially, the peaks at >3700 cm-1 are almost completely suppressed. Since the strained sites are expected to be relaxed through thermal annealing at a temperature above Tg, this result supports the argument assigning these peaks to hydrous species bonded to the strained sites in the subsurface glass network.

Figure 9-7b compares the ATR-IR spectra before and after the heat treatment. It is noted that the high wavenumber components are slightly reduced after heating above Tg. Again, this is consistent or similar to the substantial suppression of the SFG peak intensities in the >3500 cm-1 region (Figure 9-7a). Figure 9-7c compares the specular reflectance infrared (SR-IR) spectra of

1mm SLS glass before and after heating above Tg. For comparison, the SR-IR spectra of a fused quartz are shown in the Supporting Information (Figure 9-S3). After heating at 600 oC, the SLS float glass shows changes in the peak shape, while the fused quartz does not any change. This indicates that the silicate network in the subsurface region of SLS glass changes after this heat treatment.

In order to follow the chemical composition change in the top ~10 nm region of the glass,

XPS analysis was performed for pristine and 600 oC-treated SLS float glasses. The data from three random spots were averaged and corrections for surface carbonate and organic contaminants were made to obtain accurate concentrations of the constituent elements of the glass surface.73 The data shown in Figure 9-7d reveals a significant increase in the sodium content and a slight increase in the calcium content after the 600 oC treatment. The O1s high-resolution spectrum is sensitive to the distribution of NBO (528 eV) and BO (530 eV) species.352–354 The deconvolution of the O1s peaks

214 (Figures 9-7e and 9-7f) shows that the ratio of NBO/BO increases after heating at 600 oC, which is in agreement with the increased concentration of alkali and alkaline earth ions in the subsurface region. The results shown in Figure 9-7 confirm that the multiple sharp SFG peaks of SLS float glass are associated with the subsurface hydrous species in the SO2 dealkalization region. As the

Na+ ions are replenished by diffusion from the bulk during the heating treatment at 600 oC (above

(3) Tg), the contribution from the ordering induced by sodium depletion or 휒 퐸퐷퐶 to the overall SFG intensity becomes smaller (see Figure 9-1).

With confirmation of the detection of subsurface hydrous species in distinct hydrogen bonding environments in SLS float glass with SFG, one can attempt to convert the measured SFG spectral features to a radial distribution of hydrogen-bonded O-HO distances. This is possible since there is an empirical relationship between the OH stretch peak position (especially for the

-1 peaks at <3600 cm ) and the distance between two oxygen atoms (dO-HO) connected through hydrogen bonding. Figure 9-8a is the data regenerated from the Libowitzky’s work comparing the

O-H stretching vibration positions observed with 47 different minerals and the average O-HO distances in their crystal structures.101,102,355–357 Assuming the same relationship can be applied to

215 the silicate glass, the SFG spectra can be transformed to a putative radial distribution by converting the OH SFG peak positions in the x-axis to the O-HO distance using the empirical relationship shown in Figure 9-8a and plotting the square root of SFG intensities in the y-axis.

Figures 9-8b and 9-8c display the radial distribution plots obtained from the data shown in

Figures 9-3a and 9-3b. It is interesting to note that the shortest hydrogen bond distance for the subsurface hydrous species (O-HO) is about 0.268 nm, which is only ~0.01 nm longer than the average distribution between BO groups predicted from MD simulations of silica and silicate glasses.108,109,313,358Although SFG detects only the noncentrosymmetrically-ordered fraction of hydrous species in the dealkalization region (Figure 9-1), this could provide an unprecedented opportunity to measure the structural distribution of hydrous species in the SLS glass which can be compared with radial distributions calculated from theoretical simulations.108,109,313The variation between SLS glasses with different thicknesses (0.7 mm versus 1 mm) is speculated to originate from the variation in processing time for commercial float glasses of different thickness.322

Conclusion

The chemical environment of hydrous species (Si-OH and H2O) in the subsurface region of SLS float glass was investigated with SFG spectroscopy. The origin of multiple OH peaks of

SLS float glass observed in SFG must be related to the confined nature of water molecules in the dealkalized sites in the SLS glass network formed by the SO2 treatments during the float glass manufacturing process. The hydrous species of SLS float glass responsible to the multiple sharp

SFG peaks are thermodynamically more stable than physisorbed water molecules on the glass surface and are not readily exchanged with water molecules in the gas phase at temperatures below 200 oC. The non-linear optical selection rule of SFG allows detection of subtle differences in the hydrogen bonding interactions of these hydrous groups inside the dealkalized SLS glass

216 network. The structural information of hydrous species revealed by SFG could provide critical insights needed to understand chemical and mechanical properties of the SLS glass surface.

Supporting Information

Since the IR in the 2300-4000 cm-1 region can penetrate and reflect from the backside of

0.7mm and 1 mm thick of flat glass substrates,96 the interference from the backside reflection could be considered as an origin for multiple OH peaks. However, this possibility can be easily ruled out.

First, even if this happens, the back reflection can be spatially filtered out since the SFG beams from the front and back surfaces is spatially separated. Second, we have done a control experiment to eliminate or alter the back reflection. A refractive index matching liquid, CCl4, can be placed in the backside of the soda lime silica (SLS) float glass to eliminate the reflection of the backside reflection. Figure 9-S1 compares the SFG spectra collected for 1mm thick SLS glass with and without CCl4 on the back side; the multiple peak features of the 1mm thick SLS float glass do not change, which rules out the interference from the back reflection. This provides evidence that the

SFG signal recorded originate the top surface region of SLS float glass.

217

Figure 9-S1. SFG spectra of pristine 1mm SLS float glass with and without CCl4 on the back side.

Figure 9-S2 shows the anti-Stokes Raman signal when the detection signal was to the 532 nm reflection direction and the 532nm beam was filtered out with a filter (in this specific case, the

IR beam was turned off intentionally) and the monochromator was scanned in the range where SFG signal was expected (since the anti-Stokes Raman signal is at the same frequency as the SFG signal). This signal was much stronger than the SFG signal and highly directional. When the sample was tilted by one degree, then the signal intensity decreased to zero. This suggested that the origin of this anti-Stokes signal is not simple Raman scattering which would be emitted to all directions.

The same feature could be observed with a 100 fs 800nm pulses (data not shown). In normal SFG experiment with different incident angles for IR and 532 nm, this anti-Stokes signal is spatially filtered out since it is along the 532 nm reflection direction which is different from the direction of the SFG signal due to the phase matching condition. Note that the anti-Stokes signal is not sensitive to the heat treatment (Figure 9-S2), while the SFG signal of the SLS float glass is sensitive to the heat treatment (as shown in Figure 9-7a in the main text).

218

The existence of anti-stokes Raman signal from the bulk makes the phase-sensitive SFG detection challenging when visible and infrared pulses are in the collinear propagation configuration.359,360 This is because the anti-stokes Raman signal and the SFG signal are at the same frequency and cannot be spatially separated.

Figure 9-S3 shows the SR-IR spectra of a fused quartz before and after heating at 600 oC.

The data indicates that no change occurs in the silica network upon heating to 600 oC. In contrast, the silicate network in the subsurface region of SLS float glass shows discernable changes in the network vibrations (Figure 9-7c in the main text).

219

Figure 9-S3. SR-IR spectra of 1mm thick fused quartz before and after heating at 600 oC.

220

Appendix A

Wear behavior of SLS glass surfaces after leaching in acid solution

Leaching in acid treatment was performed on SLS float glass to investigate Na+ and H+ +

H2O exchange effect on the wear behavior in addition to the hydrothermal treatment discussed in

Chapter 3. Unlike hydrothermal treatment, network dissolution reactions are minimal for leaching treatment. As is expected, Na+ were leached out while hydrous species are introduced to the surface region of SLS glass. It is interesting to note that wear behavior can become drastically different when only 100 nm thick layer is modified through leaching. This confirms the hypothesis in this thesis that mobile Na+ in the surface region of SLS float glass could play a critical role in the shear- induced hydrolysis reactions.

Preparation of leached sample with ~100 nm layer leached layer

0.7 mm thick SLS float glasses from Asahi Co. were used in this study. The samples are cut to be 2.5 cm  2.5 cm. The samples were cleaned by rinsing with pure ethanol and water, and then placed in a UV-ozone chamber for 20mins.53 The cut samples were placed in a Teflon container with 50mL HCl solution (PH=1 at 25 C) at 90 C for 1 hour, 5 hours, 20 hours and 80 hours. Sample surfaces are not in contact with the Teflon during the leaching. Air sides of the leached samples were analyzed in this dissertation. For comparison with leached SLS glasses, as received SLS float glass is denoted as “pristine” and used as the reference sample.

This leaching protocol is very similar to Amma’s previous work.97 Then the thickness of leached layer can be roughly estimated from Amma’s TOF-SIMS analysis. To further confirm this estimation, XPS depth profile of samples leached for 5 hours and 20 hours have been carried out.

221 As is shown in Figure A-1, the thickness of leached layer is ~50 nm and 70 nm based on Na/Si ratio for 5 hours and 20 hours samples, which is in good agreement with the TOF-SIMS analysis.

Thus the thickness of leached layer can be as high as ~100 nm for 80 hours samples. In addition to the depletion of Na+, Ca2+ and Mg2+ are also partially depleted (<25 nm). Al content does not change much.

Figure A-1. XPS depth profile of leached SLS glass (a) 5 hours sample; (b) 20 hours sample.

Silicate network modification and introduction of hydrous species

IR analysis were performed on leached SLS glass surface to investigate the amount of hydrous species and modification of silicate network due to leaching treatment. As is shown in

Figure A-2a, the amount of hydrous species (Si-OH and H2O) increase as leaching time increases, which is consistent with the reported trends. Like hydrothermal treatment, blue shifts of Si-O-Si asymmetric stretch is observed for leached surface in Figure A-2b. For samples leached for 80 hours, a slight increase of 940 cm-1 component is observed, which can be attributed to the formation of Si-OH. This interpretation is also in agreement with the ATR-IR spectra in Figure

222 A-2a. This shift suggest that the overall distribution of Si-O bond length in the silicate network become “shorter” based on the interpretation of the MD simulation work in Chapter 8.

Figure A-2. (a) ATR-IR spectra collected with a Ge ATR crystal for pristine and leached SLS glass. (b) SR-IR spectra collected at 45 incidence angle for pristine and leached SLS glasses.

Modification of wear behavior of SLS glass with 100 nm thick surface layer

The cross sections of the wear tracks of pristine and leached SLS glasses produced at

90% RH conditions analyzed with optical profilometry is shown in Figure A-3. After 1 hour leaching treatment, the wear volume has already become larger than pristine SLS glass. For 80 hours leached sample, the wear volume becomes very similar to silica glass. Based on the XPS depth profile in Figure A-1 and TOF-SIMS profile in the literature, a ~100 nm thick dealkalized layer is expected to form on this 80 hours leached sample. In previous studies, leached layer is expected to have lower hardness than pristine SLS glass,361 which suggests that compressive stress might not be formed on leached surface. This significant increase of wear volume for leached surface confirms that leachable Na+ could play a significant role in shear-induced hydrolysis reactions at high RH conditions.

223

Figure A-3. Comparison of wear tracks formed after 400 sliding cycles with ~300 MPa Hertzian contact pressure at 90% relative humidity from pristine SLS glass, 1 hour leached sample and 82 hours leached samples.

224 Appendix B

Determining the relative abundance of SiOH/H2O in monolithic flat glass surfaces using ATR mid-IR and hydrogen depth profiles

Introduction

Hydrous species in silicate glass surfaces can have a significant impact on their chemical durability, mechanical and mechanochemical properties.31,67,187,223 For example, in soda lime silica

(SLS) glass, hydrous species exist in two forms: silanol groups (SiOH) and molecular water (H2O).

These hydrous species are generally introduced intentionally into the glass surface through surface treatments such as commercial SO2 dealkalization treatment, polishing, leaching, etc., or involuntarily through corrosion or weathering.90,97,362,363 SiOH can be formed through reactions

+ + 20,364 between non-bridging oxygen (NBO) and H /H3O and hydrolysis of bridging oxygen (BO).

365 At the same time, H2O molecules are transported into the glass matrix. This transport process can be enhanced by chemical reactions between NBO/BO and water since leached or corroded surfaces have more “open” silicate network. In some studies, it is also suggested that some of the newly formed SiOH groups can further react with each other, which creates new bridging oxygen

366,367 (BO) and H2O. Then, the relative abundance of SiOH/H2O can not only reflects changes in glass composition, but also the silicate network structure due to surface treatments. Determining the ratio of SiOH/H2O will be of great importance to understand the chemical reactions occurred during the surface treatment and modifications of surface mechanical and mechanochemical properties through these treatments.

Infrared spectroscopy is widely used to analyze the relative amount of SiOH and H2O in glasses. Titration and transmission IR have been utilized to determine the absorption coefficient of

10,11 OH and H2O in the glass bulk. Combination bands in near-IR (NIR) region is normally used to distinguish SiOH and H2O. However, combination peaks in NIR have relatively small absorption

225 cross sections, which means only large concentration of hydrous species can be determined reliably.

Also, this method might not be useful to determine the speciation of hydrous species in the glass

surface.[ref] In mid-IR region, the bending vibration of H2O (훿퐻2푂) can be contributed by water molecules alone while the OH stretching peak (휈푂퐻) is very broad due to complicated hydrogen

96,124 bonding interaction among SiOH groups and H2O molecules. Although the vibrational peaks in mid-IR region are sensitive to the total concentration of hydrous species within the probing depth, peak fittings remains challenging to separate the contribution of SiOH and H2O from broad

OH stretching vibrational bands.

Hydrous species in the surface region of glass materials has been analyzed with attenuated total reflectance infrared (ATR-IR) spectroscopy due to its excellent surface sensitivity. While the relative amount of hydrous species can be easily obtained from ATR-IR, determine the relative concentration of SiOH/H2O in leached or corroded surface is still not easy. The main complexity comes from the concentration gradient of hydrous species in these surfaces and intensity decay of evanescent wave from the ATR crystal-glass interface to the bulk. However, it is possible to obtain the SiOH/H2O ratios if these factors are properly taken into consideration.

This paper develops a method to determine the speciation of hydrous species in monolithic and flat glass surfaces based on the ATR-IR analysis and hydrogen (H) depth profile.

The relative concentration ratio of SiOH/H2O was determined for two different sets of leached

SLS glass surface: 1. a polished SLS glass with different leaching time; 2. SLS glasses with different fictive temperature under the same leaching conditions. The trend of SiOH/H2O suggests that the diffusion of water is highly dependent on the surface silicate network which changes dynamically as leaching time increases and varies for SLS glass with different fictive temperatures. The developed method was applied to hydrous speciation in the alteration layer formed at a sodium calcium aluminosilicate glass and a sodium aluminosilicate glass with similar sodium concentration.

226 Experimental details

ATR-IR spectra of liquid water was collected with a Thermo-Nicolet Nexus 670 spectrometer and a multiple reflection Si wafer ATR crystal. The Si ATR crystal was cleaned with ethanol, DI water, dried with N2 and exposed to UV/O3 for 30 minutes. The incidence angle of

ATR-IR was 45. The detailed setup can be found in previous publication.98 The information depth

(3  penetration depth) of water bending vibration region (~1640 cm-1) and OH stretching vibration region (2800-3700 cm-1) is approximately 936 nm and 462 nm, respectively.

The ATR-IR data and H-depth profile reported in this paper is reproduced from previous publications.2,368 H depth profile was obtained with time-of-flight secondary ion mass spectrometry

(ToF-SIMS). ATR-IR spectra was collected with a Bruker Vertex V70 spectrometer equipped with a DTGS detector. A diamond crystal with a refractive index of 2.4 was used. The incidence angle of this ATR-IR analysis was 45. The samples were pressed against the diamond ATR crystal with

420 N over ~1.5 mm2 area.

The details of the samples used in this study were prepared as follows. In the study time- dependent leaching behavior of SLS glass in acid solution, SLS glass was remelt from SLS float glass (Asahi Glass Co. Ltd, Tokyo, Japan). The remelt glass was annealed at 600 C and slowly cooled to the room temperature. The SLS glass was then lapped in SiC powder and polished in

CeO2 solution. The thickness of the samples were 5mm. The polished SLS glasses were then leached in a PH=1 HCl solution for 5 hours, 20 hours, 80 hours and 320hours.

For the study of fictive temperature effect on the leaching of SLS glass. SLS float glass was thinned to be 100 m by HF etching. This thin sample was then annealed at 600 C for 5 hours and slowly cooled down to room temperature. This sample would have a fictive temperature close to the glass transition temperature and is denoted as 600A. The samples with Fictive temperature lower than Tg was prepared by holding the glass at 500C for 60 hours and slowly cooled down to

227 room temperature. This sample was denoted as 500A. The samples with fictive temperatures that were higher than Tg were prepared by holding the samples at 600 C and 700 C for 15 minutes and quenched to room temperature. These two samples were denoted as 600Q and 700Q respectively.

The samples were then leached in an acid solution (PH=1, HCl) at 90 C for 600 hours.

Analysis on the OH/H2O ratio in other glasses were performed on a series of glasses with

- similar Na content and different amount of AlO4 tetrahedral. The samples were prepared with melt quench method. The compositions (mol%) of these glasses were summarized in previous publication.361 All the samples were annealed 50 C above their glass transition temperature and slowly cooled down to room temperature (cooling rate = 1 C/min). The samples were then polished in CeO2. The thickness of the samples is 1.6 mm. All these three samples were leached in PH=1 HCl solution at 90 C. The duration of leaching for SLS glass, NCAS glass and NAS glass is 320 hours, 500 hours and 100 hours respectively. Under these conditions, the thickness of leached layer are similar. Further details of the experimental set-up can be found in previous publications.2,96,97

Results and discussion

Algorithm to calculate the relative abundance of SiOH/H2O from ATR-IR spectrum

Figure B-1 shows a schematic of evanescent wave and hydrogen depth profile in the surface region of leached SLS glass. The evanescent wave decay is determined by IR incident angle, the complex refractive indices of the ATR crystal and the leached glass surface at the wavelength of the absorption band while hydrogen depth profile can be described by an inter- diffusion model.8 The overall information obtained from ATR-IR is proportional to the integral of

H depth profile  evanescent wave and described by the following equation:

228

푑 2푥 퐼 = ∑푖 훼푖 ∫ 퐶푖(푥) ∙ 퐸푥푝 (− ) 푑푥 (1) 0 푑푝

In equation (1), i is the absorption coefficient of i species in glass, x is depth from the surface to the bulk, dp is the penetration depth of the evanescent field and Ci(x) is the concentration of that species as a function of depth; in the case of hydrous species, it can be correlated with the

H depth profile.

Figure B-1. Schematics of H depth profile and evanescent wave generated from ATR-IR crystal.

Figure B-2. Schematics of the algorithm

229 Figure B-2 schematically describes methodology applied in this study. The key point here

is to properly determine the intensity ratio of 휈푂퐻, 퐻2푂 over 훿퐻2푂 in SLS glass by assuming the absorption coefficient ratio (훼 ⁄훼 ) obtained from liquid water can be used for water in 휈푂퐻, 퐻2푂 훿퐻2푂

SLS glass. The first step is to obtain this ratio from ATR-IR spectra of liquid water. Since the

penetration depth for 휈푂퐻, 퐻2푂 and 훿퐻2푂 is 154 nm and 312 nm respectively, the intensity of these two vibrational region are adjusted to 462 nm. Then, 훼 ⁄훼 can be determined to be 12.1 휈푂퐻, 퐻2푂 훿퐻2푂 from the following equation:

462 2푥 ∫ 퐸푥푝(− )푑푥 훼휈 0 푑 (훿 ) 푂퐻, 퐻2푂 퐼(휈푂퐻, 퐻2푂) 푝 퐻2푂 = 462 2푥 (1) 훼훿 퐼(훿퐻 푂) 퐻2푂 2 2 ∫ 퐸푥푝(− )푑푥 0 푑푝(휐푂퐻)

The next step is to get the intensity ratio of OH groups from SiOH and H2O. If the concentration ratio of SiOH/H2O can be assumed to be a constant (m) in the alteration layer as a function of depth, then the concentration of SiOH and H2O can be correlated with the hydrogen concentration as follows:

퐶 (푥) 퐶 (푥) 퐶(푆푖푂퐻) = 퐻 푚, 퐶(퐻 푂) = 퐻 (2) 푚+2 2 푚+2

By applying 훼 ⁄훼 ratio to the analysis of ATR-IR spectra of glass material, 휈푂퐻, 퐻2푂 훿퐻2푂

퐼(휈 ) the intensity ratio of 푂퐻, 퐻2푂 can be calculated as follows: 퐼(훿퐻2푂)

푑 2푥 ∫ 퐶 (푥)∙퐸푥푝(− )푑푥 퐼(휈 ) 훼 0 퐻 푑 (휐 ) 푂퐻, 퐻2푂 = 2 휈푂퐻, 푤푎푡푒푟 푝 푂퐻 (3) 퐼(훿퐻 푂) 훼훿 푑 2푥 2 퐻2푂 ∫ 퐶 (푥)∙퐸푥푝(− )푑푥 0 퐻 푑 (훿 ) 푝 퐻2푂

The broad OH stretching vibration peak could be properly allocated by 휈푂퐻, 퐻2푂 and

휈푂퐻,푆푖푂퐻. At last, the concentration ratio of SiOH/H2O can be obtained assuming OH stretching vibration of SiOH and H2O share the same absorption coefficient:

퐼(휐 ) 푚 = 2 푂퐻, 푆푖푂퐻 (4) 퐼(휈푂퐻, 퐻2푂)

230

Application of the SiOH/H2O calculation method to leached SLS glass analysis

Figure B-3. (a) ATR-IR spectra of DI water collected with Si crystal; (b) ATR-IR spectra of leached SLS glass with different leaching time.

To validate the methodology, SiOH/H2O was determined for SLS glass with the same thermal history but leached in hydrochloric acid solution (PH=1) at 90 C for varying periods.

ATR-IR spectra collected with diamond ATR crystal was used in previous studies (replotted in

Figure B-3b). The information depth of ATR-IR measurement is higher than the thickness of leached layer in all conditions. The amount of hydrous species in SLS glass increases as leaching time increases, which in general follow the trend of a diffusion model. Figure B-4 shows an example of how the evanescent wave can affect the ATR-IR intensity of OH and water as a function of depth. The total intensity of the ATR-IR spectra will be determined by integrating the product of evanescent wave decay and the concentration of hydrous species from glass surface to the bulk.

231

Figure B-4. Hydrogen depth profiles of leached SLS glasses with various time; evanescent wave -1 decay inside the glass for OH stretching vibration (OH, 3300 cm ) and H2O bending vibration -1 (water, 1640 cm ).

Validation the algorithm from leached SLS glass

Figure B-5. SiOH/H2O ratio for leached SLS glass shown in Figure B-2.

232

Figure B-5 shows the results of SiOH/H2O ratio in leached SLS glass with different amount of leaching time by applying the algorithm in previous section. As the leaching time increases,

SiOH/H2O concentration ratio decreases. This finding is in agreement with the idea that diffusion of water and reaction of NBOs are two relatively independent process. For leaching at short period of time (1hr& 5hr), more SiOH than H2O is introduced into the glass. At medium leaching time (20 hr & 80 hr), the ratio of SiOH/H2O is close to two. As the leaching time increases, the ratio of

SiOH/H2O becomes closer to one; at this point, the overall composition of hydrous species could

+ + be said to be stoichiometrically equivalent to H3O , replacing the leached Na ion. But, in reality, they are just an equal amount of hydroxyl and molecular water. The trend observed here suggests that formation of SiOH is a relatively faster process than water ingression.[ref] As leaching time increases, the structure of the alteration layer begins to form and change dynamically, which in general allows more water to ingress into the alteration layer. At the meantime, the rate for Na+ transport also decreases since the concentration gradient between alteration layer and the solution becomes smaller, which in turn has an impact on the SiOH formation.

233

Figure B-6. Product of hydrogen concentration and evanescent wave (~3300 cm-1) decay as a function of depth (a) 500A; (b) 600S; (c) 600Q; (d) 700Q. (e) ATR-IR spectra of leached SLS glasses with different fictive temperatures; (f) SiOH/H2O ratio in leached SLS glasses with different fictive temperatures.

To further validate this method, SiOH/H2O ratio in leached layer of SLS glass with different fictive temperatures and the same leaching conditions are calculated. These samples show quite different hydrogen depth profiles and different diffusion coefficient of hydrogen (DH) after leaching in the same conditions. In short, samples with higher fictive temperature (lower density) have larger amount of H and larger DH. The relative intensity of OH and water will be affected by the different H profiles as is shown in Figure B-6a – B-6d.

SiOH/H2O ratio is lower for SLS glass with lower density (higher fictive temperature) as is shown in Figure B-6f. From the SiOH/H2O ratio analysis, it is shown that more water will be introduced when glass density is smaller (higher fictive temperatures), which suggests the silicate network plays a greater role in the ingress of water. From hydrogen depth profile, SLS glass with density of 2.486 g/cm3 and 2.490 g/cm3 have quite similar amount of hydrogen in the leached layer and statistically indistinguishable DH as is shown in previous work. The difference in SiOH/H2O

234 ratio in these two samples might suggest more condensation reactions take place for glasses with lower density and more “silica-like” network. It can also be concluded that the transport of water is the rate-limiting step during the leaching and can be highly influenced by the thermal history of

SLS glasses.

Application of the SiOH/H2O calculation method to the leached layer on aluminosilicate glass

For some multicomponent glasses, Al is also an important network former element. During the leaching and corrosion of multicomponent glasses, AlOH group can also form. When all four

- bonds of AlO4 are hydrolyzed, Al(OH)4 can dissolve in the solution. This process can facilitate leaching and corrosion process. The degree of Al dissolution depends on the leaching and corrosion conditions. Knowing the OH/H2O ratio that remained in the leached or corroded surfaces can also help us understand the leaching or corrosion mechanism.

Here the algorithm is applied to analyze the leached multicomponent glasses with different Al concentrations and summarized in Figure B-7. Noted that SiOH and AlOH groups are assumed to have the same absorption coefficient. SLS glass shows slight lower OH/H2O ratio than the one shown in Figure B-5 for the same leaching condition. This should be due to the different thermal history and polishing protocols used before leaching treatment. The ATR-IR spectra of these two polished surfaces are also slightly different. NCAS has higher OH/H2O ratio

(~1.3) and NAS has the highest OH/H2O (~2) among these three samples. This result suggests

- that AlO4 tetrahedral will dissolve under these leaching conditions, which is also discussed and supported by XPS and indentation analysis in previous study.361

235

Figure B-7. Product of hydrogen concentration and evanescent wave (~3300 cm-1) decay as a function of depth (a) SLS glass; (b) NCAS glass; (c) NAS glass. ATR-IR spectra before and after leaching treatment (d) SLS glass; (e) NCAS glass; (f) NAS glass. The scale of y axis for (d), (e) and (f) is the same.

Sources of error limiting the accuracy of the SiOH/H2O calculation algorithm

Overall, this method is proven to be a semi-quantitative approach to determine the

SiOH/H2O concentration ratio in the surface region of SLS glass with assumptions on the absorption coefficient ratios. The ratio of OH/water obtained from liquid water is used for water molecules and hydroxyl groups in glasses. For glasses with similar composition and network structure, these assumptions will not affect the evolution of SiOH/H2O ratios determined in this study. Knowing this ratio will help us understand the reaction mechanisms that take place during the leaching or corrosion processes, including inter-diffusion of mobile cations and H+, dissolution of network formers and condensation of the silanol groups, etc. Better results could be obtained if the concentration of hydrous species of alteration layers can be determined properly.

Another possible experimental uncertainty is the H depth profile obtained from TOF-

SIMS. In principle, the top ~10 nm of SIMS profile cannot be detected reliably since the

236 sputtering equilibrium has not been reached. Mobile ions like H+ and Na+ have also been found to migrate inward during the sputtering process, which can cause uncertainty in the H depth profile.

If the shape of H depth profile is heavily distorted by these artifacts, the calculated OH/H2O ratio

+ will also be affected. In this study, C60 was used as the sputtering ions. This ion source and sputtering protocol has been found to be more reliable than other ion sources and sputtering protocols.7

Conclusion

In this study, a novel approach was developed to calculate the SiOH/H2O concentration ratio in the subsurface of SLS glass. SiOH/H2O ratio provides additional information to interpret the glass structural modification and their mechanisms during leaching treatment of SLS glass. This approach could also be applied to tackle corrosion mechanisms of nuclear waste glass, pharmaceutical containers, etc.

237 Reference

(1) Agarwal, A.; Tomozawa, M. Correlation of Silica Glass Properties with the Infrared Spectra. J. Non. Cryst. Solids 1997, 209 (1–2), 166–174. (2) Amma, S.; Luo, J.; Kim, S. H.; Pantano, C. G. Effects of Fictive Temperature on the Leaching of Soda Lime Silica Glass Surfaces. J. Am. Ceram. Soc. 2017, 100 (4), 1424–1431. (3) Pharr, G. M.; Oliver, W. C. Measurement of Thin Film Mechanical Properties Using Nanoindentation. MRS Bull. 1992, 17 (7), 28–33. (4) Wright, A. C. Neutron Scattering from Vitreous Silica. V. The Structure of Vitreous Silica: What Have We Learned from 60 Years of Diffraction Studies? J. Non. Cryst. Solids 1994, 179 (C), 84–115. (5) Urnes, S.; Andresen, A. F.; Herstad, O. Neutron Diffraction Studies of Silicate Glasses. J. Non. Cryst. Solids 1978, 29 (1), 1–14. (6) Greaves, G. N. EXAFS and the Structure of Glass. J. Non. Cryst. Solids 1985, 71 (1–3), 203–217. (7) Yamamoto, Y.; Yamamoto, K. Precise XPS Depth Profile of Soda-Lime-Silica Glass Using C60 Ion Beam. J. Non. Cryst. Solids 2010, 356 (1), 14–18. (8) Lanford, W. A.; Davis, K.; Lamarche, P.; Laursen, T.; Groleau, R.; Doremus, R. H. Hydration of Soda-Lime Glass. Journal of Non-Crystalline Solids. 1979, pp 249–266. (9) Leboeuf, V.; Blondeau, J. P.; De Sousa Meneses, D.; Véron, O. Potassium Ionic Exchange in Glasses for Mechanical Property Improvement. J. Non. Cryst. Solids 2013, 377, 60–65. (10) Stuke, A.; Behrens, H.; Schmidt, B. C.; Dupree, R. H2O Speciation in Float Glass and Soda Lime Silica Glass. Chem. Geol. 2006, 229 (1–3), 64–77. (11) Behrens, H.; Stuke, A. Quantification of H2O Contents in Silicate Glasses Using IR Spectroscopy - a Calibration Based on Hydrous Glasses Analyzed by Karl- Fischer Titration. Glass Science and Technology (Offenbach, Germany). 2003, pp 176–189. (12) Bakaev, V. A.; Bakaeva, T. I.; Pantano, C. G. On Inverse Adsorption Chromatography . 2 . Determination of Isotherms and Heats of Adsorption as Well as Energy Distributions of Adsorption Sites. J. Phys. Chem. C 2007, 111, 7473– 7486. (13) Ortiz Rivera, L.; Bakaev, V. A.; Banerjee, J.; Mueller, K. T.; Pantano, C. G. Characterization and Reactivity of Sodium Aluminoborosilicate Glass Fiber Surfaces. Appl. Surf. Sci. 2016, 370, 328–334. (14) Stapleton, J. J.; Suchy, D. L.; Banerjee, J.; Mueller, K. T.; Pantano, C. G. Adsorption Reactions of Carboxylic Acid Functional Groups on Sodium Aluminoborosilicate Glass Fiber Surfaces. ACS Appl. Mater. Interfaces 2010, 2 (11), 3303–3309. (15) Fett, T.; Guin, J. P.; Wiederhorn, S. M. Stresses in Ion-Exchange Layers of Soda- Lime-Silicate Glass. Fatigue Fract. Eng. Mater. Struct. 2005, 28 (6), 507–514.

238 (16) Wiederhorn, S. M.; Yi, F.; LaVan, D.; Richter, L. J.; Fett, T.; Hoffmann, M. J. Volume Expansion Caused by Water Penetration into Silica Glass. J. Am. Ceram. Soc. 2015, 98 (1), 78–87. (17) Lezzi, P. J.; Tomozawa, M.; Hepburn, R. W. Confirmation of Thin Surface Residual Compressive Stress in Silica Glass Fiber by FTIR Reflection Spectroscopy. J. Non. Cryst. Solids 2014, 390, 13–18. (18) Lezzi, P. J.; Xiao, Q. R.; Tomozawa, M.; Blanchet, T. a.; Kurkjian, C. R. Strength Increase of Silica Glass Fibers by Surface Stress Relaxation: A New Mechanical Strengthening Method. J. Non. Cryst. Solids 2013, 379, 95–106. (19) Lezzi, P. J.; Tomozawa, M.; Hepburn, R. W. Confirmation of Thin Surface Residual Compressive Stress in Silica Glass Fiber by FTIR Reflection Spectroscopy. J. Non. Cryst. Solids 2014, 390 (April), 13–18. (20) Ciccotti, M. Stress-Corrosion Mechanisms in Silicate Glasses. J. Phys. D. Appl. Phys. 2009, 42, 1–18. (21) WIEDERHORN, S. M.; BOLZ, L. H. Stress Corrosion and Static Fatigue of Glass. J. Am. Ceram. Soc. 1970, 53 (10), 543–548. (22) Wan, K. T.; Lathabai, S.; Lawn, B. R. Crack Velocity Functions and Thresholds in Brittle Solids. J. Eur. Ceram. Soc. 1990, 6 (4), 259–268. (23) Weber, N.; Goldstein, M. Stress-Induced Migration and Partial Molar Volume of Sodium Ions in Glass. J. Chem. Phys. 1964, 41 (9), 2898. (24) Langford, S. C.; Jensen, L. C.; Dickinson, J. T.; Pederson, L. R. Alkali Emission Accompanying Fracture of Sodium-Silicate Glasses. J. Mater. Res. 1991, 6 (6), 1358–1368. (25) Célarié, F.; Ciccotti, M.; Marlière, C. Stress-Enhanced Ion Diffusion at the Vicinity of a Crack Tip as Evidenced by Atomic Force Microscopy in Silicate Glasses. J. Non. Cryst. Solids 2007, 353 (1), 51–68. (26) Wiederhorn, S. M.; Fett, T.; Guin, J.-P.; Ciccotti, M. Griffith Cracks at the Nanoscale. Int. J. Appl. Glas. Sci. 2013, 4 (2), 76–86. (27) Girard, R.; Faivre, A.; Despetis, F. Influence of Water on Crack Self-Healing in Soda-Lime Silicate Glass. J. Am. Ceram. Soc. 2011, 94 (8), 2402–2407. (28) Singh, R. N. Kinetics of Self-Repair in Inorganic Glasses: Modeling and Experimental Verification. J. Mater. Sci. 2014, 49 (14), 4869–4879. (29) Zhuravlev, L. T. The Surface Chemistry of Amorphous Silica. Zhuravlev Model. Colloids Surfaces A Physicochem. Eng. Asp. 2000, 173 (1–3), 1–38. (30) Del Gaudio, P.; Behrens, H.; Deubener, J. Viscosity and Glass Transition Temperature of Hydrous Float Glass. J. Non. Cryst. Solids 2007, 353 (3), 223–236. (31) Luo, J.; Huynh, H.; Pantano, C. G.; Kim, S. H. Hydrothermal Reactions of Soda Lime Silica Glass – Revealing Subsurface Damage and Alteration of Mechanical Properties and Chemical Structure of Glass Surfaces. J. Non. Cryst. Solids 2016, 452, 93–101. (32) Science, A. S. Further Damage Induced by Water in Micro- Indentations in Phosphate Laser Glass. 2016, No. February 2014. (33) Bradley, L. C.; Dilworth, Z. R.; Barnette, A. L.; Hsiao, E.; Barthel, A. J.; Pantano, C. G.; Kim, S. H. Hydronium Ions in Soda-Lime Silicate Glass Surfaces. J. Am. Ceram. Soc. 2013, 96 (2), 458–463.

239 (34) He, H.; Qian, L.; Pantano, C. G.; Kim, S. H. Effects of Humidity and Counter- Surface on Tribochemical Wear of Soda-Lime-Silica Glass. Wear 2015, 342–343, 100–106. (35) He, H.; Kim, S. H.; Qian, L. Effects of Contact Pressure, Counter-Surface and Humidity on Wear of Soda-Lime-Silica Glass at Nanoscale. Tribol. Int. 2016, 94, 675–681. (36) Alazizi, A.; Barthel, A. J.; Surdyka, N. D.; Luo, J.; Kim, S. H. Vapors in the ambient—A Complication in Tribological Studies or an Engineering Solution of Tribological Problems? Friction 2015, 3 (2), 85–114. (37) He, H.; Yu, J.; Qian, L.; Pantano, C. G.; Kim, S. H. Enhanced Tribological Properties of Barium Boroaluminosilicate Glass by Thermal Poling. Wear 2017, 376–377, 337–342. (38) Yeon, J.; van Duin, A. C. T.; Kim, S. H. Effects of Water on Tribochemical Wear of Silicon Oxide Interface: Molecular Dynamics (MD) Study with Reactive Force Field (ReaxFF). Langmuir 2016, No. Md, acs.langmuir.5b04062. (39) He, H.; Qian, L.; Pantano, C. G.; Kim, S. H. Mechanochemical Wear of Soda Lime Silica Glass in Humid Environments. J. Am. Ceram. Soc. 2014, 97 (7), 2061–2068. (40) Geotti-Bianchini, F.; Preo, M.; Guglielmi, M.; Pantano, C. G. New Interpretation of the IR Reflectance Spectra of SiO2-Rich Films on Soda-Lime Glass. J. Non. Cryst. Solids 2003, 321 (1–2), 110–119. (41) Geotti-Bianchini, F.; Preo, M.; Guglielmi, M.; Pantano, C. G. Infrared Reflectance Spectra of Semi-Transparent SiO2 Rich Films on Silicate Glasses: Influence of the Substrate and Film Thickness. J. Non. Cryst. Solids 2003, 321 (1–2), 110–119. (42) Smith, N. J.; Pantano, C. G. Leached Layer Formation on Float Glass Surfaces in the Presence of Acid Interleave Coatings. J. Am. Ceram. Soc. 2008, 91 (3), 736– 744. (43) Almeida, R. M.; Guiton, T. A.; Pantano, C. G. Characterization of Silica Gels by Infrared Reflection Spectroscopy. J. Non. Cryst. Solids 1990, 121 (1–3), 193–197. (44) Tomozawa, M.; Lee, Y. K. Surface Fictive Temperature of Annealed and Rate- Cooled Soda-Lime Glasses. J. Non. Cryst. Solids 1999, 253 (1–3), 119–125. (45) Uchino, T.; Sakka, T.; Iwasaki, M. Interpretation of Hydrated States of Sodium Silicate Glasses by Infrared and Raman Analysis. J. Am. Ceram. Soc. 1991, 74 (2), 306–313. (46) Rubin, M. Optical Properties of Soda Lime Silica Glasses. Sol. Energy Mater. 1985, 12 (4), 275–288. (47) Kuzmenko, A. B. Kramers-Kronig Constrained Variational Analysis of Optical Spectra. Rev. Sci. Instrum. 2005, 76 (8), 1–9. (48) Macdonald, S. a; Schardt, C. R.; Masiello, D. J.; Simmons, J. H. Dispersion Analysis of FTIR Reflection Measurements in Silicate Glasses. J. Non. Cryst. Solids 2000, 275, 72–82. (49) Pemble, M. E.; Gardner, P. Infrared Spectroscopy from Surfacese. In Surface Analysis – The Principal Techniques; Vickerman, J. C., Gilmore, I. S., Eds.; John Wiley & Sons, Ltd: Chichester, West Sussex, PO19 8SQ, United Kingdom, 2009; pp 334–361.

240 (50) Wood, R. M. Optical Properties of Diamond: A Data Handbook. Optics and Lasers in Engineering. Springer 2004, pp 241–243. (51) Barnes, N. P.; Piltch, M. S. Temperature-Dependent Sellmeier Coefficients and Nonlinear Optics Average Power Limit for Germanium. Journal of the Optical Society of America. 1979, p 178. (52) Milosevic, M. Internal Reflection and ATR Spectroscopy; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2012. (53) Barthel, A. J.; Luo, J.; Hwang, K. S.; Lee, J.-Y.; Kim, S. H. Boundary Lubrication Effect of Organic Residue Left on Surface after Evaporation of Organic Cleaning Solvent. Wear 2016, 350–351, 21–26. (54) Kurkjian, C. R.; Kammlott, G. W.; Chaudhri, M. M. Indentation Behavior of Soda- Lime Silica Glass, Fused Silica, and Single-Crystal Quartz at Liquid Nitrogen Temperature. J. Am. Ceram. Soc. 1995, 78 (3), 737–744. (55) Varner, J. R.; Oel, H. J. SURFACE DEFECTS : THEIR ORIGIN , CHARACTERIZATION AND EFFECTS ON STRENGTH. 1975, 19, 321–333. (56) Neauport, J.; Ambard, C.; Cormont, P.; Darbois, N.; Destribats, J.; Luitot, C.; Rondeau, O. Subsurface Damage Measurement of Ground Fused Silica Parts by HF Etching Techniques. Opt. Express 2009, 17 (22), 20448–20456. (57) Kese, K. O.; Li, Z. C.; Bergman, B. Influence of Residual Stress on Elastic Modulus and Hardness of Soda-Lime Glass Measured by Nanoindentation. J. Mater. Res. 2004, 19 (10), 3109–3119. (58) Surdyka, N. D.; Pantano, C. G.; Kim, S. H. Environmental Effects on Initiation and Propagation of Surface Defects on Silicate Glasses: Scratch and Fracture Toughness Study. Appl. Phys. A 2014, 116 (2), 519–528. (59) Barthel, A. J.; Kim, S. H. Lubrication by Physisorbed Molecules in Equilibrium with Vapor at Ambient Condition: Effects of Molecular Structure and Substrate Chemistry. Langmuir 2014, 30 (22), 6469–6478. (60) Kese, K.; Rowcliffe, D. J. Nanoindentation Method for Measuring Residual Stress in Brittle Materials. Journal of the American Ceramic Society. 2003, pp 811–816. (61) Kassir-Bodon, A.; Deschamps, T.; Martinet, C.; Champagnon, B.; Teisseire, J.; Kermouche, G. Raman Mapping of the Indentation-Induced Densification of a Soda-Lime-Silicate Glass. Int. J. Appl. Glas. Sci. 2012, 3 (1), 29–35. (62) Kato, Y.; Yamazaki, H.; Yoshida, S.; Matsuoka, J. Effect of Densification on Crack Initiation under Vickers Indentation Test. J. Non. Cryst. Solids 2010, 356 (35–36), 1768–1773. (63) Freiman, S. W.; White, G. S.; Fuller, E. R. Environmentally Enhanced Crack Growth in Soda-Lime Glass. J. Am. Ceram. Soc. 1985, 68 (3), 108–112. (64) Ronchetti, C.; Lindqvist, M.; Louter, C.; Salerno, G. Stress-Corrosion Failure Mechanisms in Soda–lime Silica Glass. Eng. Fail. Anal. 2013, 35, 427–438. (65) Wong, L.; Suratwala, T.; Feit, M. D.; Miller, P. E.; Steele, R. The Effect of HF/NH4F Etching on the Morphology of Surface Fractures on Fused Silica. J. Non. Cryst. Solids 2009, 355 (13), 797–810. (66) Ezz-Eldin, F. M.; Abd-Elaziz, T. D.; Elalaily, N. A. Effect of Dilute HF Solutions on Chemical, Optical, and Mechanical Properties of Soda-Lime-Silica Glass. J. Mater. Sci. 2010, 45 (21), 5937–5949.

241 (67) He, H.; Luo, J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 2: Effects on Mechanical and Mechanochemical Properties. J. Am. Ceram. Soc. 2016, 99 (4), 1231–1238. (68) Charles, R. J. Static Fatigue of Glass. I. J. Appl. Phys. 1958, 29 (11), 1554. (69) Oehler, A.; Tomozawa, M. Water Diffusion into Silica Glass at a Low Temperature under High Water Vapor Pressure. J. Non. Cryst. Solids 2004, 347 (1–3), 211–219. (70) Liu, X.; He, J. One-Step Hydrothermal Creation of Hierarchical Microstructures toward Superhydrophilic and Superhydrophobic Surfaces. Langmuir 2009, 25 (19), 11822–11826. (71) Krohn, M. H.; Hellmann, J. R.; Shelleman, D. L.; Pantano, C. G.; Sakoske, G. E.; Corporation, F. Biaxial Flexure Strength and Dynamic Fatigue of Soda – Lime – Silica Float Glass. J. Am. Ceram. Soc. 2002, 85 (7), 1777–1782. (72) Lee, C. M.; Mohamed, N. M. A.; Watts, H. D.; Kubicki, J. D.; Kim, S. H. Sum- Frequency-Generation Vibration Spectroscopy and Density Functional Theory Calculations with Dispersion Corrections (DFT-D2) for Cellulose Iα and Iβ. J. Phys. Chem. B 2013, 117 (22), 6681–6692. (73) Smith, G. C. Evaluation of a Simple Correction for the Hydrocarbon Contamination Layer in Quantitative Surface Analysis by XPS. J. Electron Spectros. Relat. Phenomena 2005, 148 (1), 21–28. (74) Han, W.-T.; Tomozawa, M. Indentation Creep of Na2O*3SiO2 Glasses with Various Water Contents. J. Am. Ceram. Soc. 1990, 73 (197809), 3626–3632. (75) Barthel, A. J.; Kim, S. H. Lubrication by Physisorbed Molecules in Equilibrium with Vapor at Ambient Condition: Effects of Molecular Structure and Substrate Chemistry. Langmuir 2014, 30 (22), 6469–6478. (76) MICHALSKE, T. A.; FREIMAN, S. W. A Molecular Mechanism for Stress Corrosion in Vitreous Silica. J. Am. Ceram. Soc. 1983, 66 (4), 284–288. (77) Freiman, S. W.; Wiederhorn, S. M.; Mecholsky, Jr., J. J. Environmentally Enhanced Fracture of Glass: A Historical Perspective. J. Am. Ceram. Soc. 2009, 92 (7), 1371–1382. (78) Leed, E. A.; Pantano, C. G. Computer Modeling of Water Adsorption on Silica and Silicate Glass Fracture Surfaces. J. Non. Cryst. Solids 2003, 325 (1–3), 48–60. (79) Zeng, K.; Rowcliffe, D. Experimental Measurement of Residual Stress Field around Sharp Indentation in Glass. J. Am. Ceram. Soc. 1994, 77 (2), 524–530. (80) LAWN, B. R.; JAKUS, K.; GONZALEZ, A. C. Sharp vs Blunt Crack Hypotheses in the Strength of Glass: A Critical Study Using Indentation Flaws. J. Am. Ceram. Soc. 1985, 68 (1), 25–34. (81) Tsui, T. Y.; Pharr, G. M. Substrate Effects on Nanoindentation Mechanical Property Measurement of Soft Films on Hard Substrates. J. Mater. Res. 1999, 14, 292–301. (82) Milman, Y. V.; Golubenko, A. A.; Dub, S. N. Indentation Size Effect in Nanohardness. Acta Mater. 2011, 59 (20), 7480–7487. (83) Patschger, M.; Rüssel, C. Sodium/potassium Ion Exchange of a Soda-Lime-Silica Float Glass: Mechanical Strength, Hardness, Stress Profile and Thickness of the Ion Exchanged Layer. Glas. Technol. Eur. J. Glas. Sci. Technol. Part A 2015, 56

242 (5), 163–170. (84) Smedskjaer, M. M.; Bauchy, M. Sub-Critical Crack Growth in Silicate Glasses: Role of Network Topology. Appl. Phys. Lett. 2015, 107 (14). (85) Laurent, O.; Mantisi, B.; Micoulaut, M. Structure and Topology of Soda-Lime Silicate Glasses: Implications for Window Glass. J. Phys. Chem. B 2014, 118 (44). (86) Chen, L.; Yang, Y. J.; He, H. T.; Kim, S. H.; Qian, L. M. Effect of Coadsorption of Water and Alcohol Vapor on the Nanowear of Silicon. Wear 2015, 332–333, 879–884. (87) Iler, R. The Chemistry of Silica. Solubility, Polymerization, Colloid and Surface Properties, and Biochemistry; John Wiley & Sons, Inc., 1979. (88) Bunker, B. C. Molecular Mechanisms for Corrosion of Silica and Silicate Glasses. J. Non. Cryst. Solids 1994, 179 (1), 300–308. (89) Chuang, T.; Fuller, E. R. Extended Charles–Hillig Theory for Stress Corrosion Cracking of Glass. J. Am. Ceram. Soc. 1992, 75 (3), 540–545. (90) Akai, T.; Kuraoka, K.; Chen, D.; Yamamoto, Y.; Shirakami, T.; Urabe, K.; Yazawa, T. Leaching Behavior of Sodium from Fine Particles of Soda-Lime- Silicate Glass in Acid Solution. J. Am. Ceram. Soc. 2005, 88 (10), 2962–2965. (91) McGinnis, P. B.; Shelby, J. E. Diffusion of Water in Float Glass Melts. J. Non. Cryst. Solids 1994, 177, 381–388. (92) Shelby, J. E. A Limited Review of Water Diffusivity and Solubility in Glasses and Melts. J. Am. Ceram. Soc. 2008, 91 (3), 703–708. (93) Bandura, A. V. The Ionization Constant of Water over Wide Ranges of Temperature and Density. J. Phys. Chem. Ref. Data 2006, 35 (1), 15. (94) Covington, A. K.; Ferra, M. I. A.; Robinson, R. A. Ionic Product and Enthalpy of Ionization of Water from Electromotive Force Measurements. J. Chem. Soc. Faraday Trans. 1 Phys. Chem. Condens. Phases 1977, 73, 1721. (95) Luo, J.; He, H.; Podraza, N. J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda-Lime Silica Glass with Nonblocking Electrodes-Part 1: Effects of Sodium Ion Migration and Water Ingress on Glass Surface Structure. J. Am. Ceram. Soc. 2016, 99 (4), 1221–1230. (96) Amma, S.; Luo, J.; Pantano, C. G.; Kim, S. H. Specular Reflectance (SR) and Attenuated Total Reflectance (ATR) Infrared (IR) Spectroscopy of Transparent Flat Glass Surfaces: A Case Study for Soda Lime Float Glass. J. Non. Cryst. Solids 2015, 428, 189–196. (97) Amma, S. I.; Kim, S. H.; Pantano, C. G. Analysis of Water and Hydroxyl Species in Soda Lime Glass Surfaces Using Attenuated Total Reflection (ATR)-IR Spectroscopy. J. Am. Ceram. Soc. 2016, 99 (1), 128–134. (98) Asay, D. B.; Kim, S. H. Evolution of the Adsorbed Water Layer Structure on Silicon Oxide at Room Temperature. J. Phys. Chem. B 2005, 109 (35), 16760– 16763. (99) Du, Q.; Freysz, E.; Shen, Y. R. Surface Vibrational Spectroscopic Studies of Hydrogen Bonding and Hydrophobicity. Science 1994, 264 (5160), 826–828. (100) Efimov, A. M.; Pogareva, V. G.; Shashkin, A. V. Water-Related Bands in the IR Absorption Spectra of Silicate Glasses. J. Non. Cryst. Solids 2003, 332 (1–3), 93– 114.

243 (101) Libowitzky, E. Correlation of O-H Stretching Frequencies and O-H···O Hydrogen Bond Lengths in Minerals. Monatshefte fuer Chemie/Chemical Mon. 1999, 130 (8), 1047. (102) Wang, H.; DelloStritto, M. J.; Kumar, N.; Kolesnikov, A. I.; Kent, P. R. C.; Kubicki, J. D.; Wesolowski, D. J.; Sofo, J. O. Vibrational Density of States of Strongly H-Bonded Interfacial Water: Insights from Inelastic Neutron Scattering and Theory. J. Phys. Chem. C 2014, 118 (20), 10805–10813. (103) Davis, K. M. M.; Tomozawa, M. An Infrared Spectroscopic Study of Water- Related Species in Silica Glasses. J. Non. Cryst. Solids 1996, 201 (3), 177–198. (104) Navarra, G.; Iliopoulos, I.; Militello, V.; Rotolo, S. G.; Leone, M. OH-Related Infrared Absorption Bands in Oxide Glasses. J. Non. Cryst. Solids 2005, 351 (21– 23), 1796–1800. (105) Nibbering, E. T. J.; Elsaesser, T. Ultrafast Vibrational Dynamics of Hydrogen Bonds in the Condensed Phase. Chem. Rev. 2004, 104 (4), 1887–1914. (106) Zouine, A.; Dersch, O.; Walter, G.; Rauch, F. Diffusivity and Solubility of Water in Silica Glass in the Temperature Range 23 – 200 ° C. Glas. Sci. Technol. 2007, 48 (2), 85–91. (107) Davis, K. M.; Tomozawa, M. Water Diffusion into Silica Glass: Structural Changes in Silica Glass and Their Effect on Water Solubility and Diffusivity. J. Non. Cryst. Solids 1995, 185 (3), 203–220. (108) Tilocca, A.; de Leeuw, N.; Cormack, A. Shell-Model Molecular Dynamics Calculations of Modified Silicate Glasses. Phys. Rev. B 2006, 73 (10), 104209. (109) Tilocca, A.; Cormack, A. N. Modeling the Water−Bioglass Interface by Ab Initio Molecular Dynamics Simulations. ACS Appl. Mater. Interfaces 2009, 1 (6), 1324– 1333. (110) Luo, J.; Banerjee, J.; Pantano, C. G.; Kim, S. H. Vibrational Sum Frequency Generation Spectroscopy Study of Hydrous Species in Soda Lime Silica Float Glass. Langmuir 2016, 32 (24), 6035–6045. (111) Luzar, A.; Chandler, D. Hydrogen-Bond Kinetics in Liquid Water. Nature 1996, 379 (6560), 55–57. (112) Li, X.; Liu, L.; Schlegel, H. B. On the Physical Origin of Blue-Shifted Hydrogen Bonds. J. Am. Chem. Soc. 2002, 124 (32), 9639–9647. (113) Joseph, J.; Jemmis, E. D. Red-, Blue-, or No-Shift in Hydrogen Bonds: A Unified Explanation. J. Am. Chem. Soc. 2007, 129 (15), 4620–4632. (114) J., M. FTIR Techniques in Clay Minerals Studies: A Review. Vib. Spectrosc. 2003, 31, 1. (115) Della Ventura, G.; Marcelli, A.; Bellatreccia, F. SR-FTIR Microscopy and FTIR Imaging in the Earth Sciences. Rev. Mineral. Geochemistry 2014, 78 (1), 447–479. (116) Maréchal, Y. Infrared and Related Spectroscopies of H-Bonded Systems: Experimental Point of View. In The Hydrogen Bond and the Water Molecule; Elsevier, 2007; pp 77–113. (117) Du, Q.; Freysz, E.; Shen, Y. R. Vibrational Spectra of Water Molecules at Quartz/water Interfaces. Phys. Rev. Lett. 1994, 72 (2), 238–241. (118) Modig, K.; Pfrommer, B. G.; Halle, B. Temperature-Dependent Hydrogen Bond Geometry in Liquid Water. Phys. Rev. Lett. 2003, 90 (February), 75502.

244 (119) Dougherty, R. C. Temperature and Pressure Dependence of Hydrogen Bond Strength: A Perturbation Molecular Orbital Approach. J. Chem. Phys. 1998, 109 (17), 7372–7378. (120) Asay, D. B.; Kim, S. H. Effects of Adsorbed Water Layer Structure on Adhesion Force of Silicon Oxide Nanoasperity Contact in Humid Ambient. J. Chem. Phys. 2006, 124 (17), 174712. (121) Criscenti, L. J.; Kubicki, J. D.; Brantley, S. L. Silicate Glass and Mineral Dissolution: Calculated Reaction Paths and Activation Energies for Hydrolysis of a Q3 Si by H3O + Using Ab Initio Methods. J. Phys. Chem. A 2006, 110 (1), 198– 206. (122) Senturk, U.; Varner, J. R.; Lacourse, W. C. Structure-Hardness Relation for High- Temperature SO2-Dealkalized Float Glass. J. Non. Cryst. Solids 1997, 222, 160– 166. (123) Dirkes, W. E.; Rubey, W. a.; Pantano, C. G. The Formation of a Silica-Rich Surface Using Sulfur Dioxide in Drawn Glass Capillaries. J. High Resolut. Chromatogr. 1980, 3 (6), 303–305. (124) Senturk, U.; Lee, D. H.; Condrate, R. A.; Varner, J. R. ATR-FTIR Spectral Investigation of SO2-Treated Soda-Lime-Silicate Float Glass. MRS Proc. 1995, 407, 337–342. (125) Brow, R. K.; LaCourse, W. C. Fluorine Treatments of Soda-Lime Silicate Glass Surfaces. J. Am. Ceram. Soc. 1983, 66 (8), C-123-C-124. (126) Wang, J. G.; Stahley, M. E.; Pantano, C. G.; Yang, D. H.; Anderson, T.; Kuhn, L. Nanoindentation and Nanowear Studies of Sputter-Deposited Ultrathin Tin Oxide Films on Glass Substrates. MRS Proceedings. 2001, pp 2–7. (127) Schaeffer, H. A.; Stengel, M.; Mecha, J. Dealkalization of Glass Surfaces Utilizing HCI Gas. Journal of Non-Crystalline Solids. 1986, pp 400–404. (128) Geotti-Bianchini, F.; Preo, M.; Pantano, C. G. Study of Lubricating Coatings for Glass Containers with Optical Microscopy and FTIR Microscopy. Glas. Technol. 2002, 43 (4), 147–156. (129) Shallenberger, J. R.; Metwalli, E. E.; Pantano, C. G.; Tuller, F. N.; Fry, D. F. Adsorption of Polyamides and Polyamide-Silane Mixtures at Glass Surfaces. Surf. Interface Anal. 2003, 35 (8), 667–672. (130) Guiton, T. A.; Pantano, C. G. Infrared Reflectance Spectroscopy of Porous Silicas. Colloids and Surfaces A: Physicochemical and Engineering Aspects. 1993, pp 33– 46. (131) Smedskjaer, M. M.; Mauro, J. C.; Yue, Y. Ionic Diffusion and the Topological Origin of Fragility in Silicate Glasses. J. Chem. Phys. 2009, 131 (24), 1–10. (132) Karlsson, S.; Jonson, B.; Stalhandske, C. The Technology of Chemical Glass Strengthening - A Review. Glass Technology: European Journal of Glass Science and Technology Part A. 2010, pp 41–54. (133) Krieger, U. K.; Lanford, W. A. Field Assisted Transport of Na+ Ions, Ca2+ Ions and Electrons in Commercial Soda-Lime Glass I: Experimental. Journal of Non- Crystalline Solids. 1988, pp 50–61. (134) Zakel, J.; Balabajew, M.; Roling, B. On the Mechanism of Field-Induced Mixed Ionic–electronic Transport during Electro-Thermal Poling of a Bioactive Sodium–

245 calcium Phosphosilicate Glass. Solid State Ionics 2014, 265, 1–6. (135) Canagasabey, A.; Corbari, C.; Zhang, Z.; Kazansky, P. G.; Ibsen, M. Broadly Tunable Second-Harmonic Generation in Periodically Poled Silica Fibers. Opt. Lett. 2007, 32 (13), 1863–1865. (136) Quiquempois, Y.; Martinelli, G.; Duthérage, P.; Bernage, P.; Niay, P.; Douay, M. Localization of the Induced Second-Order Non-Linearity within Infrasil and Suprasil Thermally Poled Glasses. Opt. Commun. 2000, 176 (4), 479–487. (137) Moura, A. L.; Nascimento, E. M.; Araujo, M. T.; Vermelho, M. V. D.; Aitchison, J. S. Voltage Thresholdlike Evidence during Thermal-Electric Field Induction of Second Order Nonlinearity in Soda-Lime Glasses. J. Appl. Phys. 2009, 105 (3), 36106. (138) Smith, N. J.; Lanagan, M. T.; Pantano, C. G. Thermal Poling of Alkaline Earth Boroaluminosilicate Glasses with Intrinsically High Dielectric Breakdown Strength. J. Appl. Phys. 2012, 111 (8), 83519. (139) Ferreira, B.; Fargin, E.; Manaud, J. P. J.; Le Flem, G.; Rodriguez, V.; Buffeteau, T. Second Harmonic Generation Induced by Poling in Borophosphate Bulk and Thin Film Glasses. J. Non. Cryst. Solids 2004, 343 (1–3), 121–130. (140) Mukherjee, N.; Myers, R. A.; Brueck, S. R. J. Dynamics of Second-Harmonic Generation in Fused Silica. J. Opt. Soc. Am. B 1994, 11 (4), 665–669. (141) Dussauze, M.; Rodriguez, V.; Lipovskii, A.; Petrov, M.; Smith, C.; Richardson, K.; Cardinal, T.; Fargin, E.; Kamitsos, E. I. How Does Thermal Poling Affect the Structure of Soda-Lime Glass? J. Phys. Chem. C 2010, 114 (29), 12754–12759. (142) Cremoux, T.; Dussauze, M.; Fargin, E.; Cardinal, T.; Talaga, D.; Adamietz, F.; Rodriguez, V. Trapped Molecular and Ionic Species in Poled Borosilicate Glasses: Toward a Rationalized Description of Thermal Poling in Glasses. J. Phys. Chem. C 2014, 118 (7), 3716–3723. (143) Dussauze, M.; Cremoux, T.; Adamietz, F.; Rodriguez, V.; Fargin, E.; Yang, G.; Cardinal, T. Thermal Poling of Optical Glasses: Mechanisms and Second-Order Optical Properties. Int. J. Appl. Glas. Sci. 2012, 3 (4), 309–320. (144) MARIAPPAN, C.; ROLING, B. Investigation of Bioglass–electrode Interfaces after Thermal Poling. Solid State Ionics 2008, 179 (19–20), 671–677. (145) Ziemath, E. C.; Araújo, V. D.; Escanhoela, C. A. Compositional and Structural Changes at the Anodic Surface of Thermally Poled Soda-Lime Float Glass. J. Appl. Phys. 2008, 104 (5), 54912. (146) He, H.; Luo, J.; Qian, L.; Pantano, C. G.; Kim, S. H. Thermal Poling of Soda Lime Silica Glass . Part Ⅱ . Surface Mechanical Properties Change Enhanced Mechanochemical Properties of Soda Lime Silica Glass under Thermal Poling. J. Am. Ceram. Soc. 2015. (147) Zhu, X. D.; Suhr, H.; Shen, Y. R. Surface Vibrational Spectroscopy by Infrared- Visible Sum Frequency Generation. Phys. Rev. B 1987, 35 (6), 3047–3050. (148) Shen, Y. R. Surface Properties Probed by Second-Harmonic and Sum-Frequency Generation. Nature. 1989, pp 519–525. (149) Barnette, A. L.; Kim, S. H. Coadsorption of N -Propanol and Water on SiO 2 : Study of Thickness, Composition, and Structure of Binary Adsorbate Layer Using Attenuated Total Reflection Infrared (ATR-IR) and Sum Frequency Generation

246 (SFG) Vibration Spectroscopy. J. Phys. Chem. C 2012, 116 (18), 9909–9916. (150) Miller, F. A.; Wilkins, C. H. Infrared Spectra and Characteristic Frequencies of Inorganic Ions. Anal. Chem. 1952, 24, 1253–1294. (151) Kawaguchi, K.; Ikeda, H.; Sakai, D.; Funatsu, S.; Uraji, K.; Yamamoto, K.; Suzuki, T.; Harada, K.; Nishii, J. Accelerated Formation of Sodium Depletion Layer on Soda Lime Glass Surface by Corona Discharge Treatment in Hydrogen Atmosphere. Appl. Surf. Sci. 2014, 300, 149–153. (152) Lee, J.; Rovira, P. I.; An, I.; Collins, R. W. Rotating-Compensator Multichannel Ellipsometry: Applications for Real Time Stokes Vector Spectroscopy of Thin Film Growth. Rev. Sci. Instrum. 1998, 69 (4), 1800–1810. (153) Bungay, J. A. W. B. D. J. C. M. H. J. N. H. R. A. S. C. L. Overview of Variable- Angle Spectroscopic Ellipsometry (VASE): I. Basic Theory and Typical Applications. In Proc. SPIE CR72, Optical Metrology; 1999; pp 3–28. (154) Johs, B.; Herzinger, C. M. Quantifying the Accuracy of Ellipsometer Systems. Phys. Status Solidi Curr. Top. Solid State Phys. 2008, 5 (5), 1031–1035. (155) Fujiwara, H.; Koh, J.; Rovira, P.; Collins, R. Assessment of Effective-Medium Theories in the Analysis of Nucleation and Microscopic Surface Roughness Evolution for Semiconductor Thin Films. Phys. Rev. B 2000, 61 (16), 10832– 10844. (156) Collins, R. W.; Ferlauto, A. S. Optical Physics of Materials. In Handbook of Ellipsometry; 2005; pp 93–235. (157) Jbara, O.; Cazaux, J.; Trebbia, P. Sodium Diffusion in Glasses during Electron Irradiation. J. Appl. Phys. 1995, 78 (2), 868–875. (158) Cazaux, J. Electron Probe Microanalysis of Insulating Materials: Quantification Problems and Some Possible Solutions. X-Ray Spectrom. 1996, 25 (6), 265–280. (159) Shaaban, M. Field Assisted Replacement of Na+ by K+, Pb2+, Zn2+ and Sn2+ in Soda-Lime-Silica Glass. Phys. Chem. Glas. 1999, 40 (1), 34–39. (160) Private Communications with Asahi Glass Co, Ltd. 2015. (161) Tomozawa, M.; Hepburn, R. W. Surface Structural Relaxation of Silica Glass: A Possible Mechanism of Mechanical Fatigue. J. Non. Cryst. Solids 2004, 345–346, 449–460. (162) Koike, A.; Tomozawa, M. IR Investigation of Density Changes of Silica Glass and Soda-Lime Silicate Glass Caused by Microhardness Indentation. J. Non. Cryst. Solids 2007, 353 (24–25), 2318–2327. (163) Williams, C. T.; Beattie, D. A. Probing Buried Interfaces with Non-Linear Optical Spectroscopy. Surf. Sci. 2002, 500 (1–3), 545–576. (164) Raschke, M. B.; Shen, Y. R. Nonlinear Optical Spectroscopy of Solid Interfaces. Curr. Opin. Solid State Mater. Sci. 2004, 8 (5), 343–352. (165) Shen, Y. R. Nonlinear Optical Studies of Surfaces. Appl. Phys. A Solids Surfaces 1994, 59 (5), 541–543. (166) Eisenthal, K. B. Liquid Interfaces Probed by Second-Harmonic and Sum- Frequency Spectroscopy. Chem. Rev. 1996, 96 (4), 1343–1360. (167) Gragson, D. E.; McCarty, B. M.; Richmond, G. L. Ordering of Interfacial Water Molecules at the Charged Air/Water Interface Observed by Vibrational Sum Frequency Generation. J. Am. Chem. Soc. 1997, 119 (26), 6144–6152.

247 (168) Wang, H.-F.; Gan, W.; Lu, R.; Rao, Y.; Wu, B.-H. Quantitative Spectral and Orientational Analysis in Surface Sum Frequency Generation Vibrational Spectroscopy (SFG-VS). Int. Rev. Phys. Chem. 2005, 24 (2), 191–256. (169) Vidal, F.; Tadjeddine, A. Sum-Frequency Generation Spectroscopy of Interfaces. Reports Prog. Phys. 2005, 68 (5), 1095–1127. (170) Imai, H.; Horinouchi, S.; Asakuma, N.; Fukao, K.; Matsuki, D.; Hirashima, H.; Sasaki, K. Effects of Introduction of Sodium and Water on Second-Order Nonlinearity in Poled Synthetic Silica Glass. J. Appl. Phys. 1998, 84 (10), 5415– 5418. (171) An, H.; Fleming, S. Second-Order Optical Nonlinearity and Accompanying near- Surface Structural Modifications in Thermally Poled Soda-Lime Silicate Glasses. J. Opt. Soc. Am. B 2006, 23 (11), 2303. (172) Ong, S.; Zhao, X.; Eisenthal, K. B. Polarization of Water Molecules at a Charged Interface: Second Harmonic Studies of the Silica/water Interface. Chem. Phys. Lett. 1992, 191 (3–4), 327–335. (173) Lepienski, C. M.; Giacometti, J. A.; Leal Ferreira, G. F.; Freire Jr, F. L.; Achete, C. A. Electric Field Distribution Ons near-Surface Modifications in Soda-Lime Glass Submitted to a Dc Potencial. J. Non. Cryst. Solids 1993, 159, 204–212. (174) Shen, Y. R.; Ostroverkhov, V. Sum-Frequency Vibrational Spectroscopy on Water Interfaces: Polar Orientation of Water Molecules at Interfaces. Chem. Rev. 2006, 106 (4), 1140–1154. (175) Boyd, R. W. Order-of-Magnitude Estimates of the Nonlinear Optical Susceptibility. J. Mod. Opt. 1999, 46 (3), 367–378. (176) Lee, C. M.; Kafle, K.; Park, Y. B.; Kim, S. H. Probing Crystal Structure and Mesoscale Assembly of Cellulose Microfibrils in Plant Cell Walls, Tunicate Tests, and Bacterial Films Using Vibrational Sum Frequency Generation (SFG) Spectroscopy. Phys. Chem. Chem. Phys. 2014, 16 (22), 10844–10853. (177) Ju, S. S.; Wu, T.-D.; Yeh, Y.-L.; Wei, T.-H.; Huang, J.-Y.; Lin, S. H. Sum Frequency Vibrational Spectroscopy of the Liquid-Air Interface of Aqueous Solutions of Ethanol in the OH Region. J. Chinese Chem. Soc. (Taipei, Taiwan) 2001, 48 (3), 625–629. (178) Perera, P. N.; Fega, K. R.; Lawrence, C.; Sundstrom, E. J.; Tomlinson-Phillips, J.; Ben-Amotz, D. Observation of Water Dangling OH Bonds around Dissolved Nonpolar Groups. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (30), 12230–12234. (179) Tomozawa, M.; Cherniak, D. J.; Lezzi, P. J. Hydrogen-to-Alkali Ratio in Hydrated Alkali Aluminosilicate Glass Surfaces. J. Non. Cryst. Solids 2012, 358 (24), 3546– 3550. (180) Wang, C. Y.; Groenzin, H.; Shultz, M. J. Molecular Species on Nanoparticulate Anatase TiO2 Film Detected by Sum Frequency Generation: Trace Hydrocarbons and Hydroxyl Groups. Langmuir 2003, 19 (18), 7330–7334. (181) Du, Q.; Superfine, R.; Freysz, E.; Shen, Y. Vibrational Spectroscopy of Water at the Vapor/water Interface. Phys. Rev. Lett. 1993, 70 (15), 2313–2316. (182) Maréchal, Y. Infrared and Related Spectroscopies of H-Bonded Systems. In The Hydrogen Bond and the Water Molecule; Elsevier, 2007; pp 77–113. (183) Pool, R. E.; Versluis, J.; Backus, E. H. G.; Bonn, M. Comparative Study of Direct

248 and Phase-Specific Vibrational Sum-Frequency Generation Spectroscopy: Advantages and Limitations. J. Phys. Chem. B 2011, 115 (51), 15362–15369. (184) Talimian, A.; Mariotto, G.; Sglavo, V. M. Electric Field-Assisted Ion Exchange Strengthening of Borosilicate and Soda Lime Silicate Glass. Int. J. Appl. Glas. Sci. 2017, 8 (3), 291–300. (185) Abrams, M. B.; Green, D. J.; Glass, S. J. Fracture Behavior of Engineered Stress Profile Soda Lime Silicate Glass. J. Non. Cryst. Solids 2003, 321 (1–2), 10–19. (186) Yamamoto, Y.; Yamamoto, K. Precise XPS Depth Profile of Soda–lime–silica Float Glass Using C60 Ion Beam. Opt. Mater. (Amst). 2011, 33 (12), 1927–1930. (187) Fett, T.; Guin, J. P.; Wiederhorn, S. M. Stresses in Ion-Exchange Layers of Soda- Lime-Silicate Glass. Fatigue Fract. Eng. Mater. Struct. 2005, 28 (6), 507–514. (188) Sheth, N.; Luo, J.; Banerjee, J.; Pantano, C. G.; Kim, S. H. Characterization of Surface Structures of Dealkalized Soda Lime Silica Glass Using X-Ray Photoelectron, Specular Reflection Infrared, Attenuated Total Reflection Infrared and Sum Frequency Generation Spectroscopies. J. Non. Cryst. Solids 2017, 474 (August), 24–31. (189) Ghanassi, M.; Piveteau, L.; Saviot, L.; Schanne-Klein, M. C.; Ricard, D.; Flytzanis, C. Origin of the Resonant Optical Kerr Nonlinearity in Cd(S, Se)-Doped Glasses and Related Topics. Appl. Phys. B Laser Opt. 1995, 61 (1), 17–26. (190) Zhong, R. L.; Xu, H. L.; Li, Z. R.; Su, Z. M. Role of Excess Electrons in Nonlinear Optical Response. J. Phys. Chem. Lett. 2015, 6 (4), 612–619. (191) Lepienski, C.; Giacometti, J.; Leal Ferreira, G. F.; Freire, F. L.; Achete, C. A. Electric Field Distribution and near-Surface Modifications in Soda-Lime Glass Submitted to a Dc Potential. J. Non. Cryst. Solids 1993, 159 (3), 204–212. (192) Lepienski, C.; Giacometti, J.; Achete, C. Lipp Study of a Glass Sample Previously Submitted to a DC Potential. Solid State Commun. 1991, 79 (10), 825–828. (193) Matos, C.; Carvalho, I.; Silveira, E. F. d.; Margulis, W.; Pinho, R. .; Lesche, B. Charge Emission in Thermal Poling of Glasses with Carbon Film Anode. J. Non. Cryst. Solids 2000, 273 (1–3), 25–29. (194) Schmidt, B.; Nitzsche, P.; Lange, K.; Grigull, S.; Kreissig, U.; Thomas, B.; Herzog, K. In Situ Investigation of Ion Drift Processes in Glass during Anodic Bonding. Sensors Actuators A Phys. 1998, 67 (1–3), 191–198. (195) Redkov, A. V.; Melehin, V. G.; Lipovskii, A. A. How Does Thermal Poling Produce Interstitial Molecular Oxygen in Silicate Glasses? J. Phys. Chem. C 2015, 119 (30), 17298–17307. (196) Krieger, U. K.; Lanford, W. A. Field Assisted Transport of Na+ Ions, Ca2+ Ions and Electrons in Commercial Soda-Lime Glass I: Experimental. J. Non. Cryst. Solids 1988, 102 (1–3), 50–61. (197) Carlson, D. E.; Hang, K. W.; Stockdale, G. F. Ion Depletion of Glass at a Blocking Anode: II, Properties of Ion-Depleted Glasses. J. Am. Ceram. Soc. 1974, 57 (7), 295–300. (198) Lipovskii, A. A.; Melehin, V. G.; Petrov, M. I.; Svirko, Y. P.; Zhurikhina, V. V. Bleaching versus Poling: Comparison of Electric Field Induced Phenomena in Glasses and Glass-Metal Nanocomposites. J. Appl. Phys. 2011, 109 (1). (199) Dash, P.; Furman, E.; Pantano, C. G.; Lanagan, M. T. Activation Energy for

249 Alkaline-Earth Ion Transport in Low Alkali Aluminoborosilicate Glasses. Appl. Phys. Lett. 2013, 102 (8), 1–6. (200) Smith, N. J.; Pantano, C. G. Structural and Compositional Modification of a Barium Boroaluminosilicate Glass Surface by Thermal Poling. Appl. Phys. A 2014, 116 (2), 529–543. (201) Mariappan, C. R.; Roling, B. Mechanism and Kinetics of Na+ Ion Depletion under the Anode during Electro-Thermal Poling of a Bioactive Glass. J. Non. Cryst. Solids 2010, 356 (11–17), 720–724. (202) Luo, J.; Zhou, Y.; Milner, S. T.; Pantano, C. G.; Kim, S. H. Molecular Dynamics Study of Correlations between IR Peak Position and Bond Parameters of Silica and Silicate Glasses: Effects of Temperature and Stress. J. Am. Ceram. Soc. 2018, 101 (1), 178–188. (203) Skuja, L.; Güttler, B. Detection of Interstitial Oxygen Molecules in SiO2 Glass by a Direct Photoexcitation of the Infrared Luminescence of Singlet O2. Phys. Rev. Lett. 1996, 77, 2093–2096. (204) Boizot, B.; Petite, G.; Ghaleb, D.; Reynard, B.; Calas, G. Raman Study of β- Irradiated Glasses. J. Non. Cryst. Solids 1999, 243 (2–3), 268–272. (205) Max, J.-J.; Chapados, C. Isotope Effects in Liquid Water by Infrared Spectroscopy. III. H2O and D2O Spectra from 6000 to 0 cm−1. J. Chem. Phys. 2009, 131 (18), 184505. (206) Ishiuchi, S. I.; Fujii, M.; Robinson, T. W.; Miller, B. J.; Kjaergaard, H. G. Vibrational Overtone Spectroscopy of Phenol and Its Deuterated Isotopomers. J. Phys. Chem. A 2006, 110 (23), 7345–7354. (207) Sathyanarayana, D. N. Vibrational Spectroscopy: Theory and Applications; New Age International (P) Limited, 2015. (208) Flores, E.; Viallon, J.; Moussay, P.; Wielgosz, R. I. Accurate Fourier Transform Infrared (FT-IR) Spectroscopy Measurements of Nitrogen Dioxide (NO2) and Nitric Acid (HNO3) Calibrated with Synthetic Spectra. Appl. Spectrosc. 2013, 67 (10), 1171–1178. (209) Okuhara, T.; Hasada, Y.; Misono, M. In Situ Diffuse Reflectance IR of Catalytic Reduction of Nitrogen Oxides by Propene in the Presence of Oxygen over Silica- Supported Platinum. Catal. Today 1997, 35 (1–2), 83–88. (210) Hadjiivanov, K.; Bushev, V.; Kantcheva, M. Infrared Spectroscopy Study of the Species Arising during NO2 Adsorption on Ti02 ( Anatase ). Langmuir 1994, 2 (22), 464–471. (211) MORROW, B., Cheverier, P., Moran, L. An Infrared Study of the Adsorption of NO on Silica-Supported Platinum over a Wide Temperature Range. J. Catal. 1985, 91 (2), 208–215. (212) Varetti, E. L.; Pimentel, G. C. Isomeric Forms of Dinitrogen Trioxide in a Nitrogen Matrix. J. Chem. Phys. 1971, 55 (8), 3813–3821. (213) Ghiotti, G.; Garrone, E.; Boccuzzi, F. Infrared Study of Physical Adsorption. 2. Nitric Oxide on Silica Aerosil Surfaces. J. Phys. Chem. 1987, 91 (22), 5640–5645. (214) Matsumura, H.; Hayashi, T.; Chakraborty, S.; Lu, Y.; Moënne-Loccoz, P. The Production of Nitrous Oxide by the Heme/Nonheme Diiron Center of Engineered Myoglobins (Fe B Mbs) Proceeds through a Trans -Iron-Nitrosyl Dimer. J. Am.

250 Chem. Soc. 2014, 136 (6), 2420–2431. (215) Barth, B.; Mendt, M.; Pöppl, A.; Hartmann, M. Adsorption of Nitric Oxide in Metal-Organic Frameworks: Low Temperature IR and EPR Spectroscopic Evaluation of the Role of Open Metal Sites. Microporous Mesoporous Mater. 2014, 216. (216) Busaca, G., Lorenzelli, V. Infrared Study of the Adsorption of Nitrogen Dioxide, Nitric Oxide and Nitrous Oxide on Hematite. J. Catal. 1981, 72 (2), 303–313. (217) Sharp, D. W. A.; Thorley, J. 670. The Infrared Spectrum of the Nitrosonium Ion. J. Chem. Soc. 1963, 3557, 3557. (218) Gärtner, F. G.; Mollwo, E. IR Absorption of OH and OD Centres and OH/OH, OD/OD, and OH/OD Complexes in Cu-Doped ZnO Single Crystals. Phys. Status Solidi 1978, 89 (2), 381–388. (219) Tuck, J. R.; Korsunsky, A. M.; Bhat, D. G.; Bull, S. J. Indentation Hardness Evaluation of Cathodic Arc Deposited Thin Hard Coatings. Surf. Coatings Technol. 2001, 139 (1), 63–74. (220) Smedskjaer, M.; Mauro, J.; Yue, Y. Prediction of Glass Hardness Using Temperature-Dependent Constraint Theory. Phys. Rev. Lett. 2010, 105 (11), 10– 13. (221) Smedskjaer, M. M. Topological Model for Boroaluminosilicate Glass Hardness. Front. Mater. 2014, 1 (October), 23. (222) Malchow, P.; Johanns, K. E.; Möncke, D.; Korte-Kerzel, S.; Wondraczek, L.; Durst, K. Composition and Cooling-Rate Dependence of Plastic Deformation, Densification, and Cracking in Sodium Borosilicate Glasses during Pyramidal Indentation. J. Non. Cryst. Solids 2015, 419, 97–109. (223) Tadjiev, D. R.; Hand, R. J. Surface Hydration and Nanoindentation of Silicate Glasses. J. Non. Cryst. Solids 2010, 356 (2), 102–108. (224) Sellappan, P.; Rouxel, T.; Celarie, F.; Becker, E.; Houizot, P.; Conradt, R. Composition Dependence of Indentation Deformation and Indentation Cracking in Glass. Acta Mater. 2013, 61 (16), 5949–5965. (225) Luo, J.; Vargheese, K. D.; Tandia, A.; Hu, G.; Mauro, J. C. Crack Nucleation Criterion and Its Application to Impact Indentation in Glasses. Sci. Rep. 2016, 6 (1), 23720. (226) Varner, J. R.; Oel, H. J. Surface Defects: Their Origin, Characterization and Effects on Strength. J. Non. Cryst. Solids 1975, 19, 321–333. (227) Kurkjian, C. R.; Gupta, P. K.; Brow, R. K. The Strength of Silicate Glasses: What Do We Know, What Do We Need to Know? Int. J. Appl. Glas. Sci. 2010, 1 (1), 27–37. (228) Yoshida, S.; Kato, M.; Yokota, A.; Sasaki, S.; Yamada, A.; Matsuoka, J.; Soga, N.; Kurkjian, C. R. Direct Observation of Indentation Deformation and Cracking of Silicate Glasses. J. Mater. Res. 2015, 30 (15), 2291–2299. (229) Rouxel, T.; Ji, H.; Guin, J. P.; Augereau, F.; Rufflé, B. Indentation Deformation Mechanism in Glass: Densification versus Shear Flow. J. Appl. Phys. 2010, 107 (9), 94903. (230) Peter, K. W. Densification and Flow Phenomena of Glass in Indentation Experiments. J. Non. Cryst. Solids 1970, 5 (2), 103–115.

251 (231) Hermansen, C.; Matsuoka, J. Densification and Plastic Deformation under Microindentation in Silicate Glasses and the Relation to Hardness and Crack Resistance. J. Non- … 2013. (232) Cheng, W.; Finnie, I. A Prediction on the Strength of Glass Following the Formation of Subsurface Flaws by Scribing. J. Am. Ceram. Soc. 1992, 75 (9), 2565–2572. (233) Sehgal, J.; Ito, S. A New Low‐Brittleness Glass in the Soda‐Lime‐Silica Glass Family. J. Am. Ceram. Soc. 1998, 88 (190539), 2485–2488. (234) Schneider, J.; Schula, S.; Weinhold, W. P. Characterisation of the Scratch Resistance of Annealed and Tempered Architectural Glass. Thin Solid Films 2012, 520 (12), 4190–4198. (235) Talimian, A.; Sglavo, V. M. Ion-Exchange Strengthening of Borosilicate Glass: Influence of Salt Impurities and Treatment Temperature. J. Non. Cryst. Solids 2017, 456, 12–21. (236) Varshneya, A. K. Chemical Strengthening of Glass: Lessons Learned and Yet To Be Learned. Int. J. Appl. Glas. Sci. 2010, 1 (2), 131–142. (237) Svenson, M. N.; Thirion, L. M.; Youngman, R. E.; Mauro, J. C.; Rzoska, S. J.; Bockowski, M.; Smedskjaer, M. M. Pressure-Induced Changes in Interdiffusivity and Compressive Stress in Chemically Strengthened Glass. ACS Appl. Mater. Interfaces 2014, 6 (13), 10436–10444. (238) Sglavo, V. M.; Bonafini, M.; Prezzi, A. Procedure for Residual Stress Profile Determination by Curvature Measurements. Mech. Mater. 2005, 37 (8), 887–898. (239) Varshneya, A. K.; Olson, G. A.; Kreski, P. K.; Gupta, P. K. Buildup and Relaxation of Stress in Chemically Strengthened Glass. J. Non. Cryst. Solids 2015, 427, 91–97. (240) Barthel, A. J.; Gregory, M. D.; Kim, S. H. Humidity Effects on Friction and Wear between Dissimilar Metals. Tribol. Lett. 2012, 48 (3), 305–313. (241) Banerjee, J.; Kim, S. H.; Pantano, C. G. Quantification of the Areal Density of Elements and Oxygen Speciation on Flat Glass Surfaces with X-Ray Photoelectron Spectroscopy. 1–29. (242) Gy, R. Ion Exchange for Glass Strengthening. Mater. Sci. Eng. B Solid-State Mater. Adv. Technol. 2008, 149 (2), 159–165. (243) Oliver, W. C.; Pharr, G. M. Measurement of Hardness and Elastic Modulus by Instrumented Indentation: Advances in Understanding and Refinements to Methodology. J. Mater. Res. 2011, 19 (1), 3–20. (244) Widjaja, S.; Yip, T. H.; Limarga, a. M. Measurement of Creep-Induced Localized Residual Stress in Soda-Lime Glass Using Nano-Indentation Technique. Mater. Sci. Eng. A 2001, 318 (1–2), 211–215. (245) Quinn, G. D. Fractography of Glasses and Ceramics V; Varner, J. R., Quinn, G. D., Wightman, M., Eds.; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2007. (246) Quinn, G. D. Fracture Toughness of Ceramics by the Vickers Indentation Crack Length Method: A Critical Review. Mech. Prop. Perform. Eng. Ceram. II Ceram. Eng. Sci. Proc. 2008, 27 (2), 45–62. (247) Tandia, A.; Vargheese, K. D.; Mauro, J. C.; Varshneya, A. K. Atomistic Understanding of the Network Dilation Anomaly in Ion-Exchanged Glass. J. Non.

252 Cryst. Solids 2012, 358 (2), 316–320. (248) Wang, M.; Smedskjaer, M. M.; Mauro, J. C.; Sant, G.; Bauchy, M. Topological Origin of the Network Dilation Anomaly in Ion-Exchanged Glasses. Phys. Rev. Appl. 2017, 8 (5), 54040. (249) Efimov, A. M.; Pogareva, V. G. Water-Related IR Absorption Spectra for Some Phosphate and Silicate Glasses. J. Non. Cryst. Solids 2000, 275 (3), 189–198. (250) Zarubin, D. P. Infrared Spectra of Hydrogen Bonded Hydroxyl Groups in Silicate Glasses. A Re-Interpretation. Phys. Chem. Glas. 1999, 40 (4), 184–192. (251) Wang, X.; Andrews, L. Infrared Spectroscopic Observation of the Group 13 Metal Hydroxides, M (OH) 1, 2, 3 (M= Al, Ga, In, and Tl) and HAl (OH) 2. J. Phys. Chem. A 2007, 3, 1860–1868. (252) Luo, J.; Bae, S.; Yuan, M.; Schneider, E.; Lanagan, M. T.; Pantano, C. G.; Kim, S. H. Chemical Structure and Mechanical Properties of Soda Lime Silica Glass Surfaces Treated by Thermal Poling in Inert and Reactive Ambient Gases. J. Am. Ceram. Soc. 2018. (253) Yazawa, T.; Kuraoka, K.; Akai, T.; Umesaki, N.; Du, W.-F. Clarification of Phase Separation Mechanism of Sodium Borosilicate Glasses in Early Stage by Nuclear Magnetic Resonance. J. Phys. Chem. B 2000, 104 (9), 2109–2116. (254) Hamilton, J. P.; Pantano, C. G. Effects of Glass Structure on the Corrosion Behavior of Sodium-Aluminosilicate Glasses. J. Non. Cryst. Solids 1997, 222, 167–174. (255) Wassick, T. H.; Doremus, R. H.; Lanford, W. A.; Burman, C. Hydration of Soda Lime Silicate Glass, Effect of Alumina. J. Non. Cryst. Solids 1983, 54, 139–151. (256) Henderson, G. S. A Si K-Edge EXAFS/XANES Study of Sodium Silicate Glasses. J. Non. Cryst. Solids 1995, 183 (1–2), 43–50. (257) Husung, R. D.; Doremus, R. H. The Infrared Transmission Spectra of Four Silicate Glasses before and after Exposure to Water. J. Mater. Res. 1990, 5 (10), 2209– 2217. (258) Hancer, M.; Sperline, R. P.; Miller, J. D. Anomalous Dispersion Effects in the IR- ATR Spectroscopy of Water. Appl.Spectrosc. 2000, 54 (1), 138–143. (259) Berreman, D. W. Kramers-Kronig Analysis of Reflectance Measured at Oblique Incidence. Appl. Opt. 1967, 6 (9), 1519. (260) Yamamoto, K.; Ishida, H. Kramers-Kronig Analysis Applied to Reflection- Absorption Spectroscopy. Vib. Spectrosc. 1997, 15 (1), 27–36. (261) Gaskell, P. H.; Johnson, D. W. The Optical Constants of Quartz, Vitreous Silica and Neutron-Irradiated Vitreous Silica (I). J. Non. Cryst. Solids 1976, 20 (2), 153– 169. (262) Roseler, A.; Korte, E.-H. Infrared Spectroscopic Ellipsometry. In Handbook of Vibrational Spectroscopy; Chalmers, J. M., Ed.; John Wiley & Sons, Ltd: Chichester, UK, 2006. (263) Synowicki, R. A. Suppression of Backside Reflections from Transparent Substrates. Phys. Status Solidi Curr. Top. Solid State Phys. 2008, 5 (5), 1085– 1088. (264) Gin, S.; Abdelouas, A.; Criscenti, L. J.; Ebert, W. L.; Ferrand, K.; Geisler, T.; Harrison, M. T.; Inagaki, Y.; Mitsui, S.; Mueller, K. T.; et al. An International

253 Initiative on Long-Term Behavior of High-Level Nuclear Waste Glass. Mater. Today 2013, 16 (6), 243–248. (265) Hapke, B. Theory of Reflectance and Emittance Spectroscopy; Cambridge University Press: Cambridge, 2011. (266) Azzam, R. M. A. On the Reflection of Light at 45 ° Angle of Incidence. Opt. Acta (Lond). 1979, 26 (1), 113–115. (267) Yamamoto, K.; Masui, a. Complex Refractive-Index Determination of Bulk Materials from Infrared Reflection Spectra. Appl. Spectrosc. 1995, 49 (5), 639– 644. (268) Tomozawa, M.; Lee, Y.-K.; Peng, Y.-L. Effect of Uniaxial Stresses on Silica Glass Structure Investigated by IR Spectroscopy. J. Non. Cryst. Solids 1998, 242 (2–3), 104–109. (269) Shelby, R. M.; Harris, C. B.; Cornelius, P. A. The Origin of Vibrational Dephasing of Polyatomic Molecules in Condensed Phases. J. Chem. Phys. 1979, 70 (1), 34. (270) Wilson, M.; Madden, P. A.; Hemmati, M.; Angell, C. A. Polarization Effects, Network Dynamics, and the Infrared Spectrum of Amorphous SiO2. Phys. Rev. Lett. 1996, 77 (19), 4023–4026. (271) Merlin, J.-C.; Cornard, J.-P. A Pictorial Representation of Normal Modes of Vibration Using Vibrational Symmetry Coordinates. J. Chem. Educ. 2006, 83 (9), 1393. (272) Hamilton, V. E. Thermal Infrared (Vibrational) Spectroscopy of Mg-Fe Olivines: A Review and Applications to Determining the Composition of Planetary Surfaces. Chemie der Erde - Geochemistry 2010, 70 (1), 7–33. (273) Lee, C. M.; Kubicki, J. D.; Fan, B.; Zhong, L.; Jarvis, M. C.; Kim, S. H. Hydrogen-Bonding Network and OH Stretch Vibration of Cellulose: Comparison of Computational Modeling with Polarized IR and SFG Spectra. J. Phys. Chem. B 2015, 119 (49), 15138–15149. (274) HANNA, R. Infrared Absorption Spectrum of Silicon Dioxide. J. Am. Ceram. Soc. 1965, 48 (11), 595–599. (275) McMillan, P. F. Structural Studies of Silicate Glasses and Melts-Applications and Limitations of Raman Spectroscopy. Am. Mineral. 1984, 69, 622–644. (276) PARSONS, J. L.; MILBERG, M. E. Vibrational Spectra of Vitreous B2O3.xH2O. J. Am. Ceram. Soc. 1960, 43 (6), 326–330. (277) Hanna, R.; Su, G.-J. Infrared Absorption Spectra of Sodium Silicate Glasses from 4 to 30μ. J. Am. Ceram. Soc. 1964, 47 (12), 597–601. (278) White, S. Study of Sodium Silicate Glasses and Liquids by Infrared Reflectance Spectroscopy. Journal of Non-Crystalline Solids. April 1969, pp 125–130. (279) Sciacca, M. D.; Mayur, A. J.; Oh, E.; Ramdas, A. K.; Rodriguez, S.; Furdyna, J. K.; Melloch, M. R.; Beetz, C. P.; Yoo, W. S. Infrared Observation of Transverse and Longitudinal Polar Optical Modes of Semiconductor Films: Normal and Oblique Incidence. Phys. Rev. B 1995, 51 (12), 7744–7752. (280) Scott, J.; Porto, S. Longitudinal and Transverse Optical Lattice Vibrations in Quartz. Phys. Rev. 1967, 161 (1934), 903–910. (281) Lee, S. K.; Stebbins, J. F. The Distribution of Sodium Ions in Aluminosilicate Glasses: A High-Field Na-23 MAS and 3Q MAS NMR Study. Geochim.

254 Cosmochim. Acta 2003, 67 (9), 1699–1709. (282) Bertmer, M.; Züchner, L.; Chan, J. C. C.; Eckert, H. Short and Medium Range Order in Sodium Aluminoborate Glasses. 2. Site Connectivities and Cation Distributions Studied by Rotational Echo Double Resonance NMR Spectroscopy. J. Phys. Chem. B 2000, 104 (28), 6541–6553. (283) E, P. Handbook of Optical Constants of Solids, 1st Editio.; E, P., Ed.; Academic Press, 1998. (284) Varshneya, A. K. The Physics of Chemical Strengthening of Glass: Room for a New View. J. Non. Cryst. Solids 2010, 356 (44–49), 2289–2294. (285) Patschger, M.; Rüssel, C. Strengthening of a Soda–lime–silica Glass by Ion Exchange Using an Adherent Potassium Salt Coating. Glas. Technol. Eur. J. Glas. Sci. Technol. Part A 2016, 57 (1), 6–14. (286) Marshall, D. B.; Lawn, B. R. An Indentation Technique for Measuring Stresses in Tempered Glass Surfaces. J. Am. Ceram. Soc. 1977, 60 (1–2), 86–87. (287) Ahn, S. min; Park, S. Y.; Kim, Y. C.; Lee, K. S.; Kim, J. Y. Surface Residual Stress in Soda-Lime Glass Evaluated Using Instrumented Spherical Indentation Testing. J. Mater. Sci. 2015, 50 (23), 7752–7759. (288) Jannotti, P.; Subhash, G.; Ifju, P.; Kreski, P. K.; Varshneya, A. K. Influence of Ultra-High Residual Compressive Stress on the Static and Dynamic Indentation Response of a Chemically Strengthened Glass. J. Eur. Ceram. Soc. 2012, 32 (8), 1551–1559. (289) Devine, R. A. B. Ion Implantation- and Radiation-Induced Structural Modifications in Amorphous SiO2. J. Non. Cryst. Solids 1993, 152 (1), 50–58. (290) Tomozawa, M.; Hong, J. W.; Ryu, S. R. Infrared (IR) Investigation of the Structural Changes of Silica Glasses with Fictive Temperature. J. Non. Cryst. Solids 2005, 351 (12–13), 1054–1060. (291) Galeener, F. L. Band Limits and the Vibrational Spectra of Tetrahedral Glasses. Phys. Rev. B 1979, 19 (8), 4292–4297. (292) Ishikawa, K.; Suzuki, K.; Okamura, S. Asymmetric Peak Line Shape of Infrared Dielectric Function Spectra for Thermally Grown Silicon Dioxide Films. J. Appl. Phys. 2000, 88 (12), 7150. (293) Champagnon, B.; Martinez, V.; Martinet, C.; Le Parc, R.; Levelut, C. Density and Density Fluctuations Anomalies of SiO 2 Glass: Comparison and Light-Scattering Study. Philos. Mag. 2007, 87 (3–5), 691–695. (294) Ingram, M. D.; Wu, M. H.; Coats, A.; Kamitsos, E. I.; Varsamis, C. P. E.; Garcia, N.; Sola, M. Evidence from Infrared Spectroscopy of Structural Relaxation during Field Assisted and Chemically Driven Ion Exchange in Soda-Lime-Silica Glasses. Phys. Chem. Glas. 2005, 46 (2), 84–89. (295) Sharaf, N. A.; Condrate, R. A.; Ahmed, A. A. FTIR Spectral/structural Investigation of the Ion Exchange/thermal Treatment of Silver Ions into a Silicate Glass. Mater. Lett. 1991, 11 (3–4), 115–118. (296) Guldiren, D.; Erdem, İ.; Aydin, S. Influence of Silver and Potassium Ion Exchange on Physical and Mechanical Properties of Soda Lime Glass. J. Non. Cryst. Solids 2016, 441, 1–9. (297) Sglavo, V. M.; Quaranta, A.; Allodi, V.; Mariotto, G. Analysis of the Surface

255 Structure of Soda Lime Silicate Glass after Chemical Strengthening in Different KNO3 Salt Baths. J. Non. Cryst. Solids 2014, 401, 105–109. (298) Duquennoy, M.; Devos, D.; Ouaftouh, M.; Lochegnies, D.; Roméro, E. Ultrasonic Evaluation of Residual Stresses in Flat Glass Tempering: Comparing Experimental Investigation and Numerical Modeling. J. Acoust. Soc. Am. 2006, 119 (6), 3773– 3781. (299) Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E. GROMACS 4: Algorithms for Highly Efficient, Load-Balanced, and Scalable Molecular Simulation. J. Chem. Theory Comput. 2008, 4 (3), 435–447. (300) van Beest, B. W. H.; Kramer, G. J.; van Santen, R. A. Force Fields for Silicas and Aluminophosphates Based on Ab Initio Calculations. Phys. Rev. Lett. 1990, 64 (16), 1955–1958. (301) Vollmayr, K.; Kob, W.; Binder, K. Cooling-Rate Effects in Amorphous Silica: A Computer-Simulation Study. Phys. Rev. B 1996, 54 (22), 15808–15827. (302) Pedone, A.; Malavasi, G.; Menziani, M. C.; Segre, U.; Musso, F.; Corno, M.; Civalleri, B.; Ugliengo, P. FFSiOH: A New Force Field for Silica Polymorphs and Their Hydroxylated Surfaces Based on Periodic B3LYP Calculations. Chem. Mater. 2008, 20 (7), 2522–2531. (303) Chandra, A.; Patey, G. N. Dielectric Relaxation of Electrolyte Solutions: Molecular Dynamics and Theoretical Results for Ions in Simple Dipolar Solvents. J. Chem. Phys. 1994, 100 (11), 8385–8391. (304) Saiz, L.; Guàrdia, E.; Padró, J.-À. Dielectric Properties of Liquid Ethanol. A Computer Simulation Study. J. Chem. Phys. 2000, 113 (7), 2814–2822. (305) Misra, M.; Kumar, S. K. Using Time–Temperature Superposition for Determining Dielectric Loss in Functionalized Polyethylenes. ACS Macro Lett. 2017, 6 (3), 200–204. (306) Neumann, M.; Steinhauser, O. Computer Simulation and the Dielectric Constant of Polarizable Polar Systems. Chem. Phys. Lett. 1984, 106 (6), 563–569. (307) Sanders, M. J.; Leslie, M.; Catlow, C. R. A. Interatomic Potentials for SiO2. J. Chem. Soc. Chem. Commun. 1984, 4 (19), 1271–1273. (308) Lasaga, A. C.; Cygan, R. T. Electronic and Ionic Polarizabilities of Silicate Minerals. Am. Mineral. 1982, 67 (3–4), 328–334. (309) Agarwal, A.; Davis, K. M.; Tomozawa, M. A Simple IR Spectroscopic Method for Determining Fictive Temperature of Silica Glasses. J. Non. Cryst. Solids 1995, 185 (1–2), 191–198. (310) Nakamura, M.; Kanzawa, R.; Sakai, K. Stress and Density Effects on Infrared Absorption Spectra of Silicate Glass Films. J. Electrochem. … 1986, 133 (6), 1167–1171. (311) Lehmann, A.; Schumann, L.; Hubner, K. Optical Phonons in Amorphous Silicon Oxides. 1. Calculation of the Density of States and Interpretation of Lo-To Splittings of Amorphous SiO2. Phys. Status Solidi B 1983, 117 (2), 689–698. (312) Vargheese, K. D.; Tandia, A.; Mauro, J. C. Molecular Dynamics Simulations of Ion-Exchanged Glass. J. Non. Cryst. Solids 2014, 403, 107–112. (313) Du, J.; Cormack, A. N. The Medium Range Structure of Sodium Silicate Glasses: A Molecular Dynamics Simulation. J. Non. Cryst. Solids 2004, 349 (1–3), 66–79.

256 (314) Ito, S.; Taniguchi, T. Effect of Cooling Rate on Structure and Mechanical Behavior of Glass by MD Simulation. J. Non. Cryst. Solids 2004, 349 (1–3), 173– 179. (315) Brückner, R. Properties and Structure of Vitreous Silica. I. J. Non. Cryst. Solids 1970, 5 (2), 123–175. (316) Takada, A.; Richet, P.; Catlow, C. R. A.; Price, G. D. Molecular Dynamics Simulations of Vitreous Silica Structures. J. Non. Cryst. Solids 2004, 345–346, 224–229. (317) Chaplot, S. L. Morse Stretch Potential Charge Equilibrium Force Field for Ceramics: Application to the Quartz-Stishovite Phase Transition and to Silica Glass. Phys. Rev. Lett. 1999, 83 (18), 3749–3749. (318) Cowen, B. J.; El-Genk, M. S. On Force Fields for Molecular Dynamics Simulations of Crystalline Silica. Comput. Mater. Sci. 2015, 107, 88–101. (319) Wondraczek, L.; Mauro, J. C.; Eckert, J.; Kühn, U.; Horbach, J.; Deubener, J.; Rouxel, T. Towards Ultrastrong Glasses. Adv. Mater. 2011, 23 (39), 4578–4586. (320) Axinte, E. Glasses as Engineering Materials: A Review. Mater. Des. 2011, 32 (4), 1717–1732. (321) Shelby, J. E. Introduction to Glass Science and Technology, 2nd ed.; Royal Society of Chemistry: Cambridge, 2005. (322) Hülsenberg, D.; Harnisch, A.; Bismarck, A. Microstructuring of Glasses; Springer Series in Materials Science; Springer Berlin Heidelberg: Berlin, Heidelberg, 2008; Vol. 87. (323) Feldmann, M.; Weissmann, R. Initial Stages of Float Glass Corrosion. J. Non. Cryst. Solids 1997, 218, 205–209. (324) Wei, X.; Hong, S.-C.; Lvovsky, a I.; Held, H.; Shen, Y. R. Evaluation of Surface vs Bulk Contributions in Sum-Frequency Vibrational Spectroscopy Using Reflection and Transmission Geometries †. J. Phys. Chem. B 2000, 104 (14), 3349–3354. (325) Du, J.; Cormack, A. N. Molecular Dynamics Simulation of the Structure and Hydroxylation of Silica Glass Surfaces. J. Am. Ceram. Soc. 2005, 88 (9), 2532– 2539. (326) Mahadevan, T. S.; Garofalini, S. H. Dissociative Chemisorption of Water onto Silica Surfaces and Formation of Hydronium Ions. J. Phys. Chem. C 2008, 112 (5), 1507–1515. (327) Webb, E. B.; Garofalini, S. H. Relaxation of Silica Glass Surfaces before and after Stress Modification in a Wet and Dry Atmosphere: Molecular Dynamics Simulations. J. Non. Cryst. Solids 1998, 226 (1–2), 47–57. (328) Lockwood, G. K.; Garofalini, S. H. Proton Dynamics at the Water Silica Interface via Dissociative Molecular Dynamics. J. Phys. Chem. C 2014, 118, 29750–29759. (329) Tielens, F.; Gervais, C.; Lambert, J. F.; Mauri, F.; Costa, D. Ab Initio Study of the Hydroxylated Surface of Amorphous Silica: A Representative Model. Chem. Mater. 2008, 20 (10), 3336–3344. (330) Adkins, L.; Cormack, A. Large-Scale Simulations of Sodium Silicate Glasses. J. Non. Cryst. Solids 2011, 357 (14), 2538–2541. (331) Huang, C.; Cormack, A. N. The Structure of Sodium Silicate Glass. J. Chem. Phys.

257 1990, 93 (11), 8180. (332) Lowenstern, J. B.; Pitcher, B. W. Analysis of H2O in Silicate Glass Using Attenuated Total Refectance (ATR) Micro-FTIR Spectroscopy. Am. Mineral. 2013, 98 (10), 1660–1668. (333) Lambert, A. G.; Davies, P. B.; Neivandt, D. J. Implementing the Theory of Sum Frequency Generation Vibrational Spectroscopy: A Tutorial Review. Appl. Spectrosc. Rev. 2005, 40 (2), 103–145. (334) Barnette, A. L.; Bradley, L. C.; Veres, B. D.; Schreiner, E. P.; Park, Y. B.; Park, J.; Park, S.; Kim, S. H. Selective Detection of Crystalline Cellulose in Plant Cell Walls with Sum-Frequency-Generation (SFG) Vibration Spectroscopy. Biomacromolecules 2011, 12 (7), 2434–2439. (335) Lee, C. M.; Kafle, K.; Huang, S.; Kim, S. H. Multimodal Broadband Vibrational Sum Frequency Generation (MM-BB-V-SFG) Spectrometer and Microscope. J. Phys. Chem. B 2016, 120 (1), 102–116. (336) Starr, F. W.; Nielsen, J. K.; Stanley, H. E. Fast and Slow Dynamics of Hydrogen Bonds in Liquid Water. Phys. Rev. Lett. 1998, 82, 2294–2297. (337) Bonnaud, P. A.; Coasne, B.; Pellenq, R. J.-M. Molecular Simulation of Water Confined in Nanoporous Silica. J. Phys. Condens. Matter 2010, 22 (28), 284110. (338) D’Souza, A. S.; Pantano, C. G. Hydroxylation and Dehydroxylation Behavior of Silica Glass Fracture Surfaces. J. Am. Ceram. Soc. 2002, 85 (6), 1499–1504. (339) Wu, C. Stable Silicate Glasses Containing up to 10 Weight Percent of Water. J. Non. Cryst. Solids 1980, 41 (3), 381–398. (340) Oven, R.; Ashworth, D. G.; Page, M. C. On the Analysis of Field-Assisted Ion Diffusion into Glass. J. Phys. Condens. Matter 1992, 4 (16), 4089–4094. (341) Tian, C. S.; Shen, Y. R. Isotopic Dilution Study of the Water/vapor Interface by Phase-Sensitivesum-Frequency Vibrational Spectroscopy. J. Am. Chem. Soc. 2009, 131 (8), 2790–2791. (342) Mondal, J. A.; Nihonyanagi, S.; Yamaguchi, S.; Tahara, T. Structure and Orientation of Water at Charged Lipid Monolayer/Water Interfaces Probed by Heterodyne-Detected Vibrational Sum Frequency Generation Spectroscopy. J. Am. Chem. Soc. 2010, 132 (31), 10656–10657. (343) Nihonyanagi, S.; Mondal, J. A.; Yamaguchi, S.; Tahara, T. Structure and Dynamics of Interfacial Water Studied by Heterodyne-Detected Vibrational Sum- Frequency Generation. Annu. Rev. Phys. Chem. 2013, 64 (1), 579–603. (344) Dondur, V.; Fidler, D.; Adnadjević, B.; Rakić, V. The Kinetics of Water Desorption from Porous Glasses. J. Therm. Anal. 1987, 32 (2), 613–622. (345) Redhead, P. A. Thermal Desorption of Gases. Vacuum 1962, 12 (4), 203–211. (346) Travert, A.; Vimont, A.; Lavalley, J.-C.; Montouillout, V.; Rodríguez Delgado, M.; Cuart Pascual, J. J.; Otero Areán, C. Evidence for Discrepancy between the Surface Lewis Acid Site Strength and Infrared Spectra of Adsorbed Molecules: The Case of Boria−Silica. J. Phys. Chem. B 2004, 108 (42), 16499–16507. (347) Wang, X.; Andrews, L. Infrared Spectroscopic Observation of the Group 13 Metal Hydroxides, M(OH)1,2,3 (M =Al, Ga, In, and Tl) and HAl(OH)2. J. Phys. Chem. A 2007, 111 (10), 1860–1868. (348) Hadjiivanov, K. Identification and Characterization of Surface Hydroxyl Groups

258 by Infrared Spectroscopy. In Advances in Catalysis; 2014; Vol. 57, pp 99–318. (349) Wang, X.; Andrews, L. Infrared Spectra and Structures for Group 4 Dihydroxide and Tetrahydroxide Molecules. J. Phys. Chem. A 2005, 10689–10701. (350) Lu, R.; Gan, W.; Wu, B. H.; Zhang, Z.; Guo, Y.; Wang, H. F. C-H Stretching Vibrations of Methyl, Methylene and Methine Groups at the vapor/Alcohol (N = 1-8) Interfaces. J. Phys. Chem. B 2005, 109 (29), 14118–14129. (351) Sekine, T.; Suzuki, T.; Yamamoto, K. Inward Migration of Glass-Modifier Cations During Heat Treatment Under an N 2 Atmosphere. J. Am. Ceram. Soc. 2015, 98 (5), 1464–1470. (352) Jen, J. S.; Kalinowski, M. R. An ESCA Study of the Bridging to Non-Bridging Oxygen Ratio in Sodium Silicate Glass and the Correlations to Glass Density and Refractive Index. J. Non. Cryst. Solids 1980, 38–39, 21–26. (353) Miura, Y.; Matsumoto, S.; Nanba, T.; Akazawa, T. X-Ray Photoelectron Spectroscopy of Sodium Aluminosilicate Glasses. 2000, 41 (1), 24–31. (354) Sprenger, D.; Bach, H.; Meisel, W.; Gütlich, P. XPS Study of Leached Glass Surfaces. J. Non. Cryst. Solids 1990, 126 (1–2), 111–129. (355) Iogansen, A. V. Direct Proportionality of the Hydrogen Bonding Energy and the Intensification of the Stretching ν(XH) Vibration in Infrared Spectra. Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 1999, 55 (7–8), 1585–1612. (356) Braithwaite, R. S. W. Infrared Transmission Spectra of Carbonate Minerals. Mineral. Mag. 1994, 58 (392), 523–524. (357) V.C., F. The Infrared Spectra of Minerals; Farmer, V. C., Ed.; Mineralogical Society of Great Britain and Ireland: Londan, 1974. (358) Bourg, I. C.; Steefel, C. I. Molecular Dynamics Simulations of Water Structure and Diffusion in Silica Nanopores. J. Phys. Chem. C 2012, 116 (21), 11556– 11564. (359) Ji, N.; Ostroverkhov, V.; Chen, C.; Shen, Y. Phase-Sensitive Sum-Frequency Vibrational Spectroscopy and Its Application to Studies of Interfacial Alkyl Chains. 2007, 10056–10057. (360) Tian, C. S.; Shen, Y. R. Structure and Charging of Hydrophobic Material/water Interfaces Studied by Phase-Sensitive Sum-Frequency Vibrational Spectroscopy. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (36), 15148–15153. (361) Amma, S.; Luo, J.; Kim, S. H.; Pantano, C. G. Effect of Glass Composition on the Hardness of Surface Layers on Aluminosilicate Glasses Formed through Reaction with Strong Acid. J. Am. Ceram. Soc. 2018, 101 (2), 657–665. (362) Uchino, T.; Sakka, T.; Iwasaki, M. Interpretation of Hydrated States of Sodium Silicate Glasses by Infrared and Raman Analysis. J. Am. … 1991, 13 (197541), 306–313. (363) March, P.; Rauch, F. Leaching Studies of Soda-Lime-Silica Glass Using Deuterium-and 18O-Enriched Solutions. Glas. Berichte 1990. (364) P., M. B. The Structure of Na2O–Al2O3–SiO2 Glass: Impact on Sodium Ion Exchange in H2O and D2O. J. Austalia Ceram. Soc. 2001. (365) Richter, T.; Frischat, G. H.; Borchardt, G.; Scherrer, S. SHORT TIME LEACHING OF A SODA-LIME GLASS IN H2O AND D2O. Phys. Chem. Glas. 1985, 26 (6), 208–212.

259 (366) Banerjee, J.; Kim, S. H.; Pantano, C. G. Elemental Areal Density Calculation and Oxygen Speciation for Flat Glass Surfaces Using X-Ray Photoelectron Spectroscopy. J. Non. Cryst. Solids 2016, 450, 185–193. (367) Bunker, B. C.; Tallant, D. R.; Headley, T. J.; Turner, G. L.; Kirkpatrick, R. J. THe Structure of Leached Borosilicate Glass. Phys. Chem. Glas. 1988, 29 (3), 106–120. (368) Amma, S.; Kim, S. H.; Pantano, C. G. Analysis of Water and Hydroxyl Species in Soda Lime Silica Glass Surfaces Using Attenuated Total Reflection (ATR) - IR Spectroscopy. J. Am. Ceram. Soc. 2015.

260

VITA

Jiawei Luo

Jiawei Luo was born on Jan. 9th, 1989 in Dalian City, Liaoning Province, China. He received his bachelor’s degree in Chemical Engineering (thesis) and bachelor’s degree in English

Literature (non-thesis) from Dalian University of Technology in 2012. Following his undergraduate studies, he came to Penn State to pursue Ph. D. degree in chemical engineering in the fall of 2012. He joined Dr. Seong Kim and Dr. Carlo Pantano’s research group in 2013, studying the wear behavior of soda lime silica glass from surface science’s perspective. Jiawei’s interest include the surface chemistry of glass materials, principles of vibrational spectroscopy

(IR , SFG, Raman), surface mechanical properties determined from indentation analysis

(Nanoindentation, crack initiation load). He collaborated with glass industry for various projects during his graduate school study. In April 2018, he will join OFS Fitel LLC as a Research

Scientist and Engineer to develop next generation optical fiber.