Growth of Indium Nitride and Gallium Nitride on Silicon Using Metal Organic Hydride Vapor Phase Epitaxy


Material Information

Growth of Indium Nitride and Gallium Nitride on Silicon Using Metal Organic Hydride Vapor Phase Epitaxy
Physical Description:
1 online resource (214 p.)
Chaudhari, Vaibhav U
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Chemical Engineering
Committee Chair:
Anderson, Timothy J
Committee Members:
Ziegler, Kirk
Ren, Fan
Davidson, Mark R


Subjects / Keywords:
epitaxy -- gan -- hvpe -- inn
Chemical Engineering -- Dissertations, Academic -- UF
Chemical Engineering thesis, Ph.D.
Electronic Thesis or Dissertation
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )


Well aligned catalyst free InN nanorods were grown on Si by metal organic hydride vapor phase expitaxy (MO-HVPE). The effect of different surface treatments and nucleation layers was studied on the orientations of nanorods and it was found out that nanorod orientations were highly dependent on surface conditions, giving signature of expitaxy. MO-HVPE grown low temperature GaN buffer layer was found to be best surface layer for archiving vertical nanorods. Nanorods were found to be defect free crystals, growing in vertical direction with no specific rotational preference with silicon substrate. The single crystalline nature of InN nanorods was thought to be because of possible lateral relaxation in nanostructures as they grow. The strain vs dislocation energy model which used minimization of total energy stored into the system due to lattice mismatch strain was used to calculate coherency limits of nanorod. It was found that if substrate is also assumed to be flexible, the coherency limit of InN nanorod can double. The model predicts the coherent diameter of nanorod for which nanorod of any length would be defect free. It also shows that dislocations will be only confined to stained bottom of the nanostructure which is consistent with the literature. The developed vertical InN template was used for growth of 50µm thick and crack free GaN without cracks on Si. GaN was free standing but polycrystalline. It was also found that the polycrystalline nature is as a result of polycrystalline nature of LT-GaN grown on the InN template. It was also shown that formation of completely enclosed uniformly distributed nanovoids were very essential to grow crack free thick GaN which is highly textured in vertical orientation. It was also shown that GaN grown was very high quality crystal free of Indium. GaN growth on Indium metal deposited on Silicon was also studied. Depending on growth mode and conditions GaN 40µm thick film and 100nm x 5000nm GaN wafers were successfully grown in same reactor.
Statement of Responsibility:
by Vaibhav U Chaudhari.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
General Note:
Description based on online resource; title from PDF title page.
General Note:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2012.
General Note:
Adviser: Anderson, Timothy J.
General Note:

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
lcc - LD1780 2012
System ID:

This item is only available as the following downloads:

Full Text




2 2012 Vaibhav Uday Chaudhari


3 To Aai and Baba


4 ACKNOWLEDGMENTS I am very fortunate that I got to work with Dr Timothy J Anderson. He always gives s tudents lot of independence to work and encourages everyone to define their own research problem. Though it was tough initially, I think it is only through his guidance, encouragement as well as patience and trust in me, I have been able to become a ny good researcher that I am today. I will be eternally thankful to Dr Anderson for being my G uru I am also thankf ul to all my current committee members, Dr Fan Ren, Dr Kirk Ziegler, Dr Mark Davidson. I am especially thankful to Dr Kirk Ziegler that he accep ted to be on my committee on such a short notice. I am also thankful to Dr. Olga Kryliouk for introducing and guiding me in the research initially. I am thankful to Dr. Josh Mangum and Dr. Joseph Park for all the help they have provided. My research could not have taken direction without enormous amount of background work they have provided I got to work with people of diverse background and culture in Dr Anderson Materials Processing Group (EMPG) I would like to thank all the people in EMPG for help and support they have provided. My work would not have been possible without the support I received from all working at Microfab ritech as well as Chemical Engineering I would like to thank all of them espe cially to Maggie, Chuck, Ludie Shirley Debbie, Carolyn, Dennis and Jim In the end, I would like to thank my parents for their unconditional love, support and sacrifice. My father has always been my inspiration and my mother has always been my motivation I am also thankful to my wife Milan a nd brother Gaurav for their support and making my time here at Gainesville more memorable.


5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Promising GaN and InN Technology ................................ ................................ ...... 15 Growth Techniques for Nitrides ................................ ................................ .............. 16 Solution Based Methods and Involved Challenges ................................ .......... 17 Heteroepitaxy Methods and Involved Challenges ................................ ............ 18 Silicon carbide ................................ ................................ ............................ 19 Sapphire ................................ ................................ ................................ ..... 19 Silicon ................................ ................................ ................................ ........ 21 a Role of Silicon nitridation ................................ ................................ ......... 21 b Epitaxial lateral overgrowth (ELO) ................................ .......................... 24 c Nanoheteroepitaxy (NHE) ................................ ................................ ....... 25 d Use of buffer layers ................................ ................................ ................. 27 Overview of Present Work ................................ ................................ ...................... 28 2 GROWTH OF VERTICAL INDIUM NITRIDE NANORODS ON SILICON IN MO HVPE ................................ ................................ ................................ ...................... 38 InN Nanostructures in Literature ................................ ................................ ............. 38 Metal Organic Hydride Vapor Phase Epitaxy (MO HVPE) ................................ ...... 39 Reactor Setup ................................ ................................ ................................ .. 39 MO HVPE Operation ................................ ................................ ........................ 40 Indium Nitride Nanorods (InN NR ) Growth in MO HVPE ................................ ........... 42 InN NR Growth Conditions ................................ ................................ .................. 42 Use of Surface Treated Silicon ................................ ................................ ......... 43 Nitride Nucleation Layers on Silicon ................................ ................................ 44 Results and Discussion ................................ ................................ ........................... 46 Effect of Surface Cleaning Treatments ................................ ............................. 47 Effect of Nucleation Layers ................................ ................................ ............... 48 InN NR Properties ................................ ................................ ............................... 52 Growth Mechanism of InN Nanorods ................................ ............................... 52 Summary and Conclusions ................................ ................................ ..................... 64


6 3 THEORITICAL MODEL FOR FINDING LIMITS OF COHERENCY FOR INDIUM NITRIDE NANORODS GROWN ON GALLIUM NITRIDE ................................ ...... 91 Overview ................................ ................................ ................................ ................. 91 Crystal Growth Modes ................................ ................................ ...................... 91 Strain vs. Dislocation Model for Heteroepitaxy ................................ ................. 94 Formulation of Stress Strain Relations in Hookian Transversely Isotropic Solids ................................ ................................ ................................ ............ 95 Formulation of Model for Nanorod Heteroepitaxy ................................ ................... 97 Rigid Substrate Model ................................ ................................ ...................... 97 Coherent strain energy (Ec) ................................ ................................ ....... 99 Energy of d islocated system (E) ................................ .............................. 102 Flexible Substrate Model ................................ ................................ ................ 105 S K growth and its effect on coherent diameter ................................ .............. 113 Summary and Conclusions ................................ ................................ ................... 114 4 GROWTH OF GALLIUM NITRIDE ON SILICON IN MO HVPE ............................ 132 GaN U sing InN in Litrature ................................ ................................ .................... 132 Growth of GaN in MO HVPE Reactor ................................ ................................ ... 133 Reactor Setup ................................ ................................ ................................ 134 MO HVPE Operation: HVPE mode ................................ ................................ 134 MO HVPE Operation: MOCVD mode ................................ ............................. 135 Gallium Nitride Growth in MO HVPE ................................ ................................ .... 136 GaN Growth Conditions ................................ ................................ .................. 136 GaN Growth Stages ................................ ................................ ....................... 136 Growth of low temperat ure GaN capping layer for InN nanorod template 136 Growth of thick high temperature GaN ................................ ..................... 137 Results and Discussion ................................ ................................ ......................... 137 Properties of Low Temperature GaN Capping Layer ................................ ..... 138 Crystallinity of LT GaN capping layer ................................ ....................... 139 Compositional analysis of InN template capped with LT GaN ................. 145 Properties of High Temperature GaN (HT GaN) layer ................................ ... 149 Growth conditions for thin and thick HT GaN ................................ ........... 149 Properties of thinner HT GaN ................................ ................................ .. 150 Properties of thick HT GaN ................................ ................................ ...... 152 Summary and Conclusions ................................ ................................ ................... 152 5 EXPLORATORY STUDY ON GROWTH OF GALLIUM NITRIDE OVER INDIUM TEMPLATED SILICON ................................ ................................ ......................... 179 Overview ................................ ................................ ................................ ............... 179 Growth Conditions for Different Layers in Growth ................................ ................. 179 Deposition of Indium M etal Film ................................ ................................ ..... 180 Deposition of GaN in MO HVPE Reactor ................................ ....................... 180 Thin GaN films ................................ ................................ ......................... 180 Thick GaN films ................................ ................................ ........................ 181


7 Results and Discussion ................................ ................................ ......................... 181 Thin GaN films ................................ ................................ ................................ 182 Thick GaN films ................................ ................................ .............................. 183 Summary and Conclusions ................................ ................................ ................... 185 6 CONCLUSIONS AND RECOMMENDATIONS OF FUTURE WORK ................... 199 Conclusions ................................ ................................ ................................ .......... 199 Recommendations for Future Work ................................ ................................ ...... 201 Improvements in InN nanorod orientations ................................ ..................... 201 Growth of GaN and In x Ga 1 x N nanostructures in MO HVPE ........................... 202 LIST OF REFERENCES ................................ ................................ ............................. 203 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 214


8 LIST OF TABLES Table page 1 1 Common substrates for InN and GaN epitaxy ................................ .................... 31 2 1 Peak ratios corresponding to (101) ................................ ................................ ..... 6 7 2 2 Relative peak intensity ratios for various surface treatments .............................. 68


9 LIST OF FIGURES Figure page 1 1 HRTEM of GaN grown on Si(111) with 2 3 nm Si 3 N 4 as intermediate layer ....... 32 1 2 Epita xial lateral overgrowth. ................................ ................................ ................ 33 1 3 Pendeo epitaxial growth. ................................ ................................ .................... 34 1 4 Equilibrium phase diagram of Ga O N Si system 73 on Ga rich side ................... 35 1 5 Strain relaxation in Nanoheteroepitaxy ................................ ............................... 36 1 6 GaN grown on nano porous silicon 96 ................................ ................................ .. 37 2 1 MO HVPE Reactor. A) Reactor Picture, B) Reactor Schmatic ........................... 69 2 2 InN growth map from MO HVPE. ................................ ................................ ....... 70 2 3 T= 873K ................................ ................................ ................................ .............. 72 2 4 Various InN nanorod orientations. Every nanorod orientation is forms specific angle with the surface ................................ ................................ ......................... 73 2 5 cleaning ................................ ................................ ................................ .............. 74 2 6 on Si(100) showing effect of surface cleaning with HCl ................................ ................................ ................................ 75 2 7 SEM images (10000x) showing effect of surface cleaning on InNNR ................. 76 2 8 nitridation ................................ ................................ ................................ ............ 77 2 9 SEM images showing effect of nitride layers ................................ ...................... 78 2 10 InN nucleation layer (Sample D) vs MOCVD GaN as nucleation layer (Sample E1) ................................ ................................ ................................ ........ 79 2 11 for nanorods grown on HVPE GaN nucleation layer with Si(100) and Si(111) as substrates ................................ ............................... 80 2 12 SEM images of vertical InN nanorods ................................ ................................ 81 2 13 SEM images of vertical InN nanorods grown on MOCVD and HVPE GaN ....... 82


10 2 14 Summary of effects of treatments on orientations of InN NR ................................ 83 2 15 Pole figure of InNNR grown on HVPE GaN on Si (111) sample in [002] and [101] directions ................................ ................................ ................................ ... 84 2 16 rocking curve of vertical InN NR sample with Molybdenum X ray source .......... 85 2 17 TEM analysis of nanorods ................................ ................................ .................. 86 2 18 EDS spectrum InN nanorods grown on HVPE GaN on Si (100) ......................... 87 2 19 Effect of N/In ratio and Substrates on InN morphology at Cl/In=4 and T=873K .. 88 2 20 Vertical InN nanorods grown on GaN nucleation layer on Si(100) ...................... 89 3 1 Modes of epitaxial growth ................................ ................................ ................. 116 3 2 Growth of InN film on GaN film viewed in direction ............................... 117 3 3 Schematic of InN nano structure growth on Ga N film (with possible InN film) .. 118 3 4 InN nanorod growth on GaN film ................................ ................................ ...... 119 3 5 Stress field in nanorod ................................ ................................ ...................... 120 3 6 Strain Energy vs. nanorod radius ................................ ................................ ..... 121 3 7 Strai n Energy vs. nanorod height ................................ ................................ ..... 122 3 8 Coherency map for InN nanorod on rigid GaN ................................ ................. 123 3 9 Strain Energy vs. nanorod radius for flex ible substrate model ......................... 124 3 10 Interfacial lattice constant vs. nanorod radius for flexible substrate model ....... 125 3 11 Strain Ener gy vs. nanorod radius (flexible substrate) ................................ ....... 126 3 12 Strain Energy vs. nanorod height (Ho) (flexible substrate) ............................... 127 3 13 Coherenc y map for InN nanorod on GaN ................................ ......................... 128 3 14 Coherency map for InN nanorod on GaN in S K mode ................................ .... 129 3 15 SEM and COMPO images showing no evidence of films below InN ................ 130 3 16 Relaxed strain in nanorod bulk ................................ ................................ ......... 131 4 1 Schematic of MO HVPE for GaN growth ................................ .......................... 155


11 4 2 Typical growth scheme for thick HT GaN growth ................................ ............. 156 4 3 SAED pattern for LT GaN buffer in vertical InNNR template ............................ 157 4 4 Different stages of capping layer growth ................................ .......................... 158 4 5 2 GaN for 15 min ................................ ................................ ................................ ............... 160 4 6 TEM image of InN GaN core shell structure ................................ ..................... 161 4 7 HRTEM of GaN spike ................................ ................................ ....................... 162 4 8 GaN as viewed in direction represented in different models ................ 163 4 9 GaN shell lattice fringe image ................................ ................................ ........... 164 4 10 SAED patterns of InN GaN core shell structure shown in Figure 4 6 ............ 165 4 11 TEM Image showing InN GaN core shell structures filled with Indium nitride or Indium metal ................................ ................................ ................................ 166 4 12 Point scan and line scan EDS in TEM both showing absence of Indium in shell possibility of Indium liquid in dark regions ................................ ................ 167 4 13 TEM and TEM EDS of InN template completely covered in LT GaN ................ 168 4 14 Annealing of LT GaN/InN NR /Si under ammonia atmosphere for 10 minutes at different temperatures showing decomposition of InN ................................ ...... 169 4 15 Areas under GaN and InN peaks and area ratio vs. anneal t mperature 170 4 16 Annealed LT GaN/InN NR /GaN sample ................................ .............................. 171 4 17 SEM EDS of annealed LT GaN/InN NR /GaN sample ................................ ......... 172 4 18 SIMS depth profile showing In, Ga and Si for annealed LT GaN/InN NR /GaN sample ................................ ................................ ................................ .............. 173 4 19 Cracked HT GaN layer on Si (2 m) thick) ................................ ........................ 174 4 20 GaN (2 m thick) film without cracks ................................ ................................ 175 4 21 TEM cross sectional image of HT GaN sample, 700 nm grain is sin gle crystal growing in direction ................................ ................................ .............. 176 4 22 SIMS depth profile of 1.5 micron thick GaN showing no diffusion of Indium in upper HT GaN layer ................................ ................................ ......................... 177


12 4 23 SEM image of 50 m thick GaN grown on InN template ................................ 178 5 1 SEM im ages of Indium on Silicon (100) ................................ ............................ 187 5 2 SEM images of GaN grown at T=873K for 10 min ................................ ............ 188 5 3 ................ 190 5 4 SEM images of growth on In/Si at higher t emperatures for 10 min ................... 191 5 5 .............. 193 5 6 SEM images of thick MOCV D GaN growth on In/Si at 1123K for 120 min ....... 194 5 7 ................... 195 5 8 SEM images o GaN/In/Si at 1123K for 240 min ................................ ................................ ................................ ............. 197 5 9 GaN/In/Si template at 1123K ......... 198


13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GROWTH OF INDIUM NITRIDE AND GALLIUM NITRIDE ON SILICON USING METAL ORGANIC HYDRI DE VAPOR PHASE EPITAXY By Vaibhav Uday Chaudhari August 2012 Chair: Timothy J. Anderson Major: Chemical Engineering Well aligned catalyst free InN nanorods were grown on Si by metal organic hydride vapor phase expitaxy (MO HVPE) The effect of different surface treatments and nucleation layer s was studied on the orientations of nanorods and it was found out that nanorod orientations were highly dependent on surface conditions, giving signature of expitaxy. MO HVPE grown low temperature GaN buffer layer w as found to be best surface layer for archiving vertical nanorods. Nanorods were found to be defect free single crystals, growing in direction with no specific rotational preference with silicon substrate The single crystalline nature of InN nanorods was thought to be because of possible lateral relaxation in nanostructures as they grow. The strain vs dislocation energy model which used minimization of total energy approach, was used to calculate coherency limits of nanorod. It was found that if substrate is also assumed to be flexible, the coherency limit of InN nanorod can double. The model predicts the coherent diameter of nanorod for which nanorod of any length would be defect free. It also shows


14 that dislocations will be on ly confined to stained bottom of the nanostructure which is consistent with the literature and observations. The developed vertical InN template was used for growth of 50 m thick and crack free GaN without cracks on Si. GaN was free standing but polycrysta lline. It was also found that the polycrystalline nature is as a result of polycrystalline nature of LT GaN grown on the InN template. It was also shown that formation of completely enclosed uniformly distributed nanovoids were very essential to grow crack free thick GaN which is highly textured in orientation. The GaN grown was very high quality crystal free of Indium. GaN growth on Indium metal deposited on Silicon was also studied. Depending on growth mode and conditions GaN 40 m thick film and 100nm x 5000nm GaN wafers were successfully grown in same reactor using this metal film template


15 CHAPTER 1 INTRODUCTION Promising GaN and InN Technology The pseudo binary solid solution Ga x In 1 x N alloys exhibits a direct band gap in t he range 3.4 to 0.7 eV. In addition to this, group III nitrides have very good mechanical and thermal stability. As a result, III Nitrides materials have been extensively studied for various optoelectronic and photovoltaic applications 1 12 The global optoelectronics market is growing rapidly and is projected to reach whopping US$932 billion by the year 2015 13 According to Solid S tate Lighting program developed by the U. S. Department of Energy advanced solid state lighting technologies should be cost competitive as compared to con ventional lighting technologies by 2025 This would be done by developing technologies that create solid state light sources that are much mor e energ y efficient, longer lasting but cheap to scale up The set goals are to be achieved by targeting a product system efficiency of 50 percent with lighting that closely reproduces the visible portions of the sunlight spectrum 14 These incentives have motivated the recent advancements in the sectors of solid state lighting and laser technologies. Still t he continued advances in III N materials, especially GaN, InN and Ga x In 1 x N alloys, are needed to make nitride technology and applications accessible to all The high brightness light emitting diodes (LEDs) made of GaN wi th Ga x In 1 x N active layer were first demonstrated in 1995 96 by Nakamura 15 17 and Yang 18 This was followed by blue laser diodes. As of now, v arious researchers have demonstrated use of nitrides for light emitting diodes and laser technology 19 22 The recent advancement to green laser diodes 524 nm with 50 mW continuous wave output power using c plane Ga N 19,20 will pave ways to advanced technology like mini RGB laser projectors. Along


16 with optoelectronics, III nitrides also find applications in various other fields. The III N materials direct bandgap can easily co ver entire UV visible electromagnetic spectrum and are stabl e at fairly high temperatures. This makes them ideal for photovoltaic applications. Various groups have demonstrated use of III N films and nanostructures for photovoltaic applications theoretical ly 23,24 as well as experimentally 1,2,9 11,25 29 High mobil ity, high drift velocities as well as high breakdown voltages make III nitrides suitable for high power and high speed applications such as field effect transistors (FET) 5,6,12,30 high electron mobility transistors (HEMTs) 3,4,8,31 etc. In addition to this, nanostructures find applicati ons in sensing devices due to high surface area. The transformation of these demonstrated applications in nitride devices to commercial applications depend on ability to grow excellent quality nitride materials on large scale. There are still various chal lenges in the growth of excellent quality GaN, InN and Ga x In 1 x N alloys. The section below discusses the development in GaN and InN growth technology very briefly. The section mainly focuses on major challenges that growth of III V semiconductors face. Gr owth Techniques for Nitrides The main that growth of nitride materials faces is availability of lattice matched substrates. Due to the lack of native substrates for growth of GaN and Ga x In 1 x N alloys, heteroepitaxy is primary technique for growing these ma terials. It is only recently after heteroepitaxial grown seeds of better quality nitride m aterials have become available.that people have started looking at other approaches such as solution based techniques.


17 Solution B ased M ethods and I nvolved C hallenges Due to very high equilibrium pressures of nitrogen required (in the order of few GPa) bulk crystals of nitrides e.g. GaN are not possible by typical equilibrium methods such as Czochralski or Bridgman methods. Therefore, lower temperature and pressure meth ods such as ammonot hermal synthesis are employed. These methods typically involve reactive solution that dissolves metal and ammonia to form metal nitrides. These nitrides in solution then precipitate out on seed crystals. The reactor operates at lower tem peratures of the order of 773K to 873K and an order of magnitude less pressure than equilibrium pressure 32 34 The GaN crystals grown by this method are the best quality to date as the crystals grow from solution a nd do not use any expitaxy on foreign substrates. Although researchers have grown crystals with less defects, the quality of product crystal still depends on seed crystal quality, reactant purities 35 etc The product crystals size is not well controlled due to, firstly, challenges in design for such high pressures and, secondly, less understanding and control over flow and temperature patte rns under reaction conditions. Getting rid of met allic impurities and oxygen also remains challenge in this method. As a result of these limitations, high quality GaN grown by this method are orders of magnitude expensive than that grown using other methods. Indium nitride growth has similar ch allenges as GaN by this method. In addition to requirement of having high equilibrium nitrogen pressure, it has low decomposition temperatures (~ 923K ). This makes growth of InN difficult by any melt or high pressure method. Although ammonothermal method o perates at lower temperatures than decomposition temperature, no studies have been yet reported on InN ammonothermal growth.


18 Heteroepitaxy M ethods and I nvolved C hallenges The other way, which is probably more popular and traditional way, to archive InN an d GaN crystal growth is deposit material at atmospheric pressure, sub atmospheric pressures or vacuums. These processes involve little higher temperatures (ranging from approximately 773K to 1473K ). The methods in this class are hydride vapor phase epitaxy (HVPE), m etal organic chemical vapor deposition (MOCVD), molecular beam epitaxy (MBE), reactive sputtering, atomic layer deposition (ALD) etc to name a few. Every method mentioned has its advantages and disadvantages. The c ommon challenge that all these m ethods face is perhaps unavailability of native substrate. In addition to this, heteroepitaxy of InN becomes complicated owing to the low ammonia cracking efficiency at typical growth temperatures of 773 923K The absence of lattice matched substrate rem ains to be the most critical issue for heteroepitaxy of nitrides. For any heteroepitaxy finding compliant substrate is important for reducing defects like threading and screw dislocations. Lattice mismatch as well as thermal expansion coefficient mismatch between substrate and epilayer results in straining of epilayer. As epilayer grows in thickness, stress and strain in the epilayer are stored in the form of strain energy. This strain can be easily relieved partially or completely by formation of dislocat ions, thus minimizing total energy of the system. As a result, dislocations are naturally induced in epilayer in heteroepitaxy. The behavior is different and it is more prone to developing cracks as well. This is undesir able for making devices. Thus, r educing strain by finding compliant substrate or any other means such as buffer layers, developed in epilayer of GaN durin g heteroepitaxy is very essential


19 The most common substrates, being used for growth in literature as well as commercially, are sapphire and silicon carbide. Table 1 1 summarizes some main substrates that are being used with their lattice mismatches with GaN and InN. Silicon carbide Silicon carbide has a lattice mismatch of 3.3% with GaN and it also has hexagonal structure like wurtzite GaN. It is available in various polytypes, but Si terminated SiC is the most preferred one. In Si terminated polytype, there are alternate hexagonal closed packed planes of Si and carbon with surface terminated with S ilico n. It has been observed that GaN surface termination can be changed depending on the polytype of SiC used 36,37 The GaN film is N itrogen and G allium terminated respectively for Si terminated and C terminated silico n carbide substrate. The SiC substrates have high thermal and electr ical conductivity, the properties that are preferred in electronic applications. Although all these properties make it a suitable substrate for GaN growth SiC substrate has some limitatio ns. Main limitation is high cost, with more than 3000$ per 4 inch wafer 38 SiC substrate is hard to etch which limits it applicabi lity in electronic devices. Also, gallium wetting properties of SiC are poor, which results high defect density in GaN film. So, often buffer layer such as AlN has to be used 39 42 This completely defeats the purpo se of using SiC for its low lattice mismatch as AlN depositions adds one more step in GaN deposition. Sapphire Sapphire has lattice larger mismatch (~16 % ) with GaN and even higher (~28.6 % ) with InN that result in high density of dislocations of the order of 10 10 cm 1 Although sapphire has larger lattice mismatch with GaN, it is the most popular substrate. In 1983, Yoshida et al 43 first showed th at use of thin layer of AlN on s apphire substrate


20 could improve quality of GaN grown on it greatly. This was confirmed by many other studies that followed 4 4 47 This improvement in growth is attributed to two effect s of AlN buffer layer. Firstly, reduction in microscopic fluctuations in crystallite orientation due to AlN layer reduces strain between GaN and sapphire 45 That is both AlN and GaN being wurtzite structurally results in effective reduction of the strain between GaN and Al 2 O 3 surfaces. Secondly greatly reduced lattice mismatch between AlN and GaN ( ~ 2.4%) reduces surface energy at AlN GaN interface w hich promotes lateral grown of GaN. The same is true for InN growth, but reduction in mismatch to approximately 13.7 % is not as good as that in case of GaN. Thus, growth of InN film on sapphire remains more challenging. In spite of this challenge, and alt hough very uncommon, use of InN film as a buffer layer for GaN on sapphire have been reported by two authors 48,49 over span of more than a decade. Researchers have found that InN can be used as buffer layer to grow GaN on top because of soft nature of InN as well as low decomposition temperatures. Furthermore InN has same crystal structure as GaN. The relaxation of residual strain, developed during high temperature GaN growth, due to soft nature and decomposition of InN, resulted in better GaN growth. In addition to buffer layers like AlN, InN, various buffer layers like Al x In 1 x N, ZnO, BN, low temperature GaN have been reported my numerous authors. Sapphire substrates, although being popular, face various problems a s well. Sapphire has very low thermal conductivity (5.43W/mK) as compared to GaN (130W/mK). As a result sapphire is very bad heat sink for high power devices made from GaN. This is problematic for device operation as well as device lifetimes. Low electrica l conductivity of sapphire forces device contacts to be made on top or device side itself. This results in loss of important device area and complicates


2 1 device architecture. Also, thermal expansion coefficient of sapphire is more than that of GaN. This mak es growth on large diameter substrates i .e. larger than 2 inch problematic due to cracking and bowing of GaN. Silicon It has been observed that every decade cost per lumen falls by factor of 10 whereas amount of light generated per LED package increases 2 0 times 14 This progress of solid state lighting technology, calls for gr owth of better quality materials on cheaper substrates. The silicon substrate has unique advantage of availability in large sizes, high quality and lower cost. In this article, various approaches for using silicon as substrate for nitrides are discussed. T he lattice mismatch between silicon and gallium nitride is very large (17%) and thermal expansion coefficient is also very high (33%). As a result, the GaN layer grown on silicon is constantly under tensile stress, which results in various defects such as high density of dislocations, unintentional n type doping 50 as a result of silicon diffusion as well as cracks in GaN films. From the Table 1 1, silicon seems to be best substrate for InN with 7.9% lattice mismatch. But, with GaN mismatch with Si is very large (more than double that of InN). To add to this issue, silicon has diamond cubic crystal structure that does not match with GaN (more so for Si (1 0 0) than Si (1 1 1)). The I ndium nitride, on the other hand, is more lattice matched substrate for GaN as compared to S ilicon. a Rol e of Silicon nitridation Similar to nitridation of sapphire, nitridation of silicon has been shown to reduce the density of defects such as dislocations in overgrown GaN by few researchers. Nitridation of silicon produces ultra thin film (2 3 nm Figure 1 1 ) of silicon nitride 50 or si licon oxynitride depending on conditions and temperatures used. The effect of silicon


22 nitride layer depends on the thickness of this Si x N y layer. It has been observed that thickness of this layer determines the quality of Si x N y layer. As thickness increase s, the Si x N y layer tends to become amorphous and GaN grown on it tends to be polycrystalline i.e. a mixture of cubic and hexagonal GaN 50 53 Such GaN also shows multiple in plane alignments with respect to underlyi ng Si. The wurtzite GaN over layer on Si generally shows up as two in plane domains rotated by 30 54,55 in XRD pole figures The formation of Si x N y layer can be avoided altogether by introducing aluminum flux for s mall interval initially. Since thermodynamically Al N bonds formation is highly favored over Si N bond formation, introduction of Al results in very thin interlayer of AlN instead of Si x N y 50 52 In the same effect, AlN buffer layer similar to that in case of sapphire can be used to grow good quality of GaN on Si 54,56 68 The AlN layer has lower lattice mismatch, similar structure and GaN is known to wet AlN completely 69 The AlN buffer layer also brings overgrown GaN under compressive stress rather than tensile stress, the stress that is thought to produce cracks in material. As a result, GaN grown using AlN buffer layer with lesser defects (reduction from 10 10 to 10 8 cm 2 ) and crack free thickness exceeding 1 micron have been shown to be grown on silicon substrate as well. It has been shown that, the formation of Si x N y layer is not always detrimental for GaN overgrowth Ultra thin Si x N y layer has been used to grow high quality GaN by various resear chers 25,70 75 Recent studies 75 show that use of high temperature Si x N y layer grown at 1173K produced better quality GaN films than that without nitridation or with low temperature nitridation. Most of the researchers, however, agree that thick Si x N y layer can detrimental to overgrown GaN quality. Thick Si x N y layer formation can


23 be avoided by initial nitridation at lower temperature. Contrary to thick layers, thin layer tends to maintain some epitaxial relationship with silicon. For example, Si x N y whic h forms at lower temperatures, has hexagonal symmetry 76 This symmetry is similar to basal plane of wurtzite GaN with 16% lattice mismatch. Although lattice mismatch is high, GaN is able to maintain the epitaxial r elationship due to its similar crystal structure. As a result GaN grown on nitridated silicon tends to be wurtzite as compared to mixed cubic wurtzite GaN grown on bare silicon 75 In addition to that ultra thin Si x N y layer can act as barrier for Si diffusion into GaN avoiding unintenti onal n type doping of GaN by Si. As a result of this GaN grown has high structural quality as well as it is free from shallow or deep level traps. This is evident from the observations that GaN grown using Si x N y layer showed sharp XRD as well as PL peaks and PL was free of yellow shift or yellow sub peak 70,75 This technique generally involves growth of ultra thin Si x N y layer at temperatures lower than 850K followed by growth of low temperature GaN buffer layer 73 Thermodynamic analys is of Ga O N Si system was done to understand equilibrium chemistry at GaN Si interface. Oxygen was introduced in trace amount as quartz reactor tube and small impurity in all chemicals entering the reactor can act as source of oxygen. It was found that only temperatures above the 850K amorphous Si 3 N 4 is thermodynamically favorable to form ( Fig ure 1 2 ). So at lower temperature, Si surface can be partially covered with oxygen and partially covered in nitrogen. This oxinitride layer is thought to maintain e pitaxial relationship with Si as well as GaN overlayer grown at low temperature. Another view on this process is that low temperature GaN buffer acts exactly similar to AlN layer. The GaN does not allow Si N


24 bonds to form because GaN formation is favored o ver Si 3 N 4 formation at lower temperatures according to equilibrium phase diagram in figure 4. b Epitaxial lateral overgrowth (ELO) Researchers have developed more complex methods such as pendeo epitaxy to grow better quality GaN on silicon 77 83 This technique is similar to epitaxial lateral overgrowth (ELO) methods 84 86 The ELO methods generally involve growth in two to three steps. First low temperature and low quality GaN is grown on substrate. It is then covered with patterned oxide mask such as SiO 2 Then high temperature GaN seed columns grow only through small holes that expose GaN below the mask and no seed growth takes place on SiO 2 mask itself. As growth continue s, GaN from each seed column starts growing laterally to form high quality GaN. The pendeo epitaxy method involves even more steps to grow GaN on Silicon. For growth on Si, silicon is carbonized to form thin SiC layer which as very small lattice mismatch with GaN. Since GaN does not wet SiC well, it is covered with thin layer of AlN. The GaN thin layer is then grown on the AlN which can be done at higher temperature to get better quality GaN. Then in the next step, the GaN is covered with Si x N y layer and patterned to expose faces of GaN. The high temperature GaN grows from these seeds laterally. The GaN growing from pendeo seeds coalesce and grow to get very high quality GaN. The ELO and pendeo approaches are represented in cart oon in Figure 1 3 and Figure 1 4 respectively. With this method, it is possible to grow both thick and thin GaN films on silicon substrate with dislocation density as low as 10 5 10 6 cm 2 83


25 c Nanoheteroepitaxy (NHE) Parallel to the idea of patterned substrates, nanostructures suc h as nanorods, nanowires or nano patterned substrates can also be used as templates for growing GaN on silicon. The stress distribution in heated bimetal thermostat was found shown depend on the width of the thermostat 87 This idea was further applied to stresses developed in heteroepitaxy of materials with large lattice mismatch such as Si x Ge 1 x /Si system 88 Contrary to the plane 2D substrate as shown in Figure 1 5(A) if nanostructures such as seed pads, nanorods etc are used, the effective strained contact area in he teroepitaxy can be reduced to the area seed pads. Due to 3D nature of these seed pads or nanostructures, strain developed at heteroepitaxial interface can be confined to small thickness in the overgrown film. This is as a result of possible exponential lat eral as well as vertical relaxation of strain in nanostructures over the characteristic length which is proportional to smallest dimension of nanostructure or seed pad Figure 1 5(B). Due to this relaxation process, theoretically strain energy can be kept b elow the energy required to form dislocations 88,89 As results, there is reduction in dislocations in grown GaN. This process is termed as nanoheteroepitaxy (NHE). The theory of nanoheteroepitaxy and its applicatio n to the growth of system with high lattice mismatch such as GaAs/Si and GaN/Si is developed in series of publications by Zubia and Hersee 89 92 In comparison to the direct 2D growth such as growth of GaN film on thick or ultrathin silicon substrate, use of patterned Si with nano stripes is more effective in reducing the defect s in overgrown material. This later approach is based on the principles developed earlier by Luryi and Suhir 88 Due to the nano patterned substrate, the effective contact between the substrate and epilayer is reduced. As a result stress is


26 effectively accommodated within first few nanometers of epilayer as a result of lateral as well as vertical relaxation. The nanoheteroepitaxy is more effective than this approach by Luryi and Suhir and is more effective for systems with very high lattice mismatches. The main difference in the approach above and nanoheteroepitaxy is the use of nano sized seed pads in NHE instead of nano strip pattern. For example, instead of using patterned silicon, SOI with patterned silicon nano pads can be used. The flow rates and temperatures can be adjusted such that GaAs or GaN will seed selectively on Si nano pads. Due to the ability of lateral relaxation, the strain due to lattice mismatch is distributed between seeds of silicon as well as overgrown GaAs or GaN nuclei. As a result, strain as well as deformations, such as dislocations that form as a result of strain, are confined to small regions of nuclei. Also, all nuclei being away from each other, the dislocations can glide and terminate at the edge of nuc lei. The overgrown layer is relaxed and highly crystalline 90 The later approach (NHE) requires additional step of growing nano pads for nucleation over SOI, as compared to the patterning of silicon stripes which can be simply integrated with available techno logy. The NHE techniques is, however, better for very high lattice mismatched systems such as GaN on silicon. The NHE approach is used by different researchers to grow GaN on nano porous GaN on sapphire 93 SiC 94 as well as silicon 95 97 The approaches, however, do not follow NHE exactly to as described earlier, but by employing templates of substrate that is covered with seeding pads of nano porous materia ls such as silicon and GaN. These nano porous seed pads are formed by lithography techniques that used anodized porous aluminum oxide membrane as a mask. This produces porous pads on substrate


27 with uniform pore diameters typically 60nm. Use of such techniq ues has shown promise growth of quality GaN on silicon substrate ( Figure 1 6 ). d Use of buffer layers There have been various other efforts to grow GaN on silicon using different buffer layers such as AlAs films 98 ZnO films 98 102 ZnO nanorods 103 boron phosphide films 104 108 The ZnO has less than 2% lattice mismatch and same wurtzite crystal structu re. This makes ZnO highly compliant buffer layer material for GaN growth. The GaN grown with such layers is shown to be high quality with dislocation densities as low as 10 8 cm 2 Mainly two types of threading defects are more prominent in these systems, viz. stacking mismatch boundaries (SMB) and inversion domain boundaries (IDB). The SMB defects in GaN films are formed because of defective ZnO layer. The crystal structure of Si is very different from ZnO wurtzite structure. This difference in stacking se quence in substrate and ZnO epilayer often results in stacking faults in ZnO layer. The IDB defects are formed because of growth of opposite face or polarity materials on each other 99 It is often required to grow low temperature GaN layer over ZnO buffer layer. Direct exposure of ZnO to ammonia results in its deformation. As a result, Zinc diffuses out or into the GaN layer resulting in low quality GaN. Us e of low temperature GaN to cover ZnO buffer layer protects this layer. It, however, has disadvantage of forming gallium oxide at interface. The use of nanostructured ZnO template also has similar disadvantages. Due to very large surface area of ZnO nanoro ds in this case results in complete decomposition of ZnO layer after GaN growth 103 forming GaO x at interface. Not many studies are done for BP as potential compliant layer for GaN growth, but this approach becomes interesting as it provides possibility to grow cubic GaN. Boron phosphide has less than 0.6% lattice mismatch


28 with zinc blende GaN and both boron as well as phosphorus are used as dopants for silicon. It has been found that cubic GaN can be grown on Si (1 0 0) using BP buffer layers. The quality of GaN however depends on quality of BP layer. In some cases where BP layer is polycrystalline, GaN tends to be polycrystalline as wel l as cubic GaN mixed with more stable wurtzite GaN. Due to limited results as well as formation of multiple phases of GaN, this approach is still limited to research. Although not much work has been done on potential use of InN as buffer layer on silicon, the InN film as buffer on sapphire has been shown to be very promising 48,49 InN has lower lattice mismatch with silicon than sapphire. Thus, use of InN film and nanostructures as buffer on silicon is more plausibl e choice than that on sapphire. Overview of Present Work Former section presents a very brief overview of current status of GaN and InN growth. The most pressing issue in the growths of nitrides is absence of native substrates. As a result growths of both GaN and InN are predominantly via heteroepitaxy methods. Silicon technology is very mature and as a result in recent years lots of efforts are focused on growths of nitride materials on Silicon. GaN has especially created lot of interest in research commun ity due to its applicability in high brightness optics applications as well as high power electronic devices. The current work presents one such technique to grow GaN on Silicon. This study has tried to use nanostructures of InN to grow GaN on Silicon. As discussed previously, InN has lower lattice mismatch with Silicon than GaN. InN can also form pseudo binary alloy In x Ga 1 x N with GaN. Also, in parallel with NHE approach discussed previously nanostructures can be better in relieving strain. They also have better crystal quality. Both of these things can be


29 effective in stress reduction in GaN if InN is used as buffer layer. As a result, InN nanostructured templates can be used as compliant substrates for GaN growth. The study is divided into three main part s. First part of study, presented in Chapter 2, is growth of InN nanorod on Silicon. Chapter discusses in detail about effect of different surface treatments as well as buffer layer on orientation as well as morphology of InN nanorods. This study can be us ed as a guide for other similar growths. This Chapter gives unique and only recipe to grow catalyst and patterned template free vertical InN nanorod growth on Silicon. This has never been achieved before on Silicon substrates even with catalyst use. Simila r to many other nanostructures grown in different studies, InN nanorods in this study show ed high crystalline quality. When films grow, they tend to be rigid and developed strain results in high density of defects such as dislocation as well as cracks. Bot h dislocations and cracks act as strain relieving mechanisms in high strain systems such as III N hetero epitaxial growths. Research shows, however, unlike two dimensional films, nanostructures are much less prone to defects. This is also widely observed w ith nanostructures grown on foreign substrates. This is often attributed to their high aspect ratio. Because of small er diameters, and contact in only base plane, nanostructures can relax laterally as they grow. This enables them to quickly re lax strain as the y grow. Chapter 3 presents an energy minimization model for nanorod growth. It qualitatively describes critical dimensions for which nanorods will be free of defects. Although the model uses InN nanorods grown on GaN as case study, model is very genera l and can describe equilibrium limits of coherency for any hetero epitaxial system.


30 With previous Chapter s giving recipe for high quality vertical nano structures as well as idea about strain relaxation capability of the nanostructures, Chapter 4 discusses the growth of GaN on these novel nanostructured templates. Although InN films have been used as buffer layers for GaN growth on Silicon, this study distinguishes itself with use of InN nanorod template. The Chapter is divided into two subsections with low temperature GaN growth in first section and high temperature growth in the second section. The Chapter discusses various stages of GaN growth on InN as well as interesting roles that are played by InN template. Most interestingly, question of disappearanc e of Indium from final GaN growth which is also observed by other researchers, is addressed here. Also, study has successfully shown the growth of thick GaN without cracks by using InN nanorod templates. With InN decomposing, at high temperatures of GaN g rowth study of possible use of only Indium metal film for growth of GaN is discussed in Chapter 5. It will be shown in this Chapter that similar to InN nanorods as template In/Si can also be used as possible template for thick GaN growths. This Chapter als o shows interesting results about growth of GaN wafers with uniform 100nm thickness and 8 m lengths can be grown with use of In/Si templates. Indium metals role as surfactant and oxide mediated growth is believed to be responsible for these nanostructures Final Chapter along with some key conclusions, gives suggestions on future work that add to as well as improve this approach.


31 Table 1 1 Common substrates for InN and GaN epitaxy Substrate Structure Lattice Constants () Lattice Mismatch (%) a c G aN InN GaN Wurtzite 3.189 5.185 0 10.9 InN Wurtzite 3.537 05.704 9.8 0 c Al 2 O 3 Rhombohedr al 4.765 12.982 15.9* 28.6* AlN Wurtzite 3.112 4.982 2.4 13.7 6H SiC Wurtzite 3.086 15.117 3.3 14.6 Si Diamond 5.431 --------16.9 7.9


32 Figure1 1 HRTEM of GaN grown on Si(111) with 2 3 nm Si 3 N 4 as intermediate layer 50


33 Figure 1 2. Epitaxial lateral overgrowth. A) schematic of growth, B) cross sectional SEM of ELO growth 85 A B


34 Figu re 1 3 Pendeo epitaxial growth A ) schematic of growth, B )cross sectional SEM of PE growth 79 A B


35 Figure 1 4. Equilibrium phase diagram of Ga O N Si system 73 on Ga rich side


36 Figure 1 5. Strain relaxation in Nanoheteroepitaxy A) dislocation formation in 2D heteroepitaxial structures, B) strain relaxation in nanopads and seeds in NHE 90 A B


37 Figure 1 6. GaN grown on nano porous silicon 96


38 CHAPTER 2 GROWTH OF VERTICAL INDIUM NITRIDE NANORODS ON SILICON IN MO HVPE InN Nanostructures in Literature T he global optoelectronics market is growing rapidly and is projected to reach whopping US$932 billion by the year 2015 13 This projected growth, however, relies greatly on continued progress in understanding of group III nitride growth and processing. Amongst the group III nitrides InN is least stu died. It is now getting attention because of its interesting electronic properties. It has lowest direct bandgap of 0.7eV amongst the group III nitrides 109 which has extended the coverage of nitride and alloys from deep ultraviolet to far infrared region in electromagnetic spectrum. Also, due lowest effective mass amongst group III nitrides leading to superior electronic properties such as potentially higher mobility, higher drift velocities, there is increased interest u sing InN in optoelectronic applications such as laser diodes, solar cells, sensors and high frequency devices. The growth of high quality epitaxial InN has been a challenge, due to low decomposition temperatures of InN, high equilibrium nitrogen partial pr essures 110 at growth and lack of compliant substrate s for InN. Various reported results, however, suggest that growth of highly crystalline InN is possible if it is grown in the form of nanostructures rather than films. 111 117 The growth of InN nanorods and nanowires have been studied and reported by various researchers 109,111 114,116 121 .The metallic catalysts such as particles of gold, nickel etc. are most commonly used in growth as they provide preferential sites for reactio n with growth taking place via vapor liquid solid (VLS) mechanism. The metal droplet often remains unreacted in this process and can be observed on tip of the nanowire. Although this method can be promising, size of metal droplet does not always


39 ensure the uniformity of shape and size of nanorods 122,123 Also, getting rid of metal droplets after growth especially noble metals, can be a tedious as well as expensive The other most common technique is by patterning o f substrate 119 but it has not been t hat successful in achieving single crystalline InN nanostructures. There are very few reports that demonstrate self assembled growth of InN nanorods without patterning or catalyst 111,113,124 All the reports that m ention catalyst and pattern free growth, use halide based approach to get nanorods. This works also presents on such halide based but unique approach. Metal Organic Hydride Vapor Phase Epitaxy (MO HVPE) Metal Organic Hydride Vapor Phase Epitaxy (MO HVPE), as name suggests, combines more conventional Metal Organic Chemical Vapor Deposition (MOCVD) and Hydride Vapor Phase Epitaxy (HVPE) Due to uniqueness of the method, no literature is available on this method except reports by Anderson group. Hyun Jong Park has previously reported InN nanorod growth on various substrates using this method 111,112 Reactor S etup The reactor photograph and schematic are presented in Fig ure 2 1 (A) and Figure 2 1(B) respectively. It is a hot wall quartz reactor fitted in resistively heated furnace which is typically operated at atmospheric pressure The furnace has five individually controlled heating zones with heating capacity more than 1273K But in current configuration heating is res tricted to 1273K to avoid any damage to quartz reactor from possible softening at higher temperatures. Currentl y only three zones are in use and they act as source zone, mixing zone and reaction zone. The temperatures of the zones are adjusted in such a wa y that desired temperature profile is achieved. Substrates are loaded into the reactor with magnetic loading arm on tiled quartz susceptor. Load lock


40 separated by gate valve helps to minimize oxygen contamination into the reactor. The reactor can be operat ed with Nitrogen (N 2 ), Hydrogen (H 2 ), forming gas (4% H 2 balance N 2 ) and Helium (He) as sweeping and carrier gases. Typically, N 2 was used as carrier gas for InN growths. High sweep gas flows from inlet and gate valve are adjusted so as to minimize wall d epositions and confining growth to growth zone. MO HVPE O peration Typically temperature profile is adjusted in such a way that end of the inlet zone is maintained in the range of 573K to 673K. This ensures complete decomposition of metal organic as well as activation of other reactants before they enter the mixing zone. The typical overall reactions that are expected in InN formation are as follows (2 1) (2 2) (2 3) (2 4) (2 5) (2 6 ) (2 7 ) (2 8 ) Inlet consists of three concentric quartz tubes. Metal organic which in this case is Trimethyl Indium (TMI) enters the reactor from central tube. Its shorter design ensures that metal organic react s with 10% Hydrogen Chloride (HCl) gas to form Indium Chloride (InCl) ( Reaction 2 1) as well as Indi um trichloride (InCl 3 ) and other chloride


41 species before it comes out in mixing zone. However, it has been known that formation of InCl 3 and other chloride species is less favorable thermodynamically especially at temperatures used 125,126 Indium chloride then reacts with ammonia to form InN which gets deposited on hot su b strate (Reaction 2 3) If HCl gas is not used Trimethyl Indium decomposes to form Indium or reacts with ammonia directly after coming out in mi xing zone, to form Indium nitride (Reactions 2 2, 2 4, 2 5). Thus, depending on presence and absence of HCl gas same reactor can be operated in HVPE mode and MOCVD mode respectively Typical growth temperatures used for InN are in the range of 773K to 823 K. At these temperatures, ammonia decomposition is very limited. As a result, incomplete reaction of metal organic with ammonia or InN decomposition can result in Indium rich film or even Indium droplet formation (Reactions 2 2, 2 6 respectively) ork 127 however, shows that addition of H Cl gas can completely eliminat e formation of Indium droplets (Reaction 2 7). The excess use of HCl can also result in etching of InN film (Reaction 2 8). It has been however observed that the etching of InN solid by HCl is not isotropic and results in etch pits rather than uniform etch ing 126 Thermodynamic analysis of In C H Cl N, suggests that depending on HCl to Indium (HCl/TMI) ratios used, at constant V/III (i.e. a mmonia to Indium) ratio different growths of InN are expected at different temperatures 111,125 These typically include InN growth with Indium liquid droplets, InN solid growth, and InN no growth. Droplets are expe cted for HCl/TMI ratios 1 and below. When excess HCl is used i.e HCl/TMI ratio > 1, InN can either form or can be completely etched away as a function of temperature as seen in Figure 2 2. In practice, the boundary between growth and no growth is not


42 sharp as predicted by theoretical calculations. As a result, near growth etch boundary, competition between formation reactions and decomposition reactions often results in decrease in nucleation and growth. Due to large lattice mismatch, surface energetics is such that, like many other heteroepitaxial semiconductor growths, nitride growths are in Volmer Weber mode of film growth 47,128,129 In this mode of growth 130 nucleation is in the form of small islands, which grow and coalesce to form film. In fact, gas kinetics is also not simple in MO HVPE an d the amount of HCl in the reaction zone play s a role in kind of nucleation that follows Raman spectroscopy studies and DFT calculations reve a l that HCl/TMI ratio is 1 or less, InN 2D film growth is expected as InCl is predominant source of Indium. But wh en HCl/TMI ratio is 3 4, also is major source of Indium. It can form a complex with ammonia. This complex is known to po lymerize which can result in chain and ring compounds of form 131 The combination of all these factors results in growth of nanostructures. Figure 2 2(B) gives an idea on variety of possible growths of InN possible in t he reactor. In dium Nitride N anorods (InN NR ) G rowth in MO HVPE InN NR G rowth C onditions Trimethyl I ndium solution (TMI), T rimethyl Gallium ( TMG) from Epichem and 99.999% pure ammonia from Airgas South were used as indium, gallium and nitrogen source respecti vely. For InN growth TMI was reacted with 10%HCl (balance nitrogen) from Airgas South in source zone at 573K to form chlorinated species of Indium. These species were then mixed with ammonia from the concentric inlet in mixing zone. The substrate temperatu re was maintained at 833K 873K for growth. The inlet HCl/TMI molar ratio and NH 3 /TMI molar ratio were maintained at 4 and 250 respectively The N 2


43 carrier gas flow rate of 1600 sccm was also maintained all along the 1 hour runs. These growth conditions ens ured dense InN nanorod growth as established from previous studies 111,112 Silicon (100) and Silicon (111) were used as substrates and were loaded together in every run. As this study was done only to see the effect of different surface conditions of substrates, following sets of runs were carried out. For this work following surface treatments were considered. The methods that clean the surface of foreign materials, particulate entities and that are helpful in passi vation of silicon surface without adding considerable epitaxial layer of different material are treated as surface treatments. All other methods that form certain layer of material to Si surface are treated as nucleation layers. Use of S urface T reat ed Sili con A Organic solvent d egreasing and b uffered o xide e tching: In these sets of runs two types of silicon substrates were used. First set of silicon substrates were simply degreased using standard cleaning method. This method used baths of boiling trichloro ethylene (TCE), acetone and methanol. The silicon wafers were cleaned by dipping in boiling baths organic solvents for 5 minutes each in the previously mentioned sequence. Then they were washed with deionized water and dried using nitrogen gas This step r emoved any carbon or oily residues formed on silicon surface In second set of silicon, in addition to surface degrease, substrate wafers were dipped in buffered oxide etch (BO E ) for 1 min at room temperature. This step was done to remove some native oxide from silicon surface. InN samples grown on degreased and BO E cleaned S ilicon would be called sample A1 and sample A2 in further discussion. B In situ surface cleaning with 10% HCl : In these sets of runs, degreased Silicon wafers were loaded into the rea ctor and were treated with HCl for 10 min at


44 873K prior to actual InN NR growth. The flow rate of 10% HCl gas was kept at 50 sccm. The HCl is very reactive and can be used as etching agent to remove oxide layer similar to hydrofluoric acid in BO E clean in p revious section. Reaction of HCl with silicon oxide forms silicon oxychlorides which are thought to be volatile 126 As a result, HCl was t hought to be effective agent for surface cleaning of silicon. Samples of InN NR grown on HCl cleaned Si will be called sample B in further discussion. Nitride N ucleation L ayers on Silicon The results from previous studies 73,111 suggested that intermediate nitride layer can be used to grow better quality GaN and InN respectively on Silicon and Sapphire substrates These studies markedly improved quality and orientation of nitrides grown using same method. The idea was to study if the use of nucleation layer changed the orientation distribution of InN to preferred orientation. Initially, Silicon substrates were nitridated by exposure to ammonia and then the MO HVPE reactor system was used to grow films in two modes. The MOCVD mode films were grown when no HCl was used The reactions took place by thermal decomposition of MO followed by reaction with ammonia The HVPE mode films were grown when HCl was used and nitride was formed by reversible reaction between chloride sp ecies and hydride (which is ammonia in this case). In this study, in total, four types of nitride buffer layers were used, viz Silicon oxynitride, InN using HVPE growth mode and gallium nitride using both MOCVD condition and HVPE conditions. Then, InN nan orods were grown on these nucleation layers. Then their effect was studied on further InN nanorod growth and orientation using XRD and SEM. C In situ surface nitridation before growth: Nitridation of silicon is known to form amorphous silicon nitride SiN x on its surface. Due to the amorphous nature, the


45 SiN x layer is known to decrease the quality of nitrides grown. The nitridations were still carried out based on previous work. In previous work, equilibrium studies were done for Ga O N Si system as well as actual growth experiments were conducted 73 The studies revealed that SiN x formation embarked only at ~848K The SiN x formation deterred formation of good quality nitride on silicon. But, on the contrary the nitridation done at lower temperatures and for sh orter time intervals resulted in silicon oxynitride SiO 1 x N x layer. This layer was shown to improve the quality nitrides grown on top. For these reasons, nitridation of silicon was carried out followed by InN nanorod growth. In these sets of runs after sil icon substrates were loaded, in situ nitridation using ammonia at 1500 sccm was carried out. It was done by passing ammonia over hot silicon substrate at 833K for short time intervals of 10 minutes. After nitridation InN growths were carried out using cond itions mentioned earlier. These samples would be called sample C in further discussions. D Low temperature I ndium nitride film as nucleation layer : For growth of indium nitride film following conditions were used. The inlet HCl/TMI molar ratio and NH 3 /TMI molar ratio were maintained at 2 and 700 respectively. The growth zone temperature was kept at 873K while mixing and source zo n e were maintained at 723K Indium nitride was not grown without HCl because absence of HCl results in bad quality of indium nitr ide films which had indium droplets on top. The InN samples grown on these layers would be called sample D E Low temperature gallium nitride film as nucleation layer : Gallium nitrides films were grown in both HVPE mode and MOCVD mode. For HVPE mode GaN f ilms growths were carried out as follows. The TMG from Epichem and NH 3 from Airgas south


46 were used as gallium and nitrogen sources. The TMG was mixed with 10% HCl in source zone to form chloride species of gallium. The HCl/TMG molar ratio was maintained at 2. These species then reacted with NH 3 in mixing zone to form GaN film on silicon substrates. Growth zone was maintained at 873K and NH 3 /TMG molar ratio was set at 570. In separate sets of runs, GaN films in MOCVD mode were grown by maintaining high NH 3 / TMG molar ratio of about 3000 and temperature of 873K These conditions were selected based on optimizations in previous studies, the d etail s of which can be found elsewhere 73 The InN nanorod samples that use MOCVD and HVPE GaN films as buffer layers would be called sample E 1 and sample E 2 respectively in further sections. R esults and D iscussion Samples grown on bare Si substrate tend to be randomly oriented. Figure 2 3 shows typical powder X ray diffraction (XRD) pattern for randomly oriented InN nanorod sa mples grown on degreased silicon i.e.A1 samples. As seen from XRD InN nanorods showed diffraction peaks for various families of planes such as etc. The illustration in Figure 2 4 shows that to particular angle to the substrate surface. For example, vertical nanorods have ( 002) plane horizontal, that is the [002] direction makes angle of 90 with horizontal plane or substrate plane. Similarly, to get diffraction from (100), (101), and (102) planes nanorods should be tilted at angles of 0, 28, 47 to the substrate. This ang le is same as angle made by [002] direction to horizontal plane in respective cases. T wo qualitative


47 approaches are employed in order to study if any orientation is preferred. Firstly, ratios of intensity of a particular peak to a reference peak in same sa mple are calculated. These ratios are then compared to corresponding intensity ratios reference powder XRD pattern. The PDF # 50 1239 powder XRD data for InN from International Centre for Diffraction Data (ICDD) was used as a reference. Only, f irst four ma jor peaks are considered for further analysis. The (101) peak was chosen as reference peak in each sample, since it is highest intensity peak in reference PDF. In other words with (101) peak intensity as unity, the other peak intensities were calculated. T hese ratios for each sample are tabulated in Table 2 1. More the ratio for a certain peak, higher the preference of that sample to be oriented in that direction. In second approach relative intensity ratio R was determined for each peak. The ratio is defin ed as ratio of normalized intensity of observed peak to normalized intensity of same peak in reference. First approach only allows seeing preference in certain orientation as compared to (101) orientation in that particular sample. It d oes not say anything about the (101) orientation preference. The second approach, which is tabulated in T able 2 2 allow s comparison of every peak in the reference sample but d oes not tell real comparison between peak ratios of same sample. It is however worthwhile to note he re that these ratios strictly g i ve information about orientation or texture qualitatively and not quantitatively. The SEM images are taken to supplement the results observed by XRD. Effect of S urface C leaning T reatments The major difference in sample A1 a nd sample A2 is that later ha s native oxide layer partially removed. As silicon oxide is amorphous the nucleation in both cases can be different which can result in different orientations of nanorods. The XRD are shown in F igure 2 5 2 1 )


48 and relative intensity ratios (Table 2 2 ), suggested that there is not much difference in the samples. The sample B does not show any effect on any peak either (Figure 2 6). The calculations however show increased intensity of (102) peak indicating increase in that particular orientation. The reason behind this might be that HCl being anisotropic etchant, creates etch pits into the silicon. These sites are energ etically favored for nucleation As the nanorods growing from (111) Si planes from etch pits do not grow vertically, they add to the other orientation. As the random orientation would increase with more exposure to HCl prior to the reaction, no HCl clean w as done for further studies. The SEM images of A1, A2 and B showed no differences in the morphology in F igure 2 7 Although, t here is no signi ficant difference visible in XRD, InN nanorods growing on HCl cleaned Si not only are random, but also have wide d iameter distribution with diameters in the range from 50nm to 400nm. Wide distribution here also results from the fact that once nanorod becomes big enough, it serves as nucleation site, resulting in branching. Such secondary nucleations are clearly visibl e in Figure 2 7(C). The secondary nucleations or branching, however, is totally expected The same behavior was observed by previous researchers 125,126 On the other hand samples A1 and A2 in Fig ure 2 7(A) and (B) respectively show uniform distributions with InN NR with diameters nea r 200nm and 100nm respectively, which is an indicator of uniform nucleating surface. Effect of N ucleation L ayers Growth on nitridated Si: The InN NR grown on nitridated silicon show defini te improvement in (002) peak as compared to non nitridated sample A1 (Figure 2 8 ). This is because nitridation forms silicon oxynitride layer under the experimental conditions used. It was observed that nitridated Si surface consisted of a.7% Si N bonds an d 26%


49 Si O bonds and rest bulk Si Si bonds 73 Although exact mechanism of nanorod nucleation is not known on this surface, there is improved nucleation of nanorods in (002) direction due to presence of some Si N bonds. Layer of SiN x is believed to have hexagonal geometry, sim ilar to wurtzite InN and GaN 76 T he observed results are in agreement with improvements in GaN films quality that were reported previously on nitridated silicon 73 The A1 and A2 samples are perfectly random nanorods. The SEM pic ture in Figure 2 9 (A) confirms that sample C ha s visibly less number of random nanorods, and number of vertical nanorods ha s increased. This sheer increase in number of [002] nanorods increased corresponding peak intensity. Comparison between InN film, MO CVD GaN film and HVPE GaN film as nucleation layers: One of the earlier studies reported improvement in GaN film quality on silicon using silicon oxynitride layer 73 Also, there was improvement in texture in InN nanorods on silicon due to nitridation. The com bination of these observations led to studies in use of different nitride layers as nucleation layer for nanorod growth. Results from this approach are discussed next. The use of MOCVD GaN and HVPE GaN further improve s texture InN nanorods in [002] direct ion as expected. It is clear from the Table 1 and Table 2 that sample E1 and E2 are highly textured in [002]. But, the corresponding values of ratios for D samples revealed even more randomness than samples C The F igure 2 10 scan for the D and E1 samples. The [002] texture of InN nanorods in E1 is evident. This means that nucleated nanorods on InN nucleation layer are random. This randomness is consequence of two reasons. Firstly, it might just be mainly bec ause of any texture being absent in InN nucleation layers itself. It is but evident from previous observations


50 that InN by nature grows randomly on the si licon and nanorods follow texture of the film Secondly, it is known that the nanorods tend to be rand omly oriented if they originate from single big nucleation sites 132 The indium droplet wetting of InN film surface is not uniform 111 The nanorod nucleation in this typ e of reactions goes through the Cl 2 In:NH 2 adduct formati on. These adducts then undergo oligomerization to form InN 131 If a bigger indium droplet of In dium forms on film during beginning of the reaction the indium chlorine ammonia adduct forms at multiple points on same droplet, resulting in multiple nucleation from same spot. Moreover, Indium can readily wet InN as compared to GaN. As a result, many In N NR on InN film can have larger diameters. This explains observed wide size distribution of InN nanorods in Figure 2 9 (B) The Fi gure 2 11 illustrates that HVPE GaN nucleation layer can be used to get InN nanorods with very high [002] preferred orientati on on both Si (100) and Si (111). This (002) texture on GaN nucleation layer is also confirmed from cross sectional SEM images in Figure 2 12 This means that the directional nature of nanorods is governed directly by the quality of nucleation layer it nuc leates and not the actual substrate. This can be further explained by observing differences in InN nanorods grown on different GaN nucleation layers. The HVPE GaN nucleation layer growth after nitridation is highly textured in [002] directions; contrary to MOCVD GaN in which (101) GaN peak is also observed. In addition, the MOCVD grown GaN film tends to form small spherical features on surface, f rom which multiple nanorods can nucleate in different direction. One such multiple nanorod nucleation site on MO GaN film is shown in Figure 2 13 (A), whereas Figure 2 13 (B) corresponds to initial growth on HVPE GaN surface. Here, randomness is arising only due to InN tripods formation. The InN tripods are formed


51 when zinc blend cubic core is formed at the base as a result of strain relaxation mechanism and wurtzite InN NR grows from (111) faces of zinc blend core. The nanorods grown on MO GaN nucleation layer however, tend to be more uniform and smaller in diameter. The differences in polarities of GaN grown in M OCVD and HVPE could be reason behind this. T he MOCVD grown GaN is gallium terminated and is smooth In comparison, t he HVPE grown GaN films are rough er and tend to be N terminated, because they are grown in presence of HCl. The HCl acts as metal scavenger and also makes surface rough because of its anisotropic etching. Thus, in sample D, due to smoother and uniform surface, nanorod nucleation tends to be more uniform in size, resulting in more uniform nanorods. Fully grown, InN NR on MO GaN films and HVPE Ga N films are shown in 2 13 (C) and 2 13 (D) respectively. In the end a ll results are summarized graphically in Figure 2 14 It compares percentage relative ratio of each of the four major orientations considered in this study. Powder XRD only gives informa tion about the dominance of certain orientation provided that the particular planes produce constructive interference in X rays. Although, power scans show that InN nanorods on HVPE GaN films are oriented in [002] direction, pole figure gives exact rotatio n about the vertical axis and omega rocking curves gives qualitative information about certain orientation. Pole figures of vertical InN NR on Si (111) using GaN are shown in Figure 2 15. The results show that indeed GaN and InN are have their c axes align ed with [111] of Si. The phi scan for [101] peak of GaN reveals that GaN grown here does not have any preferred rotational orientation on Silicon substrate. InN NR just follow the rotation of the GaN crystal, which is expected 125 Thus, there is a texture in [002] direction, but there is no rotational preference observed


52 in these samples. The omega rocking curve done on same sample using Molybdenum X ray source is shown in Figure 2 16. The rocking curve has FWHM of 936 arcsec which is better than any reported values in literature for InN nanostructures. InN NR Properties The nanorod s are tested for crystal quality in JOEL 300CX transmission electron microscope. The F igure 2 1 7 shows individual flat ended as well as tipped nanor od. Lattice fringe image as well as SAED analysis revealed that nanorods were wurtzite with [002] growth axis along their length and {100} faces making the hexagonal shape. The TEM images also occasionally showed presence of some planer defects and evidenc e of secondary nanorod nucleation on nanorod main body. These nucleations could have also added to randomness of nanorods. The energy dispersive spectroscopy (EDS) in SEM of E2 samples ( Figure 2 1 8 ) revealed that no oxygen was present in the films in detec tion limit, however, 0.77 atomic percent of chlorine was detected. Although the detection below 1% and cannot be considered in quantitative manner, it reveals presence of chlorine in sample. We speculate that the chlorine signal is coming from HVPE GaN fil ms as nanorods by this process do not have any chlorine. Growth Mechanism of InN Nanorods As seen in previous sections, MO HVPE system is similar to HVPE system in operation when HCl is introduced into the system. The metal organic source seems to complete ly react with HCl to form Indium chloride and Indium trichloride, which act as primary Indium precursors. Increase in HCl results in increase in trichloride. At constant Indium metal flux, the value of Cl/In and N/In ratios indicate the amount of HCl and A mmonia in the system. In addition to these parameters, substrate temperatures, distance between inlet and substrate, type of substrate used, growth time etc. are the


53 parameters that influence the InN growth. As a result, for the study of effect of single parameter all other parameters need to be constant in order to see or state full effect of that parameter. However, it is to be noted that various parameters are interlinked and as a result, it is often not easy to isolate the effects of different paramete rs. The effects of change of some of the very important parameters are discussed below. Also, the effect of each variable is discussed at different conditions formed by change in other variable to better understand the complete mechanism of highly anisotro pic crystals such as nanorods. Effect of Cl/In ratio : The control on Cl/In ratio in MO HVPE system, is perhaps the most unique feature that this system offers. The Cl/In ratio is very critical in suggested mechanism of nanostructures formation in MO HVPE s ystem at constant temperature as well as N/In conditions 111,131 The chlorine to indium ratio influences the formation of chloride species of Indium. The proposed growth mechanism of InN growth depending on Cl/In ra tio are as follows. It is based on gas phase kinetics study using Raman Spectroscopy and density functional theory calculations. (2 9) For Cl/In=1 (2 10) For Cl/In=3 (2 11) (2 12) (2 13)


54 (2 14) At low Cl/In ratio of 1, as shown in equation 2 9 dominant chloride that forms is InCl. The InCl and ammonia then adsorb or chemisorb on surface as shown in equation 2 10 This results in formation InN film via layer by layer growth or islands coalescence depending on surface energetics. In case of Cl/In ratio of 3, InCl 3 is dominant gaseous specie s This species can undergo chain in of reactions as shown in equations 2 11 and 2 12 which results in monomer complex by elimination of HCl from Cl 3 In:NH 3 adduct. This gaseous monomer can undergo polymerization owing to strong H Cl polar interaction between monomer molecules. The o ligomers such as Cl 9 In 6 N 6 H 9 are ring compounds with wurtzite structure with InN at core 133 and Hydrogen and Chlorines on outside. Such oligomers are heavy and can form random nuclei on substrates. The supersaturation of Indium chloride species in gas phases then drives formation of InN nanorods. The oligomer nuclei have c axis as a polar axis. The reactivity of polar precursors of Indium is higher along polar axis than the surfaces. This results in higher reactivity along c axis, resulting in higher growth rates along direction. Further increase in Cl/In ratio increases amount of HCl in the system. The HCl is reactant that drives formation of more InCl 3 as well as it is a product in formation of monomers as well as oligomers that are responsible for InN nucleation. As a result, increase in HCl generally drives equilibrium towards formation of more InCl 3 in chloride species. Increase in HCl also results in shifting equilibrium towards monomers of InCl 3 NH 3 complex. This can be seen as more dominant etchi ng reactions by HCl. As a result, increase in Cl/In ratio results in sparse nucleations, more etching and no growths.


55 Significant increase in NH 3 is required to drive reactions towards oligomers and nanostructures. In general, at constant temperature and N /In ratio, the diameter of nanorods grown decreases with increase in Cl/In ratio. The effect can be attributed to more etching reactions. It is also seen that diameter distribution tends to become more and more uniform as Cl/In ratio increases. This effect however is complex function of gaseous oligomer formation and their interaction with substrates during heterogeneous nucleation. Effect of N/In ratio : Effect of ammonia is most straight forward of all reactants. Ammonia is always present in excess and as a result ammonia often is not a limiting reactant. As N/In ratio is increased, activity of ammonia increases and InN formation is favored according to Le Chatelier's principle. Independent of Cl/In ratio and temperature, at which InN is grown, increase in ammonia increases InN formation and as a result InN sows transition from nano or island growth to film growth. Effect of ammonia is profound on InN morphology. As seen from Figure 2 19,(taken from previous work in same reactor 125 ), although different substrate have different types of InN nanostructure morphologies, increased ammonia in the system results in transition of growth from nanostructures towards film growths. This is again result of shift of equilibrium towards pro ducts InN, HCl and ammonium chloride, due to ammonia rich atmosphere. As discussed in previous section, at high HCl concentrations, increase in ammonia can drive equilibrium towards oligomers and nanorods can form. Relatively ammonia poor conditions are n ecessary for InN nanostructure formation in case of MOCVD growths, which proceed by Vapor Liquid Solid mechanism. For


56 example, the impinging jet geometry used by Josh 126 results in low apparent N/In ratio at surface of substrate. As a result of this nanos tructures growth via VLS growth is observed. Indium flux in this case only controls the density of nucleations. As a result, most dense nu cleations are at the center of impinging jet and nanostructures become sparse towards edges of subsector. It is also worth to note here that, ammonia poor conditions here do not translate to N/In ratios of less than 1. Such ratios are bound to result in In dium metal droplet formations at the surface of films. In general it can be thus concluded that environments very rich in ammonia can result in bigger nucleations and growth. Ammonia poor environments, with optimum flows of other reactants can result in fi lm as well as nanostructures. Effect of temperature on InN growth : Temperature is always a very important parameter in material growths. InN starts to decompose above 923K, but NH3 cracking is not effective below 773K, which limits the growth window of InN to very narrow temperature range. As a result, the InN formation gets better with increase in temperature, peaks and then declines. At lower temperatures growth is limited by ammonia cracking and at higher temperatures, decomposition dominates. In MO CVD mode or when there is no enough HCl around to form Indium chloride, often indium metal droplets form. Depending on flow patterns used films as well as nanostructured growths are possible. For example, horizontal inlet results in film growth, whereas impin ging reactants can result in nanostructures via VLS mechanism. Very temperatures around 873K. At lower temperatures, no active nitrogen is available. This


57 results in metal drople t formation even at very high N/In ratios. At temperatures 973K and above, no growth is possible as InN is unstable at these temperatures. In presence of HCl, (Cl/In>1), InCl and InCl 3 are present and InN is formed by reaction between these species. Unlike MOCVD growth, reactions in presence of HCl do not take place by radical mechanism, but direct reaction of ammonia with chlorides. As a result of different mechanism, the reaction is not limited by active nitrogen concentration resulting from cracking of a mmonia. As a result, at lower temperatures, InN formation is possible in this mode. Effect of temperature; however is similar i n terms of quality and morphology of InN formed. InN quality increases with increased temperature, as increase in energy of molec ules allows them to form better crystals. InN films become more columnar, and when temperature is above 923K InN decomposition dominates and there is no growth. With increased Cl/In ratio, chemistry changes and InN grows as discontinues film or nanostructu res. This can be seen as a result of more corrosion due to increased HCl as well as corrosion due to hydrogen produced in cracking of ammonia. But effect of temperature remains similar. Thus, a general trend is that crystal quality gets better with increas e in temperature, followed by decrease in growth as temperature goes towards decomposition temperature of InN. Although crystalline quality increases, for higher Cl/In ratios nanorod diameters decrease and diameter distribution becomes narrower with increa se in the temperature. The smaller diameter crystals are better at relaxing the strains, and as a result exhibit better crystalline quality than that of bigger nanorods. This trend is clearly visible in Figure 2 2.This effect is also explained in higher de tails in work by Hyun Jong Park 112,125


58 Effect of substrate type : The effect of surface on quality and nucleation of InN film as well as nanorods is rather complex. The lattice mismatches of Sapphire, AlN, GaN and Si with InN are approximately 28%, 13%, 11% and +8% respectively. As a result Si can be seen as best lattice matched substrate for growth of InN. It was, however, seen that at constant Cl/In ratio, N/In ratio and temperature, nanorods grown on Silicon in general were bigger in diamet er than that grown on Sapphire. The diameters were smallest for nanorods grown on GaN nucleation layer. The diameters had narrower distribution in case of GaN nucleation layer as compared to any other substrate used. Use of InN nucleation layer also accoun ted for la r ger diameters and wider size distribution. All nanorods showed faster growth along polar c axis and all samples always showed some preference towards vertical orientation. The preference for vertical orientation with nucleating surface was most for GaN layer followed by Silicon and Sapphire. It has been consistently observed in this study as well as other works 125,126 .Previous study 126 also suggested that InN nanorods tend to be flat tipped in case of polar substrates like c GaN, a Al 2 O 3 c Al 2 O 3 InN nanorods have sharp tips in case of nonpolar substrate like r Al 2 O 3 and Silicon. On r Al 2 O 3 formation of nonpolar InN or GaN films is more common, but nanorods are found to be almost always oriented in these studies. In rare cases growth axis was found to be This study as well as study by Hyun Jong Park, however, did not always find the strong relation between surface polarity and shape of the tip of nanorod. It was often seen that densely nucleated nanorods tend to be flat tipped and rare nanorods tend to be sharp tipped. The flat as well as sharp tipped vertical nanorods grown on GaN nucleation layer are


59 shown in Figure 2 20. All nanorods had nitrogen polarity as Indium metal is taken away by HCl in the reactor. All these observation s are very hard to explain on the basis of gas phase nucleations alone. The random nucleation can be explained by the gas phase oligomer formation, which then gets deposited on surfaces at random locations. B ut distribution and difference in the diameter s izes of nanorods grown on different surfaces cannot be explained by random nucleation alone. Oligomers complexes can have random sizes, but if surfaces on which nanorods are grown did not have any role to play; random nanorods would show same diameter dist ributions independent of substrates used. The fact that this uniformity is not observed, points towards role of surface energetics on further growth of nanorods. The heterogeneous nucleation is highly influenced how surface energies are changed when nuclea tion occurs on foreign surface. Consider a nucleus having shape of hexagonal parallelepiped deposited on the substrate as shown in Figure 2 21. The energy change in the surface when such nucleus is deposited is given by (2 15) where is difference in chemical potential of nucleating species in vapor and condensed phase, the driving force for nucleation is molar volume of nucleated InN is s pecific free energy of nonpolar vertical faces of nucl e i is variation of surface free energy at interface due to formation of nuclei


60 is surface free energy of basal plane or polar plane is specific free energy of adhesion of InN on substrate surface The quantity can be seen as measure of how much more difficult is to split the crystal than to separate itself from substrate surface. When force of adhesion is stronger is negative and adhesion is strong whereas positive values of will indicate that adhesion is weak. The former gives rise to Frank Van der Merwe or layer by layer type of growth an d later gives rise to island growth which is also known as Volmer Waber (V W) type of growth. In some cases, initially few layers of layer by layer growth take place, but strain energy makes it more favorable to form islands thereafter. This type of growth mode is called Stranski Krastanov (S K) mode. The V W mode and S K mode are more common in semiconductor heteroepitaxy 130,134 With the help of the above equations, many of the observations regarding differences in the diameters nanostructures formed on the different surfaces can be explained. Surface wetting or adhesion of InN on Si is weaker than sapphire 135 As a result nucleation in general tends to be better on sapphire substrate. But the InN grown on Si is under tensile stress as compared to the InN grown on other substrates. This also is very important factor to be considered when InN is grown on any surface. For the over grown layer grown under tension, it is energetically more favorable to form films and bigger grains for dislocations are readily be introduced in growing material as growth takes place 136 As a result, the nanorods tend to be bigger in size when grown on Silicon as compared to the sapphire substrate. Nitridation of silicon produces layer of SiN x which is often detrimental to the further semiconductor growth. But silicon nitride is known to have hexagonal structure similar to InN basal plane. As a result the growth


61 on nitridated sample becomes more oriented vertically. As far as surface wetting is considered, the adhesion decreases and different sizes of InN nuclei can form and grown under V W growth mode. As a result, growth on nitridated silicon becomes more directional but overall diameter distribution is widened. The use of GaN and InN nucleation layers show two very opposite results when InN nanorods nucleate on these surfaces. The In N nanorods nucleating on InN will have no specific preference for layer by layer vs island growth. But, due to strain in InN layer and growth conditions used, island growth prevails. As a result nanorods with wide diameter distribution and random nucleatio ns originate. In case of GaN nucleation layer however, due to difference in the materials, the adhesion is weaker as compared to perfect InN crystal itself. As a result, nanorods with smaller and uniform nuclei nucleate. The GaN layer on which nanorods are growing was found to be oriented in polar c axis. The oligomer complexes that are responsible for nucleation of nanorods also have polar c axis. As a result, these nuclei tend to align with GaN grains to form vertical nanostructures. The alignment of exac t hexagonal basal plane to hexagonal c plane of GaN grain in nucleation layer is energetically most favorable. Thus, if GaN film is well oriented in direction and has epitaxial relationship with substrate, InN nanorods growth als o exhibits similar alignment. This is clear from the results of vertically oriented nanorods on Si in this study and that on sapphire in previous study 125 The molecular kinetic theory of crystal growth states that for film to nucleate and grow, work of nucleation has to be overcome. The principle kinetic equation for the growth rate V of a crystalline face is given by following equation.


62 (2 16) where is molar volume is edge energy which is related to specific surface energy of growing face is Boltzman constant is temperature is lattice constant is supersaturation The nucleation on kink site is energetically more rewarding than growth on step site. The nucleation on flat faces is least rewarding energetically. Because of this the closed packed faces, which tend to be ver y flat, are slowest growing faces and large supersaturations are required to increase rate drastically. As a result in crystal or film growth, slowest growing faces are the faces that generally remain at the end. As seen from the equation above, high super saturation will be required to grow InN along direction, and growth rate in direction cannot be greater than any other face. The nanorod is very anisotropic crystal and growth along c axis is certainly faster than other nonpolar m faces. As a result according to this theory growth of nanorods is highly improbable. Still nanorods are found to have formed experimentally. These contradictions in the theory and experimental results can be eliminated if ene rgetically growth of plane is favored. Often it has been found that growth on nanorods proceeds by tip formation. The tips of nanorods are typically planes like which have higher growth rates than m faces that form sidewalls of the nanorods. The se faces that form tip of the nanorods have step sites which increase the growth


63 rates. Often defects such as screw dislocation and twin defects are foun d at the tips 126 in high resolution TEM analysis. If the defect such as screw dislocation forms in c plane, a spiral step is introduced in otherwise perfect closed packed plane. This dislocation then acts as the continuous source of growth. But most of the flat tipped nanorods were highly sing crystalline with no defects such as dislocations. As a result, growth mechanism of such flat tipped nanostructured nanorods had to be different. In HVPE the c plane is often nitrogen terminated and faces such as that are nitrogen terminated tend to be rough as compared to the faces which are metal t erminated. Also due to polarity of these faces the sticking coefficients of polar molecules such as InCl which act as precursors is more on these faces. It can be seen mathematically as decrease in edge energy to a very small val ue. Decrease in edge energy, results in drastic increase in growth rate along c plane. In fact, it has been observed that, mobility of Indium on m faces is higher under ammonia rich conditions. The indium species tend to get absorbed on all surfaces as wel l as substrate and migrate to c faces which are rapidly growing faces 137 This Vapor Solid mechanism for InN nanorod growth seems to drive InN growth to form highly anisotropic crystals on various substrates The presence of V S growth mechanism can be further confirmed by different characteristics of Vapor Solid growth First of all, in any conditions, no Indium droplets are detected in SEM as well as TEM. The nanorods tend to have constant or slightly tapering diameters. This is indicator of In diffusing from side walls to c faces or tips where growth rate is maximum. The taper is especially observed on microrods. As length of rod increas es, it becomes more difficult for Indium to diffuse from faces to the


64 tip. As a result, stems tend to be fatter at the bottom and become narrower towards tips. Second feature of typical S V growth is independent growth of nanorods. Nanorods often tend to grow and intersect and still keep growing into each other. In VLS mechanism, the intersection sites are often worked as new nucleation sites, or nanorods fuse and grow into one. The bran ching is only rarely observed branches seem to grow independently like new nanorods, which is typical in V S mechanism. The formation of tripods and random nanostructures in this mechanism is discussed in detail in other parts of Chapter 2. In accordance w ith the theory developed here, branching of nanorods can be initiated by formation of oligomeric clusters of InCl 3 NH 3 that have cubic form 133 .Suc h clusters can nucleate, with (111) faces forming a tetrahedron. From each face of tetrahedron, however, more stable wurtzite InN are formed. In similar manner, different f lower like clusters of nanorods can originate from more complex ring oligomers. No liquid droplet as nucleus is necessary to explain presence of such clusters and growth can proceed via V S growth mechanism. Summary and Conclusions To summarize, the effect of different surface cleaning steps on InN nanorod orientation is presented in this Chapter The study has presented different nucleation layer approaches that can be used to get preferred orientation of InN nanorod on silicon. The nanorods cannot be grow n in preferential direction on silicon alone. Nanorods require similar material nucleation layer to control nucleation directions. By using textured nucleation layers like MOCVD GaN and HVPE GaN, the InN nanorods can be grown in [00 0 2] orientation without use of any catalyst and patterning. The orientations depend on quality and type of nucleation layer. The results also suggest


65 that the size of nuclei formed by oligomerization nucleation reaction between chlorinated species of I ndium ammonia can vary depen ding on nucleation layer. Thus, diameter of nanorods is determined by nature of nucleation layer in addition to temperature, NH 3 /TMI ratio and HCl/TMI ratio. The study also shows that, InN NR tend to follow exact texture of underlying film, if they have si milar crystal structures. The nanostructures grown tend to be very uniform dimensions on smother uniform films. In all, nanorod show excellent crystal quality. They show no oxide layer formation on the surface and are highly crystalline in nature. The defe cts tend to be agglomerated at certain locations such as bottom of nanorods. In the end it can be concluded that choosing the right surface treatments and controlling the quality of nucleation layer is critical to control the orientations of InN nanorods. The vertical single crystalline nanorod growth on silicon is possible using textured GaN layer on silicon. Key to get even better textures on InN NR lie only in improvement of nucleation layer. The effect of various parameters considering both gas phase oli gomerization mechanism as well as surface energetics is discussed at the end of the Chapter. Detailed analysis on observed results showed that possibly gas phase oligomerization and formation of InCl 3 NH 3 ring complexes was responsible for initial nucleati on at high Chlorine to Indium ratios. The nucleation then proceeded to produce InN nanorods by Vapor Solid growth mechanism. The high aspect ratio of InN crystals to form nanorods, was the result of enhanced growth rates of c direction because of various f actors such as surface roughness, possibility of surface defects such as screw dislocations The tip formation can also lead to enhanced growth rates. Possibility of higher sticking


66 coefficients of polar species such as Indium Chlorides on polar faces and h igh surface mobility of Indium species resulting in migration of reactants to c faces was also thought to be responsible for high growth rates of InN in direction. As a result, possible growth mechanism of random as well as vert ical nanorods was discussed in its entirety.


67 Table 2 1 Peak ratios corresponding to (101) P eak ratio A1 A2 B C D E 1 E2 Si(100) Si(111) (100)/( 101 ) 0.16 0.02 0.34 0.06 0.07 0.10 0.04 0.06 (002) / ( 101 ) 1.35 1.15 3.18 2.53 1.17 16 .33 68.42 11.63 (101) / ( 101 ) 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 (102) / ( 101 ) 0.43 0.29 2.81 0.20 0.18 1.30 0.00 0.06


68 Table 2 2 Relative peak intensity ratios for various surface treatments Peak ratio A1 A2 B C D E1 E2 Si(1 00) Si(111) ( 100 ) 0.25 0.04 0.2 3 0.08 0.13 0.02 0.00 0.02 (002) 2.28 2.32 2.18 3.30 2.40 4.32 4.88 4.51 ( 101 ) 0.69 0.82 0.69 0.54 0.84 0.11 0.03 0.16 ( 102 ) 1.65 1.33 1.93 0.59 0.82 0.78 0.00 0.06


69 Figure 2 1 MO HVPE Reactor A) Reac tor Picture B) Reactor Sch e matic B A


70 Figure 2 2 InN growth map from MO HVPE A) Growth etch transition temperature for InN as a function of HCl/TMI at constant V/III = 250 B) Transition temperature map with actual InN growths on c Al2O3 111,125 A


71 Figure 2 2 Continued B


72 Figure 2 3 XRD NR on Si (100) Growth conditions V/III =250, Cl/III=4.0, T= 873K


73 Figure 2 4 Various InN nanorod orientatio ns Every nanorod orientation forms specific angle with the surface 90 47 0 Nanorod Orientations (100) (001) (101) (102) 28.3


74 Figure 2 5 cleaning. Growth conditions V/III =250, Cl/III=4.0, T= 600 C for 60 minutes


75 Figure 2 6 cleaning with HCl. Growth conditions: V/III =250, Cl/III=4.0, T= 873K for 60 minutes HCl cleaning conditions :F HCl =1500sccm, T= 873K for 10 minutes


76 Figure 2 7 SEM images (1 0 000x) showing effect of surface cleaning on InN NR A ) A1: InN NR on degreased Si (100), B ) A2: InN NR nanorods on BO E cleaned Si (100), C ) B: InN nanorods on nitridated Si (100) A B C


77 Figure 2 8 for nanorods grown on Si(100) showing effect of surface nitridation Growth conditions: V/III =250, Cl/III=4.0, T= 600 C for 60 minutes Nitridation conditions:FNH3=1500sccm, T= 560 C for 10 minutes


78 Figure 2 9 SEM images showing effect of nitride layers. A) InN NR grown on nitridated Si (100), B) InN NR grown on InN film as a nucleation layer A B


79 Figure 2 10 using InN nucleation layer (Sample D) vs MOCVD GaN as nucleation layer (Sample E1) Growth conditions: Nanorods: V/III =250, Cl/III=4.0, T= 600 C for 60 minutes, InN layer: V/III =700, Cl/III=2.0, T= 600 C for 10 minutes, GaN layer: V/III =3000, Cl/III=0.0, T= 600 C for 30 minute s


80 Figure 2 1 1 with Si(100) and Si(111) as substrates. Growth conditions: Nanorods: V/III =250, Cl/III=4.0, T= 600 C for 60 minutes, GaN layer: V/III =570, Cl/III=2.0, T= 600 C for 10 minutes


81 Figure 2 12 SEM images of vertical InN nanorods. A ) cross sectional view showing InN nano/GaN/Si(100) structure B )top view Si HVPE GaN InN NR A B


82 Figure 2 1 3 SEM images of vertical InN nanorods grown on MOCVD and HVPE GaN A ) Mul tiple nucleation spot on MO GaN surface, B ) Vertical nucleations with tripods on HVPE GaN, C) Dense growth with uniform diameters on MO GaN, D) InN NR grown on HVPE GaN 120 A B C D


83 Figure 2 1 4 Summary of effect of treatments on orientations of InN NR


84 Figure 2 15. Pole figure of InNNR grown on HVPE GaN on Si (111) sample in [ 002] and [10 1] directions


85 Figure 2 1 6 rocking curve of vertical InN NR sample with Molybdenum X ray source


86 Figure 2 1 7 TEM analysis of nanorods A)lattice fringe image of nanorod in B, B) Tipped InN nanorod of 100 nm diameter, C)SAED pattern showing wurtzite pattern of nanorod in B, D)flat tipped InN nanorod with cartoon showing explanation for SAED orientation, E)Tipped nanorods with p lanar defects visible, also shows secondary nucleation. A C E D c axis (100) face B


87 Figure 2 1 8 EDS spectrum InN nanorods grown on HVPE GaN on Si (100) (Sample E2)


88 Figure 2 19 Effect of N/In ratio and Substrates on InN morphology a t Cl/In=4 and T=873K 125 100 250 500 1000 3000 7000


89 Figure 2 20 Vertical InN nanorods grown on GaN nucleation layer on Si(100) A) Conical tipped nanorods, B) Flat tipped nanorods B A


90 Figure 2 21 Heterogeneou s nucleation of hexagonal parallelepiped on substrate H L


91 CHAPTER 3 THEORITICAL MODEL FOR FINDING LIMITS OF COHERENCY FOR INDIUM NITRIDE NANORODS GROWN ON GALLIUM NITRIDE Overview Traditionally used planar hetero structures have various lim itations. They have lower surface area resulting in lower efficiencies. Due to lattice mismatches, defects like dislocations are common at hetero interfaces. Dislocations are preferred sites for impurity items. They act as high diffusivity path for dopant s and also as nonradioactive recombination centers. They can be precursors for crack formations at interfaces with higher lattice mismatch. Nanostructures such as nanowires and nanorods, on the other hand, are going to be essential building blocks of the s emiconductor technology. Due to properties like high surface area and single crystalline nature, nanostructures are one of the best candidates for applications like light emitting diodes (LEDs), laser diodes (LDs), detectors and solar cells. Various resea rchers have demonstrated formation of highly crystalline nanostructures despite of high lattice mismatches between substrate and deposited material 111,112,117,120,138,139 Moreover the use of highly crystalline nan ostructured templates for improving the quality films in heteroepitaxy has been demonstrated by various researchers 89 92,94 96 The same was true in case of InN NR grown on GaN in previous Chapter Crystal G rowth M odes In the field of crystal growth and heteroepitaxy, there are three basic accepted modes of the growth 130 The figure1 shows schematic of these modes at different coverage. The Volmer Weber (V W) growth mode ( Figure 3 1 (A) ) is characterized by island growth. This mode is a result of stronger interactions amongst ad atoms than that bet ween ad atoms and substrate atoms. When surface energies are significant and 1 >


92 2 12 1 2 12 are surface energy of substrate, surface energy of epilayer and interfacial energy between substrate and epilayer respectively, V W growth is observed. The Frank van der Merwe mode, on the contrary, is the layer by layer gr owth mode ( Figure 3 1 (B) ). In this mode, ad atoms preferentially attach to the surface of substrate forming uniform layer. The monolayers cover surface completely before the next layer grows. This results in 2D growth resulting in smooth films. The third growth mode is intermediate case between the modes explained above. It is called Stranski Krastanov (S K) growth mode 140 In this mode, the layer by layer growth till critical thickness is followed by island growth. The transition of growth mode observed here depends on properties like surface energies, interaction energies, lattice m ismatch between substrate and film. For example, in systems such as Ge/Si surface energies do not differ significantly, interfacial energy is not significant but there is a large lattice mismatch between the two. In such case, initially layer by layer gro wth occurs, which introduces strain energy. When strain energy reaches certain critical limit, it can relived by formation of dislocations or switching to V W growth (island formation) or both. Thus, in semiconductor systems, that have large lattice mismat ch, but similar structures, S K growth mode is observed e.g. Ge/Si 141 InA s/GaAs 142 In the S K mode, the island formation might be coherent or dislocated. In coherent S K, the 3D island gro wing remains coherent below certain critical size. This can occur because the interface region where islands form remains strained with some radius of curvature. Moreover, dislocations can be introduced either in layers or islands if islands are close toge ther or if islands exceed critical size.


93 In heteroepitaxy, the growth of epilayer can take place ideally in basically two ways viz. dislocated film growth and pseudomorphic film growth. Due to lattice mismatch the dislocations tend to form resulting in in terfacial dislocations or dislocated film. This is the first type of growth. In later case, however, due to epitaxial nature of the growth, epilayer grows pseudomorpically or epilayer is homogenously strained such that its lattice parameter matches with th e substrate. These two extremes are shown in Figure 3 2 for the case of InN film growth on GaN film. Ideally, observed growth tends to be the combination of both modes. That is, first few layers tend to grow pseudomorphic to the substrate. In these layers substrate and epilayer at interphase are coherently strained. This strain is stored in the form of strain energy. This pseudomorphic growth continues, until it is energetically favorable to form dislocation to relieve some strain. Over the years there hav e been various theoretical developments based on this approach that have helped engineering hetero interfaces. Initial theoretical developments by J. H. van der Merwe form fundamental approach to explain crystal interfaces in heteroepitaxy 143 146 In 1975, Mathews improved this model further by adding mesh of non interacting perpendicular dislocations to calculate critical thickness of coherency for 2D film growth 147,148 In nanostructures, however, approach that is applied to 2D film growth is not valid as it is. Extension of approach to calculate stresses in bimetal thermostats 87 to heteroepitax y on patterned substrate predicted reduction in stress and higher coherent thickness of epilayer 88 It showed that the finite dimensions of the nanostructures, the strain is much reduced as a result of lateral relaxation away from interface. This theory formed basis of nanoheteroepitaxy 89 which has been widely used since, to reduce the defects in epilayer 90 92,94 96


94 Strain vs. D islocation M odel for H eteroepitaxy On the similar lines of nanoheteroepitaxy model 89 147 model for nanostructures was developed 149 The model was primarily developed for cubic material case which is modified for hexagonal or wurtzite crystal structure here. As discussed earlier, nanorods, due t o the 1D nature of growth, tend to relieve the strain (arising due to lattice mismatch) both laterally and vertically away from interface unlike 2D films. As a result nanostructures tend to be single crystalline over wider dimensions. The heteroepitaxial f ilms, on the other hand, form dislocations. This model captures this behavior of nanostructures. The model assumes strains in all three directions unlike 2D film growth models. The strain energy is function of all these strains as well as dimensions of na nostructures such as diameter and height (if nanostructures are cylindrical). For a given strain, below certain diameter nanostructures can be defect or dislocation free for any given height. This coherency limit for nanowire diameter is out of reach for c onventional 2D film models that predict critical thickness of the films. The same model can be used to predict the critical coherent thickness by assuming very large radial spans. There are various assumptions in the model in order to simplify model for system under study. The most important assumption in the model is that the principles of linear elasticity are applicable. That is, both substrate and epilayer are assumed to be continuum and not discrete atoms. Also, at each point in continuum each compo nent of stress tensor can be expressed as linear combination of components of strain tensor. Since almost every function is linear at infinitesimal level, linear theory of elasticity can be assumed to be applicable provided deformations are small. Before discussing other


95 assumptions, it is important to revise the relationship between stress and strain as dictated by linear theory of elasticity. Formulation of S tress S train R elations in Hookian T ransversely I sotropic S olids as Ut tensio, sic vis, meaning, "As the extension, so the force". In other words, in elastic solids stresses are directly proportional to strains and each component of stress tensor can be expressed as linear combination of components of strain tensors. In tensor form it can be stated as: ( 3 1 ) where ij = stress tensor components, kl =strain tensor components and c ijkl = elements of stiffness constant tensor. Since the units of the stress tensor components are N/m 2 and th e strain tensor components are dimensionless, the units of the elasticity tensor components are N/m 2 Since c ijkl tensor relates two second rank tensors, it has 3 4 =81 components. But due to symmetry of stress tensor ( ij = ji ) and strain tensor ( ij = ji ), only 36 independent components remain to completely describe elasticity tensor. Thus, the relationship between stress and strain in matrix form is given by: (3 2) The elasticity tensor is further simplified depending on sy mmetry of the material. The materials InN and GaN have wurtzite structures. The wurtzite materials fall under


96 the category of transversely isotropic materials. Transversely isotropic materials are characterized by plane of isotropy and its properties are s ymmetric about an axis that is perpendicular to plane of isotropy. The basal plane of hexagonal lattice (wurtzite material) exhibits six folds symmetry. That is every 2 axis we get same structure with same properties. mathematically the transformation of a vector in space can be done using direction cosine matrix given by A: (3 3) Here, (a 11 a 12 a 13 ) are direction cosines of respectively. Similarly, (a 21 a 22 a 23 ) and (a 31 a 32 a 33 The transformation matrix for elasticity tensor matrix however is more complex as it involves tensor. It can be shown that it is given by M: (3 4) other a s


97 (3 5) tensor C TRANS is equal to actual elasticity tensor C. Thus, the transformation of elasticity tensor results in only five independent elements for stiffness const ant tensor viz. C 11 C 12 C 13 C 33 and C 44 As a result, in case of transversely isotropic materials the relationship between stress and strain can be given as: (3 6) If material shows rotational invariance about an axis fo r an angle of 2 about that axis 150 This implies that transversely isotropic materials like InN and GaN have equivalent properties along any direction in hexagonal basal plane (Figure 3 3(A)). Thus, these materials can be model ed with cylindrical coordinates with uniform properties in circular bases and symmetry around transverse z axis (Figure 3 3(A)). Formulation of M odel for N anorod Heteroepitaxy Rigid Substrate Model The Figure 3 4 explains model for nanostructure growth tha t allows for lateral relaxation. The Figure 3 4(A) shows InN nanostructure growing in < 000 2 Error! Bookmark not defined. > direction on GaN film. It is highly strained at interface as a


98 result of lattice mismatch. If the nanorod g rows pseudomorphically like 2D film ( Figure 3 4(B) ), then it will result in high levels stress and strain. This strain if possible will be partially relieved by edge or mixed dislocations. As a result, it is not possible to grow very long nanowires without dislocations in pseudomorpic mode. This is contrary to experimentally observed results. From literature it is also known that nanostructures have better ability to accommodate strain as compared to film because of lateral relaxation. As a result nanostru ctured is homogeneously strained and strain is relaxed as nanostructure grows, with negligible strain in bulk material ( Figure 3 4(C) ). This results in single crystal nano structures with very high aspect ratio such as nanorods and nanowires. Incorporatio n of lateral relaxation in model should be able to explain existence of very large aspect ratio single crystal nanowires. All nanowires or rods are assumed to be cylindrical to apply cylindrical coordinates. When the nanostructures grow, the epitaxial laye r is assumed to be pseudomorpic at interface i.e. the lattice of epilayer is exactly mapped on to the substrate. To allow for lateral relaxation away from interface, all the strains are assumed to exponentially vanish away from the interface. As a result, there is negligible strain in bulk material. This is achieved by incorporating characteristic relaxation length properties of material. Also, the interaction between strains fields of different nanorods are considered to be negligible. This can be assumed if nanorods are not very close to each other. Bo th the cases of rigid as well as flexible substrate are considered. First strain energy of coherent nanorod is calculated which is the function of elastic constants, lattice mismatch, strain distribution and dimensions of nanostructure such as


99 radius and height. The nanorod initially grows coherently on the substrate with interface exactly mapped like substrate. As a result this nanostructure is coherently strained and strain is stored in the form of strain energy. The nanorod can continue to grow this way until it reaches certain critical diameter R for given height and lattice mismatch. The radius R* is such that above this radius, system can deform and introduce a defect such as dislocation to partially relive its strain as well as minimize strain energ y. In this model, strain energy of coherent system (Ec) and dislocated system (E) are calculated as a function of R for known H as well as misfit f. The radius at which Ec becomes more than E is coherent radius for that H and misfit f. The model is descri bed below in detail in general terms. The epilayer material is respectively. First, calculation of coherent strain energy is discussed. Coherent s train e nergy (Ec) The lattice mismatch between substrate and epilayer is given by (3 7) The radius and height of nanorod are denoted as R and H respectively. Initially, epilayer will be coherently mapped onto the substrate. Thus, for a nan orod of coherent radius R strain will be stored only as strain energy. The uniformly stretched interface in radial direction and decaying displacement along Z axis or growth direction can be represented as follows:

PAGE 100

100 Here, f represents the total lattice mismatch, are constants related to deformation in radial directions. The unknown constants (P, Q tends to go to state of lowest energy, and c onstants will have value such that energy of system is minimized. Based on the displacements, the strains in all the directions are The stresses are related to strains through equation ( 3 6 ). Thus, stress vector is given by:

PAGE 101

101 (3 8) The hydrostatic stress field and stress field is shown in Figure 3 5 As it follows exponential decay in about the stress decreases to less than 5% of maximum stress. The strain energy density of nanorod is given by J/m 3 (3 9) J/m 3 Thus, integrating strain e nergy density over the volume of nanorod gives total strain energy of the system J This function was minimized numerically to get the values of parameters P, Q and nt of R and H as long as the two are not 0.0521,

PAGE 102

102 0.0149 and 0.1120 respectively. Minimized coherent strain energy for nanorod of radius 100nm and height 1000nm was found to be 1.91 25 x 10 12 Joules. Energy of d islocated s ystem (E) As discussed before the radius of nanorod increases above certain critical value R* system no longer remains coherent as it is energetically favorable to form dislocation. When energetics is favorable, pe rfect coherency or pseudomorphism between substrate and epilayer is broken by local degeneracy. The degeneracy is introduced by extra half plane in one of the layers. In Figure 3 2(A) for example the GaN substrate has two extra half planes (one shown by do tted yellow line). This can be also viewed as if half plane from InN layer is squeezed out due to the compressive stress. This allows for relaxation for partial relaxation for upper layers. This type of degeneracy is viewed as negative dislocation along direction, the line perpendicular to the direction as shown in figure2a of strain relief here. These types of dislocations are called edge dislocations in which burgers is along the line of dislocation are called screw dislocations. The se types of dislocations should be perpendicular to the line of dislocation as well as it should be in or parallel to interfacial plane. For example, in growth of InN on GaN both growing in direction, for dislocation line along direction, edge dislocation with direction will not relieve any strain. But an edge disloc ation direction will relieve strain.

PAGE 103

103 In this study, perpendicular sets of pairs of dislocations in direction and direction are assumed to be formed when degeneracy is introduced in coherent systems. These types of dislocations are common in hexagonal crystals 151 153 The dislocation is the introduction of localized regions of degeneracy in order to accommodate lattice mismatch strain. It requires energy to form, which is again related n pairs dislocations in direction is given by 153 (3 10) Here, b i T o minimize dislocation energy, dislocations tend to form with lowest magnitude of b possible. As a result, dislocations in closed packed direction and closed packed planes are common. In our system also, dislocations form in closed packed direction and closed packed direction. Due to fo rmation of dislocation strain is partially relaxed. The residual strain in the nanorod of radius R is (3 11)

PAGE 104

104 This strain is again stored in nanostructure as residual strain energy. Thus, total energy for dislocated syste m (E) is summation of residual strain energy and energy to form a pair of dislocation. (3 12) (3 13) As energy of dislocation only partially relieves the strain, the constants P, Q and do not change. This is also verified by again optimizing the total energy function for dislocated system. The Figure 3 6 shows the curves for coherent strain energy (green curve) and total energy of dislocated system (red curve) very high aspect ratio i .e. very large height. Initially E is greater than Ec and as a result it is not energetically favorable to form a dislocation. But as radius increases E becomes smaller than Ec and system forms dislocation rather being coherent. The coherent radius for sys tem of InN nanorod on GaN substrate was found to be 5.25nm. It means that InN nanorods with almost infinite length would be possible to grow coherently if their diameters are less than 5.25nm on GaN film purely. It is expected then that if height of nanor od is decreased then for nanorod of even higher diameters are possible to form coherently. This is indeed observed in the model. When at a constant radius coherent strain energy and total energy of dislocated system are plotted, for nanorods with radius 4n m, E is always greater than E c As a result at this

PAGE 105

105 radius any height nanorod would be coherent ( Figure 3 7(A) ). But for nanorod with radius of 6nm for examples if height is more than 2nm it will not be coherent ( Figure 3 7(B) ). The heights vs. radius can thus be plotted for various radius and heights. It is shown in Figure 3 8 Below the curve nanorods will be coherent and grow as single crystalline almost in island or Volmer Weber growth mode. Above the curve it is expected that nanorods will either have dislocations or will grow in Stranski Krastanov growth mode. It has been observed that nanorods with much greater diameter can be coherently on GaN. In S K mode, InN strained and dislocated film first forms on the GaN film and then 3D growth starts. Due to growth on partially relaxed film of same material, much larger coherent diameter is energetically favored. Another reason behind higher coherency diameter is the flexibility of substrate. It is not completely right to assume rigidity of substrate. As the InN grows on GaN, InN is elastically compressed. But, at the same time GaN film underneath itself is getting strained, thus sharing some part of strain. This case is discussed next in flexible substrate mode. The actual growth can be combination of both ca ses. F lexible S ubstrate M odel To incorporate the flexibility of substrate, InN nanorod (over layer) of radius R o is assumed to be growing on GaN substrate (under layer) such that area inscribed by nanorod basal plane in coherently matched to substrate lay er. (Note: Henceforth, epilayer and substrate are referred to as over layer and underlayer. They are denoted by subscripts o and u respectively.) Since actual lattice constant of over layer (a o ) and under layer (a u ) are different, former is under tensile stress and later is under compressive stress in this case. As a result, although coherently mapped at interface,

PAGE 106

106 the relaxed over layer radius of R o corresponds to the relaxed under layer layer radius of R u The values of R o and R u are related by overall l attice mismatch, given by f. (3 14) (3 15) Also, the overlayer and underlayer coherent mapping onto each other in a dislocation free system requires that at z=0 both radial and v ertical displacements for overlayer and underlayer should be equal. This condition can be represented numerically as follows: (3 16) The negative value of lattice mismatch ( f ) indicates over layer under compression. The strai n gets distributed, between overlayer and underlayer such that the interfacial lattice constant assumes intermediate value of a. The radial strains at the interface, in overlayer and underlayer, are given by (3 17) (3 18) To be consistent with rigid substrate model, equations would be written in terms of B instead of interface lattice constant a. The other variables to be optimized are taken tions constants for underlayer and overlayer respectively. As elastic properties of overlayer and underlayer material are

PAGE 107

107 very different strain relaxation in these materials is expected to happen over different lengths. As seen from equations ( 3 16 ) ( 3 18 ) overlayer which is located in o and o u and H u stretched. Following the formulation in rigid substrate model, various strains in overlayer and underlayer are given by (3 19) (3 20)

PAGE 108

108 Knowing the relation between stress and strain from equation ( 3 6 ), stress vectors can be expressed in terms of strains similar to expressions in ( 3 8 ). Strain energy densities in overlayer and underlayer can be calculated as (3 21) (3 22) (3 23) (3 24) It is to be noted here that elastic constants for overlayer and underlayer are represented by O 11 O 44 and U 11 U 44 respectively. Sum of integration of energy densities

PAGE 109

109 over corresponding volumes of underlayers and overlayers gives total coherent strain energy of system. (3 25) Similar to case of rigid substrate model, this function was minimized numerically to get the values of parameters P, Q, B, o u The minimum values of the most parameters are independent of R and H (except for B in some conditions) as long as the nanostructure on flexible GaN base are found to b e 0.0240, 0.0088, 0.0585, 0.1039 and 0.1171 respectively. Minimized coherent strain energy for nanorod of radius 100nm and height 1000nm was found to be 9.8357 x 10 13 Joules. It is to be noted here that values of P and Q are approximately half of that in rigid substrate model. The value of o although relaxed a little bit, has value similar to that in rigid substrate model. This can be explained easily by understanding the physical

PAGE 110

110 significance of these values. Due to flexible nature of substrate, effective lattice mismatch at interface is given by B instead of f. Thus, lattice mismatch experienced by overlayer in flexible substrate case is 0.0585 instead of 0.1079. This is the first factor that affects the values of the parameters P and Q that decide mapping at the interface. Since mappin g structure is the same, mapping parameters P and Q decrease proportional to the effective lattice mismatch at interface. Secondly, due to incorporation of flexible substrate, a part of strain is shared by underlayer. Thus, values of P and Q are also affec ted by elastic properties of both overlayer and underlayer instead of only other hand depends on elastic parameters of materials. Thus, this value remains constant. A lso, Figure 3 9 illustrates that value of total coherent strain energy when flexible substrate is assumed is decreased substantially. It means that flexible substrate structure will be better relaxed than the rigid substrate case. In coherence with rigid s ubstrate model, perpendicular sets of pairs of dislocations in direction and direction are assumed to be formed when degeneracy is introduced in coherent systems. The energy to form dislocation is given by following equation similar to equation ( 3 10 ). J (3 26) From equation (3 26), it is clear that dislocation is assumed to be introduced in the overlayer. This assumption is done on the basis that dislocation is introduce d while nanostructures grow as overlayer. As a result of formation of a pair of dislocation,

PAGE 111

111 however, total lattice mismatch is relaxed from f to f r given by equation ( 3 27 ) below which is exactly same as equation ( 3 11 ). (3 27) Again, condition that o f) assures that formed dislocation cannot e constant is dynamic and depends on the optimum value of B through equation s ( 3 17 ) and (3 18) In flexible substrate model, introduction of dislocation changes the total lattice mismatch which changes B to B r Dislocation energy in turn depends on depend s on changes in B. This interdependence of Ed on B and vice versa requires B to be optimized again on introduction of dislocation. After dislocation forms total energy of system is, thus, given by equation as follows (3 28) The expression total energy of dislocated system can be easily calculated by substituting for f r from equation (3 27) and B=B r in equation (3 25) and knowing expression for E d from (3.26). The actual expression for E will not be given here because o f length limitation of word processor. The values of all the parameters except B remain constant except at the very small values of radius. At smaller radii, dislocation formation can compensate lattice mismatch substantially. This reflects into large devi ations in P and Q values. But, at the same time for smaller radii, coherency at interface is energetically favored. As a result,

PAGE 112

112 dislocations do not tend to form at smaller radii and these deviations in values of P and Q do not affect the results. From opt imized values of B, the lattice parameters of relaxed overlayer and underlayer can be calculated using relations in equations ( 3 17 ) and ( 3 18 ). They are plotted in F igure 3 10. It shows that for smaller radius nanorods formation of dislocations will relie ve greater strain and lattice constant on both sides will tend to relax to their bulk lattice constants. But the formation of dislocation itself can be seen as introduction or reduction of plane of atoms in the material and it introduces local strain field s. This requires energy equivalent to dislocation energy. The F igure 3 1 1 shows plot of total energy of coherent as well as dislocated system for the case of flexible substrate model. Below 10.7nm, coherent strain energy is lower than strain energy of disl ocated system. As a result below 10.7nm system will be coherent. So, critical coherency diameter for InN nanorod in case of flexible substrate model is 21.4nm as compared to 10.5nm in case of rigid substrate model. The increase in critical coherency radius is due to the flexibility of substrate. It is also illustrated in F igure 3 1 2 The F igure 3 1 2(A) and F igure 3 1 2(B) show the plot of height of nanorod versus strain energies of dislocated as well as coherent systems for nanorod radius of 10nm and 12nm re spectively. Since 10nm is below critical coherency limit, coherent nanowire with any height is possible in this case. On the other hand, if nanostructure height exceeds 4nm for 12nm radius nanorod, dislocations would be introduced. Thus, for every radius, there is a critical height below which structure can be coherent. All such heights versus radii are plotted in F igure 3 1 3 The figure also compares the results for two approaches. The

PAGE 113

113 introduction of flexibility of substrate means more compliance with ove rlayer that result in increased coherent diameter. S K growth and its effect on coherent diameter The S K growth mode is simply a film growth followed by the 3D island growth. When two like compounds are grown epitaxially, initially different surface inter actions are such that 2D film growth dominates. After few layers, however, strains arising due to lattice mismatch make 3D growth more favorable. In addition to that, 3D growth also provides both lateral and vertical relaxation. As a result, 3D nanostructu res can grow strain free by acquiring S K growth mode. The S K growth can then be incorporated in given model by assuming lower lattice mismatch between substrate and epilayer. This is because initially epilayer grows either coherently or dislocated mode on the substrate. Depending on value of ( and grain size ( D ), strain is reduced below 10% of original value within distance equivalent to For example, if InN film with grains of 100nm is formed on GaN, then within 30nm the strain would be below 10% of original. This film then can be precur sor to nanostructures with at least 100 nm diameters with no dislocations. The F igure 3 1 4 shows R H coherency map for S K mode growth with different lattice mismatch. Below the curve the nanostructures tend to be coherent for any height. Above the curves the nanostructures will be dislocated. In S K mode, by formation of dislocated film has duel effect on heteroepitaxy. Firstly lattice mismatch is reduced and as a result strain is reduced. Secondly, materials become similar in terms of elastic properties, thus stress is distributed uniformly amongst the layers. The exponential length. It means that for InN with grain size of 100 nm, strain will be 0.3679 f within 30 nm. From

PAGE 114

114 F igure 3 1 4 it is clear that for S K growth with 0.3 lattice mismatch, more than 100nm radius coherent nanorod growth is possible for InN on GaN system. However, no S K growth mode was visible in the InN NR grown in Chapter 2. This can be clear from Figure 3 15 which is a cross sectional SEM image of InN NR grown on Silicon as well as GaN buffer layer. But nanorods grown have very good crystalline quality. As seen from the model it is clear that nanorods will tend to form dislocations at very early stage of growth If nanorods form pair of dislocations at 5 nm radius the relaxed nanorod would be able to grow to 20 nm radius without further dislocation formation. Al so, as seen from the strains plotted from the model, it is clear that nanorod is mostly strained at t he bottom (Figure 3 16) As a result dislocations tend to be concentrated only at the bottom center of nanorod 154,155 Result of this is maximum relaxation at the interface and grown nanorod is defect free. The s im ilar results are predicted by the theoretical model. Summary and Conclusions This Chapter presents a model for heteroepitaxial nanorod growth for transversely isotopic materials. The model predicts limit of coherency of particular nanorod growing heteroep itaxially. Unlike films, nanorod basal planes that are epitaxially bound to substrate are very small. All other faces of nanorod are free to expand and contract as no stress is applied to them on any free faces. As a result, physics of the problem requires them to relax laterally as they grow The model characterizes relaxation by relaxation constant This relaxation constant is not arbitrary, but is decided by material properties along with other parameters P and Q The model employs the minimization of free energy approach to optimize parameters and make predictions. Every system

PAGE 115

115 tends to attain least energy state. A strained nanorod can attain minimized energy state with either growing coherently or forming misfit dislocation. Misfit dislocation can be viewed as missing plane or extra plane of atoms introduced in either epilayer or substrate. When dislocation forms it creates dangling bonds as well as localized strain field. As a result, dislocation to form requires some energy. The model easily predicts state of system by energy minimization approach. Model suffers few disadvantages. It generally overestimates critical diameter. Although, model overestimates the predicted coherent region as compared to literature, it presents a qualitative picture of str ain relaxation in nanostructures. It predicts rapid lateral relaxation of strain in the nano material, such that bulk is almost strain free. Secondly model should not be used for predicting formation of second sets of dislocations as once dislocation form linear elasticity model does not hold. But baring that model perfectly predicts formation of dislocations only in basal region, which is also observed experimentally. Model is also able to qualitatively describe how SK growth mode, which is common in heter oepitaxy, is able to increase coherency in the epilayer and reduce dislocation density. In all, this model is very general and can be applied or modified for any system in the heteroepitaxy.

PAGE 116

116 Figure 3 1. Modes of epitaxial growth: A) I sland or Volmer Weber growth mode, B) Film or Frank van der Merwe growth mode and C) Stranski Krastanov growth mode 130 A B C

PAGE 117

117 Figure 3 2. Growth of InN film on GaN film viewed in direction A ) Dislocated growth with pair of dislocation at interface, yellow arrow shows Burgers vector in direction B ) pseudomorphic growth with homogeneously strained InN film A B

PAGE 118

118 Figure 3 3. Schematic of InN nano structure growth on GaN film (with possible InN film): A) hexagonal nanorod geometry with growth in z direction, hexagonal plane i s isotropic in properties, B) equivalent cylindrical geometry used for model A B

PAGE 119

119 Figure 3 4. InN nanorod growth on GaN film A) InN experiences compression at interface, B) nanorod pseudomorpic growth with high compressive strain in crystal, C) nanorod has ability of lateral relaxation results in no strain in bulk(model), is relaxation length A B C

PAGE 120

120 Figure 3 5. Stress field in nanorod. Figure shows both lateral as well as longitudinal contour. Overall stress is at its maximum at the center. There is no stress in bulk due to lateral relaxation over length of

PAGE 121

121 Figure 3 6. Strain Energy vs. nanorod radius: Red curve is total energy of dislocated system; green curve is coherent strain energy

PAGE 122

122 Figure 3 7. Strain Energy vs. nanorod height: Red curve is total energy of dislocated system, green curve is coherent strain energy, A) for R = 4nm nanorod is coherent at any height, B) for R=6nm nanorod is coherent only below 2nm height A B

PAGE 123

123 Figure 3 8 Coherency map for InN nanorod on rigid GaN In area above the curve dislocations will tend to form or system might acquire S K growth mode to increase compliance of substrate.

PAGE 124

124 Figure 3 9 Strain Energy vs. nanorod radi us for flexible substrate model. C urves show coherent energy for rigid substrate case is much more than flexible substrate case. Black line: Total coherent strain energy with rigid substrate assumption, Gold line: Total coherent strain energy with flexible substrate assumption, Blue and green lines are strain energies in overlayer and underlayer respectively

PAGE 125

125 Figure 3 10 Interfacial lattice constant vs. nanorod radius for flexible substrate model

PAGE 126

126 Figure 3 11 Strain Energy vs. nanorod radius (flexible substrate): Red curve is total energy of dislocated system; Green curve is total coherent strain ener gy

PAGE 127

127 Figure 3 12. Strain Energy vs. nanorod height (Ho) (flexible substrate): Red curve is total energy of dislocated system (E); blue curve is coherent strain energy (Ec) A ) for Ro = 1 0 nm nanorod is coherent at any height, B ) for Ro=1 3 nm nanorod is coherent only below 3.5 nm height R o =9nm H u =1000nm R o =13nm H u =1000nm A B

PAGE 128

128 Figure 3 13. Coherency map for InN nanorod on GaN. Red and blue curves are for flexible and rigid substrate case respectively. In area above the red curve dislocations will tend to form or system might acquir e S K growth mode to increase compliance of substrate.

PAGE 129

129 Figure 3 14. Coherency map for InN nanorod on GaN in S K mode. f S K = f f S K = 0.7f f S K = 0.5f f S K = 0.3f

PAGE 130

130 Figure 3 15. SEM and COMPO images showing no evidence of films below InN. A)InN NR initial stage on GaN fi lm, B) InN NR directly growing on Si B A

PAGE 131

131 Figure 3 1 6 Relaxed strain in nanorod bulk. A) Longitudinal strain, B) Radial strain, C) Shear strain A B C

PAGE 132

132 CHAPTER 4 GROWTH OF GALLIUM NITRIDE ON SILICON IN MO HVPE GaN Using InN in Litrature In the field of III V semiconductors GaN has gained lot of importance over last two decades. This is as a result of several material properties such as wide direct band gap of 3.4eV, high temperature stability, high mobility exhibited by GaN. These properties are in fact perf ect for variety of applications from public applications to military purposes. GaN has been grown on several substrate materials, among which sapphire is most widely used for several applications such as blue light emitting diodes and laser diodes. With em ergence of GaN as very important material for consumer applications, integration of GaN with silicon has beco me problem of great interests. This is because Silicon is available in fairly large diameters and relatively lower cost. Silicon also has better el ectrical and thermal conductivities as compared to sapphire. Silicon already has large share of market in semiconductor industry. As a result the production processes for Silicon devices are well developed. This has been a major driving force behind integ rating GaN with Silicon. On sapphire substrates, AlN is the most widely used buffer layer to grow GaN which has produced very good results Similar to sapphire, GaN on Si with AlN as buffer layer or by graded AlGaN layers as buffer has been attempted by va rious researchers 56 59,65 67,125,156 AlN has good lattice match with GaN, but as GaN growth on AlN goes through island growth and coalescence, often dislocation originate at grain boundaries due to slight orientat ion differences in individual grains. Earlier study also observed that due to strong bond between Si and GaN via AlN, often cracks due to high tensile strain in GaN propagate into the silicon substrate.

PAGE 133

133 Use of InN as buffer layer for providing weak bond be tween Si and GaN is very promising. InN has lattice constant between Silicon and GaN which makes it better match with Si as well as GaN. InN has low decomposition temperature of around 873K to 973K, which is perfect temperature to grow low temperature GaN 73 Just below 873K, InN bonds are weak and it can be very effective in relieving stress in GaN grown on top. Also, at higher temperatures InN can decompose into the elemental Indium and nitrogen, to reduce the bond between GaN and silicon by forming an airgap. Indium can diff use into GaN forming graded InGaN. Indium is shown to have alloy formation capability with GaN over entire compositional range previously 12 6 Both of these things can be ef fective in reducing the strain. Use of InN film as a buffer layer, getting promising result with sapphire substrate has been reported previously by various researchers 48,49,157 Thi s has been employed to GaN growth on Si by MBE more recently 158 No one has employed use of InN nanostructures as template for growth of GaN yet except previous students from our grou p 125 although use of nanostructure as an effective way of reliving strain had been shown by many researchers working on NHE 89,95,96 Second Chapter was focused on growth of vertical InN nanorod growth on Silicon substrate. In this Chapter growth of high quality thick crack free GaN on Si using InN nanorod templates will be discussed in details with all intermediate steps. Growth of GaN in MO HVPE Reactor As seen previously Metal Orga nic Hydride Va por Phase Epitaxy (MO HVPE) is hybrid between conventional Metal Organic Chemical Vapor Deposition (MOCVD) and Hydride Vapor Phase Epitaxy (HVPE). It has three metal organic source lines, two out of which are in use currently. These metalorga nic sources can be quickly swapped as

PAGE 134

134 well as turned to vent allowing rapid reactants switching. This makes MO HVPE more flexible over traditional HVPE where metal sources cannot be transient. Reactor Setup The reactor schematic for growth of GaN in reacto r is similar to that for InN. It is shown in Figure 3 1 Similar to the InN growth, reactor is capable to be operated with Nitrogen (N 2 ), Hydrogen (H 2 ), forming gas (4% H 2 balance N 2 ) and Helium (He) as sweeping and carrier gases. Typically, 4%H 2 was used as carrier gas for Ga N growths especially at high temperatures to reduce any contamination by oxygen High sweep gas flows from inlet and gate valve are adjusted so as to minimize wall depositions and confining reactions to growth zone. Typically, growth zone was maintained between 873K and 1173K for GaN growths. As seen in Chapter 2, same reactor can be operated with or without HCl. In both cases GaN forms with different mechanisms. These two modes of operation are explained briefly here MO HVPE O perati on : HVPE mode Similar to the case of InN, temperature profile is adjusted in such a way that end of the inlet zone is maintained at 573K to 8 73K. Generally, inlet is kept at higher end temperatures here for high temperature growths. This temperature profil e with shorter metal source tube ensures complete decomposition and reaction of metal organic to form chloride species. This temperature profile also ensures activation of other reactants before they enter the mixing zone. The typical overall reactions tha t are expected in HVPE of GaN are as follows. ( 3 1) ( 3 2)

PAGE 135

135 (3 3) ( 3 4) (3 5) Trimethyl Gallium c oming out of innermost tube reacts with, 10% HCl to form either GaCl or GaCl 3 as seen from equations (3 1), (3 2) and (3 3). Generally, with the temperature at the end of the inlet higher then 673K, GaCl is expected to be main Gallium specie. The reaction is generally carried out in presence of some Hydrogen. Introduction of H 2 is found to be helpful in two things. Firstly, it reduce s oxygen contamination of GaN film or Gallium oxide formation due to trace amounts of oxygen available in the reactor Secondl y, equilibrium calculations show that it can get rid of carbon contamination of the film 125 The absence of H 2 however, does not affect GaCl formation. The chlorides GaCl or GaCl 3 react with ammonia to form GaN. MO HVPE O pe ration : MOCVD mode In absence of HCl gas, the reactions proceed in completely different mechanism, similar to that observed in conventional MOCVD. In this mode, metal organic undergoes irreversible thermal decomposition and then reacts with ammonia to form GaN. In this mode since chloride species reaction with hydride is not present, it is highly dependent on active nitrogen and generally higher NH 3 /TMG molar ratios are used to avoid metal contamination and droplet formation in the film. The overall reactio n in absence of HCl can be summarized as shown in equation (3 6) below. (3 6)

PAGE 136

136 Gallium Nitride G rowth in MO HVPE GaN Growth Conditions Trimethyl Gallium (TMG) from Epichem and 99.999% pure amm onia from Airgas South were used as G allium and N itrogen source respectively. For Ga N growth TM G was reacted with 10%HCl (balance nitrogen) from Airgas South in source zone at 6 73K to form chlorinated species of Gallium These species were then mixed with ammonia from the concentric inlet in mixing zone. The substrate temperature was maintained at 833K 1173 K for growth. The inlet HCl/TMI molar ratio and NH 3 /TMI molar ratio were maintained at 2 and 570 respectively. Ultra high purity Nitrogen or 4% Hydrogen (balance Nitrogen) were used as carrier gas es or sweeping gases to avoid wall depositions. The carrier gas flow rate of 1600 sccm was also maintained throughout the runs When MO CVD conditions were used, ammonia flowrate was ramped up so that NH 3 /TMG rati o was 2000 Various substrates such as bare Si (100) and Si (111), Indium coated Si (100) were used to grow both low and high temperature GaN. The main focus of this study was to investigate various stages of growth of GaN on InN template. GaN Growth Stage s The total growth of GaN on Si involves various stages. The temperature time ramp is shown in Figure 4 2. The stages upto the growth of InN NR are covered in Chapter 2. This Chapter covers growth of GaN on InN template. It involved two different stages. Th ey are as follows: Growth of low temperature GaN capping layer for InN nanorod template Indium nitride vertical nanorod template was grown as discussed in detail in Chapter 2 at optimum conditions of HCl/TMI and NH 3 /TMI ratios of 4 and 250 respectively at 873K. Once templates were reproducible, growth of GaN was carried out

PAGE 137

137 in the same run without taking out the InN templates. This was done to avoid any exposure of InN NR to the atmosphere. Although oxidation of InN in air is not fast, year old samples only sometimes showed amorphous In 2 O 3 few nanometers thin layers on nanorods 126 This can have detrimental effect on further GaN growth. Indium nitride starts to decompose as temperatures approach 923K. As a result, it is necessary to grow GaN at low temperature first to take advantage of highly crystalline nature and flexibility of InN nanostructure d template It was shown by previous studies th at good quality of GaN growth is possible at low temperatures using MO HVPE technique 159 These optimum conditions were verified and used to grow low temperature GaN (LT GaN) capping layer. The capping layer was grown for various times for 300 to 900 seconds growths. Growth of thick high temperature GaN After first stage ensured the comp lete coverage of InN NR template, temperature was ramped up to 1123K and GaN was grown. Typical conditions used for growth are HCl/TMG molar ratio of 2 and NH 3 /TMG molar ratio of 570 at TMG flow rate of 1 to 2 sccm. After completion of growth reactor was co oled down at desired rate and ammonia was kept flowing until temperature reaches below 773K. All the materials are characterized for crystallinity by XRD and TEM, morphology by SEM and composition EDS. R esults and Discussion As seen in section Chapter 2 InN nitride nanorod template with vertically oriented nanorods was grown on Silicon The vertical nature of nanorod is confirmed by x ray diffraction (XRD) pattern and secondary electron image from scanning electron microscope image. The XRD pattern sugges ted presence of vertical nanorods because

PAGE 138

138 of presence of only InN peak absence of peaks for any other orientations such as SEM images and pole figures also showed that although vertical, nanorods do not have any particular rotational relationship between Silicon substrate. The select ed area electron diffraction (SAED) pattern is good indication of crystalline nature as well as it can be used to confirm the crystalline structure of material. The SAED pattern of LT GaN under layer shows ring pattern indicating highly textured wurtzite GaN (Figure4 3). It means that GaN is crystalline but not a single crystal. This is expected as GaN is grown at low temperature. The SAED of upper InN nanorods shows spot pattern indicating highly ordered planes or single crystalline nature of InN nanorods. The pattern indicates that the single crystals have wurtzite structure growth directio ns. Properties of L ow T emperature GaN C apping L ayer Due to high crys talline quality of InN nanorods, lower lattice mismatch as compared to silicon, similar crystal structures, it is expected that GaN growth on top would have better crystalline quality As known from previous section, d ue to lower decomposition temperatures of InN, GaN cap layer is grown at 873K on InN nanorods. Figure 4 4 shows different stages of capping layer growth in both cross sectional as well as top views. The capping layer growth fo r 15 20 minutes was found to give full coverage Figure 4 4 (E) and (F) It is also clear from the pictures that all stages the coverage of GaN is uniform. The GaN grew everywhere with similar uniformity. With progress of reaction, gradual growth on walls a nd bottom surface between nanorods can be clearly seen in Figure 4 4(A) to (D). This eventually led to completely uniform cap of LT GaN. There was no indication of voids formation. Thus, the growth does not

PAGE 139

139 seem to be limited by mass transfer at interface. The g rowth rate of GaN, however, is expected to be highly sensitive to temperature. This is because, at lower temperatures, the growth of GaN is found to be in kinetically controlled regime. Only above 800C it transits from being kinetically limited to m ass transfer limited 160 All different stages of these growths were studied to understand the growth of LT GaN on InN NR Crystallinity of LT GaN capping layer Though the coverage of LT GaN was uniform the grown layer is spikey instead of smooth coverage, giving cacti like appe arance to InN nanorods. This is very clear in Figure 4 4 (A). It was an indication that LT GaN growing on InN was having secondary nucleation after initial film growth. This is typical Stranski Krastanow growth mode that is observed in semiconductors wi th higher lattice mismatch 140,141 Since lattice mismatch between InN and GaN in more than 10%, this behavior is expected. The heterogeneous nucleation behavior is dictated by various things. Surface energetics plays impotent role. Gibbs energy of format ion of heterogeneous nuclei depends on various surface forces. It is given by by following equation 134 where V= volume of nuclei (m 3 ) =molar volume (m 3 /mol) =chemical potential differenc e for nucleating specie in vapor and condensed phase (J) = areas of base and exposed faces of nuclei respectively (m 2 ) =surface free energy (J/mol/m 2 )

PAGE 140

140 =energy of adhesion ( J/mol/m 2 ) InN has lower bond strength (6 eV) as compared to GaN (8eV). Also, bond strength of In x Ga 1 x N is expected to be between these two. As a result, GaN does not have special affinity for In N surface. So growth is either expected in island form or film followed by island form. Secondly, when similar materials with different lattice constant are deposited on each other, bonds get in tension or compression. And nature of growth is also dictated by atomic potential interactions. For our case where material of lower lattice constant is deposited on higher lattice constant material, upper layer experiences tensile force. In such case, film formation is favored with dislocation formations 136 Thirdly, every material can minimize total energy in a process if it grows with equilibrium e form of hexagonal prisms and since plane is closed pack plane, direction is generally favored growth direction. As a result secondary, nucleation is expected, especially at higher growth rates and l ower temperatures. With growth in S K mode still LT GaN on InN nanorod was also expected to be textured. The powder XRD scan of sample in Figure 4 4(E) and (F) is shown in in Figure 4 5 It shows two peaks for GaN GaN and GaN peak, confirming that indeed growth of GaN was polycrystalline. The presence of peak indicates that some of the planes are also parallel to the substrate surface or horizontal direction. If GaN grows in S K mode, first few layers that form shell on outer surface of InN nanorod are expected to be epitaxial. As a result, t he diffraction peak for GaN is expected to come from the spikes

PAGE 141

141 To see if indeed this hypothesis is true, the interfacial relationship between LT GaN over layer and InN under layer, InN GaN core shell structures was studied. The core shell structures selected for the study were formed by 5 min low temperatur e growth of GaN on InN nanor ods (sample is shown in Figure 4 4(A) and (B) ). This kept shell thin enough to be studied in Transmission Electron Microscopy (TEM) TEM is good tool to study structures which are electron transparent, typically specimens below 200nm are desirab le for such studies. The TEM bright field image of typical InN GaN core shell structure is shown in Figure 4 6 In TEM, transmitted electron beam is used to image the samples. The transmitted beam intensity varies because of mainly two reasons when it passes through sample. Firstly, the atomic densities of different atomic planes vary. Thus, depending on their orientation, incoming electron beam transmits defiantly through different planes. Secondly, higher atomic number (high Z) elements hinder or scatter electron be am more than the low Z elements. The latter effect g ives darker appearance to higher Z InN core. The difference of tints in images of different spikes as well as shell was, however, mainly because of difference in orientations of GaN planes. As seen from Figure 4 6 the nanorod core is 100 nm InN core with shell of around 30nm (not inc luding the protruding spikes). In the image of the half nanorod the dark InN core can be clearly while half of the nanorod does not have any core. On closer look, all spikes on GaN shell appear to have their tips oriented in specific sets of angles viz. ( 72 3), ( 62 3) or (42 3) approximately to the vertical direction (direction of nanorod growth). The hollowness of certain parts of the core was also surprising, but i t

PAGE 142

142 would be discussed latter on. The next section discusses study on orientation and growth direction of nanorod spikes as well as 30 nm GaN film shell. Orientation and growth directions of GaN secondary nucleation spikes : To further study the growth charac teristics of GaN such as growth directions, orientations etc. the high resolution TEM analysis was done of sample similar to that in Figure 4 6 The Figure 4 7 shows high resolution image of a spike on GaN shell. As seen from Figure 4 7(B) the spike is a t angle of 62 to the shell. The Figure 4 7(C) shows the high resolution image with arrays of atomic planes visible. On any visible array, the line scan can be done as shown. The average distance between consecutive peaks can be used to determine the d spa cing between the planes. Series of line scans were carried out to determine the d spacing between the vertical and horizontal arrays of atoms seen in the image. They were found to be 0.24757nm and 0.46215nm respectively. The ideal d spacing of is 0.2438 nm. It was closest to the smaller one of observed d spacing values. But the value of 0.46215nm did not correspond to any lower index planes in GaN. Since Figure 4 7(C ) is an atomic scale image, atomic arrangements can give idea a bout the orientations and plane in view. The CrystalMaker 2.3.1 can be used to visualize atomic arrangements in space. It was clear from the crystallographic model that regular molecular pattern in high resolution TEM image looked like atomic arrangements viewed from direction. The atomic arrangements are shown in Figure 4 8 It is clear from the models that the growth of GaN in direction occurs in ABABAB form i.e. with alternating layers of Ga and N an d after every two layers (AB) the structure repeats itself. The Figure 4 8(C), also shows the ball and stick model

PAGE 143

143 superimposed on actual lattice fringe image The model matches actual lattice fringe image to the scale. Thus, the spikes like structures on GaN shell were GaN secondary nucleation growing in direction. It was, thus, concluded that the lattice spacing of 0.24757nm corresponded to planes and other to the d spacing. This particular nano spike was growing at an angle of 62 to vertical nanorods. Since the growth direction of nanorod was direction it made parallel to the horizontal plane. In other words in powder 2 XRD pattern, it would contribute to diffraction peak Similar analysis on other spikes at other angles can be done. In GaN crystal the angle between direction and is 43. 23 and that between and directions is 73. Thus these spikes would contribute to and peaks respectively. It should be, however, noted here that the 62 spikes can also appear to be at higher or lower angle if nanorod on TEM grid is rotated towards or away from viewer respectively. As a result there can be apparent reduction or increase in d spacing also. From direct measurement of d spacing on vari ous lattice fringe images, lattice constants can be estimated. For hexagonal system the d spacing of plane is related to ( h k l ) indices and lattice constants through following relationship. Knowing the ( h k l ) indices and d sp calculated. The calculated values of lattice constants from all different spikes were consistently smaller than the actual values. Their average value was found to be 0.303 4nm and 0.4939nm respectively. The ideal v alues of lattice constants for GaN are

PAGE 144

144 a = 0.3186nm and c = 0.5189nm. It should be noted here that, the observed c/a ratio of 1.6280 was similar to ideal value of 1.6262 for GaN. As a result, t he apparent reduction in the measured value by exactly 5% could have been because of either error in TEM calibration or the slight tilt of the nanorod. Growth direction of GaN shell and its epitaxial relation with InN core : Similar to the secondary nucleation sites, lattice fringe analysis of GaN shell on InN nanorod was done. The Figure 4 9 shows lattice fringe image of shell area of structure. The observed d spacing from shell was 0.25157nm which again corresponded to the planes. To confirm the growth direction and epitaxial relationship wi th InN core selective area electron diffraction was done at hollow shell as well as core shell tip with InN core still intact. The SADPs (Selective Area Diffraction Patterns) are shown in Figure 4 10 These SADP patterns were superimposed with SADPs genera ted by kinematical simulations 161 generated by Web EMAPs. The inset images also show the simulated pattern. As seen from the Figure 4 10(A), shell with no InN core SADP matched pattern with zone axis. This means that core was growing in such a way that direction is parallel to the core and shell wall was increasing in thickness in direction. This pattern is similar to that shown for InN nanorod in Chapter 2 It confirmed that GaN shell || InN at core. The SADP of tip in Figure 4 10(B) showed double diffraction pattern as viewed from zone axis. The br ight spots in SADP correspond to reciprocal lattice space, meaning the distance between the spots is inversely proportional to the d spacing. Thus, inner and lower intensity spots corresponded to InN core and outer brighter spots corresponded to GaN. The S ADP confirmed hexagonal nature of InN and GaN and similar pattern also

PAGE 145

145 confirmed highly epitaxial growth of GaN shell on InN core. Thus, lattice relationship between InN core and GaN shell can be summarized as GaN shell || InN core After establishing the epitaxial relationship that is followed between InN and GaN, SAED measurements. The average lattice constants in G aN shell area were found to be 0.3130nm and 0.5068nm respectively with fringe measurement with c/a ratio of 1.6191. The average values of lattice constants with SAED were, however, 0.3154nm and 0.5286nm with c/a ratio of 1.6760 which was much higher than i deal value. SAED patterns were not sharp and hence these calculations become less accurate. Compositional analysis of InN template capped with LT GaN As seen from various images, in the InN GaN core shell nanostructures, not all were filled with InN nit ride. Many of the nanorods were partially filed and some of them empty ( Figure 4 11 ). It is well known that InN is not stable above 923K. In addition, when shell forms due to absence of ammonia availability, InN can decompose even at 600K to form Indium metal. As a result partially decomposed InN nanorod cores were visible in GaN shells. During TEM analysis, these dark regions sometimes behaved like liquid and liquid I ndium was forced out of shell ( Figure 4 11(E) and (F)) It confirmed that these dark re gions were Indium metal or partially decomposed InN It is known that InN and GaN are not very much miscible into each other. Also, SAED patterns showed no evidence of In x Ga 1 x N for mation. Thus, it was important to see if InN has diffused into the GaN shel l.

PAGE 146

146 Electron dispersive spectroscopy was used for qualitatively looking at the InN. In TEM and SEM, electrons that cannot come out of the sample excite and then again go back to generate X rays that are characteristics of that particular element. These X rays coming out of the electron interaction volume in sample as a result have wavelengths characteristic to the composition of that particular region. Hence, with proper detector they can be used for qualitative analysis. When line scan and point can was d one at different points in the core shell structure, no or negligible Indium signal was detected in lighter regions as shown in Figure 4 12 Also, in XRD of similar sample, shown in Figure 4 5 no In x Ga 1 x N peak was detected. For example, (0002) peak, In x G a 1 x N should be observed between InN (0002) peak and GaN (0002) peak. The EDS, thus, confirmed that there is no significant In x Ga 1 x N formation at interface of InN nanorod surface and GaN cap. Similar to the individual nanorods covered in GaN, samples show n in Figure 4 4(C) were analyzed in TEM and EDS was done. The TEM image of this sample is shown in Figure 4 13. TEM image showed lighter areas at nanorod spots indicating nanorod voids. But, t he EDS scans on darker region only detected Gallium and no Indiu m signal was found The electron transparent samples are generally 200nm and thinner. Due to such low volume, if Indium is not present in substantial amount in shell it cannot be detected. Samples are prepared in focused ion beam. And it is possible that d ue to high energy beam InN nanorods were etched away similar to electron beam removing Indium metal from core in TEM. Annealing of LT GaN grown on capped template : As seen in previous section the core shell structures showed absence of Indium in the core For samples on which GaN was grown for prolonged period of 20 min at 873K no Indium was detected. The

PAGE 147

147 InN is unstable above 923K. As InN nanorods were capped and were always below 873K for the growths InN was not expected to completely disappear. To stu dy this phenomenon, the same capped sample was annealed for 10 minutes under ammonia atmospheres from temperatures ranging from 923K to 1123K After every anneal run the XRD 2 scan was collected. For each run, new sample from same original run at 873K was used to maintain uniformity. As anneal temperature was increased, as seen from plot shown in Figure 4 14 quality of GaN improved as well as InN peak decreased. It was an indication of decomposition of InN. Also, there was no appearance of In x Ga 1 x N p eak. So even if it formed it was very limited and was not detectable. It is also clear from the Figure 4 14 that GaN peak becomes sharper till about 1073K and at 1123K InN and GaN peak heights become equal. The increase in sharp ness of GaN was due to improvement in quality due to anneal. On plots of areas under GaN and InN peaks vs. temperature ( Figure 4 15 ), it was observed that GaN peak started improving after 973K and InN peak started decreasing. The ratio of areas under the InN peak and GaN peak decreased as temperature increased. It was clear indication of InN decomposition without getting much incorporated in GaN. Estimation of amount of InN in LT GaN/InN NR /LT GaN/Si sample : The quantitative estim ation of amount of Indium in the sample was done by two methods. Firstly samples were dissolved in concentrated nitric acid overnight. The nitric acid with dissolved sample was then analyzed using ICP MS. ICP MS estimated Indium fraction to be 2 3 molar pe rcent. It was observed that nitric acid was not able to dissolve sample completely even after 48 hours. As a result, the amount of InN was estimated to be higher than 2 3%.

PAGE 148

148 In another approach, t he estimat ion of amount of Indium in sample similar to one s hown in Figure 4 4 (C) and (D) was done using SEM imaging One of the annealed samples was scraped with tweezers lightly to knock off some of the heads. The F igure 4 16 shows the SEM image of the sample with hollow nanotubes of GaN. The core radius was very uniform in diameter with average core diameter of 100nm. From the simple analysis of different images, the amount of Indium that should have been present in the sample can be easily calculated. Knowing area fraction of InN NR core in each core shell struct ure and area coverage by core shell structures, fraction of area covered by InN in the sample is known. Knowing heights of each layers in the sample, amount of InN present in fully covered InN nanorod sample as shown in Figure 4 4(E) and (F) was calculated out to be 3 5 mole percent. Even lower limit of calculation gave, more estimate than ICPMS analysis. Thus, 3 5% mole is thought to be accurate estimate for amount of InN present. Although annealing results showed that decomposition of InN core took place, there was no evidence in XRD or TEM for alloy formation. With fully covered sample, some amount out of 3 5% of Indium should be trapped in the sample in some form. The SE M EDS was used for confirming presence of Indium and its distribution in the sample. When EDS was done at the surface it showed Indium atomic fractions of 0.93 1.04% which was close to detection limit. In EDS, characteristic X rays com e out of the sample from depth of up to a micron. Although, sample was not more than 2.5 microns thick, ab sence of Indium in surface scan meant negligible I ndium is present in micron depth. But cross sectional EDS revealed presence 2.85 2.95% Indium in the sample. The point EDS at different points across cross section was done. It revealed presence of 1.25

PAGE 149

149 2.1% In, 2.37 3.70% In, 2.81 4.74% In in capping LT GaN, bottom of the nanorod and bottom LT GaN layer respectively EDS line scans revealed that all concentrations remained constant across the sample whereas along the depth of the sample Indium showed grading. Indium tends to be lower in the concentration at top and its concentration peaked at the boundary where nanorods originate. Secondary Ion Mass Spectrometry also gave similar qualitative profile for Indium The SEM EDS profiles and SIMS depth prof ile are shown in Figures 4 17 and 4 18 respectively. Properties of High Temperature GaN (HT GaN) layer The HT GaN was grown at 1123K for 1 hour and 5 hours The one hour grown film was characterized for morphology, crystalline quality, composition. The 5 hour growth was done on various types of templates mainl y to produce thick crack free GaN over 1cm x 1cm substrate. It was characterized for surface morphology and for crystalline quality. Growth conditions for thin and thick HT GaN For the 1 hour run the growth rate was low and it produced about 1500nm films, as seen in Figure 4 20(B). The HT GaN layer in HMOVPE tends to be very rough and nitrogen terminated. This effect is generally because of presence of hydrochloric acid in the system. The HCl gas acts as both scavenger for metal and etchant for GaN film. The HT GaN in 1 hour run was grown by two methods. In first method, after growth of InN template GaN was grown, GaN growth was started at 873K. But simultaneously temperature was ramped from 873K to 1 123K. During the course of 15 20 minutes that temperature ramped, GaN was continuously kept growing. This method was employed with rational that growth with continuous and gradual temperature increase will provide gradual change in quality of GaN. As a re sult,

PAGE 150

150 when HT GaN is grown at 1123K, it will be much more relaxed. In second approach, LT GaN was first grown for 20minutes. Then temperature was raised to 1123K and then HT GaN was grown for 1 hour. The samples grown from both cases are shown in Figure 4 19 and Figure 4 20 respectively. Thick GaN was grown on various samples. They included A) vertical InN nanorod templates as discussed in Chapter 2, B) random InN nanorods grown on Si, C) bare Silicon substrate and D) vertical InN nanorod templates covered with 15 min of LT GaN. All the samples were loaded into the reactor in a single run to maintain the uniformity of conditions. Initially, another layer of LT GaN was grown at 873K to form buffer layer on bare silicon and capping layer on InN nanorod samples It also gave one more coat of LT GaN on already capped sample D. Then temperature was increased to 1173K and high temperature GaN was grown for 4 hours at high growth rate of about 10 microns per hour. Properties of t hinner HT GaN In first case contrary to expectations, GaN films showed lots of cr acks and GaN films peeled off. The peeling of GaN film from substrate silicon is expected because of low bonding between LT GaN and Silicon. But in this case GaN peeled off from both silicon as well as initial LT GaN layer as seen in Figure 4 19(A) High temperature GaN tends to crack if grown on bare silicon, because of thermal expansion mismatch and lattice mismatch. But initial LT GaN gives some relaxation. In first case, when GaN was grown on InN nano rods g rown on the initial LT GaN with continuously increasing temperatures, process of capping GaN formation and decomposition of InN were going on simultaneously This resulted in lower bonding between initial LT GaN and upper GaN layer that resulted in peeli ng of later. The peeled off layers are shown in Figure 4

PAGE 151

151 19(B) and it also shows back side of uppermost layer of GaN. It is a hollow structure, with voids instead of nanorods. No nanorods were found on anywhere on the sample. The EDS was done on the same sample on both front and back side. On front side as expected Indium concentration was below 1%, while on back side it was 3% which is just above the detection limit of EDS. Figure 4 19(C) and (D) show close view underlying LT GaN buffer layer and HT GaN. While HT GaN films are rough, LT GaN show high texture with c faces of GaN crystallites visible. Grains in HT GaN are however much bigger. In second case films did not crack similar to first case, but often self separated from substrate due to low adhesio n between LT GaN buffer and silicon. This is good, if standalone thick GaN is desired product. As discussed in previous section, at higher temperatures InN decomposes Same thing was seen here. Both cross sectional SEM image in Figure 4 20 (B) and TEM imag e in Figure 4 21 showed voids of InN instead of nanorod SIMS depth profile was done to study diffusion of Indium into the film similar to in case of LT GaN growth. The SIMS profile as seen in Figure 4 22, showed that Indium that is diffused in LT GaN does not penetrate into upper GaN crystal. Due to high temperature, even In x Ga 1 x N alloy formation is not possible. Also, due to high crystalline quality, Indium may not have as much accumulation sites such as grain boundaries in HT GaN as in LT GaN. The SAED analysis of upper GaN layer produced very sharp dot pattern with was indication on grain growing in direction. This better quality growth was expected at high temperature. The grain size s w ere found to be in micron range, with 7 00nm as shown in Figure 4 21 The lattice parameters of high temperature GaN were a =0.3164

PAGE 152

152 nm and c =0.5167 and c/a ratio of 1.633 which was near to ideal ratio. The film was not, however, single crystal GaN growing in direction and various grains had different zone axes. As a result film was believed to be polycrystalline. Properties of t hick HT GaN Thick GaN grown was 50 m in thickness. Different substrates produced different results in terms of growth without cracks. The GaN grown directly on silicon and random InN nanorods cracked. The GaN grown on samples C and D did not crack, but were self separated. Both of them wer e highly crystalline and textured in direction But, only HT GaN grown on LT GaN capped InN template showed best 2 FWHM which was less than 350 arc sec. The XRD spectra however revealed that the GaN grown was polycrystalline. The powder XRD spectra showed a very small peak which is common facet that can provide strain relaxation during growth. The SEM images and powder XRD scan for best result are shown in Figure 4 23 and Figure 4 24 respectively. The results showed that if GaN is grown on LT GaN capped InN templates that have been allowed to relax at room temperature produced good results. This means that the samples of LT GaN should be allowed to relax before any extra stress is added to them by taki ng to higher temperature and growing GaN. Self separation could have been another reason why samples did not crack. Due to lower adhesion between the bottom LT layer and silicon, developing crack the samples self separated and relaxed. Summary and Conclusi ons In this Chapter crack free thick GaN was successfully grown using vertical InN NR /LT GaN/Si template developed in Chapter 2. The growth of GaN was crack free

PAGE 153

153 for nanorod templates over area of 1cm x 1cm. The GaN grown separated from Si due to weak adhe sion. Although crack free high quality GaN could be grown, the results showed the polycrystalline nature of final film, though film seems to be highly textured in direction. The polycrystalline nature of HT GaN might have origins in bottom LT GaN nature which was also found to be polycrystalline. Polycrystalline nature of LT GaN was arising from SK type growth with secondary nucleations. TEM results revealed that LT GaN growth started as epitaxial, but did not continue as epit axia l due to high strains and more favorable energetics for secondary nucleations. As a result although InN nanorods were single crystalline GaN growth was not. Main reason behind relaxation provided by InN template can be low decomposition temperatures of I nN. The decomposition of InN provided nano voids where LT GaN was able to relax as temperatures increased. This was clear from the fact that HT GaN films cracked when these conditions were not uniform. For example, when nanovoids were in non uniform distri bution, in case of random nanorods films cracked. When InN mediating layer was not used on bare Silicon, films cracked. Also, when capping of GaN was not complete and simultaneous decomposition and deposition of InN nanorods and GaN respectively was allowe d, films cracked. So uniformity of relaxation provided by nanovoids created by InN was must to get good thick GaN growth. Another relaxation mechanism can be formation of alloy of InN and GaN, though no evidence was found for that. It was also found that a lthough InN is decomposing, In dium metal was not leaving the sample. The Indium mole fraction was calculated to be 3 5% in capped LT GaN

PAGE 154

154 grown sample. The presence of Indium in the LT GaN after anneal was confirmed by SEM EDS and the concentration values m atched to calculated values. Qualitative analysis by SIMS also confirmed that Indium concentration is distributed in the LT GaN film with Indium sometimes higher at boundary where nanorods originated. Thus, this Chapter presents complete study of GaN growt h on InN with step by step investigation at different stages of growth. Although relaxation provided by InN nanostructured template can produce crack free GaN, to avoid polycrystalline nature of HT GaN, the quality and orientation of LT GaN layers must be improved.

PAGE 155

155 Figure 4 1. Schematic of MO HVPE for GaN growth

PAGE 156

156 Figure 4 2 Typical growth scheme for thick HT GaN growth

PAGE 157

157 Figure 4 3. SAED pattern for LT GaN buffer in vertical InN NR template A) Cross secti onal view (The yellow spot gives idea of position of SAED pattern taken FIB sample of similar type) B) SAED ring pattern showing polycrystalline nature of LT GaN Si InN nano LT GaN A B

PAGE 158

158 Figure 4 4 Different stages of capping layer growth A) Cross sectio nal view and B) Top view of 5 minutes LT GaN growth, C) Cross sectional view and D) Top view of 10 minutes LT GaN growth, E) Cross sectional view and F) Top view of 15 20 minutes LT GaN growth A B C D A

PAGE 159

159 Figure 4 4 E F

PAGE 160

160 Figure 4 5. XR D ( 2 pattern for vertically oriented InN nanorod capped by LT GaN for 15 min

PAGE 161

161 Figure 4 6. TEM image of InN GaN core shell structure

PAGE 162

162 Figure 4 7. HRTEM of GaN spike: A) actual core shell rod, B) magnified view spike showing spike is at 62 to shell, C) high magnification image with line scan A B C

PAGE 163

163 Figure 4 8. GaN as viewed in direction represented in different models. A ) space filled model, B ) ball stick model Ball stick model shows plane at an angle of 62 to plane C ) Space filled model superimposed on actual lattice fringes showing exact fit. < 1 2 10 > < 1 0 1 0 > ( 1 0 1 1 ) < 000 1 >

PAGE 164

164 Figure 4 9. GaN shell lattice fringe image. Image in inset shows growth direction of as well as an edge dislocation in red. Edge dislocations of |b| = a are common in wurtzite GaN.

PAGE 165

165 Figure 4 10 SAED patterns of InN GaN core shell structure shown in Figure 4 6. A) SAED of s hell showing pure GaN as viewed in zone axis, presence of some other bright spots indicate polycrystalline nature, B) SAED of tip showing double pattern as viewed in zone axis, with superimposed blue GaN pattern and green InN pattern. Simulated SAED patterns on right are generated by Web EMAPs 161 A B

PAGE 166

166 Figure 4 11. TEM Image showing InN GaN core shell structures filled with Indium nitride or Indium me tal. A) Completely filled, B) and C) partially filled, D) Empty GaN shell. E) Micro core shell filled with Indium, F) Indium forced out on focusing with electron beam A B C D A E F

PAGE 167

167 Figure 4 12. Point scan and line scan EDS in TEM both showing absence of In dium in shell possibility of Indium liquid in dark regions.

PAGE 168

168 Figure 4 1 3 TEM and TEM EDS of InN template completely covered in LT GaN. A) Lighter areas show InN voids, B) Line scan showing total absence of Indium in films. A B

PAGE 169

169 Fi gure 4 1 4 Annealing of LT GaN/InN NR /Si under ammonia atmosphere for 10 minutes at different temperatures showing decomposition of InN InN (002) GaN (002)

PAGE 170

170 Figure 4 1 5 Areas under GaN and InN peaks and area ratio vs. anneal temperature

PAGE 171

171 Figure 4 16. Annealed LT GaN/InN NR /GaN sample. Absence of InN core creates empty GaN shells with 100nm bore and 80nm thick shell. 100 nm

PAGE 172

172 Figure 4 17. SEM EDS of annealed LT GaN/InN NR /GaN sample. A) Point EDS scans at selected poi nts in different layers, B) EDS scans showing counts of different species along specified horizontal and vertical lines. B A

PAGE 173

173 Figure 4 1 8 SIMS depth profile showing In, Ga and Si for annealed LT GaN/InN NR /GaN sample. Counts/s (log scale)

PAGE 174

174 Figure 4 19. Crack ed HT GaN layer on Si (2 m) thick). A) Image showing all the layers in growth, B) Back side of thick GaN which has LT GaN grown on InN template connected to it, C) Image showing crack in lower LT GaN and morphology with hexagonal c plane faces, D) High temperature GaN image showi ng rough surface due to presence of HCl A C B D

PAGE 175

175 Figure 4 20. GaN (2 m thick) film without cracks A) Image showing surface roughness, B) Cross sectional image showing voids where nanorods were and quality difference in LT GaN and HT GaN is clear Si LT GaN/InNnano on LT GaN HT GaN A B

PAGE 176

176 Figure 4 21 TEM cross sectional image of HT GaN sample, 700 nm grain is single crystal growing in direction

PAGE 177

177 Figure 4 22. SIMS depth profile of 1.5 micron thick GaN showing no diffusion of Indium in upp er HT GaN layer Counts/s

PAGE 178

178 Figure 4 23. SEM image of 50 m thick GaN grown on InN template. A) Top view at lower magnification showing coverage, B) Top view showing big grains of 10 m size, C) Cross sectional view showing uniform crack free growth, D) Cross sectional view at higher magnification showing self se paration from Si substrate. A B C D Si Si GaN Air Gap

PAGE 179

179 CHAPTER 5 EXPLORATORY STUDY ON GROWTH OF GALLIUM NITRIDE OVER INDIUM TEMPLATED SILICON Overview From Chapter 4, it is clear that the main role InN plays in the growth of thick GaN growth is because of flexibility offered by nano template. This Chapter explores growth of GaN using very novel technique i.e. the use of thin Indium metal film as a template to grow GaN on Silicon. Although there is no report on using metal templates to deposit semiconductor material films, the re are numerous reports on growing InN and GaN nanostructures on substrates patterned with metals 120,123,124,162,163 Often in MOCVD techniques that are developed for nanowire growths, the growths take place via Va por Liduid Solid (VLS) mechanism. The metal that catalyzes the process often comes from either patterning or reaction can be self catalyzed by metal from one of the source materials 114,122,124,126,132,139,164 The m otive behind this study was to see if Indium metal deposited on Silicon can be used as template for GaN nanostructure or film growth. Metal films deposited on forging substrates when heated above melting point often self assemble and resulting semiconducto r nanowires that grow via VLS mechanism will be self assembled. Another reason for exploring Indium metal deposited on Silicon as a for GaN growth was that use of Indium metal at interface can form alloy of InN and GaN which may decrease the lattice misma tch between Silicon substrate and GaN similar to InN use. As a result, it can also prove to be a template for thick GaN growth. Growth Conditions for Different Layers in Growth Silicon (100) was used as a substrate for this study. Silicon wafer was degreas ed using RCA cleaning method described in Chapter 2 in detail. The method includes cleaning of Silicon wafer in boiling trichloroethylene, acetone, and methanol baths for 5

PAGE 180

180 minutes each. The Silicon was then cleaned in buffered oxide etch to remove any SiO 2 of which are in use currently. The wafers were then deposited with Indium metal and GaN at different conditions. They are described as follows: Deposition of Indium M etal F ilm Indium metal was deposited using electron beam physical vapor deposition sy stem located at microfab. In this method high electron beam is bombarded on anode made of material to be deposited. High energy of electron beam causes material to evaporate, the vapors then deposited on complete chamber which includes the substrate. The thickness of deposited material is measured precisely by piezoelectric crystal sensor. In this case, Indium was used as anode material; e beam was operated with 15mA emission current. The chamber pressure was maintained at 10 6 torrs. The thick ness of I ndium deposited was 100 nm. Deposition of GaN in MO HVPE R eactor GaN films were deposited in same MO HVPE reactor chamber as described in Chapter s 2 and 4. The films were grown in two steps. Thin GaN films The Indium metal films were deposited with 10 min utes GaN films first which produced about 1 m films. All films of this type were grown under MO CVD mode, i.e. no HCl was used. The HCl gas being good scavenger of metals can react with Indium to form gaseous chlorides As a result, no HCl was used to dep osit first layer of GaN. The GaN depositions were carried out at 973K as lower temperature and 1073K 1123K as higher temperature The temperatures were chosen to avoid any InN formations. The

PAGE 181

181 ammonia to TMG ratio was maintained at 1000 which was previously optimized value for good GaN deposition. Nitrogen was used as carrier gas with flow rate of 1600 sccm. Thick GaN films Once thin MOCVD film was deposited on In/Si template, they were deposited with high temperature GaN films for longer times. Thick film s of GaN were all grown at high temperature of 1123K. GaN films were deposited in both MOCVD and HVPE mode. For MOCVD mode, same conditions as described for thin GaN layer were used. MOCVD growths were done for 2hours. For HVPE mode films, HCl was introduc ed into the reactor. The HCl/TMG and NH 3 /TMG molar ratios were maintained at 2 and 600 which are previously optimized values for good HT GaN growth. Films at all the stages of growths were characterized using SEM and XRD. Results and Discussion The SEM ima ge shows deposited Indium on Si substrate in Figure 5 1. From cross sectional view (Figure 5 1(A)), it is clear that the film was precisely 100 nm. But as seen in planar view (Figure 5 1(B)), film was not continuous but formed islands for Indium metal. Thi s type of equilibrium shapes of metal are expected in various systems of heteroepitaxy as well as depositions on amorphous materials 134 The size distribution of Indium islands varied from 100nm to 700nm, with most of the islands showing hexagonal geometry. It is in fact equ ilibrium shape on Indium droplets with faces parallel to substrate surface 165,166 The substrate was approximately 70% covered with big Indium crystals. On heat treatment, Indium was expe cted melt at 433K and form In x Ga 1 x N at growth temperatures used.

PAGE 182

182 Thin GaN film s The SEM image in Figure 5 2 (A) and (B) sho ws as deposited GaN film at 973K on Si (100) and In/Si(100) samples respectively. On bare Silicon substrate film formed as expect ed, but on other sample, GaN deposited in the form of semispherical cabbage like structures. The semispherical shapes are clearly visible in SEM cross sectional image of the same sample (Figure 5 2 (C)). Also, there was a noticeable deposition of thin film of about 150nm thickness between the spheres. The cabbage type nanostructures themselves had average diameter of 700 nm. The EDS measurements showed that sample contained 56.73% Ga, 40.86% N and 2.41% Indium. The XRD spectrum of the sample is shown in Fig ure 5.3. The spectra showed that GaN film formed is polycrystalline GaN. There was a sharp peak of at 30.82 which is attributed to (In,Ga) 2 O 3 The highly textured oxide with peak was attributed to the fact that Indium metal deposited was preferentially oriented in direction. Presence of oxides is detectabl e sometimes when forming gas is not used as carrier gas for GaN growth. On the other hand, GaN grown at higher temperatures on Indium template showed completely different morphologies. The planar views of thin GaN deposited on In/Si at 1 123 K and 1 07 3K are shown in Figure 5 4 (A) and Figure 5 4(B). As seen from the images, there are apparent bright hexagonal GaN nucleations, embedded in darker matrix. The nucleations were denser but smaller in case of GaN grown at 1073K whereas nucleations in case of GaN grow n at 1123K were sparse bigger hexagonal discs. At initial judgment, the discs and matrix were thought to be different materials, but XRD spectra showed only presence of GaN as seen in Figure 5 5. Also, XRD data is

PAGE 183

183 noisy due to very weak signals. It is indi cator of sparse deposition. When In/Si samples were annealed at 1123K for 10 min under nitrogen atmosphere, the surface of Si showed morphology similar matrix structure (Figure 5 4(C)). The EDS analysis showed only presence of Gallium, Nitrogen and approx imately 5% oxygen. No Indium was found in annealed samples This meant that due to high temperatures Indium films evaporated The deposition of Gallium nitride or Gallium oxynitride on Silicon in absence of any metal organic flow was coming from parasitic wall depositions from previous runs in the hot wall reactor At higher temperatures, when metal organic was flowing; hexagonal diskettes formed might be due to new seeding of GaN. The smaller diskettes at lower temperature and bigger diskettes at higher t emperatures can be explained by higher mobility of Gallium at higher temperatures. Due to higher surface bigger diskettes are able to grow and only bigger and sparse diskettes remain. The Indium metal has been shown to work as surfactant at these temperatures 167 Because of surfactant like behavi or, reduction in substrate surface energy can in turn increase Gallium surface mobility. Because of sparse GaN diskettes formation, however a weaker XRD spectrum was also expected as the matrix only contributed to amorphous signal or noise Thi ck GaN fil m s All the thick GaN films were grown at 1123K. The MO GaN films were grown under higher TMG/NH 3 molar ratios of 1000 whereas those under HVPE mode were grown at the ratio of 600. T he SEM images of MOVPE are shown in Figure 5 6 (A) and (B) in planar view and cross sectional view respectively. The growth of GaN under these conditions was not film like. Contrary to that the GaN showed highly dense wafer structures. These wafers were approximately 8 m in height and 100 nm in thickness.

PAGE 184

184 From the cross sectional view it is clear that the GaN wafers are originating from underlying 1 m seed layer. All the seed layers, grown at temperatures ranging from 973K 1123K, showed similar results. The EDS analysi s showed presence of Gallium and nitrogen only both in seed layer as well as GaN wafers. The powder XRD scans for samples grown on both 973K LT GaN and 1123K LT GaN are shown in the Figure 5 7 (A) and (B) respectively. They confirmed absence of any Indium species. All the GaN peaks are present as wafers do not have any preferred growth direction. SEM images did not show any Gallium droplets on any part of sample either Absences of Indium meant that it is not playing any role in GaN wafer growth. Abs ence of Gallium metal droplet meant no VLS mechanism was involved in GaN growth. Absence of any stable oxide in the film confirmed that oxygen might have played role in the formation of these GaN diskettes. Growth of GaN by oxide mediated growth is studie d by v arious r esearchers 168 174 The overall reactions for oxide mediate GaN growth are given as follows: (5 1) (5 2) In oxygen poor environment Ga 2 O may readily form over more complex Ga 2 O 3. Although Ga 2 O 3 is a stable oxide, at growth conditions of high temperatures in ammonia rich environment, gaseous Ga 2 O can form. This oxide can be further reduced on reaction with ammonia either in presence or abse nce of carbon catalyst to GaN (Reactions (5 1) and (5 2) respectively ). In such Vapor Solid reactions reaction sites play important role, resulting in preferred direction or plane growths forming nanostructures.

PAGE 185

185 The other type of Ga N growth was done in HVP E mode for higher time intervals of 4 hours. From former discussion it is clear that there was no or little Indium left in the film when thin GaN was grown at 1123K. Also, deposition consisted of sparse GaN diskettes. As a result, when this template was us ed for high temperature GaN growth films cracked as expected When high temperature GaN was grown on template with 973K GaN grown for 10 minutes, films showed excellent composure. The films did not show any crack similar to films grown on InN nanorod temp late. Unlike InN nanorod template samples, most of the time Indium film underneath had not completely broken contact between silicon and upper film. As a result GaN films grown on Indium film did not self separate. The SEM planar and cross sectional view of 40 m GaN film grown on In/Si (100) template is shown in the Figure 5 8 (A) and (B) respectively. As seen from SEM images, the film grown on this template, like every other HVPE films, is very rough due to presence of HCl in growths. The powder XRD scan (Figure 5 9 ) showed results similar to HVPE thick film s grown on InN nanorod sample meaning GaN grown by this method was also polycrystalline. Summary and Conclusions This Chapter presented study about growth of GaN using novel Indium on Si template By changing the mode of growth and temperatures, it was shown that growth of GaN can be changed drastically from nanostructures growth to film growths. A thick GaN layer, similar to that presented in last Chapter was achieved, without cracks on 1cm x1c m Si (100). The GaN layer grown here did not self separate similar to that grown on InN nanorod template. This behavior was attributed to fact that Indium

PAGE 186

186 metal layer although can act as stress reliving layer; it does not prevent GaN to form bonds with Si. In MO CVD mode, high temperature growths produced nanostructures. Absence of any evidence for VLS mechanism and presence of oxygen in low temperature grown layers showed Gallium oxide mediated growth as a possible mechanism for GaN wafer growths. Wafers showed no preferred orientations, but showed uniform thickness of 100 nm. It was thought that they are grown from diskettes of GaN that form at high temperature growths on In/Si templates. The diskettes formations can be a result of high mobility of GaN at higher temperatures. In heightened motilities of Gallium molecules on substrate can be attributed to effect of Indium as a surfactant that plays part in reducing surface energies. To conclude, use of In/Si as a template for thin and thick GaN is shown he re. Although, thick GaN films were polycrystalline in nature, the preliminary results are promising and need further investigation to improve quality. Growth of GaN wafers on Si(100) has never been reported. Although wafers have shown uniformity of thickne ss, uniformity of direction is highly desired. If all wafers grow in one direction, the method can be improved on to harvest even bigger sizes of wafers. Hence, further studies will be needed to improve orientation of the GaN wafer structures.

PAGE 187

187 Figure 5 1. SEM images of Indium on Silicon (100). A) Cross sectional view showing 100nm islands of Indium, B) Top view showing self assembled Indium islands structure A B

PAGE 188

188 Figure 5 2 SEM images of GaN grown at T=873K for 10 min A) Growth on In/Si(100) as template B) Growth on Si(100) as template, C) Cross sectional view of growth on In/Si(100) template A B

PAGE 189

189 Figure 5 2 Continued C

PAGE 190

190 Figure 5 3 XRD scan of GaN grown on In/Si template for 10 min at 973K

PAGE 191

191 Figure 5 4 SEM images of growth on In/Si at higher temperatures for 10 min A) Growth at T=1123K showing bigger and sparse diskettes in matrix B) Growth at T=1073K showing sm aller but dense diskettes in matrix C) Growth under only nitrogen flow at T=1123K A B

PAGE 192

192 Figure 5 4 C

PAGE 193

193 Figure 5 5 XRD scan of GaN grown on In/Si template for 10 min at 1123 K

PAGE 194

194 Figure 5 6 SEM images of thick M OCVD GaN growth on In/Si at 1123K for 120 min A) Top view showing rough GaN wafers B) Cross sectional view showing height about 8 m A B

PAGE 195

195 Figure 5 7 XRD scan of GaN wafers grown at 1123K in MOCVD mode A) Grown on GaN/In/Si(100) template grown at 973K, B) Grown on GaN/In/Si(100) template grown at 1123K A

PAGE 196

196 Figure 5 7 Continued

PAGE 197

197 Figure 5 8 SEM images of 40 m thick HVPE GaN growth on LT GaN/In/Si at 1123K for 240 min A) Top view showing rough GaN surface B) Cross section al view showing height about 40 m free growth without self separation A B GaN Si

PAGE 198

198 Figure 5 9 XRD scan of thick GaN grown on LT GaN/In/Si template at 1123K

PAGE 199

199 CHAPTER 6 CONCLUSIONS AND RECOMMENDATION S OF FUTURE WORK Conclusions This study had final goal to achieve crack free thick and thin GaN on substrates such as Silicon which has very large lattice mismatch with GaN. The goal was successfully achieved, though various challenges still remain. The key conclusions from this study are as follows. Cha pter 2 dealt with question of how the growth direction of InN nanorods is affected by different substrate preparations for S ilicon. Chapter showed that major improvement in directionality is possible by changing surface properties structurally. Nitridation formed Silicon oxynitride layer, which changed the surface to either hexagonal geometry or amorphous. In both cases, directionality of InN nanorods in c axis direction improved. The c axis growth happens to be the preferred growth direction of nitride cry stals. If formation of oxynitride layer formed a crystalline layer, the oxynitride being geometrically more similar to basal plane of InN, it improved the growth in vertical direction. If oxynitride created the amorphous layer, then it disconnected all the epitaxial requirements for growth, and that resulted in improvement in natural growth direction This direction happened to be vertical direction in case of InN nanorods. More than nitridation, presence of good quality c directional nucleation layer of Ga N improved growth of InN in vertical direction. With optimum nucleation layer, all the nanorods with high crystallinity were grown in vertical direction. And results from pole figures showed that InN nanorods orientation was only dependent on underlying nu cleation layer. The InN nanorods grown were high quality crystals, with very few defects and dislocati ons concentrated towards bottom. This was thought to be because of possible

PAGE 200

200 lateral relaxation in nanostructures as they grow. This relaxation was modele d and model qualitatively showed relaxation into the bulk material was possible. Model also showed that coherency can be greatly improved if lower layer also shares the strain and assumed to be flexible. The results predicted were observed were confirmed w ith many similar experimental observations. InN nanorods grown on Silicon substrate were used as possible template for thick GaN growth. The thick growth without cracks was shown to be absolutely feasible. But, large lattice mismatches and lower temperatur e growths of GaN made it impossible to grow epitaxial GaN on nanorods. This resulted in secondary nucleations and GaN ended up to be highly textured in direction but polycrystalline. It can be concluded here, that only if GaN ca pping layer quality can be improved, a very big improvement in quality of thick high temperature GaN is achievable Use of InN definitely helped relaxing strain, as after cooling films did not crack but completely self separated. The importance of role of InN nanorod layer was also confirmed by the fact that, films cracked if InN nanorods decomposed before formation of capping layer. Decomposed InN when completely covered by GaN, produced porous matrix on GaN which helped relaxation of the layer. Also, Indi um was found to be graded throughout the lower part of structure, which should have influenced relaxation. So use of InN nanostructured template was successful for crack free growth of GaN was possible. Similar to InN, Indium metal was deposited on silicon and used as template for high temperature GaN growth. It was shown that In dium layer served similar purpose as InN layer. As a result, thick GaN growth was also achieved on In/Si substrates. There was no self separation, but GaN showed similar properties and polycrystalline nature. In

PAGE 201

201 absence of strong reducing agents such as hydrogen and HCl in the reactor, GaN growth on In/Si template produced GaN wafer growth. The wafer growth may have been influenced by formation of diskettes of GaN. The diskette forma tion was attributed to increased mobility of Gallium due to surfactant like effect of trace Indium atoms on substrate surface. The wafer s growth was possible due to small amount of oxide present in the reactor and reactions proceeding with oxide mediated v apor solid growth. Recommendation s for Future Work Improvements in InN nanorod orientations Although InN nanorod size and density can be controlled pretty well in MO HVPE system, major improvement is needed in the rotation al relationship with substrate. I f all the nanorods could be grown in exact same rotation with Silicon substrate, it would influence quality of material grown on it. This uniformity will also allow better devices of these nanostructures. This can be done two ways. First approach would be to use pattered substrate. If silicon is p atterned with Indium catalyst dots and other area is covered with oxide or nitride layer of silicon, resulting nanorods growth will be symmetric. Similar approach would be to use porous alumina as template. Improvement in orientation in InN nanorods will b e better for its application for devices as well as for template usage. Second way it could be achieved by using template that retains epitaxial relationship with the substrate. It is possible by layer by layer growth process such as ALD. If the ALD films of GaN can be used as substrates for InN nanorod growth, InN nanorods will have epitaxial relationship with substrate Silicon as nucleation layer will be minimal.

PAGE 202

202 Growth of GaN and In x Ga 1 x N nanostructures in MO HVPE With advance of semiconductor industry, there is major shift in how semiconductor materials are going to be grown. With arrival of 3D transistors, growth of nanostructures of materials such as GaN In x Ga 1 x N is highly desirable. Nanostructures tend to be highly crystalline and as a result devic es can perform better. GaN nanostructures growth however is challenging due to nature of GaN growth. Like many other semiconductors, GaN tends to grow polycrystalline, as it is effective way to relieve strain in material growth. But, use of pattered substr ates has shown to be effective in achieving GaN growth. In MO HVPE, there is added advantage of control over HCl to metal ratio. Due to this nanorod of InN can be grown in HVPE without any catalyst or patterning. But growth of nanostructures of GaN has not been studied in great detail. The major problem has been guessing the growth conditions for GaN nanostructure growth. The great starting point would be to grow Gallium poor In x Ga 1 x N nanostructures and then go towards GaN growth. Second hurdle has been th e substrate. With understanding of nature of GaN in this study it is clear that first growth of nucleation layer of GaN would be essential step towards getting GaN nanostructures. The path of oxide mediated growth of GaN is also worth exploring if controll ed and oriented nucleation of the GaN diskettes is achieved. If GaN nanostructures can be grown in M O HVPE with uniformity in orientation and size, it would be important step in terms of achieving GaN nano devices such as nano lasers, high brightness LEDs and nanostructured solar cells. Furthermore, well oriented GaN nanostructures will prove to be much better substrate to achieve thick single crystal growth of GaN on any substrate.

PAGE 203

203 LIST OF REFERENCES 1 Y. J. Dong, B. Z. Tian, T. J. Ke mpa, and C. M. Lieber, Nano Letters 9, 2183 (2009). 2 R. H. Horng, S. T. Lin, Y. L. Tsai, M. T. Chu, W. Y. Liao, M. H. Wu, R. M. Lin, and Y. C. Lu, Ieee Electron Device Letters 30, 724 (2009). 3 G. H. Jessen, R. C. Fitch, J. K. Gillespie, G. D. Via, N. A. Moser, M. J. Yannuzzi, A. Crespo, J. S. Sewell, R. W. Dettmer, T. J. Jenkins, R. F. Davis, J. Yang, M. A. Khan, and S. C. Binari, Ieee Electron Device Letters 24, 677 (2003). 4 M. A. Khan, A. Bhattarai, J. N. Kuznia, and D. T. Olson, Applied Physics Letter s 63, 1214 (1993). 5 M. A. Khan, Q. Chen, J. W. Yang, C. J. Sun, and M. S. Shur, Silicon Carbide and Related Materials 1995 142, 985 (1996). 6 M. A. Khan and M. S. Shur, Materials Science and Engineering B Solid State Materials for Advanced Technology 46, 69 (1997). 7 M. A. Khan, X. Hu, G. Sumin, A. Lunev, J. Yang, R. Gaska, and M. S. Shur, Ieee Electron Device Letters 21, 63 (2000). 8 V. Kumar, A. Kuliev, R. Schwindt, M. Muir, G. Simin, J. Yang, M. A. Khan, and I. Adesida, Solid State Electronics 47, 1577 (2003). 9 B. W. Liou, Japanese Journal of Applied Physics 48 (2009). 10 B. W. Liou, Ieee Photonics Technology Letters 22, 215 (2010). 11 J. K. Sheu, C. C. Yang, S. J. Tu, K. H. Chang, M. L. Lee, W. C. Lai, and L. C. Peng, Ieee Electron Device Letters 30, 2 25 (2009). 12 G. Simin, A. Koudymov, H. Fatima, J. P. Zhang, J. W. Yang, M. A. Khan, X. Hu, A. Tarakji, R. Gaska, and M. S. Shur, Ieee Electron Device Letters 23, 458 (2002). 13 S. Westwater, in Compound Semiconductors ; Vol. 16 August/September ed. ( Angel Business Communications Ltd Watford, Herts WD17 1JA, UK, 2010), p. 99. 14 J. R. Brodrick, B. Consulting., N. C. Inc., R. A. Inc., S. Consulting., and S. S. L. S. Inc., May 2011 ed., edited by U. S. D. o. Energy (Lighting Research and Development Building Technologies Program, Office of Energy Efficiency and Renewable Energy, U.S. Department of Energy, Washington, D.C, 2011).

PAGE 204

204 15 S. Nakamura, N. Senoh, N. Iwasa, and S. I. Nagahama, Japanese Journal of Applied Physics Part 2 Letters 34, L797 (1995). 16 S. Na kamura, M. Senoh, N. Iwasa, S. Nagahama, T. Yamada, and T. Mukai, Japanese Journal of Applied Physics Part 2 Letters 34, L1332 (1995). 17 S. Nakamura, Diamond and Related Materials 5, 496 (1996). 18 J. W. Yang, Q. Chen, C. J. Sun, B. Lim, M. Z. Anwar, M. A Khan, and H. Temkin, Applied Physics Letters 69, 369 (1996). 19 A. Avramescu, T. Lermer, J. Muller, C. Eichler, G. Bruederl, M. Sabathil, S. Lutgen, and U. Strauss, Applied Physics Express 3 (2010). 20 S. Lutgen, A. Avramescu, T. Lermer, D. Queren, J. Mu ller, G. Bruederl, and U. Strauss, Physica Status Solidi a Applications and Materials Science 207, 1318 (2010). 21 S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, and Y. Sugimoto, Japanese Journal of Applied Physics Part 2 Letters 35, L74 (1996). 22 S. Nakamura, Mrs Bulletin 23, 37 (1998). 23 G. F. Brown, J. W. Ager, W. Walukiewicz, and J. Wu, Solar Energy Materials and Solar Cells 94, 478 (2010). 24 H. Hamzaoui, A. S. Bouazzi, and B. Rezig, Solar Energy Materials and Sola r Cells 87, 595 (2005). 25 M. Anani, C. Mathieu, M. Khadraoui, Z. Chama, S. Lebid, and Y. Amar, Microelectronics Journal 40, 427 (2009). 26 X. Chen, K. D. Matthews, D. Hao, W. J. Schaff, and L. F. Eastman, Physica Status Solidi a Applications and Materials Science 205, 1103 (2008). 27 C. McCormick, Mrs Bulletin 34, 151 (2009). 28 L. A. Reichertz, I. Gherasoiu, K. M. Yu, V. M. Kao, W. Walukiewicz, and J. W. Ager, Applied Physics Express 2 (2009). 29 M. Song, Z. Wu, Y. Fang, R. Xiang, Y. Sun, H. Wang, C. Yu, H. Xiong, J. Dai, and C. Chen, Journal of Optoelectronics and Advanced Materials 12, 1452 (2010). 30 A. Tarakji, X. Hu, A. Koudymov, G. Simin, J. Yang, M. A. Khan, M. S. Shur, and R. Gaska, Solid State Electronics 46, 1211 (2002).

PAGE 205

205 31 I. P. Smorchkova, M. W ojtowicz, R. Sandhu, R. Tsai, M. Barsky, C. Namba, P. S. Liu, R. Dia, M. Truong, D. Ko, J. Wang, H. Wang, and A. Khan, Ieee Transactions on Microwave Theory and Techniques 51, 665 (2003). 32 R. Dwilinski, J. M. Baranowski, M. Kaminska, R. Doradzinski, J. G arczynski, and L. Sierzputowski, Acta Physica Polonica A 90, 763 (1996). 33 R. Dwilinski, R. Doradzinski, J. Garczynski, L. Sierzputowski, R. Kucharski, M. Zajac, M. Rudzinski, R. Kudrawiec, J. Serafinczuk, and W. Strupinski, Journal of Crystal Growth 312, 2499 (2010). 34 T. Hashimoto, E. Letts, M. Ikari, and Y. Nojima, Journal of Crystal Growth 312, 2503 (2010). 35 D. Ehrentraut and T. Fukuda, Journal of Crystal Growth 312, 2514 (2010). 36 T. Sasaki and T. Matsuoka, Journal of Applied Physics 64, 4531 (198 8). 37 T. Sasaki, T. Matsuoka, and A. Katsui, Applied Surface Science 41 2, 504 (1989). 38 S. A. Kukushkin, A. V. Osipov, V. N. Bessolov, B. K. Medvedev, V. K. Nevolin, and K. A. Tcarik, Reviews on Advanced Materials Science 17, 1 (2008). 39 M. E. Lin, B. Sverdlov, G. L. Zhou, and H. Morkoc, Applied Physics Letters 62, 3479 (1993). 40 R. F. Davis, S. Tanaka, L. B. Rowland, R. S. Kern, Z. Sitar, S. K. Ailey, and C. Wang, Journal of Crystal Growth 164, 132 (1996). 41 S. Tanaka, S. Iwai, and Y. Aoyagi, Journal of Crystal Growth 170, 329 (1997). 42 P. Waltereit, O. Brandt, A. Trampert, M. Ramsteiner, M. Reiche, M. Qi, and K. H. Ploog, Applied Physics Letters 74, 3660 (1999). 43 S. Yoshida, S. Misawa, and S. Gonda, Applied Physics Letters 42, 427 (1983). 44 M. A. Khan, J. N. Kuznia, D. T. Olson, and R. Kaplan, Journal of Applied Physics 73, 3108 (1993). 45 I. Akasaki, H. Amano, Y. Koide, K. Hiramatsu, and N. Sawaki, Journal of Crystal Growth 98, 209 (1989). 46 H. Amano, I. Akasaki, K. Hiramatsu, N. Koide, and N. S awaki, Thin Solid Films 163, 415 (1988). 47 H. Amano, K. Hiramatsu, and I. Akasaki, Japanese Journal of Applied Physics Part 2 Letters 27, L1384 (1988).

PAGE 206

206 48 T. Kachi, T. Tomita, K. Itoh, and H. Tadano, Applied Physics Letters 72, 704 (1998). 49 K. H. Lee, S H. Park, J. H. Kim, N. H. Kim, M. H. Kim, H. Na, and E. Yoon, Thin Solid Films 518, 6365 (2010). 50 E. Calleja, M. A. Sanchez Garcia, F. J. Sanchez, F. Calle, F. B. Naranjo, E. Munoz, S. I. Molina, A. M. Sanchez, F. J. Pacheco, and R. Garcia, Journal of Crystal Growth 201, 296 (1999). 51 A. Ohtani, K. S. Stevens, and R. Beresford, Applied Physics Letters 65, 61 (1994). 52 K. S. Stevens, A. Ohtani, M. Kinniburgh, and R. Beresford, Applied Physics Letters 65, 321 (1994). 53 D. J. As, T. Frey, D. Schikora, K Lischka, V. Cimalla, J. Pezoldt, R. Goldhahn, S. Kaiser, and W. Gebhardt, Applied Physics Letters 76, 1686 (2000). 54 F. Schulze, A. Dadgar, J. Blsing, and A. Krost, Journal of Crystal Growth 272, 496 (2004). 55 V. Lebedev, J. Jinschek, U. Kaiser, B. Sc hroter, W. Richter, and J. Krausslich, Applied Physics Letters 76, 2029 (2000). 56 J. R. Gong, M. F. Yeh, and C. L. Wang, Journal of Crystal Growth 247, 261 (2003). 57 H. P. D. Schenk, E. Frayssinet, A. Bayard, D. Rondi, Y. Cordier, and M. Kennard, Journal of Crystal Growth 314, 85 (2011). 58 F. Schulze, A. Dadgar, J. Blasing, and A. Krost, Applied Physics Letters 84, 4747 (2004). 59 J. Wan, R. Venugopal, M. R. Melloch, H. M. Liaw, and W. J. Rummel, Applied Physics Letters 79, 1459 (2001). 60 L. S. Chuah, Z Hassan, S. S. Ng, and H. Hassan, Journal of Nondestructive Evaluation 28, 125 (2009). 61 Y. Cordier, J. C. Moreno, N. Baron, E. Frayssinet, J. M. Chauveau, M. Nemoz, S. Chenot, B. Damilano, and F. Semond, Journal of Crystal Growth 312, 2683 (2010). 62 A. Dadgar, J. Blasing, A. Diez, A. Alam, M. Heuken, and A. Krost, Japanese Journal of Applied Physics Part 2 Letters 39, L1183 (2000). 63 F. Reiher, A. Dadgar, J. Blaesing, M. Wieneke, and A. Krost, Journal of Crystal Growth 312, 180 (2010).

PAGE 207

207 64 F. Schulze, A Dadgar, J. Blasing, A. Diez, and A. Krost, Applied Physics Letters 88 (2006). 65 A. Dadgar, M. Poschenrieder, J. Blasing, K. Fehse, A. Diez, and A. Krost, Applied Physics Letters 80, 3670 (2002). 66 P. Chen, R. Zhang, Z. M. Zhao, D. J. Xi, B. Shen, Z. Z. Chen, Y. G. Zhou, S. Y. Xie, W. F. Lu, and Y. D. Zheng, Journal of Crystal Growth 225, 150 (2001). 67 E. Feltin, B. Beaumont, M. Laugt, P. De Mierry, P. Vennegues, M. Leroux, and P. Gibart, Physica Status Solidi a Applied Research 188, 531 (2001). 68 S. J oblot, E. Feltin, E. Beraudo, P. Vennegues, M. Leroux, F. Omnes, M. Laugt, and Y. Cordier, Journal of Crystal Growth 280, 44 (2005). 69 D. M. Follstaedt, J. Han, P. Provencio, and J. G. Fleming, Mrs Internet Journal of Nitride Semiconductor Research 4, art no. (1999). 70 J. Y. Huang, Z. Z. Ye, L. Wang, J. Yuan, B. H. Zhao, and H. M. Lu, Solid State Electronics 46, 1231 (2002). 71 T. Lei, M. Fanciulli, R. J. Molnar, T. D. Moustakas, R. J. Graham, and J. Scanlon, Applied Physics Letters 59, 944 (1991). 72 T. Lei, T. D. Moustakas, R. J. Graham, Y. He, and S. J. Berkowitz, Journal of Applied Physics 71, 4933 (1992). 73 M. A. Mastro, O. M. Kryliouk, and T. J. Anderson, Materials Science and Engineering B Solid State Materials for Advanced Technology 127, 91 (200 6). 74 Y. E. Romanyuk, D. Kreier, Y. Cui, K. M. Yu, J. W. Ager, and S. R. Leone, Thin Solid Films 517, 6512 (2009). 75 T. N. Bhat, M. K. Rajpalke, B. Roul, M. Kumar, and S. B. Krupanidhi, Journal of Applied Physics 110, 6 (2011). 76 R. Grun, Acta Crystallo graphica Section B 35, 800 (1979). 77 R. F. Davis, T. Gehrke, K. J. Linthicum, E. Preble, P. Rajagopal, C. Ronning, C. Zorman, and M. Mehregany, Journal of Crystal Growth 231, 335 (2001). 78 R. F. Davis, T. Gehrke, K. J. Linthicum, T. S. Zheleva, P. Rajago pal, C. A. Zorman, and M. Mehregany, Zeitschrift Fur Metallkunde 92, 163 (2001). 79 R. F. Davis, T. Gehrke, K. J. Linthicum, T. S. Zheleva, E. A. Preble, P. Rajagopal, C. A. Zorman, and M. Mehregany, Journal of Crystal Growth 225, 134 (2001).

PAGE 208

208 80 R. F. Davi s, T. Gehrke, K. J. Linthicum, P. Rajagopal, A. M. Roskowski, T. Zheleva, E. A. Preble, C. A. Zorman, M. Mehregany, U. Schwarz, J. Schuck, and R. Grober, Mrs Internet Journal of Nitride Semiconductor Research 6, 1 (2001). 81 T. Gehrke, K. J. Linthicum, E. Preble, P. Rajagopal, C. Ronning, C. Zorman, M. Mehregany, and R. F. Davis, Journal of Electronic Materials 29, 306 (2000). 82 K. Linthicum, T. Gehrke, D. Thomson, E. Carlson, P. Rajagopal, T. Smith, D. Batchelor, and R. Davis, Applied Physics Letters 75, 196 (1999). 83 T. Zheleva, S. Smith, D. Thomson, K. Linthicum, P. Rajagopal, and R. Davis, Journal of Electronic Materials 28, L5 (1999). 84 N. P. Kobayashi, J. T. Kobayashi, X. G. Zhang, P. D. Dapkus, and D. H. Rich, Applied Physics Letters 74, 2836 (1999 ). 85 P. Kung, D. Walker, N. Hamilton, J. Diaz, and M. Razeghi, Applied Physics Letters 74, 570 (1999). 86 K. Y. Zang, Y. D. Wang, S. J. Chua, and L. S. Wang, Applied Physics Letters 87 (2005). 87 E. Suhir, Journal of Applied Mechanics Transactions of the Asme 53, 657 (1986). 88 S. Luryi and E. Suhir, Applied Physics Letters 49, 140 (1986). 89 D. Zubia and S. D. Hersee, Journal of Applied Physics 85, 6492 (1999). 90 S. D. Hersee, D. Zubia, X. Y. Sun, R. Bommena, M. Fairchild, S. Zhang, D. Burckel, A. Frauen glass, and S. R. J. Brueck, Ieee Journal of Quantum Electronics 38, 1017 (2002). 91 D. Zubia, S. H. Zaidi, S. R. J. Brueck, and S. D. Hersee, Applied Physics Letters 76, 858 (2000). 92 D. Zubia, S. H. Zaidi, S. D. Hersee, and S. R. J. Brueck, Journal of Va cuum Science & Technology B 18, 3514 (2000). 93 X. Wang, G. Yu, C. Lin, M. Cao, H. Gong, M. Qi, and A. Li, Electrochemical and Solid State Letters 11, H273 (2008). 94 X. Y. Sun, R. Bommena, D. Burckel, A. Frauenglass, M. N. Fairchild, S. R. J. Brueck, G. A Garrett, M. Wraback, and S. D. Hersee, Journal of Applied Physics 95, 1450 (2004). 95 J. Liang, S. K. Hong, N. Kouklin, R. Beresford, and J. M. Xu, Applied Physics Letters 83, 1752 (2003).

PAGE 209

209 96 K. Y. Zang, Y. D. Wang, L. S. Wang, S. Tripathy, S. J. Chua, a nd C. V. Thompson, Thin Solid Films 515, 4505 (2007). 97 L. S. Wang, S. Tripathy, B. Z. Wang, and S. J. Chua, Applied Surface Science 253, 214 (2006). 98 A. Strittmatter, A. Krost, V. Turck, M. Strassburg, D. Bimberg, J. Blasing, T. Hempel, J. Christen, B. Neubauer, D. Gerthsen, T. Christmann, and B. K. Meyer, Materials Science and Engineering B Solid State Materials for Advanced Technology 59, 29 (1999). 99 F. Hamdani, M. Yeadon, D. J. Smith, H. Tang, W. Kim, A. Salvador, A. E. Botchkarev, J. M. Gibson, A. Y. Polyakov, M. Skowronski, and H. Morkoc, Journal of Applied Physics 83, 983 (1998). 100 X. H. Luo, R. M. Wang, X. P. Zhang, H. Z. Zhang, D. P. Yu, and M. C. Luo, Micron 35, 475 (2004). 101 D. C. Park and S. Fujita, Physica Status Solidi a Applied Resear ch 176, 579 (1999). 102 R. F. Xiao, X. W. Sun, Z. F. Li, N. Cue, H. S. Kwok, Q. Z. Liu, and S. S. Lau, Journal of Vacuum Science & Technology a Vacuum Surfaces and Films 15, 2207 (1997). 103 K. Black, A. C. Jones, P. R. Chalker, J. M. Gaskell, R. T. Murray T. B. Joyce, and S. A. Rushworth, Journal of Crystal Growth 310, 1010 (2008). 104 V. A. Ferreira and H. W. Leite Alves, Journal of Crystal Growth 310, 3973 (2008). 105 S. Nishimura and K. Terashima, Materials Science and Engineering B Solid State Materia ls for Advanced Technology 75, 207 (2000). 106 S. Nishimura and K. Terashima, Materials Science and Engineering B Solid State Materials for Advanced Technology 82, 25 (2001). 107 S. Nishimura, S. Matsumoto, and K. Terashima, Optical Materials 19, 223 (2002 ). 108 S. Nishimura, H. Hanamoto, K. Terashima, and S. Matsumoto, Materials Science and Engineering B Solid State Materials for Advanced Technology 93, 135 (2002). 109 K. S. A. Butcher and T. L. Tansley, Superlattices and Microstructures 38, 1 (2005). 110 B. Onderka, J. Unland, and R. Schmid Fetzer, Journal of Materials Research 17, 3065 (2002). 111 O. Kryliouk, H. J. Park, Y. S. Won, T. Anderson, A. Davydov, I. Levin, J. H. Kim, and J. A. Freitas, Nanotechnology 18 (2007).

PAGE 210

210 112 H. J. Park, O. Kryliouk, T. A nderson, D. Khokhlov, and T. Burbaev, Physica E Low Dimensional Systems & Nanostructures 37, 142 (2007). 113 I. Shalish, G. Seryogin, W. Yi, J. M. Bao, M. A. Zimmler, E. Likovich, D. C. Bell, F. Capasso, and V. Narayanamurti, Nanoscale Research Letters 4, 532 (2009). 114 B. S. Simpkins, A. D. Kansal, and P. E. Pehrsson, Crystal Growth & Design 10, 3887 (2010). 115 Y. H. Kim, W. S. Yun, H. Ruh, C. S. Kim, J. W. Kim, Y. H. Shin, M. D. Kim, and J. E. Oh, Journal of Crystal Growth 312, 662 (2010). 116 C. Y. Cha ng, G. C. Chi, W. M. Wang, L. C. Chen, K. H. Chen, F. Ren, and S. J. Pearton, Journal of Electronic Materials 35, 738 (2006). 117 C. K. Chao, J. I. Chyi, C. N. Hsiao, C. C. Kei, S. Y. Kuo, H. S. Chang, and T. M. Hsu, Applied Physics Letters 88 (2006). 118 M. C. Johnson, C. J. Lee, E. D. Bourret Courchesne, S. L. Konsek, S. Aloni, W. Q. Han, and A. Zettl, Applied Physics Letters 85, 5670 (2004). 119 J. Zhang, B. L. Xu, F. H. Jiang, Y. D. Yang, and J. P. Li, Physics Letters A 337, 121 (2005). 120 C. H. Liang, L. C. Chen, J. S. Hwang, K. H. Chen, Y. T. Hung, and Y. F. Chen, Applied Physics Letters 81, 22 (2002). 121 J. Zhang, L. Zhang, X. S. Peng, and X. F. Wang, Journal of Materials Chemistry 12, 802 (2002). 122 S. N. Mohammad, Journal of Applied Physics 106, 11 (2009). 123 T. Kuykendall, P. J. Pauzauskie, Y. F. Zhang, J. Goldberger, D. Sirbuly, J. Denlinger, and P. D. Yang, Nature Materials 3, 524 (2004). 124 G. W. Xu, Z. Z. Li, J. Baca, and J. D. Wu, Nanoscale Research Letters 5, 7 (2010). 125 H. J. Park, The sis, University of Florida, 2006. 126 J. L. Mangum, Thesis, University of Florida, 2007. 127 S. W. Kang, Thesis, University of Florida, 2004. 128 S. A. Kukushkin, Reviews on advanced materials science 17, 1 (2008). 129 L. Liu and J. H. Edgar, Materials Sci ence and Engineering: R: Reports 37, 61 (2002).

PAGE 211

211 130 J. A. Venables, G. D. T. Spiller, and M. Hanbucken, Reports on Progress in Physics 47, 399 (1984). 131 Y. S. Won, Y. S. Kim, O. Kryliouk, and T. J. Anderson, Journal of Crystal Growth 310, 3735 (2008). 13 2 S. Vaddiraju, A. Mohite, A. Chin, M. Meyyappan, G. Sumanasekera, B. W. Alphenaar, and M. K. Sunkara, Nano Letters 5, 1625 (2005). 133 A. Y. Timoshkin, H. F. Bettinger, and H. F. Schaefer, Inorganic Chemistry 41, 738 (2002). 134 E. I. Givargizov, Oriented Crystallization on Amorphous Substrates Vol. 1 (Plenum Press, New York, 1991). 135 B. Maleyre, S. Ruffenach, O. Briot, B. Gil, and A. Van der Lee, Superlattices and Microstructures 36, 517 (2004). 136 C. Goyhenex, H. Bulou, J. P. Deville, and G. Trglia, Applied Surface Science 177, 238 (2001). 137 A. O. Ajagunna, A. Adikimenakis, E. Iliopoulos, K. Tsagaraki, M. Androulidaki, and A. Georgakilas, Journal of Crystal Growth 311, 2058 (2009). 138 S. D. Hersee, X. Y. Sun, and X. Wang, Nano Letters 6, 1808 (200 6). 139 N. Wang, Y. Cai, and R. Q. Zhang, Materials Science & Engineering R Reports 60, 1 (2008). 140 I. N. Stranski and V. L. Krastanow, Akademie der Wissenschaften und Literatur Mainz, Mathematisch Naturwissenschaftliche Klasse IIb 146, 797 (1939). 141 D J. Eaglesham and M. Cerullo, Physical Review Letters 64, 1943 (1990). 142 B. Elman, E. S. Koteles, P. Melman, C. Jagannath, J. Lee, and D. Dugger, Applied Physics Letters 55, 1659 (1989). 143 J. H. Vandermerwe, Discussions of the Faraday Society 201 (19 49). 144 J. H. Vandermerwe, Proceedings of the Physical Society of London Section A 63, 616 (1950). 145 J. H. Vandermerwe, Journal of Applied Physics 34, 117 (1963). 146 J. H. Vandermerwe, Journal of Applied Physics 34, 123 (1963). 147 J. W. Matthews, Jour nal of Vacuum Science & Technology 12, 126 (1975).

PAGE 212

212 148 J. W. Matthews, D. C. Jackson, and A. Chambers, Thin Solid Films 26, 129 (1975). 149 E. Ertekin, P. A. Greaney, D. C. Chrzan, and T. D. Sands, Journal of Applied Physics 97 (2005). 150 M. A. Slawinski, Waves and rays in elastic continua Vol. 1, 2 ed. (World Scientific Publishing Company, 2007). 151 S. Raychaudhuri and E. T. Yu, Journal of Vacuum Science & Technology B 24, 2053 (2006). 152 L. C. Chuang, M. Moewe, C. Chase, N. P. Kobayashi, C. Chang Hasn ain, and S. Crankshaw, Applied Physics Letters 90, 3 (2007). 153 J. P. Hirth and J. Lothe, Theory of dislocations (McGraw Hill Book Company, New York, 1968). 154 M. K. a. V. C. a. A. T. a. H. Riechert, Nanotechnology 21, 245705 (2010). 155 Y. H. Kim, H. J. Park, K. Kim, C. S. Kim, W. S. Yun, J. W. Lee, and M. D. Kim, Applied Physics Letters 95 (2009). 156 R. F. Xiang, Y. Y. Fang, J. N. Dai, L. Zhang, C. Y. Su, Z. H. Wu, C. H. Yu, H. Xiong, C. Q. Chen, and Y. Hao, Journal of Alloys and Compounds 509, 2227 (2 011). 157 A. Yamamoto, Y. Hamano, T. Tanikawa, B. K. Ghosh, and A. Hashimoto, physica status solidi (c) 0, 2826 (2003). 158 F. R. Hu, K. Ochi, Y. Zhao, B. S. Choi, and K. Hane, Journal of Crystal Growth 294, 197 (2006). 159 M. A. Mastro, Thesis, University of Florida, 2001. 160 M. D. Reed, Thesis, University of Florida, 2002. 161 J. M. Zuo and J. C. Mabon, in Microsc Microanal ; Vol. 10 (Suppl 2) (2004). 162 M. Law, J. Goldberger, and P. D. Yang, Annual Review of Materials Research 34, 83 (2004). 163 G. Sery ogin, I. Shalish, W. Moberlychan, and V. Narayanamurti, Nanotechnology 16, 2342 (2005). 164 M. Zervos and A. Othonos, Journal of Crystal Growth 312, 2631 (2010). 165 T. Yanagihara, Japanese Journal of Applied Physics Part 1 Regular Papers Short Notes & Rev iew Papers 21, 1554 (1982).

PAGE 213

213 166 M. K. Zayed and H. E. Elsayed Ali, Thin Solid Films 489, 42 (2005). 167 J. E. Northrup and J. Neugebauer, Physical Review B 60, R8473 (1999). 168 C. Xue, Y. Wu, H. Zhuang, D. Tian, Y. a. Liu, X. Zhang, Y. Ai, L. Sun, and F. Wang, Physica E: Low dimensional Systems and Nanostructures 30, 179 (2005). 169 Y. Ai, C. Xue, C. Sun, L. Sun, H. Zhuang, F. Wang, H. Li, and J. Chen, Materials Letters 61, 2833 (2007). 170 W. Han, S. Fan, Q. Li, and Y. Hu, Science 277, 1287 (1997). 171 H. Y. Peng, X. T. Zhou, N. Wang, Y. F. Zheng, L. S. Liao, W. S. Shi, C. S. Lee, and S. T. Lee, Chemical Physics Letters 327, 263 (2000). 172 P. Sahoo, J. Basu, S. Dhara, H. Fang, C. P. Liu, T. Ravindran, S. Dash, and A. Tyagi, Journal of Materials Science 47 3447 (2012). 173 B. S. Simpkins, L. M. Ericson, R. M. Stroud, K. A. Pettigrew, and P. E. Pehrsson, Journal of Crystal Growth 290, 115 (2006). 174 X. Sun and Y. Li, Angewandte Chemie International Edition 43, 3827 (2004).

PAGE 214

214 BIOGRAPHI CAL SKETCH Vaibhav Chaudhari was born in Delhi India He received his Bachelor of Chemical Engineering degree (B.Chem.Engg.) in July 2004 from Mumbai University Institute of Chemical Technology After B.Chem.Engg. he joined University of Toledo, Ohio for Master of Science in Chemical Engineering in the Fall of 2004. At Toledo, he worked under guidance of Dr. Dong Shik Kim towards h is M.S. degree. His M.S. thesis work involved use of genetically modified bacteria for production of ethanol biofuel from diff erent bio wastes. He graduated from University of Toledo in summer 2004. Right a fter M.S. degree, he joined Ph.D. program at the University of Florida in August 2006. He joined the electronic materials processing group (EMPG) to work under guidance of Dr. Tim Anderson His work at EMPG involved studying growth of gallium nitride and indium nitride on silicon substrates using merged metal organic hydride vapor phase epitaxy