Effect of Process Parameters on the Growth of N-polar GaN on Sapphire by MOCVD

Transcription

1 Effect of Process Parameters on the Growth of N-polar GaN on Sapphire by MOCVD A Thesis Submitted For the Degree of Doctor of Philosophy in the Faculty of Science by G R Krishna Yaddanapudi Department of Materials Engineering Indian Institute of Science Bangalore INDIA February 2016

2

3 Declaration I hereby declare that the work reported in this thesis is entirely original. It was carried out by me in the Department of Materials Engineering & Centre for Nano Science and Engineering (CeNSE), Indian Institute of Science, Bangalore. I further declare that it has not formed the basis for the award of any degree, diploma, membership, associateship or similar title of any university or institution. Date: (G R Krishna Yaddanapudi) iii

4 G R Krishna Yaddanapudi February 2016 All rights reserved

5 To My Loving Parents... v

6

7 Abstract Group III-Nitrides (GaN, InN & AlN) are considered one of the most important class of semiconducting materials after Si and GaAs. The excellent optical and electrical properties of these nitrides result in numerous applications in lighting, lasers, and high-power/high-frequency devices. Due to the lack of cheap bulk III- Nitride substrates, GaN based devices have been developed on foreign substrates like Si, sapphire and SiC. These technologies have been predominantly developed on the so called Ga-polarity epitaxial stacks with growth in the [0001] direction of GaN. It is this orientation that grows most easily on sapphire by metal organic chemical vapor deposition (MOCVD), the most common combination of substrate and deposition method used thus far. The opposite [000 1] or N-polar orientation, very different in properties due to the lack of an inversion centre, offers several advantages that could be exploited for better electronic and optoelectronic devices. However, its growth is more challenging and needs better understanding. The aim of the work reported in this dissertation was a systematic investigation of the relation between the various growth parameters which control polarity, surface roughness and mosaicity of GaN on non-miscut sapphire (0001) wafers for power electronics and lighting applications, with emphasis on the realization of N-polar epitaxial layers. GaN is grown on sapphire (0001) in a two-step process, which involves the deposition of a thin low temperature GaN nucleation layer (NL) on surface modified sapphire followed by the growth of high temperature device quality GaN epitaxial layer. The processing technique used is MOCVD. Various processing methods for synthesis of GaN layers are described with particular emphasis on MOCVD method. The effect of ex situ cleaning followed by an in situ cleaning on the surface morphology of sapphire (0001) wafers is discussed. The characterization tools used in this dissertation for studying the chemical bond nature of nitrided sapphire surface and microstructural evolution (morphological and structural) of GaN layers are described in detail. The effect of nitridation temperature (T N ) on structural transformation of nonmiscut sapphire (0001) surface has been explored. The structural evolution of nitrided layers at different stages of their process like as grown stage and thermal

8 annealing stage is investigated systematically. The chemical bond environment information of the nitrided layers have been examined by x-ray photoelectron spectroscopy (XPS). It is found that high temperature nitridation (T N 800 o C) results in an Al-N tetrahedral bond environment on sapphire surface. In contrast, low temperature nitridation (T N = 530 o C) results in a complex Al-O-N environment on sapphire surfaces. Microstructural evolution of low temperature GaN NLs has been studied at every stage of processing by scanning electron microscopy (SEM) and atomic force microscopy (AFM). Surface roughness evolution and island size distribution of NLs measured from AFM are discussed. It is found that NLs processed on sapphire wafers nitrided at (T N 800 o C) showed strong wurtzite [0002] orientation with sub-nanometre surface roughness. In contrast, NLs processed at (T N = 530 o C) showed zinc blende phase in the as grown stage with higher surface roughness, but acquired a greater degree of wurtzite [0002] orientation after thermal annealing prior to high temperature GaN growth. Polarity, surface quality and crystal quality of subsequently grown high temperature GaN epitaxial layers is described in relation to the structure of the transformed nitrided layers. Higher nitridation temperatures (T N 800 o C) consistently yield N-polar GaN whereas lower nitridation temperatures (T N = 530 o C) yield Gapolar GaN. It is found that the relative O atom concentration levels in nitrided layers control the density of inversion domains in N-polar GaN. The effect of various growth parameters (NH 3 flow rate, growth temperature, NL thickness) on surface morphology and mosaicity of both Ga & N-polar GaN layers is discussed in detail. We report device quality N-polar GaN epitaxial layers on non-miscut sapphire (0001) wafers by careful optimization of growth temperature. It is found that lower growth temperatures (800 o C) are favorable for obtaining smooth N- polar GaN layers. In contrast, N-polar GaN layers grown at higher temperatures (1000 to 1080 o C) are rough with hexagonal hillocks.

9 Acknowledgements It has been a great experience to be part of the successful nitride HEMT research program at IISc. I am sincerely thankful to my adviser Professor Dipankar Banerjee, working under him has indeed been a privilege for me and has also been a very rewarding experience. Sincere thanks to my second adviser Professor Srinivasan Raghavan early on gave me the opportunity to work in his world class MOCVD lab, and to get deeply involved with the hardware. Both of them gave me freedom to explore my own ideas. During the course of my work I was introduced to several scientific techniques and processes and in the process also I learned a lot about microstructural analysis from my advisers. Many thanks to my colleagues (MOCVD growers): Abheek, Hareesh and Nagaboopathy, who has taught me almost everything I know about MOCVD reactor. I would also like to thank them, with whom I have spent late growth nights, machine trouble shooting and repairs. I would also like to thank staff at micro and nano characterization facility (MNCF) at CeNSE, IISc and advanced facility for microscopy and micro analysis (AFMM), comes under the department of Materials Engineering, IISc, for keeping the facilities in good condition and providing training on the equipments. I am sincerely thankful to my mother and father for their patience and love. Last but not least, my other colleagues and friends, who has always inspired, motivated, entertained me during the good and bad times of my PhD study. To all of you. Thank you. ix

10

11 Contents Declaration of Authorship iii Abstract vii Acknowledgements ix List of Figures List of Tables Abbreviations xv xxi xxiii 1 Introduction Gallium Nitride Substrates for GaN epitaxy Structure and polarity of GaN Polarity and growth of GaN GaN for lighting and power electronic applications Effect of polarity Thesis description Experimental Techniques for GaN Synthesis & Characterization Experimental Processing techniques for GaN synthesis Metal-organic chemical vapor deposition (MOCVD) Time-Temperature (TT) process plot for GaN epitaxy on sapphire by MOCVD Characterization X-ray photoelectron spectroscopy (XPS) High resolution x-ray diffractometer Atomic force microscopy (AFM) Differential interference contrast (DIC) light microscopy.. 30 xi

12 Contents xii Scanning electron microscopy (SEM) Transmission electron microscopy (TEM) In situ reflectivity and stress monitor analysis tool Sapphire Pre-treatment & Microstructural Evolution of LT GaN Background Experimental Results In situ thermal treatment of sapphire wafers Nitridation of in situ treated sapphire wafers As nitrided sapphire wafers Annealed nitrided wafers LT GaN nucleation layer Morphological evolution Structural evolution Discussion Low temperature nitridation (T N = 530 o C) High temperature nitridation (T N = 1100 o C) Summary & conclusions Polarity & Microstructural Evolution of HT GaN Background Experimental Results Low temperature nitridation (T N = 530 o C) V/III ratio Growth temperature LT GaN NL thickness Polarity of HT GaN High temperature nitridation (T N = 1100 o C V/III ratio Polarity of HT GaN LT GaN annealing time Growth temperature Carrier gas (H 2 /N 2 ) Summary of results Low temperature nitridation (T N = 530 o C) High temperature nitridation (T N = 1100 o C) Discussion Low temperature nitridation (T N = 530 o C) and HT GaN Polarity Crystalline quality of HT GaN Surface roughness of HT GaN High temperature nitridation (T N = 1100 o C) and HT GaN. 88

13 Contents xiii Polarity Surface quality of HT GaN Crystalline quality of HT GaN Summary & conclusions Summary, Conclusions and Future Work 93 A Specifications of MOCVD Reactor 97 Bibliography 99

14

15

16

17 List of Figures 1.1 Non-centro symmetric wurtzite structure of GaN Architecture of HEMT device Block diagram of MOCVD reactor Aixtron MOCVD Reactor TT process plot Variation in the equilibrium partial pressure of Ga (P Ga ) with the input V/III ratio Variation in the equilibrium partial pressure of Ga (P Ga ) with the input V/III ratio for our MOCVD experimental conditions Surface diffusion lengths of Ga adatoms Mosaic block model: Tilt & Twist angles HRXRD goniometer rotational angles AFM tip-surface force curve DIC light microscopy optical lens diagram k-space MOSS in-situ tool k-space MOSS stress measurement Surface morphology of sapphire wafers after ex situ cleaning followed by in situ treatment in MOCVD reactor Surface morphology of sapphire wafers after ex situ cleaning followed by in situ treatment in MOCVD reactor N 1s XPS spectrum of nitrided sapphire wafers Normalized O 1s & N 1s XPS intensities from nitrided sapphire wafers De-convoluted N 1s XPS spectra from nitrided sapphire wafers De-convoluted N 1s XPS spectra from nitrided sapphire wafers SEM morphologies of LT GaN AFM morphologies of LT GaN AFM surface roughness of LT GaN High resolution x-ray (0002) ω-scan profiles for LT GaN NLs High resolution x-ray φ-scan profiles for LT GaN NLs High resolution x-ray φ-scan profiles for LT GaN NLs Optical reflectivity trace of GaN on sapphire Surface morphology of N-polar GaN xvii

18 List of Figures xviii 4.3 HRXRD rocking curve FWHM values of N-polar GaN in comparison to conventional Ga-polar GaN Surface morphology of HT GaN (V/III = 485) grown for sapphire nitrided at T N = 530 o C Surface morphology of HT GaN (V/III = 965, 1055 & 1205)) grown for sapphire nitrided at T N = 530 o C RMS surface roughness data of HT GaN (V/III = 965, 1055 & 1205)) grown for sapphire nitrided at T N = 530 o C growth rate and roughening recovery time of HT GaN samples deposited at different V/III ratios for sapphire nitridation at T N = 530 o C HRXRD rocking curve FWHM values of HT GaN samples deposited at different V/III ratios for sapphire nitridation at T N = 530 o C Optical micrographs of HT GaN grown at temperatures (a) 1000 o C, (b) 1025 o C and (c) 1050 o C for sapphire nitridation at T N = 530 o C AFM surface roughness data of HT GaN layers deposited at growth temperatures: 1000, 1025 and 1050 o CC for nitridation at T N = 530 o C Roughening recovery and growth rate of HT GaN samples grown at different growth temperatures for sapphire nitridation at T N = 530 o C HRXRD rocking curve FWHM values of HT GaN samples grown at different growth temperatures for sapphire nitridation at T N = 530 o C AFM surface roughness data of HT GaN as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C Roughening recovery time data of HT GaN as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C HRXRD rocking curve FWHM values of HT GaN samples grown as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C SEM morphologies of HT GaN samples before and after KOH etch experiment for sapphire wafers nitrided at T N = 530 o C Nomarski optical microscopy images of HT GaN layers for sapphire wafer nitrided at T N = 1100 o C Nomarski optical microscopy images of HT GaN layers (V/III = 965, 1130 & 1205) for sapphire wafer nitrided at T N = 1100 o C HRXRD rocking curve FWHM values of HT GaN layers (V/III = 965, 1130 & 1205) for sapphire wafer nitrided at T N = 1100 o C SEM morphologies of HT GaN before and after KOH etch experiment for sapphire nitridation at T N = 1100 o C SEM morphologies of HT GaN layers for sapphire nitridation at T N = 1100 o C SEM morphologies after KOH etch experiment of HT GaN layers for sapphire nitridation at T N = 1100 o C

19 List of Figures xix 4.23 Optical microscopy images of HT GaN layers grown at different growth temperatures for sapphire nitrided at T N = 1100 o C HRXRD rocking curve FWHM values of HT GaN layers grown at temperatures for sapphire nitridation at T N = 1100 o C Optical images of HT GaN layers grown at low growth temperatures for sapphire nitridation at T N = 1100 o C Surface roughness of HT GaN layers grown at low growth temperatures for sapphire nitridation at T N = 1100 o C Optical images of HT GaN layers grown under N 2 as carrier gas for sapphire nitridation at T N = 1100 o C Summary of x-ray rocking curve FWHM values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 530 o C SEM morphologies of LT GaN annealed under different flow rates of NH 3 for sapphire nitridation at T N = 530 o C Summary of surface roughness values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 530 o C Normalized O 1s intensity of nitrided sapphire wafers at different stages for sapphire nitridation at T N = 1100 o C Summary of surface roughness values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 1100 o C Summary of x-ray rocking curve FWHM values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 1100 o C

20

21 List of Tables 1.1 Properties of various commonly available foreign substrates available for III-Nitride epitaxy Structural information and bond energy of wurtzite GaN Band gaps of some typical WBG semiconducting materials in relation to Ge and Si Properties of various commonly available processing techniques for III-Nitride epitaxy Group-III & V precursors used for III-Nitride epitaxy Vapor pressure constants a and b for common metalorganic precursors Activation energy E A for decomposition of GaN, and desorption of Ga and N atoms from GaN Surface diffusion barriers for Ga & N adatoms on Ga-polar and N-polar GaN surfaces Normalized intensities of N 1s deconvoluted peaks from various possible nitrided layer structures from T N = 530, 800 & 1100 o C. Normalization has been done with respect to Al 2p peak HT GaN growth parameter space for sapphire nitridation at T N = 530 o C HT GaN growth parameter space for sapphire nitridation at T N = 1100 o C FWHM values of x-ray rocking curves for the HT GaN samples grown directly on LT GaN after ramp up, in relation to the samples grown on 4 min annealed LT GaN for sapphire nitridation at T N = 1100 o C The surface roughness values of N-polar HT GaN grown at low & high growth temperatures for sapphire nitridation at T N = 1100 o C HRXRD rocking curve FWHM values of N-polar HT GaN grown at low & high growth temperatures for sapphire nitridation at T N = 1100 o C xxi

22

23 Abbreviations MOCVD HVPE MBE LT HT NL TMGa TMAl TMIn HEMT 2DEG CTE 2D 3D SEM AFM RMS XPS TEM CBED HRXRD XRC FWHM ID Metal Organic Chemical Vapor Deposition Hydride Vapor Phase Epitaxy Molecular Beam Epitaxy Low Temperature High Temperature Nucleation Layer Tri Methyl Gallium Tri Methyl Aluminium Tri Methyl Indium High Electron Mobility Transistor 2 Dimensional Electron Gas Coefficient of Thermal Expansion 2 Dimensional 3 Dimensional Scanning Electron Microscopy Atomic Force Microscopy Root Mean Square X-ray Photoelectron Microscopy Transmission Electron Microscopy Convergent Beam Electron Diffraction High Resolution X-ray Diffractometer X-ray Rocking Curve Full Width at Half Maximum Inversion Domains xxiii

24 Abbreviations xxiv QW SQW LED DFT Quantum Well Single Quantum Well Light Emitting Diode Density Functional Theory

25 Chapter 1 Introduction 1.1 Gallium Nitride Wide band gap semiconductor Gallium Nitride (GaN) and its ternary and quaternary alloys along with InN and AlN, have excellent optical and electrical properties because of which they have found applications in a range of optoelectronic and high-power/high-frequency electronic applications [1 3]. In the absence of native bulk III-Nitride crystals for homo-epitaxy, these materials have been deposited heteroepitaxially on sapphire, Si and SiC substrates Substrates for GaN epitaxy Table 1.1 tabulates the comparison between various foreign substrates for III- Nitride epitaxy [2, 4]. Sapphire is the first substrate on which GaN technology was developed and remains the most favored substrate for optoelectronic applications such as light emitting diodes (LEDs) in the visible and the upcoming UV range. The cost is reasonable (non-miscut c-plane wafer) at $ per 2 inch wafer. The large lattice mismatch and coefficient of thermal expansion (CTE) mismatch between GaN and sapphire (0001) results in GaN films with defect density of the order of 10 8 cm 2 with residual stresses that are compressive in nature. 1

26 Chapter 1. Introduction 2 Property Si Sapphire SiC AlN GaN (111) (0001) (0001) (0001) (0001) Lattice Mismatch % Available Size Cost per 2 wafer Thermal Conductivity κ, W K 1 m Biaxial Modulus, GPa CTE ( 10 6 ), K Dislocation Density, cm Table 1.1: Properties of various commonly available foreign substrates available for III-Nitride epitaxy The drawback of sapphire is its low thermal conductivity, which makes it difficult to dissipate heat from devices during operation. The alternative substrate, SiC, was for long not suitable for RF devices due to electrical conductivity in the substrate. In 1999, the first semi-insulating SiC substrates became available, and SiC has since become the leading substrate for high-power devices, due to its high thermal conductivity. The other advantage with SiC is its low lattice mismatch with GaN. The high cost is the major drawback for SiC substrates; currently semi-insulating 4H-SiC retails for $ 1200 to 1500 per 2 inch wafer, which is nearly 20 times the cost of sapphire. The advantage of Si is its availability in large sizes and its cost ( $ 20 to 25 per 2 inch Si (111) wafer). Ga-polar GaN based HEMTs growth on Si (111) is well established. The tensile nature of residual stress (due to CTE mismatch) in GaN on Si (111) results in cracking on cool down unless specific stress mitigating strategies are adopted. In a very commonly used solution, AlGaN transition layers (graded, step-graded), are inserted between GaN and Si, to induce compressive growth stress in the GaN layer to overcome the residual tensile stress on cool down [5 8]. The other drawback with Si is that the GaN layers grown are highly defective (dislocation density is 1 to 2 orders of magnitude higher than GaN films grown on sapphire and SiC) due to the large lattice and CTE mismatch. In summary, all three substrates have their advantages and disadvantages as a result of which they find their niche areas of technological relevance.

27 Chapter 1. Introduction 3 This thesis is concerned with growth on sapphire with specific focus on the ability to control the polarity of GaN. This forms the subject matter of the next section. 1.2 Structure and polarity of GaN GaN crystallizes in a stable non-centro symmetric wurtzite structure (P63mc) in addition to a metastable zincblende structure (F 43m). The non-centro symmetric wurtzite GaN can either terminate with metal atom (Ga) as the outer layer along [0001] (+c) direction or anion (N) atom as the outer layer along [000 1] (-c) direction (Fig. 1.1 & Table 1.2). For Ga-terminated structure, each Ga atom is bonded to 3 N atoms below where as for N-terminated structure each N atom is bonded to 3 Ga atoms below (Fig. 1.1). This apparently simple crystallographic actually has dramatic consequences for practically every application. As will be shown in this thesis the growth modes of these two polarities also show very little resemblance whatsoever to each other. In particular, most current technology has been developed on Ga-polar GaN based materials as it is the default growth direction. However, the N-polar material which is more difficult to grow, has many interesting aspects that can enable better power or optoelectronic devices than the Ga-polar material. Following a brief discussion on the ability of the III-Nitrides to enable various applications, the high electron mobility transistor, a most common and simple device is used to illustrate the difference between and the advantage of the N-polar nitrides in section. 1.3 Polarity and growth of GaN There are several processing techniques available for III-Nitride materials growth. Among them metal organic chemical vapor deposition (MOCVD), molecular beam epitaxy (MBE) and hydride vapor phase epitaxy (HVPE) are commonly used. The processing technique used in this dissertation is MOCVD available at CeNSE, IISc Bangalore.

28 Chapter 1. Introduction 4 Figure 1.1: Non-centro symmetric wurtzite structured GaN with metalpolarity/ga-polarity and N-polarity. Property GaN Lattice Parameters, nm a = 0.319, c = Space Group P63mc Bond Length 1, nm Bond Length 2, nm Bond Length 3, nm Bond Length 4, nm Bond Length 5, nm Bond Energy, ev/atom 8.92 Table 1.2: Structural information and bond energy of wurtzite GaN The process parameters which are associated with the MOCVD reactor are: temperature, pressure, flow rates of group-iii precursors for Ga, Al & In, group-v precursors for N, and carrier gas flow rates (H 2 and/or N 2 ). A sapphire pre-treatment for polarity selection of nitrides, called sapphire nitridation in this dissertation, involves the transformation of sapphire (0001) surface to a thin complex unknown AlO x N 1 x [9] or AlN [10] layer prior to the growth of subsequent nitride layers. The nitrided layer reduces the chemical dissimilarity between GaN and sapphire and reduces the lattice mismatch [11, 12], both of which contribute to a reduction in interface energy, thereby promoting lateral growth. It can act as a wetting layer for the subsequently grown nitride layers and it can also aid in obtaining the orientation relationships of these with respect to the underlying sapphire substrate [12 14].

29 Chapter 1. Introduction 5 The two-step growth process, using MOCVD has been used extensively in improving the structural, optical and electrical quality of nitride epitaxial layers in particular GaN [15, 16]. This process involves the deposition of a low temperature (LT) GaN/AlN nucleation layer (NL) on sapphire wafers prior to the deposition of main high temperature (HT) GaN epitaxial layers. Each of these processing steps are associated with various growth parameters. The LT NL can be controlled with parameters like growth time, temperature and flow rates of precursors for NL deposition. Similar parameter optimisation is required for HT GaN deposition as well. In the two-step growth method, the LT layer helps to provide nuclei of an optimum density and orientation for subsequently grown HT GaN layer. The optimum nucleation density is provided by depositing a thin GaN/AlN NL at low temperature followed by its controlled annealing. The HT GaN layer then grows by addition of growth species, N and Ga, to the lateral edges of these nuclei. While the 2-step method is use for N-polar material as well, fine differences exist. The growth mechanism of N-polar GaN and its alloys is not the same as Ga-polar GaN due to their differently terminated surfaces. From total energy density functional theory (DFT) calculations it has been predicted that the Ga and N adatom surface diffusion barriers are different for Ga-polar GaN and N-polar GaN surfaces [17]. Thus, the growth parameters which are optimized for Ga-polar GaN growth on sapphire are not suitable for N-polar GaN growth. In spite of all the success achieved with Ga-polar GaN and its alloys, N-polar GaN materials grown under identical conditions yield rough surface morphology with hexagonal hillocks [18, 19]. Growth parameters which effect the surface morphology and crystalline quality of N-polar GaN are reactor pressure (P), carrier gas (H 2 and/or N 2 ), V/III ratio and nitridation temperature (T N ). The parameter V/III ratio is simply the ratio of the fluxes of group-iii to group-v precursors which are allowed into the reactor chamber. The hillock density and size of hillocks on N-polar GaN are very sensitive to V/III ratio and the density increases with the V/III ratio [18]. Longer nitridation time also seems to affect the surface morphology of N-polar GaN [19]. It has been shown that N 2 carrier gas improves the surface morphology of N- polar GaN whereas H 2 carrier gas suppresses the lateral growth of N-polar GaN

30 Chapter 1. Introduction 6 [20 22]. The above reports indicate that N-polar GaN growth is very sensitive to growth parameters, and therefore detailed understanding and careful optimization of growth parameters is required to obtain device quality layers. Keller et al., showed that device quality N-polar GaN epitaxial layers without hexagonal hillocks can be obtained by using intentionally miscut sapphire (0001) wafers of 2 o to 4 o along a- and m-directions [23]. However, the cost of highly miscut wafers are 4 times the cost of non-miscut sapphire (0001) wafers GaN for lighting and power electronic applications Optoelectronics was the first major application of III-Nitrides that was successfully commercialized following the pioneering research of Akasaki, Amano and Nakamura. Nakamura et al. demonstrated high-brightness blue, green, and yellow light emitting diodes (LEDs) with InGaN quantum well (QW) structures in 1995 [24]. Ternary InGaN alloys are used as the active layer in GaN-based LEDs and lasers [24]. The performance of blue and green single quantum well (SQW) structure LEDs has been improved and at 20 ma, the output power and the external quantum efficiency (EQE) of the blue SQW LEDs were 5 mw and 9.1%, respectively. Those of green SQW LEDs were 3 mw and 6.3%, respectively [25]. This LED epitaxial structure is still the basic foundation for all currently commercially available first-generation blue and green LEDs [25]. It is found that the efficiency of these LEDs is strongly depends on the In incorporation in the active InGaN layer [26]. Following optoelectronics, GaN based power electronics and high frequency electronics is a currently growing area. The GaN based high electron mobility transistors (HEMTs) has evolved tremendously from their first modest demonstration in 1993 [27] and is the workhorse of this group of applications. The large band gaps of GaN and AlGaN provide for large breakdown fields, and thermal stability of the materials allows for a high temperature of operation [28]. These excellent properties have led to the demonstrations of devices, with current densities as high as 2.3 A/mm [29], breakdown voltages around 0.9 kv [30], and power densities of

31 Chapter 1. Introduction 7 Material Chemical Symbol Band Gap Energy (ev) Type Germanium Ge 0.7 Indirect Silicon Si 1.1 Indirect Gallium Arsenide GaAs 1.4 Direct Silicon Carbide SiC 3.3 Indirect Zinc Oxide ZnO 3.4 Direct Gallium Nitride GaN 3.4 Direct Diamond C 5.5 Direct Aluminum Nitride AlN 6.02 Direct Table 1.3: Band gaps of some typical WBG semiconducting materials in relation to Ge and Si W/mm at 4 GHz [31]. Table 1.3 shows band gaps of some typical wide ban gap semiconductor materials in relation to Ge and Si [1 3] Effect of polarity The tremendous impact of polarity is demonstrated through its effect on the design and performance of GaN/AlGaN HEMTs. The success of III-Nitride material system is not only due to their direct WBG, intrinsic bulk material transport properties (the p-material properties do not compare well with n-properties in III- Nitride materials), but also due to the interface properties. These polar nitrides behave differently across the interface of the active layers of devices (Fig. 1.2). For example, the difference in polarization across the Al 0.25 Ga 0.75 N/GaN interface, in combination with discontinuity in conduction band yields a two dimensional electron gas (2DEG) density across the interface, well in excess of cm 2 [32] even in undoped systems. This is in contrast to AlGaAs/GaAs HEMTs, where doping is required to form a 2DEG. The enhancement in the density, mobility and quantum confinement of 2DEG formed across the interface are some of the primary requirements for HEMT devices. The following are the parameters which control the properties of 2DEG formed across the interface

32 Chapter 1. Introduction 8 ˆ 2DEG carrier density can be enhanced by choosing nitride hetero structures which have higher discontinuity in polarization across the interface (like GaN/AlN, of course polarization direction matters). ˆ The parameters which control 2DEG carrier mobility are surface roughness of base layer, alloy scattering and critical thickness of AlGaN layer for HEMT stack and dislocation density level in base layers. ˆ Quantum confinement of 2DEG can be improved by choosing nitride hetero structures which give high conduction band discontinuity across the interface. Ga-polar Al x Ga 1 x N/GaN HEMTs (the device structure is shown in Fig. 1.2) yields a 2DEG carrier density across the interface that depends on the composition x of Al in a thin Al x Ga 1 x N layer (which is under tension) sitting on GaN (Fig. 1.2a). Usually x ranges from 25 to 30%, which in turn limits the critical thickness of Al x Ga 1 x N ( 25 to 20 nm for x 25 to 30%). The density of 2DEG increases with the Al content in Al x Ga 1 x N due to increased spontaneous and piezo polarizations in the Al x Ga 1 x N layer. However, higher Al content results in formation of cracks in the Al x Ga 1 x N under tension and degrades the device performance. In contrast to the above, the design and architecture of HEMT devices for a highly confined 2DEG density (> cm 2 ) involves the deposition of nitride materials with N-polarity (Fig. 1.2b) [33]. N-polar GaN/AlGaN HEMTs improve the confinement of 2DEG due to the WBG AlGaN back barrier (Fig. 1.2b). Strong confinement results in sharp pinch off voltages. The other advantages with N-polar GaN based HEMTs are low contact resistance [34], sharp pinch off voltages due to AlGaN back barrier layers [35, 36]. Higher trans-conductance can be expected with the same gate channel separation used for Ga-polar AlGaN/GaN HEMTs [37, 38] as well as higher 2DEG density [2, 33, 39]. N-polar GaN alloy devices have shown impressive performance which are in comparable to Ga-polar AlGaN/GaN HEMTs in the basic figures of merit [33, 40].

33 Chapter 1. Introduction 9 Figure 1.2: Architectures of HEMT devices: (a) conventional Ga-polar Al- GaN/GaN HEMT, (c) shows the corresponding band structure, (b) & (d) N-polar GaN/AlGaN/GaN HEMT and its corresponding band diagram. It has been also shown that N-polar GaN enhances the In incorporation in the subsequently grown InGaN QWs on the GaN base layer for LED applications, which results in LEDs with greater efficiency [41].

34 Chapter 1. Introduction Thesis description The theme of this dissertation is the development and study of N-polar GaN epitaxial layers on non-miscut sapphire (0001) wafers for power electronics and lighting applications. The processing technique used is MOCVD, available at CeNSE, IISc Bangalore. The main scope of this thesis is to understand the relation between various growth parameters which control the polarity, surface roughness and mosaicity of GaN and epitaxial layers on non-miscut sapphire (0001) wafers. The various steps of growth such as sapphire nitridation conditions and LT GaN growth conditions is extensively studied in this work. Growth parameters such as NH 3 flow rate, growth temperature, and NL thickness have been systematically varied. Particular emphasis is placed on the correlation between the structure of these precursor layers (nitrided layer and LT GaN NL) on the quality (surface morphology and mosaicity) and polarity of the subsequently grown nitride semiconductor layers. We report for the first time device quality N-polar GaN epitaxial layers on non-miscut sapphire (0001) wafers. In chapter 2, various processing methods for synthesis of GaN layers are described with particular emphasis on MOCVD method. The characterization tools used in this dissertation for studying the chemical bond nature of nitrided sapphire surface and micro-structural evolution (morphological and structural) of LT GaN NL & HT GaN layers are described in detail. Chapter 3, starts with the effect of ex situ cleaning followed by an in situ cleaning on the surface morphology of sapphire (0001) wafers. The effect of nitridation temperature (T N ) on structural transformation of non-miscut sapphire (0001) surface has been studied systematically. The structural evolution of nitrided layers at different stages of their processing like as grown stage and thermal annealing stage is investigated systematically. The chemical bond environment information of the nitrided layers has been examined by x-ray photoelectron spectroscopy. It is

35 Chapter 1. Introduction 11 found that high temperature nitridation (T N 800 C) results in a Al-N tetrahedral bond environment on sapphire surface. In contrast, low temperature nitridation (T N = 530 C) results in a complex Al-O-N environment on sapphire surfaces. Micro-structural evolution of LT GaN NLs have been studied at every stage of processing by scanning electron microscopy and atomic force microscopy. Surface roughness evolution is measured from atomic force microscopy is described. It is found that NLs processed on sapphire wafers nitrided at (T N 800 C) showed strong [0002] orientation with sub-nanometer surface roughness. In contrast, NLs processed at (T N = 530 C) showed the presence zincblende phase in the as grown step with higher surface roughness, but acquired a greater degree of wurtzite [0002] orientation after thermal annealing prior to high temperature GaN growth. In chapter 4, polarity, surface quality and crystal quality of subsequently grown HT GaN epitaxial layers is described in relation to the structure of the transformed nitrided layers. High nitridation temperatures (T N 800 C) consistently yield N-polar GaN material whereas low nitridation temperatures (T N = 530 C) yield Ga-polar GaN. It is found that the relative oxygen atom concentration levels in nitrided layers control the density of inversion domains in N-polar GaN layers. The effect of various growth parameters such as (NH 3 flow rate, growth temperature, NL thickness) on surface morphology and mosaicity of both Ga & N-polar GaN layers is discussed in detail. First time we report device quality N-polar GaN layers at low growth temperatures 800 non-miscut sapphire (0001) wafers. Chapter 5 contains the summary of the thesis results followed by directions and suggestions for the future work.

36

37 Chapter 2 Experimental Techniques for GaN Synthesis & Characterization Group III-Nitride epitaxy on sapphire is a well-established platform and variety of experimental techniques such as metal organic chemical vapor deposition (MOCVD), molecular beam epitaxy (MBE) and hydride vapor phase epitaxy (HVPE) are available to deposit these semiconducting nitride materials. This chapter contains a brief description of various processing techniques for GaN synthesis with particular emphasis on MOCVD method followed by a brief introduction to various characterization tools which were used in this dissertation. 2.1 Experimental The non-miscut 2 inch sapphire (0001) wafers were brought from the following vendors EPISTONE, MONOCRYSTAL and KYOCERA. The unintentional miscuts of the wafers provided by vendors were in the range of 0 ± 0.3 o. The wafers were diced into 1 cm 2 size pieces by a MTI Corporation EC400 wafer dicing saw, which is furnished with diamond blades of different thicknesses varying from 300 to 100 µm. The saw is computerized with a positional accuracy of 0.01 mm. The diced wafers were then ultra-sonicated and cleaned ex-situ with acetone, isopropanol 13

38 Chapter 2. Experimental Techniques & Characterization 14 Property HVPE MBE MOCVD Growth Rate 50 µm/h 50 nm/h 2 µm/h Strength Large scale production Sharp interface Large scale production Good quality film In-situ monitor Sharp interface Very high growth rate High purity High purity H 2 free ambient No ultra high vacuum Plasma assisted growth In-situ monitor Laser assisted growth High growth rate Uniformity Weakness Complex process Expensive Expensive sources Extreme temperature Need ultra-high vacuum Hazardous sources conditions Hazardous sources Low growth rate Large quantities of NH 3 required Table 2.1: Properties of various commonly available processing techniques for III-Nitride epitaxy and de-ionized water. It is observed that the ex situ cleaning procedure is very critical and it effects the surface morphology of sapphire wafers, as described in subsequent chapter Processing techniques for GaN synthesis Polycrystalline wurtzite GaN was originally synthesized using HVPE on sapphire wafers in 1969 [42]. Subsequently several technical breakthroughs enabled single crystal wurtzite GaN, low residual background carrier concentration in undoped GaN, conductivity control of p-type GaN, epitaxial layer stacks for LEDs, LDs and HEMTs, and these led to the first modest LED and HEMT devices being introduced in 1993 [27, 43]. Table 2.1 shows the comparisons between the commonly available processing techniques used for III-Nitride epitaxy.

39 Chapter 2. Experimental Techniques & Characterization Metal-organic chemical vapor deposition (MOCVD) MOCVD is a process for the deposition of materials that utilizes volatile metal organic compounds to transport metallic atoms that are relatively non-volatile at the convenient deposition temperature. The organometallic compounds are usually mixed with other source materials such as Hydrides that react to form compound semiconductor films. Fig. 2.1 shows the simple block diagram of the horizontal flow MOCVD reactor. Fig. 2.2 shows the AIXTRON 200/4 RF-S MOCVD reactor which is installed at CeNSE, IISc. Figure 2.1: Simple block diagram of horizontal flow MOCVD reactor The group-iii precursors (TMGa, TMAl & TMIn, Table 2.2) are injected into the reactor chamber with the aid of carrier gas (H 2 /N 2 ). Group-III & group-v precursors are injected into the reactor chamber through separate nozzles. The separation plate inside the chamber prevents the pre-reaction of these precursors. Eventually, all these precursors are allowed to react onto the substrate surface which is held at typical growth temperatures. Unwanted products formed will be pumped to the scrubber for dilution before they are released to the outer atmosphere. The substrate sits on a rotating susceptor to maintain the uniformity of the deposited film. NH 3 is used as the group-v hydride source and H 2 used as the

40 Chapter 2. Experimental Techniques & Characterization 16 Figure 2.2: Aixtron AIX 200/4 RF horizontal flow MOCVD reactor available at CeNSE, IISc Bangalore. carrier gas. All the group-iii & group-v precursor bubblers and gas lines are furnished with the necessary mass flow controllers (MFCs) and pressure controllers (PCs) to control the flow rates of gases during the deposition of materials. The maximum temperature limit of the reactor is 1500 o C. The minimum reactor pressure is 10 mbar and the maximum is 1000 mbar. The reactor chamber is equipped with the removable quartz ware to prevent the reactor walls from contamination during the growth of materials. The reactor is also equipped with an in-situ real time thickness and stress monitor tool to understand the stress evolution and growth behavior of films.

41 Chapter 2. Experimental Techniques & Characterization 17 Precursor Chemical Formula Trimethyl Gallium (TMGa) (CH 3 ) 3 Ga Trimethyl Aluminium (TMAl) (CH 3 ) 3 Al Trimethyl Indium (TMIn) (CH 3 ) 3 In Ammonia NH 3 Carrier Gas H 2, N 2 Table 2.2: Group-III & V precursors used for III-Nitride epitaxy Time-Temperature (TT) process plot for GaN epitaxy on sapphire by MOCVD Fig. 2.3 shows the standard time-temperature (TT) process diagram for two-step GaN epitaxy on sapphire wafers at two different nitridation temperatures T N = 530 and 1100 o C. Figure 2.3: Typical temperature Vs time (TT) process plot for two-step GaN epitaxy on sapphire wafers The process starts with in situ thermal cleaning of sapphire wafers under the flow of purified H 2 at a temperature of 1100 o C. The other process steps which are involved in GaN epitaxy are ˆ Nitridation of sapphire at T N = 500 to 1100 o C to enable a structural transformation of sapphire surface to reduce the chemical dissimilarity between sapphire and subsequently grown GaN layers

42 Chapter 2. Experimental Techniques & Characterization 18 ˆ LT GaN NL deposition at temperature 500 to 600 o C to provide optimum nucleation density for the subsequently grown HT GaN epitaxial layers ˆ Annealing of LT GaN NL at temperature 1000 to 1080 o C involves decomposition of NLs, surface migration of decomposed atoms, incorporation and growth of new GaN nuclei. ˆ Growth of HT GaN epitaxial layer at temperature 1000 to 1080 o C involves epitaxial layer growth on annealed NLs, with adequate kinetics to grow the necessary thickness. The synthesis of GaN by using MOCVD technique is well known [44]. The formation of GaN involves the complex chemical reactions between TMGa, NH 3 and their intermediate gas phase adducts (TMGa:NH 3 ). The corresponding formation energies for each step in the reaction involves are not documented well. The formation energy of GaN can be estimated by assuming a simple model, where Ga reacts with the NH 3 according to the following reaction [44] Ga (g) + NH 3 (g) GaN (s) H 2 (2.1) The estimated equilibrium constant K for the reaction and K = a GaN. P 3/2 H 2 (2.2) P Ga. P NH3 log 10 (K) = ( ) 10 4 T log 10 (T) (2.3) Where a GaN is the activity of GaN, P is the partial pressure of the reactants which are participating in the above reaction. Fig. 2.4 shows the variation in equilibrium partial pressures of Ga (P Ga ) with the corresponding input V/III ratio [44]. V/III ratio is the ratio of flow rates of input group-v (NH 3 ) and group-iii (Ga, Al & In precursors) sources. The input V/III ratio is varied by changing

43 Chapter 2. Experimental Techniques & Characterization 19 the partial pressure P NH3 and by keeping P o Ga at a constant value. The red color points indicate the corresponding equilibrium partial pressures of Ga (P Ga ) for our MOCVD experimental conditions at two different V/III ratios 260 and Figure 2.4: Variation in the equilibrium partial pressure of Ga (P Ga ) with the input V/III ratio. Copyright: Koukitu et al., JJAP, Vol.36, L1136, The Japan Society of Applied Physics. Fig. 2.5 shows the variation in equilibrium partial pressure of Ga (P Ga ) calculated from equation (2.2) with the V/III ratio for our MOCVD experimental conditions, which we have used in this dissertation. It is found that the equilibrium partial pressure of Ga (P Ga ) decreases with the increase in V/III ratio. The driving force for the deposition of GaN is given by equation (2.4) and is controlled by the partial pressure of NH 3 (P NH3 ) (Fig. 2.4). P GaN = P o Ga P Ga (2.4) Where P Ga is the equilibrium partial pressure of Ga inside the reactor chamber, P o Ga is the input partial pressure of Ga which is kept at a constant value. For

44 Chapter 2. Experimental Techniques & Characterization 20 Figure 2.5: Variation in the equilibrium partial pressure of Ga (P Ga ) with V/III ratio for our MOCVD experimental conditions. example in our case, we have kept the corresponding input flow rate of TMGa at 4.1 sccm for GaN growth, and P v Ga is the vapor pressure of Ga. Partial pressures of reactants are varied by changing the corresponding flow rates. We do keep the TMGa (which is in liquid form) bubbler at a particular temperature where TMGa is in equilibrium with its vapor. In the subsequent process the carrier gas H 2 is passed through the TMGa bubbler to evaporate liquid TMGa, before it is fed into the reactor chamber. The input flow rate of TMGa (or P o Ga ) is controlled by the flow rate of carrier gas H 2 which is passing through the bubbler. The vapor partial pressure of metal organics (TMGa, TMAl etc.) depends upon the temperature. The relation can be expressed as [45] P v III = 10 (a b T) mbar (2.5) The vapor pressure for group-iii precursors and their corresponding temperatures are shown in Table 2.3 [45]. The driving force decreases with the decrease in V/III ratio. Three kinds of modes of deposition is possible [44] from a consideration of Fig. 2.4.

45 Chapter 2. Experimental Techniques & Characterization 21 Precursor a b (K) Vapor pressure P v III (mbar) TMGa (0 o C) TMAl (17 o C) TMIn (17 o C) Table 2.3: Vapor pressure constants a and b for common metalorganic precursors P Ga > P o Ga : Etching This mode is achieved by decreasing the V/III ratio or by reducing the partial pressure of NH 3 (P NH3 ). During this mode GaN starts decomposing and the decomposition rate is higher than the incorporation rate of Ga and N adatoms into the growing crystal. GaN decomposes at above 800 o C at a pressure of 1 atm and at lower temperature in vacuum [46, 47]. It was obtained that from mass spectroscopy and thermogravimetric experiments, GaN decomposes into Ga and NH 3. There is huge scatter in the data reported for the activation energy barrier for GaN decomposition, which spans from 0.34 to 3.1 ev and depends strongly on conditions of the ambient (H 2 and/or N 2 ) in which the decomposition takes place [46, 47]. The unit processes involved in GaN decomposition and growth are explained in the subsequent section. P Ga > P v Ga : Droplet formation As discussed earlier in this context, the equilibrium vapor is the pressure where the liquid TMGa is in equilibrium with its own vapor. This mode occurs when the equilibrium partial pressure of Ga (P Ga ) inside the reactor chamber exceeds the vapor pressure of Ga (P v Ga ). During this mode the Ga atoms tend to condense on the growing surface and eventually form liquid metallic Ga droplets on the surface.

46 Chapter 2. Experimental Techniques & Characterization 22 P Ga < P o Ga & P Ga < P v Ga : Growth This condition defines the growth window for GaN deposition. During this mode of growth, the incorporation rate of Ga and N atoms into the growing bulk crystal is higher than the decomposition rate of GaN. However, the incorporation rate of adatoms in to the growing crystal is controlled by several factors such as the life time of adatoms on the growing surface and the diffusion lengths of adatoms which in turn controlled by the corresponding surface diffusion barriers at typical MOCVD growth temperatures. The driving force for this growth window is increases with the increase in partial pressure of NH 3 (P NH3 ), which is consistent with the experimental MOCVD trends. The above results indicate that for GaN growth to occur the equilibrium partial pressure of Ga (P Ga ) is always has to be lower than the vapor pressure P v Ga and the input partial pressure P o Ga. GaN decomposition has been studied extensively and several mechanisms were proposed to explain it. The following reactions are reported for GaN decomposition [48, 49]. 2GaN (s) 2Ga (g) + N 2 (g) (decomposition) (2.6) 2GaN (s) 2Ga (l) + N 2 (g) 2Ga (g) + N 2 (g) (desorption) (2.7) 2GaN (s) GaN (g) or [GaN] x (g) (sublimation) (2.8) It was reported that H 2 could assist the GaN decomposition by reformation of NH 3 via [47]. 2GaN (s) + 3H 2 (g) 2Ga (l) + 2NH 3 (g) (2.9) Table 2.4 shows the activation energy barriers measured for the above reactions under the ambient of H 2 [47 49]. The activation energies are measured by several techniques such as thermogravimetry, mass spectrometry and reflection high energy electron diffraction (RHEED).

47 Chapter 2. Experimental Techniques & Characterization 23 Process E A ev Technique GaN decomposition 3.1 Thermogravimetry, Mass spectrometry Ga desorption from GaN RHEED, Growth rate Vs Growth temperature N desorption from GaN 6.1 RHEED Table 2.4: Activation energy E A for decomposition of GaN, and desorption of Ga and N atoms from GaN. The other unit step which takes place in GaN decomposition and growth is surface diffusion. The factor surface diffusion is critical and it controls the surface quality of the growing surface. The mode of growth is classified into two types based on surface diffusion lengths of the adatoms on the growing surface. The mode step-flow (2D mode) occurs when the diffusion length of adatoms is longer than the terrace width or step width involved during growth. In contrast, island mode (3D mode) occurs when the diffusion length is smaller than the corresponding terrace width [50]. The diffusion lengths are controlled by the corresponding surface diffusion barriers at typical process temperatures. The diffusion length can be estimated by the following equation [50] λ s = D s τ s = ( ( a 2 ν = exp E ) 1 2 sd )τ s kt (2.10) Where, D s is the diffusion coefficient, ν = is vibrational frequency of adatoms parallel to the surface, a is mean distance between the adsorption sites, E sd corresponding surface diffusion activation energy and τ s can be classified into two cases, which are life time before desorption (τ D ), which depends on the desorption rate and the life time before lattice incorporation (τ I ), which depends on growth rate. The mean life time of residence of adatoms (τ D ) on the growing surface before being re-evaporated and is given by τ D = 1 ( ) Edes exp ν kt (2.11) Where ν is the vibrational frequency of adatoms normal to the surface and E des is

48 Chapter 2. Experimental Techniques & Characterization 24 the desorption activation energy of an adatom from the growing surface. Assuming ν = ν = ν =, we have the following expression for surface diffusion length [50] ( ) Edes E sd λ s,ga = a exp 2kT (2.12) Koleske et al. [51] measured λ s,ga based on τ D using equation (2.12) and is shown in Fig. 2.6 as a solid line, which indicates λ s,ga decreases slightly as the temperature increases. This is due to the increase in desorption rate, which results in reduced τ D. But, at typical MOCVD growth pressures Ga desorption appears to be suppressed and eventually yields longer τ D. The limiting life time in such cases is the incorporation life time (τ I ) into the growing lattice. The value of τ I depends inversely on growth rate, as slower the growth rate the larger τ I. Figure 2.6: Measured surface diffusion length λ s,ga of Ga adatoms at different temperatures. The solid line corresponds to λ s,ga based on τ D and dashed lines correspond to λ s,ga based on τ I. Koleske et al., JAP, 84, 1998, Reprinted with permission. Fig. 2.6 shows the dependence of τ I on the growth rate and is λ s,ga plotted as dashed lines for different growth rates ranging from 3 to 100 nm/min. τ I was measured by the thickness per monolayer (actually bilayer, which is nm and represents one half of the lattice parameter along c-direction) by the growth rate. It was also shown that as growth rate decreases, both and increases, which eventually

49 Chapter 2. Experimental Techniques & Characterization 25 Adatom Ga-polar GaN Ga-polar GaN N-polar GaN N-polar GaN N-terminated N-terminated (ev) (ev) (ev) (ev) Ga N Table 2.5: Surface diffusion barriers for Ga & N adatoms on Ga-polar and N-polar GaN surfaces. yields a more ordered lattice because the number of adatoms that incorporate is increased. It was found that the diffusion length λ s,n for N adatoms is much smaller than Ga adatoms and is due to the high surface diffusion barrier for N adatoms when compared to Ga adatoms [17]. At the same time the other possible reason for lower λ s,n values, is due to high vapor pressure of N 2, once N migrates to next to an adjacent N, N 2 forms and desorbs. From total energy density functional theory (DFT) it is predicted that the surface diffusion barriers for Ga & N adatoms on (0001) and (000 1) surfaces are not same [17]. Table 2.5 shows the estimated surface diffusion barriers for Ga & N adatoms at typical growth temperatures. From Table 2.4 it can be suggested that the estimated diffusion lengths for Ga & N adatoms on N-polar GaN surface must be higher than on Ga-polar GaN surface. As we described in Sec of Chapter 1, the surface morphology of N-polar GaN grown under identical conditions is rough and is associated with inversion domains (IDs). In such cases the estimation of corresponding surface diffusion lengths of Ga & N adatoms on both domains (IDs) is complicated, which eventually makes difficult to understand the behavior of adatoms on the growth surface and hence the growth mechanism of N-polar GaN. 2.2 Characterization This section describes some of the standard characterization tools that have been used to analyze our samples. The primary characterization tools employed in this

50 Chapter 2. Experimental Techniques & Characterization 26 study are high resolution x-ray diffractometry, x-ray photoelectron spectroscopy, atomic force microscopy, scanning electron microscopy, differential contrast interference optical microscopy (Nomarski) and the in situ reflectivity and stress measurement tool. In some cases transmission electron microscopy has been used to examine the polarity of our samples. All these are available at the Centre for Nano Science & Engineering (CeNSE), IISc Bangalore. The transmission electron microscopy studies were done at the Advanced Facility for Microscopy and Micro analysis (AFMM), IISc Bangalore and Defense Metallurgical Research Laboratory, Hyderabad X-ray photoelectron spectroscopy (XPS) This technique has been extensively used to study the chemical nature of sapphire wafers nitrided at different temperatures. This ex situ tool was supplied by Kratos Analytical AXIS Ultra DLD, Manchester. It is a surface sensitive technique, which analyzes the kinetic energy (KE) of photoelectrons which are emitted from the sample surface via photoemission process. The spectrometer measures the binding energy (BE) of the ejected electrons from the measured KE. The BE of the photoelectrons which are emitted from the sample surface is given by the following relation. BE = hν KE φ (2.13) Where φ represents the combined electron spectrometer and sample work functions and is an instrument dependent factor normally derived for each instrument as part of a calibration procedure. The technique derives its chemical sensitivity from the fact that nearest neighbor atoms will have a direct effect upon the binding energy of the core level electrons. Therefore any change in the chemical environment such as oxidation state will lead to a modification of the KE. The instrument has the following key components: X-ray source, electron transfer lens, electron energy analyzer, and detection system. All of these components are contained within an ultra-high vacuum envelope (10 9 Torr). The spectrometer

51 Chapter 2. Experimental Techniques & Characterization 27 is equipped with monochromatic x-rays as the primary source. The x-ray gun is used in combination with a focusing monochromator due to which only the Al Kα component is diffracted from the quartz crystal. The natural line width of this component is < 0.26 ev. A charge balance option is available for insulators. An Ar ion beam was used to clean the sample surface from contamination and/or for depth profile information. Charge corrections were done with respect to C 1s (Carbon) peak located at ev. Elemental quantification of all the survey spectra and high resolution scans were carried out by CASA XPS software High resolution x-ray diffractometer HRXRD has been used to examine the crystal mosaicity of our samples. The epitaxial film is assumed to consist of single crystallites called mosaic blocks with misorientations with respect to each other [52, 53]. The out-of-plane rotation of the blocks perpendicular to the surface normal is the mosaic tilt, and the inplane rotation around the surface normal is the mosaic twist (Fig. 2.7). The Figure 2.7: Mosaic blocks in a crystal showing tilt and twist with respect to each other. Tilt corresponds to out-of-plane rotation whereas twist corresponds to in-plane rotation of mosaic blocks. Srikant et al., JAP, 82, 4286, Reprinted with permission.

52 Chapter 2. Experimental Techniques & Characterization 28 average absolute values of tilt and twist are directly related to the full width at half maximum (FWHM) of the corresponding distributions of crystallographic orientations [52, 53]. The rocking curve measurement (ω-scan) method is the most frequently employed method to analyze the mosaicity of epitaxial layered materials. In this geometry, the detector remains stationary and the sample is rotated about ω-axis (Fig. 2.8). Figure 2.8: X-ray diffractometer with different rotational angles (ω, ψ, φ) in relation to the sample reference frame. ω-scan (000l) reflections are used to measure tilt of mosaic blocks [52, 53]. Tilt is sensitive to screw and mixed dislocations but not to edge dislocations because edge dislocations do not distort (000l) planes as their burger vectors (1/3 < >) lie within those planes. Twist is caused by edge and mixed dislocations and is usually measured with ω-scans of off-axis reflections (h or k 0). Off-axis reflections occurring at higher ψ values are used, so that FWHM is dominated by in-plane twist. Thus, lower FWHM values for (000l) and off-axis reflections indicate that epitaxial films of better crystalline quality with low mosaicity. The instrument that has been used for XRC analysis was supplied by RIGAKU SmartLab, Japan. This tool was equipped with the 2 bounce and 4 bounce Ge (220) incident beam monochromators. The Cu (Kα1: with λ = nm and Kα2: with λ = nm) x-ray source is collimated and monochromated by a 4 bounce Ge (220) monochromator to achieve high angular resolution.

53 Chapter 2. Experimental Techniques & Characterization Atomic force microscopy (AFM) The AFM used in this dissertation is Dimension Icon ScanAsyst Bruker. The AFM has been used to study the surface features of LT GaN NLs and the surface roughness evolution of HT GaN epitaxial layers.afm can be operated in two modes such as contact mode and non-contact mode. Tapping mode falls in somewhere between the contact and non-contact mode. All AFM measurements were done in this dissertation using tapping mode technique. The classification of AFM working modes is described in Fig Figure 2.9: Typical inter atomic force curve between the cantilever tip of AFM and sample surface The contact mode falls in the repulsive interaction region whereas the non-contact mode falls in the attractive Van der Waals interaction region. The tapping mode falls somewhere in between the two interaction regions. In the contact mode the deflection of cantilever is kept constant. In contrast, the tip is oscillated at it s resonant frequency in non-contact mode and the amplitude of the oscillation is kept constant. Contact mode imaging is heavily influenced by frictional and adhesive forces, and can damage samples and distort image data. Non-contact imaging generally provides low resolution and can also be hampered by the contaminant (e.g., water) layer which can interfere with oscillation. Tapping mode imaging takes the positive features of both these modes and overcomes problems associated with friction, adhesion, electrostatic forces, and other

54 Chapter 2. Experimental Techniques & Characterization 30 difficulties that plague conventional AFM scanning methods, by alternately placing the tip in contact with the surface to provide high resolution and then lifting the tip off the surface to avoid dragging the tip across the surface. Tapping mode imaging is implemented in ambient air by oscillating the cantilever assembly at or near the cantilever s resonant frequency using a piezoelectric crystal. The piezo motion causes the cantilever to oscillate with a high amplitude( typically greater than 20nm) when the tip is not in contact with the surface. The oscillating tip is then moved toward the surface until it begins to lightly touch, or tap the surface. During scanning, the vertically oscillating tip alternately contacts the surface and lifts off, generally at a frequency of 50,000 to 500,000 cycles per second. As the oscillating cantilever begins to intermittently contact the surface, the cantilever oscillation is necessarily reduced due to energy loss caused by the tip contacting the surface. The reduction in oscillation amplitude is used to identify and measure surface features. When the oscillating cantilever approaches the surface the amplitude of oscillation decreases. This is due to the additional restoring force works on the cantilever and this can be seen as an increase in the spring constant of the cantilever. This drop in amplitude can be used as the feedback parameter for AFM imaging, just like the cantilever deflection in contact mode. Tapping mode inherently prevents the tip from sticking to the surface and causing damage during scanning. Unlike contact and non-contact modes, when the tip contacts the surface, it has sufficient oscillation amplitude to overcome the tipsample adhesion forces. Also, the surface material is not pulled sideways by shear forces since the applied force is always vertical Differential interference contrast (DIC) light microscopy DIC light microscopy (Nomarski) is a technique which produces impressive 3Dlike images of unstained specimens. The shadowing effects of the technique are remarkable. DIC light microscopy was used in this dissertation to provide an initial idea about surface quality of the HT GaN epitaxial layers. Nomarski microscopy

55 Chapter 2. Experimental Techniques & Characterization 31 utilizes a system of dual beam interference optics that transforms local gradients in optical path length in a specimen into regions of contrast in an image. The working principle of the Nomraski light microscope is described in Fig ˆ Light passes through a standard polarizer before entering the condenser, producing plane-polarized light ˆ This light enters a Wollaston prism located in the front focal plane of the condenser. The prism interacts with the polarized light to produce two separate wavefronts which are parallel and polarized perpendicularly to each other. These are termed the ordinary (O) and extraordinary (E) rays. Furthermore, these two wavefronts are separated by a very small difference (less than the resolution of the system) generally ranges from µm and the separation is depends upon the prism used ˆ The two wavefronts pass through the specimen, phase of one beam may be differentially shifted with respect to the other if there is a local gradient in optical path length. ˆ The light now enters a second Wollaston prism set-up which recombines the wavefronts. If there has been a phase shift between the two rays as they pass through areas of different refractive index then elliptically polarised light is the result. DIC optics are designed to convert these phase differences between the two beams of light into amplitude differences which can be visualized by the human eye ˆ Finally the light enters a second polarizing filter, termed an analyzer. The initial polarizer and this analyzer form crossed polars. The analyzer will permit the passage of some of the elliptically polarized light to form the final image. All the remaining light will be blocked by the analyzer.

56 Chapter 2. Experimental Techniques & Characterization 32 Figure 2.10: Typical optical lens diagram of differential interference contrast (DIC) light microscopy. A. Lasslett, Microscopy Division, Olympus UK Ltd, Southall, Middlesex, UK. Reprinted with permission Scanning electron microscopy (SEM) The SEM used in this study was Carl Zeiss Ultra 55 FESEM. This tool has been used to study the morphology of our samples at different stages of their process. The probe current range: 15 to 20 na, aperture size range: 7.5 to 120 µm and accelerating voltage range: 5 to 30 kv Transmission electron microscopy (TEM) TEM was used in this dissertation to examine the polarity of the grown epitaxial layers. Convergent beam electron diffraction (CBED) technique was used in combination with java electron microscope simulator (JEMS) simulated patterns to examine the polarity of our samples at different thicknesses. A 300 kv Technai

57 Chapter 2. Experimental Techniques & Characterization 33 TM G2 F30 S-TWIN TEM with FEG at Advanced Facility for Microscopy & Microanalysis (AFMM) IISc and a FEI Tecnai G2-20T 200KV TEM at Defense Metallurgical Research Laboratory (DMRL) Hyderabad were used to examine GaN layers in this dissertation. Cross sectional TEM foils were made using the regular sandwich technique, in which two pieces of the sample are cut and glued together keeping the film sides of the two pieces facing each other. The sandwich assembly is inserted and stuck inside a slotted rod and this complete assembly is pushed and glued inside a hollow tube with 3 mm outer diameter. Thin slices from this combination is cut by using Buehler ISOMET low speed saw and thinned down to 100 µm using MULTI- PREPTM equipment manufactured by Allied Instruments. TEM foils were further thinned down to electron transparency by using conventional ion-beam thinning technique using Gatan PIPS (Precision Ion Polishing System) with Cold Stage In situ reflectivity and stress monitor analysis tool The in situ tool that was used in this study is the k-space Associates multibeam optical stress sensor (MOSS). This technique has been used to monitor the growth behavior of GaN films on sapphire. The primary information that can be extracted is the real time film thickness, real time stress and growth rate. This in situ capability enables the monitoring of film stress and thickness as it develops. In the MOSS technique (Fig. 2.11a), multiple parallel laser beams illuminate the surface simultaneously and the beam positions are measured with a CCD detector. The measurement of film thickness is based on interference between the light beams reflected from the top surface of the film and from the film/substrate interface (Fig. 2.11b). If the film is not too absorbing, interference between the reflected beams results in a modulation of the reflected intensity. The interference of beams depends upon the path difference of the rays which in turn depends on growing film thickness. The period of the intensity oscillations is used to determine the growth rate of the film. In addition, the amplitude of the intensity oscillations depends on the reflectivity of the interfaces, and therefore also a probe of surface

58 Chapter 2. Experimental Techniques & Characterization 34 morphology during growth. Fig. 2.11c shows an example of GaN two-step growth on sapphire. Figure 2.11: A schematic diagram of k-space MOSS set up (a) multiple parallel laser beams fall on the substrate surface and will be reflected and detected by CCD camera, (b) interference effects for the rays which are reflected from the surface and the film/substrate interface, and (c) a sample reflectivity trace for GaN epitaxy on sapphire, which is recorded by in-situ k-space MOSS tool attached to MOCVD reactor. The growth rate is measured by λ 2η = Period of one oscillation (2.14)

59 Chapter 2. Experimental Techniques & Characterization 35 Where, λ is wavelength of light used = 660 nm. η is refractive index of the growing film and is 2.3 for GaN. The in situ tool can also monitor the real time stress evolution in films during their growth. MOSS stress monitoring is based on measuring the curvature induced in the substrate by the stressed film (Fig. 2.12). Figure 2.12: The radius of curvature of the stressed film effects the spacing between laser beams which are reflecting from the surface of the film. The relationship between the film stress, σ, and the radius of curvature, R, is given by the following equation developed originally by Stoney [54] σ = M sh 2 s 6h f R (2.15) Where h f is the film thickness, M s is the biaxial modulus of the substrate, h s is the substrate thickness, and R is radius of curvature. For flat films R is infinite, so that parallel beams of light will be reflected from the surface and the beam separation will be same as the incident beam separation. If a surface is curved, parallel beams of light that strike it at different positions will be reflected at different angles (Fig. 2.12). The amount of deflection of the light beam is related to the curvature by 1 R = δd d cosα / 2L (2.16) Where δd is the measured change in the spacing of the beams, d is the initial spacing, L is the distance from the sample to the detector and α is the angle between the incident beams and the surface normal.

60 Chapter 2. Experimental Techniques & Characterization 36 This tool is used extensively to understand the growth behavior of GaN epitaxial layers on sapphire (0001) wafers.

61 Chapter 3 Sapphire Pre-treatment & Microstructural Evolution of LT GaN It is well known that the high energy interface, lattice mismatch and CTE mismatch between HT GaN and sapphire prevents the HT GaN in growing laterally and promotes 3-dimiensional growth of GaN with poor crystalline quality. The structural quality of HT GaN layers is dramatically improved, when a thin LT GaN/LT AlN NL is deposited on sapphire prior to the growth of HT GaN. The function of the subsequently grown LT NL could be described as [55] ˆ The supply of nucleation sites of low orientational fluctuation and ˆ The promotion of lateral growth of HT GaN This chapter addresses the following two things. First one is to understand the structural transformation of sapphire surface at three different nitridation temperatures T N = 530, 800 and 1100 o C. Second one is to understand the relation between the transformed nitrided layers on sapphire surface and the microstructural evolution of the subsequently grown LT GaN NLs. 37

62 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN Background Nitridation is the foundation of N-polar nitride materials deposited on sapphire wafers. It is well known that GaN growth on non-nitrided sapphire (0001) surface results in Ga-polar GaN while that grown on sapphire (0001) surface nitrided in the temperature range above 950 o C results in N-polar GaN [19, 56, 57]. The nitridation of sapphire surface prior to the growth of LT GaN NL is known to reduce the chemical dissimilarity and lattice mismatch between GaN and sapphire [9 12]. Literature suggests that the nitrided layer is complex unknown AlO x N 1 x [9], AlN and/or AlN in O atom rich environment [10]. Sapphire surface gets damaged if it is exposed to NH 3 and H 2 ambient for longer times during nitridation at higher temperatures [9, 19]. The possible reason behind surface damage is due to the strong chemical reaction between NH 3 and sapphire [19]. Sun et al., showed that higher nitridation temperatures result in protrusions on sapphire surface during nitridation, which in turn affects the surface morphology of subsequently grown N-polar GaN epitaxial layers [19]. The chemical nature and structure of these protrusions are unknown. The structural transformation of sapphire surface critically depends upon the process conditions such as flow rates of NH 3, H 2 and/or N 2, nitridation temperature (T N ) and process time. It is well known that NH 3 decomposes at above 500 o C and the complete decomposition of NH 3 is assumed to occur easily at 1100 o C [58]. The mechanism of nitridation involves the counter diffusion of N (inward diffusion) and O (outward diffusion) atoms from the sapphire surface [58]. Diffusion data for N and O atoms in sapphire lattice is well documented [58, 59]. It has been reported that the as-grown LT GaN NL on sapphire contains cubic (zincblende) phase in addition to stable wurtzite phase [59, 60]. The usual temperature range for LT GaN growth falls in between 500 to 600 o C [59, 60]. It has been reported well in the literature that the morphology, distribution, polarity and orientation of LT GaN critically depends upon the sapphire pre-treatment called sapphire nitridation in this dissertation. The surface morphology of the as deposited LT GaN on non-nitrided sapphire wafers is very rough with 3D discrete

63 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 39 islands, in contrast the surface morphology is relatively smooth with complete coverage for the LT GaN deposited on nitrided sapphire wafers [11, 61]. It was also found that the as grown LT GaN deposited on nitrided sapphire results in Ga-polar LT GaN and the polarity is transformed to N-polar in the subsequent annealing steps [57], the polarity of LT GaN is determined by potassium hydroxide (KOH) etching methods [57]. The annealing of LT GaN NLs take place at typical GaN epitaxial layer growth temperatures 1000 to 1080 o C under the ambient of NH 3, H 2 and/or N 2 and it was found that the structural quality of subsequently grown HT GaN epitaxial layer is a strong function of annealing conditions of LT GaN [62 66]. N-polar GaN is more sensitive to the ambient present during its annealing process and it was found that N-polar GaN etches at faster rate than Ga-polar GaN during annealing period under the ambient of NH 3 and H 2 at typical MOCVD annealing temperatures [67]. The crystalline quality and electrical properties of HT GaN have also been studied as a function of growth rate/thickness of LT GaN NL [63, 68]. The distribution and density of LT GaN nuclei for the subsequent HT GaN growth can be controlled by the growth parameters such as V/III ratio and reactor pressure during the growth of LT GaN [62, 69]. This chapter therefore explores the modification of the non-miscut sapphire (0001) wafers at three nitridation temperatures T N = 530, 800 and 1100 o C to provide a link between GaN two-step growth on non-nitrided sapphire and the conventional high temperature nitridation. 3.2 Experimental Experiments were carried out in an Aixtron 200/4 RF-S horizontal flow MOCVD reactor with removable quartz ware. The in situ thermal treatment of wafers were carried out at 1100 o C under the flow of purified H 2 for 10 min before nitridation. The nitridation step was performed at three different temperatures T N = 530, 800 and 1100 o C and 200 mbar reactor pressure. Wafers were nitrided under the ambient of NH 3 and H 2 for 1 min. NH 3 flow rate was kept constant at 1500 sccm throughout the process. The nitrided wafers were taken out from the MOCVD

64 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 40 reactor chamber for ex situ XPS characterization. After sapphire nitridation, LT GaN was grown to a 60 nm thick layer to understand the effect of nitridation on microstructural evolution of NL. The thickness calibration was done by the in situ optical tool as discussed in Sec of Chapter 2. The effect of LTGaN thickness on HT GaN growth is described in Chapter 4. LT GaN NL was deposited on nitrided wafers at 530 o C for 8 min (to deposit 60 nm thick layer), 200 mbar, at a V/III ratio of 2535, TMGa flow rate of 0.6 sccm and H 2 flow rate of 6500 sccm. The wafers were then heated up from the LT GaN NL growth temperature to the LT GaN NL annealing condition in 5 minutes. The NLs were annealed thermally for 4 minutes at 1000 o C and a reactor pressure of 400 mbar in a mixture of 2000 sccm of NH 3 and 4700 sccm of H 2. During the annealing process of NL, it is possible that the nitrided sapphire surface regions which are not covered by LT GaN NLs, are exposed to NH 3 and nitrided further. We have investigated these changes by subjecting the nitrided sapphire surface to identical ramp up and hold cycles as they would encounter during the deposition of LT GaN NL and its subsequent process. Therefore, after the nitridation step, the temperature is ramped to the NL deposition temperature of 530 o C and held at 530 o C for 8 min followed by temperature ramp up to NL annealing temperature of 1000 o C in 5 minutes. Once the temperature reaches the annealing condition, it is held for 4 min. The main tool used to characterize nitrided layers was XPS as described in Sec of Chapter 2. SEM & AFM are used for morphological studies, and HRXRD is for structural analysis of LT GaN NL. 3.3 Results This section starts with a few results on the effect of ex situ and in situ cleaning on surface morphology of sapphire (0001) wafers followed by subsequent nitridation and LT GaN NL deposition steps.

65 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN In situ thermal treatment of sapphire wafers As described in Sec of Chapter 2, the diced sapphire wafers undergo ex situ cleaning with acetone, isopropanol and de-ionized water and are dried with a stream of N 2 gas to dry solvents from the surface until no droplets are found. In many cases, the surface may contain microscopic droplets which might leave organic residues on the surface upon drying with N 2. Fig. 3.1a shows the surface of sapphire which has been subjected to the ex situ cleaning process with acetone, isopropanol, de-ionized water and dried with N 2 gas prior to the in-situ thermal cleaning in the MOCVD reactor. Figure 3.1: Surface morphology of sapphire (0001) wafer which is cleaned ex situ, (a) microscopic organic remnants on sapphire surface after drying with N 2 gas, and (b) morphology of sapphire surface after in-situ treatment with H 2 at 1100 o C for 10 min in MOCVD reactor. Fig. (c) & (d) are magnified views of (b). Fig. 3.1b shows the surface morphology of sapphire after in situ treatment under the flow of purified H 2 at 1100 o C. The etched morphology is due to the reaction

66 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 42 between H 2 and the organic residues which are left after ex situ cleaning [70]. Fig. 3.1c & d are magnified images of in situ treated sapphire surface. To resolve this issue, we have optimized the ex situ cleaning procedure and it is found that the cleaning of wafers many times with de-ionized water after acetone and isopropanol helps to remove all the organic residues from the surface. Fig. 3.2 shows the surface of sapphire which is cleaned ex situ with acetone, isopropanol and multiple times with de-ionized water followed by an in situ H 2 treatment at 1100 o C in the MOCVD reactor. Figure 3.2: Surface micrographs of sapphire wafer which have undergone an optimized ex situ cleaning procedure, (a) & (b) are SEM micrographs at two different magnifications, and (c) & (d) are corresponding AFM morphologies with sub-nanometre RMS roughness 0.35 nm. AFM image shows (Fig. 3.2c & d) atomic steps on the sapphire surface with rms surface roughness 0.35 nm after in situ treatment with H 2 at 1100 o C in the MOCVD reactor.

67 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN Nitridation of in situ treated sapphire wafers As nitrided sapphire wafers Fig. 3.3 shows the high resolution x-ray N 1s photoelectron spectrum from sapphire wafers nitrided at three different temperatures T N = 530, 800 and 1100 o C. Figure 3.3: The N 1s x-ray photoelectron peak from sapphire wafers nitrided at T N = 530, 800 and 1100 o C The distinct N 1s binding energy photoelectron peak indicates that there is incorporation of N atoms into the sapphire lattice during nitridation. Three effects of nitridation temperature are discernible: the change in intensity of the N1s peak, the shift in peak positions and a peak broadening is observed with the decrease in nitridation temperature. The variation in N1s peak intensity indicates that the increase in incorporation of N atoms into the sapphire surface with the nitridation temperature. The shift in peak position and peak broadening indicates that the environment of the N atoms in the modified sapphire undergoes changes with nitridation temperature. Since surface modification is expected to occur by inward diffusion of N into sapphire to replace O atoms. Fig. 3.4 shows the normalized

68 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 44 intensities of N 1s and O 1s peaks and are plotted against nitridation temperatures. The normalization is done with respect to the Al 2p peak intensity. Increase in nitridation temperature results in greater N/O ratios in the modified sapphire surface. Figure 3.4: Normalized intensities of (a) O 1s and (b) N 1s peaks from sapphire wafers nitrided at T N = 530, 800, and 1100 o C. Normalization was done with respect to Al 2p peak The normalized O 1s intensity is compared with the O 1s intensity from nonnitrided sapphire surface as indicated by blue color line in Fig. 3.4b. It is found that the non-nitrided sapphire surface has high O atom content when compared to nitrided sapphire surfaces. Peak broadening and shift in peak positions have been explored by deconvoluting the N1s peaks. Fig. 3.5 shows the deconvoluted spectra of N 1s photoelectron peaks at different nitridation temperatures. Systematic changes in both the intensity and peak positions are observed with the change in nitridation temperature. A strong peak at ev is present for T N =800 and 1100 o C (Fig. 3.5b & c). This coincides with that of the binding energy of Al-N bond in the bulk AlN of ev [58, 71] and indicates the structural transformation of sapphire surface to AlN at T N 800 o C. There are additional peaks at ev and ev. The intensity of the former peak decreases and the latter increases with the change in nitridation temperature from T N = 800 to 1100 o C. The peaks at 395 ev correspond to sub-stoichiometric AlN x<1 with Al-Al bonds [72]. The peak at 398 ev has been attributed to incomplete

69 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 45 Figure 3.5: De-convoluted spectra of N 1s photoelectron peaks from sapphire wafers nitrided at (a) T N = 530 o C, (b) T N = 800 o C, and (c) T N = 1100 o C. T N AlN (Al-N bond) AlON (Spinel) AlO x N 1 x Sub-stoichiometric AlN (Al-Al bond) ( o C) I N1s (a. u.) I N1s (a. u.) I N1s (a. u.) I N1s (a. u.) Table 3.1: Normalized intensities of N 1s deconvoluted peaks from various possible nitrided layer structures from T N = 530, 800 & 1100 o C. Normalization has been done with respect to Al 2p peak. substitution of O by N such that tetrahedral A-N-O bonds are present [58, 73]. With decrease in nitridation temperature, peak positions shift to higher energy levels. At T N = 530 o C, the binding energy of ev indicates that N is in O atom rich environment [59, 74 77] and the peak at ev in Fig. 3.5a arises from AlON cubic spinel structure [76]. Table. 3.1 summarizes these results in the form of normalized intensities of N 1s deconvoluted peaks from various possible nitrided layer structures from T N = 530 o C to 1100 o C. Normalization has been done with respect to Al 2p peak.

70 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN Annealed nitrided wafers The XPS plots shown in Fig. 3.6 suggest that the as-nitrided layers undergo further nitridation during ramp and anneal conditions. For T N = 530 o C, the N 1s peak just after ramp up to the annealing temperature shifts toward lower energy levels indicating a transitional bonding environment with decreasing amount of O atom content, and tetrahedral Al-N bond formation (Fig. 3.6a). The peak corresponds to 405 ev (AlON cubic spinel) is no longer observed after ramp up and anneal step. It is found that after 4 min anneal step the dominant peak located at and is corresponds to the Al-N bond energy levels in O atom rich environment (Fig. 3.6c). In contrast, no significant changes are observed for the samples that have been nitrided at T N = 1100 o C as shown in Fig. 3.6b & 3.6d. The dominant peak is always found to be at ev that corresponds to tetrahedral Al-N bonds in bulk AlN [58, 71]. The minor peak located at 398 ev in as-nitrided sample is found to remain even after ramp up step (Fig. 3.6b), but has shifted to 397 ev after 4 min annealing step (Fig. 3.6d). Fig. 3.6e shows the compiled binding energy scale for the N 1s photoelectron peak with their corresponding chemical bond environments from the literature [58, 59, 71 77] LT GaN nucleation layer Morphological evolution Fig. 3.7 shows the surface morphology of LT GaN NL after the three different stages: deposition, ramp up to LT GaN annealing temperature of 1000 o C in 5 minutes (0 min annealing) and hold at 1000 o C for 4 minutes (4 min annealing). Fig. 3.8 shows the corresponding AFM morphologies. Fig. 3.9 shows the RMS surface roughness values of the NLs derived from AFM data. The most striking aspect of the LT GaN NLs evolution summarized in Fig. 3.7 to 3.9 is the significant difference in surface morphology and AFM roughness between

71 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 47 Figure 3.6: N 1s x-ray photoelectron peaks from nitrided sapphire wafers at different stages of their process: (a) & (c) are for T N = 530 o C, after ramp up and 4 min annealing, and (b) & (d) are for T N = 1100 o C, after ramp up and 4 min annealing at 1000 o C. The compilation of N 1s binding energy scale with their corresponding chemical bonding environments are shown in (e) [58, 59, 71 77] NLs deposited on surfaces nitrided at T N = 530 o C versus those deposited on surfaces nitrided at T N 800 o C. A levelling effect of the ramp up and anneal at higher temperatures can be seen from Fig. 3.7 & 3.9. On sapphire nitrided at T N = 530 o C, Fig. 3.7a, the nuclei appears to be arranged along the atomic steps on sapphire and evolve in this arrangement even during the ramping step.

72 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 48 Figure 3.7: SEM surface morphologies of LT GaN NLs: (a), (b) & (c) are for the NL at different stages of their processing on sapphire wafers nitrided at T N = 530 o C, (d), (e) & (f) are for the NL processed on sapphire wafers nitride at T N = 800 o C, (g), (h) & (i) are for the NL on sapphire wafers nitrided at T N = 1100 o C.

73 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 49 Figure 3.8: AFM surface morphologies of LT GaN NLs: (a), (b) & (c) are for the NL at different stages of their processing on sapphire wafers nitrided at T N = 530 o C, (d), (e) & (f) are for the NL processed on sapphire wafers nitride at T N = 800 o C, (g), (h) & (i) are for the NL on sapphire wafers nitrided at T N = 1100 o C.

74 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 50 Figure 3.9: AFM surface roughness evolution of LT GaN NLs deposited on sapphire wafers nitrided at T N = 530, 800 and 1100 o C at different stages of their processing. However, further exposure to NH 3 at higher temperatures 1000 o C during the hold step annihilates these preferential orientations of the islands. In the case of the samples nitrided at T N = 800 and 1100 o C the ramp up significantly smoothens the LT GaN surface. However annealing appears to lead to an element of preferential sublimation leaving behind features that appear in light contrast Structural evolution The effect of nitridation on the crystallography of the LT GaN NLs is revealed through the HRXRD scans of Fig to Fig Shows (0002) HRXRD ω-scan profiles of LT GaN NLs at different stages of their processing. It is clearly seen that the NL deposited on surfaces nitrided at T N 800 o C show strong (0002) peaks as compared to that deposited on the surface nitrided at T N = 530 o C. The LT GaN processed on sapphire nitrided at T N 800 o C is found to be wurtzite LT GaN with a strong [0002] orientation as indicated by six {10 11} asymmetric peaks shown in Fig. 3.11c and d. The sample is tilted to o about ψ-axis and is rotated 360 o about φ-axis to get diffraction from six {10 11} planes. NLs at the end of their 4 min annealing step have significantly reduced in thickness and it is

75 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 51 difficult to pick up an XRD signal from them. In contrast, the NLs grown on sapphire nitrided at T N = 530 o C does not show this strong [0002] orientation in the as-grown conditions as indicted by the φ- scan in Fig. 3.11a. This is due to the presence of zincblende phase. Fig shows the phi scan measurements for cubic LT GaN {220} peaks. Six {220} peaks which are separated by 60 o angle indicates the presence of zincblende phase in the as grown LT GaN. The sample is tilted to o about the ψ-axis and the detector is fixed at 2θ of o to get the diffraction from {220} planes. However {10 11} peaks appear after the 4 min annealing step at 1000 o C (Fig. 3.11b), indicating that decomposition and reformation of GaN during this process results in transformation of LT GaN from zincblende phase to wurtzite phase with [0002] oriented crystallography. Figure 3.10: High resolution x-ray (0002) ω-scan profiles for NL at different stages of their processing: (a) is for as grown LT GaN NL on sapphire surfaces nitrided at T N = 530, 800 and 1100 o C, (b) is for NL after the 0 min annealing conditions and (c) is for NL after 4 min annealing. The results indicate that HT nitridation of sapphire yields LT GaN NL with strong [0002] orientation. In contrast LT nitridation yields LT GaN NL with the zincblende in the as grown step and transformed to wurtzite phase with [0002] orientation in the subsequent anneal stages.

76 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 52 Figure 3.11: High resolution x-ray {10 11} diffraction φ-scan for NLs: (a), (b) are for NLs processed on sapphire wafers nitrided at T N = 530 o C, (c), (d) are for NLs processed on sapphire wafers nitrided at T N = 1100 o C. Figure 3.12: High resolution x-ray diffraction φ-scan measurements for zincblende {022} peak of as-grown LT GaN on sapphire nitrided at T N = 530 o C.

77 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN Discussion Low temperature nitridation (T N = 530 o C) The as nitrided surface at low nitridation temperatures shows a relatively high O atom content with a strong N1s peak at 401 ev accompanied by less intense peaks at 405 ev and 397 ev as shown in Fig. 3.5a. These peaks have been identified with various levels of O in AlN [58, 59, 71 77] (Fig. 3.6e). It appears plausible to state that the surface is dominated by Al 2 O 3 based structural motifs on nitridation at T N = 530 o C accompanied by small amounts of AlON cubic spinel associated with the peak at 405 ev [76]. The deposition of LT GaN on such a surface results in zincblende GaN with (111) orientation and an incomplete coverage of the surface (Fig 3.7a). The possible reason for incomplete coverage of LT GaN is due to incomplete nitridation of sapphire at low temperatures. It has been shown earlier that GaN deposited on unnitrided sapphire is rough with 3D islands and with incomplete coverage [11]. This is due to the high energy interface between GaN and sapphire [11]. Therefore we believe that the high energy interface between GaN and sapphire is not significantly affected by low temperature nitrided layer and we continue to obtain 3D islands of GaN with incomplete coverage. The possible reason for the directionality of LT GaN (Fig. 3.7a) is due to the nucleation of LT GaN along the atomic steps of the sapphire surface (Fig. 3.2). However the ramp up and anneal process provides a more complete coverage along with distinct texturing along the [0002] axis of wurtzite GaN (Fig 3.7c & 3.11b). The ramp up and anneal steps are accompanied by N1s peak shifts from the underlying nitrided surface (Fig. 3.5 and 3.6), the uncovered portions of which are exposed to H 2 and NH 3 ambient at higher temperatures of 1000 o C. As we described in Sec of Chapter 2, during ramp up and anneal stages LT GaN starts decompose and redeposits on sapphire surface, which is in the ambient of NH 3 and H 2. The decomposition and redeposition of LT GaN involves the surface migration of decomposed Ga and N adatoms, and incorporation of these adatoms into the newly formed growing

78 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 54 nuclei and/or on the existing nuclei, which eventually yields LT GaN with different surface morphology when compared to the as grown stage (Fig. 3.7a, b & c). It appears that the decomposition and redeposition of GaN occurs on further nitrided region complexes which are now rich in N atom content, given that the high intensity N1s photoelectron peak has a value of 397 ev which corresponds to Al-N bond in O atom environment (Fig. 3.6). Therefore, it is plausible to state that the increase in coverage of LT GaN during ramp up and anneal steps is due to reduction in interface energy between GaN and sapphire via changes in the transformed structure of underlying nitrided layer from AlO x N 1 x to dominant Al-N bond in O atom environment. It also appears that this process leads to the formation of wurtzite LT GaN oriented along [0002] direction (Fig. 3.11b) High temperature nitridation (T N = 1100 o C) At higher nitridation temperatures (T N = 800 and 1100 o C) there is adequate nitrogen from the decomposed NH 3 source to completely replace the O atoms in sapphire surface [58], and stable Al-N bonds will be formed due to complete exchange of N and O atoms in sapphire surface [78]. This is indicated by the high intensity XPS N1s peak at ev, which corresponds to that of Al-N binding energy in bulk AlN [58, 71]. There is no significant change in this structure of nitrided layer on annealing after ramp up as shown in Fig 3.6. As a consequence LT GaN deposited on such surfaces after high temperature nitridation results in strong [0002] oriented wurtzite LT GaN with complete surface coverage in as grown and annealed stages (Fig. 3.7d & g and 3.11c & d). The complete coverage of LT GaN is due to the transformed nitrided layer (Al-N bond environment), which results in reduced interface energy between GaN and sapphire, and eventually promotes 2D growth of LT GaN. It has been shown that nitrided layer acts as wetting layer and it reduces the interface energy between GaN and sapphire [61]. Thus, we believe that the complete coverage of LT GaN on sapphire nitrided at temperatures T N 800 o C (Fig. 3.7d & g) is due to the wetting nature of transformed nitrided layer. The surface coverage of LT GaN is complete even after ramp up stage also, but

79 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 55 the morphology contains pits of various size. The possible reason for the pits of various size is due to the polarity of LT GaN layer. It was reported that N-polar GaN decomposes rapidly in the ambient of NH 3 & H 2 than Ga-polar GaN [67]. We suspect that the as grown LT GaN contains GaN of both polarities and in the subsequent ramp up process the N-polar domains etch at faster rate than Ga-polar domains, which eventually leads to pits in the LT GaN of various size (Fig. 3.7e & h). The surface morphology at the end of the anneal step contains islands of small in size with bright contrast (Fig. 3.7f & i), are could be remnants of Ga-polar LT GaN, which have been left after etch of N-polar LT GaN. The strong [0002] orientation of wurtzite LT GaN is also due to the transformed Al-N bond nitrided layer which reduces the lattice mismatch between GaN and sapphire, and aids the NL in forming with the [0002] orientation with respect to the underlying substrate [58, 59, 71 73]. Therefore, we state that the structural evolution of nitrided layer plays an important role in determining the morphology, surface coverage and orientation of subsequently grown LT GaN NL with respect to the underlying sapphire substrate. 3.5 Summary & conclusions Chemical bond evolution of transformed nitrided layers on sapphire surfaces have been investigated systematically at different stages of their processing. It is found that HT nitridation (T N = 1100 o C) yields modified sapphire surface with Al-N bond rich environment. In contrast LT nitridation yields (T N = 530 o C) sapphire surface modified with the dominant Al 2 O 3 based structural motifs in the as grown stage and transformed to Al-N bond with O atom content. The effect of these seed layers on the microstructural evolution of subsequently grown LT GaN NLs have been studied systematically. It is observed that the morphology, surface coverage and orientation of LT GaN NL strongly depends on the evolution of the transformed nitrided layer on sapphire surface. LT GaN NLs processed on HT nitrided sapphire are found to have strong [0002] orientation with complete surface coverage. In contrast, LT GaN processed on LT nitrided sapphire showed

80 Chapter 3. Sapphire Pre-treatment & Microstructural Evolution of LT GaN 56 zincblende phase with incomplete coverage in the as grown stage and later it got transformed to wurtzite phase with [0002] orientation and with enhanced surface coverage after the subsequent ramp up and anneal steps.

81 Chapter 4 Polarity & Microstructural Evolution of HT GaN This chapter focuses on the following two points, the first one is to understand the relation between the structure of the transformed nitrided layer on sapphire and polarity of the subsequently grown two-step HT GaN epitaxial layers. The second one is to understand the relation between various growth parameters such as growth temperature, V/III ratio, LT GaN parameters and carrier gas etc., on the surface morphology and crystalline quality of HT GaN layers. 4.1 Background The influence of various growth parameters on microstructural evolution of HT GaN layers can be easily understand by monitoring the in-situ optical reflectivity traces as discussed in subsequent Sec of Chapter 2. Fig. 4.1 shows one such reflectivity trace of GaN two-step epitaxy on sapphire. Different processing stages are indicated in relation to the two-step process. ˆ Stage 1 corresponds to the pre-treatment of sapphire wafers. During this period wafers are thermally annealed at 1100 o C under the ambient of H 2 57

82 Chapter 4. Polarity & Microstructural Evolution of HT GaN 58 followed by sapphire nitridation step. During this stage the reflectivity stays constant at sapphire reflectivity ( ). The optical reflectivity trace is normalized to bare sapphire reflectivity. ˆ Stage 2, starts with the deposition of LT GaN NL at 530 o C. The rise in reflectivity is due to change in refractive index from sapphire to GaN and the reflectivity continues to increase due to interference effects of optical rays from LT GaN and sapphire. ˆ In stage 3, the as deposited LT GaN layer is annealed thermally at 1000 o C under the ambient of H 2 and NH 3. This stage involves several steps such as decomposition of LT GaN, surface migration of decomposed Ga and N adatoms and redeposition of GaN nuclei. All these steps eventually lead to roughening of LT GaN surface and reduction in the thickness of LT GaN layer [62, 79]. The reason for the drop in reflectivity shown in Fig. 4.1 is due to roughness as well as reduction in thickness of annealed LT GaN [62, 79]. ˆ Stage 4, starts with the deposition of HT GaN at 1000 o C, 400 mbar. During this stage HT GaN starts growing in island mode (3D) and the reflectivity drops to a minimum due to high surface roughness of the sample [69, 80]. The reflectivity continues to stay at a minimum until the coalescence of HT GaN islands occur. The coalescence of islands lead to recovery in optical reflectivity and after this point the HT GaN starts grow in layer by layer mode (2D) [69, 80]. The time duration between the point where the reflectivity hits a minimum and where the reflectivity recovers is called roughening recovery period. This period is a direct measure of crystalline quality of HT GaN and is sensitive to the growth parameters. ˆ In stage 5, HT GaN starts grow in layer by layer mode (2D mode). The oscillations in the reflectivity trace is due to interference effects of the optical rays which are reflected from GaN surface and GaN/sapphire interface. The period of oscillation gives the growth rate of the film as discussed in Sec of Chapter 2.

83 Chapter 4. Polarity & Microstructural Evolution of HT GaN 59 Figure 4.1: In-situ optical reflectivity trace of two-step GaN epitaxy on sapphire (0001) wafers. In spite of all the success achieved with Ga-polar GaN and its alloys, the surface morphology of N-polar GaN films grown under identical conditions, are rough with hexagonal facets and are not suitable for device fabrication [18] (Fig. 4.2). Intentionally miscut sapphire (0001) wafers (2 o to 4 o ) along a- and m-directions were used to obtain device quality N-polar GaN layers as shown in Fig. 4.2b [23]. Fig. 4.3 shows the HRXRD rocking curve FWHM values of N-polar GaN layers grown on miscut and non-miscut nitrided sapphire wafers in comparison to conventional Ga-polar GaN layers [19]. Figure 4.2: Surface morphology of N-polar GaN layers grown on (a) nonmiscut and (b) highly miscut nitrided sapphire wafers. Sumiya et al., JAP, 88, 1158, Keller et al., JAP, 102, , Reprinted with permission.

84 Chapter 4. Polarity & Microstructural Evolution of HT GaN 60 Figure 4.3: High resolution x-ray diffraction rocking curve FWHM values for N-polar GaN in comparison to conventional Ga-polar GaN. Sun et al., JCG, 311, 2948, Reprinted with permission. Although the growth mechanism of Ga-polar GaN has been clarified to a great extent, this knowledge can not be applied to N-polar GaN directly, because the surface structures of Ga-polar GaN and N-polar GaN are not the same. The surface diffusion barriers for Ga and N adatoms are different on Ga-polar GaN and N-polar GaN surfaces [17]. It is suggested that the rough surface morphology of N-polar GaN is due to mixed polar domains and the hillock formation is due to difference in growth rates of Ga-polar and N-polar domains [81]. The centre of the hillock is found to be terminate with Ga-polar domain whereas the surrounding matrix is termonated with N-polar GaN [81]. These inversion domains (IDs) are found to originate from the regions which are rich in O atom content [82, 83]. They are also found to originate from GaN/sapphire film interface [82, 83] and also from the steps on the sapphire surface [84]. However, by careful optimization of growth parameters such as nitridation temperature Sun et al. obtained relatively smoother N-polar GaN films [19]. Device quality N-polar GaN layers were obtained on intentionally miscut (2 o to 4 o ) sapphire (0001) wafers [19, 23]. Intentionally miscut wafers (2 o to 4 o ) enhances the mobility of adatoms significantly and when the step density is high the probability of incorporation Ga adatoms at the step edges is high and it results in a step flow growth mode. At lower step densities (0.5 o to 1 o ) the distance between adjacent steps is large and it results in

85 Chapter 4. Polarity & Microstructural Evolution of HT GaN 61 nucleation of islands on terraces between steps [23]. Therefore, it is plausible to state that the surface quality of N-polar GaN is controlled by the following two factors ˆ Surface mobility of adatoms on the growth surface, which is controlled by the miscut of sapphire wafers [23]. ˆ Density of IDs, which is controlled by both the O atom content on the nitrided surface and miscut of sapphire wafers [82 84]. However, to date there are no reports on systematic analysis of polarity selection and microstructural evolution of LT GaN NLs and HT GaN layers in relation to the structure of the transformed nitrided layers on sapphire wafers. This chapter addresses the factors which control the surface morphology of N-polar GaN through surface mobility of adatoms and density of IDs on non-miscut sapphire (0001) wafers. 4.2 Experimental The in situ thermal treatment of wafers were carried out at 1100 o C under the flow of purified H 2 for 10 min before nitridation. The nitridation step was performed at three different temperatures T N = 530, 800 and 1100 o C and 200 mbar reactor pressure. Wafers were nitrided under the ambient of NH 3 and H 2 for 1 min. NH 3 flow rate was kept constant at 1500 sccm throughout the process. After sapphire nitridation, LT GaN was grown to a 60 nm thick layer to understand the effect of nitridation on microstructural evolution of NL. The thickness calibration was done by the in situ optical tool as discussed in Sec of Chapter 2. LT GaN NL was deposited on nitrided wafers at T N = 530 o C for 8 min (to deposit 60 nm thick layer), 200 mbar, at a V/III ratio of 2535, TMGa flow rate of 0.6 sccm and H 2 flow rate of 6500 sccm. The wafers were then heated up from the LT GaN NL growth temperature to the LT GaN NL annealing condition in 5 minutes. The NLs were annealed thermally for 4 minutes at 1000 o C and a

86 Chapter 4. Polarity & Microstructural Evolution of HT GaN 62 Parameter A1 A2 A3 A4 A5 LT GaN Thickness, nm to 60 5 LT GaN Anneal Time, min LT GaN Anneal, NH 3 flow, sccm to HT GaN Growth Temperature, o C to HT GaN Growth Pressure, mbar HT GaN TMG flow, sccm HT GaN NH 3 flow, sccm to HT GaN V/III ratio to HT GaN Thickness, µm HT GaN Carrier gas H 2 H 2 H 2 H 2 H 2 Table 4.1: HT GaN growth parameter space for sapphire nitridation at T N = 530 o C reactor pressure of 400 mbar in a mixture of 2000 sccm of NH 3 and 4700 sccm of H 2. Several growth parameters are varied to understand the polarity and micro-structural evolution of HT GaN layers. These are tabulated in Table 4.1 & 4.2. Table 4.1 corresponds to growth parameters for sapphire nitridation at T N = 530 o C and Table 4.2 is for growth parameters at T N = 1100 o C.

87 Chapter 4. Polarity & Microstructural Evolution of HT GaN 63 Parameter B1 B2 B3 B4 B5 B6 B7 LT GaN Thickness, nm LT GaN Anneal Time, min LT GaN Anneal to NH 3 flow, sccm 5000 HT GaN Growth to Temperature, o C HT GaN Growth Pressure, mbar HT GaN TMG flow, sccm HT GaN to NH 3 flow, sccm 5000 HT GaN V/III ratio to HT GaN Thickness, µm HT GaN Carrier gas H 2 H 2 H 2 H 2 H 2 H 2 N 2 Table 4.2: HT GaN growth parameter space for sapphire nitridation at T N = 1100 o C 4.3 Results Low temperature nitridation (T N = 530 o C) Fig. 4.4a shows the surface morphology of HT GaN layer deposited on 4 min annealed LT GaN for sapphire wafer nitrided at T N = 530 o C (Fig. 3.7c of Chapter 3), as per growth conditions given in column A1 of Table 4.1, that is at a V/III ratio of 485. The morphology is quite rough and is due to incomplete coalescence of HT GaN islands. The corresponding reflectivity trace is shown in Fig. 4.4b. There is no recovery in reflectivity due to high surface roughness of the film. Red color point on the trace indicates the starting point of HT GaN growth.

88 Chapter 4. Polarity & Microstructural Evolution of HT GaN 64 Figure 4.4: Surface morphology of HT GaN grown for sapphire nitrided at T N = 530 o C, at a V/III ratio of 485 and its corresponding optical reflectivity trace. To enhance the island coalescence and to obtain device quality HT GaN epitaxial layers for the nitridation temperature of T N = 530 o C we have systematically varied several growth parameters such as V/III ratio, growth temperature of HT GaN and LT GaN thickness/growth time. The effect of each of these growth parameters on surface and microstructural evolution of HT GaN will be described in the subsequent sections V/III ratio V/III ratio is the ratio of input flow rates of group-v (NH 3 ) and group-iii precursors (TMGa, TMAl & TMIn). V/III ratio is varied in this dissertation by keeping the TMGa flow rate constantly at 4.1 sccm. NH 3 flow rate is varied from 2000 to 5000 sccm. Fig. 4.5 shows the surface morphologies of HT GaN layers grown at different V/III ratios with their corresponding reflectivity traces. The corresponding growth conditions are given in column A2 of Table 4.1. Fig. 4.6 shows the corresponding RMS surface roughness data of these samples. The surface roughness is found to be increase to 0.75 nm (V/III = 1205) from 0.5 nm (V/III = 965).

89 Chapter 4. Polarity & Microstructural Evolution of HT GaN 65 Figure 4.5: Nomarski optical micrographs of HT GaN surfaces and their corresponding in-situ optical reflectivity traces, grown at V/III ratios (a) 965, (b) 1055 and (c) 1205 for a nitridation temperature of T N = 530 o C. In all traces the red points indicate starting point of HT GaN growth. The information that we extract from the reflectivity traces shown in Fig. 4.5 are the corresponding growth rates and the roughening recovery times of HT GaN. It is found that the growth rate of HT GaN decreases with the increase in V/III ratio. The growth rate can be obtained from the period of the reflectivity oscillations as described in the Sec of Chapter 2. It is also found that the roughening recovery time decreases with the increase in V/III ratio. Fig. 4.7 shows the variation in the growth rate and roughening recovery time of HT GaN layers plotted against the V/III ratios. The mosaicity of all these HT GaN samples were characterized by HRXRD. Fig.

90 Chapter 4. Polarity & Microstructural Evolution of HT GaN 66 Figure 4.6: AFM surface roughness data for the HT GaN samples grown at different V/III ratios for sapphire nitridation at T N = 530 o C. Figure 4.7: Variation in the growth rate and roughening recovery time of HT GaN samples deposited at different V/III ratios for sapphire nitridation at T N = 530 o C. 4.8 shows the high resolution XRD rocking curve FWHM values for both symmetric (0002) and asymmetric {10 11} peaks of GaN layers. FWHM values are found to increase with the V/III ratio.

91 Chapter 4. Polarity & Microstructural Evolution of HT GaN 67 Figure 4.8: High resolution XRD rocking curve measurements for HT GaN samples grown at different V/III ratios for sapphire nitrided at T N = 530 o C. The results suggest that the surface quality and crystalline quality of HT GaN layers depends critically on the V/III ratio Growth temperature The other growth parameter which controls the surface morphology of HT GaN is the growth temperature. We have varied HT GaN growth temperature from 1000 o C to 1050 o C using the growth parameters given in column A3 of Table 4.1. Fig. 4.9 shows the surface morphology of HT GaN layers with their corresponding reflectivity traces for sapphire nitridation of T N = 530 o C. Fig shows the corresponding surface roughness evolution of HT GaN layers. The roughness is found to be decrease with the increase in growth temperature. Fig shows the corresponding changes in growth rate and roughening recovery time of the HT GaN layers. A slight increase in the growth rate of HT GaN is observed. In contrast, it is found that the roughening recovery time decreases to 1 min from 25min and this is due to the faster coalescence of HT GaN islands at higher growth temperatures.

92 Chapter 4. Polarity & Microstructural Evolution of HT GaN 68 Figure 4.9: Nomarski optical micrographs of HT GaN surfaces and their corresponding in-situ optical reflectivity traces, grown at temperatures (a) 1000 o C, (b) 1025 o C and (c) 1050 o C for sapphire nitridation at T N = 530 o C. In all traces the red points indicate starting point of HT GaN growth. Interestingly, we did not find any significant change in the crystalline quality (mosaicity) of HT GaN samples grown at different growth temperatures. Fig shows the HRXRD rocking curve FWHM values for symmetric (0002) and asymmetric {10 11} peaks.

93 Chapter 4. Polarity & Microstructural Evolution of HT GaN 69 Figure 4.10: AFM surface roughness evolution of HT GaN layers deposited at growth temperatures: 1000, 1025 and 1050 o CC for nitridation at T N = 530 o C. Figure 4.11: Variation in roughening recovery and growth rate of HT GaN samples grown at different growth temperatures for sapphire nitridation at T N = 530 o C LT GaN NL thickness In order to improve the crystalline quality of HT GaN layers further we have systematically varied the thickness/growth time of LT GaN NLs to control the

94 Chapter 4. Polarity & Microstructural Evolution of HT GaN 70 Figure 4.12: High resolution XRD rocking curve FWHM values for symmetric (0002) and asymmetric {10 11} peaks of HT GaN samples grown at different growth temperatures for sapphire nitridation at T N = 530 o C. nucleation density (process conditions: column A4 of Table 4.1). Fig shows the RMS roughness as a function of LT GaN thickness. The roughness is insensitive to thickness except at very low thicknesses where it rises sharply. Fig shows the roughening recovery rate which decreases with the increase in thickness. The FWHM values are found to be lowest for 25 nm thickness of LT GaN NL (Fig. 4.15). The optimum thickness of LT GaN and the optimum process parameters are may not be same for all the reactors and it may vary for different reactor geometries Polarity of HT GaN It is well known from the literature that the HT GaN samples grown on nonnitrided sapphire consistently yield Ga-polar GaN. Potassium hydroxide (KOH) etching and convergent electron beam diffraction (CBED) experiments were carried out on these HT GaN samples for polarity measurement. It is known from the literature that KOH etches the N-polar GaN whereas Ga-polar GaN is resistant to KOH etch [85 88]. The HT GaN samples are etched in 0.2 mol of KOH solution

95 Chapter 4. Polarity & Microstructural Evolution of HT GaN 71 Figure 4.13: AFM surface roughness data of HT GaN as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C. Figure 4.14: Roughening recovery time data of HT GaN as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C. at 60 o C for 10 min. Fig shows the surface morphology of HT GaN samples before and after KOH etching experiment. It can be seen that HT GaN samples are not significantly affected by the KOH treatment as shown in Fig. 4.16a and 4.16b. The experimental CBED patterns were taken along < > zone axis of GaN and the patterns were compared with JEMS simulated patterns. The JEMS patterns were simulated for different

96 Chapter 4. Polarity & Microstructural Evolution of HT GaN 72 Figure 4.15: High resolution XRD rocking curve FWHM values for symmetric (0002) and asymmetric {10 11} peaks of HT GaN samples as a function of LT GaN NL thickness for sapphire nitridation at T N = 530 o C. Figure 4.16: Surface morphologies (a) before and (b) after KOH etch experiments of HT GaN layers grown on annealed NLs for sapphire wafers nitrided at T N = 530 o C. The corresponding convergent beam electron diffraction patterns along with their cross-sectional images of GaN films are shown in (c).

97 Chapter 4. Polarity & Microstructural Evolution of HT GaN 73 thicknesses 110 and 145 nm from the GaN/sapphire interface. The 0002 disc indicates a bright band at the center whereas disc indicates a dark band at the center for Ga-polar GaN film. These etching results and CBED patterns shown in Fig. 4.16c indicate that the two-step HT GaN layers grown on surfaces nitrided at T N = 530 o C are Ga-polar GaN [19, 85 89] High temperature nitridation (T N = 1100 o C ) Fig shows the surface morphology of HT GaN layer deposited on 4 min annealed LT GaN NL for sapphire wafers nitrided at T N = 1100 o C (column B1 of Table 4.2 ). The morphology is quite rough and contains hexagonal faceted islands. The corresponding reflectivity trace is shown in Fig. 4.17b. There is no recovery in the reflectivity due to high surface roughness of the film. Figure 4.17: Nomarski optical microscopy images of HT GaN layers for sapphire wafer nitrided at T N = 1100 o C. The corresponding reflectivity trace is also shown in the figure. The red point on the trace indicates the starting point of HT GaN growth. Immediately after HT GaN growth reflectivity rises substantially and then drops down to a minimum after two small oscillations. This indicates that HT GaN starts with a 2D growth mode at the beginning of the growth and then in the subsequent process the growth mode is transformed to a 3D mode. We have varied several

98 Chapter 4. Polarity & Microstructural Evolution of HT GaN 74 growth parameters to enhance the surface quality of HT GaN layers. These studies are described below V/III ratio Fig shows the Nomarski optical microscopy images of HT GaN layers grown at different V/III ratios of 965, 1130 and The corresponding growth parameters are given in column B2 of Table 4.2 Figure 4.18: Nomarski optical microscopy images of HT GaN layers deposited at V/III ratios (a) 965, (b) 1130 and (c) 1205 for sapphire nitridation at T N = 1100 o C. In all cases the surface morphology is rough with faceted hexagonal islands. The corresponding reflectivity traces are not shown here because there was no recovery in the reflectivity traces due to high surface roughness. It has been shown earlier that higher V/III ratio can enhance the surface quality of HT GaN for sapphire nitridation at T N = 530 o C. In contrast, for HT GaN samples for sapphire nitridation at T N = 1100 o C, the surface morphology remains rough with faceted hexagonal islands and with small changes in the density and size of the islands for the given V/III ratio range. Fig shows the FWHM values of x-ray rocking curves of grown HT GaN samples at different V/III ratios for sapphire nitridation at T N = 1100 o C. FWHM values of (0002) and {10 11} peaks are found to increase with the V/III ratio.

99 Chapter 4. Polarity & Microstructural Evolution of HT GaN 75 Figure 4.19: High resolution XRD rocking curve FWHM values for HT GaN layer: (a) (0002) and (b) {10 11} peaks for sapphire nitridation at T N = 1100 o C Polarity of HT GaN The surface morphology of HT GaN layers for high nitridation temperatures is rough with many hexagonal hillocks as shown in Fig & Prior to further optimization of HT GaN layers, potassium hydroxide (KOH) etching and convergent electron beam diffraction (CBED) experiments were carried out on these HT GaN samples for polarity measurement. HT GaN samples are etched in 0.2 mol of KOH solution at 60 o C for 10 min. Fig shows the surface morphology of HT GaN samples before and after KOH etching experiment. The growth conditions for the samples are given in column B3 of Table 4.2 The etching and CBED studies indicate that the samples grown on annealed NL for sapphire wafers nitrided at T N = 1100 o C are N-polar GaN. As can be seen from Fig. 4.20a and 4.20b, they are substantially etched away when subjected to the same KOH etch treatment [85 88]. CBED patterns shown in Fig. 4.20c confirm that they are N-polar [19, 89]. The CBED patterns were taken along < > zone axis of GaN and the patterns were compared with the JEMS simulated patterns. The JEMS patterns were simulated for different thicknesses 100, 105 and 115 nm from the GaN/sapphire interface. In this case both 0002 disc and discs are interchanged in contrast due to N-polarity of GaN film as

100 Chapter 4. Polarity & Microstructural Evolution of HT GaN 76 Figure 4.20: Surface morphologies (a) before and (b) after KOH etch experiments of HT GaN layers grown on annealed NLs for sapphire wafers nitrided at T N = 1100 o C. The corresponding convergent beam electron diffraction patterns along with their cross-sectional images for GaN films are shown in (c). compared to the T N = 530 o C case (Fig. 4.16). The hillock remnants which are left after KOH etch are Ga-polar since they do not etch LT GaN annealing time The rough surface morphology with faceted hexagonal islands seen in the previous section may originate from mixed polarities of HT GaN. It has been suggested in the literature that the faceted hexagonal hillocks originate from inversion domains in the structure that is from Ga polar domains that grow faster than the surrounding N-polar domains [81]. It has also been suggested that Ga-polar domains form on O atom rich surfaces [82, 83]. It was thought that the annealing period at the high temperature nitridation may enhance the O atom level on the nitrided surface as a consequence of diffusion from the sapphire substrate into the nitrided layer. To enhance the surface quality of HT GaN layers we have grown these layers on samples that were ramped up to the growth temperature but not annealed.

101 Chapter 4. Polarity & Microstructural Evolution of HT GaN 77 Fig shows the Nomarski optical microscopy images of HT GaN layers grown under identical conditions except the LT GaN annealing time (column B4 of Table 4.2). The sample shown in Fig. 4.21a is deposited on 4 min annealed LT GaN as described in previous sections. In contrast, HT GaN grown on samples directly after ramp up without annealing results in a substantial reduction of the hexagonal hillock density (Fig. 4.21b). Figure 4.21: Surface morphologies of HT GaN layers deposited on (a) 4 min annealed LT GaN layer and (b) as ramped up LT GaN layers for sapphire nitridation at T N = 1100 o C. To confirm the polarity in the HT GaN that was deposited in the as-ramped up condition, the above samples are subjected to same KOH etch treatment. Fig shows the side view of SEM surface morphologies of HT GaN epitaxial layers shown in Fig. 4.21a and 4.21b, after KOH etch experiment. Both process conditions lead to N-polar GaN. However, the density of the remnant hillocks after the KOH etching were significantly less for the HT GaN grown directly after ramp up (Fig. 4.22b), that is the smoother HT GaN was associated with a lower density of remnant hillocks (Fig. 4.21b). The results suggest that device quality N-polar GaN layers can be obtained on non-miscut sapphire (0001) wafers. Table 4.3 shows the HRXRD rocking curve FWHM values for the sample grown directly on LT GaN after ramp up, in relation to the samples grown on 4 min annealed LT GaN for sapphire nitridation at T N = 1100 o C.

102 Chapter 4. Polarity & Microstructural Evolution of HT GaN 78 Figure 4.22: Surface morphologies of HT GaN layers after KOH treatment deposited on (a) 4 min annealed LT GaN layer and (b) 0 min annealed (ramped up) LT GaN layers for sapphire nitrided at T N = 1100 o C. Layer LT GaN LT GaN (0002) {10 11} Thickness Anneal Time FWHM FWHM (nm) (min) (arc-sec) (arc-sec) HT GaN HT GaN Table 4.3: FWHM values of x-ray rocking curves for the HT GaN samples grown directly on LT GaN after ramp up, in relation to the samples grown on 4 min annealed LT GaN for sapphire nitridation at T N = 1100 o C Growth temperature The other process parameter which we have varied to control the surface morphology of HT GaN is the growth temperature. We have varied HT GaN growth temperature as described in the previous section from 1000 o C to 1050 o C. The corresponding growth parameters are listed in column B5 of Table 4.2. It is found that the size and density of hexagonal islands are very sensitive to growth temperature (Fig. 4.23). The density of islands is found to increase while the size of the hexagonal islands is found to decrease with the increase in growth temperature. Fig shows the corresponding HRXRD rocking curve FWHM values. The FWHM values of (0002) and {10 11} are found to be increase with the growth

103 Chapter 4. Polarity & Microstructural Evolution of HT GaN 79 Figure 4.23: Nomarski optical microscopy images of HT GaN layers grown at (a) 1000 o C, (b) 1025 o C and (c) 1050 o C growth temperatures for sapphire nitrided at T N = 1100 o C. temperature. Figure 4.24: HRXRD rocking curve FWHM values of HT GaN layers grown at temperatures: (a) 1000 o C, (b) 1025 o C and (c) 1050 o C growth temperatures for sapphire nitridation at T N = 1100 o C. Since density and size of hexagonal islands critically depends upon the growth temperature and is found to increase with the growth temperature. In order to understand further the surface evolution of N-polar GaN layers, we have also grown HT GaN samples at lower growth temperatures < 1000 o C. Fig shows the Nomarski optical images of HT GaN layers grown at two different temperatures of 900 and 800 o C as per the growth conditions given in column B6 of Table 4.2. Interestingly, the reflectivity oscillations were observed for the sample grown at

104 Chapter 4. Polarity & Microstructural Evolution of HT GaN 80 Figure 4.25: Optical surface images of HT GaN layers grown at different temperatures (a) 900 o C, (b) 800 o C (LT GaN annealed at 800 o C) and (c) 800 o C (LT GaN annealed at 1000 o C). The corresponding in-situ optical reflectivity traces are also shown. temperature 800 o C (Fig. 4.25b). The sample surface looks mirror like. In contrast, in the sample grown at 900 o C, decay in the reflectivity oscillations has been observed (Fig. 4.25a) and the corresponding surface morphology of the sample looks relatively rougher than the sample grown at 800 o C. The results indicate that lower growth temperatures < 1000 o C seems to be favorable for obtaining N- polar GaN with good surface quality. These samples are subjected to the same

105 Chapter 4. Polarity & Microstructural Evolution of HT GaN 81 Layer LT GaN LT GaN LT GaN HT GaN Surface Thickness Anneal Anneal Growth Roughness Time Temperature Temperature (nm) (min) ( o C) ( o C) (nm) HT GaN Rough HT GaN HT GaN HT GaN HT GaN Table 4.4: The surface roughness values of N-polar HT GaN grown at low & high growth temperatures for sapphire nitridation at T N = 1100 o C Layer LT GaN LT GaN HT GaN (0002) {10 11} Thickness Anneal Growth FWHM FWHM Temperature Temperature (nm) ( o C) ( o C) (arc-sec) (arc-sec) HT GaN HT GaN HT GaN HT GaN Table 4.5: HRXRD rocking curve FWHM values of N-polar HT GaN grown at low & high growth temperatures for sapphire nitridation at T N = 1100 o C KOH treatment described in the earlier sections and are confirmed to be N-polar. The surface roughness values are indicated in Table 4.4 corresponding HRXRD rocking curve FWHM values are tabulated in Table 4.5. To enhance the crystalline quality of the sample, we have annealed the LT GaN NL at 1000 o C and reduced the temperature to 800 o C for the subsequent HT GaN growth (column B6 of Table 4.2). The FWHM values are reduced drastically (Table 4.4) and the surface morphology of the corresponding sample along with its reflectivity trace is shown in Fig. 4.25c. The corresponding AFM morphologies of samples (Fig. 4.25b & c) are shown in Fig The RMS surface roughness of the samples is found to lie in between 1 to 1.5 nm. The results suggest that device quality N-polar HT GaN layers can be obtained on non-miscut sapphire (0001) wafers by careful optimization of growth temperature.

106 Chapter 4. Polarity & Microstructural Evolution of HT GaN 82 Figure 4.26: AFM morphologies of HT GaN layers grown at temperature 800 o C for high temperature nitridation: (a) LT GaN is annealed at 800 o C (HT GaN RMS roughness 1.5 nm) and (b) LT GaN is annealed at 1000 o C (HT GaN RMS roughness 1.2 nm) Carrier gas (H 2 /N 2 ) It has been reported well in the literature that the carrier gas N 2 enhances the lateral growth of N-polar GaN. In contrast, the carrier gas H 2 suppresses the lateral growth of N-polar GaN [20 22]. In order to investigate the effect of carrier gas on surface morphology of N-polar GaN. We have deposited N-polar HT GaN layers under identical conditions while altering the carrier gas from H 2 to N 2 (column B7 of Table 4.2). Fig shows the surface morphology of HT GaN layers. The density and size of the hexagonal islands are drastically reduced for the sample grown under N 2 as carrier gas (Fig. 4.27b) Summary of results This section describes the summary of key results of HT GaN layers obtained at nitridation temperatures T N = 530 & 1100 o C.

107 Chapter 4. Polarity & Microstructural Evolution of HT GaN 83 Figure 4.27: Nomarski optical microscopy images of HT GaN layers grown under (a) H 2 as carrier gas, and (b) N 2 as carrier gas for sapphire nitrided at T N = 1100 o C Low temperature nitridation (T N = 530 o C) ˆ Low temperature nitridation (T N = 530 o C) yields Ga-polar GaN. ˆ The surface quality of Ga-polar GaN critically depends upon the parameters such as V/III ratio and growth temperature. The quality is found to increase with the growth temperature and decrease at higher V/III ratios due to statistical roughening. ˆ The crystalline quality is found to decrease with the increase in V/III ratio and LT GaN thickness, and is not significantly affected by the growth temperature High temperature nitridation (T N = 1100 o C) ˆ High temperature nitridation (T N = 1100 o C) yields N-polar GaN. ˆ The surface quality of N-polar GaN is more sensitive to the HT GaN growth temperature. The quality is found to decrease with the increase in growth temperature, and is not significantly affected by the parameter V/III ratio. ˆ Low growth temperatures (800 o C) are found to be favorable for obtaining smooth N-polar GaN materials.

108 Chapter 4. Polarity & Microstructural Evolution of HT GaN 84 ˆ The crystalline quality is found to decrease with the increase in V/III ratio and growth temperature. 4.4 Discussion The experiments described in the previous section explore a wide range of process parameters in an attempt to provide optimum combinations of polarity, surface roughness and crystalline quality of HT GaN layers. This section provides a rationale for our observations first for HT GaN growth with low temperature nitridation (T N = 530 o C) and then for HT GaN growth with high temperature nitridation (T N = 1000 o C). The experience on optimization of growth parameters for Ga-polar GaN in our MOCVD reactor has been taken as a bench mark for understanding the growth mechanism of N-polar GaN layers Low temperature nitridation (T N = 530 o C) and HT GaN Polarity Our results suggest a central role for the O atom content in the nitrided layer, and the related surface structure of the nitrided sapphire surface, in determining the polarity of the HT GaN epitaxial layers. It has been shown that HT GaN grown on non-nitrided sapphire inevitably has Ga polarity [56] and we continue to obtain Ga-polar GaN even with lower nitridation temperatures T N = 530 o C. As shown in Table 3.1 and Fig. 3.4 of Chapter 3, the as-nitrided surface at low nitridation temperatures shows a relatively high O content. Ab-initio calculations of surface stability indicate that the existence of various Al 2 O 3 based structural motifs associated with different degrees of replacement of O by N at these temperatures [78]. The as-grown LT GaN has a cubic structure (Fig & 3.12

109 Chapter 4. Polarity & Microstructural Evolution of HT GaN 85 of Chapter 3). On annealing the LT GaN prior to HT GaN deposition the underlying nitrided layer appears to change in character to a AlN based structural motifs, albeit with still high O levels. LT GaN during the anneal decomposes and redeposits, presumably on the altered nitrided layers (Fig. 3.6 of Chapter 3) with more uniform coverage (Fig. 3.7 of Chapter 3) and with a hexagaonal structure (Fig of Chapter 3). It is suggested that this redeposited hexagonal LT GaN is Ga-polar since the nitrided layer has still a high O atom content shown in Table 3.1 of Chapter 3, in similarity with the Ga-polar HT GaN grown on non-nitrided sapphire. As a consequence the subsequent HT GaN epitaxial layer is Ga-polar. Thus our results suggest Ga-polar GaN is continues to form on surfaces that have been nitrided as long as the nitridation process (at low temperatures) results in O containing AlN structural motifs Crystalline quality of HT GaN The crystalline quality of Ga-polar HT GaN has been investigated through a characterization of the HRXRD rocking curve FWHM values of (0002) and {10 11} peaks over a wide range of parametric conditions that are summarized in Fig Both these peaks show similar trends. Fig therefore provides values of just the (0002) peaks. Figure 4.28: Summary of (0002) x-ray rocking curve FWHM values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 530 o C.

110 Chapter 4. Polarity & Microstructural Evolution of HT GaN 86 The values increase with V/III ratio, go through a minimum with NL thickness and are not sensitive to the growth temperature. A comparison with roughening recovery time in Fig. 4.7 as function of V/III ratio shows that the FWHM values decrease with the increase in roughening recovery time. Since the roughening recovery time is an indicator of LT GaN nucleus density, this suggests that FWHM values improve with lower nucleus density as might be expected if coalescence of island during the recovery stage control the defect density in HT GaN [90 93]. In order to understand the effect of V/III ratio on nucleation density, we have examined the LT GaN layer formed with different V/III ratios before the growth of HT GaN was initiated. The results shown in Fig demonstrate quite convincingly that LT GaN nucleation density decreases with the decrease in V/III ratio. In analogy, it may be expected that a decrease in the LT GaN thickness (other parameters remaining constant) would imply a lower density of nucleation sites (and increase roughening recovery time as observed in Fig. 4.7 & 4.14) and a consequent decrease the defect density and thus the FWHM values, and this is what is observed in Fig However at very low nucleation density, the FWHM values rise steeply again. It is suggested this occurs because of poor or substantially incomplete coverage of the substrate by LT GaN below a certain critical thickness so that LT GaN no longer provides effective nucleation sites (Fig. 4.15). The FWHM values are insensitive to growth temperature for a given LT GaN thickness and V/III ratio. This is expected since the reduced recovery times and increasing growth rates in this case (Fig. 4.12) arise from diffusional effects due to higher growth temperatures and do not reflect changes in the underlying LT GaN nuclei size Surface roughness of HT GaN The surface roughness of HT GaN increases with V/III ratio, decreases with growth temperature and is insensitive to NL layer thickness of 25 nm below which it increases sharply, as summarized in Fig

111 Chapter 4. Polarity & Microstructural Evolution of HT GaN 87 Figure 4.29: Surface morphologies of 4 min annealed LT GaN NLs at two different NH 3 flow rates: (a) 2000 sccm and (b) 5000 sccm Figure 4.30: Summary of surface roughness values of HT GaN layers over a wide range of growth conditions for sapphire nitrided at T N = 530 o C. The data suggests that surface roughness originates from a statistical roughening process that occurs when the diffusion length is shorter than the mean distance between the binding sites [50]. The increase in surface roughness with increasing in V/III ratio (Fig. 4.6) results from a decreased mobility of Ga atoms on a surface that is increasingly dominated by excess N atoms [17] as discussed in Sec of Chapter 2. The increase in growth temperature on the other hand increases mobility in general and therefore reduces statistical roughening (Fig. 4.10). The roughening is expected to be insensitive to LT GaN nucleation layer thickness, as is observed (Fig. 4.13), at constant V/III ratio and growth temperature. The sharp increase in roughness below a critical thickness of LT GaN arises from an