Fragility, Melt/Glass Homogenization, Self-Organization in Chalcogenide Alloy Systems

Home , Liquidus

Fragility, Melt/Glass Homogenization, Self-Organization in

Chalcogenide Alloy Systems

A dissertation submitted to

Division of Research and Advanced Studies

University of Cincinnati

In partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY (Ph.D.)

In the Department of School of Electronics and Computing Systems

of the College of Engineering

September 2013

Kapila Gunasekera

M.Sc. Electrical Engineering, University of Cincinnati, Cincinnati (2010)

B.Sc. Electrical Engineering, North Dakota State University, Fargo (2008)

Thesis Advisor and Committee Chair: Dr. P. Boolchand ABSTRACT

We report on fragility and observation of the three elastic phases in the following

chalcogenides: GexSe100-x, GexSbxSe100-2x and GexSixTe100-2x. In each case, experimental evidence

shows a strong correlation between the three elastic phases and the variation of melt fragility

index, m(x. In general, m(x) is high (> 25) in the flexible and stressed-rigid phase but decrease

remarkably in the Intermediate Phase (IP) to show a global minimum (< 20). These observations

correlate the melt structure (view above Tg ) to the glass structure (view below Tg) and suggest

that the strong (low fragility) character of melts in the IP compositions is due to presence of

adaptability and extended range structural correlations in the rigid but unstressed networks

formed in the melt, features that they share with the (self-organized) networks formed in

corresponding glasses at T < Tg.

Comprehensive Raman scattering, calorimetric glass transitions, non-reversing enthalpy

* of relaxation at Tg, complex specific heat (Cp ), and volumetric measurements on the

GexSbxSe100-2x ternary were undertaken in the 0 < x < 22% range. These data provide evidence of

a rather well defined reversibility window, volumetric window and fragility window in the IP

compositions, 14.9% < x < 17.5% range. A curious local minimum of fragility, m(x), was observed in the flexible phase, and may represent presence of quasi-tetrahedral 4-fold coordinated Sb structural motifs in melts but not in corresponding glasses.

Elastic phases of GexSixTe100-2x ternary glasses are established by measurements of

non-reversing enthalpy of relaxation at Tg, and volumetric measurements. Fragility of

corresponding melts are established from Cp* measurements. The flexible phase extends in the

6% < x < 7.5% range, intermediate phase in the 7.5%< x<9% range , and stressed-rigid phase in the 9% < x < 12% range. Glasses at x > 12% are found to be chemically phase separated. A

global minimum in molar volumes of glasses-the volumetric window coincides with the

reversibility window, and confirms the space filing nature of networks formed in the IP. Non-

aging in a Telluride chalcogenide system is observed for the first time in the present study.

Retention loss of the amorphous phase, which can be attributed to physical aging in glasses, is a

key reliability issue in phase change memory devices which could be resolved by resorting to

glassy compositions inside intermediate phases where physical again is found to be minimal.

Most significantly, the physics underlying slow homogenization of chalcogenide alloy

melts/glasses is addressed in each of the three systems. Global minimum of fragility for IP melts

is responsible for inhibiting melt mixing at high temperatures. For that reason, special care was

taken to synthesize homogeneous melts/glasses in each case by reacting starting materials at high

temperatures and FT-Raman profiling melts to ascertain their homogeneity. Physical properties

of chalcogenide melts/glasses are found to vary systematically as their heterogeneity is steadily

lowered by prolonged melt reaction. These results are key to establishing the intrinsic physical

behavior of chalcogenides glasses in compositional studies both at a basic level and for real

world applications.

iii

AKNOWLEDGEMENTS

First of all I would like to thank my thesis advisor Dr. Punit Boolchand for the guidance, inspiration, valuable advice and much needed help given to me throughout my time as a graduate student at University of Cincinnati. I got the opportunity not only to work with one of the world renowned scientists in the field of glass science but also found a lifelong friend in him.

I would like to thank Steve Hall, Dr. Kadine Mohomed, Len Thomas and Dr. Steven

Aubuchon at TA Instruments for their valuable advice and assistance provided throughout the course of this study. I would also like to thank Barry Zuk at ThermoFisher scientific for being there when I needed help.

I would also like to extend my sincere gratitude to all my lab mates I got the opportunity to work with during my time in Cincinnati. I would like to thank all the past students, Dr. Deassy

Novita, Dr. Ping Chen, Jacob Wachtman, Dr, Kandasamy Vignarooban, Siddhesh Bhosle and present students Shibalik Chakraborty and Sriram Ravindren for all their help and valuable advice given throughout the course of this study.

I would also like to thank my parents for all their support throughout my life. I’m forever in their debt for the sacrifices they made for me to be where I am right now. I would like to thank my wife and my daughter for all the moral support they gave me and being there with me through thick and thin.

Finally I would like to thank my thesis committee members, Dr. Matthieu Micoulaut, Dr.

Chong Ahn, Dr. Marc Cahay, Dr. Peter Kosel and Dr. Wayne Bresser for their guidance and valuable advice. I would also like to thank University of Cincinnati for all the facilities provided

for students and the National Science Foundation for providing financial support for this body of work through the grant DMR 08-53957.

TABLE OF CONTENTS

LIST OF FIGURES AND TABLES viii

GLOSSARY OF TERMS xiv

CHAPTER 1: Introduction 1

1.1 Rigidity theory and elastic phases in network glasses 1

1.2 Slow homogenization of chalcogenide melts 2

1.3 Rigidity Theory and Phase Change Materials 3

1.4 Fragility and Elastic Phases in Glasses 4

1.5 Focus of present work 5

CHAPTER 2: Experimental Methods 7

2.1 Synthesis of glasses 7

2.2 Heat flow measurements from MDSC 8

2.3 Complex Cp measurements from mDSC 9

2.4 Molar volume measurements 12

2.5 FT-Raman scattering measurements 13

2.6 Mossbauer spectroscopy 14

CHAPTER 3: Experimental Results 15

3.1 GexSe100-x system 15

3.1.1 mDSC Cp* measurements 15

3.1.2 Slow Homogenization: FT-Raman measurements 18

3.1.3 Slow Homogenization: Molar volumes and fragility index variation 25

3.2 GexSbxSe100-x system 28

3.2.1 mDSC Cp* measurements 28

3.2.2 Slow Homogenization: FT-Raman measurements 30

3.3 GexSixTe100-2x system 31

3.3.1 mDSC heat flow measurements 31

3.3.2 mDSC Cp* measurements 33

3.3.3 Molar volume measurements 35

3.3.4 Mossbauer measurements 35

3.3.4.1 GexSixTe100-2x ternary glasses 35

3.1.4.2 GexTe100-x & SixTe100-x binary glasses 37

CHAPTER 4: Discussion 39

4.1 Melt fragility and Slow Homogenization of Ge-Se melts/glasses 40

4.2 Melt heterogeneity and Interfacial regions 47

4.3 Manifestations of melt heterogeneities in optical measurements 50

4.4 Correlating Melt Fragility and glass elastic phases in the GexSbxSe100-2x ternary 50

4.5 Melt-fragility, slow homogenization and elastic phases in the GexSixTe100-2x ternary 53

4.5.1 Melt fragility and Elastic Phases of GexSixTe100-2x 55

4.5.2 Slow homogenization of GexSixTe100-2x 57

4.5.3 Aspect of structure derived from the elastic phases and 119Sn Mossbauer 60

experiments.

CHAPTER 5: Conclusions 65

BIBLIOGRAPHY 67

APPENDIX: List of publications and conference presentations 79

vii

LIST OF FIGURES

Figure 2.1: Example of a mDSC scan at x = 13.2% of GexSbxSe100-2x. Signals corresponding to the heating cycle are indicated by right arrows and cooling cycle depicted by left arrows.

Figure 2.2: Cp* measurement of GexSe100-x, x = 10% sample. Note that as the modulation frequency is increased, the step in the In-Phase component and the endothermic peak of the

Out-of-Phase component shifts to higher temperatures.

Figure 2.3: Fragility index measurement of select compositions in GexSe100-x glasses. Slope of the curve yields fragility index. High slope curves are characteristic of fragile glassy melts whereas low slope curve are characterized as strong melts.

Figure 2.4: Raman profiling of x = 19% glass in GexSe100-2x after 6h of reacting starting materials. Raman spectra were acquired at several locations along the length of the sample column at 2.5mm intervals.

Figure 3.1: In-phase and out-of-phase components of Cp* as a function of temperature plotted for different modulation periods for select compositions in GexSe100-x binary glasses. In-phase

Cp* shows a step at Tg while the out of phase component shows an endothermic peak. Note that both in phase and out of phase signals shift to higher temperatures as the modulation period is decreased.

Figure 3.2: (a) Fragility index of GexSe100-x glasses plotted as a function of x%. (○) shows m values reported by Stolen et al.54 and () shows m values of the present study extracted from

10 Cp* measurements. (b) ΔHnr values of the present system reported by Bhosle et al.

Figure 3.3: Activation energy values derived from Cp* measurements of GexSe100-x binary.

Figure 3.4: In-situ Raman spectra obtained along the length of the sample column for

0 Ge23Se77. Sample tube was kept vertical inside a box furnace at 950 C and upon quenching

viii

Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals. Note that the intensity spread in the Se chain mode progressively decreases as reaction time is increased

Figure 3.5: In-situ Raman spectra obtained for Ge21Se79. (a-d) Sample tube was kept inside a box furnace in a vertical position and (e-h) sample tube was rocked at a rate of 6 rev/min for the duration of synthesis. Note that in both instances sample homogenizes after 144h of reaction time.

Figure 3.6: In-situ Raman spectra obtained along the length of the sample column for

0 Ge23Se77. Sample tube was kept inside a rocking furnace at 950 C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals.

Figure 3.7: Compositional variation of Raman spectra obtained along the length of the sample column for (a) Ge23Se77 and (b) Ge21Se79. Sample tube was kept vertical inside a box furnace at 9500C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals.

Figure 3.8: Compositional variation of Raman spectra obtained along the length of the sample column for (a) Ge23Se77 and (b) Ge21Se79. Sample tube was kept inside a rocking furnace at

9500C and upon quenching Raman spectra were acquired at 9 locations which were spaced at

2.5mm intervals.

10 Figure 3.9: Molar volume variation in GexSe100-x binary glasses reported by Bhosle et al. () ,

Mahadevan et al.(Δ)65, Feltz et al.(□)66, Avetikyan et al, (○)67, Ota et al. ( )68 and Yang et.

Al(○)69.

Figure 3.10: Molar volume variation as the reaction time is increased for x = 10% and x =

15% samples. Note that as the reaction time increases, molar volumes gradually increase and reach the values observed by Bhosle et al.10 The blue band depicts the range of molar volumes

reported by several different groups66, 67 ,69.

Figure 3.11: Fragility index variation of x = 10% as reaction time is increased. Note that the

variation in m decreases (see inset) as the sample homogenize and finally reaches the value

observed for the homogeneous glass.

Figure 3.12: Cp* measurements of x=14.7% sample in GexSbxSe100-2x ternary system. Top set

of curves correspond to the in-phase Cp* and bottom set of curves correspond to out of phase

Cp* signal. Note that both signals shift to higher temperatures as the modulation frequency is

increased.

Figure 3.13: (a) Non reversing enthalpy variation in GexSbxSe100-2x ternary alloys. Flexible

phase extends from 0% < x < 14.9%, intermediate phase from 14.9% < x < 17.5% and stressed

51 rigid phase from x > 17.5% . (b) Fragility index variation in GexSbxSe100-2x ternary glasses. A

global minimum in m values was observed inside the intermediate phase.

Figure 3.14: FT-Raman profiling of x = 14.2% sample after (a) 24 hrs and (b) 120 hrs of

reacting starting materials. Sample was reacted at 9500C and was kept vertical inside a box

furnace.

Figure 3.15: (a) Tg, (b) ΔHnr and (c) ΔCp variation in GexSixTe100-2x ternary glasses after 7

days of reaction time () and 14 days of reaction time (▼).

Figure 3.16: Non-reversing enthalpy (ΔHnr) variation in GexSixTe100-2x. Rejuvenated data set

(▼) shows ΔHnr values obtained soon after Tg cycling while the red () data set shows ΔHnr

values obtained 2 months after sample syntheses.

Figure 3.17: (a) Cp* measurement of x = 6% sample of the GexSixTe100-2x ternary. In-phase Cp

shows a step at Tg while Out-of-Phase component shows an endothermic peak at Tg. Note that

both signals shift to higher temperatures as the modulation frequency is increased. (b) Fragility

x index values of the present ternary.

Figure 3.18: Molar volumes variation in GexSixTe100-2x glasses. Vm values display a global minimum in 7.5% < x < 9% followed by an abrupt decrease after x > 11.5%

119 Figure 3.19: Sn Mössbauer spectra of GexSixTe100-2x ternary glasses recorded at three select compositions. Spectra can be deconvoluted in to a singlet centered at 1.65 mm/s and a doublet centered at 3.33mm/s.

Figure 3.20: Integrated intensity variation in the singlet (T site) and the doublet (O site) of three select compositions in GexSixTe100-2x ternary glasses.

119 Figure 3.21: Sn Mössbauer spectra of (a) GexTe100-x and (b) SixTe100-x binary glasses at three select compositions. Spectra can be decovoluted in to two sites, a singlet centered at 1.65 mm/s and a doublet centered at 3.33 mm/s.

Figure 3.22: Scattering strength variation of “T” site and “O” site in GexTe100-x and SixTe100-x binary glasses.

Figure 4.1: FT-Raman profiling of Ge23Se77 glass after tR=24h and tR=216h (inset). h denotes the location of the sample column from which the Raman spectra was acquired. h = 1 indicates bottom of the sample tube and h = 9 indicates the very top of the sample column. Raman spectra was acquired at 2.5mm intervals along the length of the sample column.

Figure 4.2: Plot of length h along melt column vs. stoichiometry x, illustrating batch homogenization as Ge and Se are reacted at 9500C for 216h, resulting in h(x) becoming a vertical line at 23% (■) corresponding to the weighed composition.

88 Figure 4.3: (a) Fragility index variation in GexSe100-x glasses reported by Senapati et al. (□),

54 0 0 Stolen et al. (○) and the present study (). (b) Viscosity of GexSe100-x at 950 C and 2000 C calculated using VFT model. The blue panel extends from 19.5% < x < 26% which is the IP of

this system.

Figure 4.4 Schematic of melt homogenization process of present Ge-Se chalcogenides

showing (a) growth of homogeneous regions (dark blue) of well-defined melt stoichiometry

(x) (b) at the expense of interfacial regions (multicolored slabs). In a heterogeneous melt,

regions of varying stoichiometry, x1, x2, x3, occur, but upon homogenization, a unique melt

composition x1 persists across the batch composition.

Figure 4.5 Scattering strength ratios of ES/CS modes of GexSe100-x glasses reported by Sugai

et al73(○), compared with the present study (○).

Figure 4.6: Local structures of GexSbxSe100-2x and GexAsxSe100-2x systems. Ge atoms are

shown in gray, Sb in purple, As atoms in green and Se in yellow.

Figure 4.7: Approximate glass forming region92 in Ge-Si-Te system. Glassy samples for the

present system was synthesized along GexSixTe100-2x tie line in 6% < x < 15% range.

Figure 4.8: (a) ΔHnr variation in GexSixTe100-2x glasses. Rejuvenated (▼) data set was

obtained soon after Tg cycling as-quenched glasses. () data set shows samples aged for 2

months. (b) Molar volume variation in the present ternary. Note that a global minimum in

molar volumes is observed in the IP.

Figure 4.9: Fragility index variation (○) and ΔHnr variation as a function of x%. Note that a

global minimum fragility index is observed in the IP. Fragility index shows the melt behavior

of these materials above Tg while ΔHnr describes the behavior below Tg.

Figure 4.10: Constraint counting coupled with the location of the Intermediate Phase in

compositional space can be used to estimate the concentration of 2-fold coordinated Te atoms

in SixGexTe100-2x ternary glasses.

Figure 4.11: 119Sn Mossbauer spectra of (a) Sn in c-Si and (b) c-SnTe. Note that a resonance

xii is observed at δ=1.65mm/s when Sn is tetrahedrally coordinated and δ=3.33mm/s when Sn is octahedrally coordinated.

Table 4.1: Melting temperatures (Tm), solid densities (s) and liquid densities of Si, Ge and

Te.

xiii

GLOSSARY OF TERMS

T Temperture

Tl Liquidus temperature

Tg Glass transition temperature

ΔHnr Non-reversing enthalpy at Tg

Cp Heat capacity

ΔCp Change in heat capacity at from glass to metastable liquid state at Tg

Cp* Complex heat capacity

In-phase Cp Real part of complex heat capacity

Re(Cp*) Real part of complex heat capacity

Out-of-phase Cp Imaginary part of complex heat capacity

Im(Cp*) Imaginary part of complex heat capacity

η Viscosity

G∞ Infinite frequency shear modulus

τR Shear relaxation time

tR Reaction time for starting materials to react and produce a melt

m Fragility index

nc Number of constraints per atom

̅ Mean coordination number

δ Isomer shift measured in mm/s

Δ Quadrupole splitting in mm/s

Vm Molar volume

ρs Solid density

ρl Liquid density

PCM Phase Change Memory

xiv

mDSC Modulated differential scanning Calorimetry

υ Raman mode frequency in cm-1

CS Corner sharing tetrahedra

ES Edge sharing tetrahedra

CHAPTER 1: Introduction

1.1 Rigidity theory and elastic phases in network glasses

In late 70’s J.C Phillips and M. F. Thorpe1,2 introduced constraint counting for network

glasses. Their approach was simulated by the mathematic theory of structural rigidity proposed

by J. C. Maxwell3 for mechanical struts. Phillips and Thorpe extended this idea for network

glasses where bonds between atoms resemble mechanical struts, and developed a method to

study these glassy systems in terms of their network connectivity. They predicted that flexible

networks (nc < 3) will become stressed rigid (nc > 3) when the network connectivity increases

above an average coordination number of ̅ 2.4. Transition of networks from flexible to stressed-rigid is identified as the rigidity or the mean field transition. Flexible networks are under-constrained networks where the number of constraints per atom (nc) does not satisfy the

number of degrees of freedom per atom in 3-dimentional space. Stressed rigid networks on the

other hand are over-constrained where nc > 3. Glassy networks of ̅ 2.4 are viewed as optimally connected networks with optimal glass forming tendency.

In late 90’s experiments4-6 and theoretical simulations7,8 revealed that as the connectivity

of flexible networks increases, they form a self-organized phase before becoming stressed rigid.

This self-organized phase (Intermediate Phase) is considered to be optimally constrained and

glasses in this special phase are found to possess most unusual physical properties such as non-

aging, thermally reversing Tg’s etc. Transition of networks from flexible to intermediate is

identified as the rigidity transition and intermediate to stressed-rigid is identified as the stress

transition. Since its inception in late 90’s, Intermediate Phases have been observed in many

families of glasses such as chalcogenides9-12, oxides13,14, heavy metal oxides15, and solid electrolytes16.

1 1.2 Slow homogenization of chalcogenide melts

Chalcogenide glasses have been extensively studied over the past half a century. These materials find applications in PCM devices17, grating fabrication18,19,20, photo induced waveguides18,20,21, fiber fabrication22-30, fiber amplifiers and lasers31-34, optical switching35-37 etc.

However less emphasis has been put towards the physics of sample synthesis which is crucial if not the most important factor that needs to be addressed before performing any experiment.

Perusal of the literature on chalcogenide glasses reveal that a simple measurement of a physical property such as molar volumes of GexSe100-x glasses measured by different groups render different results which are much greater than the statistical error associated with the measurement10. This type of a behavior can be best attributed to sample synthesis and tools used to ascertain homogenization of glasses. How long does one needs to react starting materials?

What experimental tools can be used to monitor sample homogenization? How does sample homogenization affect physical properties of these glasses? These are very important issues that need to be properly addressed if one seeks to understand intrinsic behavior of these materials.

Slow homogenization of chalcogenide melts and an experimental tool that can be used to monitor sample homogenization was reported in an earlier study by Bhosle et al.10. In this study, physics and kinetics of sample homogenization will be discussed in detail. Microscopic origin of slow homogenization of chalcogenide melts can be traced to their low fragility indices which lead to high viscous melts that act as diffusion barriers. A unique experimental technique based on modulated differential scanning calorimetry in complex Cp (Cp*) mode was used to measure fragility indexes. Furthermore the role of turbulent mixing of glassy melts in sample homogenization will also be discussed in detail.

2 1.3 Rigidity Theory and Phase Change Materials

Phase change memory (PCM) technology gained much interest during the past decade as

a viable solution for the rapid increase in the demand for memory storage. This unique

technology, first proposed by S. Ovshinsky in 196817, is based on storing information on

crystalline and amorphous phases of a material. Phase change (PC) technology is widely used in

the optical data storage devices such as rewritable, readable-erasable disks, and recently

companies such as Intel, Micron technologies and Samsung are moving towards utilizing this

technology for non-volatile memory devices for commercial use.

Most common PC- material that has been used in commercial devices is Ge2Sb2Te5, which is also known as 225 or GST. The structure of the crystalline phase of this material is well understood38-42 but the amorphous phase remains debatable38,43-49. Comprehensive understanding

of the structural motifs of the amorphous phase that play a role in the phase switching

phenomena is crucial to improve performance of these devices. Experiments as well as

simulations have shown GST material to be highly stressed rigid50,51, and as a result, physical

aging of this material can be expected. Physical aging can be attributed to the retention loss of

the amorphous phase which is a key reliability issue in PCM devices. Physical properties such as

the resistivity and optical constants can drift with time due to the structural relaxation of the

glassy phase which stems from aging. This phenomenon will have adverse effects on the

performance of PCM devices.

Rigidity theory can be used as a tool to address both issues mentioned above; decoding

the structure and aging of the amorphous phase. Using Rigidity theory, MD simulations50 and

experiments51 have shown that a-GST has homopolar bonds (Ge-Ge and Sb-Sb) and Sb to be

3 largely 3-fold coordinated. Tellurium is believed to be both 3-fold and 2-fold coordinated. This

study couples elastic phases and 119Sn Mössbauer spectroscopy to elucidate structural

information on GexSixTe100-2x which is a phase change material system. Ideas developed here will most certainly aid in decoding structure of amorphous phase change materials.

Even though the aging of a-GST due to its stressed rigid nature cannot be remedied, identifying new PC-compositions inside Intermediate Phases may lead to PC materials that do not age. This study presents evidence of non-aging in a phase change system for the first time and suggests that IP’s are not only peculiar to Selenides where the 8-N rule is intact but also for

Tellurides where the 8-N rule is intrinsically broken.

1.4 Fragility and Elastic Phases in Glasses

The term “fragility” is widely used to classify viscous glass forming liquids based on temperature dependence of viscosity. Glassy melts with high fragilities are classified as fragile liquids and ones with low fragilities are known as strong liquids52. Fragile glass formers are characterized by the Vogel-Fulcher behavior in the so-called Angell plot which was first introduced by Laughlin and Uhlmann in 197253. Liquids that show an Arrhenius behavior in the

Angell plot are characterized as strong glass formers. Strong to fragile variation in these viscous

liquids is quantitatively characterized by a term called fragility index, m which is defined by the

following equation.

…………………………(1) lim→ /

Melt fragilities of glasses traditionally have been studied using viscosity

measurements53,54, beam bending technique53 and dielectric spectroscopy55. Viscosity

4 measurements are quite challenging for materials with poor glass forming tendency since large

quantities of the material, typically in the range of 15-20g, are needed. Large sample sizes are

also needed for beam bending technique and obtaining such homogeneous sample sizes can be

quite challenging56. It is difficult to perform Dielectric Spectroscopy for samples of high ionic

conductivity,

mDSC experiments performed in the complex Cp (Cp*) mode provides a very attractive

way of measuring melt fragilities55,57,58 without none of the above shortcomings. Experiments are

performed with constant scan rate and modulation amplitude. Modulation period is varied over a

finite range and the imaginary part of Cp* is analyzed to obtain a relaxation time. Appropriate

values of sample size, modulation frequency range, modulation amplitude and scan rate have to be carefully determined to minimize the phase lag due to the instrument59.

Study of fragility of glassy melts opens a door to observe glassy dynamics above T > Tg while elastic phases in glasses give a picture of what goes on in the glassy phase below T < Tg.

Are the constraints which are intact in the glassy phase, broken in the liquid phase?, or are they still intact? How does strong and fragile liquids correlate with the three elastic phases and what are the consequences? How do fragility index values affect slow homogenization of glassy melts?

1.5 Focus of the present work

In the present study, microscopic origin of slow homogenization of chalcogenide melts is discussed in detail for the first time. Experimental evidence on slow homogenization of 3 chalcogenide systems, GexSe100-x, GexSbxSe100-2x and GexSixTe100-2x is discussed in detail. The

60 first system, GexSe100-x is a classic chalcogenide alloy system which has been widely studied

5 and serves as a bench mark in both experimental and theoretical studies. Variation of physical

properties such as molar volumes, fragility index, elastic phase transitions etc. is due to

heterogeneity of melts/glasses, a problem that has inhibited basic understanding and functionality of glasses in terms of topology. This issue is addressed in the present work for the first time. The latter two systems are of special interest due to their applications in phase change memory devices.

The GexSbxSe100-2x system is the Selenium counter part of GexSbxTe100-2x system in which

x = 22.22% corresponds to the famous Ge2Sb2Te5, the chalcogenide alloy widely being used in

PC memory devices. Study of GexSbxSe100-2x will most certainly shed some light on the

amorphous phase of Ge2Sb2Te5. In the present study, strong correlations between the elastic

phases and melt fragilities of GexSbxSe100-2x glasses and evidence of self-organization effects in

the liquid phase are presented for the first time.

Finally, elastic phases of GexSixTe100-2x, which is a phase change ternary system are reported. This is the first time where non-aging of a telluride chalcogenide glassy system is presented, a result which has great significance on phase change memory devices. Physical aging of the memory element is believed to accentuate retention loss of the amorphous phase which is a key reliability issue in PCM devices. Correlations between melt fragilities, elastic phases and slow homogenization of melts are discussed in detail and local environments of Ge, Si and Te in the glassy phase are decoded using the elastic phases and 119Sn Mössbauer spectroscopy.

6 CHAPTER 2: Experimental Methods

This chapter is dedicated to introduce and discuss experimental techniques employed for

sample synthesis and characterization for the present study. Details on sample synthesis, mDSC

heat flow, Cp* and fragility measurements, density measurements and FT-Raman scattering

experiments will be discussed.

2.1 Synthesis of glasses

Stoichiometric amounts of the starting materials were weighed to an accuracy of 0.1mg

and sealed in evacuated quartz tubes (5mm ID and 7mm OD) under a pressure of 1x10-7 torr.

Quartz tubes were baked inside a vacuum oven at 100oC for 24hours prior sealing to get rid of

traces of water inside the tube walls. Sealed samples were then transferred to a box furnace.

Temperature was ramped up from room temperature to 950oC at a rate of ~38oC/hr and kept at

950oC until the samples were fully homogenized. Temperature of the box furnace was lowered to

800oC prior to water quenching to realize bulk glasses.

For GexSe100-x and GexSbxSe100-x systems, sample batch sizes were kept at 2g. However,

for GexSixTe100-2x glasses, sample batch sizes were limited to 0.5g. Due to their low glass

forming tendency, the quenching rate associated with water quenching was only capable of

delivering good glasses for batch sizes of 0.5g. For sample batches of higher weight (>1g) glassy

samples realized upon water quenching were found to be partially crystalline.

To ascertain sample homogeneity, sample tubes were periodically water quenched and

Raman profiled. Details of Raman profiling experiments will be discussed in detail in section

2.5. The process of quenching and Raman profiling were carried out until a fully homogeneous

7 glass was realized. Sample reaction time varied from 7days to 14days for the three different

systems under investigation for this stydy.

2.2 Heat flow measurements from MDSC

Total, reversing and non-reversing heat flow measurements were carried out using a TA

instrument DSC2920 calorimeter. mDSC unit was calibrated for temperature and enthalpy using

In and Pb standards and heat capacity using a Sapphire disk. Glassy samples of ~20mg in weight

were hermetically sealed in aluminum pans. These pans were vacuum heated at 100oC for a day

prior using. Heat flow measurements were carried out at a scan rate of 3oC/min, modulation of

amplitude of 1oC and a period of 100s. Figure 2.1 depicts a typical mDSC scan of glassy sample.

The green curve depicts the total heat flow, a typical DSC signal which is composed of a

step and an overshoot near Tg. The advantage of modulated-DSC stems from its ability to

separate out these two features of the total heat flow signal. The step like feature at Tg is called

the reversing heat flow since its related to the “thermally reversing” component of the total heat

flow. The “thermally non-reversing” component of the total heat flow signal is called the non-

reversing heat flow and is calculated by subtracting reversing heat flow from the total heat flow.

To compensate for scan rate related effects, a subsequent cooling cycle (left arrows in Fig. 2.1)

was carried out soon after the heating cycle (right arrows in Fig. 2.1)

Tg and the change in heat capacity at Tg (ΔCp) are extracted from the reversing heat flow

liquid glass signal. Tg is extracted from the inflection point and the step height gives ΔCp, Cp – Cp .

Average of these values obtained from the heating and cooling cycle gives scan rate independent

Tg’s and ΔCps which are truly intrinsic to the glassy sample under investigation. Non-reversing

enthalpy (ΔHnr) is obtained by calculating the difference in the integrated areas of the endotherm

8 at Tg in the heating cycle and the exotherm in the cooling cycle of the non-reversing heat flow

signal. ΔHnr contains vital kinetic information of the glass transition such as role of aging,

61 impurities, non-equilibrium nature of Tg . In contrast, the reversing heat flow contains

information which is vibrational in character related to thermodynamics.

GexSbxSe100-x, x = 13.2%

0 0.003 Total heat flow 174.08 C 0.03 cal/g/0C 0.62 cal/g endotherm Non-rev heat flow 0.000

exotherm Reversing heat flow 0.38 cal/g Heat Flow (cal/sec/g) Flow Heat 173.59 0C -0.003 0 0.03 cal/g/ C

100 150 200

Temperature (0C)

Figure 2.1: Example of a mDSC scan at x = 13.2% of GexSbxSe100-2x. Signals corresponding to

the heating cycle are indicated by right arrows and cooling cycle depicted by left arrows.

2.3 Complex Cp measurements from mDSC

mDSC experiments performed in the complex Cp (Cp*) mode provides a very attractive

way of measuring melt fragilities55,57,58. Melt fragilities of glasses traditionally have been studied

53,54 53 55 using viscosity measurements , beam bending technique and dielectric spectroscopy .

Viscosity measurements and the beam bending technique require large sample sizes, typically in the range of 15-20g and synthesis of homogeneous samples of such a large scale can prove quite challenging10,56,62. Even though Dielectric Spectroscopy is a very attractive method for fragility

9 index measurements; it’s not suitable for glassy samples of high ionic conductivity. Complex Cp measurements on the other hand require very small sample sizes, typically ~10mg and can be used to characterize any sample as long as it displays a clear glass transition.

Figure 2.2: Cp* measurement of GexSe100-x, x = 10% sample. Note that as the modulation

frequency is increased, the step in the In-Phase component and the endothermic peak of the Out- of-Phase component shifts to higher temperatures.

Cp* measurements were carried out using a TA instruments Q2000 calorimeter. Glassy

samples of approximately 10mg in weight were hermetically sealed in T-Zero aluminum pans

which were vacuum heated at 100oC for 24hrs prior encapsulation of samples to remove traces of

water. Q2000 calorimeter was calibrated for temperature and heat flow using In and Pb, and for

o heat capacity using a Sapphire disk. Cp* measurements were performed at a scan rate of 1 C/min and a modulation amplitude of 1oC. Modulation period was then varied from 40sec to 140sec at

20sec increments. To avoid aging effects on the measurements, encapsulated samples were

heated above Tg and mDSC signals were then recorded upon cooling at the specified scan rate.

The scan rate, sample size, modulation amplitude and modulation period were carefully selected

10 to minimize the instrument related phase lag caused by inefficient heat transfer between the

sample and the heater head59.

Figure 2.2 shows a Cp* scan of GexSe100-x 10% sample. The real part of Cp* (in-phase

heat capacity) shows a step neat Tg and the imaginary part of Cp* (out-of-phase heat capacity)

shows an endothermic peak near the glass transition. Both these features shift to higher

temperatures as the modulation frequency is increased. Modulation period (P) sets the enthalpy relaxation time (τ) according to the following relation, τ = P/2π. Endothermic peak location of the out-of-phase heat capacity signal gives the corresponding temperature where τ = P/2π

relation is satisfied and Tg is defined as the temperature at a modulation period of 100sec.

Fragility index is obtained by estimating the slope of log (τ) plotted against T/Tg. Figure

2.3 show fragility index measurements of select compositions in GexSe100-x glasses.

1.3 x = 26% 1.2 m = 19 Ea = 192kJ/mol x = 22%

 1.1 m = 15 s

 Ea = 139kJ/mol 1.0 Log( x = 10% 0.9 m = 26

Ea = 191kJ/mol 0.8

0.975 0.980 0.985 0.990 0.995 1.000 1.005 T /T g

Figure 2.3: Fragility index measurement of select compositions in GexSe100-x glasses. Slope of

the curve yields fragility index. High slope curves are characteristic of fragile glassy melts

whereas low slope curve are characterized as strong melts.

11 2.4 Molar volume measurements

Density of synthesized glassy samples was established by using the Archimedes principle. Samples were first weighed in air and then weighed again after immersing in alcohol to an accuracy of 0.1mg. Density of the alcohol was accurately measured by using c-Si and c-Ge standards using eq. 2.1. Molar volumes were calculated according to the equations listed below.

The error in density measurements were kept at a minimum by using samples that were more than 100mg in weight. This ensures that the error in molar volumes is less than 1% of the measured value. Molar volumes measurements were repeated three times for each glassy sample to check for reproducibility.

Density of Si – 2.329 gcm-3

Density of Ge – 5.323 gcm-3

Weight of sample in air - Wair ; Weight of sample immersed in alcohol - Walc

Density of alcohol - ρalc ; Density of sample - ρsample

Molar volume of sample - Vm

// / (2.1) /

(2.2)

2.2

12 GexSe1-x, x = 19%

tR = 6h 9 9

8 8

7 7

6 6

5 5

4 4

Intensity (arb. unit) 3 3

2 2

1 1

160 200 240 280 -1 Raman shift (cm )

Figure 2.4: Raman profiling of x = 19% glass in GexSe100-2x after 6h of reacting starting

materials. Raman spectra were acquired at 9 locations along the length of the sample column at

2.5mm intervals.

2.5 FT-Raman scattering

FT-Raman scattering experiments were carried out using a Thermo Nicolet Nexus870

FT-IR with a Raman module under a resolution of 2cm-1 and a laser spot size of 50/250μm.

Homogeneity of samples was periodically monitored real time during sample synthesis using a

novel FT-Raman profiling technique by acquiring Raman spectra at several different locations

along the length of the sample column as shown in figure 2.4. A sample was deemed

homogeneous when all the acquired Raman spectra along the length of the sample column were

found to be identical.

13 In Raman profiling experiments laser spot size plays an important role. It gives a sense of

the length scale at which the sample is homogeneous. If a spot size of 50μm is used, it suggests

that the sample is spatially homogeneous on a scale of 50μm. In other words, the average Raman

signature of all the structural motifs within an area of π(50μm)2 is identical to the Raman

signature of the batch composition. The uniqueness and power of this technique stems from the

fact that it is a non-invasive method of establishing sample homogeneity with very little or no

sample preparation and it can be done in real time as sample synthesis reaction progresses.

2.6 Mössbauer spectroscopy

119Sn Mössbauer spectroscopy is an attractive experimental tool to characterize glasses.

Ge and Si based glasses can be doped with traces of isotopically enriched Sn without observing

segregation effects. Since Ge, Si and Sn are isovalent, at very low concentrations Sn replaces

available Ge and Si sites and mimics their local chemistry. This is a powerful tool to probe

structure of glasses where traditional techniques such as Raman scattering cannot be used. For

119 the present study, Sn Mössbauer spectroscopic studies were conducted on GexSixTe100-2x glasses to probe local chemistry of Ge and Si. Raman scattering could not be performed on these glasses due to their low optical gap.

14 Chapter 3: Experimental Results

In this chapter, experimental results on fragility and signatures of slow homogenization

of GexSe100-x binary and GexSbxSe100-2x ternary glasses are presented. Latter part of the chapter is

dedicated to experimental results on the elastic phases, structure and slow homogenization of

GexSixTe100-2x glasses.

3.1 GexSe100-x system

3.1.1 mDSC Cp* measurements

Figure 3.1 shows Cp* measurements of three select compositions, x=10%, 24% and

33.33% in GexSe100-x binary glasses. Top set of curves on each panel show a step which

corresponds to the in-phase component of Cp*. Bottom set of curves which show an endothermic peak comes from the out-of-phase component of Cp*. Note that both the in-phase and out-of- phase curves shift to higher temperatures as the modulation frequency is increased.

As discussed in the previous chapter, fragility index was obtained by taking slope of the

curve log(τ) vs. Tg/T. (See figure 2.3). Figure 3.2a shows the fragility index (m) variation of

GexSe100-x in compositional space. m value of 26 was measured for x = 10% and as Ge

concentration was steadily increased, a decrease in fragility index values were observed until x =

22%. A global minimum m = 14.8(0.5) was observed for x = 22%. m values steadily increase

thereafter in 23%

which were obtained from viscosity measurements are plotted in figure 3.2a for comparison. m

values reported by Stolen et al. show a global minimum at x = 22%, a behavior similar to the

current work. However these values are much higher than the m values obtained from Cp* measurements in the present study.

15 Figure 3.1: In-phase and out-of-phase components of Cp* as a function of temperature plotted for different modulation periods for select compositions in GexSe100-x binary glasses. In-phase

Cp* shows a step at Tg while the out of phase component shows an endothermic peak. Note that both in phase and out of phase signals shift to higher temperatures as the modulation period is decreased.

16 Δ

Figure 3.2: (a) Fragility index of GexSe100-x glasses plotted as a function of x%. (○) shows m

54 values reported by Stolen et al. and () shows m values of the present study extracted from Cp*

10 measurements. (b) ΔHnr values of the present system reported by Bhosle et al.

Figure 3.2b shows the non-reversing enthalpy (ΔHnr) variation in the present binary

reported by Bhosle et al.56,62. The blue panel highlights the intermediate phase which extends from x = 19.5% to x = 26%. Flexible phase extends in 0%

phase in from 26%

flexible phase followed by a global minimum inside the intermediate phase. An increase in m

values is observed in the stressed rigid phase.

Fgure 3.3 shows the variation in the activation energy for enthalpy relaxation in the

present binary. These values are derived from fragility index values according to the equation

.. ln 10. Ea variation mimics the fragility index variation displaying a global

17 minimum at x = 22%. Ea values increase as one moves away from the global minimum at x =

22%. 400

350

300

250

(kJ/mol) a

E 200

150

10 15 20 25 30 35 x%

Figure 3.3: Activation energy values derived from Cp* measurements of GexSe100-x binary.

3.1.2 Slow Homogenization: FT-Raman measurements

Recently introduced FT-Raman profiling technique 10 was used to monitor sample

homogenization of GexSe100-x binary glasses. Homogenization profiles of two select

compositions, x=23% and x=21% are presented in this section.

Fig. 3.4 shows the evolution of Raman spectra of Ge23Se77. Sample was kept at a vertical

position inside a box furnace at 9500C, quenched at 24hr intervals and Raman profiled until the

melt was completely homogeneous. After 6hr of reaction time (Fig. 3.4a) crystalline modes of α-

63 GeSe2 can be seen . As the reaction progresses, the spread in the Se chain mode steadily

decreases and all spectra overlap indicating complete sample homogenization after 216h of

reaction time. Fig 3.5(a-d) shows evolution of Raman line shapes of Ge21Se79 for different reaction times. Sample tube was kept at a vertical position inside a box furnace. Note that it takes

18 Figure 3.4: In-situ Raman spectra obtained along the length of the sample column for Ge23Se77.

Sample tube was kept vertical inside a box furnace at 9500C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals. Note that the intensity spread in the Se chain mode progressively decreases as reaction time is increased

20 f.

Figure 3.6: In-situ Raman spectra obtained along the length of the sample column for Ge23Se77.

Sample tube was kept inside a rocking furnace at 9500C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals.

21 144h for all Raman spectra along the length of the sample column to coalesce and achieve full

sample homogenization.

Experimental data obtained from FT-Raman measurements indicated in Fig. 3.4 and

3.5(a-d) were cast into a more analytical form by identifying composition of each Raman spectra obtained along the length of the sample column (Fig. 3.7). This permits one to extract compositional variation of the melt along the length of the sample column. Identification of

Raman spectra was done by referring to a library of reference Raman spectra of GexSe100-x glasses reported by Bhosle et al.64

After 24h of reaction time the compositional variation of Ge23Se77 varies from ~16.5% to

~27.5% (Fig. 3.7a). After 48hrs, compositions at the top half of the tube reach x=21.5% while

the bottom half reach x=25%. Note that at the top of the tube the glassy compositions appear to

reach the nominal composition of x=23% much faster than the bottom of the sample column, but

does not completely homogenize until tR=216h.

In Ge21Se79 glass compositional variation of the batch varies from x~16% to x~25% after

24hs of reaction time,. The spread reduces to Δx=3% after 48hrs of reaction time. Compositions

corresponding to all 9 locations of the sample column reach x=21% after 144hrs of reaction time.

In the previous two experiments, sample tubes were kept vertical inside the box furnace.

To assess the role of rocking and turbulent mixing in homogenization of these melts, the

experiments were repeated using a rocking furnace. Sample tubes were placed inside a tube

furnace and continuously rocked at rate of 6 revolutions per minute. Samples were quenched at

24hr intervals and Raman profiled to assess sample homogeneity.

9 (a) 24h 48h 8 72h 96h 7 120h 144h 168h 6 192h 216h

h 5 4 3 2 23% tube kept vertical 1

16 17 18 19 20 21 22 23 24 25 26 27 28 x%

9 (b) 24h 8 48h 72h 7 96h 120h 6 144h

h 5 4 3 2 1 21% tube kept vertical

16 18 20 22 24 x%

Figure 3.7: Compositional variation of Raman spectra obtained along the length of the sample column

0 for (a) Ge23Se77 and (b) Ge21Se79. Sample tube was kept vertical inside a box furnace at 950 C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals.

23 9 24h 48h 8 72h 96h 7 120h 144h 168h 6 192h 216h

h 5 4 3 2 23% tube rocking (a) 1

16 17 18 19 20 21 22 23 24 25 26 27 28 x%

9 21% Rocking 24h 48h 8 96h 7 120h 144h 6

h 5 4 3 2 1 (b)

16 17 18 19 20 21 22 23 24 25 x% Figure 3.8: Compositional variation of Raman spectra obtained along the length of the sample column

0 for (a) Ge23Se77 and (b) Ge21Se79. Sample tube was kept inside a rocking furnace at 950 C and upon quenching Raman spectra were acquired at 9 locations which were spaced at 2.5mm intervals.

24 Fig. 3.6 shows the Raman spectra evolution of x = 23% sample which was kept inside a

rocking furnace. Sample was quenched at 24h intervals and Raman profiled. Note that the

intensity variation in the Se chain mode at 250cm-1is considerably less at the initial stages of the

experiments compared to the previous experiment where the sample was kept vertical (Fig. 3.4).

However all Raman spectra do not coalesce after reacting starting materials for 216hrs. Fig.

3.5(e-h) shows the evolution of Raman spectra of x=21% sample which was rocked during

synthesis. Similar to the x=23%, the spread in Se chain mode is less compared to fig. 3.5a where

the sample was kept vertical. However, complete sample homogenization is achieved after

reacting starting materials for 144h. Fig. 3.8 identifies the compositional variation along the

length of the sample column for x = 23% (Fig 3.8a) and x = 21% (Fig 3.8b) samples when the

tubes were rocked during synthesis.

3.1.3 Slow Homogenization: Molar Volumes and fragility index variation

19.0

18.8 Bhosle et al.

) Mahadevan et al.

-1 18.6

mol 18.4 3 18.2 (cm

mol 18.0 V

17.8 LucasYang etet al.al. Ota et al. 17.6 Avetikyan et al. Feltz et al.

10 15 20 25 30 35

X (%)

10 Figure 3.9: Molar volume variation in GexSe100-x binary glasses reported by Bhosle et al. () ,

Mahadevan et al.(Δ)65, Feltz et al.(□)66, Avetikyan et al, (○)67, Ota et al ( )68and Yang et al.(○)69.

25 Fig. 3.9 shows molar volume variation reported by several groups including data from

Bhosle et al10 on which the present study is based on. Molar volumes of Bhosle et al. are from

glassy samples which were homogenized on a scale of 50μm. It’s noteworthy to mention that

even though the general variation of molar volumes reported by other groups are similar with a broad global minimum in 19 %< x<26% range, the values are significantly lower than the ones reported in the present study.

18.9

6d 18.6 ) -1 4d 17d 18.3

mol 4d 3

(cm 18.0 2d mol

V 2d 17.7 1d

17.4 1d

10 15 20 25 30 35 x (%)

Figure 3.10: Molar volume variation as the reaction time is increased for x = 10% and x = 15%

samples. Note that as the reaction time increases, molar volumes gradually increase and reach the

values observed by Bhosle et al.10 The blue band depicts the range of molar volumes reported by

several different groups.66,67,69.

To assess the role of sample homogeneity on molar volumes, two select compositions

(x=15% and 10%) were reacted at different time intervals and the molar volume was measured.

26 After the sample was quenched, density of 5 glass pieces weighing about 100~200mg each was measured and the molar volume was calculated according to the method described in chapter 2.

One can observe that as the reaction time is increased, molar volumes gradually increase and

10 finally reach the values observed by Bhosle et al. for homogeneous GexSe100-x glasses (Fig.

3.10). It’s also quite fascinating to observe that the initial Vm spread gradually decrease as reaction time is increased.

Fig. 3.11 show evolution of fragility index of x = 10% sample as the reaction time is increased. After quenching, 3 glass pieces weighing about ~10mg was hermetically sealed in Al pans and fragility index was measured by carrying out complex Cp measurements as described in chapter 2. Note that initial variation in m values decrease and almost vanishes as the reaction time is systematically increased. After 6 days of reacting of starting materials, m values reach the value observed for the homogeneous glass.

35 x = 10% 25 20 15 m 2 10

30 V 5 0

25 01234567

tR (days)

Fragility Index (m) 20

15 123456 t (days) R

Figure 3.11: Fragility index variation of x = 10% as reaction time is increased. Note that the variation in m decreases (see inset) as the sample homogenize and finally reaches the value observed for the homogeneous glass.

27 3.2 GexSbxSe100-x system

3.2.1 mDSC Cp* measurements

Cp* measurements were performed on GexSbxSe100-2x glasses as described in chapter 2. Figure

3.12 shows in-phase (top set of curves) and out-phase Cp (bottom set of curves) signals observed

for the x=14.7% sample. Fragility index was then extracted using the analysis method described

in chapter 2.

Figure 3.12: Cp* measurements of x=14.7% sample in GexSbxSe100-2x ternary system. Top set of curves correspond to the in-phase Cp* and bottom set of curves correspond to out of phase

Cp* signal. Note that both signals shift to higher temperatures as the modulation frequency is increased.

51 Fig 3.13a. shows ΔHnr variation in GexSbxSe100-2x glasses reported in an earlier study .

ΔHnr values decrease as x is increased and show a global minimum in 14.9%

which is identified as the intermediate phase51. Flexible phase extends from 0% < x < 14.9% and

the stressed rigid phase beyond x > 17.5%.

28 Fig. 3.13b shows the fragility index variation as a function of x(%). m values steadily

decrease as the Ge concentration is increased and shows a broad local minimum centered at

x~10%. m values slightly increase in 10%

range followed by an abrupt decrease as Ge concentration is increased beyond x > 18%.

0.8 xc(1) xc(2) = 14.9% = 17.5% 0.6 aged 2 months IP Stressed rejuvenated 0.4 Rigid Hnr(cal/g) 

0.2 Flexible

(a)

(m) Index Fragility 24

20 (b)

3 6 9 12151821

Figure 3.13: (a) Non reversing enthalpy variation in GexSbxSe100-2x ternary alloys. Flexible

phase extends from 0% < x < 14.9%, intermediate phase from 14.9% < x < 17.5% and stressed

51 rigid phase from x > 17.5% . (b) Fragility index variation in GexSbxSe100-2x ternary glasses. A

global minimum in m values was observed inside the intermediate phase.

29 3.2.2 Slow Homogenization: FT-Raman measurements

FT- Raman profiling of x = 14.2% composition of GexSbxSe100-2x system is shown in

fig.3.14. First data set (Fig.3.14a) was acquired after 24 hrs of reaction of starting materials. All

the spectra are normalized to the highest intensity peak (Ge-Se CS mode at ~200cm-1). Note that

the spread in the Se chain mode at ~250cm-1 is quite prominent after 24hrs of reaction. Fig.

3.14b shows FT-Raman profiling acquired after 5 days of reacting starting materials. Note that

the variation in mode intensity of Se chain mode at ~250cm-1 vanishes after reacting starting materials for 5 days. FT-Raman measurements were performed with a laser spot size of 250µm for GexSbxSe100-2x ternary glasses.

Figure 3.14: FT-Raman profiling of x = 14.2% sample after (a) 24 hrs and (b) 120 hrs of

reacting starting materials. Sample was reacted at 9500C and was kept vertical inside a box

furnace.

30 3.3 GexSixTe100-2x system

3.3.1 mDSC heat flow measurements

Starting materials were reacted at 9500C for periods ranging from 7 to 14 days. Due to

the low glass forming tendency of this system, sample sizes were kept at 0.5g which proved quite useful to obtain good glasses. Fig. 3.15a summarizes Tg data of GexSixTe100-2x system after 7

days () and 14 days (▼) of reacting starting materials. Tg’s monotonically increase in 6 %<

x<12% range and decrease after x>12%. Slight increases in Tgs were observed when samples

were reacted for 14days. However the threshold where Tg’s start to decrease remained

unchanged with increased reaction time.

Fig. 3.15b shows the non-reversing enthalpy (ΔHnr) variation of GexSixTe100-2x glasses.

0 After sample synthesis, all samples were Tg cycled by heating above Tg at a scan rate of 3 C/min followed by a cooling down at the same scan rate. ΔHnr variation () shows a broad global minimum centered at x~9% after reacting starting materials for 7days followed by a steady increase until x = 12% and a reduction after x > 12%. Further reaction of the starting materials for an additional 7 days (▼) shifts the ΔHnr curve low x values and the global minimum becomes

almost square-well like. However the threshold for the reduction in ΔHnr values at x = 12%

remains unchanged. ΔCp variation (Fig. 3.15c) after 7 days and 14 days of reaction times show

almost a plateau like behavior in 6 %< x < 12% range followed by a decrease in the NSPS

regime.

Fig. 3.16 shows non-reversing enthalpy variation of GexSixTe100-x measured (a) soon after

Tg cycling and (b) after aging for 2 months. Tg cycling is performed soon after sample synthesis

to remove stress that may have frozen in due to the quenching process. In comparing these two

31 a. 200 7 days

180 14 days C) o ( g 160 T

140 NSPS

120 b.

0.4

(cal/g) 0.3 nr

H 11 

0.2

0.04 c.

0.03

0.02 (cal/g/sec) p C  0.01

0.00 6 8 10 12 14 x (%)

Figure 3.15: (a) Tg, (b) ΔHnr and (c) ΔCp variation in GexSixTe100-2x ternary glasses after 7 days of reaction time () and 14 days of reaction time (▼).

32 data sets one can observe a striking feature where the ΔHnr term remains unchanged as samples age in 7.5%

Aged 2 months

0.4

(cal/g) 0.3 Rejuvenated nr 

0.2

6 8 10 12 14

Figure 3.16: Non-reversing enthalpy (ΔHnr) variation in GexSixTe100-2x. Rejuvenated data set

(▼) shows ΔHnr values obtained soon after Tg cycling while the red () data set shows ΔHnr

values obtained 2 months after sample synthesis.

3.3.2 mDSC Cp* measurements

Figure 3.17a show the in-phase and out of phase Cp signals measured for x = 6% sample

of GexSixTe100-2x ternary for several modulation frequencies. In-phase Cp signal shows a step

near the glass transition while the out-of-phase Cp shows a characteristic endothermic peak near

Tg. Both the step and the endothermic peak shifts to higher temperatures as the frequency is increased. The location of the peak of im(Cp) and the modulation frequency was then used to extract fragility index of these glassy samples. (See chapter 2 for fragility index measurement

33 technique). Fig. 3.17b show the fragility index variation in GexSixTe100-2x glasses. m values

initially decrease as x is increased and shows a global minimum centered at x = 8.5% with a minimum value of m = 26.

Figure 3.17: (a) Cp* measurement of x = 6% sample of the GexSixTe100-2x ternary. In-phase Cp

shows a step at Tg while Out-of-Phase component shows an endothermic peak at Tg. Note that

both signals shift to higher temperatures as the modulation frequency is increased. (b) Fragility

index values of the present ternary.

34 3.3.3 Molar volume measurements

Fig. 3.18 show molar volume measurements of GexSixTe100-2x ternary glasses. Molar

volumes decease steadily in 6%

22.0 ) -1 )

-3 21.6 mol 3

21.2

20.8 Molar Volume (g/cm Volume Molar Molar Volume (cm

20.4

6 8 10 12 14 x%

Figure 3.18: Molar volumes variation in GexSixTe100-2x glasses. Vm values display a global minimum in 7.5% < x < 9% followed by an abrupt decrease after x > 11.5%

3.3.4 Mössbauer measurements

3.3.4.1 GexSixTe100-2x ternary glasses

Fig. 3.19 shows the Mössbauer effect observed in GexSixTe100-2x glasses. Each line shape

can be deconvoluted into a singlet (denoted by “T”) centered at around ~1.65 mm/s and a

doublet (denoted by “O”) centered at around ~3.33 mm/s. As the concentration of Ge and Si is

increased, asymmetry of the doublet at 3.3mm/s and the intensity increases progressively while

35 the intensity of the singlet decreases. Fig. 3.20 shows the integrated intensity variation of these two site quantitatively.

O Si6Ge6Te88

Si8Ge8Te84 O

Transmission (%) Si Ge Te 10 10 80

T O

-8 -6 -4 -2 0 2 4 6 8 Veolcity (mm/s) 119 Figure 3.19: Sn Mössbauer spectra of GexSixTe100-2x ternary glasses recorded at three select

compositions. Spectra can be deconvoluted in to a singlet centered at 1.65 mm/s and a doublet

centered at 3.33mm/s.

0.9 0.8 "T" site

0.7 0.6

0.5 0.4

0.3 Integrated Intensity Integrated 0.2 "O" site

0.1 567891011 x% Figure 3.20: Integrated intensity variation in the singlet (T site) and the doublet (O site) of three select compositions in GexSixTe100-2x ternary glasses.

36 3.1.4.2 GexTe100-x & SixTe100-x binary glasses

Fig. 3.21 shows Mössbauer effect observed in three select compositions of GexTe100-x &

SixTe100-x binary glasses. Similar to the above mentioned ternary glasses, line shapes can be

deconvoluted into a singlet centerted at ~0mm/s and a doublet centered at ~3.3mm/s.

Si Te Ge15Te85 14 86 O O T T

Ge Te O 17 83 Si16Te84 O T T

O Ge18Te82 Si Te Transmission (%) 18 82

T T O

(a) (b)

-8-6-4-202468-8-6-4-202468

Velocity (mm/s) Velocity (mm/s) Figure 3.21: 119Sn Mössbauer spectra of (a) Ge Te and (b) Si Te binary glasses at three x 100-x x 100-x select compositions. Spectra can be decovoluted in to two sites, a singlet centered at 1.65 mm/s and a doublet centered at 3.33 mm/s.

Fig. 3.22 shows the integrated intensity of the doublet and the singlet plotted as a

function of x%. As the concentration of the network former (Ge & Si) is increased in the Ge-Te

binary glasses, the intensity of the doublet increases progressively while the intensity of the

singlet decreases. However this effect is reversed in the Si-Te binary where the integrated

37 intensity of the doublet decreases while the singlet intensity increases as Si concentration is

increased.

0.65 GexTe100-x "T" Site 0.60

0.55

0.50

0.45 Integrated Intensity 0.40 "O" Site (a) 0.35 14 15 16 17 18 19 x% 0.9 SixTe100-x 0.8 "T" Site 0.7 0.6 0.5 0.4 0.3 Integrated Intensity 0.2 (b) "O" Site 0.1 13 14 15 16 17 18 19 x%

Figure 3.22: Scattering strength variation of “T” site and “O” site in GexTe100-x and SixTe100-x binary glasses.

38 Chapter 4: Discussion

The Ge-Se binary has been an attractive prototype of a chalcogenide as a test system for

probing predictions of Rigidity Theory. Elemental Se and Ge are available in a rather pure form

and can be reacted rather easily in evacuated quartz tubings to synthesize bulk glasses. They

have been studied over the past 50 years. Phillips1 was attracted by the optimization of the glass forming tendency70 near x = 16.6% , close to his prediction of x = 20% based for the glass

71- condition, nc = 3. Detailed Raman scattering measurements were undertaken by several groups

73 over the last 30 years and a general picture of glass structure based on Sen chains prevailing at

x = 0, systematically crosslinked by 4-fold Ge to form a nearly fully polymerized network near x

= 33.33% emerged. The broad picture was subsequently refined as (i) Raman signature of ES

tetrahedral unit was confirmed72,74 both in theory and experiments, (ii) existence of a chemical

threshold near x = 31.5% was established10,75 in Raman, MDSC experiments leading to a small

75 but reproducible broken chemical order of the stoichiometry glass GeSe2 from Raman ,

Mössbauer75, Neutron scattering76,77 measurements. In 2001, Boolchand et al.78 showed the

existence of two elastic phase transitions, one near x = 20% and a second one near 25% in these

glasses. These results showed for the first time the existence of a compositional window in which

networks with exceptionally different properties from those compositions outside the window.

The existence of the two elastic phase transitions was confirmed rather dramatically in the work of Bhosle et al10 more recently, who for the first time used FT-Raman profiling to generate

melts/glasses in the binary of unprecedented homogeneity and observed two rather abrupt

transitions in the non-reversing enthalpy of relaxation at Tg. Abrupt rigidity and stress transitions

observed in the homogenized samples highlight the percolative nature of these elastic phase

transitions as theoretically demonstrated by Jacobs et al.79 using the “pebble game” algorithm.

39 The order of magnitude reduction of the non-reversing heat flow at Tg ( ∆Hnr) at the rigidity transition near x = 19.5% , and the term remaining minuscule in the Intermediate Phase and then increasing abruptly near x = 26% at the stress transition is fully consistent with the modeling of the ∆Hnr term as representing an excess or deficiency of mechanical constraints, |nc-

3|, as suggested by Micoulaut61. In their simplistic modeling of the non-reversing heat flow,

Micoulaut showed that the overshoot of the glass transition endotherm would be high in flexible

and stressed-rigid glasses but minimal in isostatic networks. In these experiments, for the first

time, one had an experimental tool in FT-Raman profiling to systematically control the degree of

homogeneity of batches as a whole on a scale of 10 µm56. The crucial role of melt/glass

homogeneity in probing the physical properties of non-stoichiometric chalcogenides glasses

emerged from this work for the first time.

In the past 5 years new 77Se NMR experiments were brought to bear on the issue of

structure of the Ge-Se glasses and the liquids80,81. One of these groups developed a totally different picture of a diphasic model of these glasses80, while the other group espoused the

traditional random network structure model developed many years ago. The low natural

abundance of 77Se requires large batch sizes that are not easily homogenized as widely believed.

Given these diverse results, it became necessary to more clearly elucidate the crucial role of

sample synthesis, particularly their homogeneity and its bearing on the physical behavior of non-

stoichiometric glasses and liquids. The experiments undertaken in the present study were directed towards that goal.

4.1 Melt fragility and Slow Homogenization of Ge-Se melts/glasses

Even though chalcogenide glasses have been extensively studied for close to half a century, a dialogue on sample homogenization, unfortunately, has eluded many researchers. How

40 long does one react starting materials? What experimental tools can one use to assess melt/glass

homogeneity? and on what scale should one homogenize these materials? These are very

important and relevant key questions that need to be addressed before studying physical

properties or using these materials for applications.

New insights into how melts homogenize were made feasible from FT-Raman profiling

experiments10,56,62. If a sample is fully homogeneous at a certain length scale, one can expect to

see identical Raman scattering at any given location of the sample. Results for a GexSe100-x melt

weighed at x=23% after reacting the elements at tR=24h and 216h are shown in Fig.4.1. (also see

Fig.3.4 in chapter 3). For this study sample tube was kept vertical inside a box furnace for the

duration of synthesis. Note that after 216h of reaction Raman line shapes along the length of the

tube become indistinguishable indicating complete batch homogenization.

216h h=9 8 7

150 200 250 300 4 -1 Raman shift (cm ) 3 2 1

x = 23%

(arb.unit) Intensity 24h

100 150 200 250 300 -1 Raman shift (cm )

Figure 4.1: FT-Raman profiling of Ge23Se77 glass after tR=24h and tR=216h (inset). h denotes

the location of the sample column from which the Raman spectra was acquired. h = 1 indicates bottom of the sample tube and h = 9 indicates the very top of the sample column. Raman spectra were acquired at 2.5mm intervals along the length of the sample column.

41 In chapter 3, Fig.3.4a, we show Raman spectra variation after tR=6 hours of reaction time.

It’s quite fascinating to see the Raman signature of crystalline modes of α-GeSe2 appear in early

stages of the reaction as indicated by arrows. It appears that these crystalline islands that form

initially, disappear after tR=24hrs of reaction time indicating that these crystalline compounds dissolve in the melt within the first 24hrs of reaction time. Evolution of corner sharing(CS), edge sharing(ES) GeSe2 units and Sen chain modes at tR > 24hrs is a clear indication of formation of

the network backbone. The experimental data suggests that melt homogenization is a two-step

process. The first step of homogenization deals with forming the local structural units that form the network and progresses rapidly within the first 24hrs of reaction time. The second step is formation of the network backbone and characteristic intermediate range order as the system stoichiometry equalizes the batch composition. The second step progresses at a much lower rate and acts as the rate limiting step in melt homogenization.

Fig. 4.2(also Fig. 3.7a) maps the kinetics of melt stoichiometry evolution of Ge23Se77 batch as a function of reaction time. These results are plotted as h(x) vs. x%, where h(x) denotes the height of the melt column from the tube bottom, and x- the prevailing stoichiometry at that point. Perusal of Fig. 4.2 reveals that at 24h of reaction time, melts in the lower half of the quartz tube (h<5) are Ge richer than the top half (h>5). This is expected given that the density of liquid

Ge (5.6 g/cm-3) exceeds that of liquid Se (3.99 g/cm-3), and furthermore, at the reaction

temperature of 9500C, liquid Se (melting point of 2250C) rapidly flows along the melt column

while Ge (melting point 9370C) largely remains at the bottom of the tube. As the reaction

continues, Ge atoms diffuse up while Se atoms diffuse down the melt column (Fig. 4.2). We also

note that in the top half of the melt column diffusion seems to slow down qualitatively at tR>48h

once melt stoichiometry approaches low fragility index value zone (See fig. 4.3a for fragility

42 index variation in GexSe100-x glasses). In the bottom half of the tube on the other hand, tR>48h

diffusion is not directly hindered by low fragility index composition zone (shaded area in Fig.

4.3a) and the melt stoichiometry steadily reaches the expected weighed composition of x=23%.

A similar trend in homogenization was also observed at x=21% composition as well (Fig. 3.7b).

One can expect these melts to homogenize much quicker if they were to rocked to facilitate mixing of starting materials. To assess the role of rocking, the sample tubes were

reacted at 9500C inside a rocking furnace and Raman profiled at 24 hour intervals. The main

striking difference between a sample that is rocked and kept vertical is the initial compositional

spread (ascertained by Raman profiling) along the length of the sample column. When the sample is rocked, the initial composition spread seems to decrease quite significantly because

43 convective mixing offsets the gravitational barrier. In the x=23% unrocked sample,

compositional spread along the length of the tube was found to be Δx = 11% while a spread of

Δx = 1% was observed for a sample rocked. However the rocked sample did not homogenize

even after reacting for tR=216h. Sample kept vertical reached full homogenization after 216h of

reaction time. In Ge21Se79 glass, initial compositional spread after 24h of reaction time of the

sample kept vertical was found to be Δx=9%. Rocked sample showed a compositional variation

of Δx=1%. However unlike the Ge23Se77 glass, both samples homogenized after 144h of reaction

time.

Slow diffusion of these melts can be understood by looking at fragility index variation in

GexSe100-x glasses (Fig. 4.3a). Using mathematical relations such as Vogel-Fulcher-Tammann

(VFT)82-84, Eyring85 or MYEGA86 equations one can estimate the viscosity of these glass

forming liquids at high temperatures given their fragility indices. For the sake of illustration,

viscosity calculations are calculated using time honored VFT equation which is described below.

exp / ……………….(1)

Here <τ> denotes average relaxation time of a super cooled liquid, which stems from

52,87 structural and orientation rearrangements of atoms at a given temperature . Here , B and

T0 are fitting parameters. VFT equation can be rewritten in terms of fragility index as follows,

………………...... (2)

where 16, and represents the minimum fragility index, taken for SiO2 which is the lowest value reported to date. Average relaxation time τ can be expressed in terms of a

44 viscosity value according to the following Maxwell relation. Here denotes infinite

frequency shear modulus.

……………….(3)

0 0 Figure 4.3b shows the variationa in η of GexSe100-x melts at 950 C and 2000 C. It’s quite

striking to see the change in melt viscosity as the Ge concentration is steadily increased to

x=22%. Viscosity values display a global maximum, more prominent at 20000C, at x = 22% , the

composition where fragility index shows a global minimum. This suggests that strong glasses

characterized by low m values have very high viscosities at high temperatures than fragile ones.

At x=22%, m value of 14.8(0.5) was observed, which is slightly less than the m value of SiO2, which is considered to be the strongest glass forming melt. Thus, the viscosity of the super strong composition, x=22%, according to VFT is approximately 96.2 Pa.s at 9500C, which is about

three orders of magnitude greater than the viscosity at x = 33.3% composition, the highest m

value in the GexSe100-2x binary. The large spread in viscosities severely limits diffusion of the

super strong melt compositions residing, thus slowing down the melt homogenization globally.

Evolution of fragility index as a function of melt reaction time is shown in Fig. 3.11.

Note that for inhomogeneous samples, the variance in m across the entire batch is quite large, but as the melts homogenize, m values at different locations of the batch coalesce into a single value characteristic of a homogeneous melt as shown in Fig. 3.11. Fragility index values reported earlier by Senapati et al.88 and Stolen et al.54 obtained in viscosity measurements reveal much

greater values compared to the present study (Fig. 4.3a). This large shift in m values can be

attributed to melt inhomogeneity. In viscosity measurements much larger batch compositions are

required (50g) and synthesis of such melts would take much longer to homogenize than the 7

days needed to homogenize 2 gram sized batches used in the present study.

40 Senapati et al. (a) (viscosity) 35 Stolen et al. (viscosity) 30

20 Index (m) Fragility Present data 15 (mDSC) 3  

(b) C C

0 0 2

0 at 950  at 2000

 1

) Pa.S ) Pa.S   0 -1 log( log(

10 15 20 25 30 35 x%

88 Figure 4.3: (a) Fragility index variation in GexSe100-x glasses reported by Senapati et al. (□) ,

54 0 0 Stolen et al. (○) and the present study (). (b) Viscosity of GexSe100-x at 950 C and 2000 C

calculated using VFT model. The blue panel extends from 19.5% < x < 26% which is the IP of

this system.

The observed fragility minimum appears to be intimately tied to flexible and rigid

transitions and the intermediate phase in corresponding glasses (Fig. 3.2). An issue of central

importance is how homogeneous must melts/glasses be in such studies to observe the intrinsic

46 behavior of these thresholds? Compositional width of the percolative elastic phase transitions

(stress and rigidity) provides a convenient scale. An estimate of width comes from the

reversibility window wall width, which is estimated to be at Δ̅ 0.01. Here ̅ = 2(1-x), designates the mean coordination number of GexSe100-x network taking Ge and Se to be 4- and 2-

fold coordinated. In the condition Δ̅ 0.01, translates into a Ge stoichiometry variation Δx < ½

at %, and fixes a measure of batch homogeneity at a given x. For batches of size greater than 2

grams, more care is needed to homogenize them. The variance in physical properties of

chalcogenide glasses, such as for example, molar volumes of GexSe100-x glasses reported by

different groups56 are much too large to be statistical, and provides a direct measure of purity and

heterogeneity.

4.2 Melt heterogeneity and Interfacial regions

As shown in Fig 3.10, molar volumes experiments performed as a function of tR show them to steadily increase. Also note that in the early stages of reacting the starting materials, particularly at tR < 2days, the measured Vm are quite low, in fact lower than the broad range of

3 3 values in the 18.1 cm /mol < Vm < 18.6 cm /mol band that is characteristic of homogeneous

glasses (Fig.3.9). For this reason, one cannot merely view the heterogeneous glasses (tR >1day)

to be a superposition of homogeneous domains of varying stoichiometry xi in the 0 < xi < 33.3% range. There are regions in such heterogeneous melts/glasses that are quite compacted. At rather short reaction times, tR < 1day, it is indeed true that crystalline phases form. However, such

phases steadily disappear as melts are reacted longer for tR> 1day. These Vm data are suggestive

that heterogeneous melts may be viewed as composed of homogeneous regions of well-defined

stoichiometry “xi” that are separated by heterogeneous interfacial regions as schematically

47 illustrated in Fig.4.4. The well-defined homogeneous regions can be viewed as regions composed of characteristic local structures (Sen chain fragments, GeSe2 –Corner-sharing(CS)

and Edge-Sharing(ES) tetrahedral units) with well-developed extended range structures, such as

fraction of ES/CS fixed by stoichiometry xi alone, which give rise to the appropriate mode signature in Raman experiment. On the other hand, interfacial regions are viewed as regions that

connect homogeneous regions of varying stoichiometry. They are largely composed of the same

local structures as the homogeneous regions but could include Ge-rich local structures and

broken bonds, but with the important difference that extended range structures are not

developed. Interfacial regions can be viewed as regions which possess low molar volumes and

high fragility index, features that that can be associated with absence of extended range structures. As melts homogenize upon increased tR, homogeneous regions grow by

reconstructing with interfacial ones as schematically illustrated in Fig.4.4a and b, and the process

saturates as Vm increases (Fig.3.10) and m decreases (Fig. 3.11) to acquire values characteristic

of the completely homogeneous melts/glasses.

The present finding of an increase in Vm (Fig. 3.10) and a decrease in fragility m-index

(Fig. 3.11) of Ge-Se melts/glasses as these are steadily homogenized clearly demonstrates that

some of the earlier work on these glasses, particularly those that possess a low Vm(x) must come from specimen that are intrinsically heterogeneous by virtue of synthesis. A perusal of Fig.3.9

66 67 67 69 suggests that the results of Feltz et al. , Avtikyan et al. , Ota et al. and Yang et al. display Vm

3 (x) trends that largely reside in the 17.8 – 18.1 cm /mole range across a wide range of Ge

content. This range overlaps with values we observe in our present glasses that were reacted

typically for tR < 2d (Fig.3.10), which we know from Raman profiling data to be heterogeneous.

In the work of Yang et al69, the authors synthesized 20 to 25 gm batch compositions80 and

48 reacted the elements at 700° C for 12 h in a rocking furnace. These conditions of synthesis used

by Yang et al.69, we believe, has led to heterogeneous glasses. And the diphasic model89 of these

glasses proposed from 77Se NMR has substantial fraction of the signal coming from interfacial

regions rather than the homogeneous ones. In sharp contrast, the Vm(x) trends reported by

Mahadevan et al.65 that almost straggle the results of Bhosle et al. (Fig.3.10), are on glass samples that appear reasonably homogeneous.

Figure 4.4 Schematic of melt homogenization process of present Ge-Se chalcogenides showing

(a) growth of homogeneous regions (dark blue) of well-defined melt stoichiometry (x) (b) at the expense of interfacial regions (multicolored slabs). In a heterogeneous melt, regions of varying stoichiometry, x1, x2, x3, occur, but upon homogenization, a unique melt composition x1 persists

across the batch composition.

49 4.3 Manifestations of melt heterogeneity in optical measurements

Signatures of melt heterogeneity have manifestations not only in thermal calorimetric and

mechanical measurements but also in optical measurements as well. Fig.4.5 shows scattering

strength ratios of ES/CS of the present study compared to those of Sugai et al.73. Sugai et al.

performed Raman scattering experiments using 632.8nm laser excitation which is quite close to

the excitation laser wavelength (647nm) used in Raman scattering experiments of the present study. Despite using similar experimental conditions, noticeable disparity is seen in ES/CS ratios which we believe, most likely arises from the sample heterogeneity.

0.45

0.40 Sugai et al. 0.35 ` 0.30

0.25 I (ES) / I (CS)

0.20 Present study

10 15 20 25 30 35 x%

Figure 4.5 Scattering strength ratios of ES/CS modes of GexSe100-x glasses reported by Sugai et al.73(○), compared with the present study (○).

4.4 Correlating Melt Fragility and glass elastic phases in the GexSbxSe100-2x ternary

Study of GexSbxSe100-2x ternary not only will further our understanding in the physics of

network glasses but will also serve as a stepping stone to understanding phase change materials.

50 The most widely used Ge2Sb2Te5 is chemically isolavent to Ge2Sb2Se5 which resides in the

GexSbxSe100-2x ternary. One of the major challenges in studying the corresponding Tellurides is that the amorphous phase only exists only in a thin film or in flakes obtained by splat quenching.

The corresponding selenides on the other hand form bulk glasses. The selenides conform to the

8-N bonding rule, while Tellurides, on the other hand, display resonance bonding effects where the 8-N bonding rule is intrinsically broken87. These results of the Sb bearing ternary are of great interest for several reasons. (i) They permit showing how the IP window can be altered by Ge content. (ii) They serve as a bench mark for a ternary in which bonding is covalent and in harmony to the 8-N bonding rule and (iii) these data will be extremely useful when

corresponding Tellurides are examined in a thin-film configuration, then permitting one to

90 elucidate the role of charge transfer effects . Fig. 3.13 summarizes the ΔHnr data of GexSbxSe100-

91 2x ternary and the three elastic phases are identified . In this section, correlations between melt

fragilities, elastic phases and role on sample homogeneity of these ternary glasses will be

discussed.

As presented in Fig.3.13a, flexible phase in GexSbxSe100-2x extends from 0% < x < 14.9%

range and the stressed rigid phase extends from x > 17.5% range. Both these two phases are

characterized by the presence of physical aging in which an increase in ΔHnr was observed as

glasses age. The intermediate phase forms in a narrow band in the 14.9% < x < 17.5% range and

has a unique characteristic feature of non-aging of glasses where ΔHnr remains almost constant.

Fig. 3.13b shows fragility index variation in these ternary glasses. There are two quite striking

features in fragility index values. One is the global minimum of m values inside the intermediate

phase, a result that was also observed in the GexSe100-x binary and GexSixTe100-2x ternary. This

indicates that intermediate phase compositions are strong glass forming liquids. The second

51 Figure 4.6: Local structures of GexSbxSe100-2x and GexAsxSe100-2x systems. Ge atoms are shown

in gray, Sb in purple, As atoms in green and Se in yellow.

unusual feature in fragility index values is the broad local minimum centered at around x = 10%

which corresponds to ̅ = 2.3 given Ge is 4- coordinated, Sb and Se are 3- and 2- fold

coordinated. This may well be an indication of presence Sb-Se QT units in the melt which

suggests that self-organization may also manifest in the liquid phase as well. However when bulk

glasses are realized, experimental evidence does not support the presence of Sb-Se QT units

which indicates that these units are not energetically favorable at low temperatures. Fig. 4.6 further illustrates this train of thoughts.

52 9 In GexAsxSe100-2x, IP was observed in 2.28 < ̅ < 2.40 range . The onset of the IP at

̅ 2.28 stems from As-Se QT units proliferating in the network. These units are isostatically

rigid and thus form a stress free network. IP extends to lower value of ̅ 2.28 due to As-Se

pyramidal units, which are also isostatically rigid, proliferating in the network. But as As is

replaced by Sb, the IP onset shifts to ̅ 2.42. This is a clear indication that in the Sb bearing

ternary, Sb-Se QT units are not populated in the glass. Onset of the IP at ̅ 2.42 in this case is due to Ge-Se CS and Sb-Se pyramidal units forming the backbone of the network and extends until ̅ 2.52 due to the proliferation of Ge-Se ES units51.

Similar to GexSe100-x binary, low fragility indices also play a major role in melt

homogenization in GexSbxSe100-2x ternary as well, especially strong melts in the intermediate phase. Fig. 3.14 shows Raman profiling of a glass at x = 14.2%. Not that it takes 5 days to reach complete melt homogenization which is due to slow mixing of strong melts especially in 14.9%

< x < 17.5% range.

92 4.5 Melt-fragility, slow homogenization and elastic phases in the GexSixTe100-2x ternary

Extensive studies performed on Se and S based chalcogenide systems have revealed

experimental evidence for presence of the three elastic phases; flexible, intermediate and stressed

rigid. These results strongly suggest that constraint counting is an effective tool in understanding

physical properties10,93-95 of these inorganic polymers. These systems conform to the 8-N rule.

However this 8-N rule is intrinsically broken in corresponding Tellurides50 in which Te can

possess both 2- fold and 3- fold coordination in the glassy state50,96 unlike the Selenides which

only takes 2-fold coordination. Study of these Tellurides are not only of basic scientific interest

but also of technological interest due to their application in phase change memory devices in

53 Ge 0 35

5 30

10 25

15 20

20 GexSixTe100-2x 15

10 Approx. 30 Glass Forming 5 Region 35 0 Si 65 70 75 80 85 90 95 100 Te

Figure 4.7: Approximate glass forming region92 in Ge-Si-Te system. Glassy samples for the present system was synthesized along GexSixTe100-2x tie line in 6% < x < 15% range.

which select telluride compositions are being used as the active memory element. From a basic

scientific point of view, study of these Tellurides will most certainly further our understanding of

the glassy state of matter.

Fig. 4.7 shows the glass forming region of the Ge-Si-Te system. Glassy samples were

made along GexSixTe100-2x tie line in 6% < x < 15% range. It has been experimentally verified that the glass formation is optimized in ternary glassy systems which are composed of equimolar amounts of the network formers 9. For the present system, Ge and Si serve as network formers,

thus GexSixTe100-2x tie line was chosen for the present study.

54 4.5.1 Melt fragility and Elastic Phases of GexSixTe100-2x

Fig. 4.8 summarizes (a) ΔHnr and (b) molar volume variation in GexSixTe100-2x ternary glasses.

Note that as glasses age, ΔHnr values in 7.5% < x < 9% range remains almost constant displaying

minimal aging compared to compositions outside the said window. There is a very striking

correlation between ΔHnr and molar volume variation in this compositional region where global

minimum in ΔHnr coincides with a global minimum in molar volumes. The behavior is

characteristic of intermediate phases where compacted networks form, i.e., molar volumes show

a global minimum. Non-aging of glasses in this special phase stems from stress free character of

the network. Since an increase in Tgs are observed as the intermediate phase is approached, we

are led to believe that networks are flexible in 6% < x < 7.5% range of Si and Ge. These networks become stressed rigid as Ge and Si concentration x > 9%. The reduction in ΔHnr values at x > 12% and reduction in molar volumes at x > 11.5% are most probably manifestations of nanoscale phase separation in these ternary glasses.

The non-aging behavior of the ΔHnr term is a clear indication of stability of glassy compositions in 7.5% < x < 9% range. This result suggests that atoms change their configuration minimally over time thus maintaining physical properties such as resistivity and optical constants unchanged. This unique property of the intermediate phase has immense technological significance since these materials are being used in phase change memory devices.

Retention loss of the amorphous phase is a key reliability issue of PCM devices97 which stems from the instability of the amorphous phase as physical properties drift with time.

Existence of an IP in SixGexTe100-2x demonstrates that such phases are generic i.e. they are not peculiar not to Selenides and Sulfides, but extend to Tellurides as well and can be utilized in

PCM devices to stabilize device performance.

55 x% 6 8 10 12 14 16 2 Months x = 7.5% x = 9% Aged 0.4

(cal/g) 0.3 nr

 Rejuvenated

0.2 NSPS

22.0 ) -1 )

-3 21.6 mol 3

21.2

Flexible 20.8 Molar Volume (cm Molar Volume (g/cm Stressed 20.4 Rigid

6 8 10 12 14 16 x%

Figure 4.8: (a) ΔHnr variation in GexSixTe100-2x glasses. Rejuvenated (▼) data set was obtained soon after Tg cycling as-quenched glasses. () data set shows samples aged for 2 months. (b)

Molar volume variation in the present ternary. Note that a global minimum in molar volumes is observed in the IP.

56 36 Fragile

34 0.4 Melt

Fragile Strong 30 (cal/g) Glass 0.3 nr

 28

Fragility Index (m) 0.2 26 IP Flexible Stressed Rigid 24 5 6 7 8 9 10111213 x%

Figure 4.9: Fragility index variation (○) and ΔHnr variation as a function of x%. Note that a global minimum fragility index is observed in the IP. Fragility index shows the melt behavior of these materials above Tg while ΔHnr describes the behavior below Tg.

Fig. 4.9 compares the fragility index variation m(x) with the enthalpy of relaxation,

ΔHnr(x) in the present ternary. A global minimum in fragility index was observed in the

intermediate phase, underscoring that IP melt compositions are strong and possess a high

viscosity at elevated temperatures (T>Tg). This is the key contributing factor to slow homogenization of their melts and will be discussed in details in section 4.5.2.

4.5.2 Slow homogenization of GexSixTe100-2x

Figure 3.15 shows variation in Tg, ΔHnr and ΔCp values of GexSixTe100-2x glasses after 7

days and 14 days of reaction time. Comparison between 7 days and 14 days reveal striking

57 differences in physical properties of these glasses which indicates sample homogenization takes

place over 14 days of reacting the starting materials, i.e., melts homogenize rather slowly. The

behavior is due to the strong character of IP melts as revealed by the fragility plot of 3.17b.

In Fig. 3.15b, we note that ΔHnr values display a broad minimum centered near x ~8%

after 7 days of reacting starting materials. This was the first indication that an IP could be present in this ternary. We also suspected that the broad variation is a clear indication of sample heterogeneity. The suspicion was confirmed when we noted that a square-well like ΔHnr variation emerged10,98 in melts that were reacted for 14 days. Note that the walls of the window are sharper in Fig. 3.15b. Even though a sharp rigidity transition was not observed in rejuvenated samples, upon aging we found the transition to sharpen as illustrated in Fig. 3.16 where we compare the ΔHnr(x) variation in rejuvenated glasses with those aged for 2 months. This system is a good example of using elastic phase transitions to establish sample homogeneity. FT-Raman profiling technique10 could not be employed due to the low band gap of these materials.

Tg’s are found to increase upon reacting starting materials from 7 days to 14 days (Fig.

3.15a), indicating an increase of network connectivity. However, the chemical threshold where

Tg’s begin to decrease at x = 12% remained unchanged. Reduction in Tg’s at x > 12% is an

indication of a loss in network connectivity, which more likely is due to formation of Ge rich and

Si rich nano-phases in the glasses.

ΔCp values show rather a monotonic variation in compositional space (Fig. 3.15c). 7 days

and 14 days reacted glasses yielded ΔCp values that remained almost constant in 6% < x < 12%

range where homogeneous glasses form, but the ΔCp term starts to decrease at x > 12% when

nanoscale separation occurs.

58 Several factors can contribute to slow homogenization of these melts. One is the global

minimum in fragility index values inside the intermediate phase. Low m values act as diffusion

barriers for starting materials to mix, thus slowing homogenization. Another reason behind slow

homogenization could also be the large disparity between melting points of the starting materials

and their densities. Table 4.1 shows the melting points, liquid and solid densities of Ge, Si and

Te.

Si-Ge-Te samples were reacted at 950oC, a temperature which is considerably lower than

o o o o the melting point of Si, 1414 C. Initially, at 950 C, both Ge (Tm = 935 C) and Te (Tm = 449.5 C) will be molten but Si will remain a solid. Inside the quartz tube, Te melt will sink to the bottom and on top of it will sit the Ge melt. Finally Si will float on top of the Ge layer. Since Si is a solid, the rate at which Si gets incorporated into the melt will be slow opposed to Ge which is fully molten. If Si were fully molten, diffusivity of Si atoms will be much higher which will aid in speeding up the homogenization process. The slow diffusive process of Si into the glassy melt and the inherent slow homogenization character of chalcogenide melts, are the reasons behind long reaction times required to homogenize GexSixTe100-2x glasses.

Table 4.1: Melting temperatures (Tm), solid densities (s) and liquid densities of Si, Ge and Te.

Melting Solid Density -  Solid Density -  Element Temperature - T s l m (g/cm-3) (g/cm-3) (oC)

Si 1414 2.33 2.57

Ge 935 5.232 5.6

Te 449.5 6.24 5.7

59 4.5.3 Aspect of structure derived from the elastic phases and 119Sn Mössbauer experiments.

One major drawback of studying this system was the inaccessibility to probe structure using Raman scattering, a preferred technique used to characterize chalcogenide glasses9,10,99, due to the small (< 1.16eV) optical gap of these materials. However, Mössbauer spectroscopy proved to be quite a useful tool to probe the local environment Ge and Si while the location of the Intermediate Phase helped to infer the local environment of Te.

Both experiments96 and MD simulations50 have shown that Te is both 2-fold and 3-fold

coordinated. Unlike Se, the heavier Te does not always bond conforming to the 8-N rule, a

significant property for the phase change phenomena90. The concentration of 2-fold and 3 fold

Te can be inferred by the location of the Intermediate Phase in compositional space.

If all Te atoms are 2-fold coordinated, Ge and Si 4-fold coordinated, one can expect the mean field rigidity transition to occur at x = 10% as shown by the following calculation.

nc = 7x + 7x + 2(1-2x) = 3  x = 10% ……………..(2)

However, the Intermediate Phase of SixGexTe100-2x resides below x = 10% which suggests that not all Te atoms are 2-fold coordinated. If one assumes that a fraction (1-y) of Te atoms are

3-fold coordinated, then one can write nc to be made up of contributions from both 2-fild and 3- fold coordinated Te as follows.

nc = 7x + 7x + 2y*(1-2x) + 3.5*(1-y)*(1-2x) = 3 ...... (3) (Si) (Ge) (Te-2) (Te-3) Here, the concentration of Ge and Si is given by “x” and the concentration of 2-fold coordinated Te is given by “y”. Fig. 4.10 shows a graphical solution to equation (3) where all

possible solutions of nc = 3 lie on the solid curve. Highlighted region in Fig. 4.10 shows the

60 range of “y” values for the x values in the 7.5% < x < 9% range, the region where the

Intermediate Phase is observed. This simple calculation shows that approximately 90% of Te

atoms are 2 fold coordinated.

100

70 2-fold Te%

2-fold Te fraction (y) in % in % (y) 2-fold Te fraction 60

50 024681012

Fraction of Six% or Ge (x) in % Figure 4.10: Constraint counting coupled with the location of the Intermediate Phase in

compositional space can be used to estimate the concentration of 2-fold coordinated Te atoms in

SixGexTe100-2x ternary glasses.

Local environment of Ge and Si in the present system was probed using a novel

characterization method - 119Sn Mössbauer spectroscopy. One dopes traces (~1 wt%) of 119Sn by

reacting isotopically enriched elemental Sn with the glass of interest. In dilute amounts, Sn will

replace covalent Si and Ge atoms in the network, and mimic their local chemistry. In the present

system, 1mg of Sn was doped for 100mg of the glass.

61 Before discussing 119Sn Mössbauer experimental results of the present ternary, it is

useful to identify spectra of several standard compounds of well-defined valence in this

spectroscopy. Fig. 4.11 shows Mössbauer spectra of the two well defined chemical states of Sn

that would be of interest in the present work. When Sn is in an sp3 bonded state as a dilute

substituent in c-Si, one can expect a Mössbauer resonance with unique chemical shift of

4+ δ=1.68(±0.02)mm/s (with respect to Sn as in BaSnO3). The shift is characteristic of Sn

tetrahedrally coordinated as shown in the inset (Fig. 4.11a). In c-SnTe, which represents the Sn2+ oxidation, Mössbauer spectroscopy reveals single resonance with an isomer shift of

δ=3.33(±0.02)mm/s. In this type of a chemical environment, Sn is octahedrally coordinated as show in the inset (Fig. 4.11b), and for that reason no quadrupole splitting is observed.

Fig. 3.19 shows Mössbauer spectra of 3 select compositions in the GexSixTe100-2x

ternary glasses. Note that the spectra can be deconvoluted into two sites; a singlet with a chemical shift of δ~1.60(±0.02)mm/s and a doublet with a chemical shift of δ~3.33(±0.02)mm/s.

The singlet with δ~1.6(±0.02)mm/s is a clear indication of the presence of tetrahedrally coordinated Sn in the network. This result is not surprising since one expects to have 4- fold coordinated Ge and Si sites in the network as seen in the corresponding selenides4,56,62.

Mössbauer resonance centered at δ~3.33(±0.02)mm/s highlights presence of octahedrally coordinated Sn in the glassy network. If Sn were to form a perfect octahedron with its near neighbors, one can expect to see a resonance similar to the one show in Fig.4.11b since the electric field gradient for such a local coordination vanishes. Departure from this ideal geometry indicates that Sn is in a defected octahedral configuration. Figure 3.20 shows the integrated intensity of the tetrahedral and octahedral Sn sites as a function of x. Note that the defected- octahedral site grows rapidly as the concentration of Ge and Si increase. However this result

62 T Sn in c-Si 3 Sn-SP Transmission (%) Transmission

(a) δ = 1.65mm/s

-8 -6 -4 -2 0 2 4 6 8 Velocity(mm/s)

c-SnTe O 2+ Sn Transmission (%) Transmission

(b) δ = 3.33mm/s

-8-6-4-202468 Velocity(mm/s) Figure 4.11: 119Sn Mössbauer spectra of (a) Sn in c-Si and (b) c-SnTe. Note that a resonance is observed at δ=1.65(±0.02)mm/s when Sn is tetrahedrally coordinated and δ=3.33(±0.02)mm/s when Sn is octahedrally coordinated.

does not permit one to separate the individual contribution of Ge and Si sites to the observed

119 spectra. To address this issue, Sn Mössbauer experiments were performed on binary SixTe100-x

and GexTe100-x glasses at few select compositions. Figure 3.21 shows Mössbauer results of binary

SixTe100-x and GexTe100-x glasses. Similar to the case of ternary glasses, the two sites

63 corresponding to tetrahedral and octahedral Sn are also observed in both binary glass systems. If one quantifies the site integrated intensities (Fig. 3.22), one finds that the two binary systems

display quite different behavior; the defected-octahedral site in the Ge-Te binary grows rapidly

as Ge concentration, x exceeds 17%, while the Si-Te binary the defected-octahedral intensity

steadily decreases with increasing x (Fig. 3.22b). The data suggest that Si largely goes in

tetrahedrally coordinated, while Ge prefers forming the defected-octahedral configurations and is

the main contributor to forming defected-octahedral configurations in GexSixTe100-2x ternary glasses. This result further suggests that in the GexSixTe100-2x ternary, Ge atoms are more

octahedrally coordinated than Si atoms, and majority of Si atoms are tetrahedrally coordinated.

64 Chapter 5: Conclusions

The comprehensive examination of the physical properties of the three chacogenide glass

systems in the present thesis, leads us to the following conclusions.

(1) A strong correlation is observed between melt fragility index (m) and the three elastic

phases of chalcogenide glasses. Specifically, minima in melt fragility (m) occur in Intermediate

phase (IP) glasses that display a global minimum in the non-reversing enthalpy of relaxation

(∆Hnr) at Tg. The correlation is observed in all the three chalcogenides systems GexSe100-x,

GexSbxSe100-2x, and GexSixTe100-2x investigated.

(2) The lower fragility (m < 20) or stronger nature of melts in IP compositions compared to

those compositions outside the IP, serves as a bottle neck for melt-mixing during synthesis at

elevated temperatures, and is primarily responsible for slow homogenization of melts/glasses.

(3) Melt fragility and glass molar volumes are found to systematically change as

heterogeneity of non-stoichiometric melts/glasses is steadily lowered by controlled synthesis.

These observations underscore the importance of homogenizing melts completely to establish the

intrinsic physical behavior of non-stoichiometric glasses.

(4) The coincidence in composition of glass reversibility windows with melt fragility

windows observed in each of three chalcogenides systems investigated strongly suggests that the

notion of self-organization of networks used to describe glasses must extend to corresponding

melts.

(5) Intermediate phase in GexSixTe100-2x ternary, a phase change system, is observed in

7.5%

9%12%.

65 (6) Retention loss of the amorphous phase in phase change memory devices can be addressed

by resorting to intermediate phase compositions where physical aging is minimal, thus improving stability of physical properties such as optical constants and resistivity over time.

(7) Majority of Te atoms (~90%) are found to be 2 fold coordinated. Tetrahedral and

defected octahedral sites of Ge and Si exist in GexSixTe100-2x ternary glasses and majority of defected octahedral sites are Ge based while Si is predominantly tetrahedrally coordinated.

66 BIBLIOGRAPHY [1] Phillips, J. C., Topology of covalent non-crystalline solids I: Short-range order in

chalcogenide alloys, J. of Non-Cryst. Solids 1979, 34, 153-181.

[2] Thorpe, M. F., Continuous Deformations in Random Networks, J. of Non-Cryst. Solids

1983, 57, 355-370.

[3] Maxwell, J. C., On the calculation of the equilibrium and stiffness of frames, Philos.

Mag. 1864, 27, 294-299.

[4] Selvanathan, D.; Bresser, W. J.; Boolchand, P.; Goodman, B., Thermally reversing

window and stiffness transitions in chalcogenide glasses, Solid State Commun. 1999,

111, 619-624.

[5] Boolchand, P.; Lucovsky, G.; Phillips, J. C.; Thorpe, M. F., Self-organization and the

physics of glassy networks, Philosophical Magazine 2005, 85, 3823-3838.

[6] Boolchand, P.; Georgiev, D. G.; Goodman, B., Discovery of the intermediate phase in

chalcogenide glasses, Journal of Optoelectronics and Advanced Materials 2001, 3, 703-

720.

[7] Thorpe, M. F.; Jacobs, D. J.; Chubynsky, M. V.; Phillips, J. C., Self-organization in

network glasses, J. of Non-Cryst. Solids 2000, 266, 859-866.

[8] Thorpe, M. F.; Phillips, J. C., Phase Transitions and Self-organization in Electronic and

Molecular Networks Kluwer Academic/Plenum Publishers, Cambridge, UK, 2001,

Fundamental Materials Research.

[9] Qu, T.; Georgiev, D. G.; Boolchand, P.; Micoulaut, M., in Supercooled Liquids, Glass

Transition and Bulk Metallic Glasses, edited by T. Egami et al. (Materials Research

Society, 2003), p. 157.

67 [10] Bhosle, S.; Gunasekera, K.; Chen, P.; Boolchand, P.; Micoulaut, M.; Massabrio, C.,

Meeting experimental challenges to physics of network glasses: assessing the role of

sample homogeneity, Solid State Commun. 2011, 151, 1851-1855.

[11] Georgiev, D. G.; Boolchand, P.; Eckert, H.; Micoulaut, M.; Jackson, K., The self-

organized phase of bulk PxSe1-x glasses, Europhysics Lett. 2003, 62, 49-55.

[12] Georgiev, D. G.; Boolchand, P.; Micoulaut, M., Rigidity transitions and molecular

structure of AsxSe1-x glasses, Phys. Rev. B 2000, 62, R9228-R9231.

[13] Rompicharla, K.; Novita, D. I.; Chen, P.; Boolchand, P.; Micoulaut, M.; Huff, W., Abrupt

boundaries of intermediate phases and space filling in oxide glasses, Journal of Physics:

Condensed Matter 2008, 20, 202101.

[14] Vignarooban, K.; Boolchand, P.; Micoulaut, M.; Malki, M., in APS March

MeetingBoston, MA, 2012), p. http://meetings.aps.org/Meeting/MAR12/Event/167434.

[15] Chakraborty, S., in APS March Meeting (American Physical Society, Dallas, TX, 2011),

p. http://meetings.aps.org/link/BAPS.2011.MAR.Y2031.2018.

[16] Novita, D. I.; Boolchand, P., Synthesis and structural characterization of dry AgPO3

glass by Raman scattering, infrared reflectance, and modulated differential scanning

calorimetry, Phys. Rev. B 2007, 76, 184205-184212.

[17] Ovshinsky, S. R., Reversible Electrical Switching Phenomena in Disordered Structures,

Phys. Rev. Lett. 1968, 21, 1450–1453.

[18] Ramachandran, S.; Bishop, S. G., Low loss photoinduced waveguides in rapid thermally

annealed films of chalcogenide glasses, Applied Physics Letters 1999, 74, 13-15.

[19] Slinger, C. W.; Zakery, A.; Ewen, P. J. S.; Owen, A. E., Photodoped chalcogenides as

potential infrared holographic media, Appl. Opt. 1992, 31, 2490-2498.

68 [20] Saliminia, A.; Villeneuve, A.; Galstyan, T. V.; LaRochelle, S.; Richardson, K., First- and

Second-Order Bragg Gratings in Single-Mode Planar Waveguides of Chalcogenide

Glasses, J. Lightwave Technol. 1999, 17, 837.

[21] Youden, K. E.; Grevatt, T.; Eason, R. W.; Rutt, H. N.; Deol, R. S.; Wylangowski, G.,

Pulsed laser deposition of Ga-La-S chalcogenide glass thin film optical waveguides,

Applied Physics Letters 1993, 63, 1601-1603.

[22] Quimby, R. S.; Gahagan, K. T.; Aitken, B. G.; Newhouse, M. A., Self-calibrating

quantum efficiency measurement technique and application to Pr3+-doped sulfide glass,

Opt. Lett. 1995, 20, 2021-2023.

[23] Wei, K.; Machewirth, D. P.; Wenzel, J.; Snitzer, E.; Sigel, J. G. H., Spectroscopy of

Dy3+ in Ge-Ga-S glass and its suitability for 1.3-µm fiber-optical amplifier applications,

Opt. Lett. 1994, 19, 904-906.

[24] Simons, D. R.; Faber, A. J.; de Waal, H., GeSx glass for Pr3+-doped fiber amplifiers at

1.3 μm, J. of Non-Cryst. Solids 1995, 185, 283-288.

[25] Ohishi, Y.; Mori, A.; Kanamori, T.; Fujiura, K.; Sudo, S., Fabrication of praseodymium-

doped arsenic sulfide chalcogenide fiber for 1.3-mu m fiber amplifiers, Applied Physics

Letters 1994, 65, 13-15.

[26] Marchese, D.; Jha, A., The structural aspects of the solubility of Pr3+ ions in GeS2-

based glasses, J. of Non-Cryst. Solids 1997, 213–214, 381-387.

[27] Mashinsky, V. M.; Neustruev, V. B.; Dvoyrin, V. V.; Vasiliev, S. A.; Medvedkov, O. I.;

Bufetov, I. A.; Shubin, A. V.; Dianov, E. M.; Guryanov, A. N.; Khopin, V. F.; Salgansky,

M. Y., Germania-glass-core silica-glass-cladding modified chemical-vapor deposition

69 optical fibers: optical losses, photorefractivity, and Raman amplification, Optics Letters

2004, 29, 2596-2598.

[28] Sanghera, J. S.; Aggarwal, I. D., Active and passive chalcogenide glass optical fibers for

IR applications: a review, J. of Non-Cryst. Solids 1999, 257, 6-16.

[29] Mossadegh, R.; Sanghera, J. S.; Schaafsma, D.; Cole, B. J.; Nguyen, V. Q.; Miklos, R.

E.; Aggarwal, I. D., Fabrication of Single-Mode Chalcogenide Optical Fiber, J.

Lightwave Technol. 1998, 16, 214.

[30] Brady, D. J.; Schweizer, T.; Wang, J.; Hewak, D. W., Minimum loss predictions and

measurements in gallium lanthanum sulphide based glasses and fibre, J. of Non-Cryst.

Solids 1998, 242, 92-98.

[31] Mori, A.; Ohishi, Y.; Kanamori, T.; Sudo, S., Optical amplification with neodymium-

doped chalcogenide glass fiber, Applied Physics Letters 1997, 70, 1230-1232.

[32] Cole, B.; Shaw, L. B.; Pureza, P. C.; Mossadegh, R.; Sanghera, J. S.; Aggarwal, I. D.,

Rare-earth doped selenide glasses and fibers for active applications in the near and mid-

IR, J. of Non-Cryst. Solids 1999, 256–257, 253-259.

[33] Ye, C. C.; Hewak, D. W.; Hempstead, M.; Samson, B. N.; Payne, D. N., Spectral

properties of Er3+-doped gallium lanthanum sulphide glass, J. of Non-Cryst. Solids

1996, 208, 56-63.

[34] Schweizer, T.; Samson, B. N.; Moore, R. C.; Hewak, D. W.; Payne, D. N., Rare-earth

doped chalcogenide glass fibre laser, Electronics Letters 1997, 33, 414-416.

[35] Mizrahi, V.; DeLong, K. W.; Stegeman, G. I.; Saifi, M. A.; Andrejco, M. J., Two-photon

absorption as a limitation to all-optical switching, Opt. Lett. 1989, 14, 1140-1142.

70 [36] Snyder, A. W.; Mitchell, D. J.; Poladian, L.; Ladouceur, F., Self-induced optical fibers:

spatial solitary waves, Opt. Lett. 1991, 16, 21-23.

[37] Samoc, M.; Samoc, A.; Luther-Davis, B.; Woodruff, M., The concept of guiding light

with light and negative third-order optical nonlinearities of organics, Pure and Applied

Optics: Journal of the European Optical Society Part A 1996, 5, 681.

[38] Kolobov, A. V.; Fons, P.; Frenkel, A. I.; Ankudinov, A. L.; Tominaga, J.; Uruga, T.,

Understanding the phase-change mechanism of rewritable optical media, Nature

Materials 2004, 3, 703–708.

[39] Welnic, W.; Pamungkas, A.; Detemple, R.; Steimer, C.; Blugel, S.; Wuttig, M.,

Unravelling the interplay of local structure and physical properties in phase-change

materials, Nat Mater 2006, 5, 56-62.

[40] Sun, Z.; Zhou, J.; Ahuja, R., Structure of Phase Change Materials for Data Storage,

Phys. Rev. Lett. 2006, 96, 055507.

[41] Kooi, B. J.; Groot, W. M. G.; De Hosson, J. T. M., In situ transmission electron

microscopy study of the crystallization of Ge2Sb2Te5, Journal of Applied Physics 2004,

95, 924-932.

[42] Eom, J.-H.; Yoon, Y.-G.; Park, C.; Lee, H.; Im, J.; Suh, D.-S.; Noh, J.-S.; Khang, Y.;

Ihm, J., Global and local structures of the Ge-Sb-Te ternary alloy system for a phase-

change memory device, Phys. Rev. B 2006, 73, 214202.

[43] Jovari, P.; Kaban, I.; Steiner, J.; Beuneu, B.; Schops, A.; Webb, M. A., 'Wrong bonds' in

sputtered amorphous Ge2Sb2Te5, Journal of Physics: Condensed Matter 2007, 19,

335212.

71 [44] Baker, D. A.; Paesler, M. A.; Lucovsky, G.; Agarwal, S. C.; Taylor, P. C., Application of

Bond Constraint Theory to the Switchable Optical Memory Material Ge2Sb2Te5, Phys.

Rev. Lett. 2006, 96, 255501-255503.

[45] Jovari, P.; Kaban, I.; Steiner, J.; Beuneu, B.; Schops, A.; Webb, M. A., Local order in

amorphous Ge2Sb2Te5 and GeSb2Te4, Phys. Rev. B 2008, 77, 035202.

[46] Caravati, S.; Bernasconi, M.; Kuhne, T. D.; Krack, M.; Parrinello, M., Coexistence of

tetrahedral- and octahedral-like sites in amorphous phase change materials, Applied

Physics Letters 2007, 91, 171906-171903.

[47] Akola, J.; Jones, R. O., Structural phase transitions on the nanoscale: The crucial pattern

in the phase-change materials Ge2Sb2Te5 and GeTe, Phys. Rev. B 2007, 76, 235201.

[48] Hegedus, J.; Elliott, S. R., Microscopic origin of the fast crystallization ability of Ge-Sb-

Te phase-change memory materials, Nat Mater 2008, 7, 399-405.

[49] Kohara, S.; Kato, K.; Kimura, S.; Tanaka, H.; Usuki, T.; Suzuya, K.; Tanaka, H.;

Moritomo, Y.; Matsunaga, T.; Yamada, N.; Tanaka, Y.; Suematsu, H.; Takata, M.,

Structural basis for the fast phase change of Ge2Sb2Te5: Ring statistics analogy between

the crystal and amorphous states, Applied Physics Letters 2006, 89, 201910-201913.

[50] Micoulaut, M.; Otjacques, C.; Raty, J.-Y.; Bichara, C., Understanding phase-change

materials from the viewpoint of Maxwell rigidity, Phys. Rev. B 2010, 81, 1-11.

[51] Gunasekera, K., Intermediate Phase, Molecular structure, Aging and Network Topology

of Ternary GexSbxSe100-2x Glasses, Master of Science, University of Cincinnati,

Cincinnati 2010.

[52] Angell, C. A., Relaxation in liquids, polymers and plastic crystals — strong/fragile

patterns and problems, J. of Non-Cryst. Solids 1991, 131–133, Part 1, 13-31.

72 [53] Laughlin, W. T.; Uhlmann, D. R., Viscous flow in simple organic liquids, J. of Phys.

Chem. 1972, 76, 2317-2325.

[54] Stolen, S.; Grande, T.; Johnsen, H.-B., Fragility transition in GeSe2-Se liquids, Physical

Chemistry Chemical Physics 2002, 4, 3396-3399.

[55] Carpentier, L.; Desprez, S.; Descamps, M., From strong to fragile glass- forming

systems: a temperature modulated differential scanning calorimetry investigation, Phase

Transitions 2003, 76, 787 - 799.

[56] Bhosle, S.; Gunasekera, K.; Boolchand, P., Melt Homogenization and Self Organization

in Chalcogenide Glasses - Part 2, Intl. J. App. Glass. Sci. 2012, 3, 205-220.

[57] Fukawa, Y.; Matsuda, Y.; Kawashima, M.; Kojima, S., Determination of complex-

specific heat and fragility of sodium borate glasses by temperature-modulated DSC,

Journal of Thermal Analysis and Calorimetry 2010, 99, 39-44.

[58] Matsuda, Y.; Fukawa, Y.; Matsui, C.; Ike, Y.; Kodama, M.; Kojima, S., Calorimetric

study of the glass transition dynamics in lithium borate glasses over a wide composition

range by modulated DSC, Fluid Phase Equilibria 2007, 256, 127-131.

[59] Weyer, S.; Hensel, A.; Schick, C., Phase angle correction for TMDSC in the glass-

transition region, Thermochimica Acta 1997, 304–305, 267-275.

[60] Micoulaut, M.; Kachmar, A.; Bauchy, M.; Le Roux, S.; Massobrio, C.; Boero, M.,

Structure, topology, rings, and vibrational and electronic properties of Ge_{x}Se_{1−x}

glasses across the rigidity transition: A numerical study, Phys. Rev. B 2013, 88, 054203.

[61] Micoulaut, M., Linking rigidity with enthalpic changes at the glass transition and the

fragility of glass-forming liquids: insight from a simple oscillator model, J. Phys.:

Condens. Matter 2010, 22, 1-7.

73 [62] Bhosle, S.; K.Gunasekera; Boolchand, P.; Micoulaut, M., Melt Homogenization and Self

Homogenization in Chalcogenides - Part 1, Intl. J. App. Glass. Sci. 2012, 3, 189-204.

[63] Wang, Y.; Nakamura, M.; Matsuda, O.; Murase, K., Raman-spectroscopy studies on

rigidity percolation and fragility in Ge-(S,Se) glasses, J. of Non-Cryst. Solids 2000, 266-

269, 872-875.

[64] Bhosle, S., The crucial role of sample homogeneity on the physical behavior of

chalcogenide glasses, M.S Thesis, University of Cincinnati, Cincinnati (unpublished)

2011.

[65] Mahadevan, S.; Giridhar, A.; Singh, A. K., Chemical ordering and topological effects in

chalcogenide glass systems, Indian Journal of Pure & Applied Physics 1995, 33, 643-652.

[66] Feltz, A.; Aust, H.; Blayer, Glass-Formation and Properties of Chalcogenide Systems

XXVI: Permittivity and the Structure of Glasses AsxSe1-x and GexSe1-x, J. of Non-Cryst.

Solids 1983, 55, 179-190.

[67] Avetikyan, G. B.; Baidakov, L. A., Izv. Akad. Nauk SSSR, Neorg. Mater. 1972, 8, 1489-

1490.

[68] Ota, R.; Yamata, T.; Soga, N.; Kunugi, M., Elastic properties of Ge-Se glass under

pressure, J. of Non-Cryst. Solids 1978, 29, 67-76.

[69] Yang, G.; Gueguen, Y.; Sangleboeuf, J.-C.; Rouxel, T.; Boussard-Plédel, C.; Troles, J.;

Lucas, P.; Bureau, B., Physical properties of the GexSe1-x glasses in the 0

in correlation with their structure, J. of Non-Cryst. Solids, (In press).

[70] Azoulay, R.; Thibierge, H.; Brenac, A., Devitrification characteristics of GexSe1-x

glasses, J. of Non-Cryst. Solids 1975, 18, 33-53.

74 [71] Bridenbaugh, P. M.; Espinosa, G. P.; Griffiths, J. E.; Phillips, J. C.; Remeika, J. P.,

Microscopic origin of the companion A1 Raman line in glassy Ge(S,Se)2, Phys. Rev. B

1979, 20, 4140-4144.

[72] Murase, K., in Insulating and Semiconducting Glasses, edited by P. Boolchand (World

Scientific, Singapore; River Edge, NJ, 2000), pp. 415-464.

[73] Sugai, S., Stochastic random network model in Ge and Si chalcogenide glasses, Phys.

Rev. B 1987, 35, 1345.

[74] Jackson, K.; Briley, A.; Grossman, S.; Porezag, D. V.; Pederson, M. R., Raman-active

modes of a-GeSe2 and a-GeS2: A first-principles study, Phys. Rev. B 1999, 60, R14985-

R14989.

[75] Boolchand, P.; Bresser, W. J., The structural origin of broken chemical order in GeSe2,

Philosophical Magazine B 2000, 80, 1757-1772.

[76] Massobrio, C.; Celino, M.; Salmon, P. S.; Martin, R. A.; Micoulaut, M.; Pasquarello, A.,

Atomic structure of the two intermediate phase glasses SiSe4 and GeSe4, Phys. Rev. B

2009, 79, 174201

[77] Petri, I.; Salmon, P. S.; Fischer, H. E., Defects in a disordered world: The structure of

glassy GeSe2, Phys. Rev. Lett. 2000, 84, 2413-2416.

[78] Boolchand, P.; Feng, X.; Bresser, W. J., Rigidity transitions in binary Ge-Se glasses and

the intermediate phase, J. of Non-Cryst. Solids 2001, 293, 348-356.

[79] Jacobs, D. J.; Thorpe, M. F., Generic rigidity percolation: The pebble game, Phys. Rev.

Lett. 1995, 75, 4051-4054.

[80] Bureau, B.; Troles, J.; Le Floch, M.; Smektala, F.; Lucas, J., Medium range order studied

in selenide glasses by Se-77 NMR, J. of Non-Cryst. Solids 2003, 326, 58-63.

75 [81] Gjersing, E. L.; Sen, S.; Aitken, B. G., Structure, Connectivity, and Configurational

77 Entropy of GexSe100−x Glasses: Results from Se MAS NMR Spectroscopy, J. of Phys.

Chem. C 2010, 114, 8601-8608.

[82] Vogel, H., Physik. Zeitschrift 1921, 22, 645-646.

[83] Fulcher, G. S., Analysis of recent measurements of the viscosity of glasses, J. of the Am.

Ceram.Society 1925, 8, 339-355.

[84] Tammann, G.; Hesse, W., Die Abhängigkeit der Viscosität von der Temperatur bie

unterkühlten Flüssigkeiten, Zeitschrift für anorganische und allgemeine Chemie 1926,

156, 245-257.

[85] Sipp, A.; Ritchet, P., Equivalence of volume, enthalpy and viscosity relaxation kinetics in

glass-forming silicate liquids, J. of Non-Cryst. Solids 2002, 298, 202-212.

[86] Mauro, J. C.; Yue, Y.; Ellison, A. J.; Gupta, P. K.; Allan, D. C., Viscosity of glass-

forming liquids, PNAS 2009, 106, 19780-19784.

[87] Gruber, G. J.; Litovitz, T. A., Shear and Structural Relaxation in Molten Zinc Chloride,

J. of Chem. Phys. 1964, 40, 13-26.

[88] Senapati, U.; Varshneya, A. K., Viscosity of chalcogenide glass-forming liquids: an

anomaly in the ‘strong’ and ‘fragile’ classification, J. of Non-Cryst. Solids 1996, 197,

210-218.

[89] Lucas, P.; King, E. A.; Gulbiten, O.; Yarger, J. L.; Soignard, E.; Bureau, B., Bimodal

phase percolation model for the structure of Ge-Se glasses and the existence of the

intermediate phase, Phys. Rev. B 2009, 80, 214114.

[90] Shportko, K.; Kremers, S.; Woda, M.; Lencer, D.; Robertson, J.; Wuttig, M., Resonant

bonding in crystalline phase-change materials, Nat Mater 2008, 7, 653-658.

76 [91] Gunasekera, K.; Boolchand, P.; Micoulaut, M., Elastic Phases of GexSbxSe100–2x Ternary

Glasses Driven by Topology, The Journal of Physical Chemistry B 2013, (Accepted).

[92] Feltz, A.; Büttner, H. J.; Lippmann, F. J.; Maul, W., About the vitreous systems GeTeI

and GeTeSi and the influence of microphase separation on the semiconductor behaviour

of Ge・Se glasses, J. of Non-Cryst. Solids 1972, 8–10, 64-71.

[93] Selvanathan, D.; Bresser, W. J.; Boolchand, P., Stiffness transitions in SixSe1-x glasses

from Raman scattering and temperature-modulated differential scanning calorimetry,

Phys. Rev. B 2000, 61, 15061-15076.

[94] Chen, P.; Boolchand, P.; Georgeiv, D., Long term aging of selenide glasses: evidence of

sub-Tg endotherms and pre-Tg exotherms, J. Phys.: Condens. Matter 2010, 22, 065104.

[95] Chen, P., Evidence for Intermediate Phases, boson and floppy modes, and demixing of

network molecular structure of binary As-S glasses, Ph.D., University of Cincinnati,

Cincinnati 2009.

[96] Boolchand, P.; Bresser, W. J.; Tenhover, M., Direct Evidence for Intrinsically Broken 8-

N Coordination Rule in Melt-Quenched Glasses by a Novel Method, Phys. Rev. B 1982,

25, 2971-2974.

[97] Bez, R.; Gleixner, R. J.; Pellizzer, F.; Pirovano, A.; Atwood, G., in Phase Change

Materials - Science and Applications, edited by S. Raoux, and M. Wuttig (Springler, New

York, NY, 2009), pp. 355-380.

[98] Boolchand, P.; Gunasekera, K.; Bhosle, S., Midgap states, Raman scattering, glass

homogeneity, percolative rigidity and stress transitions in chalcogenides, Physica Status

Solidi B 2012, 1-6.

77 [99] Boolchand, P.; Jin, M.; Novita, D. I.; Chakravarty, S., Raman scattering as a probe of

intermediate phases in glassy networks, Journal of Raman Spectroscopy 2007, 38, 660-

672.

78 APPENDIX

Here I append a list of journal publications and conference presentations I’ve actively taken part

in during my time as a graduate student at University of Cincinnati.

JOURNAL PUBLICATIONS: 1) K. Gunasekera, S. Bhosle, P. Boolchand, M. Micoulaut, “Topology, super-strong melts and

homogeneity of network glass forming liquids”, Journal of Chemical Physics (submitted).

2) K. Gunasekera, P. Boolchand and M. Micoualut, “Elastic phases of GexSbxSe100-2x driven by

topology”, Journal of Physical Chemistry B 117, 10027-10034 (2013).

3) R. Bhageria, K. Gunasekera, P. Boolchand and M. Micoulaut, “Fragility and molar volumes

of non-stoichiometric chalcogenides - The crucial role of melt homogenization”, Physica

Status Solidi B (submitted)

4) P. Boolchand, K. Gunasekera and S. Bhosle, “Midgap States, Raman scattering, Glass

Homogeneity, Percolative Rigidity and Stress Transitions in Chalcogenides”, Physica Status

Solidi B, 1-6 (2012) Invited.

5) S. Bhosle, K. Gunasekera, P.Boolchand, M. Micoulaut, “Melt Homogenization and self-

Organization of chalcogenide glasses - Part 1: Synthesis and homogenization of GexSe100-

x glasses”, Intl. J. of Applied Glass Science 3, 189-204(2012).

6) S. Bhosle, K. Gunasekera, P. Boolchand, M. Micoulaut, “Melt Homogenization and self-

Organization of chalcogenide glasses - Part 2: Sharply defined Intermediate phase in

structurally homogeneous glasses”, Intl. J. of Applied Glass Science 3, 205-220 (2012).

7) P. Boolchand, S. Bhosle, K. Gunasekera, K. Vignarooban and S. Chakraborty, “Glass

homogeneity precursive to self-organization”, J. of Optoelect. and Adv. Mater. 13, 1353-

1358 (2011).

79 8) S. Bhosle, K. Gunasekera, P. Chen, P. Boolchand, M. Micoulaut, C. Massobrio, “Meeting

experimental challenges to physics of network glasses: assessing role of sample

homogeneity”, Solid State Communications 151, 1851-1855 (2011).

9) K. R. Gunugunuri, K. Gunasekera, P. Boolchand, J. Dong and P. Smirniotis, “High

Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified

Ferrites”, J. Phys. Chem C 115, 7586-7595 (2011).

10) K. R. Gunugunuri, K. Gunasekera, P. Boolchand and P. Smirniotis, “Cr- and Ce-Doped

Ferrite Catalysts for the High Temperature Water-Gas Shift Reaction: TPR and Mossbauer

Spectroscopic Study”, J. Phys. Chem C 115, 920-930 (2011).

CONFERENCE PRESENTATIONS: 1) K. Gunasekera, P. Boolchand, M. Micoulaut and S. Mamedov, “Fragility, Intermediate

119 Phase, Sn Mossbauer effect in homogeneous SixGexTe100-2x glasses”, American Ceramic

Society Glass & Optical Materials Division Meeting, June 2013.

2) K. Gunasekera, P. Boolchand, M. Micoulaut and S. Mamedov, “Fragility, slow

homogenization and Intermediate Phase in SixGexTe100-2x ternary, American Physical Society

March Meeting 2013.

3) K. Gunasekera, P. Boolchand, M. Micoulaut, S. Mamedov, “Intermediate Phases and

Phase Change Materials”, American Ceramic Society Glass & Optical Materials Division

Meeting, May 2012.

4) K. Gunasekera, P. Boolchand, “Elastic Phases and Intermediate Phase in SixGexTe100-2x

glasses”, American Physical Society March Meeting 2012.

80 5) K. Gunasekera, P. Boolchand, “Elastic Phases in ternary GexSbxSe100-2x ternary glasses”,

American Physical Society March Meeting 2011.

6) K. Gunasekera, P. Boolchand, “Intermediate Phase in ternary GexSbxSe100-2x bulk alloy

glasses”, American Ceramic Society Glass & Optical Materials Division Meeting, May

2010.

7) K. Gunasekera, P. Boolchand, “On the molecular structure of GexSbxSe100-2x glasses”,

American Physical Society March Meeting 2010.