Citation
Metalorganic Chemical Vapor Deposition of Indium Nitride and Indium Gallium Nitride Thin Films and Nanostructures for Electronic and Photovoltaic Applications

Material Information

Title:
Metalorganic Chemical Vapor Deposition of Indium Nitride and Indium Gallium Nitride Thin Films and Nanostructures for Electronic and Photovoltaic Applications
Creator:
MANGUM, JOSHUA L. ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Alloys ( jstor )
Doping ( jstor )
Indium ( jstor )
Inlets ( jstor )
Nanorods ( jstor )
Nanowires ( jstor )
Oxygen ( jstor )
Photovoltaic cells ( jstor )
Silicon ( jstor )
Temperature ratio ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Joshua L. Mangum. Permission granted to University of Florida to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
7/12/2007
Resource Identifier:
659898646 ( OCLC )

Downloads

This item has the following downloads:

mangum_j ( .pdf )

mangum_j_Page_156.txt

mangum_j_Page_027.txt

mangum_j_Page_141.txt

mangum_j_Page_063.txt

mangum_j_Page_224.txt

mangum_j_Page_009.txt

mangum_j_Page_100.txt

mangum_j_Page_007.txt

mangum_j_Page_120.txt

mangum_j_Page_158.txt

mangum_j_Page_108.txt

mangum_j_Page_082.txt

mangum_j_Page_198.txt

mangum_j_Page_021.txt

mangum_j_Page_172.txt

mangum_j_Page_187.txt

mangum_j_Page_029.txt

mangum_j_Page_026.txt

mangum_j_Page_217.txt

mangum_j_Page_078.txt

mangum_j_Page_199.txt

mangum_j_Page_167.txt

mangum_j_Page_216.txt

mangum_j_Page_197.txt

mangum_j_Page_023.txt

mangum_j_Page_182.txt

mangum_j_Page_046.txt

mangum_j_Page_151.txt

mangum_j_Page_002.txt

mangum_j_Page_089.txt

mangum_j_Page_170.txt

mangum_j_Page_006.txt

mangum_j_Page_228.txt

mangum_j_Page_057.txt

mangum_j_Page_205.txt

mangum_j_Page_159.txt

mangum_j_Page_058.txt

mangum_j_Page_211.txt

mangum_j_Page_042.txt

mangum_j_Page_031.txt

mangum_j_Page_032.txt

mangum_j_Page_075.txt

mangum_j_Page_208.txt

mangum_j_Page_016.txt

mangum_j_Page_140.txt

mangum_j_Page_201.txt

mangum_j_Page_112.txt

mangum_j_Page_176.txt

mangum_j_Page_048.txt

mangum_j_Page_117.txt

mangum_j_Page_138.txt

mangum_j_Page_165.txt

mangum_j_Page_168.txt

mangum_j_Page_105.txt

mangum_j_Page_036.txt

mangum_j_Page_152.txt

mangum_j_Page_169.txt

mangum_j_Page_025.txt

mangum_j_Page_084.txt

mangum_j_Page_070.txt

mangum_j_Page_103.txt

mangum_j_Page_188.txt

mangum_j_Page_121.txt

mangum_j_Page_088.txt

mangum_j_Page_039.txt

mangum_j_Page_154.txt

mangum_j_Page_008.txt

mangum_j_Page_233.txt

mangum_j_Page_110.txt

mangum_j_Page_204.txt

mangum_j_Page_050.txt

mangum_j_Page_107.txt

mangum_j_Page_101.txt

mangum_j_Page_051.txt

mangum_j_Page_018.txt

mangum_j_Page_106.txt

mangum_j_Page_055.txt

mangum_j_Page_196.txt

mangum_j_Page_114.txt

mangum_j_Page_209.txt

mangum_j_Page_044.txt

mangum_j_Page_143.txt

mangum_j_Page_047.txt

mangum_j_Page_098.txt

mangum_j_Page_099.txt

mangum_j_Page_157.txt

mangum_j_Page_122.txt

mangum_j_Page_049.txt

mangum_j_Page_012.txt

mangum_j_pdf.txt

mangum_j_Page_202.txt

mangum_j_Page_213.txt

mangum_j_Page_043.txt

mangum_j_Page_079.txt

mangum_j_Page_166.txt

mangum_j_Page_148.txt

mangum_j_Page_060.txt

mangum_j_Page_118.txt

mangum_j_Page_061.txt

mangum_j_Page_150.txt

mangum_j_Page_017.txt

mangum_j_Page_014.txt

mangum_j_Page_091.txt

mangum_j_Page_227.txt

mangum_j_Page_226.txt

mangum_j_Page_232.txt

mangum_j_Page_004.txt

mangum_j_Page_115.txt

mangum_j_Page_062.txt

mangum_j_Page_173.txt

mangum_j_Page_147.txt

mangum_j_Page_087.txt

mangum_j_Page_220.txt

mangum_j_Page_180.txt

mangum_j_Page_068.txt

mangum_j_Page_210.txt

mangum_j_Page_113.txt

mangum_j_Page_184.txt

mangum_j_Page_175.txt

mangum_j_Page_153.txt

mangum_j_Page_131.txt

mangum_j_Page_161.txt

mangum_j_Page_092.txt

mangum_j_Page_030.txt

mangum_j_Page_073.txt

mangum_j_Page_095.txt

mangum_j_Page_116.txt

mangum_j_Page_074.txt

mangum_j_Page_223.txt

mangum_j_Page_145.txt

mangum_j_Page_219.txt

mangum_j_Page_019.txt

mangum_j_Page_221.txt

mangum_j_Page_162.txt

mangum_j_Page_041.txt

mangum_j_Page_013.txt

mangum_j_Page_022.txt

mangum_j_Page_067.txt

mangum_j_Page_218.txt

mangum_j_Page_203.txt

mangum_j_Page_065.txt

mangum_j_Page_189.txt

mangum_j_Page_059.txt

mangum_j_Page_102.txt

mangum_j_Page_215.txt

mangum_j_Page_109.txt

mangum_j_Page_081.txt

mangum_j_Page_190.txt

mangum_j_Page_206.txt

mangum_j_Page_090.txt

mangum_j_Page_034.txt

mangum_j_Page_136.txt

mangum_j_Page_080.txt

mangum_j_Page_185.txt

mangum_j_Page_040.txt

mangum_j_Page_174.txt

mangum_j_Page_020.txt

mangum_j_Page_097.txt

mangum_j_Page_194.txt

mangum_j_Page_015.txt

mangum_j_Page_076.txt

mangum_j_Page_077.txt

mangum_j_Page_005.txt

mangum_j_Page_142.txt

mangum_j_Page_119.txt

mangum_j_Page_132.txt

mangum_j_Page_064.txt

mangum_j_Page_214.txt

mangum_j_Page_144.txt

mangum_j_Page_177.txt

mangum_j_Page_130.txt

mangum_j_Page_028.txt

mangum_j_Page_054.txt

mangum_j_Page_045.txt

mangum_j_Page_139.txt

mangum_j_Page_038.txt

mangum_j_Page_128.txt

mangum_j_Page_225.txt

mangum_j_Page_069.txt

mangum_j_Page_134.txt

mangum_j_Page_155.txt

mangum_j_Page_229.txt

mangum_j_Page_011.txt

mangum_j_Page_231.txt

mangum_j_Page_230.txt

mangum_j_Page_171.txt

mangum_j_Page_181.txt

mangum_j_Page_146.txt

mangum_j_Page_104.txt

mangum_j_Page_111.txt

mangum_j_Page_137.txt

mangum_j_Page_212.txt

mangum_j_Page_127.txt

mangum_j_Page_037.txt

mangum_j_Page_052.txt

mangum_j_Page_179.txt

mangum_j_Page_178.txt

mangum_j_Page_125.txt

mangum_j_Page_183.txt

mangum_j_Page_135.txt

mangum_j_Page_056.txt

mangum_j_Page_096.txt

mangum_j_Page_094.txt

mangum_j_Page_001.txt

mangum_j_Page_093.txt

mangum_j_Page_024.txt

mangum_j_Page_053.txt

mangum_j_Page_186.txt

mangum_j_Page_133.txt

mangum_j_Page_193.txt

mangum_j_Page_085.txt

mangum_j_Page_126.txt

mangum_j_Page_086.txt

mangum_j_Page_191.txt

mangum_j_Page_129.txt

mangum_j_Page_066.txt

mangum_j_Page_035.txt

mangum_j_Page_033.txt

mangum_j_Page_003.txt

mangum_j_Page_123.txt

mangum_j_Page_010.txt

mangum_j_Page_164.txt

mangum_j_Page_160.txt

mangum_j_Page_072.txt

mangum_j_Page_195.txt

mangum_j_Page_192.txt

mangum_j_Page_222.txt

mangum_j_Page_071.txt

mangum_j_Page_163.txt

mangum_j_Page_207.txt

mangum_j_Page_124.txt

mangum_j_Page_200.txt

mangum_j_Page_149.txt

mangum_j_Page_083.txt


Full Text





METALORGANIC CHEMICAL VAPOR DEPOSITON OF INDIUM NITRIDE AND
INDIUM GALLIUM NITRIDE THIN FILMS AND NANOSTRUCTURES FOR
ELECTRONIC AND PHOTOVOLTAIC APPLICATIONS




















By

JOSHUA L. MANGUM


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2007



































2007 Joshua L. Mangum

































To Kimberly, Matt, Cassie, and Bill.









ACKNOWLEDGMENTS

First and foremost I would like to thank Dr. Tim Anderson for being my advisor and

supervisory committee chair. His extensive knowledge as a researcher and experience in

academia taught me invaluable lessons that I will carry with me throughout my career.

I would especially like to thank my supervisory committee cochair Dr. Olga Kryliouk. Dr.

Kryliouk has been a wonderful teacher and mentor, and a true friend. I feel lucky to have

worked with such a talented researcher, whose passion for research and friendly nature made

every day enjoyable. I would also like to thank my other committee members (Dr. Steven

Pearton and Dr. Jason Weaver) for their valuable insights.

My appreciation goes out to the entire staff at the Major Analytical Instrumentation Center

(MAIC), especially Dr. Amelia Dempere, Kerry Siebein, and Rosabel Ruiz, for giving me the

opportunity to work with them and for creating such an enjoyable work environment. Thanks go

to the staff at Microfabritech (especially Chuck Roland and Scott Gapinski) for always helping

me keep my equipment in proper working order.

I sincerely thank my family for their constant support during graduate school. I thank my

father for encouraging me to continue my education and I thank my mother for showing me that

anything can be accomplished with enough desire and effort. I also thank my brother for his

positive attitude and the fun times he brings into my life.

Most importantly I want to thank my fiancee, Kimberly, who has been my greatest

supporter and companion. Her patience and love has helped me achieve my goals throughout

graduate school. I also give thanks to the Gray family for being my home away from home

while I lived in Florida.









TABLE OF CONTENTS

page

A CK N O W LED G M EN T S ................................................................. ........... ............. .....

LIST OF TABLES ........................ ...........................................8

LIST OF FIGURES .................................. .. ..... ..... ................. .9

A B S T R A C T ................................ ............................................................ 17

CHAPTER

1 INTRODUCTION ............... .......................................................... 19

1.1 G group III-N itrides ........................................................... ....... .. 19
1.1.1 The III-N itride Crystal Structure.................................... ....................... 19
1.1.2 Properties of InN and GaN .............................................................................20
1.1.3 Growth of InN and InxGal-xN ..................................................................21
1.2 P h otov o ltaics ....................................................................................... .. 3 4
1.2.1 Fundamental Physics of Solar Cell Devices ............................................. 36
1.2.2 W hy InxG al-xN ? .................... ............... .... ................ .. .. ............... 38
1.2.3 Current Progress of InN and InxGal-xN Solar Cells........................................39
1.3 Terahertz Applications for InN and InxGal-xN.......................................................42
1.4 Statem ent of Thesis ................................................... ............. .... ..... 45

2 INDIUM GALLIUM NITRIDE SOLAR CELL DEVICE SIMULATIONS ........................59

2 .1 Introdu action ........................................................ .............. ................. 59
2.2 M EDICI Device Simulation Software ........................................ ....... ............... 60
2.3 Identification of InxGal-xN Solar Cell Parameters ..... ...........................60
2.4 R results .............. .. ....... ...... ........ ....... .. ....... ........... 63
2.4.1 Single-Junction InxGal-xN Cell Optimization ............................................63
2.4.2 M ulti-Junction InxGal-xN Solar Cells .................................... ............... 68
2.4.3 Phase Separation in InxGal-xN Solar Cells............................................ 69
2.4.3.1 Effect on single-junction cells ................................. ................ 69
2.4.3.2 Effect on multi-junction solar cells................... .......................... 73
2.5 Conclusions ............... ........... ....... ............ ......................... 75

3 GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN
FIMLS BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION ...............................84

3 .1 In tro d u ctio n ............................................................................................................. 8 4
3.2 E xperim mental P procedure ..................................................................... ..... ................85
3.2.1 Substrate Preparation ............................................... ............................. 85
3.2.2 The M OCVD Deposition Technique ..................................... ............... 85
3.2.3 T he M O C V D R eactor .......................................................................... ... ... 88









3.2.4 In-Situ Sapphire Substrate Surface Treatment and Buffer Layers ..................90
3 .3 R e su lts ......................................................... .................. ................ 9 0
3.3.1 Indium N itride.................. .............. .......... ..... ... .. .......... 90
3.3.1.1 G row th on silicon substrates.................................... .....................90
3.3.1.2 Film stability and aging ............................ ................................... 95
3.3.2 Indium G allium N itride............................................ ... ........................ 97
3.3.2.1 Metastable InxGal-xN alloys over the entire range (0 < x < 1) .........97
3.3.2.2 Effect of growth temperature on stability of Ino.sGa0.2N.................00
3.3.2.3 Effect of substrate on stability of InxGai-xN alloys ........................107
3.3.2.4 Determination of InxGal-xN growth rate and composition from
RB S m easurem ents ........... ..... ...... .... .... ........ ............... 113
3.3.2.5 Terahertz emission from InxGal-xN alloys....................................115
3.4 Conclusions ..................................................... .............. ........... 117

4 GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE
NANOWIRES BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION .................139

4 .1 Intro du action ................................................................................ 13 9
4.2 Experim mental Procedure ......................................................................... .. 140
4 .3 R results ................................... ...................... ................... 14 1
4.3.1 Proposed Mechanism for MOCVD Nanowire Growth................................141
4.3.2 Nanowire Morphology Dependence on Growth Parameters ........................145
4.3.2.1 G row th on p-Si (100) substrates...................................................... 145
4.3.2.2 Growth on GaN/c-A1203 substrates...............................149
4.3.3 InN Nanowire Composition and Structure................... ...................................151
4.3.4 Transport Properties of InN-In203 Core-Shell Nanowires.............................153
4.3.5 Single InN-In203 Core-Shell Nanowire for H2 Gas Sensing........................155
4 .4 C o n clu sio n s .................................................................................................. ..... 15 7

5 GROWTH AND CHARACTERIZATION OF INDIUM NITRIDE NANO- AND
MICRORODS BY HYDRIDE-METAL ORGANIC CHEMICAL VAPOR
D E P O SIT IO N ...................................... ................................................... 180

5.1 Introduction ....................................................................... ......... 180
5.2 Experim mental Procedure ......................................................................... .. 180
5.2.1 Substrate Preparation ................................................................ ............... 180
5.2.2 The H-MOCVD Reactor and Deposition Technique.................................... 180
5.3 R results .................. ....... ......... .......... .................................... 182
5.3.1 Characterization of InN Nanorods Grown on Different Substrates ..............182
5.3.2 Growth of InN Nano- and Microrods to Increase Aspect Ratio ...................185
5.3.3 Transport Properties of InN M icrorods .............. ...................... ...............187
5.4 Conclusions ....................................................... .............. .......... 188









6 EXPLORATORY STUDY OF GALLIUM NITRIDE NANO- AND
MICROSTRUCTURED GROWTH BY METAL ORGANIC CHEMICAL VAPOR
D EPO SITION ................ .... ......... .............. .............................199

6.1 Introduction ......... ........... .......................... ............................199
6.2 Experim mental Procedure ................. ......... ......................................... ............... 199
6 .3 R e su lts ............................................................................................2 0 0
6 .4 C on clu sion s ................................................... ......................... ................. 2 04

7 FUTURE WORK AND RECOMMENDATIONS ................................... .................213

7.1 R ole of O oxygen in InN ............... ........ ........ .. .... .......... ... .................. 213
7.2 Improving Crystalline Quality of InxGal-xN (1 < x < 0.3) Alloys ..........................214
7.3 P-Type Doping of InN and In-Rich InxGal-xN ...........................................................214
7.4 Growth of InxGal-xN Alloys by H-MOVPE.................................. .......................215

LIST OF REFEREN CES ......... ...................................... ......... ...... ................... 216

BIO GR A PH ICA L SK ETCH ............................................................... 233









LIST OF TABLES


Table page

1-1 Some fundamental properties of InN and GaN ............... ..................................57

1-2 Theoretical and experimental mobility and controlled doping ranges for InN and
GaN.................... ........................................ 57

1-3 Lattice and thermal expansion constants for InN substrates and buffer layers..................58

1-4 Internal quantum efficiency (IQE) of InxGal-xN p-i-n and quantum well solar cells........58

2-1 The GaN and InN material parameters used in the Medici simulations..........................82

2-2 Cell parameters for optimized cell structure with optimum band gap energy .................82

2-3 Cell characteristics for simulations allowing hot carrier collection in phase-separated
InxG al-xN solar cells......... ........... ................ ................ ...............................83

2-4 Characteristics of individual layers and overall cell for simulated multi-junction
InxG al-xN solar cells.............. .................................................................. ....... 83

3-1 Flow ratios and corresponding InxGal-xN compositions for metastable
InxGal-xN/c-Al203 grown at low temperature (530 C). .............................................136

3-2 The InxGal-xN (002) compositions) for each deposition temperature ranging from
530 to 770 C when grown at a constant inlet flow ratio of In/(In+Ga) = 0.8...............137

3-3 Thickness and compositional data obtained from RBS measurements for
InxGal-xN/c-Al203 grown for 2 hr ......................................................... ............... 138

3-4 Comparison of InxGal-xN film compositions as determined by XRD and RBS..............138

4-1 Flow velocities and Reynolds number at the outlet of the vertical inlet and inside the
quartz reactor tube .................. .................. ................... ......... .. ............ 178

4-2 The InN nanowire growth conditions for samples with varying substrate/surface
pretreatments and the resulting morphology information ................ ....... ............179

5-1 A random sampling of InN nanorod/microrod lengths and diameters for a variety of
grow th condition s........ .......................................................................... ......... ...... 198









LIST OF FIGURES


Figure page

1-1 The III-nitride zincblende crystal structure along various directions...............................47

1-2 The III-nitride wurtzite crystal structure along various directions. ..................................47

1-3 Cation-faced (Ga) and nitrogen-faced polarity for the GaN wurtzite crystal structure.....47

1-4 Velocity-field characteristics (T = 300K, n = 1017 cm-3) for wurtzite InN, GaN, A1N,
and zincblende G aA s. .......................... ........................ .. .... ................. 48

1-5 Vapor pressure of N2 in equilibrium with of InN, GaN, and A1N as a function of
tem p eratu re. ............................................................ ................. 4 8

1-6 Rhombohedral structure and surface planes of sapphire. ...............................................49

1-7 Theoretical and experimental examples of nanowire growth mechanisms....................49

1-8 Predicted binodal (solid) and spinodal (dashed) decomposition curves for InxGal-xN
assuming regular solution mixing ...................................... ......... ................... 50

1-9 The T-x phase diagrams of ternary InxGal-xN compounds........................ ................50

1-10 Several XRD spectra of phase separated InxGal-xN films at different temperatures .........51

1-11 Compositional control of InxGal-xN with respect to growth temperature..........................51

1-12 U.S. photovoltaic module implementation from 1992-2003 ...................... ................52

1-13 Comparison of world wide growth of PV production from 1990-2004. ...........................52

1-14 Market shares of different photovoltaic materials as of 2001................ .............. ....53

1-15 Schematic of light induced carrier generation in a generic solar cell...............................53

1-16 Solar irradiance vs. wavelength for extraterrestrial (AMO) and terrestrial (AM1.5)
sp ectra .......................................................... .................................. 54

1-17 Global solar irradiances averaged over three years (1991-1993) which account for
clo u d co v er ...................................... ....................................................... 5 5

1-18 Current-voltage characteristics of a generic solar cell................................................. 55

1-19 Incident solar flux for AMO (black) and AM1.5 (red) as well as the InxGal-xN band
gap energy range. ............................................................................56









1-20 Terahertz frequency range (shaded) of the electromagnetic spectrum, including
m molecular transitions .................. ......................................... .. ........ .... 56

2-1 Absorption coefficient of an InN thin film grown by MOCVD.....................................77

2-2 Absorption coefficient as a function of wavelength for several photovoltaic
sem conductors. ......................................................... ................. 77

2-3 Proposed single-junction InxGal-xN solar cell which has been modeled by Medici..........78

2-4 Solar cell structures used in single-junction solar cell simulation ...................................78

2-5 Single-junction absorber optimization steps............................................. .................. 79

2-6 Simulated solar cell efficiency vs. band gap energy of the refined InxGal-xN cell
structure for AM O and AM 1.5 illumination.................................................................... 79

2-7 Power vs. load plots for the refined solar cell structure for AMO and AM1.5
illum nation ................................................................................80

2-8 Current density vs. voltage curves for the refined solar cell structure under AMO and
A M 1 .5 illu m in action ....... .......................................................................... ... ... ............80

2-9 The efficiency and open circuit voltage as a function of junction number for
m ulti-junction InxG al-xN solar cells ............. ........................................... .... ......... 81

2-10 Energy band diagram showing the generation of electron-hole pairs and hot carriers
from incident light energy .......................................................... .. ............... 81

2-11 Phase-separated InxGal-xN p-n junction showing Ga-rich InxGal-xN precipitates
(green circles) in an In-rich InxGal-xN matrix (yellow bulk) ........... ..................81

2-12 Corresponding band diagram from the p-type region in Figure 2-11 showing the
In-rich InxGal-xN matrix and the Ga-rich precipitate (ppt)..............................................82

2-13 Band diagram showing trapping of carriers and recombination in narrow band gap
energy precipitates (ppt) in the wider band gap energy Ga-rich InxGal-xN matrix. ..........82

3-1 Solid-gas interfacial region for the reaction of TMGa and NH3 to form GaN ...............120

3-2 T he M O C V D reactor. ............................................................................ .................... 12 1

3-3 Scanning electron microscope image of direct InN growth on Si (111) with no in-situ
surface pretreatments. .............................................. ..............121

3-4 Scanning electron microscope image of InN grown on Si (111) using substrate
nitridation followed by a LT InN buffer layer.............................................................122









3-5 An XRD pattern of a polycrystalline InN thin film grown on Si (111) using a LT InN
b u ffe r lay e r ...................................... ................................................... 12 2

3-6 A TEM image of the SiOxN1-x intermediate layer used to produce high quality single
crystal G aN .............................................................................123

3-7 X-ray diffraction patterns of InN grown on Si (111) and Si (100) substrates using
substrate nitridation followed by a LT InN buffer layer ..............................................123

3-8 Cross-sectional SEM image of InN/Si (100) grown by MOCVD showing a growth
rate of 78 nm /hr ..................... .......... ......... ........ ............................124

3-9 An XRD spectrum of InN/Si (111) grown by H-MOCVD..........................................124

3-10 Scanning electron microscope images of InN/Si (111) grown by H-MOCVD...............125

3-11 Energy dispersive spectroscopy and AES analysis showing no evidence of chlorine
contamination in InN/Si (111) by H-MOCVD. ..................................... ...............125

3-12 Comparison of XRD patterns between aged and as-grown samples of InN on silicon
and sapphire substrates. ......................... ...... ..................................... .. .... .. 126

3-13 X-ray diffraction patterns for InN/c-Al203 annealed in a N2 (100 Torr) atmosphere at
500 C (20 min), 525 C (15 min), and 550 C (10 min). .........................................126

3-14 A series of XRD patterns for InxGal-xN alloys grown at low temperature (530 C) on
c-A1203 substrates. .................................................................... ......... 127

3-15 Experimental InxGal-xN compositions vs. inlet flow ratio (blue line). ..........................128

3-16 X-ray diffraction patterns of InxGal-xN thin films grown on c-A1203 substrates at
constant inlet flow ratio (In/(In+Ga) = 0.8) at temperature ranging from 530 to
770 C (blue lines). ...................................................................... 128

3-17 Magnitude of InxGal-xN (002) peak intensity vs. deposition temperature at a constant
inlet flow ratio of In/(In+G a) = 0.8 ........................................................................ .... 129

3-18 Film crystalline quality as determined by XRD FWHM of the InxGal-xN (002) peak.... 129

3-19 Magnitude of In-rich InxGa-xN (101) and Ga-rich InxGal-xN (002) peak intensity vs.
deposition temperature at a constant inlet flow ratio of In/(In+Ga) = 0.8.....................130

3-20 A series of XRD patterns for InxGal-xN alloys grown at low temperature (530 C) on
a-A120 3 substrates. ...................................... .......... .............. .. 130

3-21 Comparison of XRD spectra for InN grown on a-A1203 (blue) and c-A1203 (red)
su b state s ...................................... ...................................... .............. 13 1









3-22 X-ray diffraction FWHM InxGal-xN/a-Al203 at different values of inlet flow ratio
and the corresponding film stability. ........................................ ......................... 131

3-23 Hemholtz free energy of mixing of InxGal-xN as a function of InN mole fraction. ........132

3-24 X-ray diffraction patterns for InxGal-xN alloys grown at different inlet flow ratios on
L T InN /G aN /c-A 20 3. .......................................................... ...... 132

3-25 Growth rate of InxGa-xN/c-Al203 alloys as a function of total inlet group III flow
(sccm ) ................... .......................................................................... 13 3

3-26 Composition deviation of InxGa-xN/c-Al203 measured by RBS as a function of inlet
flow ratio ......... ...... ......... ............... ........ ..................... ............... 13 3

3-27 Time domain THz measurement system used to analyze InxGal-xN thin films.............134

3-28 Terahertz signal measured from the surface of pure InN and InxGal-xN thin films. .......134

3-29 The emitted THz frequency range and corresponding amplitude for InN and
InxG al-xN alloy s. ......................................................................... 135

3-30 Poor THz signal to noise ratio for thin (48 nm) In0.6Ga0.4N. .........................................135

4-1 Schematic of the standard horizontal inlet tube and its position with respect to the
graphite susceptor. ........................................................................159

4-2 Schematic of the vertical quartz inlet tube and its relative position with respect to the
graphite susceptor. ......................... .......................... ............ 160

4-3 Inlet tube dimentions for film and nanostructured growth. .............................................160

4-4 Schematic of the VLS mechanism for InN nanowire growth by MOCVD ....................161

4-5 Overhead view of vertical inlet tube showing the hot spot deposition zone (orange)
and the region where small In metal primary nucleation is likely to form....................161

4-6 Experimental verification of the proposed nanowire density variation due to initial
indium droplet nucleation size. ........................................ ......................................... 162

4-7 Effect of In metal wetting for InN nanowire growth on Si (100) and GaN substrates....162

4-8 Indium-Silicon binary phase diagram................................ ... .................163

4-9 Scanning electron microscope images of InN nanowires grown on p-Si (100)
substrates using nitridation followed by a LT InN buffer pretreatment ........................163

4-10 Scanning electron microscope image of a continuous polycrystalline InN thin film
grown with the horizontal inlet arrangement on p-Si (100) ..........................................164









4-11 Effect of post growth annealing on InN nanowire tip morphology .............................164

4-12 Scanning electron microscope images of InN nanowires (experiment number 68)
showing uniform dense coverage of the substrate ................................. .. .................. 165

4-13 Scanning electron microscope images of InN nanowires (experiment number 69)
showing less dense substrate coverage with uniform patches of nanowires .................165

4-14 Scanning electron images of InN nanowires (experiment number 70) showing
uniform coverage of the substrate with nanowire patches that vary in size and
nanow ire density per patch. .................................................................... ...................166

4-15 Scanning electron microscope images of InN nanowires (experiment number 71)
showing less ordered nanowire growth with significant nanowire branching...............167

4-16 Scanning electron microscope image of unsuccessful InN nanowire growth on
GaN/c-Al203 substrate. .................... ............... ..................... 167

4-17 Successful growth of InN nanowires on GaN/c-A1203 substrates by activating the
surface of GaN. ................ ............................................. 168

4-18 Scanning electron microscope image of InN nanowires grown on GaN/c-Al203
substrate at reduced inlet flow velocities ................................................ .................. 168

4-19 Scanning electron microscope images of InN nanowires grown on GaN/c-Al203
using two-step nucleation and growth approach................................... ............... 169

4-20 Scanning electron microscope images depicting the InN core-In203 shell structure
seen from broken nanowires. ...... ........................... ........................................170

4-21 X-ray diffraction spectra of InN core-shell nanowires on Si (100) substrate indicating
single crystal InN and polycrystalline In203 phases. .............. ................................. 171

4-22 Energy dispersive spectroscopy spot analysis on the outside of a core-shell
nanowires where indium, nitrogen, and oxygen are detected as well as silicon from
the underlying substrate. ............................. .........................................171

4-23 Transmission electron microscope images of a single InN-In203 core-shell nanowire...172

4-24 Selected area diffraction patterns of the InN core of an InN-In203 core-shell
nanow ire ................................................................. ................ ..... 173

4-25 Energy dispersive spectroscopy showing no oxygen detected for nanowires grown at
high inlet N/In ratios and reduced flow velocities.......... ........... ............... 173

4-26 Scanning electron microscope image of single InN-In203 core-shell nanowire
contacted by Ti/Al/Pt/Au pads for electrical measurements ................. .. ...................174









4-27 Current-voltage measurements for several InN-In203 core-shell nanowires, used for
determining the total resistance (RT) of each wire...................................... ..................174

4-28 The total measured resistance versus 1/r2 for InN-In203 nanowires .............................175

4-29 Scanning electron microscope images and dimensions of single InN-In2O3 core-shell
nanowires used in the contact and sheet resistivity calculations. ...................................175

4-30 Optical micrograph of a single InN-In2O3 core-shell nanowire after wire bonding. ......176

4-31 Current-voltage characteristics and time dependent response of InN-In203 single
wire device under N2 and 500 ppm H2 (in N2) ambients. ..............................................176

4-32 Current-voltage characteristics of single InN-In203 core-shell nanowires before and
after coating w ith 3 or 10 A of Pt. ........................................... ............................ 177

4-33 Time dependent current response of a single InN-In203 core-shell nanowires in a
500 ppm H2 ambient with different thicknesses of Pt....................................................177

5-1 Diagram of the H-MOCVD concentric inlet tube showing the arrangement of source
material flows and typical temperatures used experimentally for the source and
deposition zones ..................................................... ................ 189

5-2 A SEM image of the growth habits of InN nanorods illustrating flower-like growth
and c-axis orientation .................. ........................................ .. ........ .. .. 189

5-3 Low resolution TEM image of an InN nanorod grown on a-Al203 substrate showing
flat hexagonal cross section. ................................................ ................................ 190

5-4 Low resolution TEM image of an InN nanorod grown on r-A1203 substrate................ 190

5-5 High resolution TEM image of the rectangular shape which is seen at the end of the
nanorod when it is shaved from the substrate surface. ............. ..................................... 190

5-6 Low resolution and high resolution TEM images of the InN nanorod tip grown on
r-A 120 3 ......................................................... ................................. 19 1

5-7 High resolution TEM images showing plane bending in the tip of an InN nanorod
grow n on r-A 120 3. ..........................................................................191

5-8 Low resolution TEM images of InN nanorods grown on Si (111) substrate showing
asym m etric pencil tip ............................................................... .. ...... ... 192

5-9 High resolution TEM images of side facet roughness occurring for most nanorods
grown by H-M OCVD regardless of the substrate.......................................................... 192

5-10 Low resolution TEM image and CBED patterns of an InN nanorod grown on
a -A 20 3................................. .................................................................. ............... 1 9 3









5-11 Experimental and calculated CBED images indicating nitrogen polarity .....................193

5-12 Electron energy loss spectra with and without the detection of oxygen.......................194

5-13 High resolution TEM image showing thin amorphous oxide layer that sometimes
forms on the outside of the InN nanorods..................................................................... 194

5-14 Photoluminescence spectra of InN nanorods grown on different substrates.................195

5-15 Scanning electron microscope images of InN microrods grown on Si (111) substrate
for 90 min using Cl/In = 4.0, N/In = 250, and Tg = 600 C. .........................................195

5-16 Scanning electron microscope images of InN microrods grown on Si (111) substrate
for 90 min using Cl/In = 5.0, N/In = 250, and Tg = 650 C. .........................................196

5-17 A SEM image of InN microrods grown on Si (111) substrate for 90 min using
Cl/In = 5.0, N/In = 200, and Tg = 700 C............................................................... 196

5-18 Scanning electron microscope image of InN microrods grown on Si (111) substrate
after post growth in-situ HC1 gas etching. ............................................ ............... 196

5-19 Current-voltage measurements for several InN microrods, used for determining the
total measured resistance (RT) of each rod. ........................................ ............... 197

5-20 Total measured resistance vs. length/a2 for InN microrods. .........................................197

5-21 Scanning electron microscope images and dimensions of InN microrods used for
T L M analy sis. .......................................................................... 198

6-1 Comparison of GaN/c-Al203 grown at similar growth conditions with the horizontal
and vertical inlets. ..................................... ... ... .......... .............. .. 205

6-2 X-ray diffraction pattern of GaN grown with the vertical inlet on Si (100) substrates.. .205

6-3 Scanning electron microscope images of GaN grown at 750 oC and an inlet N/Ga
ratio of 3,000 on different substrates. ........................................ .......................... 206

6-4 Scanning electron microscope images of GaN microtubes grown at 750 oC and an
inlet N/Ga ratio of 3,000 on Si (100) substrates. ............ .............................. ........ 206

6-5 Scanning electron microscope images of GaN grown using the vertical inlet at
650 oC and N/Ga ratio of 3,000 on different substrates .............................................207

6-6 Representative EDS spectrum of single GaN microtube grown on Si (100) substrate. ..208

6-7 Scanning electron microscope images of GaN grown with the vertical inlet at the
lowest temperature tested (560-600 oC) showing excess of metal droplets on the
surface of the nanotubes................................................................................. ..... ........208









6-8 Scanning electron microscope images of GaN growth with the vertical inlet at 560 C
after a 10 min 10% HC1 wet etch showing nanotube structure ............. ... .................209

6-9 Energy dispersive spectrum of GaN nanotubes after HC1 wet etching. ........................209

6-10 Scanning electron microscope images of GaN/Si (100) with the vertical inlet at
650 C and N /G a = 20,000 ......... ................. ........................................ ...............2 10

6-11 A SEM images of GaN/Si (100) with the vertical inlet at 650 C and N/Ga = 10,000. ..210

6-12 Scanning electron microscope images of GaN/Si (100) with the vertical inlet at
650 C and N /G a = 7,000 .................................................. ............... ...............2 11

6-13 Scanning electron microscope images of curly GaN nanowires that form outside the
prim ary deposition zone of the vertical inlet. .................................................................212

6-14 Wire bonding step of GaN nanotube H2 gas sensing device where damage to the
nanotubes occurred. .......................... ...... ..................... .... .................212









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

METALORGANIC CHEMICAL VAPOR DEPOSITON OF INDIUM NITRIDE AND
INDIUM GALLIUM NITRIDE THIN FILMS AND NANOSTRUCTURES FOR
ELECTRONIC AND PHOTOVOLTAIC APPLICATIONS

By

Joshua L. Mangum

May 2007

Chair: Timothy Anderson
Cochair: Olga Kryliouk
Major Department: Chemical Engineering

Single and multi-junction InxGal-xN solar cell devices were modeled in one dimension

using MEDICI device simulation software to assess the potential of InxGal-xN-based solar cells.

Cell efficiencies of 16 and 27.4%, under AMO illumination were predicted for a single and a

5-junction InxGal-xN solar cell, respectively. Phase separation of InxGal-xN alloys is determined

to have little to no negative effects on the solar cell efficiency.

InxGal-xN alloys were grown by MOCVD over the entire compositional range (0 < x < 1)

and phase separation was analyzed with respect to substrate material and growth temperature. A

low deposition temperature of 530 C was used to produce metastable InxGal-xN/c-Al203 thin

films over the entire compositional range, which was demonstrated for the first time by

MOCVD. The use of higher deposition temperature and closely lattice matched substrates

resulted in phase separated films. Substrates with a larger lattice mismatch (c-A1203) introduce

strain in InxGal-xN which helps to stabilize the film, however, at the expense of crystalline

quality.

Growth of InN nanowires by MOCVD was controlled without the use of templates or

catalysts by varying the inlet flow pattern, N/In ratio, growth temperature, and substrate material.









A VLS growth mechanism is proposed, however, a VS growth mechanism can be achieved at

high N/In ratios. SEM and TEM analysis revealed a core-shell nanowire structure with a single

crystal InN core and a poly-crystalline In203 shell. Nanowire growth occurs along the [0002]

direction with diameters and lengths ranging from 100 to 300 nm and 10 to 40 um, respectively

for a 1 hr growth.

H-MOCVD growth of InN nano- and microrods occurred on different substrates and the

nanorod structure was studied by TEM. The polarity of the substrate directly affected the

nanorod tip shape and prismatic stacking faults are suggested as the cause for the flower-like

growth habit. Variation of growth parameters, such as temperature, N/In ratio, and Cl/In ratio

proved to be ineffective at changing the aspect ratio of the nanorods. Increased growth duration

produces microrod size dimensions regardless of the chosen growth conditions.









CHAPTER 1
INTRODUCTION

1.1 Group III-Nitrides

The III-nitrides and their alloys exhibit direct band gap values ranging from infrared to

ultraviolet wavelengths, which make them important for applications in electronic and

optoelectronic devices. Gallium nitride (GaN) and its alloy with indium nitride (InxGal-xN) have

become dominant materials for producing high brightness light emitting diodes (LEDs) and laser

diodes (LDs) that emit light in the blue region of the visible spectrum. Research on III-nitrides

has resulted in LEDs and LDs that emit light at visible, infrared (IR) and ultra-violet (UV)

wavelengths. These optoelectronic devices can be used for CD and DVD media, high brightness

displays, and solid state lighting, which has been projected to reach seven billion dollars by

2009.1 The wide band gap energy range also makes these materials candidates for absorber

layers in solar cells since the absorption edge of these materials can be varied to optimize cell

efficiency. III-nitride electronic devices are also more environmentally friendly because they do

not contain toxic elements such as arsenic, which is used to fabricate other compound

semiconductors.

1.1.1 The III-Nitride Crystal Structure

III-nitrides can exist as a cubic zincblende crystal structure, but it is thermodynamically

more stable to exist in the hexagonal wurtzite crystal structure. The cubic and hexagonal phases

differ only by the stacking sequence of close-packed III-N planes, and the energy difference

between the two structures is small. Changing the sequence during growth produces defects

such as stacking faults. The zincblende structure (space group F43m, number 216) has an

ABCABC stacking sequence of (111) close packed planes (Figure 1-1). The wurtzite structure

(space group P63mc, number 186) has an ABABAB stacking sequence with alternating layers of









close packed (0001) metal atoms (In, Ga, Al) and nitrogen pairs, where every other layer is

directly aligned (Figure 1-2).2

The zincblende and wurtzite structures both contain polar axes and therefore lack inversion

symmetry. The polarity type of grown films has been shown to significantly affect bulk and

surface properties.3'4 The bonds in the <0001> direction (wurtzite) and <111> (zincblende) are

cation-faced while the opposite direction is nitrogen-faced (Figure 1-3).5 As a general rule the

term "termination", such as N-terminated, is reserved for describing a surface property. For

example the surface of an indium nitride film can be N-terminated by depositing one monolayer

of nitrogen atoms, but the orientation of the crystal remains unchanged.6 GaN has been shown to

be Ga-faced when deposited by MOCVD on sapphire substrates and N-faced when deposited by

MBE or HVPE on sapphire substrates.7-10 Ga-faced GaN films usually have an atomically

smooth surface while N-faced GaN films have a rough surface morphology."1 The growth

parameters such as substrate material, substrate surface treatment, and N/III ratio are important

factors in controlling the polarity.6

1.1.2 Properties of InN and GaN

The properties of InN and GaN will be discussed in this section and A1N will be ignored

since it is not the primary focus of this work. Some fundamental properties of wurtzite and

zincblende InN and GaN are shown in Table 1.1. The lattice parameter of InxGal-xN can be

found by using Vegard's law as a function of the composition (x) (Eq. 1-1). The InxGal-xN film

composition can be determined experimentally by techniques such as X-ray diffraction (XRD)

and Rutherford Backscattering (RBS).

alnGal-N = xaInN + (1 x)aGaN, similarly for c-axis (1-1)









Electrical properties. As grown InN and GaN typically have large background n-type

carrier concentrations, which makes p-type doping of these materials difficult. GaN has been

more heavily studied than InN and the GaN film quality has progressed such that the background

carrier concentration is low (4 x 1016 cm-3) with relatively high electron mobility

(600 cm2/V-s).12 With these advances p-type doping of GaN has been consistently reproduced

for a number of years. It is know that the ability to dope semiconductors depends on the position

of the valence and conduction band edges with respect to a common energy reference (i.e. the

Fermi level stabilization energy, EFS).13 For this reason it is assumed that InN would be easier to

p-type dope than GaN because the position of the valence band edge is 1.1 eV closer to the

EFs.14'15 InN however still proves to be very difficult to p-type dope. Evidence has been recently

presented for bulk p-type doping of InN with magnesium by isolating the effects of the n-type

surface accumulation layer that occurs in InN structures.16'17 It is also believed that only a small

fraction of Mg acceptors are ionized at room temperature, but these recent results are promising

for the development of InN p-n junction devices. Theoretical mobilities and controlled doping

ranges of InN and GaN are listed in Table 1-2. InN and GaN have been shown to have very high

peak drift velocities at room temperature, higher than that of GaAs, and unlike GaAs the drift

velocities were shown to be fairly insensitive to temperature (Figure 1-4).18,19

1.1.3 Growth of InN and InxGal-xN

Indium Nitride Thin Films. InN was first demonstrated by Juza and Hahn in 1938 from a

InF6(NH4)3 precursor and the resulting InN had a wurtzite crystal structure.20 In the following

years, attempts at producing InN were made by several researchers.21-24 The resulting InN

samples were typically powders or small crystals and were prepared by ammonization of indium

containing molecules or thermal decomposition of more complex molecules containing indium

and nitrogen. These early experiments revealed that indium containing molecules would not









react with inactive N2 molecules, even at higher temperature25 and that the equilibrium nitrogen

vapor pressure was very high.24 During the 1970s and 80s polycrystalline InN films were being

deposited mainly by sputter deposition. Growth on silicon and sapphire substrates was done at

temperatures ranging from room temperature to 600 C with typical electron mobility of 250 +

50 cm2/V-s and carrier concentrations in the range of 5-8 x 1018 cm-3.26 Analysis of the optical

properties of InN corresponded to a band gap energy value of 2.0 eV.27-30 During this period

Tansley and Foley extensively studied RF sputtering of InN and its properties3136 and they have

grown films with the highest mobility and lowest intrinsic carrier concentration to date (2700

cm2/V-s and n = 5 x 1016 cm-3).31 It has been shown that InN dissociates at temperature greater

than 600 oC.28

During the 1990s single crystal growth of InN became more prevalent, mostly by MOCVD

and molecular beam epitaxy (MBE) growth techniques and subsequent films properties were

dramatically improved. Single crystal InN by MOCVD on sapphire substrates was first grown

by Matsuoka et al. using trimethyl indium (TMIn) and ammonia precursors.37 At about the same

time Wakahara et al. also demonstrated MOCVD single crystal InN on sapphire substrates by

reacting TMIn with microwave activated nitrogen.38'39 By 2002, single crystal InN films with

low background carrier concentrations and high mobilities had been produced, 5.8 x 1018 cm-3

and 730 cm2/V-s (for MOCVD)40 and 3.49 x 1017 cm-3 and 2050 cm2/V-s (for MBE).41 These

were significant improvements in InN film quality at the time because it was difficult to grow

InN films by MOCVD and MBE with background carrier concentrations below 1019 cm-3 with

mobility greater than 300 cm2/V-s.

By 2002 many researchers started to question the fundamental band gap energy value of

InN, because photoluminescence and absorbance data suggested that InN's band gap energy was









closer to 0.7 eV,42-49 instead of the previously accepted value of 2.0 eV. Today, the value of the

band gap energy of InN is still not agreed upon; although there is compelling evidence that the

lower band gap energy value is correct. There have been several reviews and recent papers that

try to reveal reasons for the discrepancies in the InN band gap energy value.5055 The theories

that receive the most attention for band gap energy modulation in InN are oxygen incorporation

in the films, Moss-Burstein effect, trapping levels, and film stoichiometry. Indium oxide has a

band gap energy of 3.2 eV, therefore it is assumed that oxynitrides are formed during growth and

increase the band gap energy. The Moss-Burstein effect, which was first studied by Trainer and

Rose,28 occurs when the Fermi level is pushed into the conduction band as the electron

concentration increases above the Mott critical density and therefore the band gap energy is

overestimated by optical absorption. The Mott critical density is reached when an electron gas

cannot sustain itself due to reduced electron screening in order to form bound states.56 A source

for under estimation in the InN band gap energy from PL measurements of InN films can be

linked to deep level traps, with activation energies on the order of 0.6-0.7 eV. The defect level

explanation is reinforced by the fact the highest mobility of InN to date (2700 cm2/V-s)31 is much

lower then the theoretically predicted maximum value of 4400 cm2/V-s,57 suggesting a high

concentration of compensated defects.58

Non-stoichiometric films are potential causes for band gap energy variation in InN. In-rich

film stoichiometry leads to the formation of deep level defects (0.7 eV) when indium aggregates

form and these defects states can serve as potential photoluminescence peaks.59 A low band gap

energy value of 0.7 eV can also be the result of Mie resonances from indium precipitates. It is

well known that optical losses occur from resonant light scattering and absorption by dispersed

metallic particles.60 MBE grown In-rich InN samples showed that bipolar absorption,









accompanied with transformation into heat occurred in indium clusters, as well as resonant light

scattering due to plasmon excitations.61 Absorption from these indium precipitates makes it

difficult to determine the true band gap energy value when using optical absorption

measurements. Current evidence for the InN band gap energy values range from 0.6 to 2.3 eV,

suggesting that a variety of factors are involved and that no singular source is responsible for the

disparity in the InN band gap energy.50

The band gap energy of InxGal-xN can be inferred from a linear interpolation of the

compound band gap energy values and using the bowing parameter (b, determined

experimentally), as seen in Eq. 1-2. Recent results by Davydov et al. have shown the InxGal-xN

bowing parameter to be 2.5 eV,43 while other researchers have presented evidence for a bowing

parameter that is a function of composition,62 as represented by Eq. 1-3.

E, (x) = xE, (InN) + (1 x)E (GaN) bx(1 x) (1-2)

b(x) = (1- x)[11.4 -19.4x] eV (1-3)

There are several growth challenges that make it difficult to produce high quality single

crystal InN films by MOCVD. Of the III-nitrides, InN is by far the most difficult to grow due to

its high equilibrium nitrogen vapor pressure (Figure 1-5).63 The high equilibrium vapor pressure

of InN limits the deposition temperature to less than 650 C to prevent film decomposition.64

The source materials typically used in MOCVD growth of InN are TMIn and NH3. At these

lower deposition temperatures, the extent of ammonia decomposition is very low, less than 0.1%

at 500 C.12 Due to this lack of reactive nitrogen, indium droplets can form on the surface,

therefore the inlet N/In ratio must be kept sufficiently high (-50,000) to avoid formation of

indium droplets.6









High inlet N/In ratios are only required for growth at temperature < 600 C since ammonia

decomposition occurs readily at higher temperature (> 650 C). The extent of decomposition of

ammonia, however, significantly increases the H2 partial pressure, which has been shown to

retard the InN growth rate.66 This is also the reason that a nitrogen carrier gas is preferred over a

hydrogen carrier gas. Other nitrogen sources such as hydrazine (N2H4) have been suggested

from the results of equilibrium calculations for the growth of III-nitrides.67 The analysis showed

that the growth rate of InN could be increased without the formation of indium droplets if

hydrazine was used. From a practical stand point ammonia is still used most frequently in

MOCVD growth of InN due to the explosive nature of hydrazine.

With these growth challenges, there is a narrow temperature window (400 650 C) for

successful growth of InN by MOCVD. For conventional MOCVD the growth temperature is the

most important parameter for controlling film properties such as crystalline quality, growth rate,

surface morphology and carrier concentration.68 Modified MOCVD deposition techniques, such

as plasma or laser assisted MOCVD are starting to gain popularity due to the ability to produce

more reactive nitrogen at low growth temperatures, thus avoiding some of the pitfalls of

conventional MOCVD.

High quality growth of InN is also hindered by the fact that there are no substrates that are

appropriately matched for lattice constant and thermal expansion coefficient. Sapphire substrates

(c-A1203) are most frequently used for growth of InN even though there is a large lattice

mismatch, 26%. A-, C-, and R-plane sapphire substrates were used for selected growths in this

work and the different crystal orientations of sapphire are shown in Figure 1-6.5 Silicon

substrates are becoming more popular due to their lower cost and potential for future device

integration.69 Compared to sapphire, silicon substrates have a much lower lattice mismatch (8%









for Si( 11)) with InN, but the resulting InN films are usually polycrystalline due to an

amorphous SiNx layer that forms at the Si surface during the initiation of growth.70 Typical

substrates used for growth of InN and their corresponding lattice mismatches and thermal

expansion coefficient difference are shown in Table 1-3.

A large lattice mismatch and difference in thermal expansion coefficient can lead to a large

number of structural defects. To reduce the number of defects in the heteroepitaxial InN,

substrate nitridation (for c-A1203) and buffer layers are used to improve film quality. Buffer

layers are a two step growth method that is commonly used in heteroepitaxy, which consists of a

low temperature nucleation layer followed by the main epitaxial layer. The buffer layer serves to

change the nucleation process to promote lateral growth of the subsequent film. For sapphire

substrates initial nitridation as well as buffer layers are used to improve crystal quality. Substrate

nitridation forms A1N nuclei on the c-A1203 surface, and the lattice mismatch is reduced from

26% to 14% for InN/AlN.70-72 Complementary to nitridation, InN,73'74 GaN,75'76 and A1N77

buffer layers are used to improve InN film quality on sapphire substrates. InN films with the

lowest background carrier concentration and highest electron mobility to date were grown by

MBE (3.49 x 1017 cm-3 and 2,050 cm2/V-s) using a GaN buffer layer on sapphire.41 The best

MOCVD results to date are a carrier concentration and electron mobility of 5.8 x 1018 cm-3 and

900 cm2/V-s, respectively, which were grown on c-sapphire substrates at atmospheric pressure.8

Similar to sapphire substrates, the crystalline quality of InN on Si substrates has also been

improved by the use of buffer layers.69'77'79'80

Indium Nitride Nanostructures. Since lijima discovered carbon nanotubes,81 there has

been a large interest in developing one-dimensional (ID) structures, such as nanowires,

nanorods, nanotubes, and nanobelts from other materials. Nanostructures are unique because









dimensionality and size confinement affects electrical, optical, and structural properties.

Contrary to nano-films, the additional confinement dimension of nanowires allows carrier

confinement along a specific conducting path. III-nitride nanowires have potential applications

in low power field-effect transistors (FETs), LEDs, solar cells, terahertz emitters and

detectors.82,83 These types of nanostructures are synthesized by a variety of physical and

chemical methods. The first InN nanowires were demonstrated by Dingman et al. by

decomposition of azido-indium precursors.84 The nanowires had lengths ranging from 100-1000

nm with an average diameter of 20 nm and the growth was attributed to a precursor solution-

liquid-solid (SLS) mechanism. The growth of InN nanowires and nanorods has been reported by

a number of researchers.85-106

Similar to Dingman et al. other researchers have used solvothermal methods to grow single

crystal InN nanostructures at relatively low temperature (300 oC).88 InN nanowires have been

synthesized over a wide range of temperature, and as high as 700 C.87'90 96'103,105 Other

investigators have synthesized InN nanowires at 420 oC,89 450 C,99 440-525 C,97 500 C,91,92

550 oC,95 600 oC,102 550-700 C,104 and 600-730 C.96

A variety of precursors have been used for InN nanowire growth. Single source

precursors88'95 are less common and most InN nanowire synthesis uses separate indium and

nitrogen precursors. Solid sources are typically used for the indium precursor, while nitrogen

precursors are gas sources. Indium precursors are pure indium metal,85'89'90'92'97'101 trimethyl

indium,106 indium oxide powder,87,93,96 or a combination of both indium and indium

oxide.98,103,105 The nitrogen source is typically NH3 but activated N2 has also been used.89'97

Most InN nanowire growth processes occur via vapor-solid (VS) or vapor-liquid-solid

(VLS) mechanisms. As previously mentioned InN nanowire growth via SLS mechanism has









been done, however the majority of synthesis does not proceed through this route. The VLS

mechanism uses a metal particle that acts as a catalytically active site that promotes growth of

the nanowire. The metal catalyst forms a liquid alloy with indium, where by gas phase reactions

occur with the liquid metal alloy to form a solid nanowire. The liquid alloy attracts indium vapor

which leads to solid precipitation once the alloys reaches a super saturation point. A metal

catalyst such as Au85'92 or Ni 91 have been used to create a VLS mechanism for growth and it has

also been suggested that indium acts as a catalyst (unintentionally) when no catalyst other

catalyst is used.87'90'93'99 When no metal droplet is present at the end of the nanowire it is

assumed that the reaction proceeded through a VS mechanism.103 Figure 1-7 shows a schematic

of the VLS mechanism,107 as well as experimental pictures of nanowires with92 and without103

metal droplets at the end.

InN nanowires have been grown on several different substrates, Si/SiO2,98 Si(100),85,92

Al203(0001),89 polycrystalline A1N and GaN,99 or no intentional substrate at all.87'90 In the case

of growth without an intentional substrate, nanowire samples are scratched from reactor walls or

precursor crucibles and subsequently characterized.

The diameter of InN nanowires range from 10 to 500 nm with lengths on the order of 1 to

100 microns and the growth rate varies significantly depending on the type of deposition

technique used.84,90-92,95,96,98,103 Nanowire properties, such as band gap energy, also vary

depending on the type of growth method used. The band gap values range from 0.7 to

0.9 eV,85'90,97 1.1 eV,89 and 1.7 to 1.9 eV,92,93,104 and some researchers have even reported low

(0.8 eV) and high (1.9 eV) band gap energy values for InN nanowires grown by the same

method.91'99 Lan et al.90 produced InN nanorods on Si(100) substrates using a gold catalyst

where the diameter of the nanorod influenced the band gap energy. No conclusions were









presented for the difference in the band gap energy with respect to nanorod diameter.91 Surface

electron accumulation and the Moss-Burstein effect are likely possibilities for this band gap

energy variation. Vaddiraju et al.99 suggested the higher band gap energy values of InN

nanowires are from oxygen incorporation into the nanowires, since samples often contain

mixtures of In203 and InN.

Zhang et al.105 and Yin et al. 101 have demonstrated single step growth methods that

produce core-shell nanowire structures, InN core-In203 shell and InN core-InP shell,

respectively. Core-shell structures offer the ability to study how interfacial states affect

nanowire properties and progress for developing future radial heterostructure nanowire devices.

Qian et al. 10 has grown GaN-based radial core-shell LED heterostructures that emit light over

wavelengths from 365 to 600 nm with high quantum efficiencies.

For InN nanowire device technology to progress several growth challenges must be

overcome. As in the case of InN films some fundamental properties of InN nanostructures, such

as the band gap energy, need to be studied. A variety of InN nanowire synthesis processes have

been reviewed in this section, but it is important that deposition techniques be implemented with

current technologies. It is also important that InN nanostructures be reproducible and the

deposition be precisely controlled to make progress towards more complex devices. The ability

to p-type dope InN nanostructures will also be important for future device applications.

Indium Gallium Nitride Thin Films. The first InxGal-xN alloys were grown by

Osamura et al.27 and optical absorption measurements were given to reveal the relationship

between band gap energy and alloy composition and theoretical bowing parameter of 1.05 eV

was determined. Osamura et al. later noticed that InxGal-xN alloys phase separated after

annealing in an argon atmosphere at 700 C.29 Nagatomo et al. 109 grew InxGal-xN alloys by









MOCVD on sapphire substrates with 0 < x < 0.42 at 500 oC, while Yoshimoto et al. 110 produced

single crystal InxGal-xN alloys on sapphire substrates with 0 < x < 0.22 at 800 C. The higher

temperature growth of InxGal-xN sufficiently improved the quality to allow for PL to be observed

for the first time from an InxGal-xN alloy.

The primary focus of InxGal-xN research has been for Ga-rich solutions that are important

for applications in light emitting diodes (LEDs). Shuji Nakamura pioneered the development of

visible LEDs in the early 1990s, especially high brightness blue LEDs based on III-nitride

heterstructures.111 Adding small amounts of indium to GaN became the optimal material for

active layers in LEDs. InxGal-xN active regions are usually highly defective due to a large

number of threading dislocations, yet the LEDs remain highly efficient.112 Other III-V

compound semiconductors have a much more sensitive relationship between extended defect

density and device performance. Blue, green, amber, and UV LEDs have been demonstrated

using InxGal-xN active regions in either double heterostructure or quantum well structures.113

The indium composition of these active layers are usually less than 10%, with the exception of

some quantum well structures where the indium content can be as high as 45%. Considerably

less studies of In-rich InxGal-xN alloys have been made because AlGaAs and AlGaInP LEDs

cover this wavelength range. LED devices typically use very thin layers of InxGal-xN, as small

as 25 A, and the indium content not usually greater than 50%. The InN band gap energy

controversy, which started in 2002, has increased the interest in In-rich InxGal-xN alloys as a

result of device applications now possible with the smaller band gap energy of InN such as full

spectrum solar cells14 and terahertz emitters and detectors.114 Most of the recent research on

In-rich InxGal-xN has been focused on understanding the fundamental properties of the alloys,









such as structural, electrical properties, and optical properties as well as improving film

quality.115-122

There are several growth challenges that must be overcome for successful growth of

InxGal-xN alloys. Perhaps most challenging is phase separation occurs in the alloy due to an

11% lattice mismatch between InN and GaN. Ho and Stringfellow analyzed the solid phase

miscibility gap in the InxGal-xN system and determined the maximum equilibrium incorporation

of In into GaN (or Ga into InN) is less than 6% at a typical deposition temperature of 800 C.123

The binodal and spinodal curves were calculated using a modified valence-force-field (VFF)

model (Figure 1-8) to estimate the interaction energy assuming the solid solution behaves as a

regular solution. It can be seen that InxGal-xN alloys are theoretically unstable or metastable

over a large compositional range at typical growth temperatures (Figure 1-8). The critical

temperature at which complete miscibility exists was calculated to be 1250 oC, which is greater

than the melting temperature of InN. Karpov performed a similar calculation to predict phase

separation of InxGal-xN alloys and showed that the miscibility gap could be reduced by

introducing compressive lattice strain (compared to unstrained films, Figure 1-9).124 Karpov's

results have been experimentally confirmed as evidence for the observation of single phase

InxGal-xN with up to 30% In incorporation at deposition temperature of 700-800 oC, for

InxGal-xN heterostructures or quantum dots.125-129 Phase separation can be identified by TEM

analysis or by observing peak shouldering or separation from XRD measurements (Figure 1-10).

More recently researchers have shown that single phase metastable InxGal-xN can be

grown over the entire compositional range by MBE when a low temperature is used.47'130

Compositional modulation also occurs in InxGal-xN films where nano-domains of In-rich or

Ga-rich sections can be formed.118'120'131'132 This non-uniform distribution has been shown to









affect the photoluminescence properties of the as grown films by decreasing the FWHM and

reducing the peak intensity.133

The vapor pressure difference between InN and GaN (Figure 1-5) is another problem that

affects high quality growth of InxGal-xN alloys. At lower deposition temperature indium

incorporation can be increased, however, higher crystalline quality is achieved at higher

deposition temperature. Figure 1-11 shows how indium incorporation is affected by deposition

temperature when using TMIn and TEGa.12 The distribution coefficient of In between the vapor

and sold phases is considerably greater than unity at 800 C because of the large difference in

decomposition pressure at elevated temperature along with near equilibrium conditions at the

growth interface at this higher temperature. At the lower temperature of 500 C the distribution

coefficient is close to unity suggesting non-equilibrium (reaction limited) conditions are

expected. It is also evident that at 800 C, control of the composition becomes difficult for

intermediate compositions given the rapid change in solid composition with that in the vapor.

Choosing the optimal N/III inlet ratio is directly affected by the specified deposition

temperature. Ammonia decomposition efficiency will determine the actual N/III ratio, however

it is very difficult to know the exact NH3 decomposition efficiency since the value relies heavily

on the reactor design as well as temperature. For this reason the inlet flow ratio of

NH3/(TMIn+TEGa) is commonly listed as the N/III ratio for MOCVD growth. In this work, a

N/III ratio of 50,000 is frequently used; however the actual N/In ratio at the substrate surface will

be lower, depending on the deposition temperature used. For simplicity the N/III ratio

mentioned in this work represents the NH3/(TMIn+TEGa) molar flow ratio into the reactor.

When the deposition temperature is low (< 600 C) the inlet N/III ratio must be high enough to

maintain sufficient levels of active nitrogen and avoid In droplet formation. As the temperature









is increased above 650 C the N/III ratio must be appropriately decreased so the excess hydrogen

partial pressure doesn't inhibit In incorporation into the film. Growth of InxGal-xN alloys

(especially for In-rich compositions) is fairly difficult due to the narrow growth regimes of InN

coupled with phase separation and vapor pressure differences that occur with the addition of

gallium to InN.

Indium Gallium Nitride Nanostructures. Growth of InxGal-xN nanostructures offers the

promise for improving the efficiency of III-nitride based LEDs. The main loss of efficiency in

III-nitride LEDs is through non-radiative recombination due to threading dislocations formed

during GaN and InxGal-xN film growth.134 Nanowire or nanorod growth is a way to practically

eliminate threading dislocations and significantly reduce the non-radiative recombination

centers. InxGal-xN nanostructures have been studied far less than InN nanostructures; there are

only a few reports for InxGal-xN nanostructured growth .135139

The first InxGal-xN nanostructures were produced by Kim et al. using hydride vapor phase

epitaxy at a low temperature of 510 C.138,139 These InxGal-xN nanorods on sapphire (0001)

substrates were approximately 50 nm in diameter, 10 tm in length, and oriented in the c-axis

(also known as "well aligned"). All the InxGal-xN nanorods were Ga-rich alloys with maximum

indium composition of 20%.

Chen et al. 136,137 produced InxGal-xN nanorings/nanodots and ordered InxGal-xN

nanolines/nanodots, respectively, by using selective area nitride growth on patterned SiO2 masks

on GaN substrates. These InxGal-xN nanostructures, approximately 80 nm in diameter (or

across) were also Ga-rich showing PL emissions at 420 nm (2.95 eV)136 and 450-500 nm (2.76 -

2.48 eV).137









Cai et al.135 was the first to demonstrate In-rich InxGal-xN nanostructure growth. InxGal-xN

straight and helical nanowires were grown in a tube furnace using elemental Ga and In which

were evaporated in an NH3/Ar flow and deposited on a Au covered Si(100) substrate. This

method is similar to that has been used to produce InN nanowires.92 These InxGal-xN

nanostructures also show a high In-content core (30-60%) and low In-content shell (4-10%) with

either a hexagonal or cubic structure. Possible mechanisms for the helical growth structure was

attributed to lateral displacement of the Au catalyst compared to the nanowire central axis or

differing growth rates between the core and shell of the nanowire. The core-shell structure was

believed to occur due to phase separation of the InxGal-xN alloy.

Growth of InxGal-xN nanostructures involves the difficulties related to InxGal-xN film

growth such as compositional control and phase separation as well as reproducibility and

controlled synthesis related to nanowire growth. As previously mentioned growth synthesis can

be precisely controlled using patterning of SiO2 masks, however, future device specifications

may not always allow for pattering schemes to take place.

1.2 Photovoltaics

Photovoltaics (PV) is the field of study where electricity is directly produced from solar

radiation (sunlight). The photovoltaic effect was first discovered by Edmund Baquerel in 1839

when he observed the effect of light on silver coated platinum electrodes that were immersed in

an electrolyte.140 By the 1950s developments in the silicon electronics industry made it possible

to fabricate silicon p-n junctions, which served as the first solar cell devices.141 During the 1950s

and 60s silicon solar cell technology was developed for space and satellite applications where

delivery of fuel was difficult. Interest in solar cells peaked again in the western world in the

1970s due to the oil-dependent energy crisis. During this time other cheaper solar cell

technologies were explored, such as polycrystalline Si, amorphous Si, thin film, and organic









materials. The 1990s was a time where the need was recognized for renewable energy sources

and alternatives to fossil fuels, promptly significant growth in solar cell research and technology

occurred at this time. By the late 1990s the photovoltaics industry was growing by 15-25% per

year, and for the first time solar cell power generation became competitive with remote low

power applications (rural electrification, navigation, and telecommunications).142

Today solar cell technology and its implementation is of global interest and some countries

such as Germany and Japan are leading the way. By 2003 the solar electricity industry provided

employment for over 35,000 people world wide (European Photovoltaic Industry Association,

Brussels, Belgium, EPIA Roadmap, 2004, http://www.epia.org/documents/Roadmap_EPIA.pdf,

accessed April 2007). From 2004 to 2005 the PV industry doubled and annual sales have

surpassed $10 billion, which is drawing significant capital investment. U.S. implementation of

solar cell modules has increased in recent years, however, the U.S. has been losing market share

to other countries in Europe as well as Japan.143 The loss in market share is directly related to

the lack of U.S. growth in grid connected systems, which has grown significantly (world wide) in

recent years. One third of the 90 Megawatts (MW) that were installed in the U.S. in 2004 came

from California alone.143 The progress of U.S. photovoltaic module implementation in recent

years can be seen in Figure 1-12,144 and a comparison to world wide markets can be seen in

Figure 1-13.143

Over 95% of current solar cell technologies in use are silicon based (Figure 1-14).145,146

There are however a variety of technologies and materials being actively researched to produce

more cost effective solar cells. Some of the new technologies being pursued are solar

concentration systems that focus light onto solar cell absorbers, thin film organic and inorganic









materials to reduce materials cost and processing time and multi-junction solar cells to absorb

more of the solar spectrum's light.

The highest efficiency solar cells to date are based on III-V materials, specifically

GaInP/GaInAs/Ge tandem cells that achieve efficiencies of 39%.146 InxGal-xN alloys are

currently only proposed as a potential material for high efficiency solar cells.

1.2.1 Fundamental Physics of Solar Cell Devices

For light energy to be converted into electricity a solar cell first absorbs incident photons.

If the energy of the photon is greater than the electronic band gap energy of the semiconductor

then the energy of the photon is transferred to the semiconductor which generates electron-hole

pairs. A built in electric field created by a p-n junction is used to separate the electron-hole pair,

and the electrons and holes are collected via external metal contacts of the solar cell device. A

schematic of a p-n junction solar cell device, including light absorption and generation of

electron-hole pairs, is shown in Figure 1-15.147 The p-n junction can be formed by like

semiconductor materials (homojunctions) or unlike materials (heterojunctions). Light absorption

in the solar cell device depends on the absorption coefficient (a), which is an inherent property of

the chosen semiconductor material.

Solar radiation, the raw material for photovoltaic energy conversion, is emitted by the sun

over a range of wavelengths from ultraviolet to infrared. The power density at the surface of the

sun is 62 MW/m2 and the solar flux is reduced to 1335 W/m2 at the earth's atmosphere, mostly

due to the reduced angular range of the sun.141 The power density available on the surface of

earth is reduced to approximately 900 W/m2 due to atmospheric absorption of light. The

extraterrestrial and terrestrial solar spectrums are shown in Figure 1-16. The atmospheric

absorption due to water molecules occurs primarily at 900, 1100, 1400, and 1900 nm and for









CO2 molecules at 1800 and 2600 nm (Figure 1-16).141 The atmospheric absorption is quantified

by using an air mass factor, nAirMass, which is defined as

optical path length to Sun
nAlrMass = cosec (ys) (1-4)
optical path length if Sun directly overhead

where ys is the angle of elevation of the sun. From this definition the extraterrestrial solar

spectrum corresponds to Air Mass 0 or AMO. The AMO spectrum can also be modeled by

assuming black body radiation at 5760K.141 The standard terrestrial solar spectrum is AM1.5,

however, actual solar irradiances varies depending on the season, daily variation of the suns

position, orientation of the earth, and cloud conditions. Average global solar irradiances vary

from below 100 to over 300 W/m2, where the higher values are usually found at inland desert

areas (Figure 1-17). The black disks located on the map represent the area needed to meet

today's total global energy supply, producing 18 terawatts of electricity (TWe), assuming a

modest conversion efficiency of 8% (Figure 1-17).

A solar cell device can be considered a forward biased diode whose current flows in the

opposite direction of the built-in bias. The rectifying behavior of solar cell devices is needed to

separate the carriers that are generated. A solar cells output power is determined by the open

circuit voltage (Vo), short circuit current (Is,), and fill factor (FF). Figure 1-18a shows the I-V

characteristics for a solar cell under dark and illuminated conditions and Figure 1-18b shows the

maximum power rectangle formed by the I-V characteristics where the boxed area is

proportional to the power output of the cell.148 The maximum power, Pmax, is the area of the

maximum power rectangle or the product of maximum power current (Imp) and maximum power

voltage (Vmp), Pmax = Imp*Vmp. The intersection of Imp and Vmp is the maximum power operating

point. The FF describes the squareness of the I-V characteristics and is always less than one.147









Equations for the fill factor and efficiency (r) of a solar cell device is given by Eqs. 1-5 and 1-6,

respectively.

P I, p V
FF max mpVm (1-5)
I V ~ I V
sc oc scVoc

Pma Im V FF IscV
max mp mp (1-6)


Quantities such as FF, q, Is,, and Vo are the main parameters used to characterize a solar cell

device, but these quantities are dependent on the semiconductor structural quality and electronic

properties. Two objectives for obtaining an efficient solar cell involve minimizing

recombination of generated electron-hole pairs and maximizing the absorption of photons.

1.2.2 Why InxGal-xN?

With the discovery of a lower band gap energy of InN,42-49 the wavelength range of

InxGal-xN now spans from infrared (0.7 eV) to ultraviolet (3.4 eV). This wavelength range

covers virtually the entire solar spectrum and creates the opportunity for InxGal-xN based high

efficiency solar cells (Figure 1-19). To exploit the ability of InxGal-xN to absorb light at multiple

wavelengths it has been proposed to fabricate multi-junction tandem solar cells using different

compositions of InxGal-xN (Wladek Walukiewicz, 2002, Full Solar Spectrum Photovoltaic

Material Identified, www.lbl.gov/msd/PIs/Walukiewicz/02/02 8 FullSolar Spectrum.html,

accessed April 2007). Typically most high efficiency tandem solar cells are based on monolithic

In-Ga-As-P-based layers grown on germanium substrates.145 When compared to these systems,

InxGal-xN-based devices have has the ability to be grown on inexpensive silicon substrates and

use fewer and less toxic elements, which make fabrication more flexible and cost effective.

InN has been shown to have a high absorption coefficient, 105 cm-1 (this work), which

means only thin absorber layers are needed to absorb the majority of incoming photons. Thinner









films mean reduced manufacture time and higher throughput, therefore reducing final solar cell

costs. Another benefit of InxGai-xN alloys is that they are radiation hard, which is beneficial for

space PV because it is difficult to replace solar cells in space, so long device lifetimes are

important.14

1.2.3 Current Progress of InN and InxGal-xN Solar Cells

Indium nitride (InN) as a possible photovoltaic (PV) material is a relatively new idea that

was first suggested by Yamamoto et al. in 1994.149 At that time it was suggested that InN be

used as the top cell in a 2-junction tandem solar cell since the band gap energy of InN (1.95 eV)

was close to the optimum value for AMO illumination. Throughout the 1990s little attention was

given to InN as a PV material, instead research efforts focused on single crystal growth of InN

on different substrates using mostly MOCVD and MBE growth techniques.68 Later Malakhov150

proposed a InN/Si heterojunction for a high performance and cost effective solar cell (again

considering a InN band gap energy value of 2.0 eV).

Modeling InxGal-xN tandem solar cell structures has recently been done by Hamzaoui

et al.151 Simulations were done for two, three, four, five, and six InxGal-xN junction tandem

solar cells. The efficiency ranged from 27%, for a two junction cell, to 40% for a six junction

cell. A maximum theoretical efficiency greater than 50% could be achieved using the InxGal-xN

ternary alloy to produce a multi-junction solar cell (using optimum band gaps under

concentration).152 InxGal-xN heterojunctions on p-Si and p-Ge substrates were simulated by Neff

et al. for In compositions ranging from 0.4 to 1.0.153 Best calculated cell efficiencies under

AM1.5 illumination were 18 and 27% for InxGal-xN on p-Ge and p-Si, respectively. Pure InN

heterostructures showed a reduced efficiency of 2.5%.

Although the majority of single crystal InN growth has occurred on sapphire substrates,68

growth on other substrates such as silicon or germanium is preferred for PV applications. Silicon









substrates are desired for their low cost and large area capabilities. Initially single crystal growth

on Si substrates was not possible due to an amorphous SiNx layer forming during growth,68

however, more recent MBE results have produced single crystal InN and InxGal-xN films on

Si( 11) substrates by performing brief silicon substrate nitridation for 3 min.154 These recent

MBE results also demonstrated rectifying characteristics of n-InN/p-Si (on most samples).

Earlier CVD of n-GaxIni-xN/p-Si films also showed rectifying behavior.155

Germanium substrates provide adequate lattice matching to InN (11.3% for

InN(0001)/Ge(1 11)) and also can be used to fabricate vertical conduction PV device designs,

which lead to higher efficiencies than top-connected cells.156 Trybus et al. has recently produced

single crystal InN on Ge(l 11) substrates using MBE,157 and also assessed some aspects of

InN/Ge based PV.156 I-V tests of their n-InN/p-Ge showed only ohmic behavior. In their

assessment it is pointed out that without p-type doping of InN the traditional tandem solar cell

structure must be redesigned, specifically involving the tunnel junctions. Tunnel junctions are

degenerately doped semiconductors that serve as interconnects between the absorber layers of

the solar cells as well as collection areas for minority carriers. In efforts to bypass the need for

p-InN, Trybus et al. suggests using a layer of epitaxial Al between the Ge substrate and InN

layer as a sub-cell interconnect and collecting junction. The Al layer also prevents an In-Ge

eutectic from forming at the interface.156

Another problem that hinders the progress of InN-based PV is the high background n-type

doping. The high intrinsic electron concentration is one of the main issues that makes p-type

doping more difficult. InN has displayed strong band bending at the surface and

heterointerfaces,158-160 which coupled with the high n-type background makes it more difficult to

form rectifying solid-state junctions. Very recently Jones et al. has presented evidence for p-type









doping of InN using Mg acceptors.16 It was shown from capacitance-voltage measurements that

Mg-doped InN contains a bulk p-type region that is electrically suppressed by a thin surface

inversion n-type layer. The surface accumulation layer is believed to be caused by donor-like

surface defects and chemical or physical treatments have been shown to be ineffective at

removing this layer.159,161 Initial p-type doping results are promising for eventual progress

towards InN based devices. With these present device limitations of InN only one reference can

be found where InxGal-xN solar cell devices have been tested.162 Two types of devices were

tested, p-i-n solar cells with i-Ino.o7Ga0.93N layers (using n,p-GaN capping layers) and a quantum

well device with In0.4Ga0.6N layers as the wells. The internal quantum efficiency (IQE) results

for the two p-i-n cells and the quantum well device are listed in Table 1-4. The IQE is the ratio

of the number of collected carriers to the number of photons that enter the cell. The IQE is

related to the collection efficiency but it takes into account losses associated to the limited

thickness of the absorber layer.163 The low IQE for cell 2 was attributed to poor crystalline

quality, implying a large number of defects. X-ray data for cell 1 showed that the indium

composition in the i-layer ranged from 3-16% with a maximum intensity at 7%. Lower

efficiency of the quantum well structure was due to incomplete adsorption of the light due to the

low thickness of the wells (1 nm).

In summary, theoretical predictions have shown that InxGal-xN is a promising absorber

material for high efficiency solar cells. Current progress includes growth on Si and Ge substrates

which are beneficial for PV devices as well as progress towards p-type doping of InN.

Challenges that must be addressed for developing InxGal-xN-based solar cells include successful

p-type doping of In-rich InxGal-xN and understanding the alloy phase separation that occurs

during growth and its effect on cell characteristics. These current device limitations have lead to









the demonstration of InxGal-xN-based solar cell devices with atypical structures, such as thin

quantum wells.

1.3 Terahertz Applications for InN and InxGai-xN

Terahertz (THz = 1012 Hz) is a term used to describe waves with a spectrum ranging from

0.1 to 10 THz, or wavelengths from 3 mm to 30 [tm, respectively. A frequency of 1 THz is

equivalent to wavelengths of 0.3 mm (300 [tm), wavenumber of 33 cm-1, or energy of 4.14 meV.

The THz frequency region is located between the infrared and millimeter wavelengths on the

electromagnetic spectrum (Figure 1-20).164

The THz frequency regime lies between the electronic and photonic domains and therefore

a mixture of optical and electronic mechanisms is often used to generate THz emission.165

Semiconductor photonic devices, which are dominated by inter-band transitions are limited to

the high end of the THz range (- 10s of THz), while today's electronic devices can only reach

frequencies up to a few hundred gigahertz, past which causes circuit failure.164 THz fields are

not well-developed even though the region sits between the well developed (by comparison)

regimes of photonics and microwave technology.166 In recent years, however, THz technology

has made significant developments mostly due to the improvement and availability of

femtosecond lasers.167

THz frequency devices have applications in a variety of fields including astrophysics,

plasma physics, spectroscopy, medical imaging (T-rays), biology, and communications.165 There

is thus considerable motivation to improve the current THz technology, which has several

disadvantages. Current photonic THz emitters are bulky and expensive or must be operated at

cryogenic temperature, as in the case of quantum tunneling lasers.168 THz detectors are also

bulky, expensive, and lack precision.164









InN was established as an ideal material for high frequency terahertz devices once the

theoretically predicted values of the peak electron velocity was shown to be higher than that of

GaAs and GaN, under moderate electric fields.19,169 Other favorable electron transport properties

of InN are large intervalley energy separation, large polar optical phonon energy and a small

effective mass.18'19'170'171

There has been little experimental work analyzing terahertz emission and detection of

InN114,172 and no know work on InxGal-xN thin film alloys. A few researchers have investigated

terahertz emission from InxGal-xN/GaN multi-quantum wells167'173-175 and more researchers have

theoretically investigated THz frequency aspects of InN176-180 and III-nitride

heterstructures. 181-184

Brazis et al.184 performed Monte Carlo simulations of the third harmonic generation (THG)

efficiency of GaAs, InP, InN, and GaN to compare to Si. Their results showed that the III-V

semiconductors predicted THG efficiency exceeds experimental values for Si by two orders of

magnitude. Cooling to liquid nitrogen temperature (77K) increased the THG efficiency. It was

also noted that InP and InN showed superior maximum efficiencies for the materials examined.

Shiktorov et al. 178 examined frequency multiplication by higher order odd harmonic generation

with Monte Carlo simulations and confirmed that InN was a more ideal material than InP or

GaAs.

InN n+nn+ or n+nnn+ structures have been modeled by Monte Carlo simulations.176,177,180

Optical phonon emission was shown to be the dominant scattering mechanism in a n+nn

structure and that a free carrier grating can be formed in the n-region.180 InN n+nnn+ structures

were predicted to emit 50 [tW of microwave power in the 1.1 to 1.2 THz range when connected

to an external circuit and operated at liquid nitrogen temperature.177 Starikov et al.176 showed









that noise enhancement in n nn structures of InN and Ga In,. As can predict the onset of

instability.

O'Leary et al.179 compared the THz emission of bulk InN vs. thin films by using Monte

Carlo simulation. Bulk InN (10 tm) showed possible emission at 10 GHz while 100 nm thick

InN had emission frequencies up to 2.5 THz. Korotyeyev et al.183 also modeled the benefits of

III-N heterostructures vs. bulk materials and attributed the benefit to electron pinning.

Promising experimental results for THz emission have been found for InxGal-xN/GaN

heterostructures173'175 and it was shown that emission efficiency in InxGal-xN/GaN

heterostructures is better than rectification processes in bulk crystals like ZnTe.167 Stanton

et al.174 presented evidence that acoustic phonons could be used to image surfaces and interfaces

in nanostructures from experiments on InxGal-xN/GaN heterostructures.

The first THz emission from InN was demonstrated by Ascazubi et al.114 by optical

excitation of ultrashort radiation pulses from the surface of InN (at room temperature). The

semiconductor was unbiased and excitation was generated by femtosecond Ti:sapphire laser

pulses at 800 nm. THz radiation from InN was compared to that of optimized p-type InAs and

showed radiation on the same order of magnitude. This is a promising result since the InN thin

films contained high defect densities and carrier concentrations (1010 cm-2 and 1018 cm-3,

respectively), and also because InAs is currently one of the best THz semiconductor surface

emitters.185 Optimized InN layers are believed to produce much higher THz emission since

lower carrier concentrations will lead to less free carrier absorption. Meziani et al.172

investigated THz emission from high quality InN epitaxial layers under a range of temperatures

(2-300K) and magnetic fields up to 13 Tesla. Higher THz transmission was noticed for higher

magnetic fields, which was confirmed by simulations, and helicon waves were considered an









important contribution to emission. It was shown that the carrier concentration and momentum

scattering rate, thus the film quality, could be determined by using this contactless method of

THz transmission.

Also, THz emission and detection by GaN HEMTs was recently reported by several

investigators.185-187 These recent results suggest that InxGal-xN alloys are promising materials for

small size, room temperature and high performance THz emitters and detectors.

1.4 Statement of Thesis

Properties such as a high absorption coefficient, radiation hardness, and absorption over a

wide range of wavelengths make InxGal-xN an attractive photovoltaic material (Section 1.2.2).

To better understand InxGal-xN's potential as a PV absorber layer, InxGal-xN-based solar cell

device simulations were performed using the MEDICI device simulation software package

(Chapter 2). Optimized single junction and multi-junction InxGal-xN solar cells were simulated

and the potential effects of phase separation are assessed.

InN and InxGal-xN are materials that show a great deal of potential in applications such as

high efficiency solar cells and terahertz electronic devices, and InxGal-xN has already been

proven to be a dominant material for LED applications. Although significant progress has been

made in the field of InxGal-xN based LEDs, the properties of InN and In-rich InxGal-xN are still

not very well understood. It is for this reason that growth InxGal-xN thin films have been grown

over the entire compositional range (0 < x < 1), so that fundamental issues such as phase

separation can be better understood. For the applications of solar cells, InxGal-xN alloy

composition is very important for the design of high efficiency multi-junction solar cells.

Understanding and optimizing growth conditions for different InxGal-xN compositions is

important for THz applications since THz emission is inversely proportional to doping

concentration due to free carrier absorption, which is natively very high in nitrides. Phase









separation in InxGal-xN alloys might also prove interesting for THz applications because residual

strain affects the internal bias and thus the emission efficiency. The growth of InN and

InxGal-xN thin films is presented in Chapter 3.

The benefits of semiconductor nanostructures were discussed in Section 1.1.3, and these

nanostructures offer unique abilities to understand the properties of InN that cannot be gleaned

from bulk samples. InN and InxGal-xN nanostructures are specifically advantageous for PV and

THz applications due to their large surface to volume ratio. For solar cells a larger surface area

can result in greater collection efficiency and THz emission efficiency should increase with

increased surface area since the emission mechanism results from interfacial interactions. The

growth of InN nanostructures has been presented in the literature more frequently in recent years,

however, growth is typically not highly controllable. There is a need for nanostructured growth

to be incorporated with current growth techniques and process requirements, which is why

controlled MOCVD or H-MOCVD growth of nanostructures is preferred. Growth of InN

nanowires by MOCVD and InN nano- and micro-rods by H-MOCVD are presented in Chapter 4

and 5, respectively.

Chapter 6 contains an exploratory study of GaN nanostructures by MOCVD. This section

is only discussed briefly since GaN nanostructures are not the primary focus of this work,

however, this area is scientifically interesting. Recommendations for future work are presented

in Chapter 7.





















A B C

Figure 1-1. The III-nitride zincblende crystal structure along various directions. A) [100] (1 unit
cell). B) [110] (2 unit cells). C) [111] (2 unit cells) (Ref 2).


Sc


~0


p


B C


Figure 1-2. The III-nitride wurtzite crystal structure along various directions. A) [0001]. B)
[1120]. C) [1010] (Ref 2).


Ga-face


Substrate


N-face


Substrate


Figure 1-3. Cation-faced (Ga) and nitrogen-faced polarity for the GaN wurtzite crystal structure
(Ref. 5).
















7






1o"


1I 100


1000


Electric Field [kV/cmj

Figure 1-4. Velocity-field characteristics (T = 300K, n = 1017 cm-3) for wurtzite InN, GaN, A1N,
and zincblende GaAs (Ref 19).

TEmperatura T I C1
U7 2210 1ce 0o6 70 wa 440 3Ms

= 2146 K
i I
2

10_






1J Ok J 7 K \


2 A a a 10 12 14 M
'ell W)


Figure 1-5. Vapor pressure of N2 in equilibrium with of InN,
temperature (Ref. 63).


GaN, and A1N as a function of










(C)
cUoo,~ I


(m)
tioToi


Figure 1-6. Rhombohedral structure and surface planes of sapphire (Ref 5).


A


rapornowre
m ital albly
caalyst liquid



I U Ii


Figure 1-7. Theoretical and experimental examples of nanowire growth mechanisms.
A) Theoretical VLS growth mechanism (Ref 107). B) Experimental evidence for a
VLS mechanism showing metal catalyst at tip of nanowire (Ref. 92).
C) Experimental evidence for a VS mechanism, showing no metal droplet (Ref 103).











1200-
1000o

BOO


400

200

0


0 0.2
GaN


0.4 0.6 0.8
Xn


Figure 1-8. Predicted binodal (solid) and spinodal (dashed) decomposition curves for InxGal-xN
assuming regular solution mixing (Ref. 123).


1a00 A Binodal

1200 Spinrodal

800 nSan wi
I nGaN
400 relaxed

1r,_


C.O 0.2 0.4 0.6

GaN Solid phase
composition


08 1.0
InN


1200 I
l oo B
1000 B
800 Wurtzle
800 Intr-face /
pepen dlcular /
400 toC-axs I/
200 / t

C.0 0.2 0.4 0.6

Ga N Solid phase
composition


Figure 1-9. The T-x phase diagrams of ternary InxGal-xN compounds. A) Relaxed InxGal-xN
layers. B) Strained InxGal-xN layers with the interface orientation perpendicular to
the hexagonal axis of the crystal (Ref 124).


IJ I If I


'-

rGaN
rained
I i
0I8 1.0
InN



































Figure 1-10. Several
(Ref. 12).


21 22 .33 34 35 36
DIFFRACTION ANGLE 20(deg.)


XRD spectra of phase separated InxGal-xN films at different temperatures


O




aI II
z -









Z 02 D. h 0.B 1.00


TMI
TMI.TEG


Figure 1-11. Compositional control of InxGal-xN with respect to growth temperature (Ref. 12).












1nnnn1


250000


CO I1UUUUU ---------------T -- -- H -- a *e



50000



0
31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec. 31 Dec.
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003


Figure 1-12. U.S. photovoltaic module implementation from 1992-2003 (Ref. 144).


a.
-o


SUnited States
__ Japan
SEurope
D Rest of World



















199 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2W92 2013 2004


Figure 1-13. Comparison of world wide growth of PV production from 1990-2004 (Ref. 143).


E grid-connected centralized
* grid-connected distributed
S off-grid non-domestic
0 off-grid domestic


E.uuu +


UUUU '


200











Wortd Markt 2D01


by Technology
a Fr -1
n@'m M&


Siiue Cryalal &
$5.17%


Figure 1-14. Market shares of different photovoltaic materials as of 2001 (Ref 145).


I | s ,,,SL i h i


Metal grid


n-type layer


fit e

p-type layer


h+ "



Metal contact


Antireflective layer


Figure 1-15. Schematic of light induced carrier generation in a generic solar cell (Ref 147).


PPON~p lawa~
4Y.~Q4i

















1.75

4 II- Terrestrial (AM1.5)
E 1.50

E
1.25
U




t; 0.75


0.50 -


0.25


0.00
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750
Wavelength (nm)


Figure 1-16. Solar irradiance vs. wavelength for extraterrestrial (AMO) and terrestrial (AM1.5)
spectra (American Society for Testing and Materials, ASTM, Reference Spectra for
Photovoltaic Performance Evaluation, ASTM G-173-03, 2003,
http://rredc.nrel.gov/solar/spectra/, accessed April 2007).















-~ 'A a
^^^ ^^TT^~
-z ,. --'
.r. .- -I I


-~P---if~I~C

34 -Jl


I
t,
1\


-'~


Marlrias Loster 2006


... ........... ::EEEE .

< .
Sii0 150 200 250 300 35 Wm


0 50 100 150 200 250 300 350 W/mi


Z- 18 TWe


Figure 1-17. Global solar irradiances averaged over three years (1991-1993) which account for
cloud cover (Matthias Loster, 2006, Total Primary Energy Supply: Land Area
Requirements, http://www.ez2c.de/ml/solar landarea/, accessed April 2007).


Curant, I


Current


Vrmp Vc0


Valloge


Figure 1-18. Current-voltage characteristics of a generic solar cell. A) under dark (solid line)
and illuminated (dashed line) conditions. B) The resulting maximum power rectangle
(under illumination) (Ref. 151).
















)AM1.5
I I lkI -AM1-

E












0.0 0.5 1.0 1.5 2.0 25 10 5 4.0 4.5

InN Bandgap(eV) GaN



Figure 1-19. Incident solar flux for AMO (black) and AM1.5 (red) as well as the InxGal-xN band
gap energy range (colored box).


S- 14--- *- -oi.I -I@- --.10 -.
Microwave Terahertz Mid-infrared Near-
..... infrared
--I I "


Molecular ratalona (gai


0 '
04'


0.01
0.01


0.1
0.1


Lowtrrequelyo
bond vibratons
I


CrytatlneHW phonpn
VIbrationa (molid


HVdrogmrborndilg trl
and tarlonae (gn nd Il




1 1
frequency / THz
10 1
wavenumber / cm


Overlones and
combinallon bands





-4
Iches
*ulds,

High frequency bond
lrbrallons


100


1000


Figure 1-20. Terahertz frequency range (shaded) of the electromagnetic spectrum, including
molecular transitions (Ref 168).


I


r
i
3


II









Table 1-1. Some fundamental properties of InN and GaN (Ref. 50a, 188, 189b, 190).
Property GaN (wurtzite) InN (wurtzite)
Room temperature band gap energy (eV) 3.39 0.6-2.1a
Temperature coefficient (eV/K) dEg/dT = -6.0 x 10-4 dEg/dT = -1.8 x 10-4
Lattice constants (A) a =3.189 a =3.5377b
c =5.185 c = 5.7037b
Thermal expansion (K 1) Aa/a = 5.59 x 10-6 Aa/a = 4 x 10-6
Ac/c = 3.17 x 10-6 Ac/c = 3 x 10-6
Thermal conductivity (W/cm-K) 1.3 0.8 + 0.2
Index of refraction 2.67 2.9-3.05
GaN (zincblende) InN (zincblende)
Room temperature band gap energy (eV) 3.2-3.3 0.6c
Lattice constants (A) a = 4.52 a = 4.986c

Table 1-2. Theoretical and experimental mobility and controlled doping ranges for InN and
GaN.
Property InN Ref. GaN Ref

Theoretical maximum electron
mobility (cm2 V-s)
300K 4,400 57 1,350 192
77K 30,000 19,200
Experimental maximum hole
mobility (cm2 V-s)
300K no available data 13 193
Controlled doping range (cm-3)
n-type (Si) 5x1016 to 5x1020 194 1016 to 4x1020
16 t 195
p-type (Mg) no available data 1016 to 6xl 0 _8










Table 1-3. Lattice and thermal expansion constants for InN substrates and buffer layers (Ref 5,
196).
Substrate material or Space group Thermal expansion
lattice parameters c
buffer Layer symmetry coefficients
a c
a(A) c(A) (106 K-1) (106 K-1)

GaN P63mc 3.189 5.185 5.59 3.17
A1N P63mc 3.112 4.982 4.20 5.30
c-A1203 R3c 4.759 12.991 7.30 8.50
Si(111) Fd3m 5.430 ------ 3.59
6H-SiC P63mc 3.080 15.117 4.46 4.16
GaAs F43m 5.653 ------ 6.00
InP F43m 5.869 ------ 4.50
GaP F43m 5.451 ------ 4.65
ZnO P63mc 3.252 5.213 2.90 4.75
MgO Fm3m 4.216 ------ 10.50
MgAl204 Fd3m 8.083 ------ 7.45
LiA102 P41212 5.169 6.268 15.00 7.10
LiG 2 P2 5.406 5.013 6.00 7.00
LiGaO2 Pna21
b = 6.3786 b= 9.00


Table 1-4. Internal quantum
(Ref. 162).


efficiency (IQE) of InxGal-xN p-i-n and quantum well solar cells


Cell # Type InxGal-xN (x) IQE (%)
1 p-i-n 0.07 19.0
2 p-i-n 0.40 1.0

3 quantum well 0.40 8.0


J









CHAPTER 2
INDIUM GALLIUM NITRIDE SOLAR CELL DEVICE SIMULATIONS

2.1 Introduction

InxGal-xN is an optimal material for absorber layers in solar cells because the InN-GaN

alloy system covers a wide spectral range that spans the majority of wavelengths in the solar

spectrum. In-rich InxGal-xN has a high absorption coefficient and has also been proven to be

highly radiation resistant.14 A more detailed explanation of InxGal-xN's material qualifications

for PV applications is given in Section 1.2.2. Development of InxGal-xN based photovoltaics has

been delayed by the difficulty in p-type doping of InN and In-rich InxGal-xN alloys. The recent

results for Mg doped InN, however, are encouraging for eventual p-type doping of InN and thus

In-rich InxGal-xN alloys.16

In this chapter, single and multi-junction InxGal-xN solar cells are modeled using MEDICI

device simulation software. The optimum cell parameters are determined for a single-junction

InxGal-xN solar cell and multi-junction solar cells are simulated for two, three, four, and five

junctions. The implications of phase separation and the effect on cell efficiencies are also

discussed for both single and multi-junction InxGal-xN solar cells.

The intention of this simulation study is to estimate the efficiency of InxGal-xN solar cells

by applying realistic limits (thickness, doping levels, band gap energy, etc) to the cell layers and

to use the most recent material properties from the available literature. No p-n junction

InxGal-xN solar cell has been demonstrated to date therefore it is not possible to match the

predictions to real devices. Applying realistic cell parameters is currently the best method to

measure the possible potential of InxGal-xN solar cells.









2.2 MEDICI Device Simulation Software

Medici is a 2-dimensional device simulator that models the electrical, thermal, and optical

properties of semiconductor devices. The Taurus-Medici program is part of the Technology

Computer-Aided Design (TCAD) software distributed by the Synopsys Corporation. Medici is

frequently used to model MOS and BJT devices. A 2D non-uniform mesh grid is used that can

model planar or nonplanar surface features. For each grid section Medici solves Poisson's

equations as well as the electron and hole continuity equations and predicts the two-dimensional

distributions of potential and carrier concentrations at arbitrary bias conditions. The program can

model several phenomenons such as recombination, photogeneration, impact ionization, band

gap narrowing, band-to-band tunneling, mobility, and carrier lifetimes (Synopsis TCAD Medici

user manual version 2001.4).

2.3 Identification of InxGal-xN Solar Cell Parameters

Material Properties. Table 2-1 lists the input parameter values for InN and GaN which

are used in the Medici simulation. The appropriate references are listed for each input parameter

and if no values are available in the literature then a realistic value was estimated or assumed

(values in italics). For example, no definitive p-type doping has been demonstrated for InN;

therefore no hole mobility values have been published. It can, however, be estimated because

mobility ([t) of a semiconductor is inversely proportional to the effective mass (m*) (Eq. 2-1),197


/ =q (2-1)
m

where is the average time between collisions and q is the unit of charge. The electron

effective mass of GaN (0.22mo)198 is larger than the effective mass of InN (0.07mo),41 where mo

is the electron rest mass (9.11 x 10-31 kg). Prior to the InN band gap controversy, theoretical

calculations were performed that showed the hole effective mass of InN was approximately the









same or less than the values for GaN (for light hole and heavy holes).199 It has been suggested

that the hole effective mass be revised (and decreased) due to the commonly accepted lower InN

band gap energy value198. From the available theoretical calculations it is safe to assume that the

hole mobility of InN is the same or slightly greater than the value obtained experimentally for

GaN (200 cm2/V-s).200 For the simulations presented here a hole mobility of 200 cm2/V-s was

chosen for InN, although the real value is likely higher. A wide range of electron mobility data

has been published for InN which ranges from 200 to 2700 cm2/V-s depending on the type of

deposition technique.31'41'42,58'201'202 High quality InN layers grown by MOCVD have electron

mobilities close to 1000 cm2/V-s, therefore this was the value used for the Medici simulations.

Optimization for the growth of InN thin films on c-A1203 by MOCVD, can be found in

Ref 203. High quality single crystal InN films were grown on c-A1203 substrates using substrate

nitridation and a low temperature InN buffer layer, resulting in a XRD FWHM of 1339 arcsec.203

The absorption coefficient (a) of this high quality InN film was measured (Figure 2-1). The

absorption of InN ranges from 7 x 104 to 1.5 x 105 cm-1 in the incident energy range of 0.7 to

3.4 eV with an average value of 9 x 10-4 cm-1 (up to the band gap energy of GaN). The

absorption dropped quickly to zero at an approximate band gap energy value of 0.65 eV (not

shown in Figure 2-1). For the absorption measurements, an incandescent lamp inside a quartz

tube was used as the illumination source and a diffraction grating was used to filter light into

different spectral ranges. A Hitachi 330 spectrophotometer was used to detect the transmitted

light. The absorption coefficient of GaN was measured by Muth et al.204 with an average value

of 1 x 105 cm-1. For simplicity a common value for the absorption coefficient of 9.5 x 104 cm-1

was used for all InxGal-xN compositions.









The properties in Table 2-1 were varied as a linear function based on InxGal-xN

composition, with the exception of the band gap energy. A variable band gap energy bowing

parameter is used that has been determined by fitting data to several sets of InxGal-xN alloys.62

This bowing parameter is believed to be the most accurate because other bowing parameters are

usually determined from limited composition ranges (i.e. either Ga-rich or In-rich compositions

only). The equations determining the band gap energy and variable bowing parameter were

previously shown (Eqs. 1-2 and 1-3).

Illumination sources. Terrestrial (AM1.5) and extraterrestrial (AMO) solar spectra were

used as illumination sources for the InxGal-xN cells. The AM1.5 spectrum was simulated by

using spectral irradiance values (as a function of wavelength) from the American Society for

Testing and Materials reference spectrum (G-173-03). The AMO spectrum was modeled by the

Medici software by assuming black body radiation at a temperature of 5800K (Eq. 2-2).


AI = Rs 27Zc2 AA (2-2)

exp A-1
kT

where Rs is the radius of the sun (7 x 108 m), RSE is the distance between the sun and the earth

(1.5 x 1011 m), T is the temperature of the black body, and h is the wavelength.

Recombination Models. The "CONSRH" and "AUGER" Medici recombination models

are used to describe the recombination in the InxGal-xN layers. CONSHR refers to a

concentration dependent Shockley-Read-Hall model, where the carrier concentration determines

the carrier lifetimes and the probability for recombination and generation. The Shockley-Read-

Hall model was developed in 1952 and used to describe the recombination and generation

statistics for holes and electrons by using a distribution of trap states in the forbidden gap as well

as kinetic transport models for electrons and holes.205207 With the CONSHR model, Medici









determines a probability for recombination through trap states by predicting a distribution of trap

states within the band gap. The AUGER model refers to Auger recombination, which is band to

band recombination that occurs when two like carriers collide. When two electrons collide in the

conduction band one carrier loses its energy and is recombined with a hole in the valence band,

while the surviving carrier receives the energy released by recombination and becomes a higher

energy electron. The excited electron subsequently loses its energy to lattice vibrations through

collisions with the semiconductor lattice.197 Other than these two defect models, no specific

defect levels or interface states were used in the simulation.

2.4 Results

2.4.1 Single-Junction InxGal-xN Cell Optimization

Cell parameters such as n-side thickness, p-side thickness, doping concentrations, and band

gap energy were optimized to obtain the maximum efficiency of a single InxGal-xN p-n junction

solar cell. The cell parameters were limited by design rules typically assigned to thin film

single-junction solar cells. A fundamental design rule of thin film solar cells is that the thickness

of the solar cell must be on the order of a few microns. The motivation for thin film solar cells is

to obtain the best cell efficiency while keeping layers thin to reduce processing time and the cost

of fabrication. Thin film silicon solar cell technologies are approximately 10 [tm thick compared

to the bulk cells which are 250 [tm thick. When the cell layer thickness decreases light trapping

techniques such as surface roughing and texturing are important to enhance absorption in the

cell.208 This is done for Si solar cells because the absorption coefficient is low compared to other

photovoltaic materials (Figure 2-2). More recently investigated photovoltaic materials such as

CuInSe2 (CIS) have a higher absorption coefficient over a broad range of wavelengths (Figure

2-2). Absorber layers fabricated with CIS then have lower thicknesses (-2 m).147 Since









InxGal-xN has a similar absorption coefficient the same maximum thickness design rule

(approximately 2-3 [tm thick absorber layers) was applied.

Typically p-n junction solar cells employ asymmetric doping between the p-type and

n-type layers, which also differ in thickness. Generally the thickest layer (- 2 [tm) is lowly

doped (1017 cm-3) while the thinner layer is highly doped (1020 cm-3).197 It is also preferred that

the thicker layer be p-type because of the lower mobility holes. Since minority carriers are

collected instead of majority carriers, it is preferred to have higher mobility electrons traveling

through the thickest part of the solar cell, thus requiring a thick p-type region. This is beneficial

because a higher mobility means carriers are more likely to be collected before recombination

can occur. The motivation for the asymmetric doping and thickness structure is to minimize the

light absorption by free carriers while still producing a large depletion region to minimize

diffusion lengths (available time for recombination) and increase charge carrier separation. Free

carrier absorption reduces efficiency because this absorption does not generate electron hole

pairs, but instead promotes carriers in the conduction and valence bands to higher energies,

which is then lost in the form of heat.209

It is assumed for InxGal-xN that p-type doping will be achieved though it is currently very

difficult. For this work the acceptable p-type doping range is limited to 1 x 1016 to 1 x 1017 cm3.

This doping range was chosen because similar controlled doping ranges have been exhibited for

InP MOCVD.210 A comparison with InP must be made since no doping levels have been

measured for p-type InN. InP is a III-V semiconductor with a band gap energy (1.34 eV) that is

close to InN's band gap energy (0.7-1.0 eV). The controllable n-type doping range of InN has

been shown to be 1 x 1018 to 1 x 1020 cm-3.194,211 From these doping characteristics it is ideal to

use a low doped thick p-type InxGal-xN layer with a highly doped thin n-type InxGal-xN layer for









the structure of the solar cell (Figure 2-3). The anti-reflecting (AR) coating is used to prevent

light reflection from the top surface of the semiconductor. Electrical connections to the cell are

made by the back and front contacts, where the back contact could be metal contact to a p-type

silicon substrate and the front contact is also a metal contact deposited in a finger arrangement to

maximize light transmission. For the Medici simulation no exact substrate properties (such as

those for Si) were included for the calculations, instead a generic contact resistance was used.

The resistance for the back and top contacts are modeled by including a contact resistance of 2 x

10-6 (Medici user manual version 2001.4) and 1 x 10-4 Q-cm2,212 respectively.

All single-junction simulations were p-InxGai-xN/n-InxGai-xN junctions and were modeled

by ID Medici simulations. Arbitrary doping and thickness conditions were initially set to

determine the optimal band gap energy for the maximum single-junction cell efficiency. The

thickness and doping level of the p- and n-type layers were then optimized, and then the band

gap energy value was re-evaluated in an iterative process. When band gap energy values were

varied during optimization the same band gap energy value was used for both p- and n-type

layers. During all optimization steps only one variable (such as n-side thickness, p-side doping

concentration, etc.) was varied while all other parameters remained fixed.

The initial absorber structure (Figure 2-4a) consisted of a homoepitaxial stack of a 300 nm

n-type InxGal-xN layer with a doping concentration of 1019 cm-3 and a 2 [tm thick p-type

InxGal-xN layer with a doping concentration of 1016 cm-3. The efficiency calculations for the

various steps of the optimization procedure are plotted in Figure 2-5. The efficiency of the initial

structure was low with an efficiency of about 4%. As previously shown in Eq. 1-6 the efficiency

is the maximum power produced by the solar cell divided by the inlet power (incident energy

from the simulated solar spectrum). The dominant factor for the low efficiency of this cell is the









thickness of the highly doped n-type layer that gives strong free carrier absorption and

recombination in this layer. The band gap energy of InxGal-xN for the initial structure was varied

from 1.1 to 1.6 eV (Figure 2-5, represented by blue diamond symbols) and the maximum

efficiency occurred at a band gap energy of 1.5 eV, which is close to the optimal ideal band gap

energy of approximately 1.4 eV.145

In the next stage of optimization, the thickness of the n-type layer was varied (at a constant

band gap energy value of 1.5 eV). A reduction from 300 to 25 nm produced a significant

increase in the efficiency due to reduced free carrier absorption and carrier scattering (square

symbols in Figure 2-5). A further increase in efficiency is obtained when a linear graded carrier

concentration in the n-type layer is used instead of a uniform doping profile (triangle symbols in

Figure 2-5).

The p-side doping concentration was varied from 1016 cm-3 to 1017 cm-3 and the best cell

efficiencies were obtained at a doping concentration 1017 cm-3. The doping concentration of the

p-type layer did not exceed 1017 cm-3 due to the previously assumed design rules. Finally the

absorption coefficient of InxGal-xN is sufficiently high that increasing the p-side thickness will

not provide any significant improvement in efficiency, therefore the thickness of the p-side layer

was decreased to determine the reduction in efficiency. When the p-side thickness is reduced

from 2 to 1.5 to 1 [m the change in efficiency with respect to the 2 [tm thickness is -0.21% (1.5

tm) and -0.62% (1 tm). For the purposes of this study a p-side thickness of 2 tm was used in

order to obtain the maximum efficiencies possible, however it can clearly be seen that material

processing times can be significantly reduced (for absorber layer growth) with only a minimal

loss in efficiency. Also, adding thickness to the p-side layer will only increase the series

resistance, leading to a decrease in voltage.









The maximum efficiency band gap energy was re-evaluated using the refined absorber

layer structure (Figure 2-4b). A plot of the cell efficiency versus band gap energy for both AMO

and AM1.5 illumination is shown in Figure 2-6. The maximum calculated efficiency was

obtained at a band gap energy of 1.44 eV for both AMO and AM1.5 illumination. Plots of the

cell power versus load and J-V curves for AMO and AM1.5 are shown in Figures 2-7 and 2-8,

respectively, and the cell characteristics (Js, Voc, maximum power, FF, and collection f) are

listed in Table 2-2. Both illumination conditions show similar values for the fill factor (86%)

and the efficiency is slightly higher for the AM1.5 (16% compared to 15.3% for AMO). The

highest collection efficiency is obtained for the intermediate band gap energy values tested and

that the efficiency decreases for large or small band gap energy values (Figure 2-6). As the band

gap energy increases the voltage of the solar cell will increase, however the solar flux decreases

at these higher band gap energies (Figure 1-19). As the band gap energy of the solar cell

increases all the light energy below the cell band gap energy cannot be collected. This creates a

trade off between band gap energy and the available light energy for collection, which is why a

maximum efficiency is obtained at a band gap energy value of 1.4-1.5 eV.

It is difficult to assess the validity of these simulations because no p-n InxGal-xN solar cells

have been fabricated. The InxGal-xN simulations presented here were compared to copper

indium gallium diselenide (CIGS) Medici simulations that use the properties that are measured

for real CIGS solar cell devices, such as absorption coefficients, layer thicknesses, and doping

concentrations.213 Comparing these two materials is ideal because both materials are thin film

absorber structures with similar absorption coefficients. Both simulated solar cells were

assigned a band gap energy of 1.24 eV and illuminated with AMO light. At this band gap energy

value the efficiency of the CIGS devices was 16% (FF = 73.8%) and the InxGal-xN cell had an









efficiency of 14.6% (FF = 83.8%). Current laboratory best efficiencies for CIGS are 19.5%

while larger scale module efficiency range from 13 to 15%.214 The Medici solar simulator in

conjunction with the recombination models used produces cell efficiencies in between the best

laboratory cells and module efficiencies. The difference in FF was attributed to a CdS buffer

layer and ZnO transparent conducting oxide (TCO) layer, which was not present in the InxGal-xN

structure. No p-n InxGal-xN solar cells have been fabricated to date, however, this comparison

was used to validate the model used for the Medici InxGal-xN simulations. As previously

mentioned the InxGal-xN cell efficiencies could be improved by increasing the level of p-type

doping which would provide similar or higher efficiency values as CIGS. However, the doping

levels were capped at lower levels to represent anticipated near future InxGal-xN p-type doping

capabilities. The same recombination models and contact resistances were used for the CIGS

simulation for accuracy in the comparison.

2.4.2 Multi-Junction InxGal-xN Solar Cells

Multi-junction InxGal-xN solar cells were next simulated using the refined single-junction

solar cell structure previously described as a starting point. Two, three, four, and five solar cell

junctions were modeled. The p- and n-side thickness of each p-n junction remained fixed at 2

[tm and 40 nm, respectively, for each absorber layer simulated. It was previously assumed that

the achievable levels of p-type doping for InxGal-xN are much below degenerate doping levels.

For this reason the solar cell junctions must be connected in a mechanical stack arrangement

instead of monolithically using tunnel junctions. It is possible, however, that an alternative

material system could be used for the tunnel junction. Each layer of the mechanical stack multi-

junction cell is simulated separately thus requiring two connections per cell in the stack.

The efficiency of each multi-junction solar cell is simply the summation of each individual

junction within the cell, with only the spectrum of the incident radiation changing, depending on









the characteristics of the cell above it. The same approach is used for determining the cell open

circuit voltage. A plot of the solar cell efficiency and open circuit voltage is shown in Figure 2-9

as a function junction number. The optimized p-n junction band gap energy values for each

multi-junction cell are shown in Table 2-4. The efficiency increases from 15.4% for a

single-junction cell to 27.4% for a five junction solar cell (Figure 2-9). These solar cell

efficiencies are slightly less than current III-V space solar cells produced by Spectrolab based

on GaInP2/GaAs/Ge cells, which have efficiencies up to 28.3%, which use three cell junctions

(Spectrolab Inc, Ultra Triple Junction Space Solar Cell data sheets,

www.spectrolab.com/prd/prd.htm, accessed April 2007).

For these simulations conservative estimates were made for the structure of the p-n

junction absorber so it is not surprising that the efficiencies are lower. It was also assumed that

no light greater than the band gap energy of a specific junction was allowed to transmit to

subsequent junctions. Other researchers have simulated InxGal-xN solar cells by applying a less

constrictive model to the cell parameters and have predicted efficiencies of 39% for a five

junction monolithic cell under AM1.5 illumination.151 Therefore, InxGal-xN multi-junction solar

cell absorbers have been predicted to reach similar efficiencies as the current best cells in

production even after assigning a constrictive model to the simulations. There is also potential

for obtaining much higher efficiencies if InxGal-xN absorber layers can be grown with superior

material properties.

2.4.3 Phase Separation in InxGal-xN Solar Cells

2.4.3.1 Effect on single-junction cells

Since phase separation is known to occur in InxGal-xN alloys it is important to assess how

solar cell characteristics might be affected. Upon initial inspection it is not clear how phase

separation will affect cell characteristics because there are several carrier dynamics to consider









when a solar cell is illuminated. For example, consider a uniform direct band gap energy

semiconductor p-n junction that is connected to an external circuit. Electron-hole pairs are

created when the p-n junction is illuminated with photon energy greater than the band gap energy

of the semiconductor (Figure 2-10). When photons with energy much greater than the band gap

energy are absorbed then electrons can be promoted high into the conduction band (Figure 2-10).

The excess absorbed energy can be transferred to either of the two bands (or both), but typically

the majority of the excess energy is transferred to the band with the smaller carrier effective

mass.215 The p-type InxGal-xN region is the bulk of the solar cell absorber and the electrons have

the smallest effective mass, therefore the majority of the energy is transferred to the conduction

band (Figure 2-10). The energetic electron is called a hot electron because the electron

temperature (Te) is higher than the semiconductor lattice temperature (To). Typically this

energetic electron loses its energy through a series of phonon scattering events within the

semiconductor lattice, where the electron relaxes to the band edge with an electron temperature

To. The time constant (z) for hot electron relaxation to the band edge is on the order of

picoseconds216'217 while band to band recombination is on the order of microseconds.207

Now consider an InxGal-xN p-n junction which contains phase segregated regions of

InyGalyN where x > y. This structure forms an In-rich InxGal-xN matrix with Ga-rich

precipitates (Figure 2-11). A representative band diagram is shown in Figure 2-12 for the matrix

and precipitate in the p-type region of the p-n junction. A "type I" band offset applies to

InN/GaN interface where the valence and conduction band edges of InN lie within the band

edges of GaN.218 There is a greater band gap discontinuity in the conduction band (AE, 1.7

eV) than the valence band (AEv 1 eV), however there is still some debate about the exact offset

values.42,122,219,220 Minority carriers (electrons) in the phase-separated structure that are









generated under illumination will be accelerated towards the n-type region, assuming that this

region is within the depletion region of the junction. This type of phase-separated structure will

be beneficial for the generation of hot carriers (compared to carriers generated in the matrix)

since the band gap energy of the precipitates are greater than the matrix. If it is assumed that the

hot carriers generated in the precipitates quickly relax to the band edges of the matrix before

being collected by external contacts then the efficiency of the solar cell will simply take on the

electrical characteristics of the matrix material. The efficiency of a phase-separated solar cell

might be slightly reduced due to an increased transit time around the wider band gap energy

regions, resulting in a higher probability of recombination. Regardless of a slight decrease in

efficiency a single-junction phase-separated solar cell should closely model a uniform solar with

the same band gap energy and doping characteristics as the matrix material in the

phase-separated solar cell. This predicted result is based on the assumption that all hot carriers

are relaxed before being collected.

If it is assumed that hot carriers are allowed to remain hot and this energy can be collected

without relaxing to the band edge of the matrix then the efficiency of a phase-separated solar cell

can be increased compared to a uniform single-junction solar cell. For this situation a device

structure was assumed to contain an In-rich InxGal-xN matrix with a band gap energy of 1.0 eV

and precipitate phase with a band gap energy of 2.0 eV. The overall thickness of the cell was

approximately 2 [tm and the same doping profiles were used as the previously optimized

single-junction cell. The precipitate volume fractions were varied from 0.1 to 0.25 for the

phase-separated InxGal-xN solar cells, assuming a uniform distribution of precipitates. The

matrix and precipitate band gap energy values were assigned because experimental compositions

(from XRD) corresponding to these band gap energies were seen for phase-separated InxGal-xN









thin film samples. No experimental evidence has been determined for the volume fraction of the

precipitates therefore these values were assumed. For these simulations hot carriers were

allowed to remain hot and be collected at their hot energy and it was assumed that no increased

recombination occurred due to increased transit time around the wide band gap energy

precipitates. It was also assumed that hot carriers were only contributed from the wider band gap

energy phase-separated regions. As expected the maximum efficiency occurred at the highest

volume fraction of precipitates because more hot carriers were produced and the cell

characteristics are shown in Table 2-3. Illumination from both AMO and AM1.5 produced

efficiencies of 19% for a precipitate volume fraction of 0.25. This is an impressive result for a

single-junction solar cell since the efficiency increased 3% compared to a uniform structure,

however this result assumes that the hot carriers do not relax from their elevated state.

In certain cases it is safe to assume that some carriers can remain hot because the energy

bands are capable of sustaining the extra energy. For example in semiconductors majority

carriers can relax back to the band edge while minority carriers remain hot.221 This situation can

be appropriately applied to the 2 [tm p-type InxGal-xN absorber layer, where the majority of light

absorption takes place. As previously mentioned, the majority of energy is transferred to the

band with the smaller effective mass which is the conduction band and the electrons are also the

minority carriers in the p-type region. These conditions are ideal for maintaining hot electrons in

the solar cell. However, even if the minority carrier electrons can remain hot significant phonon

scattering will likely occur at the metal contact and cause the hot electrons to relax to the band

edge. Other researchers have suggested wide band gap energy semiconductors with very narrow

band widths (<< kTo) as contacts, which would allow for isentropic cooling of hot electrons.222

The voltage for this type of cell would be determined by the summation of the free energy









difference between each hot electron hole pair. This is an attractive approach, however, not

practical for real devices due to the assumptions that must be made for the isentropic contacts.

Another potential increase in cell efficiency that can be applied to phase-separated solar

cells is the generation of more electron hole pairs through impact ionization. Impact ionization

occurs when an electron is accelerated to the point where it collides with an atom in the lattice

and the energy transfer frees a bound electron, creating an electron hole pair.197 When high

electric fields are present impact ionization leads to carrier multiplication (avalanching) that can

cause semiconductor devices to fail (breakdown). This effect would be beneficial for solar cells

because the increased carrier generation would lead to an increased current. An impact

ionization model from the Medici simulation program was applied to the same phase-separated

solar cell structure used for modeling the hot electrons. The results revealed that there was little

to no change in the device efficiency due to impact ionization. Since impact ionization occurs

most frequently at high electric fields it is likely that the built in bias in the solar cell device is

too low to accelerate carriers enough to ionize atoms in the lattice.

2.4.3.2 Effect on multi-junction solar cells

A phase-separated solar cell structure could potentially have negative effects on the

efficiency of a multi-junction absorber structure. A lower efficiency is possible as a result of the

reduced absorption of light in the precipitate regions of bottom cells because high energy

wavelengths will have already been absorbed. This will essentially create optically transparent

regions which will not produce carriers. Efficiency reduction due to phase separation will

become more pronounced when the number of junctions is increased because the band gap

energy values are precisely tuned for certain wavelengths. If these light wavelengths are

absorbed before reaching the appropriately designed cell then inactive areas (dead volumes) are

produced in the cell.









To show how a two junction solar cell can be unaffected by phase separation consider a top

absorber layer with a matrix band gap energy of 1.8 eV (with a higher band gap energy

precipitate phase) and a bottom absorber layer with a matrix band gap energy of 1.0 eV. As long

as the precipitate band gap energy in the bottom cell has a band gap energy below 1.8 eV then

the cell characteristics will be unaffected. The negative effects of phase separation in multi-

junction solar cells can be avoided by making cell layers sufficiently thick to compensate for any

inactive area in the absorber layers. Previous single-junction results show that there is no

significant change in absorber efficiency when the p-side thickness is reduced from 2.0 to

1.0 pm. This indicates that the majority of the absorption takes place in the first pm of the p-type

layer and that the additional micron just assures that all of the light is absorbed. If it is assumed

that 30% (by volume) of a phase-separated solar cell is made up of wide band gap energy

(optically inactive) areas, then absorption will be limited to 70% of the solar cell material. If a

2 pm absorber layer thickness is used then there will be approximately 1.4 pm of active area.

This thickness should be sufficient for absorbing the majority of incident light on the solar cell

and the only expected reduction in efficiency is attributed to increased recombination, due to

longer transit times for carriers around the inactive regions.

Wider band gap energy junction layers are important for obtaining high efficiency multi-

junction solar cells. For example the largest band gap energy of the five junction solar cell

previously simulated was 2.6 eV. To achieve these wider band gap energies more gallium must

be incorporated into InxGal-xN. When the alloy composition becomes Ga-rich then it is likely

that In-rich segregated phases will form, which will have a smaller band gap energy then the

Ga-rich matrix. These smaller band gap energy precipitates will trap carriers, leading to more

recombination and a reduction in collection efficiency (Figure 2-13). Recombination in the









narrow gap precipitate region can lead to light emission which can be re-absorbed by junctions

further down the cell. However, not all recombination will lead to the production of light and

this is not an efficient way to collect the light energy because the solar cell takes high energy

light and converts it to a reduced amount of lower energy light. Therefore phase separation in

the form of smaller band gap energy precipitates in a wide band gap energy matrix will lead to a

reduced efficiency. It is important to mention that the probability of phase separation is reduced

as the In composition decreases far below 50%, therefore the negative effects of phase separation

for a multi-junction solar cell is limited to a small compositional range (0.5 < x < 0.3).

2.5 Conclusions

In this solar cell modeling approach, conservative cell properties were assigned to absorber

layers to anticipate efficiency of InxGal-xN-based solar cells. The main assumption of this model

is that low p-type doping capabilities will be achieved for In-rich InxGal-xN. Single-junction

simulations revealed that the optimal band gap energy for both AMO and AM1.5 illumination

conditions is 1.44 eV, yielding an efficiency of 15.3% and 16%, respectively. The structure for

the refined single-junction was a 40 nm n-type layer with a graded carrier concentration of

1 x 1018 to 1 x 1020 cm-3 on top of a 2 [tm thick p-type layer with a uniform carrier concentration

of 1 x 1017 cm-3. Multi-junction solar cells that were connected in a mechanical stack

arrangement were simulated using the same structure as the refined single-junction solar cell.

The maximum efficiency (27.4%) was obtained for a five junction InxGal-xN solar cell, which is

comparable to current III-V solar cells based on Ga In,. P/GaAs/Ge. Comparing the

conservative efficiency estimates with other simulation approaches shows that the predicted

efficiency for a five junction InxGal-xN cell can range from 27-39%. The impressive upper limit

can only be achieved if material growth of InxGal-xN progresses to achieve the optimum

properties required for these cells. If the lower of the two cell efficiencies is only possible, as









assumed in this simulation, then the efficiency is still close to the best solar cells currently

available.

The effect of phase separation which occurs in InxGal-xN was also assessed to determine

the positive or negative effects on cell efficiency. For a single-junction solar cell, phase

separation could theoretically have a positive effect through the efficient generation of hot

electrons, however, it is mostly likely that phase separation will have a negligible effect on

single-junction cell efficiency. The refined InxGal-xN absorber structure presented here was

shown to be ideal for sustaining hot electrons. Even if the hot electrons do not relax, phonon

scattering at metal contacts would prevent the extra energy from being collected. Methods for

collecting hot electrons have been suggested by other researchers, however, these methods are

not practical. If an absorber layer phase separates, then the solar cell will take on the efficiency

characteristics of the lowest band gap energy material, which is assumed to be the majority

component or matrix material in an In-rich InxGal-xN alloy. The possibility of impact ionization

was also considered for phase-separated solar cells, however, no increase in efficiency was

produced most likely due to the lack of a high electric field in the semiconductor depletion

region.

Single-junction solar cells are predicted to practically unaffected by phase separation

unless the gallium composition becomes greater than 50%, leading to the formation In-rich

InxGal-xN precipitates. These smaller band gap energy precipitates will likely form carrier wells

where recombination will occur. Typically higher gallium contents are required for the top

absorber layers in multi-junction cells which require a wide band gap energy. For this reason it

is believed that phase separation could reduce the efficiency of top junctions in multi-junction












cell when gallium rich InxGal-xN alloys make up the matrix phase, however, only in a small


compositional range.


1.3E+05


1.1E+05 -


9.0E+04


7.E+04 -


5.0E+04


3.0E+04 -


1.OE+04
0.7 1 1.3 1.6 1.9 2.2 2.5 2.8 3.1 3.4
Band gap energy (eV)

Figure 2-1. Absorption coefficient of an InN thin film grown by MOCVD (Ref 203).


Wavelength (om]
1.2 1.0 0.9 0.8 0.7 0.6 0.5



CulnSe2
10*
Cerra


1 104 / /a-Si:H


GaAs mono.Si






10 / dS





1.0 1.5 2.0 2.5
Photon energy [eVI

Figure 2-2. Absorption coefficient as a function of wavelength for several photovoltaic
semiconductors (Ref 140).









Light


Front grid lines
V


Anti-reflecting coating



I-I- n-lnxGal-xN


Substrate/back contact


Figure 2-3. Proposed single-junction InxGal-xN solar cell which has been modeled by Medici.


Initial abmorber structurm


Refined absorber structure


n-ln Ga, N n = 10-" cm- t= 10 nm
n= 10"' cm- n-lnxGa-xN n = 101-'cm-: t= 10 nm
t= 300 nm n = 101- cm-: t= 20 nm


p-In Ga,_ N p-In Ga,. N
p = 10" cm- p = 10' cm--
t = 2 pm Drawing not t = 2.0pm
A to scale B
Figure 2-4. Solar cell structures used in single-junction solar cell simulation. A) Initial structure
used as a starting point for the simulations. B) Absorber structure after refinement of
cell parameters.












16%


14'.. -


12%


o0I -':



E
c



Z 6%


4'',. -


2%


nor,


0 1 2 3 4 5 6 7 8
Optimization step


1 1.25


1.5 1.75


9 10 11 12 13 14


Figure 2-5. Single-junction absorber optimization steps.


18%


16%


14%


12%


S10%
C

E 8%-


6%


4%


2%


0% -
0


x= 0


x=1.0


.5 0.75


Band gap (eV)

Figure 2-6. Simulated solar cell efficiency vs. band gap energy of the refined InxGal-xN cell
structure for AMO and AM1.5 illumination.


25 nm


50 nrn/ lu nm
n-side thickness
va -ation / p-side thickness
variation
100 nm

// n-type carrier
concentration grading
200 Layeri 1020 cnm
Layer2 1019 cm3
0 / Layer3 '108 cm3
300 n

-1 i 1 e eV V 1 5 eV 16eV
Eg = 1 1 eV


Initial band gap
tuning


x=0.82
x = 0.85-


x= 0.68 x=


x = 0.1


x = 0.91


7 x = 0.15
=0.94 x0.13
x =0.12'

.97


InGalxN
Band gap based on compositional dependent
bowng parameter [Bec03]


2 2.25


UJ u















20 -I-



E 16




L



5



0
0 100 200 300 400
Load (Ohm/cm2)
Figure 2-7. Power vs. load plots for the refined solar cell structure for
illumination.


10

AMO (1.44 eV)
5 --AM1.5 (1.44 eV)


0
0 -------------------------
-" 0.)0 0.20 0.40 0.60 0.80
E
I -5


-10


S-15 I
S J,, -20.11


500


AMO and AM1.5


Voltage (V)

Figure 2-8. Current density vs. voltage curves for the refined solar cell structure under AMO and
AM1.5 illumination.


















15%


1 2 3 4 5
Junction number
Figure 2-9. The efficiency and open circuit voltage as a function of junction number for
multi-junction InxGal-xN solar cells.


,


hp > E9
hi'
hrvv*


Figure 2-10. Energy band diagram showing the generation of electron-hole pairs and hot carriers
from incident light energy, adapted from Ref. 197.

n-type
layer 0 0-'

S--n I"1 _lnGa,,N
O' --precipitates

p-type x > y
layer
InGal-,N
matrix
0

Figure 2-11. Phase-separated InxGal-xN p-n junction showing Ga-rich InxGal-xN precipitates
(green circles) in an In-rich InxGal-xN matrix (yellow bulk).


5.44 V
4.32 V
3.27 V
2.13V




V, = 0.92 V













4-_


Figure 2-12. Corresponding band diagram from the p-type region in Figure 2-11 showing the
In-rich InxGal-xN matrix and the Ga-rich precipitate (ppt).


Figure 2-13. Band diagram showing trapping of carriers and recombination in narrow band gap
energy precipitates (ppt) in the wider band gap energy Ga-rich InxGal-xN matrix.


Table 2-1.


The GaN and InN material parameters used in the Medici simulations. Estimated
ara meters in italics (Ref 42a 47b 2000 21 5h


Electron Valence band Conduction band Electron Hole
Band gap affinity Dielectric density of states density of states mobility mobility
Material energy (eV) (eV) constant (Ny, cm-3) (Nc, cm-3) (cm2/V-s) (cm2/V-s)
GaN 3.46e 4.1d 8.9b 4.62 x 1019g 2.23 x 1018g 1000h 200c
InN 0.70a 5.8f 15.3b 5.20 x 1019g 9.15 x 1017g 1000 200


Table 2-2. Cell parameters for optimized cell structure with optimum band gap energy.
Max Incident
Fill Factor Efficiency
Band gap (eV) Illumination Jso (mA/cm2) Voc (V) Power Poweractor iciency
(mW/cm2) (mW/cm2) (%) (%)
1.44 AMO -24.95 0.918 19.72 128 86.09 15.36
1.44 AM1.5 -20.11 0.912 15.77 98 86.01 16.03











Table 2-3. Cell characteristics for simulations allowing hot carrier collection in phase-separated
InxGal-xN solar cells.
Precipitate Max Power Incident
Matrix/Precipitate I JSC Max Power er Fill Factor Efficiency
SPia volume Illumination Voc (V) Power (
band gap energy (eV) fraction (mA/cm2) (mW cm2) (mWm2 (%) (%)

1.0/2.0 0.10 AMO -39.16 0.593 18.93 128 81.53 14.75
1.0/2.0 0.10 AM1.5 -30.20 0.586 14.40 98 81.36 14.64
1.0/2.0 0.20 AMO -39.20 0.693 22.64 128 83.34 17.64
1.0/2.0 0.20 AM1.5 -30.22 0.686 17.25 98 83.20 17.54
1.0/2.0 0.25 AMO -39.23 0.774 24.50 128 80.69 19.09
1.0/2.0 0.25 AM1.5 -30.24 0.736 18.69 98 83.97 19.00

Table 2-4. Characteristics of individual layers and overall cell for simulated multi-junction
InxGal-xN solar cells.
Individual
Optimized Individual AMO Input Overall
Junction cell max Overall .
S band gap absorber power efficiency
number power 2 Voc (V)
number energy (eV) Voc (V) poe2 (mW/cm2) (%)
(mW/cm )
1 1.44 0.918 19.72 128 0.918 15.36

2 2.00 1.460 15.95 128 2.127 22.28
1.20 0.667 12.65 128
2.30 1.738 12.50 128
3 1.60 1.060 11.73 128 3.274 24.25
1.00 0.476 6.90 128
2.50 1.930 10.24 128
1.80 1.253 11.31 128 4.316 26.38
4 4.316 26.38
1.30 0.764 8.46 128
0.90 0.369 3.85 128
2.60 2.025 9.22 128
2.00 1.445 9.35 128
5 1.60 1.048 7.72 128 5.447 27.41
1.20 0.661 6.39 128
0.80 0.268 2.50 128









CHAPTER 3
GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN FIMLS BY
METAL ORGANIC CHEMICAL VAPOR DEPOSITION

3.1 Introduction

InN and InxGal-xN are less studied materials when compared to semiconductors such as Si

or GaAs. Ga-rich InxGal-xN (0 < x < 0.3) alloys have become dominant materials as active

layers in visible LEDs even though the mechanisms for light emission are not completely

understood. InN and In-rich InxGal-xN have been studied far less than Ga-rich InxGal-xN

primarily due to the few new device applications in the 2 eV band gap energy range and the

inability to grow high quality single crystal InN. More recently single crystal InN has been

consistently reproduced by MBE and MOCVD growth techniques.68 Available high quality InN

epitaxial layers have made it possible to investigate the properties of InN in greater detail. The

band gap energy controversy has sparked more interest in InN and In-rich InxGal-xN films

because it created new applications for these materials. These new applications include terahertz

emitters and detectors as well as high efficiency solar cells.

Before these new applications can be exploited it would be helpful if the fundamental

properties such as relationships between growth conditions and the resulting phase separation,

film composition, doping level, and crystalline quality were established. In this chapter

InxGal-xN phase separation will be discussed with respect to substrate material, buffer layer,

growth temperature, and inlet composition (film composition).

Currently high quality Ga-rich InxGal-xN alloys are grown at high temperature (800 C) for

LED applications and In-rich InxGal-xN is being grown for the fundamental study of In-rich

InxGal-xN alloy properties. A current summary of InxGal-xN alloy growth progress is reviewed

in Section 1.1.3. It has been shown experimentally that approximately 30% indium can be

incorporated into GaN at temperature above 700 oC.125-129 Low temperature growth by MBE has









demonstrated metastable InxGal-xN over the entire compositional range.47'130 No results have

been published for InxGal-xN over the entire compositional range by MOCVD. Also, no analysis

has specifically examined InxGal-xN growth parameters as they relate to phase separation. In

general most literature work merely states that phase separation was noticed and no correlation to

the growth conditions is made. This work aims to understand how changes in deposition

temperature, substrate material, and, substrate pretreatments (including buffer layers) affect

InxGal-xN phase separation.

3.2 Experimental Procedure

3.2.1 Substrate Preparation

InN and InxGal-xN thin films were typically deposited on c-A1203 substrates, however,

a-A1203, GaN/c-Al203, Si (100), and Si (111) substrates were also included in selected runs. The

GaN/c-A1203 substrates used in this study were obtained from Uniroyal Optoelectronics GaN

(5 [tm) grown by MOCVD on c-A1203. No chemicals were used to clean the GaN/c-Al203

surface and the substrates were cleaned with a high pressure nitrogen gun to remove any

particles from the surface before loading. The sapphire and silicon substrates were cleaned in

boiling trichloroethylene (TCE), acetone, and methanol each for 5 min to remove organic

material. A high pressure nitrogen gun was used to quickly dry the substrates after removal from

the methanol cleaning step.

3.2.2 The MOCVD Deposition Technique

Metal organic sources used in MOCVD are typically liquids, finely crushed solids, or

dissolved in a solvent and are contained in stainless steel bubblers. The pressure and temperature

of the bubblers is adjusted to control the partial pressure of the precursor species inside the

bubbler. Adjusting the total pressure and flow rate of the carrier gas controls the transport to the

reactor. This dilute stream of metal organic vapor and carrier gas is then mixed with the other









source materials, typically hydrides, and delivered to a substrate placed on a heated susceptor.

The susceptor is commonly heated by RF induction, radiative (lamp), or resistance heating. The

gaseous sources undergo complex homogeneous and heterogeneous reactions in the hot region at

and near the heated substrate to produce film growth.226'227

MOCVD reaction chambers are typically made of stainless steel or quartz and the shape of

the reactor is designed to produce laminar flow over the susceptor. Most MOCVD systems

operate at low pressure (50-200 Torr) because it reduces closed streamline flows which decrease

dead volumes or vortices. This type of growth technique is popular for III-V materials and is

typically used to grow epitaxial single crystal films, however, polycrystalline and amorphous

films can also be grown.226'227

Reaction Chemistry for MOCVD GaN and InN. The formation of GaN or InN on the

substrate surface from the reaction between the group III (TEGa or TMIn) and group V (NH3)

sources. The overall reactions are given by Eqs. 3-1 and 3-2.

(C2H5)3Ga (g) + NH3 (g) -* GaN (s) + 3C2H6 (g) (3-1)

(CH3)3In (g) + NH3 (g) InN (s) + 3CH4 (g) (3-2)

A general reaction for GaN or InN is represented by

R3M + NH3 MN (s) + 3RH (g) (3-3)

Where M = Ga or In, R = CH3 or C2H5.228,229

This expression does not reflect the actual reaction pathways because the details are not very

well established and the available information suggests the reactions are complex. Jacko and

Price were the first to study the pyrolysis of TMIn and suggested the following mechanism.230

In(CH3)3 In(CH3)2 + CH3- (3-4)

In(CH3)2 -- In(CH3) + CH3- (3-5)









In(CH3) -- In + CH3* (3-6)

DenBaars et al. 231 later studied the decomposition of trimethyl gallium in the growth of epitaxial

GaAs and concluded the same mechanism as Jacko and Price. More recently the decomposition

of TMIn was studied by in-situ Raman spectroscopy.232 The results of this study suggest several

reaction intermediates appear during the decomposition of TMIn,

nMMIn (MMIn)n (3-7)

2DMIn (DMIn)2 (3-8)

DMIn + MMIn DMIn-MMIn (3-9)

DMIn-MMIn CH3InCH2 + HInCH3 (3-10)

TMIn + MMIn (DMIn)2 (3-11)

where MMIn is monomethyl indium and DMIn is dimethyl indium. The most energetically

favorable intermediate was calculated to be (DMIn)2 followed by the slightly less favorable

production of DMIn-MMIn.

The decomposition of ammonia at high growth temperature is assumed to take place on the

substrate surface or reactor walls to yield atomic nitrogen or a nitrogen containing radical.

Thermodynamically it is known that NH3 decomposes completely into N2 an H2 at temperatures

above 300 C. When temperature is below 650 C and no catalyst is used, NH3 decomposition is

slow and strongly depends on growth conditions and reactor design.233 It is also believed that the

removal of the first hydrogen bond is the rate limiting step in ammonia decomposition.228

NH3 (g) NH3-x (g) + xH (g) (3-12)

From the reaction in Eq. 3-12 a possible growth mechanism for InN and GaN at a solid gas

interface is,

M(CH3) (s/g) + NH (s/g) -* M-N (s) + CH4 (3-13)









where M = Ga or In.

MOCVD growth is not an equilibrium process, therefore thermodynamics can only

determine the overall driving force for the reaction and reactor kinetics or transport rates define

the rate at which film growth will proceed. Thus the reaction pathways and rate constants, along

with the flow velocities and temperature gradients near the substrate are very important. A

schematic of the precursors at the solid-gas interface and boundary layer regions for the reaction

oftrimethyl gallium (TMGa) and ammonia to form GaN (Figure 3-1). A similar diagram is

expected for TMIn and NH3 sources to form InN at the solid-gas interface. TMGa must diffuse

through the boundary layer, possibly pyrolize, and then adsorb on to the substrate surface where

the adsorbed Ga-containing molecule or atom reacts with ammonia that also diffused through the

boundary layer with possible reactions (Figure 3-1). The growth rate of these MOCVD reactions

is controlled by the reaction rate at low temperature and diffusion of the group III source though

the boundary layer at higher temperatures where the reaction rates become high, since NH3 is

usually in great excess. Adduct formation such as Ga(CH3)2-NH2 can also be important for high

temperature growth (Figure 3-1). The growth kinetics and the growth mechanisms that occur at

the solid gas interface are not very well understood, however, this has not prevented empirical

growth studies.228

3.2.3 The MOCVD Reactor

MOCVD experiments were performed in a horizontal, low pressure (100 Torr), cold-

walled quartz reactor with RF-inductive heating of a tilted graphite susceptor. This MOCVD

system was originally designed by Nippon Sanso for low temperature GaAs growth, but has

subsequently been modified for InN, GaN, or InxGal-xN growth.203'234 Solid trimethyl indium

(TMIn, 99.9995%, Rhom and Haas Electronic Materials), liquid triethyl gallium (TEGa,









99.9995%, Rhom and Haas Electronic Materials), and ammonia (NH3, 99.9999%, Air Products)

are used as precursors with a nitrogen carrier gas (UF Microfabritech building LN2 boil off).

The MOCVD reactor schematic and a photo are shown in Figure 3-2. Samples are loaded

onto a quartz wafer tray and placed in a load-lock which is evacuated and purged 8 to 10 times to

minimize oxygen contamination from reaching the reactor. A mechanical fork is used to load the

wafer tray onto the susceptor before starting the reaction. During the experiment N2 is delivered

to the metal organic precursor bubbler, which is then combined with a dilution N2 stream and the

pure NH3 stream, just before entering the quartz chamber. The metal organic precursor

temperature is controlled by NESLAB RTE (Thermo Electron Corporation) refrigerated

bath/circulators. In these studies, the susceptor temperature was in the range of 450-850 C and

the quartz wall was maintained at a constant 25 C with cooling water. Down stream waste gases

are pumped by a dry vacuum pump (BOC Edwards XDS 10) and removed through the

Microfabritech exhaust system.

The growth parameters such as susceptor growth temperature, N/III ratio, substrate

nitridation time and temperature, buffer layer material thickness, total flow rate and reactor

pressure are chosen depending on the desired epitaxial layer properties. The N/III ratio is the

inlet molar ratio of NH3/TMIn (or NH3/TEGa) and the N/III ratio is controlled by adjusting the

flow rates of nitrogen through the TMIn (TEGa) bubbler and independently by the ammonia

flow rate. Growth temperatures is controlled by a PID temperature controller (Gultan West

2070) that uses a quartz insulated thermocouple to measure the susceptor temperature and adjust

the power of a Lepal T-15 RF generator. The thermocouple is located inside the center of the

graphite susceptor, via a drilled hole in the susceptor.









3.2.4 In-Situ Sapphire Substrate Surface Treatment and Buffer Layers

In-situ pretreatments such as nitridation of sapphire substrates or the addition of a low

temperature (LT) InN buffer layers are used for the growth of InN and InxGal-xN thin films.

Nitridation of sapphire substrates is applied to grade the lattice mismatch between the III-nitride

main layer and the sapphire substrate by forming an A1N nucleation layer.70 LT InN buffer

layers have also been shown to improve subsequent film quality compared to samples without

buffer layers.73'74 Aspects of substrate nitridation and buffer layers are discussed in greater detail

in Section 1.1.3.

Sapphire substrates were nitridated under an ammonia flow (1600 sccm, 4 SLM N2

dilution) at high temperature (850 C) for duration of 15 min prior to growth. After nitridation

the reactor was allowed to cool (in a low pressure N2 atmosphere) to the appropriate growth

temperature selected for buffer layer growth or the main layer growth. LT InN buffer layers

were only used for some InxGal-xN samples because the primary focus of this study was not to

produce high crystalline quality but to understand how phase separation is related to growth

conditions. When used, LT InN buffer layers were grown for 15 min at a temperature of 450 C

and a N/III ratio of 50,000. MOCVD optimization of InN deposition temperature, N/III ratio,

buffer layer temperature/duration, and substrate nitridation temperature/duration has been done

elsewhere.203'234

3.3 Results

3.3.1 Indium Nitride

3.3.1.1 Growth on silicon substrates

Silicon is widely used in the semiconductor industry and the integration of nitride-based

devices with Si IC technology is a driver for using Si substrates. Si (111) substrates have a

lattice mismatch of -8% with InN, which is less than nitridated A1203 (A1N, 13%) and GaN









(11%), while Si (100) has a lattice mismatch of 35% with InN. InN thin films were grown on

p-Si (100) and n-Si (111) substrates at 530 C and an inlet N/III ratio of 50,000. Surface

pretreatment conditions such as nitridation and the use of an InN buffer layer were varied to

refine the growth conditions of InN on silicon substrates.

Growth of InN directly on silicon substrates can lead to growth of incomplete films or

polycrystalline films with very small and broad XRD peak intensities. Standard TCE, acetone,

and methanol chemical cleaning steps were used to clean the Si substrates prior to growth.

Direct growth on Si (111) substrates without the use of a LT InN buffer layer or substrate

nitridation forms an incomplete film of InN (Tg = 530 C and N/III = 50,000, Figure 3-3). When

a LT InN buffer layer is used a continuous film is formed (SEM, Figure 3-4), however the

resulting films is polycrystalline, as seen from the appearance of the (002), (101), and (102)

reflections in the XRD pattern (Figure 3-5). InN is believed to be polycrystalline when grown

directly on silicon substrates due to an amorphous SiNx layer that forms during the initial stage

of growth from the exposure of Si to NH3.68 The formation of SiNx can be suppressed by

growing a protective non-nitride heterostructured buffer layer, such as GaAs followed by

subsequent growth of InN, yielding a continuous and smooth surface morphology.235

Intentional substrate nitridation to form a crystalline 3-Si3N4 layer by reacting NH3 with

the silicon substrate at high temperature has also been tested.236 When brief substrate nitridation

is used (< 15 min), however, it is likely that a silicon oxynitride (SiOxN-x) forms instead of a

pure P-Si3N4. This result has been shown by Kryliouk et al. using identical deposition

equipment and substrate nitridation conditions.237 The formation of a silicon oxynitride

intermediate layer has been shown to be an ideal method for producing single crystal GaN on

silicon substrates. A TEM image of GaN/Si (111) shows a thin amorphous structure for the









silicon oxynitride layer (Figure 3-6). Details about the growth and characterization of the

SiOxNi-x intermediate layer can be found elsewhere.234'238 This result encourages performing

silicon substrate nitridation followed by growth of a LT InN buffer layer to grow single crystal

InN on Si (111) and (100). The same sapphire nitridation step (850 C, 1600 sccm NH3 for 15

min) was applied to silicon substrates (as described for sapphire in Section 3.2.4), and InN was

grown for 1 hr at 530 C (N/III = 50,000). XRD patterns of InN on Si (111) and Si (100) are

shown in Figure 3-7. Comparing the patterns for InN/Si (111) (Figures 3-5 and 3-7) it can be

seen that the use of substrate nitridation produces a highly textured InN film with a preferred

growth orientation in the [0001] direction and the InN (101) reflection can no longer be detected.

When substrate nitridation and a LT InN buffer layer are used for InN growth on Si (100) only

the InN (002) reflection is detected and no polycrystalline peaks appear. The Si (200) substrate

peak occurs at a 20 value which is close to the polycrystalline InN (101) peak, which can cause

the polycrystalline InN peak to go unnoticed. Adding a slight tilt (3 or 4) in the omega angle

(normal to the film surface) during an XRD scan will cause the single crystal silicon peak to be

undetected while the less crystalline InN remains unchanged. With the Si substrate peak

removed, no polycrystalline InN (101) peak is detected (Figure 3-7). For the nitridated Si

substrates, there is little difference in the magnitude of the peak intensities of the (002) reflection

for InN/Si (111) and InN/Si (100) even though there is a large difference in lattice mismatch with

respect to InN. The XRD peak intensity is strongly dependent upon the specimen interaction

volume (film thickness and surface area) and the crystallinity of the film. The growth rate of

InN/Si (100) was determined to be 80 nm/hr from cross-sectional SEM (Figure 3-8) and a

similar growth rate was found for InN/Si (111). Therefore the film thickness can be ignored as

contributing factor to any difference in InN (002) peak intensity. Films were typically grown on









cut 10 x 10 mm2 substrates, which eliminates any sample size affect on the XRD peak intensity.

This result suggests that the oxynitride layer has more influence over the film crystallinity than

the silicon substrate orientation. The intermediate SiOxNi-x layer was previously determined to

have an amorphous structure,234 which may account for the small InN (002) peak intensities

when grown on Si.

Even with the use of substrate nitridation the InN (002) peak intensities are relatively small

compared to InN/c-A1203, grown under otherwise identical conditions and therefore the

crystalline quality of InN/Si could be significantly improved. The growth rate of InN/c-Al203 is

-100 nm/hr and was previously determined for the same deposition equipment used in this

study.203 Since InN has a similar growth rate and similar sample sizes (10 x 10 mm2) were

tested, then the difference in InN (002) peak intensity can be attributed to a difference in

crystalline quality. InN/Si was grown for two hr to account for any growth rate difference for

c-A1203 and Si substrates, which again yielded a much larger peak for InN/c-A1203.

For this reason, InN thin films were grown on Si (111) substrates by using a hydride-

MOCVD growth technique. In this technique, HC1 reacts with the metal organic species (TMIn)

to form volatile InCl, which then reacts with NH3 to form InN. A more detailed description of

the H-MOCVD deposition technique is given in Section 5.2.2. This technique is beneficial for

growth because the presence of HC1 cleans residual SiOx on the Si surface and takes advantage

of the small lattice mismatch (-8%) between InN (002) and Si (111). The oxynitride cannot form

because HC1 reacts with SiOx to form volatile SiHnC14-n and H20. Other benefits of the

H-MOCVD technique include the ability to grow InN at low N/III ratios due to preferential

etching of In droplets by HC1.234 H-MOCVD growth of InN also allows for high growth rate of









InN without sacrificing growth quality, which would occur if a conventional MOCVD used a

similar high concentration of reactants.

Thus InN growth on Si was attempted in a different reactor by H-MOCVD as described

below. After standard chemical cleaning, the Si surface was cleaned in-situ (prior to growth) by

annealing in a H2 atmosphere for 10 min at 850 C. The reactor was allowed to cool in a H2

atmosphere to the buffer layer growth temperature of 450 C. HC1 gas was allowed to flow into

the reactor for 2 min prior to buffer layer growth (450 C) with no other source materials flowing

to provide secondary cleaning of the silicon substrate. A LT InN buffer layer was grown (in N2

carrier) for 15 min (N/In = 2,500 and 450 C) followed by InN growth at 560 C (N/In = 2,500)

for 2 hr. When grown on Si (111) in the H-MOCVD system, InN is highly textured in the (002)

direction and with a large and narrow peak intensity signifying increased crystalline quality

compared to InN/Si (111) grown by conventional MOCVD (XRD, Figure 3-9). H-MOCVD

InN/Si (111) grows in a columnar structure with a very rough surface where grain sizes range

from 150 to 500 nm (SEM, Figure 3-10). It is suggested that the columnar structure is made up

of vertical InN (002) columns since the XRD results show a highly textured film in the [0002]

direction.

Another benefit of H-MOCVD besides increased crystalline quality is that the growth rate

of InN (750 nm/hr) is an order of magnitude higher compared to traditional MOCVD (78 nm/hr).

It must be noted that a significantly higher TMIn flow rate is used in the H-MOCVD (0.67 sccm)

compared to conventional MOCVD (0.03 sccm). It is difficult to scale up a conventional

MOCVD system to achieve this high growth rate because it would require an extremely large

NH3 flow rate (35 SLM) to maintain the required inlet N/III ratio of 50,000 for preventing

indium droplet formation. The addition of HC1 using the H-MOCVD system allows for a









moderate NH3 flow rates of 1.7 SLM to be used. The results of energy dispersive spectroscopy

(EDS) and Auger electron spectroscopy (AES) analysis confirmed that no chlorine was detected

in the film (within the available detection limits, 0.5-1 atomic %, Figure 3-11).

Growth of InN on Si substrates is desired for several reasons including the reduced cost of

silicon, large area substrates, and future device integration. Substrate nitridation and the

formation of a SiOxNi-x intermediate layer is shown to be critical to form highly textured InN

(002) when grown on Si (111) and Si (100) by MOCVD. The amorphous SiOxNi-x layer,

however, affects the ability to form high quality InN. H-MOCVD proves to be an ideal

technique for growth of InN/Si (111) because the addition of HCl prevents the oxynitride from

forming and the lattice mismatch remains at -8%. The low lattice mismatch produces a highly

textured InN (002) film and a higher growth rate was possible with H-MOCVD at a low N/III

ratio without the formation of indium droplets. This is a huge advantage compared to traditional

MOCVD where InN is typically growth at N/III ratios of 50,000.

3.3.1.2 Film stability and aging

Long term film stability is important for future device integration of InN based materials.

If InN is not stable in oxygen containing environments such as air then fabrication steps such as

annealing of electrical contacts becomes more difficult. It has been suggested that oxygen is

incorporated into InN films after air exposure at room temperature, for example sputtered InN

films have been shown to incorporate oxygen into the grain boundaries of InN and form

crystalline In203 phases at room temperature.50 This room temperature annealing process was

determined over a period of months to years. To independently verify this room temperature

annealing process, InN thin films were grown on a-,c-, and r-orientations of A1203 as well as Si

(100) and Si (111) by MOCVD. For all orientations of A1203 and Si substrates, substrate

nitridation and a LT InN buffer layer were used. Post growth characterization was done by XRD









and then the samples were stored in air for a period of 12 to 15 months. After aging (i.e., room

temperature annealing) in air the same samples were analyzed again by XRD (Figure 3-12).

There is no evidence of In203 crystalline phase formation when InN samples (on c-A1203 or

Si (100) substrates) are aged in air for periods exceeding one year (Figure 3-12). It is important

to mention that amorphous oxide phases are undetectable by XRD. This analysis does not

support the claim by Butcher and Tansley5s that crystalline indium oxide phases can form in InN

after room temperature aging. However, sputtered films were used in the reference case and

MOCVD grown InN was used for this work. It is possible that grain boundaries play a role for

oxygen incorporation into InN films. Sputtered films are polycrystalline which lead to a large

number of grain boundaries and impurity diffusion occurs more rapidly through grain boundaries

compared to the bulk for single crystal films.239 No differences in In203 formation were noticed

for single crystal InN (InN/c-A1203) and polycrystalline InN (InN/Si (100)), which differ by the

amount of grain boundary densities and area available for oxide growth and oxygen transport.

From this analysis it remains unclear how crystalline oxide phases could form in InN at room

temperature.

From the results above, formation of crystalline indium oxides does not occur at room

temperature to an extent detectable by XRD, at higher temperature, however, crystalline oxide

phases are more likely to form. Yodo et al.240 found that crystalline In203 (222) forms in InN

upon annealing at 500 C for 5 min in a N2 atmosphere (1 atm). These MBE InN samples were

deposited at 500 C for a much greater time than the annealing time and no In203 XRD peak was

evident in the as grown film. SIMS analysis determined a 1% residual oxygen concentration in

the as-grown film. These MBE samples were deposited under In-rich conditions and therefore

In-droplets formed on the surface. It was concluded that the indium droplets reacted with









oxygen to form amorphous indium oxides upon air exposure and these phases will crystallize

after annealing at elevated temperature.

For comparison InN/c-A1203 substrates were grown by MOCVD using substrate

nitridation and a LT InN buffer layer and then exposed to air. These samples were then aged for

6 months to ensure that sufficient time was given for oxygen to diffuse into the film. A high

N/In ratio (50,000) was used for growth therefore no In droplets were present on the surface.

After aging, InN was annealed in a low pressure (100 Torr) N2 atmosphere for 20 min at 500 oC,

15 min at 525 C and 10 min at 550 C. The as-grown samples are compared to the annealed

samples (Figure 3-13). The In203 (222) peak occurs at 20 = 30.6 and it is clear (Figure 3-13)

that no crystalline indium oxide is formed when InN is annealed up to 550 C in a low pressure

N2 atmosphere. For all three annealing conditions the InN (002) peak intensity is increased when

the same XRD setup was used, suggesting that the film crystallinity was improved. A slight

narrowing (AFWHM = -144 arcsec) of the InN (002) peak was found for annealed samples

compared to the as-grown sample. From this analysis it can be determined that annealing in low

pressure N2 atmospheres can prevent the formation of indium oxide crystalline phases in InN

(without In droplets), even after exposure to air. These results also suggest that the growth of

stoichiometric InN films is important for preventing the formation crystalline In203.

3.3.2 Indium Gallium Nitride

3.3.2.1 Metastable InxGal-xN alloys over the entire range (0 < x < 1)

InxGal-xN alloys were grown on c-A1203 substrates without an InN buffer layer at 530 C

at varying composition over the entire compositional range (0 < x < 1). Each film has

approximate thickness of 100 nm, based on previous growth rate analysis.203 Typically a fixed

deposition temperature is not used for growth over the entire compositional range because higher

Ga compositions achieve better crystallinity at higher temperature. Low deposition temperature,









however, has been used to grow metastable InxGal-xN alloys by MBE over the entire

compositional range.47,130 These metastable films have only been grown by MBE techniques and

currently no evidence has been published for metastable InxGal-xN films grown over the entire

compositional range by MOCVD.

In this work a low temperature approach was used to suppress phase separation. The

InxGal-xN film composition was controlled by varying the inlet group III flux (TEGa or TMIn

flow rate). The inlet flow ratio, TMIn/(TMIn + TEGa) or simplified as In/(In+Ga), is the molar

ratio of the TMIn molar flow to the total inlet group III molar flow. The ammonia flow rate was

adjusted to maintain a constant N/III ratio of 50,000 for each experiment. XRD patterns of pure

InN, pure GaN, and seven different compositions of InxGal-xN are shown in Figure 3-14. The

pure InN (002) reflection located at 20 = 31.4 shifts to the right (towards the GaN (002) peak) as

the flow ratio is changed to lower values (more Ga-rich) (Figure 3-14). Also, it is important to

notice that there is no second reflection in the pattern that would suggest microscopic phase

separation. These results are consistent with the hypothesis that a low growth temperature will

suppress phase separation. This is the first time single phase deposition of InxGal-xN has been

demonstrated over the entire compositional range using MOCVD. It is also evident that the low

growth temperature significantly affects the crystalline quality, since high crystalline quality for

GaN usually occurs at temperatures at or above 850 C.

A table of inlet flow ratio and the corresponding film composition (x) as determined by

XRD for each run, is shown in Table 3-1. Bragg's law (Eq. 3-14) is used to determine the

distance between atomic planes, d-spacing, by using the experimental 20 values recorded for the

InxGal-xN (002) peak at each flow ratio.

nA= 2dsin(O) (3-14)









In the equation for Bragg's law, n is an integer and X is the wavelength of the incident X-rays

(CuKa = 1.54056 A). Using the d-spacing value calculated from Bragg's law and the

corresponding crystal plane indices, the a and c lattice parameters can be determined (Eq. 3-15).

Simplifications ofEq. 3-15 are shown for InxGal-xN (002) and InxGal-xN (101) (Eqs. 3-16 and 3-

17, respectively). The relationship between cubic indices and hexagonal indices are given in Eq.

3-18.

1 4 h2 + hk + k2 I2
For hexagonal structures: 3 + (3-15)
d2 3 a2 ) c

1 12 4 2 4sinO
InxGal-xN (002): 2 =2 2 or c (3-16)
d c c d nA


InxGa-xN (101): 12 4 +- +2 (3-17)
d 3 a2 c

Cubic (h k 1) Hexagonal (h k i 1), where i = -(h + k) (3-18)

The InxGal-xN composition approximately follows a linear dependence with a slope of one for all

compositions tested (Figure 3-15). The compositional dependence shows a negative deviation

(xfilm < xiniet) for In/(In+Ga) > 0.5 and a positive deviation filmm > Xinlet) for In/(In+Ga) < 0.5. A

similar linear dependence on composition has been shown by Matsuoka et al. for Ga-rich

InxGal-xN alloys grown by MOCVD (Figure 1-11).12 The positive and negative compositional

deviations can be explained by the suppression of InN decomposition by GaN coverage and

surface site blocking by methyl radicals, respectively. It is know that InN is reaction limited at

the low deposition temperature used (530 C) due to decreased NH3 decomposition

efficiency.241'242 Talaleav et al. studied kinetic effects limiting the growth rate of III-V

compounds by MOCVD and concluded that the blocking of group III species by methyl radicals

was a dominant mechanism for reducing the growth rate.243 In this case of InxGal-xN growth at









low temperature (530 C), the addition of TEGa (for In/(In+Ga) > 0.5) is believed to reduce the

growth rate of InN due to site blocking by Ga atoms and methyl radicals. This leads to a

negative deviation from the linear relationship between composition and inlet flow ratio. As the

inlet flow ratio becomes greater than 0.5 it is believed that surface coverage by GaN suppresses

thermal decomposition of InN, which increases the solid indium composition in the film. The

suppression of InN decomposition produces a positive deviation from the linear dependence of

composition to the inlet flow ratio.

3.3.2.2 Effect of growth temperature on stability of Ino.sGa.z2N

To understand the temperature dependence on phase separation, InxGal-xN alloys were

grown at a constant flow ratio (In/(In+Ga) = 0.8) and deposition temperature was varied in the

range from 530 to 850 C in 30 C intervals. Films were grown on c-A1203 substrates without an

InN buffer layer with an approximate thicknesses of 300 nm, as estimated by previous growth

rate analysis.203 A constant N/III ratio of 50,000 was maintained at each growth temperature. A

value of In/(In+Ga) = 0.8 was chosen for this study because phase separation is typically not

seen with films grown at this inlet In fraction.119,244 As the indium inlet fraction decreases below

0.8 phase separation is more likely to occur.245 The results given in the previous section suggests

that phase separation is suppressed when using a low growth temperature and it is postulated that

phase separation would be more likely to occur at higher deposition temperature. At higher

deposition temperatures Ga is more likely to incorporate into the solid solution due to the

increased thermal decomposition of InN. This result has been verified experimentally (Figure

1-11) for a deposition temperature of 800 C.12 Increasing the deposition temperature will

change the InxGal-xN composition for the same In-rich flow condition used to grow stable

Ino.8Ga0.2N (at 530 C). It is believed that higher deposition temperature will lead to phase

separation since the Ga content will increase (at a constant inlet flow ratio), and the InxGal-xN




Full Text

PAGE 1

1 METALORGANIC CHEMICAL VAPOR DEPO SITON OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN FILMS AND NANOSTRUCTURES FOR ELECTRONIC AND PHOTOVOL TAIC APPLICATIONS By JOSHUA L. MANGUM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Joshua L. Mangum

PAGE 3

3 To Kimberly, Matt, Cassie, and Bill.

PAGE 4

4 ACKNOWLEDGMENTS First and foremost I would like to thank Dr. Tim Anderson for being my advisor and supervisory committee chair. His extensive kno wledge as a researcher and experience in academia taught me invaluable lessons that I will carry with me throughout my career. I would especially like to thank my supervis ory committee cochair Dr. Olga Kryliouk. Dr. Kryliouk has been a wonderful teacher and mentor and a true friend. I feel lucky to have worked with such a talented researcher, whos e passion for research a nd friendly nature made every day enjoyable. I would also like to thank my other committee members (Dr. Steven Pearton and Dr. Jason Weaver) for their valuable insights. My appreciation goes out to the entire staff at the Major Analytical Instrumentation Center (MAIC), especially Dr. Amelia Dempere, Kerr y Siebein, and Rosabel Ruiz, for giving me the opportunity to work with them a nd for creating such an enjoyable work environment. Thanks go to the staff at Microfabritech (especially Chuc k Roland and Scott Gapinski) for always helping me keep my equipment in proper working order. I sincerely thank my family for their consta nt support during graduate school. I thank my father for encouraging me to continue my educa tion and I thank my mother for showing me that anything can be accomplished with enough desire a nd effort. I also thank my brother for his positive attitude and the fun times he brings into my life. Most importantly I want to thank my fian ce, Kimberly, who has been my greatest supporter and companion. Her patience and love has helped me achieve my goals throughout graduate school. I also give thanks to the Gr ay family for being my home away from home while I lived in Florida.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .........9 ABSTRACT....................................................................................................................... ............17 CHAPTER 1 INTRODUCTION..................................................................................................................19 1.1 Group III-Nitrides.........................................................................................................19 1.1.1 The III-Nitride Crystal Structure.......................................................................19 1.1.2 Properties of InN and GaN................................................................................20 1.1.3 Growth of InN and InxGa1-xN...........................................................................21 1.2 Photovoltaics.............................................................................................................. ...34 1.2.1 Fundamental Physics of Solar Cell Devices.....................................................36 1.2.2 Why InxGa1-xN?................................................................................................38 1.2.3 Current Progress of InN and InxGa1-xN Solar Cells..........................................39 1.3 Terahertz Applications for InN and InxGa1-xN..............................................................42 1.4 Statement of Thesis.......................................................................................................45 2 INDIUM GALLIUM NITRIDE SOLA R CELL DEVICE SIMULATIONS........................59 2.1 Introduction............................................................................................................... ....59 2.2 MEDICI Device Simulation Software..........................................................................60 2.3 Identification of InxGa1-xN Solar Cell Parameters........................................................60 2.4 Results.................................................................................................................... .......63 2.4.1 Single-Junction InxGa1-xN Cell Optimization...................................................63 2.4.2 Multi-Junction InxGa1-xN Solar Cells...............................................................68 2.4.3 Phase Separation in InxGa1-xN Solar Cells........................................................69 2.4.3.1 Effect on single-junction cells............................................................69 2.4.3.2 Effect on multi-junction solar cells.....................................................73 2.5 Conclusions................................................................................................................ ...75 3 GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN FIMLS BY METAL ORGANIC CHEM ICAL VAPOR DEPOSITION...............................84 3.1 Introduction............................................................................................................... ....84 3.2 Experimental Procedure................................................................................................85 3.2.1 Substrate Preparation........................................................................................85 3.2.2 The MOCVD Deposition Technique................................................................85 3.2.3 The MOCVD Reactor.......................................................................................88

PAGE 6

6 3.2.4 In-Situ Sapphire Substrate Surface Treatment and Buffer Layers....................90 3.3 Results.................................................................................................................... .......90 3.3.1 Indium Nitride...................................................................................................90 3.3.1.1 Growth on silicon substrates...............................................................90 3.3.1.2 Film stability and aging......................................................................95 3.3.2 Indium Gallium Nitride.....................................................................................97 3.3.2.1 Metastable InxGa1-xN alloys over the entire range (0 x 1).........97 3.3.2.2 Effect of growth temperature on stability of In0.8Ga0.2N..................100 3.3.2.3 Effect of substrate on stability of InxGa1-xN alloys..........................107 3.3.2.4 Determination of InxGa1-xN growth rate and composition from RBS measurements...........................................................................113 3.3.2.5 Terahertz emission from InxGa1-xN alloys........................................115 3.4 Conclusions................................................................................................................ .117 4 GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE NANOWIRES BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION.................139 4.1 Introduction............................................................................................................... ..139 4.2 Experimental Procedure..............................................................................................140 4.3 Results.................................................................................................................... .....141 4.3.1 Proposed Mechanism for MOCVD Nanowire Growth...................................141 4.3.2 Nanowire Morphology Dependence on Growth Parameters..........................145 4.3.2.1 Growth on p-Si (100) substrates.......................................................145 4.3.2.2 Growth on GaN/c-Al2O3 substrates..................................................149 4.3.3 InN Nanowire Composition and Structure......................................................151 4.3.4 Transport Properties of InNIn2O3 Core-Shell Nanowires.............................153 4.3.5 Single InNIn2O3 Core-Shell Nanowire for H2 Gas Sensing..........................155 4.4 Conclusions................................................................................................................ .157 5 GROWTH AND CHARACTERIZATION OF INDIUM NITR IDE NANOAND MICRORODS BY HYDRIDE-META L ORGANIC CHEMICAL VAPOR DEPOSITION..................................................................................................................... ..180 5.1 Introduction............................................................................................................... ..180 5.2 Experimental Procedure..............................................................................................180 5.2.1 Substrate Preparation......................................................................................180 5.2.2 The H-MOCVD Reactor and Deposition Technique......................................180 5.3 Results.................................................................................................................... .....182 5.3.1 Characterization of InN Nanorods Grown on Different Substrates................182 5.3.2 Growth of InN Nanoand Micror ods to Increase Aspect Ratio.....................185 5.3.3 Transport Properties of InN Microrods...........................................................187 5.4 Conclusions................................................................................................................ .188

PAGE 7

7 6 EXPLORATORY STUDY OF GALLIUM NITRIDE NANOAND MICROSTRUCTURED GROWTH BY M ETAL ORGANIC CHEMICAL VAPOR DEPOSITION..................................................................................................................... ..199 6.1 Introduction............................................................................................................... ..199 6.2 Experimental Procedure..............................................................................................199 6.3 Results.................................................................................................................... .....200 6.4 Conclusions................................................................................................................ .204 7 FUTURE WORK AND RECOMMENDATIONS..............................................................213 7.1 Role of Oxygen in InN................................................................................................213 7.2 Improving Crystalline Quality of InxGa1-xN (1 x 0.3) Alloys............................214 7.3 P-Type Doping of InN and In-Rich InxGa1-xN............................................................214 7.4 Growth of InxGa1-xN Alloys by H-MOVPE................................................................215 LIST OF REFERENCES.............................................................................................................216 BIOGRAPHICAL SKETCH.......................................................................................................233

PAGE 8

8 LIST OF TABLES Table page 1-1 Some fundamental prope rties of InN and GaN..................................................................57 1-2 Theoretical and experime ntal mobility and controlled doping ranges for InN and GaN............................................................................................................................ ........57 1-3 Lattice and thermal expansion constants for InN substrates and buffer layers..................58 1-4 Internal quantum efficiency (IQE) of InxGa1-xN p-i-n and quantum well solar cells........58 2-1 The GaN and InN material parameters used in the Medici simulations............................82 2-2 Cell parameters for optimized cell structure with optimum band gap energy...................82 2-3 Cell characteristics for simulations allowi ng hot carrier collecti on in phase-separated InxGa1-xN solar cells...........................................................................................................83 2-4 Characteristics of individual layers and overall cell for simulated multi-junction InxGa1-xN solar cells...........................................................................................................83 3-1 Flow ratios and corresponding InxGa1-xN compositions for metastable InxGa1-xN/c-Al2O3 grown at low temperature (530 oC)...................................................136 3-2 The InxGa1-xN (002) composition(s) for each deposition temperature ranging from 530 to 770 oC when grown at a constant inlet flow ratio of In/(In+Ga) = 0.8.................137 3-3 Thickness and compositional data obtained from RBS measurements for InxGa1-xN/c-Al2O3 grown for 2 hr....................................................................................138 3-4 Comparison of InxGa1-xN film compositions as determined by XRD and RBS..............138 4-1 Flow velocities and Reynolds number at th e outlet of the vertical inlet and inside the quartz reactor tube............................................................................................................178 4-2 The InN nanowire growth conditions for samples with varying substrate/surface pretreatments and the resulting morphology information................................................179 5-1 A random sampling of InN nanorod/microrod lengths and diameters for a variety of growth conditions.............................................................................................................198

PAGE 9

9 LIST OF FIGURES Figure page 1-1 The III-nitride zincblende crysta l structure along vari ous directions................................47 1-2 The III-nitride wurtzite crystal structure along various directions....................................47 1-3 Cation-faced (Ga) and nitr ogen-faced polarity for the GaN wurtzite crystal structure.....47 1-4 Velocity-field characte ristics (T = 300K, n = 1017 cm-3) for wurtzite InN, GaN, AlN, and zincblende GaAs.........................................................................................................48 1-5 Vapor pressure of N2 in equilibrium with of InN, GaN, and AlN as a function of temperature.................................................................................................................... ....48 1-6 Rhombohedral structure and surface planes of sapphire...................................................49 1-7 Theoretical and experimental exampl es of nanowire growth mechanisms.......................49 1-8 Predicted binodal (solid) and spinoda l (dashed) decomposition curves for InxGa1-xN assuming regular solution mixing......................................................................................50 1-9 The T-x phase diagrams of ternary InxGa1-xN compounds................................................50 1-10 Several XRD spectra of phase separated InxGa1-xN films at different temperatures.........51 1-11 Compositional control of InxGa1-xN with respect to growth temperature..........................51 1-12 U.S. photovoltaic module implementation from 1992-2003.............................................52 1-13 Comparison of world wide grow th of PV production from 1990-2004............................52 1-14 Market shares of different photovoltaic materials as of 2001............................................53 1-15 Schematic of light induced carrier generation in a generic solar cell................................53 1-16 Solar irradiance vs. wavelength for extrat errestrial (AM0) and terrestrial (AM1.5) spectra........................................................................................................................ ........54 1-17 Global solar irradiances averaged over three years (1991-1993) which account for cloud cover.................................................................................................................... .....55 1-18 Current-voltage characteristi cs of a generic solar cell.......................................................55 1-19 Incident solar flux for AM0 (bl ack) and AM1.5 (red) as well as the InxGa1-xN band gap energy range............................................................................................................... .56

PAGE 10

10 1-20 Terahertz frequency range (shaded) of the electromagnetic spectrum, including molecular transitions..........................................................................................................56 2-1 Absorption coefficient of an InN thin film grown by MOCVD........................................77 2-2 Absorption coefficient as a functi on of wavelength for several photovoltaic semiconductors................................................................................................................. .77 2-3 Proposed single-junction InxGa1-xN solar cell which has be en modeled by Medici..........78 2-4 Solar cell structures used in single-junction sola r cell simulation.....................................78 2-5 Single-junction absorber optimization steps......................................................................79 2-6 Simulated solar cell efficiency vs. band gap energy of the refined InxGa1-xN cell structure for AM0 and AM1.5 illumination.......................................................................79 2-7 Power vs. load plots for the refine d solar cell structure for AM0 and AM1.5 illumination................................................................................................................... .....80 2-8 Current density vs. voltage curves for th e refined solar cell structure under AM0 and AM1.5 illumination............................................................................................................80 2-9 The efficiency and open circuit voltage as a function of junction number for multi-junction InxGa1-xN solar cells...................................................................................81 2-10 Energy band diagram showing the generati on of electron-hole pairs and hot carriers from incident light energy..................................................................................................81 2-11 Phase-separated InxGa1-xN p-n junction showing Ga-rich InxGa1-xN precipitates (green circles) in an In-rich InxGa1-xN matrix (yellow bulk).............................................81 2-12 Corresponding band diagram from the p-t ype region in Figure 2-11 showing the In-rich InxGa1-xN matrix and the Ga-rich precipitate (ppt)................................................82 2-13 Band diagram showing trapping of carri ers and recombination in narrow band gap energy precipitates (ppt) in the wider band gap energy Ga-rich InxGa1-xN matrix...........82 3-1 Solid-gas interfacial region for the reaction of TMGa and NH3 to form GaN................120 3-2 The MOCVD reactor.......................................................................................................121 3-3 Scanning electron microscope image of dir ect InN growth on Si (111) with no in-situ surface pretreatments.......................................................................................................121 3-4 Scanning electron microscope image of InN grown on Si (111) using substrate nitridation followed by a LT InN buffer layer.................................................................122

PAGE 11

11 3-5 An XRD pattern of a polycrystalline In N thin film grown on Si (111) using a LT InN buffer layer................................................................................................................... ....122 3-6 A TEM image of the SiOxN1-x intermediate layer used to produce high quality single crystal GaN.................................................................................................................... ..123 3-7 X-ray diffraction patterns of InN grown on Si (111) and Si (100) substrates using substrate nitridation followed by a LT InN buffer layer..................................................123 3-8 Cross-sectional SEM image of InN/Si (100) grown by MOCVD showing a growth rate of 78 nm/hr............................................................................................................... .124 3-9 An XRD spectrum of InN/ Si (111) grown by H-MOCVD..............................................124 3-10 Scanning electron microscope images of InN/Si (111) grown by H-MOCVD...............125 3-11 Energy dispersive spectroscopy and AES analysis showing no evidence of chlorine contamination in InN/Si (111) by H-MOCVD................................................................125 3-12 Comparison of XRD patterns between aged and as-grown samples of InN on silicon and sapphire substrates....................................................................................................126 3-13 X-ray diffraction patterns for InN/c-Al2O3 annealed in a N2 (100 Torr) atmosphere at 500 oC (20 min), 525 oC (15 min), and 550 oC (10 min).................................................126 3-14 A series of XRD patterns for InxGa1-xN alloys grown at low temperature (530 oC) on c-Al2O3 substrates............................................................................................................127 3-15 Experimental InxGa1-xN compositions vs. inlet flow ratio (blue line).............................128 3-16 X-ray diffraction patterns of InxGa1-xN thin films grown on c-Al2O3 substrates at constant inlet flow ratio (In/(In+Ga) = 0.8) at temperatur e ranging from 530 to 770 oC (blue lines)...........................................................................................................128 3-17 Magnitude of InxGa1-xN (002) peak intensity vs. deposit ion temperature at a constant inlet flow ratio of In/(In+Ga) = 0.8..................................................................................129 3-18 Film crystalline quality as determined by XRD FWHM of the InxGa1-xN (002) peak....129 3-19 Magnitude of In-rich InxGa1-xN (101) and Ga-rich InxGa1-xN (002) peak intensity vs. deposition temperature at a constant in let flow ratio of In/(In+Ga) = 0.8.......................130 3-20 A series of XRD patterns for InxGa1-xN alloys grown at low temperature (530 oC) on a-Al2O3 substrates............................................................................................................130 3-21 Comparison of XRD spect ra for InN grown on a-Al2O3 (blue) and c-Al2O3 (red) substrates..................................................................................................................... .....131

PAGE 12

12 3-22 X-ray diffraction FWHM InxGa1-xN/a-Al2O3 at different values of inlet flow ratio and the corresponding film stability................................................................................131 3-23 Hemholtz free energy of mixing of InxGa1-xN as a function of InN mole fraction.........132 3-24 X-ray diffraction patterns for InxGa1-xN alloys grown at different inlet flow ratios on LT InN/GaN/c-Al2O3.......................................................................................................132 3-25 Growth rate of InxGa1-xN/c-Al2O3 alloys as a function of total inlet group III flow (sccm)......................................................................................................................... ......133 3-26 Composition deviation of InxGa1-xN/c-Al2O3 measured by RBS as a function of inlet flow ratio..................................................................................................................... .....133 3-27 Time domain THz measurement system used to analyze InxGa1-xN thin films...............134 3-28 Terahertz signal measured from the surface of pure InN and InxGa1-xN thin films........134 3-29 The emitted THz frequency range and corresponding amplitude for InN and InxGa1-xN alloys...............................................................................................................135 3-30 Poor THz signal to noise ratio for thin (48 nm) In0.6Ga0.4N............................................135 4-1 Schematic of the standard horizontal inle t tube and its position with respect to the graphite susceptor............................................................................................................159 4-2 Schematic of the vertical quartz inlet tube and its relativ e position with respect to the graphite susceptor............................................................................................................160 4-3 Inlet tube dimentions for film and nanostructured growth..............................................160 4-4 Schematic of the VLS mechanis m for InN nanowire growth by MOCVD....................161 4-5 Overhead view of vertical inlet tube showing the hot spot deposition zone (orange) and the region where small In metal pr imary nucleation is likely to form......................161 4-6 Experimental verification of the proposed nanowire density vari ation due to initial indium droplet nucleation size.........................................................................................162 4-7 Effect of In metal wetting for InN nanowire growth on Si (100) and GaN substrates....162 4-8 Indium-Silicon binary phase diagram..............................................................................163 4-9 Scanning electron microscope images of InN nanowires grown on p-Si (100) substrates using nitridation followe d by a LT InN buffer pretreatment..........................163 4-10 Scanning electron microscope image of a continuous polycrystalline InN thin film grown with the horizontal inle t arrangement on p-Si (100).............................................164

PAGE 13

13 4-11 Effect of post growth ann ealing on InN nanowire tip morphology.................................164 4-12 Scanning electron microscope images of InN nanowires (experiment number 68) showing uniform dense cove rage of the substrate...........................................................165 4-13 Scanning electron microscope images of InN nanowires (experiment number 69) showing less dense substrate coverage with uniform patches of nanowires...................165 4-14 Scanning electron images of InN na nowires (experiment number 70) showing uniform coverage of the substrate with nanowire patches that vary in size and nanowire density per patch..............................................................................................166 4-15 Scanning electron microscope images of InN nanowires (experiment number 71) showing less ordered nanowire growth with significant nano wire branching.................167 4-16 Scanning electron microscope image of unsuccessful InN nanowire growth on GaN/c-Al2O3 substrate.....................................................................................................167 4-17 Successful growth of InN nanowires on GaN/c-Al2O3 substrates by activating the surface of GaN.................................................................................................................168 4-18 Scanning electron microscope imag e of InN nanowires grown on GaN/c-Al2O3 substrate at reduced in let flow velocities.........................................................................168 4-19 Scanning electron microscope images of InN nanowires grown on GaN/c-Al2O3 using two-step nucleation and growth approach..............................................................169 4-20 Scanning electron microscope images depicting the InN core-In2O3 shell structure seen from broken nanowires............................................................................................170 4-21 X-ray diffraction spectra of InN core-shell nanowires on Si (100) substrate indicating single crystal InN a nd polycrystalline In2O3 phases........................................................171 4-22 Energy dispersive spectr oscopy spot analysis on the outside of a core-shell nanowires where indium, nitrogen, and oxygen are detected as well as silicon from the underlying substrate...................................................................................................171 4-23 Transmission electron microsco pe images of a single InN-In2O3 core-shell nanowire...172 4-24 Selected area diffraction pattern s of the InN core of an InN-In2O3 core-shell nanowire....................................................................................................................... ....173 4-25 Energy dispersive spectroscopy showing no oxygen detected for nanowires grown at high inlet N/In ratios and reduced flow velocities...........................................................173 4-26 Scanning electron microscope image of single InNIn2O3 core-shell nanowire contacted by Ti/Al/Pt/Au pads for electrical measurements............................................174

PAGE 14

14 4-27 Current-voltage measurements for several InNIn2O3 core-shell nanowires, used for determining the total resistance (RT) of each wire...........................................................174 4-28 The total measured resistance versus l/r2 for InNIn2O3 nanowires................................175 4-29 Scanning electron microscope images and dimensions of single InNIn2O3 core-shell nanowires used in the contact a nd sheet resistivit y calculations.....................................175 4-30 Optical micrograph of a single InNIn2O3 core-shell nanowire after wire bonding.......176 4-31 Current-voltage characteristics a nd time dependent response of InNIn2O3 single wire device under N2 and 500 ppm H2 (in N2) ambients.................................................176 4-32 Current-voltage characteristics of single InNIn2O3 core-shell nanowires before and after coating with 3 or 10 of Pt....................................................................................177 4-33 Time dependent current response of a single InNIn2O3 core-shell nanowires in a 500 ppm H2 ambient with different thicknesses of Pt......................................................177 5-1 Diagram of the H-MOCVD concentric inle t tube showing the arrangement of source material flows and typical temperatures used experimentally for the source and deposition zones...............................................................................................................189 5-2 A SEM image of the growth habits of InN nanorods illustrating flower-like growth and c-axis orientation.......................................................................................................189 5-3 Low resolution TEM image of an InN nanorod grown on a-Al2O3 substrate showing flat hexagonal cross section.............................................................................................190 5-4 Low resolution TEM image of an InN nanorod grown on r-Al2O3 substrate..................190 5-5 High resolution TEM image of the rectangular shape which is seen at the end of the nanorod when it is shaved fr om the substrate surface.....................................................190 5-6 Low resolution and high resolution TEM images of the InN nanorod tip grown on r-Al2O3.............................................................................................................................191 5-7 High resolution TEM images showing plan e bending in the tip of an InN nanorod grown on r-Al2O3.............................................................................................................191 5-8 Low resolution TEM images of InN nanor ods grown on Si (111) substrate showing asymmetric pencil tip.......................................................................................................192 5-9 High resolution TEM images of side f acet roughness occurring for most nanorods grown by H-MOCVD regardless of the substrate............................................................192 5-10 Low resolution TEM image and CBED patterns of an InN nanorod grown on a-Al2O3.............................................................................................................................193

PAGE 15

15 5-11 Experimental and calculated CBED images indicating nitrogen polarity.......................193 5-12 Electron energy loss spectra with and without the detection of oxygen..........................194 5-13 High resolution TEM image showing thin amorphous oxide layer that sometimes forms on the outside of the InN nanorods........................................................................194 5-14 Photoluminescence spectra of InN na norods grown on different substrates...................195 5-15 Scanning electron microscope images of InN microrods grown on Si (111) substrate for 90 min using Cl/In = 4.0, N/In = 250, and Tg = 600 oC.............................................195 5-16 Scanning electron microscope images of InN microrods grown on Si (111) substrate for 90 min using Cl/In = 5.0, N/In = 250, and Tg = 650 oC.............................................196 5-17 A SEM image of InN microrods grown on Si (111) substrate for 90 min using Cl/In = 5.0, N/In = 200, and Tg = 700 oC.........................................................................196 5-18 Scanning electron microscope image of InN microrods grown on Si (111) substrate after post growth in-situ HCl gas etching........................................................................196 5-19 Current-voltage measurements for severa l InN microrods, used for determining the total measured resistance (RT) of each rod......................................................................197 5-20 Total measured resistance vs. length/a2 for InN microrods.............................................197 5-21 Scanning electron microscope images and dimensions of InN microrods used for TLM analysis...................................................................................................................198 6-1 Comparison of GaN/c-Al2O3 grown at similar growth c onditions with the horizontal and vertical inlets............................................................................................................ .205 6-2 X-ray diffraction pattern of GaN grown with the vertical in let on Si (100) substrates...205 6-3 Scanning electron microscope images of GaN grown at 750 oC and an inlet N/Ga ratio of 3,000 on different substrates...............................................................................206 6-4 Scanning electron microscope imag es of GaN microtubes grown at 750 oC and an inlet N/Ga ratio of 3,000 on Si (100) substrates..............................................................206 6-5 Scanning electron microscope images of GaN grown using the vertical inlet at 650 oC and N/Ga ratio of 3,000 on different substrates...................................................207 6-6 Representative EDS spectrum of single GaN microtube grown on Si (100) substrate...208 6-7 Scanning electron microscope images of GaN grown with the vertical inlet at the lowest temperature tested (560-600 oC) showing excess of metal droplets on the surface of the nanotubes...................................................................................................208

PAGE 16

16 6-8 Scanning electron microscope images of GaN growth with the vertical inlet at 560 oC after a 10 min 10% HCl wet etch showing nanotube structure........................................209 6-9 Energy dispersive sp ectrum of GaN nanotubes after HCl wet etching...........................209 6-10 Scanning electron microscope images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 20,000...............................................................................................210 6-11 A SEM images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 10,000...210 6-12 Scanning electron microscope images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 7,000.................................................................................................211 6-13 Scanning electron microscope images of curly GaN nanowires that form outside the primary deposition zone of the vertical inlet...................................................................212 6-14 Wire bonding step of GaN nanotube H2 gas sensing device where damage to the nanotubes occurred..........................................................................................................212

PAGE 17

17 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy METALORGANIC CHEMICAL VAPOR DEPO SITON OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN FILMS AND NANOSTRUCTURES FOR ELECTRONIC AND PHOTOVOL TAIC APPLICATIONS By Joshua L. Mangum May 2007 Chair: Timothy Anderson Cochair: Olga Kryliouk Major Department: Chemical Engineering Single and multi-junction InxGa1-xN solar cell devices were modeled in one dimension using MEDICI device simulation softwa re to assess the potential of InxGa1-xN-based solar cells. Cell efficiencies of 16 and 27.4%, under AM0 illu mination were predicted for a single and a 5-junction InxGa1-xN solar cell, respectively. Phase separation of InxGa1-xN alloys is determined to have little to no negative effects on the solar cell efficiency. InxGa1-xN alloys were grown by MOCVD over the entire compos itional range (0 x 1) and phase separation was analyzed with respect to substrate material and gr owth temperature. A low deposition temperature of 530 oC was used to produce metastable InxGa1-xN/c-Al2O3 thin films over the entire compositional range, which was demonstrated for the first time by MOCVD. The use of higher deposition temperat ure and closely lattice matched substrates resulted in phase separated films. Substrates with a larger lattice mismatch (c-Al2O3) introduce strain in InxGa1-xN which helps to stabilize the film, however, at the expense of crystalline quality. Growth of InN nanowires by MOCVD was cont rolled without the use of templates or catalysts by varying the inlet flow pattern, N/In ratio, growth temper ature, and substrate material.

PAGE 18

18 A VLS growth mechanism is proposed, however, a VS growth mechanism can be achieved at high N/In ratios. SEM and TEM analysis revealed a core-shell nanowire structure with a single crystal InN core and a poly-crystalline In2O3 shell. Nanowire growth occurs along the [0002] direction with diameters and lengths ranging from 100 to 300 nm and 10 to 40 m, respectively for a 1 hr growth. H-MOCVD growth of InN nanoand microrods occurred on di fferent substrates and the nanorod structure was studied by TEM. The polarit y of the substrate directly affected the nanorod tip shape and prismatic stacking faults ar e suggested as the cause for the flower-like growth habit. Variation of growth parameters, such as temperature, N/In ratio, and Cl/In ratio proved to be ineffective at chan ging the aspect ratio of the nanor ods. Increased growth duration produces microrod size dimensions regard less of the chosen growth conditions.

PAGE 19

19 CHAPTER 1 INTRODUCTION 1.1 Group III-Nitrides The III-nitrides and their alloys exhibit direct band gap values ranging from infrared to ultraviolet wavelengths, which make them im portant for applications in electronic and optoelectronic devices. Gallium nitride (GaN ) and its alloy with indium nitride (InxGa1-xN) have become dominant materials for producing high brightness light emitting diodes (LEDs) and laser diodes (LDs) that emit light in the blue region of the visible spectrum. Research on III-nitrides has resulted in LEDs and LDs that emit light at visible, infrared (IR) and ultra-violet (UV) wavelengths. These optoelectroni c devices can be used for CD and DVD media, high brightness displays, and solid state lighting, which has been projected to reach se ven billion dollars by 2009.1 The wide band gap energy range also make s these materials candidates for absorber layers in solar cells since the absorption edge of these materials can be varied to optimize cell efficiency. III-nitride electronic devices are al so more environmentally friendly because they do not contain toxic elements such as arsenic, which is used to fabricate other compound semiconductors. 1.1.1 The III-Nitride Crystal Structure III-nitrides can exist as a cubic zincblende crystal structure, but it is thermodynamically more stable to exist in the he xagonal wurtzite crystal structure. The cubic and hexagonal phases differ only by the stacking sequence of close-pack ed III-N planes, and the energy difference between the two structures is small. Cha nging the sequence during growth produces defects such as stacking faults. The zincblende stru cture (space group F43m, number 216) has an ABCABC stacking sequence of (111) close packed pl anes (Figure 1-1). Th e wurtzite structure (space group P63mc, number 186) has an ABABAB stacking sequence with altern ating layers of

PAGE 20

20 close packed (0001) metal atoms (In, Ga, Al) a nd nitrogen pairs, where every other layer is directly aligned (Figure 1-2).2 The zincblende and wurtzite structures both co ntain polar axes and therefore lack inversion symmetry. The polarity type of grown films has been shown to significantly affect bulk and surface properties.3,4 The bonds in the <0001> direction (wurtzite) and <111> (zincblende) are cation-faced while the opposite direc tion is nitrogen-faced (Figure 1-3).5 As a general rule the term termination, such as N-terminated, is reserved for describing a surface property. For example the surface of an indium nitride film can be N-terminated by depositing one monolayer of nitrogen atoms, but the orientation of the crystal remains unchanged.6 GaN has been shown to be Ga-faced when deposited by MOCVD on sapphire substrates and N-faced when deposited by MBE or HVPE on sapphire substrates.7-10 Ga-faced GaN films usually have an atomically smooth surface while N-faced GaN films have a rough surface morphology.11 The growth parameters such as substrate material, substrat e surface treatment, and N/ III ratio are important factors in controlling the polarity.6 1.1.2 Properties of InN and GaN The properties of InN and GaN will be discusse d in this section and AlN will be ignored since it is not the primary focus of this work. Some fundamental prope rties of wurtzite and zincblende InN and GaN are shown in Table 1.1. The lattice parameter of InxGa1-xN can be found by using Vegards law as a function of the composition (x) (Eq. 1-1). The InxGa1-xN film composition can be determined experimentally by techniques such as X-ray diffraction (XRD) and Rutherford Backscattering (RBS). GaN InN N Ga Ina x xa ax x) 1 (1 similarly for c-axis (1-1)

PAGE 21

21 Electrical properties. As grown InN and GaN typical ly have large background n-type carrier concentrations, which make s p-type doping of these materials difficult. GaN has been more heavily studied than InN and the GaN film quality has progressed such that the background carrier concentration is low (4 x 1016 cm-3) with relatively high electron mobility (600 cm2/V-s).12 With these advances p-type doping of GaN has been cons istently reproduced for a number of years. It is know that the ab ility to dope semiconductors depends on the position of the valence and conduction ba nd edges with respect to a co mmon energy reference (i.e. the Fermi level stabilization energy, EFS).13 For this reason it is assumed that InN would be easier to p-type dope than GaN because the position of the valence band edge is 1.1 eV closer to the EFS.14,15 InN however still proves to be very difficult to p-type dope. Evidence has been recently presented for bulk p-type doping of InN with magne sium by isolating the effects of the n-type surface accumulation layer that occurs in InN structures.16,17 It is also believed that only a small fraction of Mg acceptors are ionized at room temp erature, but these recent results are promising for the development of InN p-n junction device s. Theoretical mobilit ies and controlled doping ranges of InN and GaN are listed in Table 1-2. InN and GaN have been shown to have very high peak drift velocities at room te mperature, higher than that of GaAs, and unlike GaAs the drift velocities were shown to be fairly in sensitive to temperature (Figure 1-4).18,19 1.1.3 Growth of InN and InxGa1-xN Indium Nitride Thin Films. InN was first demonstrated by Juza and Hahn in 1938 from a InF6(NH4)3 precursor and the resulting InN had a wurtzite crystal structure.20 In the following years, attempts at producing InN were made by several researchers.21-24 The resulting InN samples were typically powders or small crystals and were prepared by ammonization of indium containing molecules or thermal decomposition of more complex molecules containing indium and nitrogen. These early experiments revealed that indium containi ng molecules would not

PAGE 22

22 react with inactive N2 molecules, even at higher temperature25 and that the equilibrium nitrogen vapor pressure was very high.24 During the 1970s and 80s polycry stalline InN films were being deposited mainly by sputter deposition. Growth on silicon and sapphire s ubstrates was done at temperatures ranging from room temperature to 600 oC with typical elec tron mobility of 250 50 cm2/V-s and carrier concentratio ns in the range of 5-8 x 1018 cm-3.26 Analysis of the optical properties of InN corresponded to a band gap energy value of 2.0 eV.27-30 During this period Tansley and Foley extensively studied RF sputtering of InN and its properties31-36 and they have grown films with the highest mobility and lowest intrinsic carrier con centration to date (2700 cm2/V-s and n = 5 x 1016 cm-3).31 It has been shown that InN di ssociates at temperature greater than 600 oC.28 During the 1990s single crystal growth of In N became more prevalent, mostly by MOCVD and molecular beam epitaxy (MBE) growth tech niques and subsequent films properties were dramatically improved. Single crystal InN by MOCVD on sapphire substrates was first grown by Matsuoka et al using trimethyl indium (TMI n) and ammonia precursors.37 At about the same time Wakahara et al also demonstrated MOCVD single cr ystal InN on sapphire substrates by reacting TMIn with microwave activated nitrogen.38,39 By 2002, single crystal InN films with low background carrier concentrations and high mobilities had been produced, 5.8 x 1018 cm-3 and 730 cm2/V-s (for MOCVD)40 and 3.49 x 1017 cm-3 and 2050 cm2/V-s (for MBE).41 These were significant improvements in InN film quality at the time because it was difficult to grow InN films by MOCVD and MBE with bac kground carrier concentrations below 1019 cm-3 with mobility greater than 300 cm2/V-s. By 2002 many researchers starte d to question the fundamental band gap energy value of InN, because photoluminescence and absorbance data suggested that InNs band gap energy was

PAGE 23

23 closer to 0.7 eV,42-49 instead of the previously accepted va lue of 2.0 eV. Today, the value of the band gap energy of InN is still not agreed upon; although there is compelling evidence that the lower band gap energy value is correct. There have been several reviews and recent papers that try to reveal reasons for the discrepa ncies in the InN band gap energy value.50-55 The theories that receive the most attenti on for band gap energy modulation in InN are oxygen incorporation in the films, Moss-Burstein effect, trapping leve ls, and film stoichiometry. Indium oxide has a band gap energy of 3.2 eV, therefor e it is assumed that oxynitrides are formed during growth and increase the band gap energy. Th e Moss-Burstein effect, which wa s first studied by Trainer and Rose,28 occurs when the Fermi level is pushe d into the conduction band as the electron concentration increases above th e Mott critical density and th erefore the band gap energy is overestimated by optical absorption. The Mott cri tical density is reached when an electron gas cannot sustain itself due to re duced electron screening in order to form bound states.56 A source for under estimation in the InN band gap energy fr om PL measurements of InN films can be linked to deep level trap s, with activation energi es on the order of 0.6-0.7 eV. The defect level explanation is reinforced by the fact the highest mobili ty of InN to date (2700 cm2/V-s)31 is much lower then the theoretically pr edicted maximum value of 4400 cm2/V-s,57 suggesting a high concentration of compensated defects.58 Non-stoichiometric films are potential causes fo r band gap energy variation in InN. In-rich film stoichiometry leads to the formation of deep level defects (0.7 eV) when indium aggregates form and these defects states can se rve as potential photoluminescence peaks.59 A low band gap energy value of 0.7 eV can also be the result of Mie resonances from indium precipitates. It is well known that optical losses occur from resona nt light scattering and absorption by dispersed metallic particles.60 MBE grown In-rich InN samples showed that bipolar absorption,

PAGE 24

24 accompanied with transformation into heat occurred in indium clusters, as well as resonant light scattering due to plasmon excitations.61 Absorption from these indium precipitates makes it difficult to determine the true band gap energy value when using optical absorption measurements. Current evidence for the InN band gap energy values range from 0.6 to 2.3 eV, suggesting that a variety of fact ors are involved and that no singular source is responsible for the disparity in the InN band gap energy.50 The band gap energy of InxGa1-xN can be inferred from a linear interpolation of the compound band gap energy values and using the bowing parameter (b, determined experimentally), as seen in Eq. 1-2. Recent results by Davydov et al. have shown the InxGa1-xN bowing parameter to be 2.5 eV,43 while other researchers have presented evidence for a bowing parameter that is a function of composition,62 as represented by Eq. 1-3. ) 1 ( ) ( ) 1 ( ) ( ) ( x bx GaN E x InN xE x Eg g g (1-2) ] 4 19 4 11 )[ 1 ( ) ( x x x b eV (1-3) There are several growth challenges that ma ke it difficult to produce high quality single crystal InN films by MOCVD. Of the III-nitrides, InN is by far the most difficult to grow due to its high equilibrium nitrogen va por pressure (Figure 1-5).63 The high equilibrium vapor pressure of InN limits the deposition temperature to less than 650 oC to prevent film decomposition.64 The source materials typically used in MOCVD growth of InN are TMIn and NH3. At these lower deposition temperatures, the extent of amm onia decomposition is very low, less than 0.1% at 500 oC.12 Due to this lack of reactive nitroge n, indium droplets can form on the surface, therefore the inlet N/In ratio must be kept sufficiently high (~50,000) to avoid formation of indium droplets.65

PAGE 25

25 High inlet N/In ratios are only requ ired for growth at temperature 600 oC since ammonia decomposition occurs readily at higher temperature ( 650 oC). The extent of decomposition of ammonia, however, significantly increases the H2 partial pressure, which has been shown to retard the InN growth rate.66 This is also the reason that a ni trogen carrier gas is preferred over a hydrogen carrier gas. Other nitr ogen sources such as hydrazine (N2H4) have been suggested from the results of equilibrium calcula tions for the growth of III-nitrides.67 The analysis showed that the growth rate of InN could be increase d without the formation of indium droplets if hydrazine was used. From a practical stand poin t ammonia is still used most frequently in MOCVD growth of InN due to the e xplosive nature of hydrazine. With these growth challenges, there is a narrow temperature window (400 650 oC) for successful growth of InN by MOCVD. For conve ntional MOCVD the growth temperature is the most important parameter for controlling film prope rties such as crystalline quality, growth rate, surface morphology and carrier concentration.68 Modified MOCVD depos ition techniques, such as plasma or laser assisted MOCVD are starting to gain popularity due to the ability to produce more reactive nitrogen at low growth temperatur es, thus avoiding some of the pitfalls of conventional MOCVD. High quality growth of InN is also hindered by th e fact that there are no substrates that are appropriately matched for lattice co nstant and thermal expansion coe fficient. Sapphire substrates (c-Al2O3) are most frequently used for growth of InN even though there is a large lattice mismatch, 26%. A-, C-, and R-plane sapphire substr ates were used for selected growths in this work and the different crystal orientati ons of sapphire are shown in Figure 1-6.5 Silicon substrates are becoming more popular due to thei r lower cost and potential for future device integration.69 Compared to sapphire, silicon substrates have a much lower lattice mismatch (8%

PAGE 26

26 for Si(111)) with InN, but the resulting InN films are usually polycrystalline due to an amorphous SiNx layer that forms at the Si surf ace during the initiation of growth.70 Typical substrates used for growth of InN and th eir corresponding lattice mismatches and thermal expansion coefficient differe nce are shown in Table 1-3. A large lattice mismatch and difference in ther mal expansion coefficient can lead to a large number of structural defects. To reduce the number of defects in the heteroepitaxial InN, substrate nitridation (for c-Al2O3) and buffer layers are used to improve film quality. Buffer layers are a two step growth me thod that is commonly used in he teroepitaxy, which consists of a low temperature nucleation layer followed by the main epitaxial layer. The buffer layer serves to change the nucleation process to promote lateral growth of the subsequent film. For sapphire substrates initial nitridation as well as buffer layers are used to improve crystal quality. Substrate nitridation forms AlN nuclei on the c-Al2O3 surface, and the lattice mismatch is reduced from 26% to 14% for InN/AlN.70-72 Complementary to nitridation, InN,73,74 GaN,75,76 and AlN77 buffer layers are used to improve InN film qua lity on sapphire substrates. InN films with the lowest background carrier concentration and high est electron mobility to date were grown by MBE (3.49 x 1017 cm-3 and 2,050 cm2/V-s) using a GaN buffer layer on sapphire.41 The best MOCVD results to date are a carrier con centration and electron mobility of 5.8 x 1018 cm-3 and 900 cm2/V-s, respectively, which were grown on c-sapphi re substrates at at mospheric pressure.78 Similar to sapphire substrates, the crystalline qu ality of InN on Si substrates has also been improved by the use of buffer layers.69,77,79,80 Indium Nitride Nanostructures. Since Iijima discovered carbon nanotubes,81 there has been a large interest in developing one-dime nsional (1D) structures such as nanowires, nanorods, nanotubes, and nanobelts from other mate rials. Nanostructur es are unique because

PAGE 27

27 dimensionality and size confinement affects electrical, optical, and structural properties. Contrary to nano-films, the a dditional confinement dimension of nanowires allows carrier confinement along a specific conducting path. III-n itride nanowires have potential applications in low power field-effect transistors (FETs), LEDs, solar cells, terahertz emitters and detectors.82,83 These types of nanostructures are s ynthesized by a variety of physical and chemical methods. The first InN nano wires were demons trated by Dingman et al by decomposition of azido-indium precursors.84 The nanowires had lengths ranging from 100-1000 nm with an average diameter of 20 nm and the growth was attributed to a precursor solutionliquid-solid (SLS) mechanism. The growth of InN nanowires and nanorods has been reported by a number of researchers.85-106 Similar to Dingman et al. other researchers have used sol vothermal methods to grow single crystal InN nanostructures at relatively low temperature (300 oC).88 InN nanowires have been synthesized over a wide range of temperature, and as high as 700 oC.87,90,96,103,105 Other investigators have synthesi zed InN nanowires at 420 oC,89 450 oC,99 440-525 oC,97 500 oC,91,92 550 oC,95 600 oC,102 550-700 oC,104 and 600-730 oC.96 A variety of precursors have been used for InN nanowire growth. Single source precursors88,95 are less common and most InN nanowire synthesis uses separate indium and nitrogen precursors. Solid sources are typically used for the indium precursor, while nitrogen precursors are gas sources. Indi um precursors are pure indium metal,85,89,90,92,97,101 trimethyl indium,106 indium oxide powder,87,93,96 or a combination of both indium and indium oxide.98,103,105 The nitrogen source is typically NH3 but activated N2 has also been used.89,97 Most InN nanowire growth processes occur vi a vapor-solid (VS) or vapor-liquid-solid (VLS) mechanisms. As previously mentione d InN nanowire growth via SLS mechanism has

PAGE 28

28 been done, however the majority of synthesis do es not proceed through this route. The VLS mechanism uses a metal particle that acts as a ca talytically active site th at promotes growth of the nanowire. The metal catalyst forms a liquid alloy with indium, where by gas phase reactions occur with the liquid metal alloy to form a solid nanowire. The liquid alloy attracts indium vapor which leads to solid precipitation once the a lloys reaches a super saturation point. A metal catalyst such as Au85,92 or Ni 91 have been used to create a VL S mechanism for growth and it has also been suggested that indium acts as a ca talyst (unintentionally) when no catalyst other catalyst is used.87,90,93,99 When no metal droplet is present at the end of the nanowire it is assumed that the reaction pro ceeded through a VS mechanism.103 Figure 1-7 shows a schematic of the VLS mechanism,107 as well as experimental pictures of nanowires with92 and without103 metal droplets at the end. InN nanowires have been grown on se veral different substrates, Si/SiO2,98 Si(100),85,92 Al2O3(0001),89 polycrystalline AlN and GaN,99 or no intentional substrate at all.87,90 In the case of growth without an intentional substrate, nanowire samples are scratched from reactor walls or precursor crucibles and subsequently characterized. The diameter of InN nanowires range from 10 to 500 nm with lengths on the order of 1 to 100 microns and the growth rate varies signi ficantly depending on th e type of deposition technique used.84,90-92,95,96,98,103 Nanowire properties, such as band gap energy, also vary depending on the type of growth method use d. The band gap values range from 0.7 to 0.9 eV,85,90,97 1.1 eV,89 and 1.7 to 1.9 eV,92,93,104 and some researchers have even reported low (0.8 eV) and high (1.9 eV) band gap energy valu es for InN nanowires grown by the same method.91,99 Lan et al.90 produced InN nanorods on Si(100) s ubstrates using a gold catalyst where the diameter of the nanorod influenced the band gap energy. No conclusions were

PAGE 29

29 presented for the difference in the band ga p energy with respect to nanorod diameter.91 Surface electron accumulation and the Moss-Burstein effect are likely possibilities for this band gap energy variation. Vaddiraju et al.99 suggested the higher band gap energy values of InN nanowires are from oxygen incorporation into the nanowires, since samples often contain mixtures of In2O3 and InN. Zhang et al.105 and Yin et al.101 have demonstrated single step growth methods that produce core-shell nanowire structures, InN core-In2O3 shell and InN core-InP shell, respectively. Core-shell structures offer the ability to study how interfacial states affect nanowire properties and progress for developing futu re radial heterostruct ure nanowire devices. Qian et al.108 has grown GaN-based radial core-shell LED heterostruct ures that emit light over wavelengths from 365 to 600 nm with high quantum efficiencies. For InN nanowire device technology to progr ess several growth challenges must be overcome. As in the case of InN films some fu ndamental properties of InN nanostructures, such as the band gap energy, need to be studied. A va riety of InN nanowire synthesis processes have been reviewed in this section, but it is important that deposition techniques be implemented with current technologies. It is al so important that InN nanostruc tures be reproducible and the deposition be precisely controlled to make progress towards more complex devices. The ability to p-type dope InN nanostructures will also be important for future device applications. Indium Gallium Nitride Thin Films. The first InxGa1-xN alloys were grown by Osamura et al .27 and optical absorption measurements we re given to reveal the relationship between band gap energy and alloy compositi on and theoretical bowing parameter of 1.05 eV was determined. Osamura et al later noticed that InxGa1-xN alloys phase separated after annealing in an argon atmosphere at 700 oC.29 Nagatomo et al.109 grew InxGa1-xN alloys by

PAGE 30

30 MOCVD on sapphire substrates with 0 x 0.42 at 500 oC, while Yoshimoto et al.110 produced single crystal InxGa1-xN alloys on sapphire substrates with 0 x 0.22 at 800 oC. The higher temperature growth of InxGa1-xN sufficiently improved the quality to allow for PL to be observed for the first time from an InxGa1-xN alloy. The primary focus of InxGa1-xN research has been for Ga-rich solutions that are important for applications in light emitting diodes (LEDs). Shuji Nakamura pioneered the development of visible LEDs in the early 1990s, especially high brightness bl ue LEDs based on III-nitride heterstructures.111 Adding small amounts of indium to GaN became the optimal material for active layers in LEDs. InxGa1-xN active regions are usually hi ghly defective due to a large number of threading dislocations, yet the LEDs remain highly efficient.112 Other III-V compound semiconductors have a much more sensitive relationship between extended defect density and device performance. Blue, green, amber, and UV LEDs have been demonstrated using InxGa1-xN active regions in either double hetero structure or quantum well structures.113 The indium composition of these active layers are usually less than 10%, with the exception of some quantum well structures where the indium content can be as high as 45%. Considerably less studies of In-rich InxGa1-xN alloys have been made b ecause AlGaAs and AlGaInP LEDs cover this wavelength range. LED devices typically use very thin layers of InxGa1-xN, as small as 25 and the indium content not usually greater than 50%. The InN band gap energy controversy, which started in 2002, has increased the intere st in In-rich InxGa1-xN alloys as a result of device applications now possible with the smaller band gap energy of InN such as full spectrum solar cells14 and terahertz emitters and detectors.114 Most of the recent research on In-rich InxGa1-xN has been focused on understanding the fundamental properties of the alloys,

PAGE 31

31 such as structural, electrical properties, and optical properties as well as improving film quality.115-122 There are several growth challenges that mu st be overcome for successful growth of InxGa1-xN alloys. Perhaps most challenging is phase separation occurs in the alloy due to an 11% lattice mismatch between InN and GaN. Ho and Stringfellow analyzed the solid phase miscibility gap in the InxGa1-xN system and determined the maximum equilibrium incorporation of In into GaN (or Ga into InN) is less than 6% at a typical depos ition temperature of 800 oC.123 The binodal and spinodal curves were calculated using a modified vale nce-force-field (VFF) model (Figure 1-8) to estimate the interacti on energy assuming the solid solution behaves as a regular solution. It can be seen that InxGa1-xN alloys are theoretically unstable or metastable over a large compositional range at typical grow th temperatures (Figure 1-8). The critical temperature at which complete miscibility exists was calculated to be 1250 oC, which is greater than the melting temperature of InN. Karpov pe rformed a similar calcula tion to predict phase separation of InxGa1-xN alloys and showed that the misc ibility gap could be reduced by introducing compressive lattice strain (com pared to unstrained films, Figure 1-9).124 Karpovs results have been experimentally confirmed as evidence for the observation of single phase InxGa1-xN with up to 30% In incorporati on at deposition temperature of 700-800 oC, for InxGa1-xN heterostructures or quantum dots.125-129 Phase separation can be identified by TEM analysis or by observing peak shouldering or separati on from XRD measurements (Figure 1-10). More recently researchers have show n that single phase metastable InxGa1-xN can be grown over the entire compos itional range by MBE when a low temperature is used.47,130 Compositional modulation also occurs in InxGa1-xN films where nano-domains of In-rich or Ga-rich sections can be formed.118,120,131,132 This non-uniform distri bution has been shown to

PAGE 32

32 affect the photoluminescence pr operties of the as grown film s by decreasing the FWHM and reducing the peak intensity.133 The vapor pressure difference between InN and GaN (Figure 1-5) is another problem that affects high quality growth of InxGa1-xN alloys. At lower depos ition temperature indium incorporation can be increased, however, higher crystalline quality is achieved at higher deposition temperature. Figure 1-11 shows how indium incorpor ation is affected by deposition temperature when using TMIn and TEGa.12 The distribution coefficien t of In between the vapor and sold phases is considerab ly greater than unity at 800 oC because of the large difference in decomposition pressure at elevated temperature along with near equilibrium conditions at the growth interface at this higher temperat ure. At the lower temperature of 500 oC the distribution coefficient is close to unity suggesting nonequilibrium (reaction limited) conditions are expected. It is also evident that at 800 oC, control of the composition becomes difficult for intermediate compositions given the rapid change in solid com position with that in the vapor. Choosing the optimal N/III inlet ratio is dir ectly affected by the specified deposition temperature. Ammonia decomposition efficiency will determine the actual N/III ratio, however it is very difficult to know the exact NH3 decomposition efficiency since the value relies heavily on the reactor design as well as temperature. For this reason the inlet flow ratio of NH3/(TMIn+TEGa) is commonly listed as the N/III ratio for MOCVD growth. In this work, a N/III ratio of 50,000 is frequently used; however the ac tual N/In ratio at the substrate surface will be lower, depending on the deposition temper ature used. For simplicity the N/III ratio mentioned in this work represents the NH3/(TMIn+TEGa) molar flow ratio into the reactor. When the deposition temperature is low ( 600 oC) the inlet N/III ratio must be high enough to maintain sufficient levels of ac tive nitrogen and avoid In droplet formation. As the temperature

PAGE 33

33 is increased above 650 oC the N/III ratio must be appropriate ly decreased so the excess hydrogen partial pressure doesnt inhibit In inco rporation into the film. Growth of InxGa1-xN alloys (especially for In-rich compositions ) is fairly difficult due to th e narrow growth regimes of InN coupled with phase separation and vapor pressure differences that occur with the addition of gallium to InN. Indium Gallium Nitride Nanostructures. Growth of InxGa1-xN nanostructures offers the promise for improving the efficiency of III-nitride based LEDs. The main loss of efficiency in III-nitride LEDs is through non-ra diative recombination due to th reading dislocations formed during GaN and InxGa1-xN film growth.134 Nanowire or nanorod grow th is a way to practically eliminate threading dislocatio ns and significantly reduce th e non-radiative recombination centers. InxGa1-xN nanostructures have been studied far less than InN nanostructures; there are only a few reports for InxGa1-xN nanostructured growth .135-139 The first InxGa1-xN nanostructures were produced by Kim et al. using hydride vapor phase epitaxy at a low temperature of 510 oC.138,139 These InxGa1-xN nanorods on sapphire (0001) substrates were approximately 50 nm in diameter, 10 m in length, and orient ed in the c-axis (also known as well aligned). All the InxGa1-xN nanorods were Ga-rich alloys with maximum indium composition of 20%. Chen et al.136,137 produced InxGa1-xN nanorings/nanodots and ordered InxGa1-xN nanolines/nanodots, respectively, by using selective area nitrid e growth on patterned SiO2 masks on GaN substrates. These InxGa1-xN nanostructures, approximately 80 nm in diameter (or across) were also Ga-rich showing PL emissions at 420 nm (2.95 eV)136 and 450-500 nm (2.76 2.48 eV).137

PAGE 34

34 Cai et al .135 was the first to demonstrate In-rich InxGa1-xN nanostructure growth. InxGa1-xN straight and helical nanowires were grown in a tube furnace using elemental Ga and In which were evaporated in an NH3/Ar flow and deposited on a Au covered Si(100) substrate. This method is similar to that has been used to produce InN nanowires.92 These InxGa1-xN nanostructures also show a high In-content core (30-60%) and low In-content shell (4-10%) with either a hexagonal or cubic structure. Possible m echanisms for the helical growth structure was attributed to lateral displacemen t of the Au catalyst compared to the nanowire central axis or differing growth rates between the core and shell of the nanowire. The core-shell structure was believed to occur due to phase separation of the InxGa1-xN alloy. Growth of InxGa1-xN nanostructures involves th e difficulties related to InxGa1-xN film growth such as compositional control and pha se separation as well as reproducibility and controlled synthesis related to nanowire growth. As prev iously mentioned growth synthesis can be precisely controlled using patterning of SiO2 masks, however, future device specifications may not always allow for pattering schemes to take place. 1.2 Photovoltaics Photovoltaics (PV) is the field of study where electricity is directly produced from solar radiation (sunlight). The phot ovoltaic effect was first discovered by Edmund Baquerel in 1839 when he observed the effect of light on silver co ated platinum electrodes that were immersed in an electrolyte.140 By the 1950s developments in the sili con electronics industry made it possible to fabricate silicon p-n junctions, which served as the first solar cell devices.141 During the 1950s and 60s silicon solar cell technology was develope d for space and satellite applications where delivery of fuel was difficult. Interest in sola r cells peaked again in the western world in the 1970s due to the oil-dependent energy crisis. During this time othe r cheaper solar cell technologies were explored, such as polycrysta lline Si, amorphous Si, thin film, and organic

PAGE 35

35 materials. The 1990s was a time where the need was recognized for renewable energy sources and alternatives to fossil fuels, promptly signif icant growth in solar cell research and technology occurred at this time. By the late 1990s th e photovoltaics industry was growing by 15-25% per year, and for the first time solar cell power ge neration became competitive with remote low power applications (rural electrificat ion, navigation, and telecommunications).142 Today solar cell technology and its implementati on is of global interest and some countries such as Germany and Japan are leading the way. By 2003 the solar electricity industry provided employment for over 35,000 people world wide (E uropean Photovoltaic Industry Association, Brussels, Belgium, EPIA Roadmap, 2004, http://www.epia.org/documents/Roadmap_EPIA.pdf accessed April 2007). From 2004 to 2005 the PV industry doubled and annual sales have surpassed $10 billion, which is drawing significant capital investment. U.S. implementation of solar cell modules has increased in recent years, however, the U.S. has been losing market share to other countries in Europe as well as Japan.143 The loss in market shar e is directly related to the lack of U.S. growth in grid connected system s, which has grown significantly (world wide) in recent years. One third of the 90 Megawatts (MW) that were installed in the U.S. in 2004 came from California alone.143 The progress of U.S. photovolta ic module implementation in recent years can be seen in Figure 1-12,144 and a comparison to world wide markets can be seen in Figure 1-13.143 Over 95% of current solar cell technologies in use are silicon based (Figure 1-14).145,146 There are however a variety of technologies a nd materials being activel y researched to produce more cost effective solar cells. Some of the new technologies being pursued are solar concentration systems that focus light onto solar cell absorbers, thin film organic and inorganic

PAGE 36

36 materials to reduce materials cost and processi ng time and multi-junction solar cells to absorb more of the solar spectrums light. The highest efficiency solar cells to date are based on III-V materials, specifically GaInP/GaInAs/Ge tandem cells that achieve efficiencies of 39%.146 InxGa1-xN alloys are currently only proposed as a potential material for hi gh efficiency solar cells. 1.2.1 Fundamental Physics of Solar Cell Devices For light energy to be converted into electricity a solar cell fi rst absorbs incident photons. If the energy of the photon is greater than the electronic band gap energy of the semiconductor then the energy of the photon is transferred to the semiconductor which generates electron-hole pairs. A built in electric field created by a p-n junction is used to separate the electron-hole pair, and the electrons and holes are co llected via external metal contact s of the solar cell device. A schematic of a p-n junction solar cell device, including light absorp tion and generation of electron-hole pairs, is shown in Figure 1-15.147 The p-n junction can be formed by like semiconductor materials (homojunctions) or unlike materials (heterojunctions ). Light absorption in the solar cell device depends on the absorption coefficient ( ), which is an inherent property of the chosen semiconductor material. Solar radiation, the raw material for photovol taic energy conversion, is emitted by the sun over a range of wavelengths from ultraviolet to in frared. The power density at the surface of the sun is 62 MW/m2 and the solar flux is reduced to 1335 W/m2 at the earths atmosphere, mostly due to the reduced angul ar range of the sun.141 The power density available on the surface of earth is reduced to approximately 900 W/m2 due to atmospheric absorption of light. The extraterrestrial and terrestria l solar spectrums are shown in Figure 1-16. The atmospheric absorption due to water molecules occurs primarily at 900, 1100, 1400, and 1900 nm and for

PAGE 37

37 CO2 molecules at 1800 and 2600 nm (Figure 1-16).141 The atmospheric absorption is quantified by using an air mass factor, nAirMass, which is defined as overhead directly Sun if length path optical Sun to length path optical nAirMasscosec ( s) (1-4) where s is the angle of elevation of the sun. From this definition the extraterrestrial solar spectrum corresponds to Air Mass 0 or AM0. The AM0 spectrum can also be modeled by assuming black body radiation at 5760K.141 The standard terrestr ial solar spectrum is AM1.5, however, actual solar irradiances varies dependi ng on the season, daily variation of the suns position, orientation of the earth, and cloud conditi ons. Average global solar irradiances vary from below 100 to over 300 W/m2, where the higher values are usually found at inland desert areas (Figure 1-17). The black disks located on the map represen t the area needed to meet todays total global energy suppl y, producing 18 terawatts of el ectricity (TWe), assuming a modest conversion efficiency of 8% (Figure 1-17). A solar cell device can be considered a forwar d biased diode whose current flows in the opposite direction of the built-in bias. The rectifyi ng behavior of solar cell devices is needed to separate the carriers that are ge nerated. A solar cells output power is determined by the open circuit voltage (Voc), short circuit current (Isc), and fill factor (FF). Figure 1-18a shows the I-V characteristics for a solar cell under dark and i lluminated conditions and Figure 1-18b shows the maximum power rectangle formed by the I-V characteristics where the boxed area is proportional to the power output of the cell.148 The maximum power, Pmax, is the area of the maximum power rectangle or the pr oduct of maximum power current (Imp) and maximum power voltage (Vmp), Pmax = Imp*Vmp. The intersection of Imp and Vmp is the maximum power operating point. The FF describes the squareness of the IV characteristics and is always less than one.147

PAGE 38

38 Equations for the fill factor and efficiency ( ) of a solar cell device is given by Eqs. 1-5 and 1-6, respectively. oc sc mp mp oc scV I V I V I P FF max (1-5) in oc sc in mp mp inP V I FF P V I P P max (1-6) Quantities such as FF, Isc, and Voc are the main parameters used to characterize a solar cell device, but these quantities are dependent on the semiconductor structural quality and electronic properties. Two objectives for obtaining an efficient solar cell involve minimizing recombination of generated electron-hole pa irs and maximizing the absorption of photons. 1.2.2 Why InxGa1-xN? With the discovery of a lower band gap energy of InN,42-49 the wavelength range of InxGa1-xN now spans from infrared (0.7 eV) to ultr aviolet (3.4 eV). This wavelength range covers virtually the entire solar spec trum and creates the opportunity for InxGa1-xN based high efficiency solar cells (Figure 1-19) To exploit the ability of InxGa1-xN to absorb light at multiple wavelengths it has been proposed to fabricate multi-junction tandem solar cells using different compositions of InxGa1-xN (Wladek Walukiewicz, 2002, Full Solar Spectrum Photovoltaic Material Identified, www.lbl.gov/msd/PIs/Walukiewic z/02/02_8_Full_Solar_Spectrum.html accessed April 2007). Typically most high effici ency tandem solar cells are based on monolithic In-Ga-As-P-based layers grown on germanium substrates.145 When compared to these systems, InxGa1-xN-based devices have has the ability to be grown on inexpensive silicon substrates and use fewer and less toxic elements, which make fa brication more flexible and cost effective. InN has been shown to have a high absorption coefficient, ~ 105 cm-1 (this work), which means only thin absorber layers are needed to ab sorb the majority of incoming photons. Thinner

PAGE 39

39 films mean reduced manufacture time and highe r throughput, therefore reduc ing final solar cell costs. Another benefit of InxGa1-xN alloys is that they are radiat ion hard, which is beneficial for space PV because it is difficult to replace sola r cells in space, so l ong device lifetimes are important.14 1.2.3 Current Progress of InN and InxGa1-xN Solar Cells Indium nitride (InN) as a possible photovoltaic (PV) material is a relatively new idea that was first suggested by Yamamoto et al in 1994.149 At that time it was suggested that InN be used as the top cell in a 2-j unction tandem solar cell since the band gap energy of InN (1.95 eV) was close to the optimum value for AM0 illumi nation. Throughout the 1990s little attention was given to InN as a PV material, instead research efforts focused on single crystal growth of InN on different substrates using mostly MOCVD and MBE growth techniques.68 Later Malakhov150 proposed a InN/Si heterojunction for a high perf ormance and cost effective solar cell (again considering a InN band gap energy value of ~ 2.0 eV). Modeling InxGa1-xN tandem solar cell structures ha s recently been done by Hamzaoui et al .151 Simulations were done for tw o, three, four, five, and six InxGa1-xN junction tandem solar cells. The efficiency ranged from 27%, for a two junction cell, to 40% for a six junction cell. A maximum theoretical efficiency grea ter than 50% could be achieved using the InxGa1-xN ternary alloy to produce a multi-junction solar cell (using optimum band gaps under concentration).152 InxGa1-xN heterojunctions on p-Si and p-Ge substrates were simulated by Neff et al. for In compositions ranging from 0.4 to 1.0.153 Best calculated cel l efficiencies under AM1.5 illumination were 18 and 27% for InxGa1-xN on p-Ge and p-Si, respectively. Pure InN heterostructures showed a reduced efficiency of 2.5%. Although the majority of single crystal InN growth has occu rred on sapphire substrates,68 growth on other substrates such as silicon or germ anium is preferred for PV applications. Silicon

PAGE 40

40 substrates are desired for their low cost and large area capabilities. Initia lly single crystal growth on Si substrates was not possible due to an amorphous SiNx layer forming during growth,68 however, more recent MBE results have produced single crystal InN and InxGa1-xN films on Si(111) substrates by performing brief silicon substrate nitridation for 3 min.154 These recent MBE results also demonstrated rectifying charact eristics of n-InN/p-Si (on most samples). Earlier CVD of n-GaxIn1-xN/p-Si films also showed rectifying behavior.155 Germanium substrates provide adequate lattice matching to InN (11.3% for InN(0001)/Ge(111)) and also can be used to fa bricate vertical conduction PV device designs, which lead to higher efficienci es than top-connected cells.156 Trybus et al. has recently produced single crystal InN on Ge(111) substrates using MBE,157 and also assessed some aspects of InN/Ge based PV.156 I-V tests of their n-InN/p-Ge s howed only ohmic behavior. In their assessment it is pointed out that without p-type doping of InN th e traditional tandem solar cell structure must be redesigned, specifically involving the tunnel junctions. Tunnel junctions are degenerately doped semiconductors that serve as interconnects between the absorber layers of the solar cells as well as collecti on areas for minority carriers. In efforts to bypass the need for p-InN, Trybus et al. suggests using a layer of epitaxial Al between the Ge substrate and InN layer as a sub-cell inte rconnect and collecting junction. The Al layer also prevents an In-Ge eutectic from forming at the interface.156 Another problem that hinders the progress of InN-based PV is the high background n-type doping. The high intrinsic electron concentration is one of the main issues that makes p-type doping more difficult. InN has displayed strong band bending at the surface and heterointerfaces,158-160 which coupled with the high n-type background makes it more difficult to form rectifying solid-state junctions. Very recently Jones et al. has presented evidence for p-type

PAGE 41

41 doping of InN using Mg acceptors.16 It was shown from capacitance-voltage measurements that Mg-doped InN contains a bulk p-type region that is electrically suppr essed by a thin surface inversion n-type layer. The surface accumulation layer is believed to be caused by donor-like surface defects and chemical or physical treatmen ts have been shown to be ineffective at removing this layer.159,161 Initial p-type doping results ar e promising for eventual progress towards InN based devices. With these presen t device limitations of In N only one reference can be found where InxGa1-xN solar cell devices have been tested.162 Two types of devices were tested, p-i-n solar cells with i-In0.07Ga0.93N layers (using n,p-GaN cappi ng layers) and a quantum well device with In0.4Ga0.6N layers as the wells. The intern al quantum efficiency (IQE) results for the two p-i-n cells and the quantum well device are listed in Table 1-4. The IQE is the ratio of the number of collected carriers to the numbe r of photons that enter the cell. The IQE is related to the collection effici ency but it takes into account losses associated to the limited thickness of the absorber layer.163 The low IQE for cell 2 was at tributed to poor crystalline quality, implying a large number of defects. X-ray data for cell 1 showed that the indium composition in the i-layer ranged from 3-16% with a maximum intensity at 7%. Lower efficiency of the quantum well structure was due to incomplete adsorption of the light due to the low thickness of the wells (1 nm). In summary, theoretical pred ictions have shown that InxGa1-xN is a promising absorber material for high efficiency sola r cells. Current progress includes growth on Si and Ge substrates which are beneficial for PV devices as well as progress towards p-type doping of InN. Challenges that must be addressed for developing InxGa1-xN-based solar cells include successful p-type doping of In-rich InxGa1-xN and understanding the alloy pha se separation that occurs during growth and its effect on cell characteristics. These current device limitations have lead to

PAGE 42

42 the demonstration of InxGa1-xN-based solar cell devices with at ypical structures, such as thin quantum wells. 1.3 Terahertz Applications for InN and InxGa1-xN Terahertz (THz = 1012 Hz) is a term used to describe waves with a spectrum ranging from 0.1 to 10 THz, or wavelengths from 3 mm to 30 m, respectively. A frequency of 1 THz is equivalent to wavelengths of 0.3 mm (300 m), wavenumber of 33 cm-1, or energy of 4.14 meV. The THz frequency region is located between the infrared and millimeter wavelengths on the electromagnetic spect rum (Figure 1-20).164 The THz frequency regime lies between the el ectronic and photonic domains and therefore a mixture of optical and electr onic mechanisms is often used to generate THz emission.165 Semiconductor photonic devices, which are dominat ed by inter-band transitions are limited to the high end of the THz range (~ 10s of THz), while todays electronic devices can only reach frequencies up to a few hundred gigahert z, past which causes circuit failure.164 THz fields are not well-developed even though the region sits between the well developed (by comparison) regimes of photonics and microwave technology.166 In recent years, however, THz technology has made significant developments mostly due to the improvement and availability of femtosecond lasers.167 THz frequency devices have applications in a variety of fields including astrophysics, plasma physics, spectroscopy, medical imag ing (T-rays), biology, and communications.165 There is thus considerable motivation to improve the current THz tec hnology, which has several disadvantages. Current photonic THz emitters are bulky and expensive or must be operated at cryogenic temperature, as in the case of quantum tunneling lasers.168 THz detectors are also bulky, expensive, and lack precision.164

PAGE 43

43 InN was established as an ideal material for high frequency terahertz devices once the theoretically predicted values of the peak electron velocity was shown to be higher than that of GaAs and GaN, under moderate electric fields.19,169 Other favorable elec tron transport properties of InN are large intervalley energy separati on, large polar optical phonon energy and a small effective mass.18,19,170,171 There has been little experimental work an alyzing terahertz emission and detection of InN114,172 and no know work on InxGa1-xN thin film alloys. A few researchers have investigated terahertz emission from InxGa1-xN/GaN multi-quantum wells167,173-175 and more researchers have theoretically investigated TH z frequency aspects of InN176-180 and III-nitride heterstructures.181-184 Brazis et al .184 performed Monte Carlo simulations of the third harmonic generation (THG) efficiency of GaAs, InP, InN, and GaN to compare to Si. Their results showed that the III-V semiconductors predicted THG efficiency exceeds experimental values for Si by two orders of magnitude. Cooling to liquid nitrogen temperatur e (77K) increased the THG efficiency. It was also noted that InP and InN showed superior ma ximum efficiencies for the materials examined. Shiktorov et al .178 examined frequency multiplication by higher order odd harmonic generation with Monte Carlo simulations a nd confirmed that InN was a more ideal material than InP or GaAs. InN n+nn+ or n+n-nn+ structures have been modeled by Monte Carlo simulations.176,177,180 Optical phonon emission was shown to be th e dominant scattering mechanism in a n+nn+ structure and that a free carrier gr ating can be formed in the n-region.180 InN n+n-nn+ structures were predicted to emit 50 W of microwave power in the 1.1 to 1.2 THz range when connected to an external circuit and operate d at liquid nitrogen temperature.177 Starikov et al .176 showed

PAGE 44

44 that noise enhancement in n+nn+ structures of InN and GaxIn1-xAs can predict the onset of instability. OLeary et al.179 compared the THz emission of bulk InN vs. thin films by using Monte Carlo simulation. Bulk InN (10 m) showed possible emission at 10 GHz while 100 nm thick InN had emission frequencies up to 2.5 THz. Korotyeyev et al.183 also modeled the benefits of III-N heterostructures vs. bulk materials and attributed the benefit to electron pinning. Promising experimental results for THz emission have been found for InxGa1-xN/GaN heterostructures173,175 and it was shown that emission efficiency in InxGa1-xN/GaN heterostructures is better than rectifica tion processes in bulk crystals like ZnTe.167 Stanton et al.174 presented evidence that acoustic phonons could be used to image surfaces and interfaces in nanostructures from experiments on InxGa1-xN/GaN heterostructures. The first THz emission from InN was demonstrated by Ascazubi et al.114 by optical excitation of ultrashort radiation pulses from th e surface of InN (at room temperature). The semiconductor was unbiased and excitation was generated by femtosecond Ti:sapphire laser pulses at 800 nm. THz radiation from InN was co mpared to that of optimized p-type InAs and showed radiation on the same orde r of magnitude. This is a prom ising result since the InN thin films contained high defect densitie s and carrier concentrations (1010 cm-2 and 1018 cm-3, respectively), and also because InAs is curren tly one of the best THz semiconductor surface emitters.185 Optimized InN layers are believed to produce much higher THz emission since lower carrier concentrations will lead to less free carrier absorption. Meziani et al.172 investigated THz emission from high quality InN epitaxial layers under a range of temperatures (2-300K) and magnetic fields up to 13 Tesla. Higher THz transmission was noticed for higher magnetic fields, which was confirmed by simulati ons, and helicon waves were considered an

PAGE 45

45 important contribution to emission. It was shown that the carrier concentration and momentum scattering rate, thus the film quality, could be determined by using this contactless method of THz transmission. Also, THz emission and detection by GaN HEMTs was recently reported by several investigators.185-187 These recent results suggest that InxGa1-xN alloys are promising materials for small size, room temperature and high pe rformance THz emitters and detectors. 1.4 Statement of Thesis Properties such as a high absorption coeffici ent, radiation hardness, and absorption over a wide range of wavelengths make InxGa1-xN an attractive photovoltaic material (Section 1.2.2). To better understand InxGa1-xNs potential as a PV absorber layer, InxGa1-xN-based solar cell device simulations were performed using the MEDICI device simulation software package (Chapter 2). Optimized singl e junction and multi-junction InxGa1-xN solar cells were simulated and the potential effects of phase separation are assessed. InN and InxGa1-xN are materials that show a great deal of potential in applications such as high efficiency solar cells and te rahertz electronic devices, and InxGa1-xN has already been proven to be a dominant material for LED applic ations. Although signifi cant progress has been made in the field of InxGa1-xN based LEDs, the properties of InN and In-rich InxGa1-xN are still not very well understood. It is for this reason that growth InxGa1-xN thin films have been grown over the entire compositional range (0 x 1), so that fundamental issues such as phase separation can be better understood. For the applications of solar cells, InxGa1-xN alloy composition is very important for the design of high efficiency multi-junction solar cells. Understanding and optimizing growth conditions for different InxGa1-xN compositions is important for THz applications since THz em ission is inversely proportional to doping concentration due to free carrier absorption, which is natively very high in nitrides. Phase

PAGE 46

46 separation in InxGa1-xN alloys might also prove interesting for THz applications because residual strain affects the internal bias and thus th e emission efficiency. The growth of InN and InxGa1-xN thin films is presented in Chapter 3. The benefits of semiconductor nanostructures were discussed in Section 1.1.3, and these nanostructures offer unique abilities to understand the properties of InN that cannot be gleaned from bulk samples. InN and InxGa1-xN nanostructures are specifical ly advantageous for PV and THz applications due to their la rge surface to volume ratio. For solar cells a larger surface area can result in greater co llection efficiency and THz emission efficiency should increase with increased surface area since the emission mechanis m results from interfacial interactions. The growth of InN nanostructures has be en presented in the literature mo re frequently in recent years, however, growth is typically not highly controllable. There is a need for nanostructured growth to be incorporated with curre nt growth techniques and proc ess requirements, which is why controlled MOCVD or H-MOCVD gr owth of nanostructures is pr eferred. Growth of InN nanowires by MOCVD and InN nanoand micro -rods by H-MOCVD are presented in Chapter 4 and 5, respectively. Chapter 6 contains an explor atory study of GaN nanostructures by MOCVD. This section is only discussed briefly since GaN nanostructu res are not the primary focus of this work, however, this area is scientifica lly interesting. Recommendations for future work are presented in Chapter 7.

PAGE 47

47 Figure 1-1. The III-nitride zincbl ende crystal structure along various directions. A) [100] (1 unit cell). B) [110] (2 unit cells). C) [111] (2 unit cells) (Ref 2). Figure 1-2. The III-nitride wurt zite crystal structure along vari ous directions. A) [0001]. B) ] 0 2 11 [. C) ] 0 1 10 [ (Ref. 2). Figure 1-3. Cation-faced (Ga) a nd nitrogen-faced polarity for the GaN wurtzite crystal structure (Ref. 5).

PAGE 48

48 Figure 1-4. Velocity-field ch aracteristics (T = 300K, n = 1017 cm-3) for wurtzite InN, GaN, AlN, and zincblende GaAs (Ref. 19). Figure 1-5. Vapor pressure of N2 in equilibrium with of InN, GaN, and AlN as a function of temperature (Ref. 63).

PAGE 49

49 Figure 1-6. Rhombohedral structure and surface planes of sapphire (Ref 5). Figure 1-7. Theoretical and experimental examples of na nowire growth mechanisms. A) Theoretical VLS growth mechanism (Ref 107). B) Experimental evidence for a VLS mechanism showing metal catalyst at tip of nanowire (Ref. 92). C) Experimental evidence for a VS mechan ism, showing no metal droplet (Ref. 103).

PAGE 50

50 Figure 1-8. Predicted binodal (solid) and sp inodal (dashed) decomposition curves for InxGa1-xN assuming regular solution mixing (Ref. 123). Figure 1-9. The T-x phase diagrams of ternary InxGa1-xN compounds. A) Relaxed InxGa1-xN layers. B) Strained InxGa1-xN layers with the interface or ientation perpendicular to the hexagonal axis of the crystal (Ref. 124).

PAGE 51

51 Figure 1-10. Several XRD spectra of phase separated InxGa1-xN films at different temperatures (Ref. 12). Figure 1-11. Compositional control of InxGa1-xN with respect to growth temperature (Ref. 12).

PAGE 52

52 Figure 1-12. U.S. photovoltaic module implementation from 1992-2003 (Ref. 144). Figure 1-13. Comparison of wo rld wide growth of PV pr oduction from 1990-2004 (Ref. 143).

PAGE 53

53 Figure 1-14. Market shares of different photovoltaic ma terials as of 2001 (Ref. 145). Figure 1-15. Schematic of light induced carrier generation in a generi c solar cell (Ref. 147).

PAGE 54

54 ASTM G173-03 Reference Spectra 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 25050075010001250150017502000225025002750 Wavelength (nm)Spectral Irradiance (W m-2 nm -1) Extraterrestrial (AM0) Terrestrial (AM1.5) Figure 1-16. Solar irradiance vs. wavelength for extraterrestrial (AM0) and terrestrial (AM1.5) spectra (American Society for Testing and Materials, ASTM, Reference Spectra for Photovoltaic Performance Eval uation, ASTM G-173-03, 2003, http://rredc.nrel.gov/solar/spectra/ accessed April 2007).

PAGE 55

55 Figure 1-17. Global solar irradi ances averaged over three ye ars (1991-1993) which account for cloud cover (Matthias Loster, 2006, To tal Primary Energy Supply: Land Area Requirements, http://www.ez2c.de/ml/solar_land_area/ accessed April 2007). Figure 1-18. Current-voltage charac teristics of a generic solar cell. A) under dark (solid line) and illuminated (dashed line) conditions. B) The resulting maximum power rectangle (under illumination) (Ref. 151).

PAGE 56

56 Figure 1-19. Incident sola r flux for AM0 (black) and AM 1.5 (red) as well as the InxGa1-xN band gap energy range (colored box). Figure 1-20. Terahertz frequenc y range (shaded) of the elec tromagnetic spectrum, including molecular transitions (Ref. 168).

PAGE 57

57 Table 1-1. Some fundamental pr operties of InN and GaN (Ref. 50a, 188, 189b, 190c). GaN (wurtzite)InN (wurtzite) Room temperature band gap energy (eV)3.390.6-2.1aTemperature coefficient (eV/K) dEg/dT = -6.0 x 10-4 dEg/dT = -1.8 x 10-4 Lattice constants ()a = 3.189a = 3.5377bc = 5.185c = 5.7037bThermal expansion (K-1) a/a = 5.59 x 10-6 a/a = 4 x 10-6 c/c = 3.17 x 10-6 c/c = 3 x 10-6Thermal conductivity (W/cm-K)1.30.8 0.2 Index of refraction2.672.9-3.05 GaN (zincblende)InN (zincblende) Room temperature band gap energy (eV)3.2-3.30.6cLattice constants ()a = 4.52 a = 4.986cProperty Table 1-2. Theoretical and e xperimental mobility and cont rolled doping ranges for InN and GaN. InNRef.GaNRef 300K4,4001,350 77 K 30,000 o 19,200 o 300Kno available data13193 Controlled doping range (cm-3) n-type (Si)5x1016 to 5x10201016 to 4x1020p-type (Mg)no available data 1016 to 6x1018Experimental maximum hole mobility (cm2/V-s) 194195 Property Theoretical maximum electron mobility (cm2/V-s) 57192

PAGE 58

58 Table 1-3. Lattice and thermal e xpansion constants for InN substrat es and buffer layers (Ref. 5, 196). Substrate material or buffer Layer Space group symmetry a ()c () a (10-6 K-1) c (10-6 K-1) GaN P63mc 3.189 o 5.185 oo 5.59 oo 3.17 AlN P63mc 3.112 o 4.982 oo 4.20 oo 5.30 c-Al2O3R3c4.759 o 12.991 oo 7.30 oo 8.50 Si (111)Fd3m5.430 o ------3.59 oo -----6H-SiC P63mc 3.080 o 15.117 oo 4.46 oo 4.16 GaAsF43m5.653 o ------6.00 oo -----InPF43m5.869 o ------4.50 oo -----GaPF43m5.451 o ------4.65 oo -----ZnO P63mc 3.252 o 5.213 oo 2.90 oo 4.75 MgOFm3m4.216 o ------10.50 oo -----MgAl2O4Fd3m 8.083 o -----7.45 oo -----LiAlO2P41212 5.169 o 6.268 oo 15.00 oo 7.10 5.406 o 5.013 oo 6.00 oo 7.00 LiGaO2Pna21b = 6.3786b = 9.00 lattice parameters Thermal expansion coefficients Table 1-4. Internal quant um efficiency (IQE) of InxGa1-xN p-i-n and quantum well solar cells (Ref. 162). Cell #Type InxGa1-xN (x) IQE (%) 1p-i-n0.0719.0 o 2p-i-n0.401.0 3quantum well0.408.0

PAGE 59

59 CHAPTER 2 INDIUM GALLIUM NITRIDE SOLA R CELL DEVICE SIMULATIONS 2.1 Introduction InxGa1-xN is an optimal material for absorber la yers in solar cells because the InN-GaN alloy system covers a wide spectral range that spans the majority of wavelengths in the solar spectrum. In-rich InxGa1-xN has a high absorption coefficient and has also been proven to be highly radiation resistant.14 A more detailed explanation of InxGa1-xNs material qualifications for PV applications is given in Section 1.2.2. Development of InxGa1-xN based photovoltaics has been delayed by the difficulty in p-type doping of InN and In-rich InxGa1-xN alloys. The recent results for Mg doped InN, however, are encouragi ng for eventual p-type doping of InN and thus In-rich InxGa1-xN alloys.16 In this chapter, single and multi-junction InxGa1-xN solar cells are modeled using MEDICI device simulation software. The optimum cell pa rameters are determined for a single-junction InxGa1-xN solar cell and multi-junction solar cells are simulated for two, three, four, and five junctions. The implications of phase separati on and the effect on cell efficiencies are also discussed for both single and multi-junction InxGa1-xN solar cells. The intention of this simulation study is to estimate the efficiency of InxGa1-xN solar cells by applying realistic limits (thic kness, doping levels, band gap energy, etc) to the cell layers and to use the most recent material properties from the available literature. No p-n junction InxGa1-xN solar cell has been demonstrated to date therefore it is not possible to match the predictions to real devices. A pplying realistic cell parameters is currently the best method to measure the possible potential of InxGa1-xN solar cells.

PAGE 60

60 2.2 MEDICI Device Simulation Software Medici is a 2-dimensional device simulator that models the electrical, thermal, and optical properties of semiconductor devices. The Taur us-Medici program is part of the Technology Computer-Aided Design (TCAD) so ftware distributed by the Synopsys Corporation. Medici is frequently used to model MOS and BJT devices. A 2D non-uniform mesh grid is used that can model planar or nonplanar surface features. Fo r each grid section Medi ci solves Poissons equations as well as the electron and hole continu ity equations and predic ts the two-dimensional distributions of potential and carrier concentrations at arbitr ary bias conditions. The program can model several phenomenons such as recombinat ion, photogeneration, impact ionization, band gap narrowing, band-to-band tunneling, mobility, and carrier lifetimes (Synopsis TCAD Medici user manual version 2001.4). 2.3 Identification of InxGa1-xN Solar Cell Parameters Material Properties. Table 2-1 lists the input parameter values for InN and GaN which are used in the Medici simulation. The appropria te references are listed for each input parameter and if no values are available in the literature then a realistic value was estimated or assumed (values in italics). For example, no definitive p-type doping has been demonstrated for InN; therefore no hole mobility values have been publis hed. It can, however, be estimated because mobility ( ) of a semiconductor is inversely pr oportional to the effective mass (m*) (Eq. 2-1),197 *m q (2-1) where < > is the average time between collisions a nd q is the unit of charge. The electron effective mass of GaN (0.22mo)198 is larger than the e ffective mass of InN (0.07mo),41 where mo is the electron rest mass (9.11 x 10-31 kg). Prior to the InN band gap controversy, theoretical calculations were performed that showed the hole effective mass of InN was approximately the

PAGE 61

61 same or less than the values for GaN (for light hole and heavy holes).199 It has been suggested that the hole effective mass be revised (and decr eased) due to the commonly accepted lower InN band gap energy value198. From the available theoretical calcula tions it is safe to assume that the hole mobility of InN is the same or slightly gr eater than the value obtained experimentally for GaN (200 cm2/V-s).200 For the simulations presented here a hole mobility of 200 cm2/V-s was chosen for InN, although the real value is likely higher. A wide range of electron mobility data has been published for InN which ranges from 200 to 2700 cm2/V-s depending on the type of deposition technique.31,41,42,58,201,202 High quality InN layers grown by MOCVD have electron mobilities close to 1000 cm2/V-s, therefore this was the value used for the Medici simulations. Optimization for the growth of InN thin films on c-Al2O3 by MOCVD, can be found in Ref. 203. High quality single crys tal InN films were grown on c-Al2O3 substrates using substrate nitridation and a low temperatur e InN buffer layer, re sulting in a XRD FW HM of 1339 arcsec.203 The absorption coefficient ( ) of this high quality InN film was measured ( Figure 2-1). The absorption of InN ranges from 7 x 104 to 1.5 x 105 cm-1 in the incident energy range of 0.7 to 3.4 eV with an average value of 9 x 10-4 cm-1 (up to the band gap energy of GaN). The absorption dropped quickly to zero at an appr oximate band gap energy value of 0.65 eV (not shown in Figure 2-1). For the absorption measurements, an incandescent lamp inside a quartz tube was used as the illumination source and a diffraction grating was used to filter light into different spectral ranges. A Hitachi 330 spectro photometer was used to detect the transmitted light. The absorption coefficien t of GaN was measured by Muth et al.204 with an average value of 1 x 105 cm-1. For simplicity a common value for th e absorption coefficient of 9.5 x 104 cm-1 was used for all InxGa1-xN compositions.

PAGE 62

62 The properties in Table 2-1 were va ried as a linear function based on InxGa1-xN composition, with the exception of the band gap energy. A va riable band gap energy bowing parameter is used that has been determin ed by fitting data to several sets of InxGa1-xN alloys.62 This bowing parameter is believed to be the mo st accurate because other bowing parameters are usually determined from limited composition ranges (i.e. either Ga-rich or In-rich compositions only). The equations determining the band ga p energy and variable bowing parameter were previously shown (Eqs. 1-2 and 1-3). Illumination sources. Terrestrial (AM1.5) and extraterre strial (AM0) solar spectra were used as illumination sources for the InxGa1-xN cells. The AM1.5 spectrum was simulated by using spectral irradiance values (as a function of wavelength) from the American Society for Testing and Materials reference spectrum (G -173-03). The AM0 spectrum was modeled by the Medici software by assuming black body radiation at a temp erature of 5800K (Eq. 2-2). 1 exp 25 2 2kT hc hc R R ISE S (2-2) where RS is the radius of the sun (7 x 108 m), RSE is the distance between the sun and the earth (1.5 x 1011 m), T is the temperatur e of the black body, and is the wavelength. Recombination Models. The CONSRH and AUGER Medici recombination models are used to describe th e recombination in the InxGa1-xN layers. CONSHR refers to a concentration dependent Shockley -Read-Hall model, where the carri er concentration determines the carrier lifetimes and the probability for reco mbination and generation. The Shockley-ReadHall model was developed in 1952 and used to describe the recombination and generation statistics for holes and electrons by using a distribution of trap stat es in the forbidden gap as well as kinetic transport models for electrons and holes.205-207 With the CONSHR model, Medici

PAGE 63

63 determines a probability for recombination through trap states by predicting a distribution of trap states within the band ga p. The AUGER model refers to Auge r recombination, which is band to band recombination that occurs when two like carri ers collide. When two electrons collide in the conduction band one carrier loses it s energy and is recombined with a hole in the valence band, while the surviving carrier recei ves the energy released by reco mbination and becomes a higher energy electron. The excited elec tron subsequently loses its ener gy to lattice vibrations through collisions with the semiconductor lattice.197 Other than these two defect models, no specific defect levels or interface states were used in the simulation. 2.4 Results 2.4.1 Single-Junction InxGa1-xN Cell Optimization Cell parameters such as n-side thickness, pside thickness, doping c oncentrations, and band gap energy were optimized to obtain th e maximum efficiency of a single InxGa1-xN p-n junction solar cell. The cell parameters were limited by design rules typically as signed to thin film single-junction solar cells. A fundame ntal design rule of thin film solar cells is that the thickness of the solar cell must be on the or der of a few microns. The motivation for thin film solar cells is to obtain the best cell efficiency while keeping la yers thin to reduce processing time and the cost of fabrication. Thin film silicon so lar cell technologies are approximately 10 m thick compared to the bulk cells which are 250 m thick. When the cell layer thickness decreases light trapping techniques such as surface r oughing and texturing are important to enhance absorption in the cell.208 This is done for Si solar cel ls because the absorption coefficient is low compared to other photovoltaic materials (Figure 2-2) More recently investigated photovoltaic materials such as CuInSe2 (CIS) have a higher absorption coefficient over a broad range of wavelengths (Figure 2-2). Absorber layers fabricated wi th CIS then have lower thicknesses (~2 m).147 Since

PAGE 64

64 InxGa1-xN has a similar absorption coefficient the same maximum thickness design rule (approximately 2-3 m thick absorber layers) was applied. Typically p-n junction solar cells employ asymmetric doping between the p-type and n-type layers, which also differ in thickne ss. Generally the thickest layer (~ 2 m) is lowly doped (1017 cm-3) while the thinner layer is highly doped (1020 cm-3).197 It is also preferred that the thicker layer be p-type because of the lo wer mobility holes. Since minority carriers are collected instead of majority carriers, it is pref erred to have higher mobility electrons traveling through the thickest part of the so lar cell, thus requiring a thick p-t ype region. This is beneficial because a higher mobility means carriers are more likely to be collected before recombination can occur. The motivation for the asymmetric doping and thickness struct ure is to minimize the light absorption by free carriers while still producing a large depletion region to minimize diffusion lengths (available time for recombinati on) and increase charge car rier separation. Free carrier absorption reduces effici ency because this absorption do es not generate electron hole pairs, but instead promotes ca rriers in the conduction and vale nce bands to higher energies, which is then lost in the form of heat.209 It is assumed for InxGa1-xN that p-type doping will be achie ved though it is currently very difficult. For this work the acceptable p-type doping range is limited to 1 x 1016 to 1 x 1017 cm-3. This doping range was chosen because similar cont rolled doping ranges have been exhibited for InP MOCVD.210 A comparison with InP must be made since no doping levels have been measured for p-type InN. InP is a III-V semiconductor with a band gap energy (1.34 eV) that is close to InNs band gap energy (0 .7-1.0 eV). The controllable n-type doping range of InN has been shown to be 1 x 1018 to 1 x 1020 cm-3.194,211 From these doping charact eristics it is ideal to use a low doped thick p-type InxGa1-xN layer with a highly doped thin n-type InxGa1-xN layer for

PAGE 65

65 the structure of the solar cell (F igure 2-3). The anti-re flecting (AR) coating is used to prevent light reflection from the top surf ace of the semiconductor. Electri cal connections to the cell are made by the back and front contacts, where the b ack contact could be metal contact to a p-type silicon substrate and the front cont act is also a metal contact depos ited in a finger arrangement to maximize light transmission. For the Medici si mulation no exact substrat e properties (such as those for Si) were included for the calculations, instead a generic contact resistance was used. The resistance for the back and top contacts are modeled by including a c ontact resistance of 2 x 10-6 (Medici user manual version 2001.4) and 1 x 10-4 -cm2,212 respectively. All single-junction simulations were p-InxGa1-xN/n-InxGa1-xN junctions and were modeled by 1D Medici simulations. Arbitrary doping an d thickness conditions we re initially set to determine the optimal band gap energy for the maximum single-junction cell efficiency. The thickness and doping level of the pand n-type layers were then optimized, and then the band gap energy value was re-evaluated in an iterative process. When band gap energy values were varied during optimization the same band gap energy value was used fo r both pand n-type layers. During all optimization steps only one va riable (such as n-side thickness, p-side doping concentration, etc.) was varied while all other parameters remained fixed. The initial absorber structure (Figure 2-4a) consisted of a ho moepitaxial stack of a 300 nm n-type InxGa1-xN layer with a doping concentration of 1019 cm-3 and a 2 m thick p-type InxGa1-xN layer with a doping concentration of 1016 cm-3. The efficiency calculations for the various steps of the optimization pr ocedure are plotted in Figure 2-5. The efficiency of the initial structure was low with an efficien cy of about 4%. As previously shown in Eq. 1-6 the efficiency is the maximum power produced by the solar ce ll divided by the inlet power (incident energy from the simulated solar spectrum). The dominant factor for the low efficiency of this cell is the

PAGE 66

66 thickness of the highly doped n-type layer that gives str ong free carrier absorption and recombination in this layer. The band gap energy of InxGa1-xN for the initial structure was varied from 1.1 to 1.6 eV (Figure 2-5, represen ted by blue diamond symbols) and the maximum efficiency occurred at a band gap energy of 1.5 eV, which is close to the optimal ideal band gap energy of approximately 1.4 eV.145 In the next stage of optimization, the thickness of the n-type layer was varied (at a constant band gap energy value of 1.5 eV). A reduc tion from 300 to 25 nm produced a significant increase in the efficiency due to reduced free carrier absorption and carrier scattering (square symbols in Figure 2-5). A further increase in efficiency is obtain ed when a linear graded carrier concentration in the n-type laye r is used instead of a uniform doping profile (triangle symbols in Figure 2-5). The p-side doping concentr ation was varied from 1016 cm-3 to 1017 cm-3 and the best cell efficiencies were obtained at a doping concentration 1017 cm-3. The doping concentration of the p-type layer did not exceed 1017 cm-3 due to the previously assumed design rules. Finally the absorption coefficient of InxGa1-xN is sufficiently high that incr easing the p-side thickness will not provide any signifi cant improvement in efficiency, theref ore the thickness of the p-side layer was decreased to determine the reduction in effi ciency. When the p-si de thickness is reduced from 2 to 1.5 to 1 m the change in efficiency with respect to the 2 m thickness is -0.21% (1.5 m) and -0.62% (1 m). For the purposes of this study a p-side thickness of 2 m was used in order to obtain the maximum efficiencies possible, however it can clearly be seen that material processing times can be significan tly reduced (for absorber laye r growth) with only a minimal loss in efficiency. Also, adding thickness to the p-side layer will only increase the series resistance, leading to a decrease in voltage.

PAGE 67

67 The maximum efficiency band gap energy was re-evaluated using the refined absorber layer structure (Figure 2-4b). A plot of the cell efficiency ve rsus band gap energy for both AM0 and AM1.5 illumination is shown in Figure 2-6. The maximum calculated efficiency was obtained at a band gap energy of 1.44 eV for bot h AM0 and AM1.5 illumination. Plots of the cell power versus load and J-V curves for AM0 and AM1.5 are shown in Figures 2-7 and 2-8, respectively, and the cell characteristics (Jsc, Voc, maximum power, FF, and collection ) are listed in Table 2-2. Both illumination conditions show similar values for the fill factor (86%) and the efficiency is slightly higher for the AM1.5 (16% compared to 15.3% for AM0). The highest collection efficiency is obtained for the intermediate ba nd gap energy values tested and that the efficiency decreases for large or small band gap energy values (Figure 2-6). As the band gap energy increases the voltage of the solar cel l will increase, however the solar flux decreases at these higher band gap energies (Figure 119). As the band gap energy of the solar cell increases all the light energy below the cell band gap energy cannot be collected. This creates a trade off between band gap energy and the availabl e light energy for collection, which is why a maximum efficiency is obtained at a band gap energy valu e of 1.4-1.5 eV. It is difficult to assess the validity of these simulations because no p-n InxGa1-xN solar cells have been fabricated. The InxGa1-xN simulations presented here were compared to copper indium gallium diselenide (CIGS) Medici simulati ons that use the properties that are measured for real CIGS solar cell devices, such as abso rption coefficients, layer thicknesses, and doping concentrations.213 Comparing these two materials is idea l because both materials are thin film absorber structures with similar absorption coe fficients. Both simulated solar cells were assigned a band gap energy of 1.24 eV and illuminate d with AM0 light. At this band gap energy value the efficiency of the CIGS devices was 16% (FF = 73.8%) and the InxGa1-xN cell had an

PAGE 68

68 efficiency of 14.6% (FF = 83.8%). Current laboratory best e fficiencies for CIGS are 19.5% while larger scale module effi ciency range from 13 to 15%.214 The Medici solar simulator in conjunction with the recombinati on models used produces cell effi ciencies in between the best laboratory cells and module effi ciencies. The difference in FF was attributed to a CdS buffer layer and ZnO transparent conducting oxide (T CO) layer, which was not present in the InxGa1-xN structure. No p-n InxGa1-xN solar cells have been fabricated to date, however, this comparison was used to validate the m odel used for the Medici InxGa1-xN simulations. As previously mentioned the InxGa1-xN cell efficiencies could be improve d by increasing the level of p-type doping which would provide similar or higher efficiency values as CIGS. However, the doping levels were capped at lower levels to represent anticipated near future InxGa1-xN p-type doping capabilities. The same recombination models an d contact resistances were used for the CIGS simulation for accuracy in the comparison. 2.4.2 Multi-Junction InxGa1-xN Solar Cells Multi-junction InxGa1-xN solar cells were next simulate d using the refined single-junction solar cell structure previously described as a star ting point. Two, three, four, and five solar cell junctions were modeled. The pand n-side thickness of each pn junction remained fixed at 2 m and 40 nm, respectively, for each absorber layer simulated. It was previously assumed that the achievable levels of p-type doping for InxGa1-xN are much below degenerate doping levels. For this reason the solar cell junctions must be connected in a mechanical stack arrangement instead of monolithically using tunnel junctions. It is possible, however, that an alternative material system could be used for the tunnel junc tion. Each layer of the mechanical stack multijunction cell is simulated separately thus requiri ng two connections per cell in the stack. The efficiency of each multi-junction solar cell is simply the summation of each individual junction within the cell, with only the spectrum of the incident radiation changing, depending on

PAGE 69

69 the characteristics of the cell above it. The same approach is used for determining the cell open circuit voltage. A plot of the solar cell efficien cy and open circuit voltage is shown in Figure 2-9 as a function junction number. The optimized p-n junction band gap energy values for each multi-junction cell are shown in Table 2-4. The efficiency increases from 15.4% for a single-junction cell to 27.4% for a five junction solar cell (Fig ure 2-9). These solar cell efficiencies are slightly less th an current III-V spa ce solar cells produced by Spectrolab based on GaInP2/GaAs/Ge cells, which have efficiencies up to 28.3%, which use three cell junctions (Spectrolab Inc, Ultra Triple Junction Space Solar Cell data sheets, www.spectrolab.com/prd/prd.htm accessed April 2007). For these simulations conservative estimates were made for the structure of the p-n junction absorber so it is not surp rising that the efficiencies are lo wer. It was also assumed that no light greater than the band gap energy of a specific junction was allowed to transmit to subsequent junctions. Other researchers have simulated InxGa1-xN solar cells by applying a less constrictive model to the cell parameters and ha ve predicted efficiencies of 39% for a five junction monolithic cell under AM1.5 illumination.151 Therefore, InxGa1-xN multi-junction solar cell absorbers have been predicted to reach sim ilar efficiencies as the current best cells in production even after assigning a constrictive model to the simulations. Th ere is also potential for obtaining much higher efficiencies if InxGa1-xN absorber layers can be grown with superior material properties. 2.4.3 Phase Separation in InxGa1-xN Solar Cells 2.4.3.1 Effect on single-junction cells Since phase separation is known to occur in InxGa1-xN alloys it is important to assess how solar cell characteristics might be affected. Upon initial inspection it is not clear how phase separation will affect cell characteristics because there are several carrier dynamics to consider

PAGE 70

70 when a solar cell is illuminated. For exampl e, consider a uniform direct band gap energy semiconductor p-n junction that is connected to an external ci rcuit. Electron-hole pairs are created when the p-n junction is illuminated with photon energy grea ter than the band gap energy of the semiconductor (Figure 2-10). When photons with energy much greater than the band gap energy are absorbed then electrons can be prom oted high into the conduc tion band (Figure 2-10). The excess absorbed energy can be transferred to either of the tw o bands (or both), but typically the majority of the excess energy is transferred to the band with the smaller carrier effective mass.215 The p-type InxGa1-xN region is the bulk of the solar cell absorber and the electrons have the smallest effective mass, therefore the majority of the energy is tran sferred to the conduction band (Figure 2-10). The energe tic electron is called a hot electron because the electron temperature (Te) is higher than the semic onductor lattice temperature (To). Typically this energetic electron loses its en ergy through a series of phonon s cattering events within the semiconductor lattice, where the el ectron relaxes to the band edge with an electron temperature To. The time constant ( ) for hot electron relaxation to the band edge is on the order of picoseconds216,217 while band to band recombinati on is on the order of microseconds.207 Now consider an InxGa1-xN p-n junction which contains phase segregated regions of InyGa1-yN where x > y. This structure forms an In-rich InxGa1-xN matrix with Ga-rich precipitates (Figure 211). A representative band diagram is shown in Figure 2-12 for the matrix and precipitate in the p-type re gion of the p-n junction. A t ype I band offset applies to InN/GaN interface where the valence and conduc tion band edges of InN lie within the band edges of GaN.218 There is a greater band gap di scontinuity in the conduction band ( Ec ~ 1.7 eV) than the valence band ( Ev ~ 1 eV), however there is still so me debate about the exact offset values.42,122,219,220 Minority carriers (electrons) in th e phase-separated structure that are

PAGE 71

71 generated under illumination will be accelerated towards the n-type region, assuming that this region is within the depletion regi on of the junction. This type of phase-separated structure will be beneficial for the generation of hot carriers (compared to ca rriers generated in the matrix) since the band gap energy of the prec ipitates are greater than the matr ix. If it is assumed that the hot carriers generated in the prec ipitates quickly relax to the ba nd edges of the matrix before being collected by external contacts then the effi ciency of the solar cell will simply take on the electrical characteristics of the matrix material. The efficien cy of a phase-separated solar cell might be slightly reduced due to an increas ed transit time around the wider band gap energy regions, resulting in a higher proba bility of recombination. Regardless of a slight decrease in efficiency a single-junction phase -separated solar cell should closely model a uniform solar with the same band gap energy and doping characteri stics as the matrix material in the phase-separated solar cell. This predicted result is based on the assumption that all hot carriers are relaxed before being collected. If it is assumed that hot carriers are allowed to remain hot and this energy can be collected without relaxing to the band edge of the matrix then the efficiency of a phase-separated solar cell can be increased compared to a uniform single-ju nction solar cell. For this situation a device structure was assumed to contain an In-rich InxGa1-xN matrix with a band gap energy of 1.0 eV and precipitate phase with a band gap energy of 2.0 eV. The overall th ickness of the cell was approximately 2 m and the same doping profiles were us ed as the previously optimized single-junction cell. The precip itate volume fractions were varied from 0.1 to 0.25 for the phase-separated InxGa1-xN solar cells, assuming a uniform di stribution of preci pitates. The matrix and precipitate band gap energy values were assigned because experimental compositions (from XRD) corresponding to these band gap en ergies were seen for phase-separated InxGa1-xN

PAGE 72

72 thin film samples. No experimental evidence ha s been determined for the volume fraction of the precipitates therefore these valu es were assumed. For these simulations hot carriers were allowed to remain hot and be collected at thei r hot energy and it was assumed that no increased recombination occurred due to increased tr ansit time around the wide band gap energy precipitates. It was also assu med that hot carriers were only contributed from the wider band gap energy phase-separated regions. As expected th e maximum efficiency occurred at the highest volume fraction of precipitates because more hot carriers were produced and the cell characteristics are shown in Table 2-3. Il lumination from both AM0 and AM1.5 produced efficiencies of 19% for a precipita te volume fraction of 0.25. This is an impressive result for a single-junction solar cell since th e efficiency increased 3% comp ared to a uniform structure, however this result assumes that the hot carri ers do not relax from th eir elevated state. In certain cases it is safe to assume that some carriers can remain hot because the energy bands are capable of sustaining the extra energy. For example in semiconductors majority carriers can relax back to the band e dge while minority carriers remain hot.221 This situation can be appropriately applied to the 2 m p-type InxGa1-xN absorber layer, where the majority of light absorption takes place. As prev iously mentioned, the majority of energy is transferred to the band with the smaller effective mass which is the conduction band a nd the electrons are also the minority carriers in the p-type region. These cond itions are ideal for maintaining hot electrons in the solar cell. However, even if the minority carrier electron s can remain hot significant phonon scattering will likely occur at the metal contact an d cause the hot electrons to relax to the band edge. Other researchers have s uggested wide band gap energy se miconductors with very narrow band widths ( kTo) as contacts, which would allow for isentropic cooling of hot electrons.222 The voltage for this type of cell would be determined by the summation of the free energy

PAGE 73

73 difference between each hot electron hole pair. This is an attractive approach, however, not practical for real devices due to the assumptions that must be made for the isentropic contacts. Another potential increase in cell efficiency th at can be applied to phase-separated solar cells is the generation of more electron hole pairs through impact ionization. Impact ionization occurs when an electron is accelerated to the poin t where it collides with an atom in the lattice and the energy transfer frees a bound elec tron, creating an electron hole pair.197 When high electric fields are present impact ionization lead s to carrier multiplication (avalanching) that can cause semiconductor devices to fail (breakdown). This effect would be beneficial for solar cells because the increased carrier ge neration would lead to an in creased current. An impact ionization model from the Medici simulation program was applied to the same phase-separated solar cell structure used for mode ling the hot electrons. The results revealed that there was little to no change in the device efficiency due to im pact ionization. Since impact ionization occurs most frequently at high el ectric fields it is likely that the bui lt in bias in the solar cell device is too low to accelerate carriers enough to ionize atoms in the lattice. 2.4.3.2 Effect on multi-junction solar cells A phase-separated solar cell structure coul d potentially have ne gative effects on the efficiency of a multi-junction absorber structure. A lower efficiency is possible as a result of the reduced absorption of light in the precipitate regions of bottom cells because high energy wavelengths will have already been absorbed. Th is will essentially create optically transparent regions which will not produce carriers. Efficiency reduction due to phase separation will become more pronounced when the number of junctions is increased because the band gap energy values are precisely tuned for certain wavelengths. If these light wavelengths are absorbed before reaching the appropriately design ed cell then inactive areas (dead volumes) are produced in the cell.

PAGE 74

74 To show how a two junction sola r cell can be unaffected by pha se separation consider a top absorber layer with a matrix band gap ener gy of 1.8 eV (with a higher band gap energy precipitate phase) and a bottom ab sorber layer with a matrix band gap energy of 1.0 eV. As long as the precipitate band gap ener gy in the bottom cell has a band gap energy below 1.8 eV then the cell characteristics will be unaffected. The negative effects of phase separation in multijunction solar cells can be avoide d by making cell layers sufficiently thick to compensate for any inactive area in the absorber layers. Previous single-junction results show that there is no significant change in absorber efficiency when the p-side thickness is reduced from 2.0 to 1.0 m. This indicates that the majority of the absorption takes place in the first m of the p-type layer and that the additional micron just assures that all of the light is absorbed. If it is assumed that 30% (by volume) of a phase-separated so lar cell is made up of wide band gap energy (optically inactive) areas, then absorption will be limited to 70% of the solar cell material. If a 2 m absorber layer thickness is used then there will be approximately 1.4 m of active area. This thickness should be sufficient for absorbing th e majority of incident light on the solar cell and the only expected reduction in efficiency is attributed to in creased recombination, due to longer transit times for carrier s around the inactive regions. Wider band gap energy junction layers are im portant for obtaining high efficiency multijunction solar cells. For example the largest ba nd gap energy of the fi ve junction solar cell previously simulated was 2.6 eV. To achieve these wider band gap energies more gallium must be incorporated into InxGa1-xN. When the alloy composition b ecomes Ga-rich then it is likely that In-rich segregated phases will form, which will have a smaller band gap energy then the Ga-rich matrix. These smaller band gap energy precipitates will trap carriers, leading to more recombination and a reduction in collection effi ciency (Figure 2-13). Recombination in the

PAGE 75

75 narrow gap precipitate region can lead to light emission which can be re-absorbed by junctions further down the cell. However, not all recombin ation will lead to th e production of light and this is not an efficient way to collect the lig ht energy because the so lar cell takes high energy light and converts it to a reduced amount of lower energy light. Therefore phase separation in the form of smaller band gap en ergy precipitates in a wide band gap energy matrix will lead to a reduced efficiency. It is important to mention th at the probability of phase separation is reduced as the In composition decreases far below 50%, th erefore the negative eff ects of phase separation for a multi-junction solar cell is limited to a small compositional range (0.5 x 0.3). 2.5 Conclusions In this solar cell modeling appr oach, conservative cell propertie s were assigned to absorber layers to anticipate efficiency of InxGa1-xN-based solar cells. The main assumption of this model is that low p-type doping capabilit ies will be achieved for In-rich InxGa1-xN. Single-junction simulations revealed that the optimal band gap energy for both AM0 and AM1.5 illumination conditions is 1.44 eV, yielding an efficiency of 15.3% and 16%, re spectively. The structure for the refined single-junction was a 40 nm n-type layer with a grad ed carrier concentration of 1 x 1018 to 1 x 1020 cm-3 on top of a 2 m thick p-type layer with a uniform carrier concentration of 1 x 1017 cm-3. Multi-junction solar cells that were connected in a mechanical stack arrangement were simulated using the same struct ure as the refined single-junction solar cell. The maximum efficiency (27.4%) wa s obtained for a five junction InxGa1-xN solar cell, which is comparable to current III -V solar cells based on GaxIn1-xP/GaAs/Ge. Comparing the conservative efficiency estimates with other si mulation approaches shows that the predicted efficiency for a five junction InxGa1-xN cell can range from 27-39%. The impressive upper limit can only be achieved if material growth of InxGa1-xN progresses to achieve the optimum properties required for these cells. If the lower of the two cell efficiencies is only possible, as

PAGE 76

76 assumed in this simulation, then the efficiency is still close to the best solar cells currently available. The effect of phase separation which occurs in InxGa1-xN was also assessed to determine the positive or negative effect s on cell efficiency. For a single-junction solar cell, phase separation could theoretically have a positive e ffect through the efficient generation of hot electrons, however, it is mostly likely that pha se separation will have a negligible effect on single-junction cell effici ency. The refined InxGa1-xN absorber structure presented here was shown to be ideal for sustaining hot electrons. Even if the hot elec trons do not relax, phonon scattering at metal contacts would prevent the extra energy from being collected. Methods for collecting hot electrons have b een suggested by other researcher s, however, these methods are not practical. If an absorber la yer phase separates, then the sola r cell will take on the efficiency characteristics of the lowest band gap energy material, which is assumed to be the majority component or matrix mate rial in an In-rich InxGa1-xN alloy. The possibility of impact ionization was also considered for phase-separated solar cells, however, no increase in efficiency was produced most likely due to the lack of a high electric field in the semiconductor depletion region. Single-junction solar cells are predicted to practically un affected by phase separation unless the gallium composition becomes greater than 50%, leading to the formation In-rich InxGa1-xN precipitates. These smaller band gap energy precipitates will likely form carrier wells where recombination will occur. Typically high er gallium contents are required for the top absorber layers in multi-junction cells which requ ire a wide band gap energy. For this reason it is believed that phase separation could reduce th e efficiency of top junctions in multi-junction

PAGE 77

77 cell when gallium rich InxGa1-xN alloys make up the matrix phase, however, only in a small compositional range. Figure 2-1. Absorption coe fficient of an InN thin film grown by MOCVD (Ref. 203). Figure 2-2. Absorption coeffi cient as a function of wavelength for several photovoltaic semiconductors (Ref. 140).

PAGE 78

78 Figure 2-3. Proposed single-junction InxGa1-xN solar cell which has been modeled by Medici. Figure 2-4. Solar cell structures used in single-junction solar cell si mulation. A) Initial structure used as a starting point for the simulations. B) Absorber structure after refinement of cell parameters.

PAGE 79

79 Figure 2-5. Singlejunction absorber optimization steps. Figure 2-6. Simulated solar cell efficien cy vs. band gap energy of the refined InxGa1-xN cell structure for AM0 and AM1.5 illumination.

PAGE 80

80 Figure 2-7. Power vs. load plots for the refined solar ce ll structure for AM0 and AM1.5 illumination. Figure 2-8. Current density vs. voltage curves for the refined solar cell structure under AM0 and AM1.5 illumination.

PAGE 81

81 Figure 2-9. The efficiency a nd open circuit voltage as a f unction of junction number for multi-junction InxGa1-xN solar cells. Figure 2-10. Energy band diagra m showing the generation of elect ron-hole pairs and hot carriers from incident light energy, adapted from Ref. 197. Figure 2-11. Phase-separated InxGa1-xN p-n junction showing Ga-rich InxGa1-xN precipitates (green circles) in an In-rich InxGa1-xN matrix (yellow bulk).

PAGE 82

82 Figure 2-12. Corresponding band diagram from th e p-type region in Figure 2-11 showing the In-rich InxGa1-xN matrix and the Ga-rich precipitate (ppt). Figure 2-13. Band diagram showing trapping of carriers and recombination in narrow band gap energy precipitates (ppt) in the wider band gap energy Ga-rich InxGa1-xN matrix. Table 2-1. The GaN and InN material parameters used in the Medici simulations. Estimated parameters in italics (Ref. 42a, 47b, 200c, 219d, 220e, 223f, 224g, 225h). Material Band gap energy (eV) Electron affinity (eV) Dielectric constant Valence band density of states (Nv, cm-3) Conduction band density of states (Nc, cm-3) Electron mobility (cm2/V-s) Hole mobility (cm2/V-s) GaN 3.46e4.1d 8.9b4.62 x 1019 g2.23 x 1018 g1000h200c5.20 x 1019 g9.15 x 1017 g1000200 InN 0.70a5.8f15.3b Table 2-2. Cell parameters for optimized cell structure with op timum band gap energy. Band gap (eV)Illumination Jsc (mA/cm2) Voc (V) Max Power (mW/cm2) Incident Power (mW/cm2) Fill Factor (%) Efficiency (%) 1.44AM0-24.950.91819.7212886.0915.36 1.44AM1.5-20.110.91215.779886.0116.03

PAGE 83

83 Table 2-3. Cell characteristics for simulations a llowing hot carrier collec tion in phase-separated InxGa1-xN solar cells. Matrix/Precipitate band gap energy (eV) Precipitate volume fraction Illumination Jsc (mA/cm2) Voc (V) Max Power (mW/cm2) Incident Power (mW/cm2) Fill Factor (%) Efficiency (%) 1.0 / 2.00.10 AM0 -39.160.59318.9312881.5314.75 1.0 / 2.00.10 AM1.5 -30.200.58614.409881.3614.64 1.0 / 2.00.20 AM0 -39.200.69322.6412883.3417.64 1.0 / 2.00.20 AM1.5 -30.220.68617.259883.2017.54 1.0 / 2.00.25 AM0 -39.230.77424.5012880.6919.09 1.0 / 2.00.25 AM1.5 -30.240.73618.699883.9719.00 Table 2-4. Characteristics of individual laye rs and overall cell for simulated multi-junction InxGa1-xN solar cells. Junction number Optimized band gap energy (eV) Individual absorber Voc (V) Individual cell max power (mW/cm2) AM0 Input power (mW/cm2) Overall Voc (V) Overall efficiency (%) 1 1.440.91819.72128 0.91815.36 2.001.46015.95128 1.200.66712.65128 2.301.73812.50128 1.601.06011.73128 1.000.4766.90128 2.501.93010.24128 1.801.25311.31128 1.300.7648.46128 0.900.3693.85128 2.602.0259.22128 2.001.4459.35128 1.601.0487.72128 1.200.6616.39128 0.800.2682.50128 22.12722.28 33.27424.25 44.31626.38 55.44727.41

PAGE 84

84 CHAPTER 3 GROWTH OF INDIUM NITRIDE AND INDIUM GALLIUM NITRIDE THIN FIMLS BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION 3.1 Introduction InN and InxGa1-xN are less studied materials when comp ared to semiconductors such as Si or GaAs. Ga-rich InxGa1-xN (0 x 0.3) alloys have become dominant materials as active layers in visible LEDs even though the mechanisms for light emission are not completely understood. InN and In-rich InxGa1-xN have been studied far less than Ga-rich InxGa1-xN primarily due to the few new device applications in the ~ 2 eV band gap energy range and the inability to grow high quality single crystal In N. More recently single crystal InN has been consistently reproduced by MBE and MOCVD growth techniques.68 Available high quality InN epitaxial layers have made it possible to investigat e the properties of InN in greater detail. The band gap energy controversy has sparked more interest in InN and In-rich InxGa1-xN films because it created new applications for these materials. These new applications include terahertz emitters and detectors as well as high efficiency solar cells. Before these new applications can be exploited it would be helpful if the fundamental properties such as relationships between growth conditions and the resulting phase separation, film composition, doping level, and crystalline quality were established. In this chapter InxGa1-xN phase separation will be discussed with respect to substrate material, buffer layer, growth temperature, and inlet composition (film composition). Currently high quality Ga-rich InxGa1-xN alloys are grown at high temperature (800 oC) for LED applications and In-rich InxGa1-xN is being grown for the fundamental study of In-rich InxGa1-xN alloy properties. A current summary of InxGa1-xN alloy growth progr ess is reviewed in Section 1.1.3. It has been shown experime ntally that approximately 30% indium can be incorporated into GaN at temperature above 700 oC.125-129 Low temperature growth by MBE has

PAGE 85

85 demonstrated metastable InxGa1-xN over the entire compositional range.47,130 No results have been published for InxGa1-xN over the entire compositional range by MOCVD. Also, no analysis has specifically examined InxGa1-xN growth parameters as they re late to phase separation. In general most literature work merely states that phase separation was notic ed and no correlation to the growth conditions is made. This work aims to understand how changes in deposition temperature, substrate material and, substrate pretreatments (including buffer layers) affect InxGa1-xN phase separation. 3.2 Experimental Procedure 3.2.1 Substrate Preparation InN and InxGa1-xN thin films were typically deposited on c-Al2O3 substrates, however, a-Al2O3, GaN/c-Al2O3, Si (100), and Si (111) subs trates were also included in selected runs. The GaN/c-Al2O3 substrates used in this study were obta ined from Uniroyal Optoelectronics GaN (5 m) grown by MOCVD on c-Al2O3. No chemicals were used to clean the GaN/c-Al2O3 surface and the substrates were cleaned with a high pressure nitrogen gun to remove any particles from the surfac e before loading. The sapphire and silicon substrates were cleaned in boiling trichloroethylene (TCE), acetone, and me thanol each for 5 min to remove organic material. A high pressure nitrogen gun was used to quickly dry the substrat es after removal from the methanol cleaning step. 3.2.2 The MOCVD Deposition Technique Metal organic sources used in MOCVD are t ypically liquids, finely crushed solids, or dissolved in a solvent and are cont ained in stainless steel bubblers. The pressure and temperature of the bubblers is adjusted to c ontrol the partial pressure of th e precursor species inside the bubbler. Adjusting the total pressure and flow rate of the carrier ga s controls the transport to the reactor. This dilute stream of metal organic va por and carrier gas is th en mixed with the other

PAGE 86

86 source materials, typically hydrides, and delivered to a substrate placed on a heated susceptor. The susceptor is commonly heated by RF induction, radiative (lamp), or resistance heating. The gaseous sources undergo complex homogeneous and he terogeneous reactions in the hot region at and near the heated substrat e to produce film growth.226,227 MOCVD reaction chambers are typically made of stainless steel or quartz and the shape of the reactor is designed to produce laminar flow over the susceptor. Most MOCVD systems operate at low pressure (50-200 Torr) because it reduces closed streamlin e flows which decrease dead volumes or vortices. This type of growth technique is popular for III-V materials and is typically used to grow epitaxi al single crystal films, howeve r, polycrystalline and amorphous films can also be grown.226,227 Reaction Chemistry for MOCVD GaN and InN. The formation of GaN or InN on the substrate surface from the reaction between the group III (TEGa or TMIn) and group V (NH3) sources. The overall reactions ar e given by Eqs. 3-1 and 3-2. (C2H5)3Ga (g) + NH3 (g) GaN (s) + 3C2H6 (g) (3-1) (CH3)3In (g) + NH3 (g) InN (s) + 3CH4 (g) (3-2) A general reaction for GaN or InN is represented by R3M + NH3 MN (s) + 3RH (g) (3-3) Where M = Ga or In, R = CH3 or C2H5.228,229 This expression does not reflect the actual reac tion pathways because the details are not very well established and the availabl e information suggests the reactio ns are complex. Jacko and Price were the first to study the pyrolysis of TMIn and suggested th e following mechanism.230 In(CH3)3 In(CH3)2 + CH3 (3-4) In(CH3)2 In(CH3) + CH3 (3-5)

PAGE 87

87 In(CH3) In + CH3 (3-6) DenBaars et al.231 later studied the decomposition of trimet hyl gallium in the growth of epitaxial GaAs and concluded the same mechanism as Jac ko and Price. More recently the decomposition of TMIn was studied by in-situ Raman spectroscopy.232 The results of this study suggest several reaction intermediates appear dur ing the decomposition of TMIn, nMMIn (MMIn)n (3-7) 2DMIn (DMIn)2 (3-8) DMIn + MMIn DMIn-MMIn (3-9) DMIn-MMIn CH3InCH2 + HInCH3 (3-10) TMIn + MMIn (DMIn)2 (3-11) where MMIn is monomethyl indium and DMIn is dimethyl indium. The most energetically favorable intermediate was calculated to be (DMIn)2 followed by the slightly less favorable production of DMIn-MMIn. The decomposition of ammonia at high growth te mperature is assumed to take place on the substrate surface or reactor walls to yield atomic nitrogen or a nitrogen containing radical. Thermodynamically it is known that NH3 decomposes completely into N2 an H2 at temperatures above 300 oC. When temperature is below 650 oC and no catalyst is used, NH3 decomposition is slow and strongly depends on grow th conditions and reactor design.233 It is also believed that the removal of the first hydrogen bond is the rate limiting step in ammonia decomposition.228 NH3 (g) NH3-x (g) + xH (g) (3-12) From the reaction in Eq. 3-12 a possible growth mechanism for InN and GaN at a solid gas interface is, M(CH3) (s/g) + NH (s/g) M-N (s) + CH4 (3-13)

PAGE 88

88 where M = Ga or In. MOCVD growth is not an e quilibrium process, therefore thermodynamics can only determine the overall driving force for the reaction and reactor kinetics or transport rates define the rate at which film growth will proceed. Thus the reaction pa thways and rate constants, along with the flow velocities and temperature gradie nts near the substrate are very important. A schematic of the precursors at the solid-gas in terface and boundary layer regions for the reaction of trimethyl gallium (TMGa) a nd ammonia to form GaN (Figure 3-1). A similar diagram is expected for TMIn and NH3 sources to form InN at the solid-g as interface. TMGa must diffuse through the boundary layer, possibly pyrolize, and then adsorb on to the substrate surface where the adsorbed Ga-containing molecule or atom r eacts with ammonia that also diffused through the boundary layer with possibl e reactions (Figure 3-1). The grow th rate of these MOCVD reactions is controlled by the reaction ra te at low temperature and di ffusion of the group III source though the boundary layer at higher temperatures wher e the reaction rates become high, since NH3 is usually in great excess. A dduct formation such as Ga(CH3)2-NH2 can also be important for high temperature growth (Figure 3-1). The growth kinetics and the grow th mechanisms that occur at the solid gas interface are not very well understo od, however, this has not prevented empirical growth studies.228 3.2.3 The MOCVD Reactor MOCVD experiments were performed in a horizontal, low pressure (100 Torr), coldwalled quartz reactor with RF-inductive heating of a tilted graphite susceptor. This MOCVD system was originally designed by Nippon Sanso for low temperature GaAs growth, but has subsequently been modified for InN, GaN, or InxGa1-xN growth.203,234 Solid trimethyl indium (TMIn, 99.9995%, Rhom and Haas Electronic Materials), liqui d triethyl gallium (TEGa,

PAGE 89

89 99.9995%, Rhom and Haas Electronic Materials), and ammonia (NH3, 99.9999%, Air Products) are used as precursors with a nitrogen carrier gas (UF Microfabritech building LN2 boil off). The MOCVD reactor schematic and a photo are s hown in Figure 3-2. Samples are loaded onto a quartz wafer tray and placed in a load-lock which is evacuated and purged 8 to 10 times to minimize oxygen contamination from reaching the reacto r. A mechanical fork is used to load the wafer tray onto the susceptor before star ting the reaction. During the experiment N2 is delivered to the metal organic precursor bubbler, whic h is then combined with a dilution N2 stream and the pure NH3 stream, just before entering the quart z chamber. The metal organic precursor temperature is controlled by NESLAB RTE (T hermo Electron Corporation) refrigerated bath/circulators. In these studies, the susceptor temperat ure was in the range of 450-850 oC and the quartz wall was maintained at a constant 25 oC with cooling water. Down stream waste gases are pumped by a dry vacuum pump (BOC Edwards XDS 10) and removed through the Microfabritech exhaust system. The growth parameters such as susceptor growth temperature, N/III ratio, substrate nitridation time and temperature, buffer layer ma terial thickness, total flow rate and reactor pressure are chosen depending on the desired epitaxial layer prope rties. The N/III ratio is the inlet molar ratio of NH3/TMIn (or NH3/TEGa) and the N/III ratio is controlled by adjusting the flow rates of nitrogen through the TMIn ( TEGa) bubbler and indepe ndently by the ammonia flow rate. Growth temperatures is controlled by a PID temperature c ontroller (Gultan West 2070) that uses a quartz insulated thermocouple to measure the susceptor temperature and adjust the power of a Lepal T-15 RF generator. The thermocouple is located in side the center of the graphite susceptor, via a dr illed hole in the susceptor.

PAGE 90

90 3.2.4 In-Situ Sapphire Substrate Surfa ce Treatment and Buffer Layers In-situ pretreatments such as nitridation of sapphi re substrates or th e addition of a low temperature (LT) InN buffer layers ar e used for the growth of InN and InxGa1-xN thin films. Nitridation of sapphire s ubstrates is applied to grade the lat tice mismatch between the III-nitride main layer and the sapphire substrat e by forming an AlN nucleation layer.70 LT InN buffer layers have also been shown to improve subseq uent film quality compared to samples without buffer layers.73,74 Aspects of substrate nitridation and bu ffer layers are discussed in greater detail in Section 1.1.3. Sapphire substrates were nitridated under an ammoni a flow (1600 sccm, 4 SLM N2 dilution) at high temperature (850 oC) for duration of 15 min prior to growth. After nitridation the reactor was allowed to cool (in a low pressure N2 atmosphere) to the appropriate growth temperature selected for buffer layer growth or the main layer growth. LT InN buffer layers were only used for some InxGa1-xN samples because the primary focus of this study was not to produce high crystalline quality but to understand how phase separation is related to growth conditions. When used, LT InN buffer layers we re grown for 15 min at a temperature of 450 oC and a N/III ratio of 50,000. MOCVD optimization of InN depositi on temperature, N/III ratio, buffer layer temperature/duration, and substrate nitridation temp erature/duration has been done elsewhere.203,234 3.3 Results 3.3.1 Indium Nitride 3.3.1.1 Growth on silicon substrates Silicon is widely used in the semiconductor i ndustry and the integration of nitride-based devices with Si IC technology is a driver for us ing Si substrates. Si (111) substrates have a lattice mismatch of -8% with InN, which is less than nitridated Al2O3 (AlN, 13%) and GaN

PAGE 91

91 (11%), while Si (100) has a lattice mismatch of 35% with InN. InN thin films were grown on p-Si (100) and n-Si ( 111) substrates at 530 oC and an inlet N/III ratio of 50,000. Surface pretreatment conditions such as nitridation and the us e of an InN buffer layer were varied to refine the growth conditions of InN on silicon substrates. Growth of InN directly on sili con substrates can lead to growth of incomplete films or polycrystalline films with very small and broad XRD peak intensities. Standard TCE, acetone, and methanol chemical cleaning steps were used to clean the Si substrates prior to growth. Direct growth on Si (111) subs trates without the use of a LT InN buffer layer or substrate nitridation forms an incomplete film of InN (Tg = 530 oC and N/III = 50,000, Figure 3-3). When a LT InN buffer layer is used a continuous film is formed (SEM, Figure 3-4), however the resulting films is polycrystalline, as seen from the appearance of the (002), (101), and (102) reflections in the XRD pattern (Figure 3-5). In N is believed to be pol ycrystalline when grown directly on silicon substrat es due to an amorphous SiNx layer that forms during the initial stage of growth from the exposure of Si to NH3.68 The formation of SiNx can be suppressed by growing a protective non-nitride heterostructured buffer laye r, such as GaAs followed by subsequent growth of InN, yielding a continuous and smooth surface morphology.235 Intentional substrate nitridation to form a crystalline -Si3N4 layer by reacting NH3 with the silicon substrate at high temp erature has also been tested.236 When brief substrate nitridation is used ( 15 min), however, it is likel y that a silicon oxynitride (SiOxN1-x) forms instead of a pure -Si3N4. This result has been shown by Kryliouk et al. using identi cal deposition equipment and substrate nitridation conditions.237 The formation of a silicon oxynitride intermediate layer has been shown to be an ideal method for produci ng single crystal GaN on silicon substrates. A TEM imag e of GaN/Si (111) shows a th in amorphous structure for the

PAGE 92

92 silicon oxynitride layer (Figure 36). Details about the growth and characterization of the SiOxN1-x intermediate layer can be found elsewhere.234,238 This result encourages performing silicon substrate nitridation followed by growth of a LT InN buffer layer to grow single crystal InN on Si (111) and (100). The sa me sapphire nitridation step (850 oC, 1600 sccm NH3 for 15 min) was applied to silicon substrates (as descri bed for sapphire in Section 3.2.4), and InN was grown for 1 hr at 530 oC (N/III = 50,000). XRD patterns of InN on Si (111) and Si (100) are shown in Figure 3-7. Comparing the patterns fo r InN/Si (111) (Figures 3-5 and 3-7) it can be seen that the use of substrate nitridation produces a highly textured InN film with a preferred growth orientation in th e [0001] direction and the InN (101) re flection can no longer be detected. When substrate nitridation and a LT InN buffer la yer are used for InN growth on Si (100) only the InN (002) reflection is detect ed and no polycrystalline peaks a ppear. The Si (200) substrate peak occurs at a 2 value which is close to the polycryst alline InN (101) peak, which can cause the polycrystalline InN peak to go unno ticed. Adding a slight tilt (3 or 4o) in the omega angle (normal to the film surface) during an XRD scan will cause the single crystal silicon peak to be undetected while the less crystalline InN rema ins unchanged. With the Si substrate peak removed, no polycrystalline InN (101) peak is detected (Figure 3-7). For the nitridated Si substrates, there is little differe nce in the magnitude of the peak intensities of the (002) reflection for InN/Si (111) and InN/Si (100) even though there is a large diffe rence in lattice mismatch with respect to InN. The XRD peak intensity is strongly dependent upon the specimen interaction volume (film thickness and surface area) and the crystallinity of the film. The growth rate of InN/Si (100) was determined to be ~ 80 nm/hr from cross-sectional SE M (Figure 3-8) and a similar growth rate was found for InN/Si (111). Therefore the film thic kness can be ignored as contributing factor to any differe nce in InN (002) peak intensity. Films were typically grown on

PAGE 93

93 cut 10 x 10 mm2 substrates, which eliminates any sample size affect on the XRD peak intensity. This result suggests that the oxynitride layer has more influen ce over the film crystallinity than the silicon substrate orientation. The intermediate SiOxN1-x layer was previously determined to have an amorphous structure,234 which may account for the sma ll InN (002) peak intensities when grown on Si. Even with the use of substrate nitridation the InN (002) peak intensities are relatively small compared to InN/c-Al2O3, grown under otherwise identica l conditions and therefore the crystalline quality of InN/Si c ould be significantly improved. The growth rate of InN/c-Al2O3 is ~ 100 nm/hr and was previously determined for the same deposition equipment used in this study.203 Since InN has a similar growth rate and similar sample sizes (10 x 10 mm2) were tested, then the difference in InN (002) peak in tensity can be attribut ed to a difference in crystalline quality. InN/Si was grown for two hr to account for any growth rate difference for c-Al2O3 and Si substrates, which again yiel ded a much larger peak for InN/c-Al2O3. For this reason, InN thin films were gr own on Si (111) substrates by using a hydrideMOCVD growth technique. In this technique, HCl reacts with th e metal organic species (TMIn) to form volatile InCl, which then reacts with NH3 to form InN. A more detailed description of the H-MOCVD deposition technique is given in Section 5.2.2. This technique is beneficial for growth because the presence of HCl cleans residual SiOx on the Si surface and takes advantage of the small lattice mismatch (-8%) between InN (002) and Si (111). The oxynitride cannot form because HCl reacts with SiOx to form volatile SiHnCl4-n and H2O. Other benefits of the H-MOCVD technique include the ability to grow InN at low N/III ratios due to preferential etching of In droplets by HCl.234 H-MOCVD growth of InN also allows for high growth rate of

PAGE 94

94 InN without sacrificing growth quality, which would occur if a conventional MOCVD used a similar high concentration of reactants. Thus InN growth on Si was attempted in a different reactor by HMOCVD as described below. After standard chemical cleaning, the Si surface was cleaned in-sit u (prior to growth) by annealing in a H2 atmosphere for 10 min at 850 oC. The reactor was allowed to cool in a H2 atmosphere to the buffer layer growth temperature of 450 oC. HCl gas was allowed to flow into the reactor for 2 min prior to buffer layer growth (450 oC) with no other source materials flowing to provide secondary cleaning of the silicon s ubstrate. A LT InN buffer layer was grown (in N2 carrier) for 15 min (N/In = 2,500 and 450 oC) followed by InN growth at 560 oC (N/In = 2,500) for 2 hr. When grown on Si (111) in the H-MOC VD system, InN is highly textured in the (002) direction and with a la rge and narrow peak intensity signi fying increased crystalline quality compared to InN/Si (111) grown by conven tional MOCVD (XRD, Fi gure 3-9). H-MOCVD InN/Si (111) grows in a columnar structure wi th a very rough surface where grain sizes range from 150 to 500 nm (SEM, Figure 3-10). It is sugge sted that the columnar structure is made up of vertical InN (002) columns since the XRD resu lts show a highly textured film in the [0002] direction. Another benefit of H-MOCVD besides increased crystalline quality is that the growth rate of InN (750 nm/hr) is an order of magnitude high er compared to traditi onal MOCVD (78 nm/hr). It must be noted that a signifi cantly higher TMIn flow rate is used in the H-MOCVD (0.67 sccm) compared to conventional MOCVD (0.03 sccm). It is difficult to s cale up a conventional MOCVD system to achieve this high growth ra te because it would require an extremely large NH3 flow rate (35 SLM) to maintain the requ ired inlet N/III ratio of 50,000 for preventing indium droplet formation. The addition of HCl using the H-MOCVD system allows for a

PAGE 95

95 moderate NH3 flow rates of 1.7 SLM to be used. Th e results of energy dispersive spectroscopy (EDS) and Auger electron spectroscopy (AES) analysis confirmed that no chlorine was detected in the film (within the available detection limits, ~ 0.5-1 atomic %, Figure 3-11). Growth of InN on Si substrates is desired for several reasons including the reduced cost of silicon, large area substrates, and future devi ce integration. Substr ate nitridation and the formation of a SiOxN1-x intermediate layer is shown to be critical to form hi ghly textured InN (002) when grown on Si (111) and Si (100) by MO CVD. The amorphous SiOxN1-x layer, however, affects the ability to form high quali ty InN. H-MOCVD proves to be an ideal technique for growth of InN/Si (111) because the addition of HCl prevents the oxynitride from forming and the lattice mismatch remains at -8%. The low lattice mismatch produces a highly textured InN (002) film and a higher growth rate was possible with H-MOCVD at a low N/III ratio without the formation of indi um droplets. This is a huge a dvantage compared to traditional MOCVD where InN is typically grow th at N/III ratios of 50,000. 3.3.1.2 Film stability and aging Long term film stability is important for future device integration of InN based materials. If InN is not stable in oxygen containing environmen ts such as air then fa brication steps such as annealing of electrical contacts becomes more diffi cult. It has been s uggested that oxygen is incorporated into InN films after air exposure at room temperature, for example sputtered InN films have been shown to incorporate oxygen into the grain boundaries of InN and form crystalline In2O3 phases at room temperature.50 This room temperatur e annealing process was determined over a period of mont hs to years. To independently verify this room temperature annealing process, InN thin films were grown on a-,c-, and rorientations of Al2O3 as well as Si (100) and Si (111) by MOCVD. For all orientations of Al2O3 and Si substrates, substrate nitridation and a LT InN buffer layer were used. Post growth characte rization was done by XRD

PAGE 96

96 and then the samples were stored in air for a pe riod of 12 to 15 months. After aging (i.e., room temperature annealing) in air the same samples were analyzed again by XRD (Figure 3-12). There is no evidence of In2O3 crystalline phase formation when InN samples (on c-Al203 or Si (100) substrates) are aged in air for periods exceeding one year (Figure 3-12). It is important to mention that amorphous oxide phases are und etectable by XRD. This analysis does not support the claim by Butcher and Tansley50 that crystalline indium oxi de phases can form in InN after room temperature aging. However, sputte red films were used in the reference case and MOCVD grown InN was used for this work. It is possible that grain bounda ries play a role for oxygen incorporation into InN films. Sputtered films are polycryst alline which lead to a large number of grain boundaries and impurity diffusion occurs more rapidly through grain boundaries compared to the bulk for single crystal films.239 No differences in In2O3 formation were noticed for single crystal InN (InN/c-Al2O3) and polycrystalline InN (InN/Si (100)), which differ by the amount of grain boundary densities and area availa ble for oxide growth and oxygen transport. From this analysis it remains unclear how crysta lline oxide phases could form in InN at room temperature. From the results above, formation of crysta lline indium oxides does not occur at room temperature to an extent detectable by XRD, at higher temperature, however, crystalline oxide phases are more likely to form. Yodo et al.240 found that crystalline In2O3 (222) forms in InN upon annealing at 500 oC for 5 min in a N2 atmosphere (1 atm). These MBE InN samples were deposited at 500 oC for a much greater time than the annealing time and no In2O3 XRD peak was evident in the as grown film. SIMS analysis determined a 1% residual oxygen concentration in the as-grown film. These MBE samples were deposited under In-rich co nditions and therefore In-droplets formed on the surface. It was conc luded that the indium droplets reacted with

PAGE 97

97 oxygen to form amorphous indium oxides upon air exposure and these phases will crystallize after annealing at elevated temperature. For comparison InN/c-Al2O3 substrates were grown by MOCVD using substrate nitridation and a LT InN buffer layer and then exposed to air. These samples were then aged for 6 months to ensure that sufficient time was gi ven for oxygen to diffuse into the film. A high N/In ratio (50,000) was used for growth therefore no In droplets were present on the surface. After aging, InN was annealed in a low pressure (100 Torr) N2 atmosphere for 20 min at 500 oC, 15 min at 525 oC and 10 min at 550 oC. The as-grown samples ar e compared to the annealed samples (Figure 3-13). The In2O3 (222) peak occurs at 2 = 30.6o and it is clear (Figure 3-13) that no crystalline indium oxide is fo rmed when InN is annealed up to 550 oC in a low pressure N2 atmosphere. For all three ann ealing conditions the InN (002) peak intensity is increased when the same XRD setup was used, suggesting that th e film crystallinity was improved. A slight narrowing ( FWHM = -144 arcsec) of the InN (002) peak was found for annealed samples compared to the as-grown sample. From this anal ysis it can be determined that annealing in low pressure N2 atmospheres can prevent the formation of indium oxide crystalline phases in InN (without In droplets), even after exposure to air. These result s also suggest that the growth of stoichiometric InN films is important fo r preventing the formation crystalline In2O3. 3.3.2 Indium Gallium Nitride 3.3.2.1 Metastable InxGa1-xN alloys over the entire range (0 x 1) InxGa1-xN alloys were grown on c-Al2O3 substrates without an InN buffer layer at 530 oC at varying composition over the entire compositional range (0 x 1). Each film has approximate thickness of 100 nm, based on previous growth rate analysis.203 Typically a fixed deposition temperature is not used for growth over the entire compositi onal range because higher Ga compositions achieve better crystallinity at higher temperature. Low deposition temperature,

PAGE 98

98 however, has been used to grow metastable InxGa1-xN alloys by MBE over the entire compositional range.47,130 These metastable films have onl y been grown by MBE techniques and currently no evidence has been published for metastable InxGa1-xN films grown over the entire compositional range by MOCVD. In this work a low temperature approach wa s used to suppress phase separation. The InxGa1-xN film composition was controlled by varyi ng the inlet group III flux (TEGa or TMIn flow rate). The inlet flow ratio, TMIn/(TMIn + TEGa) or simplified as In/(In+Ga), is the molar ratio of the TMIn molar flow to the total inlet group III molar flow. The ammonia flow rate was adjusted to maintain a constant N/III ratio of 50,000 for each experiment. XRD patterns of pure InN, pure GaN, and seven di fferent compositions of InxGa1-xN are shown in Figure 3-14. The pure InN (002) reflection located at 2 = 31.4o shifts to the right (towards the GaN (002) peak) as the flow ratio is changed to lower values (more Ga-rich) (Figure 3-14). Also, it is important to notice that there is no second re flection in the pattern that wo uld suggest microscopic phase separation. These results are c onsistent with the hypothesis that a low growth temperature will suppress phase separation. This is th e first time single phase deposition of InxGa1-xN has been demonstrated over the entire compositional range usi ng MOCVD. It is also evident that the low growth temperature significantly affects the crysta lline quality, since high crystalline quality for GaN usually occurs at temp eratures at or above 850 oC. A table of inlet flow ratio and the corres ponding film composition (x) as determined by XRD for each run, is shown in Table 3-1. Bra ggs law (Eq. 3-14) is used to determine the distance between atomic planes, d-sp acing, by using the experimental 2 values recorded for the InxGa1-xN (002) peak at each flow ratio. ) sin( 2 d n (3-14)

PAGE 99

99 In the equation for Braggs law, n is an integer and is the wavelength of the incident X-rays (CuK = 1.54056 ). Using the d-spacing valu e calculated from Braggs law and the corresponding crystal plane indices, the a and c lattice parameters can be determined (Eq. 3-15). Simplifications of Eq. 3-15 are shown for InxGa1-xN (002) and InxGa1-xN (101) (Eqs. 3-16 and 317, respectively). The relations hip between cubic indices and he xagonal indices are given in Eq. 3-18. For hexagonal structures: 2 2 2 2 2 23 4 1 c l a k hk h d (3-15) InxGa1-xN (002): 2 2 2 24 1c c l d or n d c sin 4 2 (3-16) InxGa1-xN (101): 2 2 2 2 23 4 1 c l a h d (3-17) Cubic (h k l) Hexagonal (h k i l), where i = -(h + k) (3-18) The InxGa1-xN composition approximately follows a linear dependence with a slope of one for all compositions tested (Figure 3-15). The compos itional dependence shows a negative deviation (xfilm < xinlet) for In/(In+Ga) > 0.5 an d a positive deviation (xfilm > xinlet) for In/(In+Ga) < 0.5. A similar linear dependence on composition has been shown by Matsuoka et al. for Ga-rich InxGa1-xN alloys grown by MOCVD (Figure 1-11).12 The positive and negative compositional deviations can be explained by the suppressi on of InN decomposition by GaN coverage and surface site blocking by methyl radi cals, respectively. It is know that InN is reaction limited at the low deposition temperature used (530 oC) due to decreased NH3 decomposition efficiency.241,242 Talaleav et al studied kinetic effects limiting the growth rate of III-V compounds by MOCVD and concluded that the bloc king of group III species by methyl radicals was a dominant mechanism for reducing the growth rate.243 In this case of InxGa1-xN growth at

PAGE 100

100 low temperature (530 oC), the addition of TEGa (for In/(In+G a) > 0.5) is believed to reduce the growth rate of InN due to site blocking by Ga atoms and methyl radicals. This leads to a negative deviation from the linear relationship betw een composition and inlet flow ratio. As the inlet flow ratio becomes greater than 0.5 it is believed that surface coverage by GaN suppresses thermal decomposition of InN, which increases th e solid indium composition in the film. The suppression of InN decomposition produces a positiv e deviation from the linear dependence of composition to the inlet flow ratio. 3.3.2.2 Effect of growth temperature on stability of In0.8Ga0.2N To understand the temperature dependence on phase separation, InxGa1-xN alloys were grown at a constant flow ratio (In/(In+Ga) = 0.8) a nd deposition temperature was varied in the range from 530 to 850 oC in 30 oC intervals. Films were grown on c-Al2O3 substrates without an InN buffer layer with an approximate thicknesse s of 300 nm, as estimated by previous growth rate analysis.203 A constant N/III ratio of 50,000 was main tained at each growth temperature. A value of In/(In+Ga) = 0.8 was chosen for this study because phase sepa ration is typically not seen with films grown at this inlet In fraction.119,244 As the indium inlet fraction decreases below 0.8 phase separation is more likely to occur.245 The results given in th e previous section suggests that phase separation is suppressed when using a low growth temperature and it is postulated that phase separation would be more likely to occur at higher deposition temperature. At higher deposition temperatures Ga is more likely to incorporate into the solid solution due to the increased thermal decomposition of InN. This re sult has been verified experimentally (Figure 1-11) for a deposition temperature of 800 oC.12 Increasing the deposition temperature will change the InxGa1-xN composition for the same In-rich fl ow condition used to grow stable In0.8Ga0.2N (at 530 oC). It is believed that higher depos ition temperature will lead to phase separation since the Ga content will increase (at a constant inlet flow ratio), and the InxGa1-xN

PAGE 101

101 solid solution will approach unstable compositions (x > 0.3). The inlet flow value of 0.8 was chosen as a base line to analyze the effect of temperature on phase separation, since it has been demonstrated to be stable at Tg = 530 oC. The majority of incident solar energy lies in the band gap energy range of 0.7 to 2.0 eV, therefore for solar cell device applications In-rich InxGa1-xN compositions are de sired. Generally it is desired to fabricate semiconductor layers with low defect density and high crystalline quality for electronic applications. Hi gh quality materials make it easier to control the device performance, and for solar cells the efficiency of a high quality device incr eases efficiency due to the more efficient collection of genera ted carriers. One way to increase InxGa1-xN film quality is to increase the deposition temperature. This increases atom surface mobility and allows for atoms to find a lower free energy adsorption site leading to less defects and a higher degree of crystallinity. There is a trade off for increasing the InxGa1-xN deposition temperature to achieve better crystallinity because higher temperature leads to more rapid InN decomposition. For this reason it is important to study the effect s of crystalline quality of In-rich InxGa1-xN alloys at elevated temperatures to predict the appropria te deposition temperatur e and the occurrence of phase separation. InxGa1-xN thin films were grown on nitridated c-Al2O3 substrates without the addition of a LT InN buffer layer. It is possible for an In N buffer layer to influence phase separation by creating more lattice matched nucleation sites for InN deposition compared to GaN, therefore the buffer layer was omitted. Substrate nitridation occurred at 850 oC for 15 min as previously described in Section 3.2.4. Several InxGa1-xN thin films were grown with a constant inlet flow ratio of In/(Ga+In) = 0.8 at 530, 560, 590, 620, 650, 680, 710, 740, and 770 oC and characterized by

PAGE 102

102 XRD (Figure 3-16). InN (530 oC) and GaN (850 oC) samples grown with the same sequence are included as a reference (Figure 3-16). At first glance these XRD patterns seem complex, however, the peaks can be systemically separate d to understand the phase separation that occurs at different growth temperatur es. It is noted that InN th in films that are grown on c-Al2O3 substrates without the use of LT InN buffer layers typically produce polycrystalline films, due to the large lattice mismatch between the substrat e and thin film. Reflections from other orientations such as InN (100) and InN (101) can occur at 2 = 28.99o and 33.18o, respectively. Examining the XRD pattern for pure InN (F igure 3-16), a peak is evident at 2 = 33.2o, consistent with a small amount of InN (101). A similar result is noticed for InxGa1-xN grown at 530 oC, where InxGa1-xN (100), InxGa1-xN (002), and InxGa1-xN (101) peaks are observed at 2 = 29.89o, 31.87o and 33.93o, respectively (Figure 3-16). The peak intensity ratios indicate that the film is highly textured InxGa1-xN (002). It is important to acknowledge the presence of polycrystalline peaks, especially InxGa1-xN (101) peaks, because they can be misinterpreted as Ga-rich InxGa1-xN (002) since the range of possible 2 values for the solid solution contains the reflections of the other planes. The most important region to analyze lies between the InN (002), 2 = 31.4o and the GaN (002), 2 = 34.6o vertical lines (Figure 3-16). This is the entire range where InxGa1-xN (002) peaks are expected. The peaks to the right of GaN (002) correspond to the c-Al2O3 substrate and are labeled as such. The XRD peaks cl ose to the InN (002) reflection at 2 = 31.4o are In-rich InxGa1-xN (002) phase while the peaks closer to the GaN (002) 2 values (34.6o) are either Ga-rich InxGa1-xN (002) or In-rich InxGa1-xN (101) peaks. To clarif y, it is possible that InxGa1-xN (100) reflections can also appear close to the 2 value of InN (002), 2 = 31.4o, if the

PAGE 103

103 Ga content is sufficiently high. These (100) reflections are not lik ely to be confused with (002) reflections because the intensity ratio is si gnificantly higher for the (002) reflection. InxGa1-xN thin films grown at 530 oC do not show microscopic phase separation (Figure 3-16), which is consistent with the results di scussed in Section 3.3.1. When the deposition temperature is increased, peak splitting occurs which is direct evidence for phase separation in the film. Separate InxGa1-xN (002) peaks at different values of 2 (composition) are resolved for deposition temperature in the range 560 to 620 oC. In the temperature range from 650 to 680 oC the XRD peaks broaden, making it more difficult to resolve individual peaks. At deposition temperature of 710 oC or above, specific peaks are again ea sier to resolve and phase separation occurs with two very different compositions Table 3-2 lists the peak positions and corresponding InxGa1-xN composition(s) (as determined by XRD) for each deposition temperature. It is difficult to compare the effect of de position temperature on film crystalline quality when XRD data is plotted on a logarithmic scale (F igure 3-16). For this reason the magnitude of the InxGa1-xN (002) peak intensity was plotted for each deposition temperature (Figure 3-17). Each experiment used the exact same flow cond itions and the sapphire subs trates were the same size, therefore an increase in the XRD peak intens ity can be attributed to an increase in film crystallinity or structural quality. It is also possible that higher temperat ures could increase the growth rate and produce higher XRD peak inte nsities due to thickness changes. Adachi et al. found that the MOCVD InN growth rate incr eased over the temper ature range of 500-620 oC when the TMIn supply was constant.246 The growth rate, however, cannot be the only contributing factor the XRD peak intensity sin ce InN film decomposition also occurs in this same temperature range.

PAGE 104

104 The stable InxGa1-xN film grown at 530 oC has a low XRD peak inte nsity (Figure 3-17). The crystallinity increases significantly when the deposition temperature is increased to 560 oC, which is attributed to a reduction in strain si nce the film becomes phase separated and slight increase in growth rate. It is also possible that the higher temperature increases surface atom mobility leading to better cr ystallinity. No significant decomposition occurs at 560 oC and other researchers grow pure InN by MOCVD at this temperature.74,247,248 Since many factors contribute to the XRD peak intensity, the FWHM of the InxGa1-xN (002) peaks are plotted as a function of deposition temperature for 530, 560, 590, and 620 oC (Figure 3-18). Phase separation is noticed for the three highest temperatures (Figure 3-18) and the FWHM was measured for each of the InxGa1-xN (002) phase separated peaks. The data points for the phase separated temp eratures represents the averag e FWHM, while the extremes of the error bars represents the FWHM of each phase separated InxGa1-xN (002) peak. The highest FWHM (3672 arcsec) is noticed for the InxGa1-xN film deposited at 530 oC (Figure 3-18). The FWHM reaches a minimum value of 936 and 1440, for the two highest intensity (dominant) phase separated peaks at a deposition temperature of 560 oC. The values for the FWHM increases for deposition temperature of 590 and 620 oC, and the value for the FWHM, however, is still lower than the metastable InxGa1-xN (530 oC). These results suggest that the crystalline quality of the phase separated InxGa1-xN films are better than the metastable InxGa1-xN film grown at 530 oC. The results indicate a deposition temperature of 560 oC should be used to obtain the best crystalline quality at the inlet fl ow ratio tested (In/(In+Ga) = 0.8). If phase separation cannot be tolerated, th e low deposition temperature of 530 oC must be used. A decrease in crystallinity is seen at 590 oC (compared to 560 oC) possibly due to increased decomposition of InN, or a higher growth that leads to 3D island growth instead of the

PAGE 105

105 more desirable 2D growth (from a structural qua lity perspective). It has been show by other researchers that InN film quality degrades at temperatures ~ 600 oC.249,250 Similar film degradation (InN decomposition and 3D growth) mechanisms can be used to describe the decreased crystal line quality at 620 oC. At deposition temperature > 650 oC the decomposition rate of InN becomes rapid, even though ammonia also decomposes rapi dly in this temperature regime,66,246 thus reducing the InxGa1-xN (002) peak intensity (Figure 3-17). Highly efficient decomposition of NH3 leads to an increase of partial pressure of hydrogen at the higher deposition temperature (> 650 oC), which retards the growth rate of InN. Therefore when using an inlet flow ratio of In/(In+Ga) = 0.8 the deposition temperature must remain near 530 oC to produce stable InxGa1-xN films. At a growth temperature of 560 oC the InxGa1-xN film becomes phases separate d, however, the crystalline quality increases for the phase separated regions The intermediate temperature range 650 to 710 oC produces very small InxGa1-xN peak intensities which are slightly above the background XRD intensity. The small intensities are most lik ely due to InN decomposition as well as a low GaN growth rate at these temper atures. Higher growth rate (and higher quality) GaN is achieved at deposition temperature of 850 to 1150 oC.234,251 Polycrystalline In-rich InxGa1-xN (101) peaks occur at 2 = 33.85o for low deposition temperatures (< 650 oC) (Figure 3-16). When the depos ition temperature is increased to 710 oC or above, a Ga-rich InxGa1-xN (002) phase starts to form in the region where In-rich InxGa1-xN (101) peaks would be expected. This resu lt can be seen from the magnitude of InxGa1-xN (101) peak intensities for each deposition temperature (F igure 3-19). There is an increase in peak intensity (in the region of 2 = 33.8o) from T = 710 to 740 oC, and this is attributed to the formation of Ga-rich InxGa1-xN (002) phases since In-rich InxGa1-xN phases are less likely to

PAGE 106

106 occur at these temperatures. Even though In-rich InxGa1-xN (101) and Ga-rich InxGa1-xN (002) XRD peaks can overlap, In-rich InxGa1-xN (101) peaks are dominant at lower (< 650 oC) deposition temperatures and Ga-rich InxGa1-xN (002) are dominant at higher (> 710 oC) deposition temperatures. This result can be attr ibuted to the increase growth rate of GaN at higher deposition temperatures as well as the increas ed InN thermal decomposition. It is also not likely that the polycrystalline In-rich InxGa1-xN (101) peak would be detected with a significant peak intensity when the In-rich InxGa1-xN (002) peak intensity is al so low, since the films are textured in the [002] direction. At 740 oC the intensity of the Ga-rich InxGa1-xN (002) phase increases, which is not surprising because GaN is more favorable at higher temperature. It is surprising, however, that the XRD intensity of the In-rich InxGa1-xN (002) also increases at th is temperature compared to 710 oC. It is not clear why the In-rich phase show ed an increase in XRD intensity at a higher temperature, but it is possible that an increase in the GaN growth rate allows more surface coverage of InN and inhibits thermal decompos ition. An increase in XRD intensity due to thickness is also not likely, especially for In-ri ch phases due to excess hydrogen partial pressure inhibiting the growth rate of InN. A 30 degree in crease in temperature (770 oC) no longer shows the presence of an In-rich InxGa1-xN (002) phase, only Ga-rich InxGa1-xN (002) is present. This result is consistent with the work done by Matsuoka et al. who found that Ga-rich InxGa1-xN alloys still form at high deposition temperatur e even when inlet flow conditions are In-rich.12 To summarize, InxGa1-xN films can have been grown in the temperature range 530-770 oC using a constant inlet flow ratio of TM In/(TMIn + TEGa) = 0.8. Growth at 530 oC produces a stable InxGa1-xN film, but as the temperature is increased phase separation occurs. At lower deposition temperature (560, 590, and 620 oC) the separate InxGa1-xN phases have compositions

PAGE 107

107 that lie in the In-rich regime. These deposition te mperatures also show an increase in crystalline quality, attributed to a reducti on in strain from phase separa tion, and the best quality was obtained at a depositi on temperature of 560 oC. At Tg = 590 oC, two InxGa1-xN (002) phases can be resolved from the XRD data that correspond to In0.86Ga0.14N and In0.71Ga0.29N, and a similar result occurs at 560 and 620 oC. In the intermediate temperature range (650 to 710 oC) produces InxGa1-xN films with very broad XRD peaks and poor structural quality. At 740 oC, separate InxGa1-xN (002) phases are noticed with one phase In-rich (In0.86Ga0.14N) and the other Ga-rich (In0.27Ga0.77N) phase. The highest deposition temperature used, 770 oC, produced an InxGa1-xN film with no distinct In-rich peak, only a Ga-rich peak (In0.28Ga0.72N) with a broad shoulder to the left, which signifies phase separation with a lesser gallium content. Films deposited at 560 oC, had three distinct InxGa1-xN (002) compositional phases that could be resolved from XRD data, In0.85Ga0.15N, In0.7Ga0.3N and In0.4Ga0.6N. These phases are easily identified for Tg = 560 oC because this temperature produced the highest peak intensities of all the temperatures examined. It is possible that film s deposited at other depos ition temperature also had several InxGa1-xN (002) compositional phases present, however, these phases cannot be specifically identified due to XRD p eak broadening and low intensities. 3.3.2.3 Effect of substrate on stability of InxGa1-xN alloys An early prediction of the miscibility gap of InxGa1-xN by Ho and Stringfellow123 indicates the limit is ~ 6% Ga incorporation into InN or In incorporation into GaN, at typical growth temperature (800 oC). More recent analysis has shown that the introduction of compressive strain into in InxGa1-xN alloy films reduces the miscibilit y gap (when coherently deposited on unstrained GaN).124 The amount of strain (compressive or te nsile) in an epitaxial film is directly related to the lattice mismatch between the substr ate or buffer layer and the film. To understand the substrate effect on the film stability, InxGa1-xN alloys were grown on a-Al2O3 and

PAGE 108

108 GaN/c-Al2O3 substrates and compared to metastable InxGa1-xN alloys grown on c-Al2O3. The same growth temperature was used for each substrate, Tg = 530 oC along with N/III = 50,000. Substrate nitridation and a LT InN buffer layer we re used for each growth with overall film thicknesses of ~ 100 nm. A series of InxGa1-xN films grown at 530 oC on the a-plane Al2O3 during which the inlet flow ratio (TMIn/(TMIn+TEGa) was varied from 0 to 1 were characterized by XRD (Figure 3-20). Immediately it can be seen that a-Al2O3 substrate (Figure 3-20) di ffers significantly from c-Al2O3 substrate (Figure 3-14) with respect to ph ase separation. The films grown on c-plane Al2O3 substrates are single phase films while th e films grown on a-plane sapphire show obvious phase separation from XRD peak separation. Al though, phase separation occurs most frequently in the compositional range of 0.3 < x < 0.7,245 this is not the result seen for InxGa1-xN grown on a-Al2O3 substrates. Phase separation occurs at inlet flow ratios of 0.7 and 0.6 but not at 0.5. Interestingly, values of flow ratio of 0.1 and 0. 3 show phase separation, while the flow ratio of 0.2 does not show phase separation. The lattice mismatch between a-Al2O3 and InN or GaN is -8.8% or 1.9%,5 respectively. This gives a lattice matched condition at In0.8Ga0.2N. In the case of InN/a-Al2O3 the reduced lattice mismatch is beneficial because the structural quality of InN is improved due to reduced strain in the heteroepitaxial film (compared to c-Al2O3). This prediction has been verified experimentally by growing InN on both aand c-Al2O3 substrates during the same growth. Substrate nitridation was used as well as a LT InN buffer layer for the growth of InN on aand c-Al2O3 (XRD, Figure 3-21). The FWHM of the In N (002) peak decreases from 1080 (c-Al2O3) to 864 arcsec (a-Al2O3) confirming that the structural qual ity of InN is improved when growth occurs on a-plane sapphire. The difference in cr ystalline quality for the aand c-planes of

PAGE 109

109 sapphire would be even larger if substrate ni tridation was not used (due to improving the mismatch between InN and c-Al2O3). The a-plane of sapphire w ould seem to be the preferred substrate for growth of InxGa1-xN alloys from a structural quality perspective due to a decreased lattice mismatch between the substrate and the film From the experimental results, however, the lack of substrate induced strain shifts the binodal decomposition more towards ideal solution behavior, leading phase separation to occur ove r a wide range of compositions. Assuming a regular solution model predicts a large InxGa1-xN alloy miscibility gap and a high critical temperature (1250 oC).123 These experimental results sugges t that substrate selection is an important parameter when considering the effects of phase separation for specific InxGa1-xN compositions. In the case of c-Al2O3 the lattice mismatch between the substrate and the alloy film is sufficiently high, ranging from 33 to 26% when x = 0 and 1, re spectively. The large lattice mismatch produces elastic strain (compr essive) which helps to suppress phase separation at the compositions tested (based on the results presented in Section 3.3.2.1). Therefore it is suggested that the elasti c strain induced by the substrate stabilizes the InxGa1-xN alloy film from phase separation. Thermodynamic analysis of the InxGa1-xN alloys system has shown that phase separation can be suppressed (miscibility gap reduction and also a lowering of the critical temperature) when compressive or tensile stress is include d in the Hemholtz free energy calculation.252 The Hemholtz free energy (F, Eq. 3-19) of a system can be described by contributions from the macroscopic (Fo, Eq. 3-20) and microscopic ( F, Eq. 3-21) mixtures252 F = Fo + F (3-19) Fo(x,T) = (1-x)FGaN(T) + xFInN(T) (3-20) F(x,T) = U(x,T) T S(x,T) (3-21)

PAGE 110

110 J j J j jx x M M T x U0 0) 1 ( ) ( (3-22) where U is the mixing internal energy (identifie d with the mixing enthalpy of the alloy), S is the mixing entropy, M is the cluster number, is the cluster energy, and j is the cluster class index (when using a quasichemical approach to disorder and composition and a pseudopotentialplane-wave approximation for the total energy). The Fo term can be calculated by determining the standard Hemholtz energy of FGaN and FInN using standard thermodynamic properties of binary compounds. The internal energy (Eq. 3-22) is a function of elastic energy ( ) and introducing strain into the system causes the elas tic energy to decrease and therefore the excess free energy ( F) decreases. Lowering the excess free en ergy reduces the critical temperature to overcome phase separation and reduces the miscibility gap (when assuming a uniform distribution of strain). The critical temperature is decreased by 120K (to 1130K) when -5% biaxial strain is included and the miscibility gap is reduced more for In-rich compositions due to a compositional dependent interac tion parameter that produces asym metry. Allowing the biaxial strain to be distributed i nhomogeneously further reduces F so the values become negative over the entire compositional range (at T > 700K). The significant lowering of F occurs because the strain energy is proportional to the square of the in-plane strain en ergy (biaxial strain). In this case the miscibility gap and the critical temperature are significantly reduced.252 These theoretical predictions suggest that a lack of substrate induced st rain (due to the smaller lattice mismatch) favors a phase separation for InxGa1-xN when deposited on a-Al2O3. Although the strain from the substrate helps prevent phase separation, the crystalline quality is reduced due to the high level of strain in the alloyed film. An increase in InxGa1-xN strain, as determined by a decrease in XRD FWHM of the InxGa1-xN (002) peak, is in good agreement with the suppressi on of phase separation for InxGa1-xN/a-Al2O3 (Figure 3-22). As the

PAGE 111

111 flow ratio changes to include more Ga in the film (In/In+Ga = 0.9 and 0.8) the strain energy increases, which increases the InxGa1-xN (002) FWHM, signifying a decreased crystalline quality (Figure 3-22). At a flow ratio of In/In+Ga = 0.7 the film strain cannot be stabilized by the substrate and phase separation occurs. Upon phase separation the FWHM decreases, suggesting an increase in crystalline quality. A similar tren d can be seen for the remaining inlet flow ratios (Fig. 3-22), where the stable films (at In/In+Ga = 0.5 and 0.2) are less crystalline compared to the phase separated films where th e strain has been relaxed (In/In+Ga = 0.7, 0.6, 0.4, 0.3, and 0.1). Recently, Zheng et al .253 determined the Hemholtz ener gy of mixing for unstrained and strained 32-atom InxGa1-xN clusters at 0K from ab-initio calculation by using the Vienna Abinitio Simulation Package (VASP) within the fram e work of density functional theory (DFT) and the local density approximation. The Hemholtz energy was calculated using Eq. 3-19. Strain was introduced into the structures by replacing the InxGa1-xN equilibrium lattice parameter (ao) with the lattice parameter of InN (aInN) or GaN (aGaN). Substituting the lattice parameters for InN or GaN produces the maximum possible tensile (aInN) or compressive (aGaN) strain that the InxGa1-xN system can endure. It is important to mention that it is not likely that these fully strained cases could be realized experimentally, however, the calculations add insight to the affect of tensile or compressive strain on the mixing free energy. The resulting Hemholtz energy was calculated for strained and unstrained InxGa1-xN clusters as a function of InN mole fraction (Figure 3-23). When the InxGa1-xN equilibrium lattice parameter is used the Hemholtz energy is positive over the entire compositional range and th e area under the curve represents the spinodal decomposition region (where phase separation will occur, Figure 3-23a). Adding compressive and tensile strain to the system causes the Hemholtz energy to become negative over the entire

PAGE 112

112 compositional range for both types of strain (Fig ure 3-23b). These result s suggest that elastic strain plays a significant role in reducing the miscibility gap of InxGa1-xN, therefore it is suggested that the lack of subs trate induced strain from more closely lattice matched substrates (a-Al2O3) increases the probability of phase separation (at a given composition) compared to cAl2O3. A similar phase separation analysis can be applied to InxGa1-xN alloys grown on GaN (5 m)/c-Al2O3. InxGa1-xN films were grown at 530 oC and a N/III ratio of 50,000 with a LT InN buffer layer for all inlet flow compositions and ch aracterized by XRD (Figur e 3-24). Analysis of the XRD peaks is much more difficult when GaN/c-Al2O3 substrates are used due to the large peak intensity of the 5 m GaN layer. Nevertheless it can be seen that a narrow peak of phase separated InN (002) is detected for all inlet flow ratios (2 = 31.4o). This peak is assumed to be phase separated InN which is formed during InxGa1-xN growth. It is possible, however unlikely, that this peak is detected due to the presence of an InN buffer layer. InN buffer layers are deposited for 15 min at low temperature (450 oC) resulting in a 5-10 nm film (based on RBS analysis), which is polycrysta lline or amorphous (from XRD). The XRD FWHM of the phase separated InN is 288 arcsec, which is lower than the value obtained for optimized InN thin films (574 arcsec) using a LT InN buffer la yer and the same GaN substrates.203 It has shown previously in this work that a reduction in strain improves the crystalline quality of the as grown films (from the reduction in the XRD FWHM). Therefore, it is suggested that the phase separated InN regions have very li ttle residual strain due to occurrence of strain relaxation after phase separating, which leads to a higher degree of crystallinity. Phase separation is believed to be more fa vorable on the GaN substrates when a LT InN buffer layer is included because no phase separa tion occurs at similar (Ga-rich) compositions

PAGE 113

113 when no LT InN buffer layer is used.254 It is postulated that th e LT InN buffer layer promotes phase separation by providing nucle ation sites for InN, which are more energetically favorable sites compared to a compressively strained bond with GaN (when considering a Ga-rich sample). For the growth of stable (non-phase separated) InxGa1-xN thin films an ideal substrate is one that has sufficient lattice mismatch to de velop enough compressive or tensile (at a given composition) to lower the mixing free energy for InxGa1-xN. This conclusion has been verified with the growth of InxGa1-xN on c-Al2O3 substrates by producing metastable films over the entire compositional range. Regardless of the composition the lattice mismatch between c-Al2O3 and InxGa1-xN is sufficiently high to produce compressi ve strain in the film, therefore phase separation can be suppressed. In contrast, InxGa1-xN alloys were grown on the more closely lattice matched substrates of a-Al2O3 and GaN/c-Al2O3. Phase separation proved to be much more favorable on these substrates since strain coul d not be sustained in the alloy films. It has also been shown that increase in growth temp erature kinetically favors phase separation for InxGa1-xN/c-Al2O3 (from Section 3.3.2.2). The increased temperature gives atoms in the alloy film enough energy to find an energetically favorable site (i.e. a lower strained bond). 3.3.2.4 Determination of InxGa1-xN growth rate and composition from RBS measurements InxGa1-xN thin films were grown at 530 oC (N/III = 50,000) on sa pphire substrates with varying inlet flow ratio and subsequently analyz ed by RBS to determine the relationship between inlet flow ratio and solid solution composition (x). The RBS data was also used to predict the film thickness based on inlet flow conditions. Inlet flow ratio values of In/(In+Ga) = 0.2, 0.35, 0.5, 0.65, 0.75, and 0.8 were used for this anal ysis and films that included substrate nitridation and a LT InN buffer layer. RBS measurements were performed by exposing samples to a monoenergetic beam of 2 MeV 4He+ ions, which were generated by a 2.5 MeV Van de Graaff accelerator. A silicon surface barrier detector wa s positioned at a backscattered angle of 165o

PAGE 114

114 with respect to the incident beam to collect the backscattered He ions. The energy of the backscattered He ions is dictated by de pth and mass of the target material (InxGa1-xN). Therefore RBS can determine the depth profil e of individual elements in th e solid, giving the thickness and composition of InxGa1-xN samples. RBS analysis was done fo r the entire thic kness of the films tested ( 500 nm), meaning that the depth channe ling was stopped once the substrate material was detected. The thickness and composition as determined by RBS for each inlet flow ratio are shown in Table 3-3. InxGa1-xN films were grown for 2 hr and it was determined from RBS that the LT InN buffer layer has an approximate thickness of 5 to 7 nm, which was not included in the thickness measurement (Table 3-3). All growth s were performed at the same temperature therefore the extent of NH3 decomposition efficiency should be similar for each experiment. When the total group III inlet fl ow rate increases, the amount of ammonia is increased to maintain a constant N/III ratio prior to each run. Since all gr owth occurred at a constant temperature in a reaction limited regime and the group III species is the limiting reactant the growth rate should be plotted as a function of total group III flow. The experimental thicknesses as a function of total group III (TMIn + TEGa) flow rate is plotted (Figure 3-25), where the error in the group III flows is due to variation in the bu bbler pressure during the experiment. It can be seen that the growth rate of InxGa1-xN increases by increasing the gr oup III flow rate. At this deposition temperature (530 oC) the reaction is kinetically limite d, therefore a first order reaction is expected between the group III sources and the growth rate. A linear regression is fit to the data where the intercept occurs at y = 0 (Fi gure 3-25), showing the e xperimental relationship between the group III flow and InxGa1-xN growth rate, where the slope of the treadline is analogous to the reaction rate constant.

PAGE 115

115 The RBS compositional data was used to determ ine the solid solution alloy deviation from the gas phase inlet flow ratio. The difference between the inlet flow ratio and the RBS solid solution composition is plotted with a linear relationship between inlet flow ratio and composition (Figure 3-26). When the value of th e inlet flow ratio decrease below 0.7 then the film composition starts to deviate for a linear re lationship, where the so lid solution compositions contain more In than expected from the inlet flow ratio. This deviation mo st likely occurs due to GaN surface coverage which suppresses InN thermal decomposition. The solid solution composition as determined from RBS data are compared to solid solution composition determined from XRD data (Table 3-4). The XRD data from the metastable InxGa1-xN/c-Al2O3 film (Section 3.3.2.1) was used in this co mparison. The XRD and RBS compositions match fairly closely for flow ratios of 0.1, 0.2, 0.8, and 0.9, which correspond to InxGa1-xN compositions close to pure GaN and pure InN, resp ectively. The largest discrepancy is seen for flow ratios of 0.4, 0.5, and 0.6, for which RBS measurements always gave a higher value of x. The difference in this flow ratio region is believ ed to be due to the poor crystalline quality and thus any error in interpreting the spread of the XRD peaks for these flow ratios. 3.3.2.5 Terahertz emission from InxGa1-xN alloys As previously discussed in Section 1.3, In N has been shown to emit in the terahertz frequency range after optical excitation.114 There are many potential advantages for using InxGa1-xN alloys for THz emitters and detectors in cluding reduced size, operation at room temperature, better signal detec tion, and cheaper manufacturing co sts. Some work has been done on THz emission from InxGa1-xN/GaN heterstructures,173,175 however, there is no known evidence of THz emission from InxGa1-xN thin films. In this section THz emission from InxGa1-xN thin films with up to 30% gallium content is presented for the first time.

PAGE 116

116 THz emission measurements were performed on InxGa1-xN alloys given on c-Al2O3 substrates at 530 oC (N/III = 50,000) that includes substrat e nitridation and a LT InN buffer layer for preparation. The alloy composition was varied by changing the inlet flow ratio, In/(In+Ga), from 0.0 to 0.4. A time domain THz measurem ent system was used to test the THz emission from InxGa1-xN thin films (Figure 3-27). THz emission from InxGa1-xN thin films occurs by exposing the surface to titanium-sapphire (Ti:S) laser pulses with a duration of 130 fs at a wavelength of 800 nm. A 1 mm thick ZnTe <110> cr ystal is used to dete ct the THz radiation by electro-optic sampling. The THz signal was measured from the surface of InN, In0.9Ga0.1N, In0.8Ga0.2N, In0.7Ga0.2N, and In0.6Ga0.4N samples grown on c-Al2O3 (Figure 3-28). The sample with the largest signal, labeled as InN surface (Figure 3-28) is a 1 m thick InN reference sample grown by MBE for comparison. The THz signal for the InN and InxGa1-xN alloys grown by MOCVD are much smaller than the MBE reference InN, which is at tributed to the difference in film thickness. The MOCVD samples have a thickness of approximately 50 to 70 nm, which prevents a significant por tion of the pulsed laser energy from being absorbed. The highest THz emission intensity was produced by In0.8Ga0.2N (Figure 3-29). It is not completely clear as to why the 80% indium co mposition registers the hi ghest THz signal but very likely that it is due to thickness variations between the InxGa1-xN samples. The thicknesses of the In0.9Ga0.1N, In0.8Ga0.2N, In0.7Ga0.2N, and In0.6Ga0.4N films are 65, 94, 62, and 48 nm, respectively. Sample thickness was determined from the group III inlet flow values and using the relationship to growth rate descri bed in Section 3.3.2.4. Therefor e it is certainly possible that the increased thickness of In0.8Ga0.2N allowed for more absorptio n of the laser and thus a

PAGE 117

117 corresponding increase in signal intensity. Film thickness (48 nm ) might also account for the low signal to noise ratio when In0.6Ga0.4N was tested (Figure 3-30). An indirect indication of the fundamental band gap energy of InN can be gleaned from THz emission of InxGa1-xN. The titanium-sapphire used to stimulate THz emission from the surface of the semiconductor samples has a wavelength of 800 nm or 1.55 eV. The band gap energy of tested materials must be below 1.55 eV or the sample becomes opt ically transparent to the laser. For the InxGa1-xN samples described here, THz emission was detected for up to 30% Gallium incorporation, which implies that the ba nd gap energy is below 1.55 eV. If the band gap energy of InN is assumed to be a value of 0.7 eV and a linear relationship is used for band gap energy changes with composition then In0.7Ga0.3N would have a band gap energy value of 1.51 eV. THz emission from In0.7Ga0.3N is significant evidence that the band gap energy of InN is below 1.0 eV (assuming a linear relationship between EgInN and EgGaN). The increased band gap energy of In0.6Ga0.4N is another reason for why the alloy composition showed no THz emission. A band energy gap value of ~ 2.0 eV has been independently verified by photoluminescence measurements for In0.6Ga0.4N. 3.4 Conclusions MOCVD growth of InN was performed on Si (100) and Si (111) substrates and it was shown that the use of nitridation as well as a LT InN buffer layer were critical to forming highly textured InN (002). Although highly textured growth occurred on silicon substrates by MOCVD, XRD results showed low peak intensities and broad peak widths. It was determined that the formation of a silicon oxynitride laye r was the most influential parameter for the subsequent films quality. The H-MOCVD growth technique was used to grow InN on Si (111) substrates at a low N/III ratio without the forma tion of indium droplets. The presence of HCl also inhibited the formation of a silicon oxynitride layer and theref ore direct growth of InN on Si

PAGE 118

118 (111) could occur. XRD and SEM results confir m the growth of a highly textured InN (002) film with a growth rate 10X greater than MOCVD (0.75 m/hr compared to 0.078 m/hr). The group III flow was significantly increased using the H-MOCVD technique, however, no reduction in film quality occurred. Also, this type of group III flow scaling would not be possible using the previously descri bed MOCVD technique (Section 3.2.3) The stability of InN thin films with respect to the formation of crys talline oxide phases was assessed by aging samples at room temperat ure and low pressure annealing at higher temperature. No evidence was found for the fo rmation of crystalline indium oxide phases at room temperature or after annealing at temperature up to 550 oC for short periods of time (10 to 20 min). Other researchers have found that crystalline indium oxide phases were formed after only 5 min anneals at 500 oC in a nitrogen atmosphere.240 This aging study showed that crystalline oxides are extremely unlikely to form at room temperature and that crystalline oxide formation can be avoided at higher temper ature by annealing at low pressure. Growth of InxGa1-xN alloys was performed at low temperature over the entire compositional range and growth occurred on a-Al2O3, c-Al2O3, and GaN/c-Al2O3 substrates. Microscopic single phase metastable InxGa1-xN over the entire compositional range (0 x 1) was grown by MOCVD for the first time, using a low deposition temperature. Single phase InxGa1-xN can be grown, however, the crystalline quality suffers due to the low deposition temperature of 530 oC. The stability of InxGa1-xN alloys is also affected by the chosen substrate when grown at low temperature. Phase separati on is more likely to occur in InxGa1-xN when more closely lattice matched substrates are used because less substrate induced strain can be introduced into the film. Compressive or tensile strain has been shown to decrease the miscibility gap and reduce the

PAGE 119

119 critical temperature to overcome phase separation.252,253 Using a substrate with a large lattice mismatch (c-Al2O3) over the entire compositional range, bu ilds a enough compressive strain into the film to suppress phase separation between In N and GaN. When substrates such as GaN/cAl2O3 or a-Al2O3 are used, the probability of phase sepa ration increases because less substrate induced strain is produce (due to the lower lattice mismatch with InxGa1-xN). InxGa1-xN/c-Al2O3 stability was assessed with resp ect to growth at higher deposition temperature when using a constant inlet flow rati o. The inlet flow conditions were set for an approximate composition of In0.8Ga0.2N, when grown at 530 oC. Phase separation was seen in all the films grown above 530 oC and the compositional difference between the phase separated regions became more pronounced as the temperat ure increased. The composition of the In-rich phase separated region remained fairly constant with increased temperature until thermal decomposition of InN became dominant. The composition of the phase separated region with a higher Ga-content was shifted toward GaN as th e deposition temperature increased. It was concluded that avoiding pha se separation of In-rich InxGa1-xN alloys becomes increasingly difficult at higher depos ition temperatures (Tg > 530 oC). Once phase separation occurs the crystalline quality of the phase separated regions is increased. Terahertz emission frequencies were shown for InxGa1-xN/c-Al2O3 samples at compositions of 0.7 x 1. Emission signals were weak for these samples mostly due to their thickness (50 to 90 nm) and no si gnal was detected for alloys with a 40% indium composition because these samples were believed to be optical ly transparent to the stimulating laser light. These results show promise for InxGa1-xN thin films as candidates for THz emitter and detector applications, especially since current t echnologies can be drastically improved.

PAGE 120

120 Figure 3-1. Solid-gas interfacial re gion for the reaction of TMGa and NH3 to form GaN. Showing boundary layer diffusion, pryolsi s of reactant spec ies, absorption, desorption, adduct formation, and surface reactions (Ref. 228).

PAGE 121

121 Figure 3-2. The MOCVD reactor. A) A photo of the MOCVD react or. B) The process reactor schematic. Figure 3-3. Scanning electron microscope imag e of direct InN growth on Si (111) with no in-situ surface pretreatments. Resulting in incomplete film growth.

PAGE 122

122 Figure 3-4. Scanning electron microscope imag e of InN grown on Si (111) using substrate nitridation followed by a LT InN buffer laye r. Resulting in a continuous film. Figure 3-5. An XRD pattern of a polycrystalline InN thin film grown on Si (111) using a LT InN buffer layer.

PAGE 123

123 Figure 3-6. A TEM image of the SiOxN1-x intermediate layer used to produce high quality single crystal GaN (Ref. 237). Figure 3-7. X-ray diffraction pa tterns of InN grown on Si (111) and Si (100) substrates using substrate nitridation followed by a LT In N buffer layer. Intentional substrate nitridation at high temperature (850 oC) was used to form SiOxN1-x intermediate layer. Data plotted on a logarithmic scale.

PAGE 124

124 Figure 3-8. Cross-sectional SEM image of InN/Si (100) grown by MOCVD showing a growth rate of 78 nm/hr. Figure 3-9. An XRD spectrum of InN/Si (111) grown by H-MOCVD.

PAGE 125

125 Figure 3-10. Scanning electron microscope im ages of InN/Si (111) grown by H-MOCVD. Figure 3-11. Energy dispersive spectroscopy and AES analysis s howing no evidence of chlorine contamination in InN/Si (111) by H-MOCVD.

PAGE 126

126 Figure 3-12. Comparison of XRD patterns betw een aged and as-grown samples of InN on silicon and sapphire substrates. Blue lineas-grown, Red lineaged. Figure 3-13. X-ray diffrac tion patterns for InN/c-Al2O3 annealed in a N2 (100 Torr) atmosphere at 500 oC (20 min), 525 oC (15 min), and 550 oC (10 min). Blue lineas-grown, Red lineannealed.

PAGE 127

127 Figure 3-14. A series of XRD patterns for InxGa1-xN alloys grown at low temperature (530 oC) on c-Al2O3 substrates. Substrate nitridation followed by a LT InN buffer layer (N/III = 50,000) were used in each experiment.

PAGE 128

128 Figure 3-15. Experimental InxGa1-xN compositions vs. inlet flow ratio (blue line). Also plotted is a theoretical direct rela tionship between composition and flow ratio with a slope of 1 (red line) Figure 3-16. X-ray diffraction patterns of InxGa1-xN thin films grown on c-Al2O3 substrates at constant inlet flow ratio (In/(In+Ga) = 0.8) at temp erature ranging from 530 to 770 oC (blue lines). Pure InN (530 oC) and pure GaN (850 oC) standards are plotted (red).

PAGE 129

129 Figure 3-17. Magnitude of InxGa1-xN (002) peak intensity vs. deposition temperature at a constant inlet flow ratio of In/(In+Ga) = 0.8. Figure 3-18. Film crystalline quality as determined by XRD FWHM of the InxGa1-xN (002) peak

PAGE 130

130 Figure 3-19. Magnitude of In-rich InxGa1-xN (101) and Ga-rich InxGa1-xN (002) peak intensity vs. deposition temperature at a constant inlet flow ratio of In/(In+Ga) = 0.8. Figure 3-20. A series of XRD patterns for InxGa1-xN alloys grown at low temperature (530 oC) on a-Al2O3 substrates. Substrate nitridati on followed by a LT InN buffer layer (N/III = 50,000) were used in each experiment.

PAGE 131

131 Figure 3-21. Comparison of XR D spectra for InN grown on a-Al2O3 (blue) and c-Al2O3 (red) substrates. Figure 3-22. X-ray diffraction FWHM InxGa1-xN/a-Al2O3 at different values of inlet flow ratio and the corresponding film stability.

PAGE 132

132 Figure 3-23. Hemholtz free energy of mixing of InxGa1-xN as a function of InN mole fraction. A) Equilibrium lattice parameter (unstrained bonds). B) Complete tensile or compressive atom bond straining (Ref. 253). Figure 3-24. X-ray diffraction patterns for InxGa1-xN alloys grown at diffe rent inlet flow ratios on LT InN/GaN/c-Al2O3.

PAGE 133

133 Figure 3-25. Growth rate of InxGa1-xN/c-Al2O3 alloys as a function of total inlet group III flow (sccm). Films were deposited at at 530 oC (N/III = 50,000). Figure 3-26. Composition deviation of InxGa1-xN/c-Al2O3 measured by RBS as a function of inlet flow ratio.

PAGE 134

134 Figure 3-27. Time domain THz meas urement system used to analyze InxGa1-xN thin films (Diagram used with permission from Dr. Ingrid Wilke, Rensselaer Polytechnic Institute, Center for Terahertz Research, Troy, NY). Figure 3-28. Terahertz signal measured from the surface of pure InN and InxGa1-xN thin films.

PAGE 135

135 Figure 3-29. The emitted THz frequency range and corresponding amplitude for InN and InxGa1-xN alloys. Figure 3-30. Poor THz signal to noise ratio for thin (48 nm) In0.6Ga0.4N.

PAGE 136

136 Table 3-1. Flow ratios and corresponding InxGa1-xN compositions for metastable InxGa1-xN/c-Al2O3 grown at low temperature (530 oC). Composition determined by XRD measurements. Sample InxGa1-xN (0002) 2 (o) Composition (x) Pure InN31.41.00 In/(In+Ga) = 0.931.80.85 In/(In+Ga) = 0.832.10.75 In/(In+Ga) = 0.632.80.52 In/(In+Ga) = 0.533.20.41 In/(In+Ga) = 0.433.10.45 In/(In+Ga) = 0.233.60.30 In/(In+Ga) = 0.134.00.17 Pure GaN34.60.00

PAGE 137

137 Table 3-2. The InxGa1-xN (002) composition(s) for each deposition temperature ranging from 530 to 770 oC when grown at a constant inlet flow ratio of In/(In+Ga) = 0.8. Multiple InxGa1-xN compositions are listed for phase separated samples. Sample XRD peak separation (phase separation) 2 (o) Composition (x) by XRD T = 530 oC, InN (002) n/a31.41.00 T = 530 oC, InxGa1-xN (002) n/a31.870.78 peak 131.650.85 peak 232.110.70 peak 333.050.40 peak 131.610.86 peak 232.090.71 peak 131.550.88 peak 232.050.72 T = 650 oC, InxGa1-xN (002) n/a31.890.77 T = 680 oC, InxGa1-xN (002) n/a31.650.85 peak 131.650.85 peak 233.510.27 peak 131.630.86 peak 233.510.27 T = 770 oC, InxGa1-xN (002) n/a33.470.28 T = 850 oC, GaN (002) n/a34.60.00 T = 740 oC, InxGa1-xN T = 560 oC, InxGa1-xN (002) T = 590 oC, InxGa1-xN T = 620 oC, InxGa1-xN T = 710 oC, InxGa1-xN

PAGE 138

138 Table 3-3. Thickness and compositional data obtained from RBS measurements for InxGa1-xN/c-Al2O3 grown for 2 hr. Sample number Growth temperature (oC) Inlet flow ratio: In/(In+Ga) Measured composition (by RBS) Measured thickness (by RBS, in nm) Calculated thickness based on inlet flows (nm) 1328500.000.00490 o -----------1365300.200.3270101 o 1375300.350.4990102 o 1355300.490.59110 o 107 o 1385300.640.68110 o 108 o 1335300.750.709099 1345300.800.78125 o 123 o 1315301.001.0070 -----------Table 3-4. Comparison of InxGa1-xN film compositions as determined by XRD and RBS. Inlet flow ratio: In/(In+Ga) (x) in InxGa1-xN by XRD (x) in InxGa1-xN by RBS 0.90.850.91 0.80.750.78 0.60.520.65 0.50.410.59 0.40.450.52 0.20.300.32 0.10.170.16

PAGE 139

139 CHAPTER 4 GROWTH OF INDIUM NITRIDE AND INDI UM GALLIUM NITRIDE NANOWIRES BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION 4.1 Introduction Growth of one-dimensional semiconductor nanostru ctures, such as nanowires is of interest to better understand how dimensionality and size confinement affect semiconductor properties. From a device point of view nanostructures offer the benefit small feature size in two dimensions and thus a specific conduction pathway. III-nitri de nanowires have applications in low-power and high-density field-effect transistors (FETs), so lar cells, and terahertz emitters and detectors. For these device applications to be integrated with cu rrent technologies, cont rolled synthesis of III-nitride nanostructures must take place. The majority of the III-nitri de nanowire literature (reviewed in Section 1.1.3) i nvolves uncontrolled nanowire synthesi s that would be difficult to scale to production levels. Simple uncontrolled nanowire growth is beneficial for initial analysis of semiconductor properties, howev er, controlled growth on a subs trate is preferred for future device integration. In this chapter, results of controlled In N nanowire growth usi ng a traditional MOCVD reactor will be presented. The effects of select ed growth parameters such as N/In ratio, relative flow velocity, annealing conditions, in-situ surf ace pretreatments, and type of substrate are specifically investigated. Subs trate nitridation and low temperature InN buffer layers were selectively used as in-situ surface treatments on Si (100) substrates. Growth on GaN/c-Al2O3 substrates also occurred but substrate nitridat ion and LT InN buffer layer were not used. The surface treatments and chosen substr ate directly affect the nuclea tion density of the nanowires on the substrate surface and the nanowire dimensions Other growth parameters such as post growth annealing, the N/III ratio, and relative flow velocity ar e used to control the morphology and core-shell structure of the nanowires.

PAGE 140

140 4.2 Experimental Procedure Nanowires were synthesized in a Nippon Sa nso quartz MOCVD reactor that uses RF inductive heating of a tilted graphite suscepto r. The MOCVD reactor is described in greater detail in Section 3.2.3. The quartz inlet tube in the MOCVD r eactor was modified for nanowire growth. For standard III-nitrid e film growth a horizontal quartz inlet tube is used where source gases are delivered in front of the tilted suscepto r (Figure 4-1). This tilted susceptor design with laminar flow horizontal to the substrate surf ace allows enhanced mass transfer rate from boundary layer thinning along the susceptor length to compensate for reactant depletion. The boundary layer thinning is a resu lt of the inverse velocity caused by the decreasing cross sectional area. InN nanowires, however, were gr own using a vertical flow quartz inlet tube where source gases are delivered above the suscepto r, directly over the wafer tray (Figure 4-2). A similar vertical inlet tube design was used to concentrate reactants over the susceptor to improve the growth rate of InN.203 It was noticed experimentally that incomplete films and nanostructures were formed on the edges of the wafer tray when this vertical inlet tube was used. Therefore this previous vertical inlet tube was slightly modified to produce incomplete films (i.e. nanostructures) over the entire wafer tray. The corresponding dimensi ons of the respective horizontal and vertical inlet t ubes are shown in Figure 4-3. In this study InN nanowires were grown on p-Si (100) and GaN/c-Al2O3 substrates. Silicon substrates were selected because of th eir low cost and potential for future device integration. GaN substrates were used to anal yze any differences in growth with respect to silicon substrates, therefore adding insight to the nanowire growth mechanism. Silicon substrates were degreased in warm trichlor oethylene (TCE), acetone and methanol each for 5 min and then blow dry with a st atic free nitrogen gun. For GaN/c-Al2O3 substrates no chemical cleaning was performed and a nitrogen gun was used to en sure that the surface was free

PAGE 141

141 of particulates before loading. For Si, subs trate nitridation and LT InN buffer layers were selectively included in growth experiments to understand how these surface pretreatments affects the nanowire nucleation and growth As previously mentioned, no substrate nitridation or LT InN buffer layers were used for InN nanowire growth on GaN/c-Al2O3 substrates. The same LT InN buffer layer and substrate nitridation proced ures were used for InN nanowire growth that which were described in gr eater detail in Section 3.2.4 (InN film growth). 4.3 Results 4.3.1 Proposed Mechanism for MOCVD Nanowire Growth Nanowire growth commonly occurs via vaporliquid-solid (VLS) growth mechanism and this mechanism is described in greater detail in Section 1.1.3. The VLS growth mechanism was first suggested by Wagner and Ellis in the 1960s to account for the growth of mm-sized silicon whiskers in the presence of Au particles.255 Confirmation of the VLS mechanism was made by Wu and Yang using real time TEM imaging of Ge nanowires grown in the presence of Au catalyst particles.256 In this mechanism, anisotropic crystal growth is promoted by the presence of a liquid-solid interface. A liquid metal laye r resides at the growing tip of the nanowire, significantly enhancing the growth rate of the tip through a redu ction in the activation of the growth process. A schematic of a typical VLS mechanism was previously given (Figure 1-7a), where a semiconductor nanowire is grow n via gold catalyst-assisted growth. A vapor-liquid-solid (VLS) growth mechanism is proposed for the growth of InN nanowires by MOCVD. For InN nanowires grow n by MOCVD, the VLS mechanism is different than the process described above because the metal catalyst, indium is self-seeded and consumed by the formation of InN. This differs from a seeded VLS mechanism that uses a separate metal catalyst that does not incorporate into the growing rod. In traditional inert catalyzed VLS, the size of the semiconductor na nowire is determined by the diameter of the

PAGE 142

142 catalyst droplet, which is determined by the ca talyst amount and surface tension. For MOCVD InN nanowires, the size of the initial indium dr oplet formed on the substrate surface does not determine the diameter of the nanowire, but th e number density of nanowires formed. InN nanowire nucleation occurs at the solid-liquid interface and then carri es an In droplet with it. The diameter of this droplet determines the di ameter of the subsequent InN nanowire growth. This growth mechanism can be better understood schematically (Figure 4-4). The formation of indium droplets on the surface of the substrate an d the indium droplets can differ in size (as shown by an increase in droplet si ze from left to right, Figure 44a). Larger indium droplets are formed by droplet coalescence in the near inlet region where the TMIn is concentrated. This concentrated region occurs primar ily in the substrate zone direc tly under the inlet tube. This region is referred to as the deposi tion hot spot (Figure 4-5). In the hot spot region (orange colored region, Figure 4-5) large indium droplets are li kely to form, while the light blue region represents regions where smalle r indium droplets are nucleated presumably due to the In depletion by the droplet formation in the hot spot region. The surface contact area of this initial indium droplet determines the resulting density of the nanowire pa tches (the larger the area the greater number of nanowires, Figure 4-4b). Each nanowire growing from the primary droplet contains a smaller nucleat ion droplet on the tip of the na nowire which acts as a catalyst for continued growth of the nanowire (Figure 4-4c). This growth relationship was e xperimentally verified by SEM (F igure 4-6). The left image (Figure 4-6) represents a single nanowire growth, which occurs out side the hot spot deposition region (light blue region, Figure 4-5). A single na nowire is formed because the initial or primary indium droplet size was too sma ll to produce more nanowires from the same droplet. The center image of Figure 4-6 shows a nanowire patch of a few nanowires which co rresponds to the center

PAGE 143

143 schematic in Figure 4-4. This type of nuclea tion and growth would be found in the transition region from low to high concentration of TMIn (from light blue to orange, Figure 4-5). The nanowire density increases due to an increase in indium droplet size on the surface of the substrate as the position of the substrate beco mes closer to the inlet tube and higher TMIn concentration. Nanowire patches with a high de nsity of nanowire nucleation are found in this hot spot region (far right image, Figure 4-6). Previous experiments showed that InN thin f ilms can be grown without the formation of In droplets on the film surface by using inlet N/In molar ratio greater than 50,000 (at a deposition temperature of 530 oC).203 InN nanowire growth occurred at the same deposition temperature (530 oC) and N/In ratio (50,000) when using the vert ical inlet tube and In droplets formed at nanowire tips. Thus it is suggest ed from these experimental observ ations that the apparent N/In ratio at the substrate surface is less when the ve rtical inlet tube is used compared to the horizontal inlet tube design as evidenced by the form ation of In droplets. It is believed that a decrease in residence time of sour ce materials in the region of the susceptor is responsible for the discrepancy between inlet N/In ra tio and apparent N/In ratio at the substrate surface. The average velocity and Reynolds number for the vertical inlet and th e quartz reactor were calculated using Eqs. 4-1 and 4-3, respectively, and these values are listed in Table 4-1. 2 r V A V v where Vis the volumetric flow rate (4-1) RT P Mw (4-2) D v Re, (4-3)

PAGE 144

144 where D is the pipe diameter and is the dynamic viscosity. Although the horizon tal inlet has the same outlet velocity as the vertical inlet, the gas exiting the horizontal inlet expands in the larger dimensions of the quartz reactor. This expansion occurs to fully developed flow in a length of one diameter. For the ve rtical inlet, the velocity at th e outlet of the tube is assumed fully developed and the flow regime is an impingi ng jet. The average flow velocity at the outlet of each tube is two orders of ma gnitude higher than the average velo city of flow in the horizontal inlet (Table 4-1). Furthermore, the boundary la yer that develops in an impinging jet flow scheme of this design is thinner than that developed in the bounda ry layer over a tilted horizontal plate. These conditions gi ve shorter reactant residence time a nd less time at temperature thus gas phase reactions such as NH3 decomposition and adduct formation s hould occur to a lesser extent. TMIn decomposes quickly at this temperature and is therefore less likely affected by the difference in boundary layer thickness. These co mbined effects create a lower effective N/In ratio when the vertical inlet tube is used and leads to the formation of indium droplets, which nucleate nanowire growth. The formation of indium droplets is not the on ly requirement for successful growth of InN nanowires. The choice of substrate material si gnificantly affects the ability to form InN nanowires. Previous experiments using the horizontal inlet showed that the InN film growth with reduced N/In ratios in the range of 6,000 to 50,000 showed no indication of nanowire formation when grown on Si (100) or LT GaN/c-Al2O3 substrates.203 Substrate selection should affect the In metal wetting of the substrate, whic h directly affects the ability of the nanorods to form. InN nanowires were succe ssfully grown on p-Si (100) subs trates when using the vertical inlet tube, however, the same growth conditions (Tg = 530 oC, N/In = 50,000) were unsuccessful when applied to GaN/c-Al2O3 substrates. The same growth conditions yield different results

PAGE 145

145 when applied to p-Si (100) substrates (s uccessful nanowire growth) and on GaN/c-Al2O3 substrates (no nanowire growth ) (Figure 4-7). For GaN/c-Al2O3 substrates, crooked nanowires formed across the substrate surface and no nanowire growth occurred in the vertical direction, leading to the formation of an incomplete film. Indium wetting is know to be difficult on Si substrates which is one reason that In N film growth is challenging on silicon,257 however the lack of wetting is beneficial for nanowire growth beca use it creates an indium droplet which acts as a nucleation site for growth. For growth on GaN/c-Al2O3 substrates, indium forms a continuous wetting layer across the surface inst ead of droplets. It has been shown elsewhere that In-rich conditions on GaN substrates produce a m onolayer of In bonded to a GaN bilayer.258 The wetting promotes 2D growth along the surface instead of 1D growth verti cally (Figure 4-7). 4.3.2 Nanowire Morphology Dependence on Growth Parameters 4.3.2.1 Growth on p-Si (100) substrates As previously mentioned, vertical (1 D) growth of InN nanowires on GaN/c-Al2O3 substrates is more difficult than on Si (100) substrates due to th e difference in surface tension of indium on the substrate surface. The lack of wetting across the silicon surface is favorable for 1D growth. For growth on silic on substrates it is important to mention the possible formation of an In-Si eutectic phase. Silicon and indium exhi bit a nearly degenerate, simple eutectic phase diagram with a eutectic temperature of 156.6 oC (Figure 4-8). It is no ted that at the growth temperature of 530 oC the solubility of Si is limited (< 1% ). InN nanowire growth on p-Si (100) at higher deposition temperature wa s briefly explored, specifically at temperature greater than 650 oC. Deposition temperature in the range of 550 to 730 oC was used by other researchers to grow InN nanowires,96,104 since it was believed that highe r temperatures would improve NH3 decomposition efficiency. Increased NH3 decomposition would increase the N/In ratio and reduce the relative amount of indium. Therma l decomposition proved to be dominant at these

PAGE 146

146 temperatures and no InN nanowires were succes sfully grown, theref ore all InN nanowires discussed in this work were grown at 530 oC. The growth conditions such as substrate nitridation, inclusion of a LT InN buffer layer, po st growth annealing, in let N/In ratio, and group V or group III preference at the beginning of grow th were investigated to understand the effects on nanowire structure and morphol ogy. Group V or group III preference simply means that either NH3 or TMIn was allowed to flow into the r eactor for several seconds prior to the other source material at the start of growth. InN nanowires were grown on p-Si (100) substrat es using the vertical inlet tube previously described (SEM, Figure 4-9). Substrate nitridat ion followed by a LT InN buffer layer were used as surface pretreatments before the nanowire growth These are the same pretreatments used to grow continuous InN/Si thin f ilms with the horizont al inlet arrangement. For comparison, a SEM image of an InN film grown at these conditions with the horiz ontal inlet is provided (Figure 4-10). The diameter of the InN nanowires ra nged from 100 to 300 nm (see Figure 4-9a for a typical nanowire) with an average diameter of 200 nm and lengths ranged from 10 to 25 m for 1 hr growth. The In metal drop let is clearly seen at the tip of the nanowire (Figure 4-9b), consistent with a VLS growth mechanism. The use of a LT InN buffer layer allows a high density of uniform nucleation sites to form, wh ich promotes a high density nanowire growth on the substrate (Figure 4-9c and 49d). A continuous InN film is not formed by the LT InN buffer layer therefore the wetting of In on silicon (or lack there of) does not significantly change allowing for 1D growth to remain favorable. Post growth annealing in a low pressure (100 Torr) NH3 atmosphere was employed to determine the effect on the indium droplet at the tip of the nanowire. Annealing was done in-situ immediately after growth at 530 oC without exposing the nanowires to air. The ammonia flow

PAGE 147

147 was kept at a constant fl ow of 1600 sccm (4 SLM N2 dilution) for a duration up to 15 min. Ammonia reacts with the residual indium in th e droplet to form a sharpened InN nanowire (SEM, Figure 4-11). Morphological differences are clearly seen be tween InN grown with the horizontal versus the vertical inlet (at the same growth conditions), therefore InN nanowire growth conditions were varied to understand the effect of using substrate nitridation or a LT InN buffer layer prior to InN growth. The growth conditions and the morphol ogical changes produced by variation in buffer layer and nitridation pretreatments, as well as group III or V source preference are summarized in Table 4-2. For simplicity the experimental run numbers (68 through 71) will be used to discuss the growth results as th ey pertain to substrate pretreat ments used in this study. SEM images (Figures 4-12 through 4-15) are also re ferenced in Table 4-2 for the corresponding growth conditions. It is noted th at the magnification of the images varies in these figures and the images were taken from a similar position in th e hot spot of the fl ow pattern. Run 68 reproduced the InN nanowire results previously mentioned (Figure 49) that used both substrate nitridation followed by a LT InN buffer layer, except this run included post growth NH3 annealing for 10 min. Similar results are obtaine d that show dense InN nanowires with fairly uniform diameters ranging from 100 to 300 nm (F igure 4-12). When the LT InN buffer is not used (run 69) the nanowire patch nucleation densit y is significantly reduc ed (Figure 4-13). The resulting nanowires also show a larger range of nanowire diam eters (100 to 700 nm) within the same patch, however, the overall size of the nano wire patch remains nearly constant. Little variation in the nanowire patch size suggests that the nucleation processes are similar. The growth of nanowires in run 70 is done without the aid of substrate nitridation or a LT InN buffer layer. In this case the NH3 was allowed to flow into the reactor for 10 sec prior to the initiation

PAGE 148

148 of the TMIn flow and thus growth. This step is used to ensure th at indium does not interact with the substrate without the presence of NH3 and no significant nitridation of the substrate is believed to occur in 10 sec at the lower depos ition temperature (based on the comparison between substrates which were intentionally nitridated at 850 oC for 15 min). The growth results from run 70 show a uniform nucleation distributi on across the substrate surface, yet the nanowire patches differ in size and hence the number of nanowires (Figure 4-14). The nanowire diameter within a single patch range form 100 nm to 1 m. No buffer layer or nitridation is used for run 71 and in contrast to run 70, TMIn was al lowed to flow into the reactor without NH3 present for 30 sec (Figure 4-15). In this case the nucleati on sites are less ordered a nd the excess indium at the beginning of the reac tion leads to significant branching w ithin the nanowire patches. The branching leads to a larger num ber of smaller nanowires with respect to length and diameter. The density of nucleation sites and nanowire si ze (diameter and length) can be modified by using particular silicon substrat e pretreatments before growth. The use of a LT InN buffer layer produces densely nucleated nanowires that grow toge ther and resulting in a field of nanowires. Substrate nitridation only leads to sparse nucleation sites a nd a larger range of nanowire diameters. When no surface trea tment is performed prior to gr owth the nucleation size is less controlled therefore there is a large variation in nanowire patch sizes. Branching occurs when indium droplets are deposited on the surface in the absence of ammonia and the nucleation patches vary in size and number of nanowires. The nucleation of liquid In on the surface is clear ly important to the density and size of the patches, which are related to the density and si ze (length and diameter) of the nanowires. The base growth conditions selected for this study lie on the boundary of forming InN directly and forming In droplets. Thus NH3 serves to shift this boundary at this conditions (these experiments

PAGE 149

149 occur slightly within the 2-phase region). Nucl eation of In droplets can occur initially only or continuously. If it is only initia lly and the surface prope rties are uniform, then a single size type distribution would be expected. If it is continuous, th en the size would be greater for the early nucleation and smaller for the later ones. At higher nucleation dens ity it is possible for secondary nucleation occur (droplet coalescence) to give size variation leading to a nanowires with a high extent of liquid fractions. Thus r un 68 may be extremely high nucleation density or complete wetting (favorable InN coverage), while run 69 appears to have uniform nuclei (suggestion nucleation initially then on ce liquid In forms it catalyzes the NH3 decomposition. Run 70 gives poorer surface conditions for nucleati on (due to the low temperature nitridation) and run 71 gives higher nuclei density c onsistent with favorable In nucleation 4.3.2.2 Growth on GaN/c-Al2O3 substrates The growth of InN nanowires also occurred on GaN/c-Al2O3 substrates, however the growth conditions had to be modified compared to those of the previous s ection. As previously shown the increased In wetting on the GaN su rface causes InN nanowires to grow along the surface of the substrate instead of vertically (Figure 4-7). This result occurs when NH3 is allowed to flow into the reactor slightly before TMIn. A sim ilar result showing nanowire growth along the surface of the substrate (l ittle to no vertical nanowire gr owth) is seen when TMIn is introduced into the reactor prior to NH3 (Figure. 4-16). In this case, however, some small nanowires were found on the edges of the incomplete film regions on the GaN substrate, suggesting that the introduction of TMIn fi rst is beneficial for forming nanowires. To induce nanowire formation in the vertical direction the GaN substrate was preheated to 850 oC for 1 min prior to growth (4 SLM N2 atmosphere, 100 Torr) to activate the surface of GaN. This process creates nitrogen vacancies in the surface of the film which change the metal wetting properties on the surface. Once the reac tor cooled to the deposition temperature (530 oC)

PAGE 150

150 TMIn was introduced into the reactor in the absence of NH3 for 20 sec. Activation of the GaN surface proved critical to the formation of nanowires (Figure 4-17). Small pits in the GaN surface are clearly noticed (Figure 4-17), which we re not seen in other attempts to grown InN nanowires on GaN (without the pr eheating step). Previously m odeling work has shown that In can form a single monolayer in the top layer of GaN and even pene trate into the second monolayer.259 Chen et al. verified the theoretical results from STM measurements and also determined that strain due to the occupation of in dium in the first two monolayers caused pits to form in the GaN surface.260 It is believed that the pit formation is the main mechanism that allows for the formation of InN nanowires on Ga N substrates because the In metal wetting on the substrate is reduced, making droplet forma tion more likely. InN nanowires grown on GaN/c-Al2O3 substrates shows similar nanowire diamet er and length as nanowires grown on Si substrates, 200-400 nm and 10-20 m, respectively. The inlet flow velocity was studied to determ ine the effects of the effective N/In ratio on the resulting morphology of the nanowires. Fo r this study the nitroge n carrier gas and the ammonia flow rates were reduced by half and the TMIn flow was adjusted to maintain a constant inlet N/In ratio of 50,000. The ammonia flow was reduced from 1600 to 800 sccm and the N2 carrier gas was reduced from 4000 to 2000 sccm. Th e reduced flow velocity out of the vertical inlet tube slightly increases the residence time of NH3, which increases decomposition efficiency, therefore increasing the effective N/In ra tio. The increase in the effective N/In ratio reduces the diameter of the nanowires (averagi ng 150 nm) and increases their length up to 40 m for 1 hr growth (Figure 4-18). Interestingly th e nanowires also exhibit very sharp tips without having to employ post growth annealing. An in crease in growth rate and reduction in nanowire diameter is expected with an increase in the e ffective N/In ratio. Most of the nanowire growth

PAGE 151

151 occurs in an In-rich regime, th erefore increasing the amount of active nitrogen decreases the size of the catalyst droplet at the e nd of the nanowire, thus making th e nanowire diameter smaller. Since active nitrogen is the limiting reactant then the growth rate will increase if the relative N/In ratio is increased. To produce more stoichiometric growth conditions, close to single phase InN (i.e. less evidence of excess indium after growth), the inle t N/In ratio was set to 73,000, while the inlet carrier gas flow and NH3 flow were also cut by half. With these conditions the overall increase in the N/In ratio was too high and no nanowir es were produced (only film), and it was determined that there needed to be some In-rich step to initiate nanowire growth. A two step growth method was used in which nanowires were nucleated at a N/In ra tio of 50,000 (at reduced flow velocity) for 5 min and then the inlet N/In was increased to 73,000 for the duration of the 1 hr growth. For this two step growth no indium droplet is apparent at the tip of the nanowire (Figure 4-19). The roughness of the nanowire walls is also an indication that growth did not occur via a VLS mechanism. Once the indium dr oplet is consumed from the initial nucleation step the reaction is believed to progress via solid-vapor mechanism. The nanowire diameters average 150 nm and the lengths are significantly shorter (< 10 m) than the nanowires grown by VLS for the same growth duration, due to the high er N/In ratio and different growth mechanism. 4.3.3 InN Nanowire Composition and Structure The InN nanowire composition and structure were investigated by scanning electron microscopy (SEM), energy dispersive spectro scopy (EDS), x-ray diffraction (XRD), and transmission electron microscopy (TEM). Char acterization by these t echniques indicates the nanowires exist as a core-shell structure whic h contains a single crystal InN core with a polycrystalline In2O3 shell (regardless of the substrate) (Figure 4-21). XRD results confirm the InN coreIn2O3 shell structure (Figure 4-21). A singl e crystal InN (002) reflection exist along

PAGE 152

152 with In2O3 (222) and In2O3 (400) polycrystalline reflections, at 2 = 31.4, 30.6, and 35.45o, respectively. Further confirma tion of the composition of the co re-shell structure was obtained from EDS (Figure 4-22). The electron beam wa s focused on a single core-shell nanowire where oxygen, nitrogen, and indium are detected as well as a silicon peak from the underlying substrate. The crystalline quality of the core -shell was analyzed using TEM and single crystal InN core and the transition to the polycrystalline In2O3 shell can be clearly distinguished (Figure 4-23). Selected area diffraction pa tterns (SADP) were taken of the InN core (Figure 4-24). The single diffraction spots reinforces that the InN core is single crystal and th e pattern indicates that [00.2] is the preferred growth direction. Evidence of the core-shell structure require s that the proposed VLS growth mechanism must account for the inclusion of oxygen. Two li kely sources of oxygen are native oxides from the substrate and residual oxygen in the reaction chamber, or a combination of both sources. Oxygen from the native oxides on the substrate is po ssible especially when silicon substrates are used. This hypothesis was tested by comparing InNIn2O3 core-shell nanowires grown on native oxide Si, buffered oxide etched Si and GaN/c-Al2O3 substrates at the same growth conditions. XRD, SEM, and EDS analysis revealed that th ere was no significant difference in the oxygen content of the nanowires when grown on thes e different substrates. However, these characterization techniques have an oxygen sens itivity limits ~ 1%. Ne vertheless this result suggests that residual oxygen from the reaction ch amber (or an air leak) is also a source for oxygen contamination. This is possible since the nanowire growth experime nts are carried out at low pressure, 100 Torr. The Gibbs energy of reaction (Gf) at nanowire growth conditions for the formation of InN and In2O3 are given in Eqs. 4-4 and 4-5, respectively.261

PAGE 153

153 In + NH3 InN + 3/2H2 Gf = -86 KJ/mol (4-4) 2In + 3/2O2 In2O3 Gf = -669.5 KJ/mol (4-5) It is clear that the formation of indium oxide is more favorable due to the larger negative Gibbs free energy value. However there is a significa ntly larger amount of ammonia in the reaction chamber than residual oxygen and NH3 decomposition forms H2 or H2O which will help reduce oxygen. Therefore it is li kely that the oxide layer forms upon cooling when NH3 is kinetically limited for reducing the oxide. During coo ling the ammonia decomposition reduces to essentially zero242 and the reaction between indium and oxygen is still thermodynamically favorable, especially when there is excess indi um metal on the outside of the nanowire. Excess indium can be eliminated from the na nowire tip and walls of the nanowire by using a two step nucleation and growth method that adju sts the inlet N/In ratio, as described in the previous section. The elimination of the ex cess indium makes it less favorable for oxygen to form on the outside of the nanowire, and no oxygen signal is detected by EDS for nanowires grown at these conditions (Figure 4-25). In this case, InN nanowires were grown on a GaN/c-Al2O3 substrate to minimize oxygen incorporation from the substrate (compared to Si). From this analysis it is believed that the indium oxide layer forms on the outside of the nanowire during cooling and that the primary source of oxygen is residual oxygen in the reaction chamber (e.g. carrier gas, small leak, desorption). It is important to mention that oxygen from the substrate could also play a role, however it is believed to be less significant. 4.3.4 Transport Properties of InNIn2O3 Core-Shell Nanowires Single InNIn2O3 core-shell nanowires devices were fa bricated and the contact resistivity and sheet resistivity of the nanowires were meas ured using a transmission line model (TLM). The TLM method is often used to describe current transport in lateral contacts of semiconductors.262,263 InNIn2O3 core-shell nanowires were grown on nitridated Si (100)

PAGE 154

154 substrates at 530 oC for 1 hr and then annealed in NH3 for 15 min in-situ immediately after growth. The core-shell nanowires were remove d from the silicon substrate by sonication and then transferred to an insulating substrate. A shadow mask was used to evaporate Ti/Al/Pt/Au (20/80/40/80 nm) ohmic contacts. Metal was deposited in the openings of the shadow mask (50 x 50 m2 or 115 x 115 m2) leaving 10 m gaps between metal contact s to provide electrical contacts to the nanowires (Figure 4-26). To determine contact and sheet resistivity, the total resistance of several InN-In2O3 coreshell nanowires was measured from I-V measur ements (Figure 4-27 and Eq. 4-6). The total measured resistance (RT) for a circular wire is given by Eq. 4-7.264 TIR V (4-6) 22r L R Rs c T (4-7) Rc is the total contact resistance, s is the sheet resistivity of the InN In2O3 wire, L is the length of the wire, and r is the radius. The total meas ured resistance from I-V data for 4 samples was plotted versus L/r2 (Figure 4-28). The nanowire dimensi ons were determined by SEM (Figure 4-29). A linear relationship was f it to the experimental data from Figure 4-28. It can be seen from Eq. 4-7 that the slope of the treadline in Figure 4-28 is equal to s/ while the y-intercept is equal to 2Rc. Therefore the sheet and contact ( c) resistivity are calculated from Eqs. 4-8 and 4-9.265 slopes (4-8) s c cR r 2 2 32 (4-9) From the experimental data presented, the shee t resistivity and contact resistivity for a single InNIn2O3 nanowire is 1.5 x 10-3 -cm and 1.16 x 10-5 -cm2. The value for contact resistivity

PAGE 155

155 is only an approximation because small changes in the slope of the linear fit of the TLM method can significantly change the magnitude of c. The sheet resistivity of the InNIn2O3 core shell nanowires presented here were compared to InN nanowires grown by ammonization of indium metal on gold patterned silicon substrates at 500 oC in a tube furnace.92 InN nanowires grown by this me thod produced a sheet and contact resistivity of 4 x 10-4 -cm and 1.09 x 10-7 -cm2, respectively.85 The sheet resistivity of the InNIn2O3 core-shell nanowires is one order of ma gnitude higher than the referenced InN nanowires. The oxide shell is believed to be re sponsible for the increase in resistivity because it reduces the available cross sectiona l area for current transport. It is believed that the oxide shell does not have a more significant impact on the sh eet resistivity because the InN core is very conductive. Increased conductivity in the InN core could be due to an In-rich InN layer between the nitride and the oxide or poten tial silicon doping of the InN. As previously mentioned silicon and indium form a eutectic phase at 156.6 oC. Even though silicon is an amphoteric dopant in many III-V compound semiconductors, Si has been shown to be an n-type dopant in InN.211 TLM data also suggests that c ontact resistivity of the InNIn2O3 core-shell nanowires is two orders of magnitude higher than the referenced InN nanowires. It was previously mentioned that experimental errors could signifi cantly affect these value so veri fying the two order of magnitude difference in contact resistivity is not possible. It is safe to assu me, however, that the addition of an oxide shell on the nanowire would increase the contact resistivity. 4.3.5 Single InNIn2O3 Core-Shell Nanowire for H2 Gas Sensing Using nanowires for gas sensing devices has pr oved to be advantageous due to their large surface to volume ratio, low power consumption, selectivity at room temperature, and low weight.266 Development of selective nanowire H2 gas sensors is important for use with proton-exchange membrane (PEM) and solid oxide fuel cells (SOF Cs), which are proposed for

PAGE 156

156 use in space and future transportation applications.267 The mechanism for gas detection involves the removal or addition of atoms on the surface of the nanowire, which changes the conductivity of the nanowire. Oxygen mol ecules will absorb onto the In2O3 shell surface of the InNIn2O3 core-shell nanowire after air exposure. Once H2 is introduced into the system a reduction of the absorbed oxygen surface states will occur, leadin g to a drop in the nanowire resistance due to a creation of surface donor s ites (oxygen vacancies). To test the capability of InNIn2O3 core shell nanowires for selective H2 gas sensing, single wires devices were fabricat ed by the same method as descri bed in Section 4.3.4. Single InN In2O3 core shell nanowires were wire bonded before testing (Figure 4-31). The I-V characteristics for time depende nt current response under N2 and 500 ppm H2 (in N2) ambients of the single wire device were determined (Figure 4-32). The I-V characteristics are identical for both ambients, which indicates that there is no selective response to H2 (Figure 4-32a). This result is confirmed with time dependent curr ent measurements (Figure 4-32b) that show no signal response when the nanowire is exposed to hydrogen. The H2 sensitivity of a gas sensor can be increased by adding a catalytic metal co ating or by doping the sensor material with a transition metal.268,269 A catalyst such as Pt will func tionalize the surface of the nanowire and promote the dissociation of H2. For this reason individual core-s hell nanowires were sputter coat ed with Pt clusters in an attempt to improve the H2 signal response from the nanowires Core-shell nanowires were coated with 3 and 10 of Pt and the corres ponding I-V characteristics and time dependent response to 500 ppm of H2 were measured (Figures 4-33 and 434, respectively). It can be seen from the I-V characteristics (Figure 4-33) that th e bare and 3 Pt coated nanowire show similar conductivity, but the 10 Pt coat ed nanowire shows an order of magnitude increase in the

PAGE 157

157 conductivity. The increase in c onductivity after the 10 coating is most likely due to the formation of a continuous Pt film instead of Pt cl usters. This possibility is reinforced when the 10 Pt coated core-shell nanowire registers a significantly higher current response to the introduction of 500 ppm H2 (in N2)compared to the 3 Pt coa ting (Figure 4-34b). The 3 coating does increase the signal response to H2 compared to the bare core-shell nanowire (Figure 4-34a), however, the signal only changes 1.3 A (0.5% change). The signal response for the bare or 3 coated nanowire is too low for these InNIn2O3 core-shell nanowires to be effective for detecting H2. One possible reason that the InNIn2O3 core shell nanowires were not able to selectively detect hydrogen is because the wires were intrinsically too conductive. This causes the depletion effect to be too sm all thus causing a low sensitivity. 4.4 Conclusions Controlled growth of InN nanowires was su ccessfully demonstrated using a standard MOCVD reactor by changing the flow pattern with a modified inlet tube. The modified vertical flow regime reduced the residence time of NH3 over the hot susceptor, thus reducing decomposition and decreasing the effective N/In ratio A lower N/In ratio leads to excess indium and the stability of In droplets. Poor wetting of the substrate surface by liquid indium proved to be critical to the formation of nanowires. The excess indium acts as a catalyst site and remains at the tip of the nanowire to promote subsequent grow th. It is important to mention that lowering the N/In ratio while using the horizontal inle t tube does not lead to nanowire formation. Nanowires were grown on Si (100) substrates and the mo rphology of the nanowires was sensitive to the growth conditions. The use of nitridation followed by growth of a LT InN buffer layer produces dense fields of uniform diameter na nowires that completely cover the substrate. Growth without a LT InN buffer leads to less dense growth, which forms patches of InN nanowires. Nanowire length, diameter, and de gree of branching cha nged with the use of

PAGE 158

158 substrate nitridation and/or initial exposure to only TMIn or NH3 source materials for a determined length of time prior to the start of growth. Post-growth NH3 annealing was shown to sharpen the tip of the nanowire and eliminate the amount of indium metal at the tip of the nanowire. The dimensions of the nanowires de pend on the growth conditions applied, but the InN nanowires have average diameters in the ra nge of 100 to 400 nm with lengths ranging from 10 to 25 m. It was also found that a reduction in flow velocity incr eased the relative N/In ratio during growth, which leads to reduced nanowire di ameter (~ 150 nm) and increased length, up to 40 m for a 1 hr growth. Characterization of the nanowir es revealed that an InNIn2O3 core-shell structure was formed during growth. The InN core was found to be single crystal with th e preferred growth in the [002] direction while the In2O3 shell is polycrystalline. Pote ntial sources of oxygen are from the substrate surface (especially in the case of Si) as well as residual oxygen in the reaction chamber (e.g. inlet gas source, air leak). It is believed that the oxide shell is formed during cooling when NH3 can no longer reduce residual oxygen specie s at lower temperatures (due to decreased thermal decomposition efficiency). GaN/c-Al2O3 substrates required that the growth conditions used for Si substrates be modified to reduce the amount of indium wettin g on the GaN surface. It was also found that using a two-step growth method of nucleation at a lower N/In ratio followed by growth at a higher N/In ratio eliminated the indium droplet form the tip of the nanowire. The growth method is believed to change from a VLS to a VS gr owth mechanism when the excess indium from the nucleation step is consumed, producing a rough surface morphology with no indium droplet tip. These nanowires also were much shorter (5 m), which is additional evidence for a solid-vapor growth mechanism. The lack of excess indium on the surface of the nanow ires also eliminated

PAGE 159

159 the oxide shell from forming on the nanowire. The surface morphology of the nanowires grown by the VS mechanism shows a similar surface mo rphology to InN thin films grown by MOCVD (using the horizontal inlet). This is further c onfirmation that a VS mech anism occurs at higher N/In ratios. Single nanowire transport measurements were done using the tran smission line method (TLM), which showed that the core-shell nanowir es were highly conductive and that the oxide shell increased the contact resistance. The nanow ires were believed to be highly conductive due to n-type silicon doping or a possible In-metal ri ch layer between the oxide shell and the nitride core. This high conductivity also proved to be detrimental for using these nanowires as single nanowire gas sensing devices. It is believed a reduce field effect led to a small change upon exposure to dilute H2 (in N2) gas, which was not significant enough for the required device applications. Nanowires without th e presence of the oxide shell were unable to be tested due to the difficulty in testing nanowires with lengths below 10 m. Figure 4-1. Schematic of the standard horizontal inlet tube and its position with respect to the graphite susceptor. Stream lines ar e shown for the expected flow pattern.

PAGE 160

160 Figure 4-2. Schematic of the vert ical quartz inlet tube and its rela tive position with respect to the graphite susceptor. Stream lines are show n to demonstrate the e xpected flow pattern. Figure 4-3. Inlet tube dimentions for film and nanostructured growth. A) horizontal (film) inlet tube. B) vertical (nanostructure) inlet tube.

PAGE 161

161 Figure 4-4. Schematic of the VLS mech anism for InN nanowire growth by MOCVD. A) Primary nucleation of indium droplets wi th varying sizes on the substrate surface. B) Initial nanowire formation plus secondary formation of nanowire In metal catalyst. C) Nanowire density as a function of primary indium droplet size. Figure 4-5. Overhead view of vertical inlet tube showing the hot spot deposition zone (orange) and the region where small In metal primary nuc leation is likely to form (light blue).

PAGE 162

162 Figure 4-6. Experimental verifi cation of the proposed nanowire de nsity variation due to initial indium droplet nucleation size. Upon clos e inspection the metal indium droplet can be seen at the tip of each nanowire. Figure 4-7. Effect of In metal wetting for InN nanowire growth on Si (100) and GaN substrates. Successful nanowire growth on silicon substr ates (poor In metal wetting leading to droplet formation) and unsu ccessful growth on GaN/c-Al2O3 substrates (complete In metal wetting, no droplet formation).

PAGE 163

163 Figure 4-8. Indium-Silicon binary phase diagra m (obtained from the Center for Research in Computational Thermochemistry website using the compound and solution Fact database, http://www.crct.polymtl.ca/fact/documentation/ retrieved March 2007, site authored by Christopher W. Bale). Figure 4-9. Scanning electron microscope im ages of InN nanowires grown on p-Si (100) substrates using nitridation followed by a LT InN buffer pretreatment. A) Single InN nanowire with 250 nm diameter. B) Single InN nanowire showing indium droplet at tip. C and D) High density of nanowires grow ing together and cove ring the substrate.

PAGE 164

164 Figure 4-10. Scanning electron microscope image of a continuous polycryst alline InN thin film grown with the horizontal in let arrangement on p-Si (100) Substrate nitridation followed by a LT InN buffer layer pretreatment was used for this experiment. Figure 4-11. Effect of post growth annealing on InN nanowire tip morphology. A) No anneal showing indium droplet. B) Sharpe ned InN nanowire tip after 15 min NH3 anneal (T = 530 oC, 1600 sccm, 100 Torr).

PAGE 165

165 Figure 4-12. Scanning electron microscope imag es of InN nanowires (experiment number 68) showing uniform dense coverage of the subs trate. Growth cond itions are summarized in Table 4-2. Figure 4-13. Scanning electron microscope imag es of InN nanowires (experiment number 69) showing less dense substrate coverage with uniform patches of nanowires. Growth conditions are summarized in Table 4-2.

PAGE 166

166 Figure 4-14. Scanning electron images of In N nanowires (experiment number 70) showing uniform coverage of the substrate with nanowire patches that vary in size and nanowire density per patch. Individual nanowires vary in diameter and length. Growth conditions are summ arized in Table 4.2.

PAGE 167

167 Figure 4-15. Scanning electron microscope imag es of InN nanowires (experiment number 71) showing less ordered nanowire growth with significant nanowire branching. Growth conditions are summarized in Table 4.2. Figure 4-16. Scanning electron microscope im age of unsuccessful InN nanowire growth on GaN/c-Al2O3 substrate. TMIn was allowed to flow into the reactor in the absence of NH3 for 20 sec.

PAGE 168

168 Figure 4-17. Successful growth of InN nanowires on GaN/c-Al2O3 substrates by activating the surface of GaN. Surface activation occurred at 850 oC (in N2 atmosphere) for 1 min before growth at 530 oC. Post growth NH3 annealing was done for 15 min at 530 oC for this experiment. Figure 4-18. Scanning electron microscope image of InN nanowires grown on GaN/c-Al2O3 substrate at reduced inlet flow velocities. No post-growth NH3 anneal was used.

PAGE 169

169 Figure 4-19. Scanning electron microscope images of InN nanowires grown on GaN/c-Al2O3 using two-step nucleation and growth appro ach. Nucleation and growth used a N/In of 50,000 and 73,000, respectively. N/In rati os are for reduced flow rate conditions, NH3 = 800 sccm, and N2 carrier = 2000 sccm.

PAGE 170

170 Figure 4-20. Scanning electron microsc ope images depicting the InN core-In2O3 shell structure seen from broken nanowires.

PAGE 171

171 Figure 4-21. X-ray diffraction spectra of InN core-shell na nowires on Si (100) substrate indicating single crystal In N and polycrystalline In2O3 phases. Figure 4-22. Energy dispersive spectroscopy spot analysis on the outside of a core-shell nanowires where indium, nitrogen, and oxygen are detected as well as silicon from the underlying substrate.

PAGE 172

172 Figure 4-23. Transmission electron mi croscope images of a single InN-In2O3 core-shell nanowire. The core and she ll can be clearly identified and are labeled appropriately.

PAGE 173

173 Figure 4-24. Selected area diffraction pa tterns of the InN core of an InN-In2O3 core-shell nanowire. The patterns indicate the core is single crystal and th at growth occurs on the [00.2] direction. Figure 4-25. Energy dispersive spectroscopy showing no oxygen de tected for nanowires grown at high inlet N/In ratios and reduced flow velocities. No In droplet at the nanowire tip was evident (Figure 4-18).

PAGE 174

174 Figure 4-26. Scanning electron microscope image of single InNIn2O3 core-shell nanowire contacted by Ti/Al/Pt/Au pads for electrical measurements. Figure 4-27. Current-voltage measurements for several InNIn2O3 core-shell nanowires, used for determining the total resistance (RT) of each wire.

PAGE 175

175 Figure 4-28. The total measured resistance versus l/r2 for InNIn2O3 nanowires. Figure 4-29. Scanning electron microscope images and dimensions of single InNIn2O3 core-shell nanowires used in the cont act and sheet resistivity calculations.

PAGE 176

176 Figure 4-30. Optical micrograph of a single InNIn2O3 core-shell nanowire after wire bonding. Figure 4-31. Current-voltage characteristics and time dependent response of InNIn2O3 single wire device under N2 and 500 ppm H2 (in N2) ambients. A) Current-Voltage characteristics. B) Ti me dependent response.

PAGE 177

177 Figure 4-32. Current-voltage ch aracteristics of single InNIn2O3 core-shell nanowires before and after coating with 3 or 10 of Pt. Figure 4-33. Time dependent cu rrent response of a single InNIn2O3 core-shell nanowires in a 500 ppm H2 ambient with different thicknesses of Pt. A) Bare and 3 Pt coated core-shell nanowire. B) 10 Pt coated core-shell nanowire.

PAGE 178

178 Table 4-1. Flow velocities and Reynolds number at the outlet of th e vertical inlet and inside the quartz reactor tube (Ref. 203). Vertical inletQuartz reactor Carrier gas, N24000 sccm4000 sccm Ammonia flow1600 sccm1600 sccm Total flow 5600 sccm5600 sccm Pressure100 Torr100 Torr Pipe diameter0.386 cm8.5 cm Kinematic viscosity, n1.116 x 10-6 m 2/s1.116 x 10-6 m 2/s Average velocity, 7.98 m/s0.0164 m/s Reynolds number27612.5

PAGE 179

179 Table 4-2. The InN nanowire growth conditions for samples with varying substrate/surface pretreatments and the resulting morphology information. Experiment number 68697071 InN buffer layer (duration / temperature) 15 min / 450 oC nonenonenone Nitridation (duration / temperature) 15 min / 850 oC15 min / 850 oC nonenone Source valve opened first (duration) NH3 (5 sec)NH3 (5 sec)NH3 (10 sec)TMIn (30 sec) Growth temperature 530 oC530 oC530 oC530 oC Inlet V/III ratio50,00050,00050,00050,000 Post growth annealing (Duration / temperature 10 min / 530 oC10 min / 530 oC10 min / 530 oC10 min / 530 oC Reference figureFigure 4-10Figure 4-11Figure 4-12Figure 4-13 Separation between nanowire patches ~ 0 20 100 m0 10 m0 20 m Nanowire patch size variation NoNoYesYes Nanowire diameter range 100 300 nm100 700 nm100 1000 nm60 400 nm Nanowire length range* 10 25 m10 25 m8 20 m1 20 m *based on 1 hr growth

PAGE 180

180 CHAPTER 5 GROWTH AND CHARACTERIZATION OF INDI UM NITRIDE NANOAND MICRORODS BY HYDRIDE-METAL ORGANIC CHEMICAL VAPOR DEPOSITION 5.1 Introduction As discussed in Chapter 4, the growth of InN nanowires can be achieved by MOCVD, however the inlet N/In ratio must be precisely controlled during growth because indium droplet formation is sensitive to this growth parameter. The presence of indium metal also leads to the formation of indium oxide, which may not be de sired for particular a pplications such as III-nitride heterostructure nanorods. Theref ore there is a motivation to produce InN nanostructures without excess indium that can react to form oxides. In this chapter the growth of InN nanoand microrods were grown by H-MOCVD on GaN/c-Al2O3, c-Al2O3, and silicon substrates. The effect of growth parameters, such as Cl/In ratio and temperature were specifically investigated and the resulting nanostructures were characterized to investigate the nanor od structural and optical properties. 5.2 Experimental Procedure 5.2.1 Substrate Preparation Silicon substrates were cleaned in warm TC E, acetone, and methanol each for 5 min and blown dry with a nitrogen gun before loading. Sapphire substrates were only cleaned in warm methanol for 5 min and blown dry, while no chemical cleaning was done for GaN/c-Al2O3 substrates. A nitrogen gun was used to remove any particulates from the GaN substrate surface before loading. 5.2.2 The H-MOCVD Reactor and Deposition Technique The H-MOCVD growth technique is simila r to conventional MOCVD because metal organic sources are used, yet the H-MOCVD reactor has the ability include HCl in the source gas stream. The addition of HCl forms metal chlori des by reacting with the metal organic species

PAGE 181

181 and the chlorides subsequently r eact with the group V species (NH3). In traditional MOCVD, metal organic molecules react directly with amm onia. For this reason the technique has been dubbed hydride or H- MOCVD. When HCl is not used the reactor can be operated as a standard MOCVD reactor. This H-MOCVD technique is beneficial becau se the MOCVD design of fers the ability to do precise source switching, which is not possibl e with standard HVPE growth techniques that use liquid metal sources, due to metal chloride out gassing. Similar to HVPE growth, the H-MOCVD technique has been shown to produc es high growth rates of high quality GaN (20 m/hr).270 InN nanorods have also been demonstr ated by this growth technique and the nanostructures have been shown to be bene ficial for selective gas detection devices.271 H-MOCVD is also beneficial becau se the polarity of the grown films can be controlled. When HCl is used N-faced films are grown (producin g a rough surface) and Ga-faced films are grown (resulting in a smooth surface) when conve ntional MOCVD mode is used (on c-Al2O3 substrates). It was also mentioned in Chapter 3 that the use of HCl allows InN films to be grown at low N/In ratios without the forma tion of indium droplets on the surface. The H-MOCVD operates at atmospheric pressure and is a hot walled reactor that uses six individually controlled heating el ements to control the temperatur e of the quartz reactor tube. Quartz starts to become less rigid when the temperature is above the 1000 oC, therefore the maximum reactor temperature is limited to 1000 oC. A load lock is used to minimize oxygen contamination from reaching the reactor and the s ubstrates are loaded into the growth zone on a tilted quartz susceptor with a magnetic load arm. Waste gases pass through Novapure S407 scrubber media for removal of unreacted ammonia and then vented to the atmosphere. TMIn, TMGa, NH3, and 10% HCl (in nitrogen) are used as source materials and N2, H2 or 4% H2 (in

PAGE 182

182 nitrogen) can be used as a carrier gas. The H-MO CVD reactor uses a triply concentric inlet tube to deliver the source gases (Figure 5-1). The overall reaction and decomposition mechanisms are shown below in Eq. 5-1 through 5-3 and 5-4 though 5-6, respectively. In(CH3)3(g) + HCl (g) InCl(g) + 3CH4 (g) + C2H6(g) (5-1) InCl(g) + NH3(g) InN(s) + HCl(g) + H2(g) (5-2) In(CH3)3(g) + NH3(g) InN(s) + 3CH4(g) (5-3) InN(s) In (l) + N2(g) (5-4) In(l) + HCl InCl + H2 (5-5) InN(s) + HCl(g) InCl(g) + H2(g) + N2(g) (5-6) TMIn flows out of the smallest diameter tube at the center of the inlet where it reacts with HCl to form InCl (Eq. 5-1). The InCl leaves the inlet tube and reacts with NH3 to form InN in the hot substrate zone (Eq. 52). It is possible for InCl3 to form along with of InCl, however InCl3 has been determined to be thermodynamically unfavorable (compared to InCl) when the temperature is gr eater than 450K.272 The overall reaction for MO CVD mode (i.e. no HCl flow) is shown in Eq. 5-3. Decomposition reactions can also occur, such as thermal decomposition of InN at the growth temperature (Eq. 5-4) or etch ing of InN by HCl (Eq. 5-6). Complex chemical equilibrium calculations of the In -C-H-Cl-N system show that In droplets are preferentially etched (Eq. 5-5) before InN (Eq. 5-6) at typical InN growth temperatures (500-600 oC).234 Reactions in an H2 ambient were not considered since InN is typically grown in a nitrogen carrier gas. 5.3 Results 5.3.1 Characterization of InN Nanoro ds Grown on Different Substrates InN nanorods were grown by H-MOCVD on a, c-, and r-plane orientations of Al2O3 as well as Si (111) and GaN (0001) substrates. InN nanorods were grown at a deposition

PAGE 183

183 temperature of 600 oC, N/In ratio of 250, and a Cl/In ratio of 4.0. Growth on different substrates was investigated to understand how the polarity of the chosen substrate affects the structural properties of the nanorods. After growth InN na norods were shaved from the surface of the surface and placed onto lacey carbon grids for analysis by TEM. Growth on a-, c-, and r-Al2O3 substrates yield single crysta l InN nanorods, however they differ by shape and faceting. Some of the nanor ods are orientated in the c-axis however many nanorods grow in flower-like arra ngement which results in randoml y oriented growth directions of the nanorods (Figure 5-2). The most noti ceable difference between InN nanorods grown on a-, c-, and r-Al2O3 is the different morphologies of the nanorod tips. Growth on aand c-Al2O3 produces hexagonal cross sectioned tips (Figure 53), which are atomically flat while growth on non-polar r-Al2O3 generates faceted pencil-lik e tips (Figure 5-4). The la ck of diffraction contrast in the TEM images indicates a high level of st ructural perfection in the nanorods. When the nanorod is separated from the substrate surfac e the bottom edge is rectangular and is not atomically flat. This helps determine the growth direction of the nanorods (Figure 5-3 and 5-4). The tip facets form at a 30o angle to the long facets on the si de of the nanorod (Figure 5-4). When the nanorod is separated from the substrate a rectangular end is formed (TEM, Figure 5-5). The pencil-like tip of InN nanorods grown on r-plane Al2O3 are not atomically sharp, an approximately 20 nm plateau in the c-axis is found at the nanorod tip (Figure 5-6). Plane bending occurs in the tip of the sharpened nanor od (Figure 5-7). InN nanorods grown on Si (111) substrates have a similar tip shape as nanorods grown on r-Al2O3 while nanorods grown on GaN/c-Al2O3 show a similar morphology to nanorods grown on a-,c-Al2O3. Nanorods grown on Si (111) substrate show slightly asymmetric tips compared to r-Al2O3 (Figure 5-8). InN nanorods grown on all five differe nt substrates show some so rt of atomic roughness along the

PAGE 184

184 long side facets of the nanorod (Figure 5-9) Some nanorod specimens exhibit atomic smoothness along one of the side facets, however roughness is usually observed. Convergence beam electron diffraction ( CBED) was done on an InN nanorod grown on a-Al2O3. The CBED analysis revealed that the atom ically flat facets at the top end of the nanorod form at 60o angles with respect to the c-axis (Fig ure 5-10). Occasionall y the diameter of InN nanorods grown on a-Al2O3 varies along the length, shown by the v-shaped notch in region II of Figure 5-10a. CBED patterns of region II (Figure 5-10c) shows the presence of two crystals which are rotated 60o from each other. This 60o rotation corresponds to the [ 11 1 0] direction which has the same displacement vector as pr ismatic stacking faults (PSFs) which commonly occur in InN films grown on sapphire substrates.273 It is possible that these PSFs are responsible for the flower-like growth structure seen fo r these InN nanorods. The InN nanorod is grown with nitrogen polarity, which is consistent with previous results (Figure 5-11).272 Electron energy loss spectroscopy (EELS) was used to verify composition of the InN nanorods. The majority of InN nanorods show a pure In and N composition however oxygen is occasionally detected due to a small oxide laye r on the outside of the InN nanorods. The oxygen detection usually occurs when Al2O3 substrates are used. EELS spectra are shown for InN nanorods for grown on c-Al2O3, with and without the presence of a thin oxide layer (Figure 5-12). The amorphous oxide layer seldom form s, however, when formed it is very thin compared to the size of the nanowire (Figure 5-13 ). Residual oxygen in the reactor is a potential source of oxygen contamination that occasionally leads to the formation of a thin amorphous surface oxide. This is the most likely source of oxygen contamination since the experiments are carried out at atmospheric pressure It is also possible that a sm all oxide layer could form after the rods are exposed to air, however, not likel y since the oxide layer is seldom detected.

PAGE 185

185 InN nanorods were measured by photoluminescen ce (PL) and optical properties of the nanorods differ slightly depending on the substr ate used for growth. The position of the PL peaks for all nanorods tested fell in the range of 0.9-0.95 eV (Figure 5-14), which corresponds to a carrier concentration of the mid 1019 (based on the Moss-Burstein effect and a true band gap energy of 0.7 eV). These PL results are slight ly lower than energy values obtained previously (1.08 eV).272 The properties of these nanorods are c onsistent with bulk InN because the nanorod diameters are typically greater than 100 nm. The crystalline quality and the diameter of the nanorod are expected to affect PL intensity as well as position of th e peak. Larger sized diameter nanorods are expected to yield a stronger intens ity and a lower PL energy because there is less electron accumulation at the surface. A surface electron accumulation layer typically forms in InN and is expected to over estimate the ba nd gap energy of InN due to the Moss-Burstein effect.16,17,158,159,160 A larger diameter nanorod will have a lower concentration of electron accumulated at the surface and therefore produce a larger intensity. InN nanorods grown on Si (111) substrates, which have the largest diamet er (~ 600 nm) of samples tested, produces the strongest peak intensity and lowe st peak energy (Figure 5-14). 5.3.2 Growth of InN Nanoand Microrods to Increase Aspect Ratio. Previous growth of InN nanor ods by H-MOCVD has shown that typical diameters of InN nanorods range from 100 to 300 nm and lengths of about 1 m for a 1 hr growth.272 Therefore the typical lengths to width (asp ect) ratios range from 10:1 to 3.3:1. Single rod device fabrication of individual nanorods is cumbersome when the lengths are less than 10 microns. When increasing the length of the nanorod it is desired that the diameter remains in the nanometer range so the properties of the nanostruc ture remains significan tly different than bulk materials. For this reason growth parameters such as deposition temperature, Cl/In ratio, and

PAGE 186

186 N/In ratio were varied in an effort to incr ease the aspect ratio of InN nanorods grown by H-MOCVD. The Cl/In ratio is the ratio of inlet molar flows of HCl and TMIn. InN nanorods with the dimensions prev iously mentioned were grown at 600 oC, N/In ratio of 250 and a Cl/In ratio of 4.0 for 1 hr. The growth duration was increased to 90 min and the microrods (no longer nanometer sized) aspect rati o remained consistent with an average aspect ratio of 7:1 (Figure 5-15). The gr owth temperature was increased to 650 oC and the Cl/In ratio was increased to 4.5, the N/In ratio (250) re mained constant and nanorods were grown for 90 min. The density of the microrods was slightly reduced and the lengths ranged from 10 to 30 m, but the diameter also ranged from roughly 1 to 5 m, resulting in no change in the overall aspect ratio. The Cl/In ratio was again in creased to 5.0 without changing the growth temperature, N/In ratio, or growth duration (650 oC, 250, 90 min, respectively). Again the nanorods exhibited a similar aspect ratio and onl y the density of the na norods on the substrate was reduced (Figure 5-16). The growth temperature was increased to 700 oC and the N/In ratio was reduced to 200 while the Cl/In ratio remained at 5.0 for a 90 min growth. At these growth conditions etching becomes dominant and the micr orods are almost comple tely etched (Figure 5-17). Post growth HCl etchi ng was done to see if the InN mi crorods could be controllably etched to reduce the diameter to a nanometer range. Post growth et ching (1.1 sccm HCl, Tanneal = 550 oC, 30 min) of InN microrods (Tg = 650 oC, N/In = 250, Cl/In = 4.5) revealed that non-uniform etch pits formed on the microrod surface (Figure 5-18). From this study it is shown that the le ngth of H-MOCVD InN nanorods can only be increased at the expense of increasing the diam eter, which usually pro ceeds into the microrod range. A random sampling of InN mircorod lengths and diameters are given for a variety of growth conditions (Table 5-1). From this sample set it can be seen that the aspect ratio never

PAGE 187

187 exceeds 10:1 and that the average aspect ratio is 8:1. Post growth etching reveals that even though the nanorods have been shown to uniformly grow in the growth-etch transition region272 it is not possible to uniformly etch the microrods. 5.3.3 Transport Properties of InN Microrods Single InN microrod devices were fabricated and the contact resistivity and sheet resistivity of the microrods were determined by using a transmission line model. This is a similar method that was used to test the transport properties of InNIn2O3 core-shell nanowires (Section 4.3.4). InN microrods were grown on Si (111) substrates at 650 oC for 90 min at a N/In ratio = 250 and Cl/In = 4.5. The InN microrods were removed from the silicon substrate by sonication and then transferred to an insulating substrate. A sha dow mask was used to evaporate Ti/Al/Pt/Au (20/80/40/80 nm) oh mic contacts. Metal was depos ited in the openings of the shadow mask (50 x 50 m2 or 115 x 115 m2) leaving 10 m gaps between metal contacts. The total measured resistance (RT) for a hexagonal wire (Eq. 5-7)264 was determined from I-V measurements (Figure 5-19) using Eq. 5-8. 23 3 2 2a L R Rs c T (5-7) TIR V (5-8) 2 3 3 slopes (5-9) s c cR a 2 33 27 (5-10) The total measured resistance from I-V da ta was plotted versus microrod length/a2, where a is the distance from the center of the microrod to the facet edge (Figure 5-20). The microrods and their corresponding dimensions were determined by SE M (Figure 5-21). A linear fit was assigned to the data (Figure 5-20) where the y-intercept corresponds to half the contact resistance (Rc) and the relationship between th e slope of the line and sheet resistiv ity is shown in Eq. 5-9. Once the

PAGE 188

188 sheet resistivity and contact resistance are determ ined the contact resistivity can be calculated (Eq. 5-10). The sheet and contact resistivity for InN microrods was determined to be 2 x 10-2 and 4 x 10-6 -cm, respectively. When compared to the transport measurements done for InNIn2O3 in Chapter 4, the InN microrods have a higher sheet resistivity and a lower contact resistivity. It is not useful to compare the contact resistivity of the microrods to the nanowires because small changes in the slope of the linear f it could significantly change these values. It is likely that the InN microrods have a higher sheet resistivity be cause the core-shell na nowires are so conductive (thus giving a lower sheet resistiv ity). It was already suggested that the core-shell nanowires high conductivity was at fault for the lack of response in H2 gas detection, so it is likely that the high quality InN microrods have a lower carrier concentration. InN nanorods have proven to be successful at selectively detecting H2,271 which is more confirmation that core-shell nanowires are too conductive. 5.4 Conclusions The growth of InN nanorods on diffe rent substrates (a-, c-, r-Al2O3, Si (111), and GaN/c-Al2O3) by H-MOCVD reveals different nanorod tip st ructures based on the polarity of the substrate. When InN nanorods are grow n on polar substrates such as a-, c-Al2O3 the nanorod tips are flat while grow th on non-polar r-Al2O3 yields sharp nanorod tips. Growth on silicon substrates also produce pointed nanorod tips, ex cept the tips are asymme tric. When sharpened tips are noticed atomic plane bending occurs and a 20 nm plateau forms at the top of the nanorod tip. CBED analysis revealed that prismatic stacking faults may be responsible for the flower-like growth habit of the InN nanorods. TEM analysis of the side faceting of the nanorods shows atomic roughness and occasionally a thin oxide la yer can be detected, however, the nanorods are usually oxide free.

PAGE 189

189 By varying the growth conditions of InN nanor ods it was determined that the aspect ratio of the nanorods cannot be increased above 10:1, regardless of the growth conditions. Therefore the nanorods will eventually become microrods with increased growth duration. It was also shown that in-situ etching did not preferentially etch InN in one gr owth direction. Instead etch pits were formed over the surface of the microrod. Transport measurements were done for singl e InN microrods by us ing the transmission line method and it was shown that th e sheet resistivity of the InN microrods was higher than the core-shell nanowires discussed in Chapter 4. Th is added further confirma tion that the core-shell nanowires were too conductive for use as single wire gas sensing devices. Figure 5-1. Diagram of the H-MOCVD concentric inlet tube showing th e arrangement of source material flows and typical temperatures used experimentally for the source and deposition zones. Figure 5-2. A SEM image of the growth habits of InN nanorods illustrating flower-like growth and c-axis orientation (Ref. 272).

PAGE 190

190 Figure 5-3. Low resolution TEM im age of an InN nanorod grown on a-Al2O3 substrate showing flat hexagonal cross section. Also representative of growth on c-Al2O3. Figure 5-4. Low resolution TEM image of an InN nanorod grown on r-Al2O3 substrate. The pencil-like tip as well as faceting angle between tip and edge of the nanorod are shown. Figure 5-5. High resolution TEM im age of the rectangular shape whic h is seen at the end of the nanorod when it is shaved from the substrate surface.

PAGE 191

191 Figure 5-6. Low resolution and high resoluti on TEM images of the InN nanorod tip grown on r-Al2O3. The flat 20 nm plateau and and side facet of the InN nanorod can be seen. Figure 5-7. High resolution TEM images showin g plane bending in the tip of an InN nanorod grown on r-Al2O3.

PAGE 192

192 Figure 5-8. Low resolution TEM images of In N nanorods grown on Si ( 111) substrate showing asymmetric pencil tip. Figure 5-9. High resolution TEM images of si de facet roughness occurring for most nanorods grown by H-MOCVD regardless of the substrate.

PAGE 193

193 Figure 5-10. Low resolution TEM image and CBED patterns of an InN nanorod grown on a-Al2O3. A) Low resolution image of a singl e InN nanrod. B-D) Convergence beam electron diffraction images for labeled nanorod regions I-III, respectively. Figure 5-11. Experimental and calculated CBED images indicating nitrogen polarity.

PAGE 194

194 Figure 5-12. Electron energy loss spectra with and without the detection of oxygen. A) No oxygen detection. B) Nanowire wher e thin oxide layer is detected. Figure 5-13. High resolution TEM image showi ng thin amorphous oxide layer that sometimes forms on the outside of the InN nanorods.

PAGE 195

195 Figure 5-14. Photoluminescence spectra of InN nanorods grown on different substrates. A) Normalized linear scale. B) logarithmic scales. Figure 5-15. Scanning electron microscope im ages of InN microrods grown on Si (111) substrate for 90 min using Cl/I n = 4.0, N/In = 250, and Tg = 600 oC.

PAGE 196

196 Figure 5-16. Scanning electron microscope im ages of InN microrods grown on Si (111) substrate for 90 min using Cl/I n = 5.0, N/In = 250, and Tg = 650 oC. Figure 5-17. A SEM image of InN microrods grown on Si (111) substrate for 90 min using Cl/In = 5.0, N/In = 200, and Tg = 700 oC. Figure 5-18. Scanning electron microscope image of InN microrods grown on Si (111) substrate after post growth in-situ HCl gas etching.

PAGE 197

197 Figure 5-19. Current-voltage measurements for several InN microrods, used for determining the total measured resistance (RT) of each rod. Figure 5-20. Total measur ed resistance vs. length/a2 for InN microrods.

PAGE 198

198 Figure 5-21. Scanning electron microscope images and dimensions of InN microrods used for TLM analysis. Table 5-1. A random sampling of InN nanorod/mi crorod lengths and diameters for a variety of growth conditions. Tgrowth (oC) V/IIICl/III Growth duration (min) Substrate Length ( m) Diameter ( m) Aspect ratio 6002504.090Si (111)7.8 oo 1.65.0 6002504.090Si (111)10.5 oo 1.19.4 6502504.590Si (111)5.0 oo 0.77.1 6502504.590Si (111)38.2 oo 5.27.3 6502504.590Si (111)18.3 oo 2.38.1 6502504.590 c-Al2O312.9 oo 1.77.6 6502504.590 GaN/c-Al2O312.3 oo 1.48.6 6502505.090Si (111)37.2 oo 4.87.8 6502505.090Si (111)26.3 oo 3.28.2 6502505.090Si (111)17.9 oo 1.89.8 6502505.090 GaN/c-Al2O328.0 oo 3.09.3 Average =8.0

PAGE 199

199 CHAPTER 6 EXPLORATORY STUDY OF GALLIUM NITRIDE NANOAND MICROSTRUCTURED GROWTH BY METAL ORGANIC CHEMICAL VAPOR DEPOSITION 6.1 Introduction GaN nanostructures have gained attenti on for fundamental understanding of how dimensionality affects the electrical and optical properties, but also for nanodevice applications. Device applications involve nano-sized UV opt oelectronics, high speed FETs, and high temperature microelectronics.274-277 GaN nanostructures have been grown by a variety of methods including catalytic CVD (typically using Ni or Au catalyst), thermal CVD, MBE and pulsed laser deposition (a lso using catalysts).86,278-282 Most CVD processes use pure gallium metal, GaN powder, Ga2O3 oxide powder, or some combination of the three. As in the case of InN nanowires, it is bene ficial to controllable grow GaN nanostructures on substrates by MOCVD, since this growth tech nique is commonly used in industry. Also, III-nitride alloy growth can be more easily c ontrolled compared to the majority of growth methods described above. In this chapter the growth of GaN nanostruc tures by using a traditional MOCVD reactor is discussed. Growth parameters such as inlet N/ Ga ratio, deposition temperature, and substrate material were investigated with regards to the GaN structure. 6.2 Experimental Procedure For the growth of GaN nanotubes and nanowires a vertical inlet tube was used in the MOCVD reactor. Detailed descriptions of the MO CVD growth technique and vertical inlet tube setup are presented in Chapter 3 and 4, respec tively. For MOCVD growth of GaN nanowires and nanotubes, triethyl gall ium (TEGa) and ammonia (NH3) were used as precursors. The deposition temperature was varied from 560 to 850 oC and the N/Ga ratio ranged from 3,000 to 30,000. Growth occurred on Si (100), c-Al2O3, and GaN (5 m)/c-Al2O3 substrates. Silicon

PAGE 200

200 substrates were degreased in warm trichlor oethylene (TCE), acetone and methanol each for 5 min and then blow dry with a static free nitrogen gun. Sapphire substrates were only cleaned in methanol for 5 min and then blown dry. No chemical cleaning was done for GaN/c-Al2O3 substrates and a nitrogen gun was used to ensure that the surface was free of particulates before loading. Substrate nitridation at 850 oC for 15 min (1600 sccm NH3 in 4 SLM N2 dilution) was done for all samples and no low temperature buffer layers were used. 6.3 Results As in the case of InN nanowire growth by MOCVD, GaN nanowire growth occurs by a VLS mechanism. Once the inlet N/Ga ratio is increased beyond a certain threshold value, GaN nanowire growth proceeds by a vapo r-solid (VS) mechanism. This is a similar result seen for InN nanowires (see Section 4.3.2.2). GaN thin film growth optimization previously occurred and the ideal growth conditions for silicon substrate included subs trate nitridation and LT GaN buffer layer at 560 oC (N/Ga = 3,000) for 15 min followed by HT GaN at 850 oC (N/Ga = 3,000).234 These previous growth optimizations were made using the standard horizontal inlet tube (Figure 4-1). For compar ison a SEM image of a GaN film grown at ideal conditions using the horizontal in let tube is compared to GaN gr own by the vertical inlet tube (Figure 6-1). Both GaN films were grown on c-Al2O3 substrates. When the vertical inlet was used, no LT GaN buffer layer was used after substrat e nitridation and prior to growth. GaN thin films have previously been grown on c-Al2O3 substrates without a LT buffer (using the horizontal inlet) and the morphology di d not significantly change as in this instance. It is clear that the morphology of the two stru ctures is different (Figure 6-1) The horizontal inlet grown GaN shows a relatively smooth surface compared to the extremely rough surface produced by the vertical inlet, (tilted SE M image, Figure 6-1c). At 850 oC it is assumed that Ga droplets do not form since ammonia decomposition is rapid at these temperatures. Therefore the vertical

PAGE 201

201 island growth is believed to be due to significan t increase in the growth rate by the concentration of reactants directly over the susceptor and wafer tray. Triangular tips on the ends of the GaN are an indication that cubic phases might also be present (Figure 6-1d). XRD verified that cubic and hexagonal phases are present in the GaN film grown on Si ( 100) with the vertical inlet (Figure 4-2). It is difficult to distinguish between the hexagon al (0002) and cubic (111) phases of GaN because the 2 values overlap. It is likely that both of these phases coexist, since lower intensity polycrystall ine peaks are detected for both cubic and hexagonal phases. The polycrystalline peaks can be positively identi fied because the peaks occur at different 2 values. The SEM images in Figure 6-1 are representative for GaN grown with the vertical inlet on Si (100), c-Al2O3, and GaN/c-Al2O3, therefore the presence of cubi c phases is expected at these conditions regardless of the substrate and this result was confirmed by XRD. In order to produce more favorable conditions for GaN nanowire growth without the aid of catalysts the deposition temperatur e was reduced while maintaining a constant inlet N/Ga ratio of 3,000. At a deposition temperature of 750 oC the growth rate of GaN is reduced and films are formed on GaN/c-Al2O3 and c-Al2O3 that are less rough comp ared to growth at 850 oC with the vertical inlet (SEM, Figure 63). It is not completely clear why the GaN deposited on GaN/c-Al2O3 shows small voids or pits on the surface of th e film. It is likely that Ga wetting of the surface increases st rain in the first one or two monolayers on the surface, as in the case of In wetting on GaN which was previously described in Chapter 4. Evidence for the formation of Ga-droplets is verified by GaN growth on Si (100) substrates (a t the same temperature). At 750 oC the GaN deposition is significan tly different on Si (100), which is believed to be due to gallium wetting of the substrate surface, similar to the results presented in Chapter 4. On Si (100) substrates the GaN forms mi crowires with Ga droplets at the tip, (Figure 6-4a). Other

PAGE 202

202 sections on the Si (100) substr ate can be found where the GaN dr oplet has been consumed which reveals a tube structure (Figure 6-4b,c). Decreasing the deposition temperature to 650 oC with a constant inle t N/Ga ratio of 3,000 reduces the effective N/Ga ratio becaus e the decreased temperature reduces NH3 decomposition efficiency. With these growth conditions the decrease in active nitroge n creates an excess of gallium metal which promotes microstructure growth via VLS mechanism. Figure 6-5 shows SEM images of GaN nanowires grown on GaN/c-Al2O3, c-Al2O3, and Si (100). The microtubes range from 1 to 3 m thick where the diameter decreases as the tube get longer due to the consumption of the Ga metal dropl et. Lengths range from 2 to 15 m where the size of the remaining Ga droplet at the end of growth is inversely proportional to th e length. EDS was done on the GaN microtube and only Ga and N were detected, which suggests that there is no significant carbon incorporation that is responsible for the formati on of the tube structure (Figure 6-6). This EDS spectrum is representative for GaN microtubes grown at all growth conditions (i.e. 750 oC, 650 oC, etc.). GaN was also deposited with th e vertical inlet using an in let N/Ga ratio of 3,000 at 600 oC and 560 oC. As expected the lower te mperature further reduced the NH3 decomposition and an increase in Ga metal at the tip of the microtubes is noticed (Figure 6-7). To determine the extent of nitride growth below the excess metal, the Ga droplets were selectively wet etched in a 10% HCl (90% water) by volume for 10 min. After etching the excess gallium is removed, which reveals a GaN nanotube structur e (Figure 6-8). EDS was done after etching which confirmed that the nanotubes were compos ed of Ga and N (Figure 6-9). The deposition temperature that provided th e most uniform GaN microstructure for all three substrates tested was 650 oC. For this reason the inlet N/Ga ratio was increase at this

PAGE 203

203 deposition temperature to see the morphological e ffect on the microwire st ructure. The highest inlet N/Ga ratio tested (20,000) produced mostly GaN thin films with o ccasional sites where 1D growth occurred (Figure 6-10). Th is result was consistent for all substrates tested (GaN/c-Al2O3, c-Al2O3, and Si (100)). The growth at this high N/Ga ratio forms small microrods that have grown by a VS mechanism instead of a VLS mechanism. At th is high N/Ga ratio the group V (active N) species becomes the excess source materi al and the growth rate of the nanostructure is limited by the group III flux. The sparse microrods are approximately 1 m in diameter and 2-3 m in height. Lowering the N/Ga ratio by half to 10,000 (at Tg = 650 oC) increases the number of 1D nucleation because TEGa incr eases possibility of metal rich nucleation sites (Figure 6-11). The density of 1D growth is increased when th e inlet N/Ga ratio is reduced to 7,000 (Figure 6-12). For these growth conditions nanowires lengths range from 3-5 m with diameters ranging from 100-200 nm. Reducing the N/Ga ratio below 7,000 at a depos ition temperature 650 oC results in similar nanowire structures, however the samples with the greatest uniformity across the substrate were grown at this condition. Frequently, small areas of curly GaN nanow ires were found on substrates whose edges were placed outside of the prim ary deposition zone or hot spot of the vertical inlet tube. A schematic of the vertical inlet and the associated hot spot were previously discussed in Chapter 4 (Figure 4-5). Outside the hot spot curly GaN nanowires typically form (Figure 6-13). These nanowires are typically seen for depo sition temperatures at or below 650 oC and they occur at the edges of the graphite susceptor. It is not completely clear wh y these curly GaN nanowires form but it is possible that the relative N/Ga ratio is significantly changed far away from the primary deposition zone. From a quality control point of view the region furthest from the vertical inlet tube is the most difficult re gion where growth can be cont rol growth. These curly GaN

PAGE 204

204 nanowires have lengths (uncur led) in the range of 10-20 m and diameters of 100-300 nm. Producing the growth of these nanowires in th e primary deposition zone proved to be very difficult and could not be reproducibly controlled. GaN nanotubes were grown at 650 oC on c-Al2O3 to be fabricated in to a gas sensing device for selective room temperature H2 detection. The wire bonding st ep (Figure 6-13) proved to be detrimental to the device operation because th e thin walled nanotubes were crushed upon contacting. This contacting method proved ineffective for ma king a successful Ohmic contact and therefore the device test could not be completed. 6.4 Conclusions From this exploratory study it has been show n that GaN microstructures can be grown by traditional MOCVD by using vertical inlet tube to deliver the source materials. The most reproducible microstructures were produ ces at a deposition temperature of 650 oC and an inlet N/Ga ratio of 7,000. For these growth conditions the effect of the chosen substrate had little influence on the morphology of the microstructure This was also the case for high temperature growth (850 oC) even though only island grow th occurred (instead of micro or nanostructures). When a deposition temperature was between 650 and 750 oC the chosen substrate material had a greater effect on the microstructure likely due to the wetting nature (surface tension) of gallium on the substrate.

PAGE 205

205 Figure 6-1. Comp arison of GaN/c-Al2O3 grown at similar growth c onditions with the horizontal and vertical inlets. A) horizontal inlet s howing smooth GaN film. B-D) vertical inlet showing extremely rough GaN surface morphology. Figure 6-2. X-ray diffraction pattern of GaN grown with the vertical inlet on Si (100) substrates. Cubic and hexagonal phases are detected.

PAGE 206

206 Figure 6-3. Scanning el ectron microscope images of GaN grown at 750 oC and an inlet N/Ga ratio of 3,000 on different substrates. A) c-Al2O3 substrate. B) GaN/c-Al2O3 substrates. Figure 6-4. Scanning electr on microscope images of GaN microtubes grown at 750 oC and an inlet N/Ga ratio of 3,000 on Si (100) substrates. A) Ga droplet at tip of microwire. B,C) Microtube structure of GaN.

PAGE 207

207 Figure 6-5. Scanning electron mi croscope images of GaN grown using the vertical inlet at 650 oC and N/Ga ratio of 3,000 on diffe rent substrates. A) GaN/c-Al2O3 substrate. B) c-Al2O3 substrates. C,D) Si (100) substrate.

PAGE 208

208 Figure 6-6. Representative EDS spectrum of single GaN microtube grown on Si ( 100) substrate. Figure 6-7. Scanning electron micr oscope images of GaN grown with the vertical inlet at the lowest temperature tested (560-600 oC) showing excess of metal droplets on the surface of the nanotubes.

PAGE 209

209 Figure 6-8. Scanning electron mi croscope images of GaN growth with the vertical inlet at 560 oC after a 10 min 10% HCl wet etch showing nanotube structure. Figure 6-9. Energy dispersi ve spectrum of GaN nanotube s after HCl wet etching.

PAGE 210

210 Figure 6-10. Scanning electron microscope images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 20,000. Figure 6-11. A SEM images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 10,000.

PAGE 211

211 Figure 6-12. Scanning electron microscope images of GaN/Si (100) with the vertical inlet at 650 oC and N/Ga = 7,000.

PAGE 212

212 Figure 6-13. Scanning electron microscope images of curly GaN nanowires that form outside the primary deposition zone of the vertical inlet. Figure 6-14. Wire bonding step of GaN nanotube H2 gas sensing device where damage to the nanotubes occurred.

PAGE 213

213 CHAPTER 7 FUTURE WORK AND RECOMMENDATIONS 7.1 Role of Oxygen in InN Currently it is not well understood how oxygen incorporates into InN and the resulting effect on electrical and optical properties. There are many debata ble claims made about the role of oxygen in InN and these claims are summarized in reviews, for example see Ref. 50. It has been suggested that oxygen is a possible ca use for increasing the band gap energy of InN44 and that oxygen can contribute to n-type doping of InN.283 Oxygen is believed to exist in InN samples in amorphous phases, which is w hy oxygen may not be detected from XRD measurements. InN films presumed to contain amorphous oxygen form crystalline In2O3 phases after annealing at 500 oC in a nitrogen atmosphere.240 Room temperature aging of InN films in air for periods of months and even years has produced XRD peaks of crystalline In2O3.50 It is difficult to study the role of oxygen fr om current literature because much of the evidence is contradictory. For example in this work we unsucce ssfully tried to reproduce the results of Hur et al.284 in which InN films were annealed at 550 oC for 10 min under a high vacuum (10-5 Torr) to form crystalline In2O3. We annealed single crystal InN/c-Al2O3 thin films in a low pressure (100 Torr) nitrogen atmosphere (after being exposed to air for several months) at 500, 525, 550 oC for a duration greater or equal to 10 mi n. XRD results showed that InN films did start to decompose (due to the ap pearance of an indium peak) at Tanneal > 500 oC, however no crystalline In2O3 peaks were produced. Since current literature results are scattere d and analysis has occurred on a variety of different substrates and growth techniques it is suggested that oxygen incorporation into InN be studied. Understanding the role of oxygen is important for improving film quality, identifying the ideal deposition techniques and how aging in ai r will affect material quality. The air-surface

PAGE 214

214 interface is also important for InN because it is believed that this in terface affects the surface charge accumulation seen in InN.16,17 The surface charge accumul ation layer is one of the hurdles that must be ove rcome for successful p-type doping of InN to occur. 7.2 Improving Crystalline Quality of InxGa1-xN (1 x 0.3) Alloys Growth of InxGa1-xN alloys is difficult due to the 11% lattice mismatch and resulting phase separation that occurs between InN and GaN. Differences in equilibrium nitrogen vapor pressures also make it difficult to incorporate in dium at high growth temperatures (preferred for GaN) and crystalline quality must be sacrificed at low temperatures for indium rich alloys. In recent years Ga-rich alloys (up to 20-30% indi um) have been produced with high structural qualities and good photoluminescence properties, mostly due to a pplications in visible LEDs (Technologies and Devices International, Inc. press release, TDI Demo nstrates Novel InGaN Epitaxial Materials at the In ternational Workshop on Nitrid e Semiconductors (IWNS) 2006, http://www.tdii.com/release_html_b1.html accessed April 2007). In-rich InxGa1-xN alloys have gained more attention in recen t years however improvements in film quality and reproducibility must occur, especially in the compositional range where phase sepa ration is most prevalent (0.7 x 0.3). The relationship between growth temp erature, alloy composition, and N/III ratio will be crucial for improving crystalline qu ality. Developing high quality InxGa1-xN alloys over the entire compositional range is beneficial for a variet y of applications such as high efficiency solar cells, terahertz electronics, comm unications, and optoelectronics. 7.3 P-Type Doping of InN and In-Rich InxGa1-xN InN and In-rich InxGa1-xN based electronic and optoelectron ic devices are severely limited by the inability to form p-n homoj unctions, since InN and In-rich InxGa1-xN alloys have not successfully been p-type doped. Some research ers are making progress in Mg doping of InN to make p-type films,16,17 but even slight evidence of p-type doping in the bulk is difficult to

PAGE 215

215 achieve. It is suggested that Mg and Zn are used for p-type d opants since Mg is a successful ptype dopant in GaN and Zn has been successful for InP. Both of these elements should be tried independently and also together via co-doping. Mg acceptors are difficult to ionize in GaN since the activation energy is higher for the deep acceptor so annealing of InN doped films will become important. If possible a high pressure annealing cham ber should be designed and built such that InN samples can be annealed at higher temperatures to prevent film decomposition. 7.4 Growth of InxGa1-xN Alloys by H-MOVPE As discussed in Chapter 5 the H-MOVPE technique offers the unique ability to be operated as a conventional MOVPE reactor or a HVPE reactor that uses metal organic sources instead of metal only precursor. H-MOVPE is beneficial because it has already been proven to be a valuable technique for high quality growth of InN on silicon substrates and has exceptionally high growth rates (compared to conventional MOVPE). Theref ore it is suggested that InxGa1-xN alloys are grown by H-MOVPE which can then be compared to the growth of InxGa1-xN by MOVPE (presented in this work). Specifically it would be interesting to determine the effect on phase separation when H-MOVPE is used, since the deposition technique operates closer to equilibrium than MOVPE. On several occasions during this work the film thickness of MOVPE InxGa1-xN limited the amount of characte rization and analysis that could be done, such as PL and THz measurements. Once InxGa1-xN growth is optimized for H-MOVPE a 1 m thick film could be produce in one hr, which would take appr oximately 18 hr by MOVPE. Thick (1-3 m) layer will also be required for future solar cell device applications which simply cannot be obtained by our MOVPE system.

PAGE 216

216 LIST OF REFERENCES 1 P. C. H. Chan, 6th IEEE International Conference on Electronic Packaging Technology, Shenzhen, China, 2005, p. 13. 2 L. Liu and J. H. Edgar, Mat. Sci. Eng. R 37, 61 (2002). 3 C. J. Sun, P. Kung, A. Saxier, H. Ohsato, E. Bigan, M. Razeghi, and D. K. Gaskill, J. Appl. Phys. 76, 236 (1994). 4 T. Sasaki and S. Zembutzu, J. Appl. Phys. 61, 2533 (1987). 5 O. Ambacher, J. Phys. D 31, 2653 (1998). 6 E.S. Hellman, MRS Internet. J. Nitride Semicond. Res. 3, 11 (1998). 7 F. A. Ponce, D. P. Bour, W. T. Young, M. Saunders, and J. W. Steeds, Appl. Phys. Lett. 69, 337 (1996). 8 A. R. Smith, R. M. Feenstra, D. W. Greve, J. Neugebauer, and J. E. Norhrup, Phys. Rev. Letters 79, 3934 (1997). 9 A. Kazimirov, G. Scherb, J. Zegenhagen, T. L. Lee, M. J. Bedzyk, M. K. Kelly, H. Angerer, and O. Ambacher, J. Appl. Phys. 84, 1703 (1998). 10 M. Sumiya and S. Fuke, MRS Inte rnet J. Nitride Semicond. Res. 9, 1 (2004). 11 M. Seelmann-Eggebert, J. L. Weyher, H. Obl oh, H. Zimmermann, A. Rar, and S. Porowski, Appl. Phys. Lett. 71, 2635 (1997). 12 T. Matsuoka, Superlatt. Microstruc. 37, 19 (2005). 13 W. Walukiewicz, Physica B 302-303, 123 (2001). 14 J. Wu, W. Walukiewicz, K. M. Yu, W. Shan, J. W. Ager III, E. E. Haller, H. Lu, W. J. Schaff, W. K. Metzger, and S. Kurtz, J. Appl. Phys. 94, 6477 (2003). 15 J. Furthmuller, P. H. Hahn, F. Fuchs, and F. Bechstedt, Phys. Rev. B 72, 205106 (2005). 16 R. E. Jones, K. M. Yu, S. X. Li, W. Walukiewic z, J. W. Ager, E.E. Haller, H. Lu, and W. J. Schaff, Phys. Rev. Lett. 96, 125505 (2006). 17 V. Cimalla, M. Niebelschutz, G. Ecke, O. Amba cher, R. Goldhahn, H. Lu, and W. J. Schaff, Phys. Stat. Sol. C 3, 1721 (2006). 18 E. Bellotti, B. K. Doshi, K. F. Brennan, J. D. Albrecht, and P. P. Ruden, J. Appl. Phys. 85, 916 (1999).

PAGE 217

217 19 B. E. Foutz, S. K. OLeary, M. S. Shur, and L. F. Eastman, J. Appl. Phys. 85, 7727 (1999). 20 R. Juza and H. Hahn, Zeitschr. Anorgan. Allgem. Chem. 239, 282 (1938). 21 R. Juza and R. Rabenau, Zeits chr. Anorgan. Allgem. Chem. 285, 212 (1956). 22 T. Renner, Zeitschr. Anorgan. Allgem. Chem. 298, 22 (1959). 23 G. V. Samsonov, The Nitrides, Na ukova Dumka, Kiev, Ukraine, 1969. 24 J. B. MacChesney, P. M. Bridenbaugh, and P. B. OConnor, Mat. Res. Bull. 5, 783 (1970). 25 N.A. Gorjunova, Slosnye Almazopodobnye Poluprovodniki, Moscow, Russia, 1964. 26 H. J. Hovel and J. J. Cuomo, Appl. Phys. Lett. 20, 71 (1972). 27 K. Osamura, K. Nakajima, Y. Murakami, P. H. Shingu, and A. Oh tsuki, Sol. St. Comm. 11, 617 (1972). 28 J. W. Trainor and K. Rose, J. Electron. Mater. 3, 821 (1974). 29 K. Osamura, S. Naka, and Y. Murakami, J. Appl. Phys. 46, 3432 (1975). 30 N. Puychevrier and M. Menoret, Thin Sol. Films 36, 141 (1976). 31 T. L. Tansley and C. P. Foley, Electron. Lett. 20, 1066 (1984). 32 T. L. Tansley and C. P. Foley, Proceedings of the 3rd IEEE International Conference on Semi-Insulating III-V Materials, Warm Springs, Oregon, 1984, p. 497. 33 C.P. Foley and T.L. Tansley, Appl. Surf. Sci. 22/23, 663 (1985). 34 C.P. Foley and T.L. Tansley, Phys. Rev. B 33, 1430 (1986). 35 T. L. Tansley and C. P. Foley, J. Appl. Phys. 59, 3241 (1986). 36 T. L. Tansley and C. P. Foley, J. Appl. Phys. 60, 2092 (1986). 37 T. Matsuoka, H. Tanaka, T. Sasaki, and A. Katsui, Proceedings of the 16th International Symposium Gallium Arsenide and Related Co mpounds, Karuizawa, Japan, 1990, p. 141. 38 A. Wakahara and A. Yoshida, Appl. Phys. Lett. 54, 709 (1989). 39 A. Wakahara, T. Tsuchiya, and A. Yoshida, J. Cryst. Growth 99, 385 (1990).

PAGE 218

218 40 A. Yamamoto, T. Tanaka, K. Koide, and A. Hashimoto, Phys. Stat. Sol. A 194, 510 (2002). 41 J. Wu, W. Walukiewicz, W. Shan K. M. Yu, J. W. Ager III, E. E. Haller, Hai Lu, and W. J. Schaff, Phys. Rev. B 66, 201403 (2002). 42 V. Y. Davydov, A. A. Klochikhin, R. P. Sesyan, V. V. Emtsev, S. V. Ivanov, F. Bechstedt, J. Furthmuller, H. Harima, A. V. Mudryi, J. Ader hold, O. Semchinova, and J. Graul, Phys. Stat. Sol. B 229, R1 (2002). 43 V. Y.Davydov, A. A. Klochikhin, V. V. Emtsev, S. V. Ivanov, V. V. Vekshin, F. Bechstedt, J. Furthmuller, H. Harima, A. V. Mudryi, A. Hashimoto, A. Yamamoto, J. Aderhold, J. Graul, E. E. Haller, Phys. Stat. Sol. B 230, R4 (2002). 44 V. Y. Davydov, A. A. Klochikhin, V. V. Em tsev, D. A. Kurdyukov, S. V. Ivanov, V. A. Vekshin, F. Bechstedt, J. Furthmuller, J. Ad erhold, J. Graul, A. V. Mudryi, H. Harima, A. Hashimoto, A. Yamamoto, and E. E. Haller, Phys. Stat. Sol. B 234, 787 (2002). 45 J. Wu, W. Walukiewicz, K. M. Yu, J. W. Ager III, E. E. Haller, Hai Lu, W. J. Schaff, Y. Saito, and Y. Nanishi, Appl. Phys. Lett. 80, 3967 (2002). 46 T. Matsuoka, H. Okamoto, M. Nakao, H. Harima, and E. Kurimoto, Appl. Phys. Lett. 81, 1246 (2002). 47 M. Hori, K. Kano, T. Yamaguchi, Y. Saito, T. Araki, Y. Nanishi, N. Teraguchi, and A. Suzuki, Phys. Stat. Sol. B 234, 750 (2002). 48 Y. Saito, H. Harima, E. Kurimoto, T. Yamaguchi N. Teraguchi, A. Suzuki, T. Araki, and Y. Nanishi, Phys. Stat. Sol. B 234, 796 (2002). 49 T. Miyajima, Y. Kudo, K. L. Liu, T. Uruga, T. Honma, Y. Saito, M. Hori, Y. Nanishi, T. Kobayashi, and S. Hirata, Phys. Stat. Sol. B 234, 801 (2002). 50 K. S. A. Butcher, T.L. Tansley, Superlatt. Microstruc. 38, 1 (2005). 51 B. Monemar, P. P. Paskov, and A. Kasic, Superlatt. Microstruc. 38, 38 (2005). 52 D. Alexandrov, K. S. A. Butcher, an d T. L. Tansley, J. Cryst. Growth 288, 261 (2006). 53 P. Carrier and S. H. Wei, J. Appl. Phys. 97, 033707 (2005). 54 K. M. Yu, Z. Liliental-Weber, W. Walukiewicz, W. Shan, J. W. Ager III, S. X. Li, R. E. Jones, and E. E. Haller, Appl. Phys. Lett. 86, 071910 (2005). 55 P. Carrier and S. H. Wei, 27th International Conf erence on the Physics of Semiconductors, Flagstaff, AZ, 2004, p. 287.

PAGE 219

219 56 T. Ouisse, Europhys. Lett. 46, 519 (1999). 57 V. W. L. Chin, T. L. Tansley, and T. Osotchan, J. Appl. Phys. 75, 7365 (1994). 58 D. C. Look, H. Lu, W. J. Schaff, J. Jasi nski, and Z. Liliental-Weber, Appl. Phys. Lett. 80, 258 (2002). 59 K. S. A. Butcher, M. Wintrebert-Fouquet, S. K. Shrestha, H. Timmer, K. E. Prince, and T. L. Tansley, Mater. Res. Soc. Symp. Proc. 743, 707 (2003). 60 G. Mie, Ann. Phys. 25, 277 (1908). 61 S. V. Ivanov, t. V. Shubina, V. N. Jmerik, V. A. Vekshin, P. S. Kopev, and B. Monemar, J. Cryst. Growth 269, 1 (2004). 62 F. Bechstedt, J. Furthmuller, M. Ferhat, L. D. Teles, L. M. R. Scolfaro, J. R. Leite, V. Y. Davydov, O. Ambacher, and R. Goldhahn, Phys. Stat. Sol. A 195, 628 (2003). 63 O. Ambacher, M. S. Brandt, R. Dimitrov, T. Me tzger, M. Stuzmann, R. A. Fischer, A. Miehr, A. Bergmaier, and G. Dollinger, J. Vac. Sci. Technol. B 14, 3532 (1996). 64 A. Yamamoto, Y. Murakami, K. Koide, M. ad achi, and A. Hashimoto, Phys. Stat. Sol. B 228, 5 (2001). 65 T. Matsouka, in GaN and Related Materials edited by S. J. Pearton (Gordon and Breach, New York, 1997), p. 53. 66 A. Koukitu, T. Taki, N. Takahash i, H. Seki, J. Cryst. Growth 197, 99 (1999). 67 A. Koukitu, Y. Kumagai, N. Kubota, and H. Seki, Phys. Stat. Sol. B 216, 707 (1999). 68 A. G. Bhuiyan, A. Hashimoto, and A. Yamamoto, J. Appl. Phys. 94, 2779 (2003). 69 J. Grandal and M. A. Sanch ez-Garcia, J. Cryst. Growth 278, 373 (2005). 70 A. Yamamoto, M. Tsujino, M. Ohkubo, and A. Hashimoto, J. Cryst. Growth 137, 415 (1994). 71 Y. C. Pan, W. H. Lee, C. K. Shu, H. C. Lin, C. I. Chiang, H. Chang, D. S. Lin, M. C. Lee, and W. K. Chen, Jpn. J. Appl. Phys. 38, 645 (1999). 72 T. Tsuchiya, H. Yamano, O. Miki, A. Wakahara, and A. Yo shida, Jpn. J. Appl. Phys. 38, 1884 (1999). 73 T. C. P. Chen, C. Thomidis, J. Abell, W. Li, and T. D. Moustakas, J. Cryst. Growth 288, 254 (2006).

PAGE 220

220 74 J. W. Ju, J. I. Lee, and I. H. Lee, Thin Sol. Films 515, 2291 (2006). 75 S. Yamaguchi, M. Kariya, S. Nitta, T. Takeuchi C. Wetzel, H. Amano, and I. Akasaki, J. Appl. Phys. 85, 7862 (1999). 76 K. Xu, W. Terashima, T. Hata, N. Hashimoto, Y. Ishitani, and A. Yoshikawa, Phys. Stat. Sol. C 0, 277 (2002). 77 H. Lu, W. J. Schaff, J. Hwang, H. Wu, G. Koley, and L. F. Eastman, Appl. Phys. Lett. 79, 1489 (2001). 78 A. Yamamoto, K. Sugita, H. Takatsuka, Y. Hamano, N. Imai, and A. Hashimoto, Proceedings of the International Workshop on Nitride Semiconductors, Pittsburgh, PA, 2004, p. 113. 79 S. Gwo, C. L. Wu, C. H. Shen, W. H. Chang, T. M. Hsu, J. S. Wang, and J. T. Hsu, Appl. Phys. Lett. 84, 3765 (2004). 80 C. L. Wu, C. H. Shen, H. W. Lin, H. M. Lee, S. Gwo, Appl. Phys. Lett. 87, 241916 (2005). 81 S. Iijima, Nature 354, 56 (1991). 82 M. Law, L. E. Greene, J. C. Johnson, R. Saykally, and P. Yang, Nature Mater. 4 455 (2005). 83 A. Motayed, A.V. Davydov, M. He, J. Melngailis, and S. N. Mohammad, 10th Nanotech Conference, Santa Clara, CA, 2007. 84 S. D. Dingman, N. P. Rath, P. D. Markow itz, P. C. Gibbons, and W. E. Buhro, Angew. Chem. Int. Ed. 39 1470 (2000). 85 C. Y. Chang, G. C. Chi, W. M. Wang, L. C. Chen, and K. H. Chen, Appl. Phys. Lett. 87, 093112 (2005). 86 C. Y. Chang, G. C. Chi, W. M. Wang, L. C. Chen, K. H. Chen, F. Ren, and S. J. Pearton, J. Electron. Mater. 35, 738 (2006). 87 G. Cheng, E. Cimpoiasu, E. Stern, R. Munde n, N. Pradhan, A. Sanders, and M. A. Reed, Proc. 5th IEEE Conf. on Nanotech. 2, 447 (2005). 88 U. K. Gautam, K. Sardar, F. L. Deepak, and C. N. R. Rao, Pramana 65, 549 (2005). 89 X. H. Ji, S. P. Lau, H. Y. Yang, and S. F. Yu, Nanotechnology 16, 3069 (2005). 90 M. C. Johnson, C. J. Lee, E. D. Bourret-Cour chesne, S. L. Konsek, S. Aloni, W. Q. Han, and A. Zettl, Appl. Phys. Lett. 85, 5670 (2004).

PAGE 221

221 91 Z. H. Lan, C. H. Liang, C. W. Hsu, C. T. Wu H. M. Lin, S. Dhara, K. H. Chen, L. C. Chen, and C. C. Chen, Adv. Funct. Mat. 14, 233 (2004). 92 C. H. Liang, L. C. Chen, J. S. Hwang, K. H. Chen, Y. T. Hung, and Y. F. Chen, Appl. Phys. Lett. 81, 22 (2002). 93 S. Luo, W. Zhou, Z. Zhang, L. Liu, X. Dou, J. Wang, X. Zhao, D. Liu, Y. Gao, L. Song, Y. Xiang, J. Zhou, and S. Xie, Small 1, 1004 (2005). 94 Z. Qian, S. Hou, J. Zhang, R. Li, Z. Shen, X. Zhao, and Z. Xue, Physica E 30, 81 (2005). 95 K. Sardar, M. Dan, B. Schwenzer, and C. N. R. Rao, J. Mat. Chem. 15, 2715 (2005). 96 B. Schwenzer, L. Loeffler, R. Seshadri, S. Ke ller, F. F. Lange, S. P. DenBaars, and U. K. Mishra, J. Mat. Chem. 14, 637 (2004). 97 T. Stoica, R. J. Meijers, R. Calarco, T. Richther, E. Sutter, H. Luth, Nano Lett. 6, 1541 (2006). 98 T. Tang, S. Han, W. Jin, X. Liu, C. Li D. Zhang, and C. Zhou, J. Mater. Res. 19, 423 (2004). 99 S. Vaddiraju, A. Mohite, A. Chin, M. Meyyappa n, G. Sumanasekera, B. W. Alphenaar, and M. K. Sunkara, Nano Lett. 5, 1625 (2005). 100 S. Yin, Z. Z. Sun, J. Lu, and X. R. Wang, Appl. Phys. Lett. 88, 233110 (2006). 101 L. W. Yin, Y. Brando, Y. C. Zhu, D. Golberg, and M. S. Li, Appl. Phys. Lett. 84, 1546 (2004). 102 L. W. Yin, Y. Bando, D. Golber g, and M. S. Li, Adv. Mater. 16, 1833 (2004). 103 J. Zhang, L. Zhang, X. Peng, and X. Wang, J. Mat. Chem. 12, 802 (2002). 104 J. Zhang, B. Xu, F. Jiang, Y. Yang, and J. Li, Phys. Lett. A 337, 121 (2005). 105 J. Zhang, F. Jiang, Y. Yang, B. Xu, J. Li, X. Wang, and S. Wang, Mater. Lett. 60, 2153 (2006). 106 H. J. Park, O. Kryliouk, T. Anderson, D. Khokhlov, and T. Burbaev Physica E (In press, corrected proof, 2006). 107 Y. Xia, P. Yang, Y. Sun, Y. Wu, B. Mayers, B. Gates, Y. Yi n, F. Kim, and H. Yan, Adv. Mater. 15, 353 (2003). 108 F. Qian, S. Gradecak, Y. Li, C. Y. Wen, and C. Lieber, Nano Lett. 5, 2287 (2005).

PAGE 222

222 109 T. Nagatomo, T. Kuboyama, H. Minamino, and O. Omoto, Jpn. J. Appl. Phys. 28, L1334 (1989). 110 N. Yoshimoto, T. Matsuoka, T. Sasa ki, and A. Katsui, Appl. Phys. Lett. 59, 2251 (1991). 111 S. Nakamura, T. Mukai, and Senoh, Appl. Phys. Lett. 64, 1687 (1994). 112 S. D. Lester, F. A. Ponce, M. G. Craf ord, D. A. Steigerwald, Appl. Phys. Lett. 66, 1249 (1995). 113 T. Mukai, S. Nagahama, N. Iwasa, M. Senoh, and T. Yamada, J. Phys. C 13, 7089 (2001). 114 R. Ascazubi, I. Wilke, K. Denniston, H. Lu, and W.J. Schaff, Appl. Phys. Lett. 84, 4810 (2004). 115 G. Franssen, T. Suski, P. Perlin, H. Teisseyre, A. Khackapuridze, L.H. Dmowski, J.A. Plesiewicz, A. Kaminska, M. Kurouchi and Y. Nanishi, Appl. Phys. Lett. 89, 121915 (2006). 116 P. Schley, R. Goldhahn, A. T. Winzer, G. G obsch, V. Cimalla, O. Ambacher, M. Rakel, C. Cobet, N. Esser, H. Lu, and W. J. Schaff, Phys. Stat. Sol. B 243, 1572 (2006). 117 S. X. Li, K. M. Yu, J. Wu, R. E. Jones, W. Walukiewicz, J. W. Ager III, W. Shan, E. E. Haller, H. Lu, and W. J. Schaff, Physica B 376-377, 432 (2006). 118 Z. Liliental-Weber, D. N. Zakharov, K. M. Yu, J. W. Ager III, W. Walukiewicz, E. E. Haller, H. Lu, and W. J. Schaff, Physica B 376-377, 486 (2006). 119 W. Walukiewicz, R. E. Jones, S. X. Li, K. M. Yu, J. W. Ager III, E. E. Haller, H. Lu, and W. J. Schaff, J. Cryst. Growth 288, 278 (2006). 120 E. Lioudakis, A. Othonos, E. Dimakis, E. Ili opoulos, and A. Georgakilas, Appl. Phys. Lett. 88, 121128 (2006). 121 H. Naoi, M. Kurouchi, D. Muto, S. Takado, T. Araki, T. Miyajima, H. Na, and Y. Nanishi, Phys. Stat. Sol. A 203, 93 (2006). 122 S. K Lin, K. T. Wu, C. P. Hua ng, C. T. Liang, Y. H. Chang, Y. F. Chen, P. H. Chang, N. C. Chen, C. A. Chang, H. C. Peng, C. F. Shih, K. S. Liu, and T. Y. Lin, J. Appl. Phys. 97, 46101 (2005). 123 I. H. Ho and G. B. Stringfellow, Appl. Phys. Lett. 69, 2701 (1996). 124 S. Y. Karpov, MRS Internet J. of Nitride Semicond. Res. 3, 16 (1998). 125 T. Matsuoka, N. Yoshimoto, T. Sasaki, and A. Katsui, J. Electron. Mater. 21, 152 (1992).

PAGE 223

223 126 S. Nakmura, T. Mukai, M. Senoh, S. I, Nagahama, and N. Iwasa, J. Appl. Phys. 74, 3911 (1993). 127 S. Nakamura, Microelectr. J. 25, 651 (1994). 128 M. Shimizu, K. Hiramatsu, and N. Sawaki, J. Cryst. Growth 145, 209 (1994). 129 Y. Narukawa, Y. Kawakami, M. Funato, S. Fujita, S. Fujita, and S. Nakamura, Appl. Phys. Lett. 70, 981 (1997). 130 E. Iliopoulos, A. Georgakilas, E. Dimakis, A. Adikimenakis, K. Tsagraaki, M. Androulidaki, and N. T. Pelekanos, Phys. Stat. Sol. A 203, 102 (2006). 131 L. K. Teles, L. G. Ferreira, J. R. Leite, L. M. R. Scolfaro, A. Kharchenko, O. Husberg, D. J. As, D. Schikora, and K. Lischka, Appl. Phys. Lett. 82, 4274 (2003). 132 Y. Kangawa, K. Kakimoto, T. Ito, and A. K oukitu, J. Cryst. Growth (In press, corrected proof, 2006). 133 F. B. Naranjo, M. A. Sanchez-Garcia, F. Ca lle, E. Calleja, B. Jenichen, and K. H. Ploog, Appl. Phys. Lett. 80, 231 (2002). 134 D. Cherns, S. J. Henley, and F. A. Ponce, Appl. Phys. Lett. 78, 2691 (2001). 135 X. M. Cai, Y. H. Leung, K. Y. Cheung, K. H. Tam, A. B. Djurisic, M. H. Xie, H. Y. Chen, and S. Gwo, Nanotechnology 17, 2330 (2006). 136 P. Chen, S. J. Chua, Y. D. Wang, M. D. Sander, and C. G. Fonstad, Appl. Phys. Lett. 87, 143111 (2005). 137 A. Chen, S. J. Chua, P. Chen, X. Y. Chen, L. K. Jian, Nanotechnology 17, 3903 (2006). 138 H. M. Kim, W. C. Lee, T. W. Kang, K. S. Chung, C. S. Yoon, and C. K. Kim, Chem. Phys. Lett. 380, 181 (2003). 139 H. M. Kim, T. W. Kang, and K. S. Chung, J. Ceram. Process. Res. 5, 241 (2004). 140 H.J. Moller, Semiconductors for Solar Cells (Artech House, Boston, MA, 1993). 141 J. Nelson, The Physics of Solar Cells (Imperial College Press, London, England, 2003). 142 D. Anderson, Clean Electricity from Photovoltaics edited by M. D. Archer and R. D. Hill (Imperial College Press, London, England, 2001). 143 A. Jager-Waldau and J. Non-Cryst. Solids 352, 1922 (2006).

PAGE 224

224 144 K. L. Shum and C. Watanabe, Energ. Policy 35, 1186 (2007). 145 A. Goetzberger, C. Hebling, and H.W. Schock, Mater. Sci. Eng. R 40, 1 (2003). 146 S. Kurtz, D. Friedman, J. Geisz, and W. McMahon, J. Cryst. Growth (In press corrected proof, 2006). 147 A. Luque and S. Hegedus, Handbook of Photovoltiaic Science and Engineering (John Wiley & Sons, Hoboken, NJ, 2003). 148 S. M. Sze, Modern Semiconductor Device Physics (John Wiley & Sons, New York, 1998). 149 A. Yamamoto, M. Tsujino, M. Ohkubo, and A. Hashimoto, Sol. Energ. Mat. Sol. C. 35, 53 (1994). 150 V. Y. Malakhov, Sol. Energ. Mat. Sol. C. 76, 637 (2003). 151 H. Hamzaoui, A.S. Bouazzi, and B. Rezig, Sol. Energ. Mat. Sol. C. 87, 595 (2005). 152 C.H. Henry, J. Appl. Phys. 51, 4494 (1980). 153 H. Neff, O. K. Semchinova, A. M. N. Lima, A. Filimonov, and G. Holzhueter, Sol. Energ. Mat. Sol. C. 90, 982 (2006). 154 T. Yamaguchi, C. Morioka, K. Mizuo, M. Hori T. Araki, and Y. Nanishi, International Symposium on Compound Semic onductors: Post-Conference Proceedings, San Diego, CA, 2003, p. 214. 155 S. E. Aleksandrov, V. A. Zykov, T. A. Gavrikova, and D. M. Krasovitski Semiconductors 32, 412 (1998). 156 E. Trybus, G. Namkoong, W. He nderson, S. Burnham, W. A. Doolittle, M. Cheung, and A. Cartwright, J. Cryst. Growth 288, 218 (2006). 157 E. Trybus, G. Namkoong, W. He nderson, W. A. Doolittle, R. Liu, J. Mei, F. Ponce, M. Cheung, F. Chen, M. Furis, and A. Cartwright, J. Cryst. Growth 279, 311 (2005). 158 T. D. Veal, I. Mahboob, F. J. Piper, C. F. McCo nville, H. Lu, and W. J. Schaff, J. Vac. Sci. Technol. B 22, 2175 (2004). 159 I. Mahboob, T. D. Veal, C .F. McConville, H. Lu, and W. J. Schaff, Phys. Rev. Lett. 92, 036804 (2004). 160 K. A. Rickert, A. B. Ellis, F. J. Himpsel, H. Lu, W. Schaff, J. M. Redwing, F. Dwikusuma, and T. F. Kuech, Appl. Phys. Lett. 82, 3254 (2003).

PAGE 225

225 161 S. X. Li, K. M. Yu, J. Wu, R. E. Jones, W. Walukiewicz, J. W. Ager III, W. Shan, E. E. Haller, H. Lu, and W. J. Schaff, Phys. Rev. B 71 161201 (2005). 162 O. Jani, C. Honsberg, A. Asghar, D Nicol, I. Fergurson, A. Doolittle, and S. Kurtz, 31st IEEE Photovoltaic Specialists Conference, Lake Buena Vista, FL, 2005, p. 37. 163 R. J. Van Overstraeten and R. P. Mertens, Physics, Technology and Use of Photovoltaics (Adam Hilger Ltd, Accord, MA, 1986). 164 E. Pickwell and V. P. Wallace, J. Phys. D 39, R301 (2006). 165 D. Dragoman and M. Dragoman, Prog. Quant. Electron. 28, 1 (2004). 166 A. G. Davies and E. H. Linfield, Phys. World 14 37 (2004). 167 D. Truchinovich, M. Walther, H. Helm, M. Koch, S. Lahmann, A. Hangleiter, and P.U. Jepsen, 10th International IEEE Conference on Terahert z Electronics Proceed ings, San Diego, CA, 2002, p. 63. 168 R. Kohler, A. Tredicucci, F. Belt ram, H. E. Beere, E. H. Linfield, A. G. Davies, D. A. Richie, R. C. Lotti, F. Rossi, Nature 417, 156 (2002). 169 C. G. Rodrigues, V. N. Freire, A. R. Vasconcellos, R. Luzzi, Appl. Phys. Lett. 74, 1893 (2000). 170 S. K. OLeary, B. E. Foutz, M. S. Shur, U. V. Bhapkar, and L. F. Eastman, J. Appl. Phys. 83, 826 (1998). 171 K. T. Tsen, C. Poweleit, D. K. Ferry, H. Lu, and W. J. Schaff, Appl. Phys. Lett. 86, 222103 (2005). 172 Y. M. Meziani, B. Maleyre, M. L. Sadowski, S. Ruffenach, O. Briot, and W. Knap, Phys. Stat. Sol. A 202, 590 (2005). 173 K. H. Lin, G. W. Chern, Y. K. Huang, and C. K. Sun, Phys. Rev. B 70, 073307 (2004). 174 C. J. Stanton, G. D. Sanders, R. Liu, G. W. Chern, C. K. Sun, J. S. Yahng, Y. D. Jho, J. Y. Sohn, E. Oh, and D. S. Kim, Superlatt. Microstruc. 34, 525 (2003). 175 D. Turchinovich, P. U. Jepsen, B. S. Monozon, M. Koch, S. Lahmann, U. Rossow, and A. Hangleiter, Phys. Rev. B 68, 241307(R) (2003). 176 E. Starikov, P. Shiktorov, V. Gruzinskis, L. Reggiani, L. Varani, and J. C. Vaissiere, 18th International Conference on Noise and Fluctu ations, Salamanca, Spain, 2005, p. 791.

PAGE 226

226 177 E. Starikov, V. Gruzinskis, and P. Shiktorov, Phys. Stat. Sol. A 190, 287 (2002). 178 P. Shiktorov, E. Starikov, V. Gruzinskis, S. Perez, T. Gonzales, L. Reggiani, L. Varani, and J. C. Vaissiere, IEEE Trans. Electr. Dev. 50, 1171 (2003). 179 S. K. OLeary, B. E. Foutz, M. S. S hur, and L. F. Eastman, Appl. Phys. Lett. 88, 152113 (2006). 180 V. Gruzinskis, P. Shiktorov, E. Starikov, L. Reggiani, L. Varani, and J. C. Vaissiere, Semicond. Sci. Technol. 19, S173 (2004). 181 E. Starikov, P. Shiktorov, V. Gruzinskis, L. Re ggiani, L. Varani, J. C. Vaissiere, and J. H. Zhao, Physica B 314, 171 (2002). 182 E. Starikov, P. Shiktorov, V. Gruzinskis, L. Re ggiani, L. Varani, J. C. Vaissiere, and J. H. Zhao, J. Phys. C 13, 7159 (2001). 183 V. V. Korotyeyev, V. A. Kochelap, K. W. Kim, and D. L. Woolard, Appl. Phys. Lett. 82, 2643 (2003). 184 R. Brazis, R. Raguotis, P. Moreau, and M. R. Siegrist, Int. J. Infrared Milli. 21, 593 (2000). 185 J. N. Heyman, P. Neocleous, D. Hebert, P. A. Crowell, T. Muller, and K. Unterrainer, Phys. Rev. B 64, 085202 (2001). 186 W. Knap, J. Lusakowski, F. Teppe, N. Dyakonova and A. El Fatimy, IEICE Trans. Electron. E89 926 (2006). 187 F. Teppe, V. Y. Dachorovski, A. P. Dmitriev, X. Xie, X. C. Zhang, S. Rumyantsev, W. Knap, and M. S. Shur, Appl. Phys. Lett. 87, 022102 (2005). 188 F. Teppe, W. Knap, D. Veksler, M. S. Shur, A. P. Dmitriev, V. Y. Kachorovskii, and S. Rumyantsev, Appl. Phys. Lett. 87, 052107 (2005). 189 H. Morkoc, S.Strite, G. B. Gao, M. E. Lin, B. Sverdlov, and M. Burns, J. Appl. Phys. 76, 1363 (1994). 190 K. S. A. Butcher, Advanced Materials in Electronics edited by Q. Guo (Research Signpost, Trivandrum, 2004). 191 V. Cimalla, U. Daiser, I. Cimalla, G. Ecke, J. Pezoldt, L. Spiess, O. Ambacher, H. Lu, and W. Schaff, Superlatt. Microstruc. 36, 487 (2004). 192 D. C. Look and J. R. Sizelove, Appl. Phys. Lett. 79, 1133 (2001).

PAGE 227

227 193 U. Kaufmann, P. Schlotter, H. Obloh, K. Kohler, and M. Maier, Phys. Rev. B 62, 10867 (2000). 194 R. E. Jones, S. X. Li, L. Hsu, K. M. Yu, W. Walukiewicz, Z. Liliental-Weber, J. W. Ager III, E. E. Haller, H. Lu, and W. J. Schaff, Physica B 376-377, 436 (2006). 195 D. Doppalapudi and T. D. Moustakas, Handbook of Thin Film Materials edited by H. S. Nalwa (Academic Press, San Diego, 2002), p. 57. 196 S. Strite and H. Morkoc, J. Vac. Sci. Technol. B 10, 1237 (1992). 197 R. F. Pierret, Semiconductor Device Fundamentals (Addison-Wesley, Boston, 1996). 198 I. Vurgaftman and J. R. Meyer, J. Appl. Phys. 94, 3675 (2003). 199 S. K. Pugh, D. J. Dugdale, S. Brand, R. A. Abram, Semicond. Sci. Technol. 14, 23 (1999) 200 D. K. Gaskill, L. B. Rowland, and K. Doverspike, Properties of Group III Nitrides edited by J. Edgar (EMIS Data reviews Series No. 11, Stevenage, 1995). 201 T. Inushima, V. V. Mamutin, V. A. Vekshin, S. V. Ivanov, T. Sakon, M. Motokowa, and S. Ohoya, J. Cryst. Growth 227-228, 481 (2001). 202 Z. L. Xie, R. Zhang, B. Liu, L. Li, C. X. Liu, X. Q. Xiu, H. Zhao, P. Han, S. L. Gu, Y. Shi, and Y. D. Zheng, J. Cryst. Growth 298, 409 (2007). 203 Tae Woong Kims doctoral dissertation. A dvisor: Dr. Tim Anderson, Department of Chemical Engineering, Univ ersity of Florida (2006) 204 J. F. Muth, J. H. Lee, I. K. Shmagin, R. M. Kolbas, H. C. Casey Jr., B. P. Keller, U. K. Mishra, U. K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 71, 2572 (1997). 205 R.N. Hall, Phys. Rev. 87, 387 (1952). 206 W. Shockley and W. T. Read, Phys. Rev. 87, 835 (1952). 207 J. Singh, Electronic and Optoel ectronic Properties of Semic onductor Structures (University Press, Cambridge, 2003). 208 E. Yablonovitch and G. D. Cody, IEEE Trans. Electr. Dev. 29, 300 (1982). 209 D. A. Cluston and P. A. Basore, Prog. Photovoltaics 5, 229 (1997). 210 P. R. Berger, S. N. G. Chu, R. A. Logan, E. Byrne, D. Coblentz, J. Lee III, N. T. Ha, and N. K. Dutta, J. Appl. Phys. 73, 4095 (1993).

PAGE 228

228 211 M. Higashiwaki, T. Inushima, and T. Matsui, Phys. Stat. Sol. B 240, 417 (2003). 212 J. S. Jang, S. J. Park, and T. Y. Seong, Phys. Stat. Sol. A 194, 576 (2002). 213 Woo Kyoung Kims doctoral dissertation. A dvisor: Dr. Tim Anderson, Department of Chemical Engineering, Univ ersity of Florida (2006). 214 N. G. Dhere and R. G. Dhere, J. Vac. Sci. Technol. A 23, 1208 (2005) 215 J. Shah, Sol. St. Electr. 21, 43 (1978). 216 C. V. Shank, R. L. Fork, R. F. Leheny, and J. Shah, Phys. Rev. Lett. 42, 112 (1979). 217 H. Ye, G. W. Wicks, P. M. Fauchet, Appl. Phys. Lett. 74, 711 (1999). 218 J. Wu and W. Walukiewicz, Superlatt. Microstruc. 34 63 (2003). 219 V. Bougrov, Properties of Advanced Semiconductor Ma terials: GaN, AlN, InN, BN, SiC, SiGe edited by M. E. Levinshtein, S. L. Rumy antsev, and M. S. Shur (John Wiley & Sons, New York, 2001). 220 B. Monemar, Fundamental Energy Gap of GaN from Photoluminescence Excitation Spectra Phys. Rev. B 10, 676 (1974). 221 R. T. Ross and A. J. Nozik, J. Appl. Phys. 53, 3813 (1982). 222 P. Wurfel, Sol. Energ. Mat. Sol. C. 46, 43 (1997). 223 W. Walukiewicz, K. M. Yu, S. X. Li, R. E. Jone s, J. W. Ager III, E. E. Haller, H. Lu, and W. J. Schaff, Presentation abstract from E-MRS Fall Meeting, (M aterials Research Society, Warsaw 2005). 224 M. Levinshtein, S. Rumyantsev, and M. Shur (eds.), Handbook Series on Semiconductor Parameters (Vol. 1 & 2), (World Sc ientific Publishing Londen, 1999). 225 D. C. Look, C. E. Stutz, R. J. Molnar, K. Saarinen, Z. Liliental-Weber, Sol. St. Comm. 117, 571 (2001). 226 S. Franssila, Introduction to Microfabrication (John Wiley & Sons, Hoboken, NJ, 2004). 227 T. D. Steiner, Semiconductor Materials and Devices Library (Artech House Inc., Boston, MA, 2004). 228 J. I. Pankove and T. D. Moustaka (eds.) Gallium Nitride (GaN) I: Semiconductors and Semimetals 50 (Academic Press, San Diego, CA, 1998).

PAGE 229

229 229 A. Koukitu, N. Takahashi, and H. Seki, Jpn. J. Appl. Phys. 36, L1136 (1997). 230 M. G. Jacko and S. J. W. Price, Can. J. Chem. 42, 1198 (1964). 231 S. P. DenBaars, B. Y. Maa, P. D. Dapkus, A. D. Danner, and H. C. Lee, J. Cryst. Growth 77, 188 (1986). 232 Jang Yeon Hwangs doctoral dissertation. Advi sor: Dr. Tim Anderson, Department of Chemical Engineering, Univ ersity of Florida (2004). 233 V.S. Ban, J. Electrochem. Soc. 119, 761 (1972). 234 Sang Won Kangs doctoral dissert ation. Advisor: Dr. Tim Anderson, Department of Chemical Engineering, Univ ersity of Florida (2004). 235 A. Yamamoto, Y. Yamauchi, M. Ohkubo, A. Ha shimoto, and T. Saitoh, Sol. St. Electr. 41, 149 (1997). 236 X. S. Wang, G. Zhai, J. Yang, and N. Cue, Phys. Rev. B 60, R2146 (1999). 237 O. Kryliouk, T. Anderson, and M. Mastro U.S. Patent 6906351 (Issued June 14, 2005). 238 Mike Mastros doctoral disserta tion. Advisor: Dr. Tim Anders on, Department of Chemical Engineering, University of Florida (2001). 239 A. P. Sutton and R. W. Balluffi, Interfaces in Crystalline Materials (Clarendon Press, New York, 1995). 240 T. Yodo, Y. Kitayama, K. Miyaki, H. Yona, Y. Ha rada, K. E. Prince, and K. S. A. Butcher, J. Cryst. Growth 269, 145 (2004). 241 A. Yamamoto, M. Adachi, and A. Hashimoto, J. Cryst. Growth 230, 351 (2001). 242 M. Mesrine, N. Grandjean, and J. Massies, Appl. Phys. Lett. 72, 350 (1997). 243 R. A. Talalaev, E. V. Yakovlev, S. Yu. Kar pov, and Yu. N. Makarov, J. Cryst. Growth 230, 232 (2001). 244 H. Komaki, T. Nakamura, R. Katayama, K. Onabe, M. Ozeki, and T. Ikari, J. Cryst. Growth (In press, corrected proof, 2007). 245 T. Matsuoka, J. Cryst. Growth 189/190, 19 (1998). 246 M. Adachi, Y. Murakami, A. Hashimoto, and A. Yamamoto, Proceedings of the International Workshop on Nitride Semiconducto rs (IPAP Conference Series) 1, 339 (2000).

PAGE 230

230 247 H. Wang, Y. Huang, Q. Sun, J. Chen, J. J. Z hu, L. L. Wang, Y. T. Wang, H. Yang, M. F. Wu, Y. H. Qu, and D. S. Jiang, Mat. Lett. 61, 516 (2007). 248 J. C. Lin, Y. K. Su, S. J. Chang, W. H. Lan, W. R. Chen, Y. C. Cheng, W. J. Lin, Y. C. Tzeng, H. Y. Shin, C. M. Chang, Opt. Mat. (In Press, corrected proof, 2007). 249 S. Keller, I. Ben-Yaacov, S. P. DenBaars, and U. K. Mishra, Proceedings of the International Workshop on Nitride Semiconductors, Tokyo, Japan, 2000, p. 233. 250 W. K. Chen, Y. C. Pan, H. C. Lin, J. Ou, W. H. Chen, M. C. Lee, Jpn. J. Appl. Phys. 36, L1625 (1997). 251 K. Hoshino, N. Yanagita, M. Araki, and K. Tadatomo, J. Cryst. Growth 298, 232 (2007). 252 L. K. Teles, J. Furthmuller, L. M. R. Scolfa ro, J. R. Leite, and F. Bechstedt, Phys. Rev. B 62, 2475 (2000). 253 J. Zheng and J. Kang, Mat. Sci. Semicon. Proc. 9, 341 (2006). 254 V. N. Jmerik, A. M. Mizerov, T. V. Shubi na, M. Yagovkina, V. B. Listoshin, A. A. Sitnikova, S. V. Ivanov, M. H. Kim, M. Koike, and B. J. Kim, J. Cryst. Growth 301-302, 496 (2007). 255 R. S. Wagner and W. C. Ellis, Appl. Phys. Lett. 4, 89 (1964). 256 Y. Wu and P. Yang, J. Am. Chem. Soc.123, 3165 (2001). 257 B. Maleyre, S. Ruffenach, O. Briot, B. Gil, and A. Van der Lee, Superlatt. Microstruc. 36, 517 (2004). 258 H. Chen, A. R. Smith, R. M. Feenstra, D. W. Greve, and J. E. Northrup, MRS Internet J. Nitride Semicond. Res. 4S1, G9.5 (1999). 259 J. E. Northrup and J. Neugebauer, Phys. Rev. B 60, 8473 (1999). 260 H. Chen, R. M. Feenstra, J. E. Northrup, T. Zywietz, J. Neugebauer, and D. W. Greve, J. Vac. Sci. Technol. B 18, 2284 (2000). 261 O. Knacke, O. Kubaschewski, and K. Hesselman (Eds.), Thermochemical Properties of Inorganic Substances (Springer, New York, 1991). 262 H. Murrmann and D. Widmann, IEEE Trans. on Electr. Dev. 16, 1022 (1969). 263 H. H. Berger, Sol. St. Electr. 15, 145 (1972).

PAGE 231

231 264 D. K. Schroder, Semiconductor Materials and Device Characterization (Wiley, New York, 1990). 265 S. E. Mohney, Y. Wang, M. A. Cabassi, K. K. Le w, S. Dey, J. M Redwing, and T. S. Mayer, Sol. St. Electr. 49, 227 (2005). 266 D. Zhang, C. Li, X. Liu, S. Han, T. Tang, and C. Zhou, 3rd IEEE Conference on Nanotechnology San Francisco, CA, 2003, p. 44. 267 L. C. Tien, H. T. Wang, B. S. Kang, F. Ren, P. W. Sadik, D. P. Norton, S. J. Pearton, and J. Lin, Electrochem. Solid St. 8, G230 (2005). 268 Y. Lu, J. Li, J. Han, H. T. Ng, C. Binder, C. Partridge, and M. Meyyappan, Chem. Phys. Lett. 391, 344 (2004). 269 I. Sayago, E. Terrado, E. Lafuente, M. C. Horrillo, W. K. Maser, A. M. Benito, R. Navarro, E. P. Urriolabeitia, M. T. Martinez, and J. Gutierrez, Synthetic Met. 148, 15 (2005). 270 O. Kryliouk, M. Reed, T. Dann, T. Ande rson, and B. Chai, Mat. Sci. Eng. B 66, 26 (1999). 271 O. Kryliouk, H. J. Park, H. T. Wang, B. S. Kang, T. J. Anderson, F. Ren, and S. Pearton, J. Vac. Sci. Technol. B 23, 1891 (2005). 272 Hyun Jong Parks doctoral dissertation. A dvisor: Dr. Tim Anderson, Department of Chemical Engineering, Univ ersity of Florida (2004). 273 Z. Liliental-Weber, H. Lu, W. J. Schaff, O. Kr yliouk, H. J. Park, J. Mangum, T. Anderson, Phys. Stat. Sol. B (To be published, 2007). 274 X. Duan, and C. M. Lieber, J. Am. Chem. Soc. 122, 188 (2000). 275 Y. Huang, X. Duane, Y. Cui, L. J. Lauhon, K. H. Kim, and C. M. Lieber, Science 294, 1313 (2001). 276 Y. Huang, X. Duan, Y. Cui, and C. M. Lieber, Nano Lett. 2, 101 (2002). 277 L. J. Lauhon, M. S. Gudiksen, D. Wang, C. M. Lieber, Nature 420, 57 (2002). 278 C. Y. Chang, S. J. Pearton, P. J. Huang, G. C. Chi, H. T. Wang, J. J. Chen, F. Ren, K. H. Chen, and L. C. Chen, Appl. Surf. Sci. 253, 3196 (2007). 279 J. R. Kim, H. M. So, J. W. Park, J. J. Kim, J. Kim, C. J. Lee, and S. C. Lyu, Appl. Phys. Lett. 80, 3548 (2002). 280 C. C. Chen, C. C. Yeh, C. H. Chen, M. Y. Yu, H. L. Liu, J. J. Wu, K. H. Chen, L. C. Chen, J. Y. Peng, and Y. F. Chen, J. Am. Chem. Soc. 123, 2791 (2001).

PAGE 232

232 281 D. K. T. Ng, M. H. Hong, L. S. Yan, Y. Zhou, and G. X. Chen, J. Alloy. Compd. (In Press, corrected proof, 2006). 282 M. Niebelschutz, V. Cimalla, O. Ambacher, T. M achleidt, J. Ristic, and E. Calleja, Physica E (In Press, corrected proof, 2006). 283 M. Yoshimoto, H. Yamamoto, W. Huang, H. Harima, and J. Saraie, Appl. Phys. Lett. 83, 3480 (2003). 284 T. B. Hur, I. J. Lee, H. L. Park, Y. H. Hwang, and H. K. Kim, Sol. Stat. Comm. 130, 397 (2004).

PAGE 233

233 BIOGRAPHICAL SKETCH Josh Mangum was born in Greensboro, North Carolina on May 13th, 1981. He is the son of William and Cassandra Mangum and has an olde r brother, Matthew. In 1999, he graduated from Walter Hines Page High School, where his intere st in chemistry first developed with the aid of a wonderful teacher, Mrs. Peggy Stevens. After High School, Josh attended North Carolina State University in Raleigh, NC to major in chemical engineering a nd he graduated with a bachelor of science in May 2003. Immediatel y after receiving his undergraduate degree in chemical engineering Josh moved to Gainesville, FL to pursue a PhD in chemical engineering at the University of Florida. Upon joining Dr. Tim Andersons research group, his research interests mainly involved crystal growth of nitride semiconductors (specifically InN, InxGa1-xN, and GaN). While at UF he was lucky enough to move in across the street from a beautiful graduate student, Kimberly Gray, who later became his fiance, and the two were married shortly after the completion of Joshs Ph.D.