Citation
Investigation of Antiphase Domain Boundary Energetics in GaAs-on-Si(001)

Material Information

Title:
Investigation of Antiphase Domain Boundary Energetics in GaAs-on-Si(001)
Creator:
Barrett, Caleb S
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (170 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Materials Science and Engineering
Committee Chair:
JONES,KEVIN S
Committee Co-Chair:
PEARTON,STEPHEN J
Committee Members:
PERRY,SCOTT S
LAW,MARK E
REN,FAN

Subjects

Subjects / Keywords:
antiphase -- defects -- epitaxy -- gaas -- semiconductors
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Materials Science and Engineering thesis, Ph.D.

Notes

Abstract:
GaAs was once called the material of the future for the semiconductor industry. Significant effort was directed towards the integration of high-performance GaAs on large Si wafers, but GaAs-on-Si devices never became economically feasible compared to Si devices. With the recent push for researching alternative transistor channel materials to replace Si, interest has been renewed in GaAs and the other III-V compound semiconductors. However, many of the same challenges to integrate GaAs on Si still remain. In particular, during the epitaxial growth of GaAs on Si, antiphase domain boundaries (APBs) can nucleate at the GaAs/Si interface and propagate throughout the entire GaAs layer. Much effort has been placed towards achieving APB-free GaAs through various avenues, e.g. substrate preparation and changing growth conditions. Still little is known about how different factors affect the energetic stability of APBs in GaAs. This work investigates these factors at different stages of processing of GaAs films grown on Si(001) by metal-organic chemical vapor deposition (MOCVD). First, it is shown that APB density is quantitatively correlated to the presence of contamination at the GaAs/Si interface. Second, it is demonstrated that the MOCVD growth temperature directly affects the tendency of APBs to change habit planes and self-annihilate. Third, post-growth annealing is shown to drive the dissolution of grown-in APBs. Finally, Fermi level-induced modulations in antisite defect formation energies are investigated as a method of affecting the APB stability. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2017.
Local:
Adviser: JONES,KEVIN S.
Local:
Co-adviser: PEARTON,STEPHEN J.
Statement of Responsibility:
by Caleb S Barrett.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
LD1780 2017 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

INVESTIGATION OF ANTIPHASE DOMAIN BOUNDARY ENERGETICS IN GAAS ON SI(001) By CALEB SHUAN CHIA BARRETT 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 2017

PAGE 2

2017 Caleb Shuan Chia Barrett

PAGE 3

To all those who came before

PAGE 4

4 ACKNOWLEDGMENTS Nobody is truly self made. Support and mentorship are integral in attaining success. This is especially the case in the pursuit of a Ph D. Th ere are many people who deserve to be acknowledged for helping me get to this point. motivation for me to pursue my Ph mother always supported me in everything I did. My brother set a great e xample for me And of course, I have to thank all of my friends that I grew up with that have left positive impacts on my life. I would like to thank my Ph D advisor Kevin Jones for giving me this opportunity and being a great source of guidance and information. The staff at NRF have also been immeasurably helpful. Thank you to Andres and Brent specifically for taking time out of your day for everything from assisting me wi th an ALD recipe to asking how my project was goin g and offering advice Also, thanks to Nick Rudawski for teaching me how to operate a FIB and TEM and giving me tips for sample preparation. from the past and present Special thanks to Aaron for being a great mentor to me when I firs t got to Gainesville and offering me a place to live. Ethan, Henry, Ryan, Tom, Zach, Will, Kathy, David, Emily, Gibson, Amalie, Van; thanks for helping to make our lab group one of the

PAGE 5

5 TA BLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRA CT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 1.1 Motivation ................................ ................................ ................................ ......... 17 1.2 The Role of GaAs in Future Transistors ................................ ............................ 19 1.3 History of GaAs Epitaxy on Si ................................ ................................ ........... 20 1.3.1 Two Step Growth Process ................................ ................................ ....... 21 1.3.2 Substrate Surface Preparation ................................ ................................ 23 1.3.3 Substrate Orientation ................................ ................................ ............... 24 1.3.4 Growth Related Defects ................................ ................................ .......... 24 1.4 Antiphase Boundaries in GaAs on Si ................................ ................................ 25 1.4.1 GaAs Crystal Structure ................................ ................................ ............ 25 1.4.2 Formation of Antiphase Boundaries ................................ ........................ 26 1.4.3 Suppression of Antiphase Boundaries ................................ ..................... 28 1.5 Antiphase Boundary Energetics ................................ ................................ ........ 31 1.5.1 Calculations of Formation Energies ................................ ......................... 3 1 1.5.2 Fermi Level Effects ................................ ................................ .................. 34 1.5.3 Kinetics of Transformation ................................ ................................ ....... 35 1.6 Thermal Effects on III V Antiphase Boundaries ................................ ................ 36 1.6.1 Growth Temperature Effects ................................ ................................ ... 37 1.6.2 Annealing Effects ................................ ................................ .................... 38 1.7 Summary and Statement of Research Goals ................................ .................... 39 2 EXPERIMENTAL METHODS ................................ ................................ ................. 45 2.1 Ma terials Processing ................................ ................................ ......................... 45 2.1.1 Substrate Preparation ................................ ................................ .............. 45 2.1.2 GaAs Film Growth ................................ ................................ ................... 46 2.1.3 Thermal Processing ................................ ................................ ................. 47 2.1. 3.1 Prevention of surface degradation ................................ ................. 48 2.1.3.2 Furnace anneals ................................ ................................ ............ 49 2.1.3.3 Rapid thermal anneals ................................ ................................ ... 50 2.2 Sample Characterization ................................ ................................ ................... 50

PAGE 6

6 2.2.1 Transmission Electron Microscopy ................................ .......................... 50 2.2.1.1 Identification of antiphase boundaries ................................ ............ 53 2.2.1.2 Preparation of samples ................................ ................................ .. 54 2.2.2 Chemical Etching and Scanning Electron Microscopy ............................. 55 2.2.3 X Ray Diffraction ................................ ................................ ..................... 58 2.2.4 Ellipsometry ................................ ................................ ............................. 59 2.2.5 Seco ndary Ion Mass Spectrometry ................................ .......................... 60 2.2.6 Hall Effect ................................ ................................ ................................ 60 3 THE EFFECT OF INTERFACIAL CONTAMINATION ................................ ............. 68 3.1 Background ................................ ................................ ................................ ....... 68 3.2 Experiment al ................................ ................................ ................................ ..... 69 3.3 Results and Discussion ................................ ................................ ..................... 71 3.3.1 SIMS Measurements ................................ ................................ ............... 71 3.3.2 APB Density Measurements ................................ ................................ .... 73 3.3.3 XTEM Characterization of Interface ................................ ......................... 74 3.3.4 Discussion of Source of Particles ................................ ............................ 75 3.3.5 Correlation or Causation? ................................ ................................ ........ 76 3.4 Summary ................................ ................................ ................................ .......... 78 4 THE EFFECT OF GROWTH TEMPERATURE ................................ ...................... 86 4.1 Background ................................ ................................ ................................ ....... 86 4.2 Experimental ................................ ................................ ................................ ..... 88 4.3 Results ................................ ................................ ................................ .............. 90 4.3.1 APB Density Depth Profiles ................................ ................................ ..... 90 4.3.2 APB Annihilation Rates ................................ ................................ ........... 91 4.3.3 XTEM Analysis ................................ ................................ ........................ 93 4.4 Discussion ................................ ................................ ................................ ........ 94 4.4.1 Driving Force for Annihilation ................................ ................................ .. 94 4.4.2 Discussion of Mechanism ................................ ................................ ........ 95 4.4.3 Comparison with Other Samples ................................ ............................. 96 4.5 Summary ................................ ................................ ................................ .......... 97 5 THE EFFECT OF POST GROWTH ANNEALING ................................ ................ 105 5.1 Dissolution of Antiphase Domains by Annealing ................................ ............. 105 5.1.1 Background ................................ ................................ ........................... 105 5.1.2 Experimental ................................ ................................ ......................... 107 5.1.3 Re sults and Discussion ................................ ................................ ......... 109 5.1.3 Summary ................................ ................................ ............................... 113 5.2 Effect of Sample Conditions ................................ ................................ ............ 113 5.2.1 Background ................................ ................................ ........................... 113 5.2.2 Expe rimental ................................ ................................ ......................... 114 5.2.3 Results ................................ ................................ ................................ .. 114 5.2.3.1 High density samples ................................ ................................ ... 114

PAGE 7

7 5.2.3.2 Low temperature bulk growth sample ................................ .......... 116 5.2.4 Discussion ................................ ................................ ............................. 118 5.2.5 Summary ................................ ................................ ............................... 119 5.3 Effect of Annealing Conditions ................................ ................................ ........ 120 5.3.1 Background ................................ ................................ ........................... 120 5.3.2 Expe rimental ................................ ................................ ......................... 120 5.3.3 Results ................................ ................................ ................................ .. 121 5.3.4 Discussion ................................ ................................ ............................. 122 5.3.5 Summary ................................ ................................ ............................... 123 5.4 Effect of APB Density on Si Diffusion ................................ .............................. 124 5.4.1 Background ................................ ................................ ........................... 124 5.4.2 Experimental ................................ ................................ ......................... 124 5.4.3 Results and Discussion ................................ ................................ ......... 125 5.4.4 Summary ................................ ................................ ............................... 127 5.5 Summ ary of Findings for the Chapter ................................ ............................. 127 6 THE EFFECT OF DOPING ................................ ................................ ................... 142 6.1 Background ................................ ................................ ................................ ..... 142 6.2 Experimental ................................ ................................ ................................ ... 143 6.3 Result s ................................ ................................ ................................ ............ 144 6.4 Discussion ................................ ................................ ................................ ...... 145 6.5 Summary ................................ ................................ ................................ ........ 146 7 CONCLUSIONS AND FUTURE WORK ................................ ............................... 148 7.1 Major Findings ................................ ................................ ................................ 148 7.2 Future Work ................................ ................................ ................................ .... 150 A PPENDIX: SUPPLEMENTARY INFORMATION ................................ ..................... 152 A.1 Particle Distribution Calculation ................................ ................................ ...... 152 A.2 Hybrid Al 2 O 3 ALD Recipe ................................ ................................ ............... 153 LIST OF REFERENCES ................................ ................................ ............................. 156 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 170

PAGE 8

8 LIST OF TABLES Table page 1 1 GaAs APB formation energy on different habit planes as determined by DFT calculations. Values from Rubel and Baranovskii [92]. ................................ ....... 41

PAGE 9

9 LIST OF FIGURES Figure page 1 1 GaAs unit cell along various viewing directions. ................................ ................. 41 1 2 Plan view SEM image of GaAs layer grown on Si(001) with high APB density. 42 1 3 Schematic drawing of dimer bonds on the Si(001) surface showing the rotation of dimers by 90 across single steps. ................................ .................... 42 1 4 <110> projection of an epitaxial GaAs layer on a Si(001) substrate with APBs. ................................ ................................ ................................ ................. 43 1 5 <110> projection of a Si(001) surface showing the two types of terraces that will exist. ................................ ................................ ................................ ............. 43 1 6 Composite XTEM image showing t he meandering path of an APB through a GaAs on Si(001) layer. ................................ ................................ ....................... 44 1 7 Arbitrary Gibbs free energy diagram for a GaAs on Si syste m with two different states: a GaAs layer with APBs propagati ng vertically through the film and a GaAs layer with self annihilated APBs suppressed in the film. .......... 44 2 1 Simplified schematic drawing of MOCVD chamber. ................................ ........... 63 2 2 Simulated diffraction pattern for GaAs along the [110] zone axis. ...................... 64 2 3 DF XTEM image of a GaAs layer on Si(001) made using the g 002 Bragg reflection showing t he contra st reversal across an APB ................................ .... 64 2 4 Procedure for Fourier filtering a HAADF STEM image of a GaAs on Si film. ..... 65 2 5 BF XTEM images of a GaAs on Si(001) film before and after etching with HF/HNO 3 solution. ................................ ................................ .............................. 66 2 6 HRTEM image of a developed etch pit at the site of an APB in a GaAs on Si layer demonstrating the selectivity of the HF/HNO 3 etchant for APBs. ............... 66 2 7 Example of tracing procedure of APBs in plan view SEM images. ..................... 67 2 8 BF XTEM image of a GaAs on Si sample after using the non selective H 3 PO 4 /H 2 O 2 etchant ................................ ................................ .......................... 67 3 1 Representative plan view SEM images of GaAs film grown on Si(001) with varying pre growth conditions. ................................ ................................ ............ 80 3 2 Illustration showing the variation in pre cleaning procedure to affect the amount of residual contamination across the Si(001) wafer. .............................. 80

PAGE 10

10 3 3 Diag ram of GaAs layer structure for surface contamination experiment. ............ 81 3 4 BF XTEM images of GaAs layer at different wafer positions. ............................. 81 3 5 SIMS profiles of contamination sources at the GaAs/Si interface with varying wafer position. ................................ ................................ ................................ .... 82 3 6 I ntegrated i nterfacial oxygen and carbon doses as a function of wafer position. ................................ ................................ ................................ .............. 82 3 7 Representative plan view SEM images of the APB density in the GaAs film at different wafer positions. ................................ ................................ ..................... 8 3 3 8 Plot of all APB density counts as a function of wafer position. ........................... 83 3 9 HRTEM image of GaAs/Si interface of a sample from the wafer edge showing the spatial distribution of amorphous interfacial particles .................... 84 3 10 HRTEM images at higher magnification of a series of amorphous particles observed at the GaAs/Si interface. ................................ ................................ ..... 84 3 11 STEM images of the GaAs/Si interface in an edge sample showing an APB associated with an amorphous particle. ................................ .............................. 85 3 12 Plot of APB density as a function of interfacial carbon dose for measured samples. ................................ ................................ ................................ ............. 85 4 1 Illustration of APBs oriented along various habit planes in a GaAs on Si film for a <110> projection ................................ ................................ ........................ 99 4 2 Illustration of GaAs layer structure for the growth temperature experiment. ....... 99 4 3 Representative plan view SEM images of the APB density in the sample with bulk growth at 570 C at varying film thickness points: ................................ ..... 100 4 4 Depth profiles of APB density for all samples with varying bulk growth temperature. ................................ ................................ ................................ ..... 101 4 5 Example of fitting procedure for APB density depth profile curves for sample with bulk growth at 610 C. ................................ ................................ ............... 101 4 6 Plot of the ext racted APB annihilation rates for all samples versus the bulk growth temperature. ................................ ................................ ......................... 102 4 7 Arrhenius plot of APB annihilation rates for samples with bulk growth from 530 to 610 C. ................................ ................................ ................................ ... 102 4 8 Pair of DF XTEM images showing an antiphase domain in a sample with bulk growth at 630 C. ................................ ................................ .............................. 103

PAGE 11

11 4 9 Pair of DF XTEM images showing an antiphase domain in the sample with bulk growth at 610 C. ................................ ................................ ...................... 103 4 10 Illustration of a n annihilating antiphase domain showing how t he kinking of APBs leads to a net reduction in the number of atoms involved in an APB in the (001) growth plane. ................................ ................................ ..................... 104 4 11 Comparison of APB density depth profiles for samples grown at different times with similar conditions ................................ ................................ ............ 104 5 1 BF XTEM image of of a GaAs on Si sample annealed at 600 C for 30 min with 20 nm of Al 2 O 3 as a capping layer ................................ ............................ 129 5 2 Representative plan view SEM images showing the e volution of APB density and domain size with annealing. ................................ ................................ ....... 129 5 3 Plot of surface APB densities for samples with varying annealing times and temperatures. ................................ ................................ ................................ ... 130 5 4 Depth profiles of APB density for samples as grown and with varying annealing conditions. ................................ ................................ ........................ 130 5 5 Example of the fitting procedure for APB density versus annealing time. ......... 131 5 6 Illustration of possible mechanism for the movement of an APB segment from a { 110} configuration to {111} via the diffusion of a Ga vacancy (V Ga ). ..... 132 5 7 Representative plan view SEM images of APB density in highly defective sample as grown and after annealing. ................................ .............................. 133 5 8 Plot of surface APB densities for high starting APB density sample for various annealing conditions. ................................ ................................ ........... 133 5 9 Representative plan view SEM images of high starting APB density sample after annealing showing rounding of APBs. ................................ ...................... 134 5 10 Representative plan view SEM images of the APB density in a highly defective sample with more discrete domains as grown and after annealing .. 134 5 11 Plot of surface APB densities for sample with high starting density and discrete domains after various annealing treatments. ................................ ...... 135 5 12 Evolution of large and small antiphase domains in high density sample with annealing. ................................ ................................ ................................ ......... 135 5 13 Representative plan view SEM image of APB density in a sample grown entirely at a low temperature of 530 C. ................................ ........................... 136

PAGE 12

12 5 14 APB density depth profiles for the low temperature growth sample as grown and after various annealing treatments. ................................ ........................... 136 5 15 DF XTEM image s of p inched off antiphase domain in the low temperatu re growth sample after annealing. ................................ ................................ ......... 137 5 16 Representative plan view SEM images of the APB density in the low temperature growth sample after annealing with different capping conditions. 137 5 17 Plot of surface APB densities for the low temperature sample after annealing with a proximity cap and Al 2 O 3 cap at 600 C for various times. ....................... 138 5 18 Plot of surface APB densities for the low temperature growth sample compared with a high temperature growth sample (610 C) for similar annealing conditions with a p roximity cap. ................................ ....................... 138 5 19 Representative plan view SEM images showing the grown in nature of APBs as a function of growth temperature. ................................ ................................ 139 5 20 Representative plan view SEM images of the as grown surface APB densities in the samples for the Si diffusion study. ................................ ........... 139 5 21 BF XTEM images of the low and high APB density samples for the Si diffusion s tudy. ................................ ................................ ................................ .. 140 5 22 SIMS profiles of Si concentration in the GaAs layers for samples with different starting APB densities as grown and after annealing. ........................ 140 5 23 Comparison of Si concentration profiles for the low and high APB density samples after annealing at 650 C for 2 hrs. ................................ .................... 141 5 24 (002) and (004) XRD rocking curves of the low and high APB density samples as grown. ................................ ................................ ........................... 141 6 1 Representative plan view SEM samples of the APB density in doped and undoped samples as grown. ................................ ................................ ............ 147 6 2 Plots of APB density as a function of doping conditions and Fermi level for the p type, undoped, and n type samples. ................................ ....................... 147

PAGE 13

13 LIST OF ABBREVIATIONS ABF Annular bright field AFM Atomic force microscopy ALD Atomic layer deposition APB Antiphase domain boundary BF Bright field CMOS Complimentary metal oxide semiconductor DEZn Diethylzinc DF Dark field DFT Density functional theory EDS Energy dispersive spectroscopy EELS Electron energy loss spectroscopy FIB Focused ion beam FWHM Full width at half maximum HAADF High angle annular dark field HRTEM High resolution transmission electron microscopy MBE Molecular beam epitaxy MOCVD Metal organic chemical vapor deposition MOSFET Metal oxide semiconductor field effect transistor nMOS n type metal oxide semiconductor pMOS p type metal oxide semiconductor RTA Rapid thermal anneal SEM Scanning electron microscopy SIMS Secondary ion mass spectrometry STEM Scanning transmission electron microscopy

PAGE 14

14 TBAs Tertiarybutylarsine TEM Transmission electron microscopy TMGa Trimethylgallium XRD X ray diffraction XTEM Cross sectional transmission electron microscopy

PAGE 15

15 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 INVESTIGATION OF ANTIPHASE DOMAIN BOUNDARY ENERGETICS IN GAAS ON SI(001) By Caleb Shuan Chia Barrett August 2017 Chair: Kevin Jones Major: Materials Science and Engineering GaAs was once called the material of the future for the semiconductor industry. Significant effort was directed towards the integration of high performance GaAs on large Si wafers, but GaAs on Si devices never became economically feasible compared to Si devices. With the recent push for researching alt ernative transistor channel materials to replace Si, interest has been renewed in GaAs and the other III V compound semiconductors. However, many of the same challenges to integrate GaAs on Si still remain. In particular, during the epitaxial growth of GaA s on Si, antiphase domain boundaries (APBs) can nucleate at the GaAs/Si interface and propagate throughout the entire GaAs layer. Much effort has been placed towards achieving APB free GaAs through various avenues, e.g. substrate preparation and changing g rowth conditions. Still little is known about how different factors affect the energetic stability of APBs in GaAs. This work investigates these factors at different stages of processing of GaAs films grown on Si(001) by metal organic chemical vapor deposi tion (MOCVD). First, it is shown that APB density is quantitatively correlated to the presence of contamination at the GaAs/Si interface. Second, it is demonstrated that the MOCVD growth temperature directly affects the tendency of APBs to change habit pla nes and

PAGE 16

16 self annihilate. Third, post growth annealing is shown to drive the dissolution of grown in APBs. Finally, Fermi level induced modulations in antisite defect formation energies are investigated as a method of affecting the APB stability

PAGE 17

17 CHAPTER 1 INTRODUCTIO N 1.1 Motivation Technological progress in the microelectronics industry has traditionally been predicted that the number of transistors on a single silicon chip would double every one to two years [1] shrinking the individual transistor siz e. This decrease in size brings a proportional increase in electrical performance, as explained by the Dennard scaling law [2] A key component of the Dennard scaling law is that as transistors are scaled down the power density should also remain constant. However, this rule has broken down in recent years as leakage currents in devices have become a greater concern [3 5] Classical Dennard scaling of Si longer sustainable. In an attempt to extend the life of towards non planar device architectures as well as alternative transistor channel materials to replace Si. The most likely group of materials to replace Si in future transistors are the III V compound semiconductors, so called because they are composed of elements from groups III and V of the periodic table. III V materials such as GaAs, InAs, and their ternary alloy InGaAs have much better electron transport properties than Si, including higher electron mobilities [6] Transistors made from these materials can theoretically obtain much higher drive currents at similar supply voltages relative to Si based devices. However, it is impractical to use III V wafers as the basis for future devices because they are generally more expensive to produce [7] cannot be made as large as

PAGE 18

18 Si wafers [8] and are fragile compared to Si wafers [9] Therefore, research has been devoted to integrating III V materials on Si to combine their respective beneficial properties. Epitaxial growth is a common method of fabricating III V layers on Si wafers. The growth of GaAs on Si specifically has been well studied since the 1980s. Much of the early work was focused on obtaining good quality GaAs on Si to replace more expensive bul k GaAs in photonics applications, with the goal of eventually combining these devices with transistors in integrated circuits [10] However, as with other III Vs, the epitaxial layers continued to be plagued by a variety of growth related defects and the cost and performance of GaAs on Si devices never rivaled that of bulk GaAs. Therefore, work in the area of GaAs growth on Si was largely shelved until the recent push for new transistor materials in microelectronics applications. This renewed interest has sparked a need for research to fill the gaps left previo usly in the understanding of how these different defects can be suppressed or eliminated. One type of defect that is unique to the epitaxy of a polar material such as GaAs on a nonpolar material such as Si is antiphase domain boundaries (APBs). APBs are high energy planar defects composed of antisite bonds and can be treated as a special type of grain boundary separating regions of crystal with reversed polarity. They can propagate through the entire thickness of an epitaxial layer and lead to degradation of electrical properties [11] Therefo re, from an engineering standpoint, it is critical to develop processes that either prevent the nucleation of APBs or achieve complete suppression of antiphase domains within the epitaxial film but that are also practical for implementation in complex indu strial process flows. These breakthroughs require

PAGE 19

19 knowledge of the fundamental characteristics of APBs and how their energetics are affected by different conditions. Even though several decades of research have been spent on GaAs growth on Si, still little is known about the energetics and thermal stability of APBs. Successful integration of GaAs and other III Vs on Si will never occur without a thorough understanding of these factors. 1.2 The Role of GaAs in Future Transistors In its most common usage in modern electronic devices, the transistor functions when it is off. Therefore, the two most important properties that define a transistor are the on state drive curre nt, I on and the off state leakage current, I off A good transistor must have a high ratio of I on to I off in order for the two states to be distinguishable. The breakdown in scaling of Si based devices has arisen largely because I off has gotten very large due to decreases in the gate dielectric thickness and the transistor threshold voltage [4] As well, increasing the supply voltage to increase I on is not feasible due to heat dissipation concerns [6] On paper, the III V compound semiconductors have much more attrac tive properties for use in metal oxide semiconductor field effect transistors (MOSFETs) over Si. For n type MOS (nMOS) devices, the electron mobility at room temperature of InGaAs and InAs is more than 10 times that observed for Si [6] III V compounds may also be used for p type MOS (pMOS), as the hole mobility of materials such as GaSb is greater than that of Si [6, 12] Thus, marked improvements in I on can theoretically be made for III V devices over Si devices even at lower operating voltages. In practice, integrating III Vs on Si for future devices comes with a number of challenges, not least of which is the issue of growth related defects. The lattice constant mismatch in

PAGE 20

20 particular can be quite large. For example, InAs and Si have a lattice mismatch of ~11%. The lattice mismat ch can lead to a high density of threading dislocations that will, in combination with the antiphase disorder, degrade the quality of the layer and adversely affect device performance [13] Buffer layers are commonly implemented in the epitaxial growth of III V compoun ds on Si to alleviate growth related defects [12, 14] The buffer layer material must be cho sen such that it has an intermediate lattice constant between that of the substrate and the subsequent layer. It must also form good interfaces with the substrate material and the next growth layer. GaAs is a good candidate for a buffer layer material. Whi le future transistor devices could be based on GaAs, larger gains in performance would be made by using other III V compounds, such as InAs, instead. In the role of a buffer layer material, GaAs has several beneficial properties. In particular, it has a la ttice constant larger than Si but smaller than InAs, it can make ternary alloys with InAs with variable lattice constant, and there is already a wealth of history of research on the growth of GaAs on Si. Another advantage of using GaAs as a buffer material is its relatively large bandgap (1.42 eV at room temperature) which makes it semi insulating when it is undoped [15] This property provides electrical isolation for the active channel layer in a transistor. Thus, the fi rst step towards realizing III V devices on Si substrates is the growth of GaAs layers on Si with good crystal quality and suppression of defects. 1.3 History of GaAs Epitaxy on Si Since the early reports on the semiconducting nature of III V compounds in the 1950s [16] GaAs has displayed significant promise for device applications. It has many attractive properties relative to Si, particularly for high frequency, high temperature, and high radiation environ ment applications [17] However, the technology for Si in general

PAGE 21

21 integrated circuits applications matured much more rapidly and it became the standard of the semiconductor industry. The cos t of Si devices was also much lower compared to GaAs in part due to the larger wafer size of Si which allowed for higher throughput. It was realized that in order to make GaAs devices economically feasible for large scale manufacturing, GaAs would need to be integrated onto Si wafers [13] Thus, significant effort towards researching GaAs epitaxy on Si began in the 1980s. There are two primary growth methods for GaAs epitaxy on Si. One is molecular beam epitaxy (MBE) and the other is metal organic chemical vap or deposition (MOCVD). Briefly, MBE takes place in ultrahigh vacuum (~10 9 Torr) and evaporates the elemental sources onto a substrate, while MOCVD takes place at higher pressure (~1 Torr to 760 Torr) and uses gas based metal organic precursors that are flowed across t he substrate [18, 19] MBE is considered to be a technique that is further from equilibrium and can be used to form high purity films with good stoichiometric control. However, the ultrahigh vacuum environment of MB E makes it inefficient for large scale production compared to MOCVD [19] Therefore, MOCVD has been the dominant method used in industry. 1.3.1 Two Step Growth Process Successf ul growth of GaAs films on Si with good morphology was demonstrated for both MBE [20] and MOCVD [21] systems in 1984. The critical breakthrough in both of these studies was the use of a two step growth process for GaAs to reduce defect densities and achieve good crystal quality. The first step is a thin nucleat ion layer grown at a relatively low temperature, typically 400 600 C [21] The grow th temperature is then raised to the range of conventional GaAs homoepitaxy, 600 700 C, in the second

PAGE 22

22 step for the remainder of the film growth. Attempts to nucleate GaAs on Si at typical GaAs homoepitaxy temperatures resulted in rough films with high def ect densities [22] The advantage of a low temperature nucleation layer is mani fested in the number and size of nuclei that form during the initial stages of growth. There are three generally accepted modes of growth for epitaxial films [23] : two dimensional (2 D) layer by layer growth (Frank van der Merwe mode), three dimensional (3 D) island growth (Volmer Weber mode), and a combination of the two (Strankski Krastanov mode). For highly strained films, such as GaAs on Si, it is difficult to maintain stability of 2 D layer growth. This can be seen in a wetting surface energy inequality relationship. Layer by layer growth is only favorable if the following is true: (1 1) substrate film interface are the surface energies of the substrate, film, and substrate film interface, respectively [23, 24] film which can then reverse t his inequality and thus favor the growth of 3 D islands. The number of nuclei that will form at the initial stages of growth can be related to the degree of undercooling, set by the growth temperature, T [25 ] : (1 2) where r is the critical nucleus radius and T m is the equilibrium solidification temperature of the film. A low temperature presents a large undercooling and thus promotes the formation of many, small nuclei allowing for earlier coalescence of the film [26 28] As well, the islands at a low growth temperature are more likely to still be fully strained, i.e. pseudomorphic, before the film coalesces [22, 28] For a high growth temperature, fewer islands form and can grow quite large and form a high density of dislocations

PAGE 23

23 before they meet and coalesce [22] Misoriented nuclei can also form and lead to a polycrystalline film. Thus, the nucleation temperature should be kept as low as possible. Nucleation of GaAs on Si at temperatures as low as 225 C has been shown for MBE [26] The nucleation temperatur e for GaAs on Si by MOCVD is limited to ~400 C since the temperature at the substrate surface must still be high enough to crack, or pyrolyze, the precursor molecules [29, 30] It should also be noted t hat once the film is nucleated, it no longer is advantageous to grow at a low temperature as this hinders the growth rate and the annihilation of any existing defects [31] 1.3.2 Substrate Surface Preparation The preparation of the Si substrate is also critical in achieving successful growth of GaAs on Si. Contaminants present on the Si surface prior to growth can degrade the quality of the epitaxial GaAs layer and lead to higher defect densities [13] Interfacial amorphous co ntaminant particles have been directly linked to the formation of growth defects [32, 33] and it is speculated that they may also be able to nucleate APBs [13] The primary sources of contamination on the Si surface are oxygen from the native oxide of Si and carbon from the environment. Carbon contamination is typically more difficult to remove than oxygen [34] Most early methods of Si substrate preparation involved some combination of degreasing in organic solvents then dipping in HF to remove the native oxide [13, 21] Other studies found it was advantageous to intentionally oxidize and etch the Si surface in repeated steps to gradually reduce the amount of carbon [35 38] After a number of oxidation and etch cycles the substrates would be transferred to the growth chamber and annealed at 800 1000 C to thermally desorb the remaining oxide prior to film growth. Eventually, it became widely accepted that a wet chemical treatment involving HF followed by a high temper ature anneal near

PAGE 24

24 1000 C in the growth chamber was sufficient in reducing the surface contamination [38, 39] 1.3.3 Substrate Orientation Early work by Kroemer et al. [40 43] proposed the use of (112) oriented Si wafers to reduce the antiphase disorder in GaAs and other III V compound films. The supporting theory was that the pattern of dangling bonds on the (112) Si surf ace would cause a preference for either Ga or As adatoms to stick to specific sites. Therefore, GaAs islands nucleating on different areas of the (112) surface would have the same orientation and not form APBs when they coalesced. However, since Si technol ogy in industry was and still is based on (001) oriented wafers, the (112) solution was not adopted. The focus of this work will also remain on (001) oriented Si for its technological relevancy. 1.3.4 Growth Related Defects As mentioned above, the various strategies for GaAs growth on Si are all aimed at reducing the density of defects in the GaAs layer. The primary growth related defects that typically occur, in order of defect formation energy, are [44, 45] : threading dislocations, stacking faults, twin boundaries, and APBs. Stacking faults and twin boundaries are planar defects that result from errors in the stacking sequence of atomic planes during growth. These errors can usually be avoided by optimizing growth conditions. The nucleation o f threading dislocations (from interfacial misfit dislocations) and APBs is mostly unavoidable for conventional GaAs on Si epitaxy. There are already numerous studies on the reduction of threading dislocation densities in III V layers grown on Si. The main strategies are the use of strained layer superlattices [46 51] to deflect the propagation of threading dislocations away from the film surface and, more

PAGE 25

25 recently, aspect ratio trapping [52 59] to encourage the termination of dislocations and other defects at oxide sidewalls. The elimination of APBs in III V films is more complex and not as well understood. They are also particularly interesting defects b ecause of their large formation energy and uniqueness in semiconductor processing. Literature relating to APB formation and energetics will be discussed herein. 1.4 Antiphase Boundaries in GaAs on Si 1.4.1 GaAs Crystal Structure To understand the formation of APBs in GaAs grown on Si, one must begin by analyzing the crystal structure of GaAs as compared to Si. GaAs has the zinc blende crystal structure and Si has the diamond cubic crystal structure. While differing in name, these crystal structures are very similar. Each atom in the lattice shares four bonds with neighboring atoms. The basic unit cell for GaAs is a face centered cubic (FCC) sublattice of one atom type (Ga or As) with additional atoms that are of the opposing type at the four tetrahedral site s. The four tetrahedral atoms are a part of an equivalent FCC sublattice with the origin at a 0 (, , ), where a 0 is the lattice constant. Figure 1 1A shows a GaAs unit cell based on the Ga sublattice with As atoms at the tetrahedral sites. The diamond cub ic structure is identical in terms of atom positioning, but the two FCC sublattices are composed of the same type of atom. Since the sublattices in GaAs are of opposing atom types, not all configurations are structurally alike. This difference is most nota ble in the <110> projections of the GaA s crystal structure. Figure 1 1B shows a GaAs unit cell along the [110] viewing GaAs with As polarity in the +z direction. If the l attice is rotated by 90 (or, equivalently, inverted) and viewed along t he [ 110] direction (Figure 1 1C ), the direction of bond

PAGE 26

26 polarity is reversed. Therefore, if GaAs in nucleated in regions with orientations offset by 90, the crystal structures will h ave opposing polarity and the Ga and As sublattice positions are switched. For Si homoepitaxy this is a nonissue as the sublattices are identical. Likewise, for GaAs homoepitaxy, Ga and As adatoms will energetically prefer to be incorporated in the correc t sublattice and follow the ordering of the substrate. However, during the heteroepitaxy of GaAs on Si, this ordering preference no longer applies and structural defects can result. It is also important to no te that in comparing Figure 1 1B and Figure 1 1C if the two unit cells are offset by a single atomic layer e quivalent (002) planes will be offset by one quarter of the lattice constant, or one half of the (002) plane spacing in the +z direction. Thus, the (002) plane is sensitive to any dispersio ns in the GaAs crystal ordering arising from such monolayer offsets This factor will become relevant for the detection of APBs in GaAs. 1.4.2 Formation of Antiphase Boundaries The possible formation of APBs in GaAs was first predicted by Holt in 1969 [60] This prediction was based on the characteristic dual sublattices of the GaAs zinc blende structure which allows for regions of crystal with opposite polarity, i.e. are antiphase with respect to the other. Thus, if these two types of domains formed in a b ulk crystal, a planar boundary would form at their intersection. This boundary was termed an antiphase domain boundary, or antiphase boundary for short. The atomic bonds across the boundary would be forced to be antisite bonds, i.e. Ga Ga or As As bonds, d ue to the differing allocation of Ga and As atoms in the sublattices in the two orientations. Antiphase domains were first experimentally observed in GaAs epitaxial layers by Morizane in 1977 [61] for GaAs grown on Ge. The degree of anti phase disorder in

PAGE 27

27 epitaxial GaAs lay ers can be extensive. Figure 1 2 shows a plan view scanning electron microscope (SEM) image of a GaAs layer grown on Si(001) with a large areal density of APBs. APBs in GaAs must form closed loops, i.e. domains, and cann ot intersect in triple points since there are only two sublattices [44] The surface of the substrate is the key to the formation of APBs in epitaxial GaAs layers. The (001) surface of Si is known to exhibit a 2x1 dimerized reconstruction [13] The dimers form between atoms with dangling bonds at the surface in order to lower the surface energy [62, 63] The orientation of the dimer bond on the Si surface will depend on the sublattice that terminates the surface. Thus, two distinct domains can exist on the surface. This difference is evident with the presence of a step on the Si(001) surface that is a single atom in height. These types of steps will be referre d to as single steps. Figure 1 3 shows how the orientation of dimerization is rotated by 90 across single steps. Surface steps will always be present on a real Si(001) surface [10, 43] mainly because of entropy and the fact that slices of Si wafers from the boule are never perfectly flat. It follows that if GaAs growth is nucleated on either side of a single step on the Si surface, the crystal will be rotated by 90 to follow the orientation of the substrate and the respective crystal po larities will be reversed. Therefore, when the two nucleated islands coalesce, an APB will form above the surface step. Figure 1 4 shows a <110> projection of an epitaxial GaAs layer on a Si(001) substrate with the presence of surface steps. An APB is nucl eated at a single step in Figure 1 4 A and propagates vertically along a {110} type plane. The reversal of bond polarity across the APB from As polar to Ga polar can be observed, along with the alternating antisite bonds of Ga Ga and As As

PAGE 28

28 along the APB pla ne. If an APB propagates along a habit plane that is inclined relative to the substrate normal, e.g. {111}, then it may self annihilate with another APB, as seen in Figure 1 4 B The antisite bonds along the {111} habit plane are all the same type; in this case, As As. This type of habit plane is considered to be non stoichiometric and may not be energetically favorable. In the case of a surface step that is two atomic layers in height, i.e. a double step, the sublattice ordering of the epitaxial GaAs layer is reset and no APB forms, as seen in Figure 1 4 C In addition to physical surface steps, it is suspected that the presence of contaminant particles on the Si substrate may act analogously to steps and contribute to the nucleation of APBs [13] 1.4.3 Suppression of Ant iphase Boundaries As depicted in Figure 1 4 C the presence of all double steps on the Si(001) surface would theoretically prevent the nucleation of APBs. Therefore, most of the initial efforts to suppress antiphase disorder in epitaxial GaAs films grown o n Si were based on achieving a perfect step doubling of the Si substrate. For as prepared Si(001) wafers the most common step height is a single atomic layer [64, 65] One approach focused on deliberately misorienting the Si(001) wafer s by off cutting the wafer some angle from the (001) plane which would increase the step density. It was found that these vicinal Si(001) wafers exhibited higher percentages of double steps with increasing annealing temperature (up to 1000 C) in ultra hig h vacuum environments [66 68] The preference for double steps with s ubstrate misorientation was theorized to stem from a reduction in the elastic interaction energy between steps since the terrace distance is reduced with greater misorientation and higher step density [69, 70] The reconstruction of adjacent single steps into a double steps increases the average terrace width and creates a single domain surface.

PAGE 29

29 Other studies found that step doubling could be achieved even for nominally on axis Si(001) wafers, i.e. no off cut, with high temperature annealing. Sakam oto and Hashiguchi [71] first showed a two domain Si(001) surface with single steps reconstructed to a single domain surface via conversion of the single steps to double steps after annealing at 1000 C for 20 min in a MBE system. Later work speculated that this result, and the step doubling result from off cut substrates, was due to an energetic preference for a specific step edge [64, 72] Figure 1 5 shows the two types of terraces that will exist on a Si(001) surface with steps. Type A terraces have dan gling bonds running perpendicular to the step edge and Type B terraces have dangling bonds running parallel to the step edge. Type B terraces and the corresponding step edges were thought to be more energetically favorable [64] and in particular T ype B double steps were theorized to be favored over either type single steps [65, 72] More recent work has corroborated this energetic preference for double height steps with dangling bonds parallel to the step edge [73, 74] Therefore, there is a driving force for a surface made of single steps of either type, i.e. a normal Si(001) surface, to reconstruct to form only Type B double step edges with annealing. Misorien ting the substrate simply reduces the distance that atoms at unfavorable step edges need to diffuse to reach a favorable step edge. The accepted method of Si substrate preparation for optimal GaAs growth and reduced antiphase disorder involves some degree of off cutting followed by a high temperature pre growth anneal near 1000 C [38] The pre growth anneal has the additional benefit of reducing the amount of surface contamination, including native oxide, that may contribute to higher densities of APBs and other defects. Several

PAGE 30

30 studies in the late 1980s were able to demonstrate the growth of APB free GaAs films by MBE and MOCVD using this procedure [75 77] and it continues to be employed for the growth of GaAs and other III V compounds in more recent years [78 80] However, the use of misoriented substrates and high temperature anneals is not necessarily practical for integration into large scale industrial processes. In particular, established process flows for integrated circuit manufacturing are based on nominally on axis Si(001) wafers. Moving to misoriented subs trates would require a complete revamping of these process flows and so is cost prohibitive. High temperature anneals should also ideally be avoided since there may be other existing layers or devices on the Si wafer before the deposition of GaAs that woul d be affected [10] Thus, it is unlikely that the nucl eation of APBs can be completely prevented for GaAs films grown on Si(001) substrates with no intentional off cut. Film thickness is a variable that may help to suppress any nucleated APBs in GaAs on Si. Georgakilas et al. [11] showed through depth profiling of a GaAs layer grown on a misorient ed Si(001) substrate that the APB density decreased with increasing GaAs thickness. Therefore, APBs must be self annihilating during growth. This evidence corresponds with experimental observations in transmission electron microscopy (TEM) that APBs in GaA s have a tendency to change habit planes during growth [81 85] which is conducive for self annihilation. Figure 1 6 is a composite cross sectional TEM (XTEM) image showing the meandering path of an APB propagating through a GaAs layer gr own on Si(001). It is likely that the early studies of GaAs growth on Si that grew good quality films took advantage of this effect, whether intentionally or

PAGE 31

31 not. GaAs on Si film thicknesses for device fabrication were commonly on the order of a few micron s [13, 75] In theory, based on the evidence for APB self annihilation, APBs can be completely suppressed in a film regardless of the nucleated density if growth is continued indefinitely. However, there is an upper limit on GaAs film thick ness due to the difference in the coefficient of thermal expansion with Si. The cool down from the high growth temperature can induce large stresses in the film and lead to cracking, depending on the film thickness [38] For typical GaAs on Si heteroepitaxial growth ranges, the critical thic [86 88] In addition to the potential for film cracking, growing thick GaAs films on the order of microns should be avoided to reduce cost and dwell time in industrial processes. Growing another group IV layer on top of the GaAs layer also will not necessarily reset the atomic ordering and suppress antiphase disorder. Jia and Fitzgerald [89] showed that APBs will propagate through a GaAs film with Ge interlayers. Thus, the misordering induced by single steps is being preserved even through the Ge layers. Self anni hilation of APBs must then be achieved completely in the GaAs film with minimal t hickness. To this end, it becomes important to understand the energetics of APBs and how their propagation can be affected. 1.5 Antiphase Boundary Energetics 1.5.1 Calculations of Formation Energies The first attempt at calculations of APB energetics in GaAs were made by Petroff in 1986 [90] These calculations were based on a simple analysis of the wrong bond energy and wron g bond densities for given APB planes. Petroff found that {110} and {112} APBs should have a lower formation energy than {111} or {100} APBs. However,

PAGE 32

32 subsequent work by Vanderbilt and Lee in 1992 [91] showed that this approach was insufficient because it neglected the electrostatic contribution of wrong bonds to APB formation energy. The Ga Ga bond has a net charge of q/2 and the As As bond has a net charge of +q/2, where q is the elementary electron charge. APBs on {110} type planes are self compensated in that they are stoichiometric and will have equal numbers of Ga Ga and As As bonds. On the other hand, {111} APBs are non stoichiometric and will be fully formed of the same type of wrong bonds, either As As or G a Ga. Thus, APBs that are stoichiometric should generally be energetically favored over non stoichiometric APBs. The first principles calculations by Vanderbilt and Lee found that {110} APBs had lower formation energies than non stoichiometric {111} or {10 0} APBs. The most recent analysis of APB formation energies for GaAs was made by Rubel and Baranovskii in 2009 [92] using density functional theory (DFT) calculations. They expanded upon the work of Vanderbilt and Lee by calculating formation energies for {112} APBs which are fully stoichiometric and {113} APBs which are partially stoichiometric. As with {111} APBs, these higher index planes would also be conducive for self annihilation. The APB formation energies calculated by Rubel and Baranovskii are listed in Table 1 1. It was once again determined that {110} APBs have the lowest formation energy (28 meV/ 2 ). APBs along {112} and {113} have an identical formation energy (39 meV/ 2 ) that is lower than that found for {111} APBs (43 meV/ 2 ). The low formation energy of {110} APBs agrees with TEM observation s that show {110} planes as the most common for APB propagation [83, 93] However, the evidence for APB self annihilation [11] indicates that APBs must be kinking to higher index planes, even if they

PAGE 33

33 are higher formation energy. Thus, {112} and {113} planes may be good candidates as APB habit planes to facilitate annihilation. It should be noted th at while {110} APBs are commonly observed in TEM and have relatively low formation energies, APBs can theoretically exist on any set of crystal orientations. When observed in plan view, such as in Figure 1 2 or in the work of Cohen and Carter [94] APBs are not always faceted along {110} planes forming re ctangular domains. Rather, the APBs are faceted along various planes including {110} and sometimes are apparently curved. Vertical APBs connecting {110} segments could exist in any configuration between {110} and {100} e.g. {120} or {130} The formation e nergies of APBs along these planes have not been calculated, nor have they been calculated for curved APBs, likely in part since these orientations do not facilitate self annihilation. The work of Beyer et al. [95] also showed that nucleated APBs can be templated by the underlying Si surface structure. It is possible that APBs nucleate along vari ous facets including {110} to follow the Si step structure and then otherwise follow the shortest distance to form a closed loop. This configuration would minimize the total line length of APBs that form and lead to the sometimes curved and wandering natur e of APBs seen in Figure 1 2. There is an energy balance between the formation energy of APBs along these orientations and the overall energy added to the system by the total line length of APBs T he lowe st energy configuration of APBs may be revealed by annealing experiments. In general, the existence of any APB in GaAs should be unfavorable due to the large associated energy. As a comparison, the formation energy of a stacking fault in GaAs is ~3 meV/ 2 [45, 96] about an order of magnitude lower than the formation

PAGE 34

34 energy for APBs. Only zero dimensional defects, i.e. point defects, should be pres ent at thermal equilibrium [44] In GaAs on Si APBs are forced to form because of sin gle steps on the Si surface. It follows from the thermodynamic instability of APBs that there should be an energetic driving force for the self annihilation of APBs. This driving force is likely what leads to the kinking of APBs from {110} planes to higher index planes, even if these planes are theoretically a larger energy cost. 1.5. 2 Fermi Level Effects Since APBs are composed entirely of antisite bonds their energetic stability may be affected by the relative stability of antisite defects in the GaAs l ayer. The Ga Ga wrong bond will have an associated Ga As defect and the As As wrong bond will have an associated As Ga defect. Recall that these bonds have a net charge; Ga Ga bonds will act as acceptors and As As bonds will act as donors. Walukiewicz presen ted a model in which the Fermi energy in heavily doped GaAs is stabilized by amphoteric native defects that compensate the do pant species [97] In other words, the formation energy is lowered for a given native defect, depending on the type of dopant used. As the Fermi level approaches the conductio n band edge, i.e. f or more n type GaAs, the formation of Ga As antisite defects and Ga vacancies (V Ga ) that act as compensating acceptors is favored. T owards the valence band edge, i.e. for more p type GaAs, formation of As Ga antisite defects and As vacancies (V As ) that act as compensating donors is favored. These native defects in compound semiconductors work to stabilize the Fermi energy and shift it away from the band edges closer to an equilibrium charge neutral level [98] The Fermi level effect on point defect energies was later corroborated with DFT calculations by Chroneos et al. [99] This work focused only on the forma tion energies of Ga As and As Ga antisite defects as a function of Fermi level, but showed a similar trend

PAGE 35

35 to the work by Walukiewicz. The formation energy of As Ga is lower near either the valence or conduction band edge. The formation energy of Ga As tends t o decrease moving from the valence band edge to the conduction band edge, where it is lower overall in energy compared to the As Ga defect. Exp erimental work has supported this model of dopant com pensation by native point defects for Si doped InGaAs [100] However, no study has reported on the effect of doping on APB energetic stability in GaAs or other III V arsenides. 1.5.3 Kinetics of Transformation While APBs are a unique defect for polar on nonpolar systems in semiconductor epitaxy, they are hardly unique to compound semiconductor material systems as a whole. APBs can exist in any crystal in which there are multiple equivalent sublattices such that the polarity can be reversed from one grain to a neighboring grain. They are well documented in other compounds, including metals and m etal oxides. Some of the earliest reports on the possible existence of APBs were made for copper gold alloys in the 1930s [101 103] Cu 3 Au, for example, has a FCC lattice with four equivalent simple cubic sublattices [103] Various studies found that these alloys exhibited antiphase domains in their ordered states [104 106] Notably, these studies also found that the antiphase domain size changed with annealing treatments [106, 107] The domain growth for these systems typically obeys the following equation [107, 108] : (1 3) where t is the annealing time, D is the domain size at time t, D 0 is the initial domain size, and k is a rate constant. Equation 1 3 is identical to the equation used to describe the kinetics of grain growth [109] The temperature dependence of domain growth can be then described by an Ar rhenius relationship:

PAGE 36

36 (1 4) where k is the rate constant from Equation 1 3 at a given temperature T, k B is a is the activation energy. As is the case in grain growth, Allen and Cahn [110] theorized that the evolution of APBs is driven by a reduction in curvature. Over time, as domains grow, both the mean curvature and total surface area of APBs s hould decrease. This transformation lowers the overall free energy of the system. Allen and Cahn derived a equation for this transformation similar to Equation 1 3: (1 5) where S v dimensionless constant, and M is a rate constant similar to k from Equation 1 3 with units of m 2 /s. S v can be related to APB density which is simply the line length of APBs per unit area in plan view. Therefore, the change in APB density should also have a t 1/2 dependence. This time dep endent APB density evolution has been observed during growth of epitaxial Fe 3 O 4 films [111] and with annealing [112] but has not been studied for APBs in GaAs. 1.6 Thermal Effects on III V Antiphase Boundaries Although there has been extensive work on the effects of thermal treatments on APBs in other material systems, thermal effects on APB evolution in GaAs and ot her III V compounds is not well understood. There are two primary avenues for affecting APB energetics in epitaxial GaAs films grown on Si via application of thermal energy. The first is to increase the growth temperature which may affect the tendency of A PBs to change habit planes. The second is to anneal the samples after growth which may allow

PAGE 37

37 neighboring APBs to diffuse towards each other and self annihilate, similar to grain growth. Studies on the effects of growth temperature and annealing on APBs in GaAs and other III V compounds will now be discussed. 1.6 1 Growth Temperature Effects If APBs in GaAs on Si(001) layers propagated entirely parallel to the substrate normal on the lowest energy {110} planes, they would never self annihilate. However, the work of Georgakilas et al. [11] strong ly indicated that APBs have a tendency to change their plane of propagation during growth. It follows that the tendency of APBs kink to higher index planes that are conducive for self annihilation may be a function of the growth temperature. While MOCVD is not an equilibrium process [30] increasing the temperature in the system should bring it closer to equilibrium conditions. Based on the high formation energy of APBs, the equilibriu m state for a GaAs on Si layer should be APB free or, at the very least, with minimized APB interfacial area. Figure 1 7 shows an arbitrary free energy diagram for a GaAs on Si system in two different states: B, a GaAs layer with APBs propagating verticall y along {110} planes, and A, a GaAs layer with self annihilated APBs. The transition between the two states may be kinetically limited by different factors during growth. Thus, there is an activation energy, E a that must be satisfied by increasing the gro wth temperature. The effect of growth temperature on APBs has been examined for GaP films grown on Si(001) by MOCVD [113, 114] For bulk growth temperatures near 450 C, APB propagation along {110} planes was favored and these APBs extended to the surface of the film. When the bulk growth temperature was raised to 625 C, the majorit y of APBs were oriented along {111} and {112} type planes leading to complete

PAGE 38

38 suppression of antiphase domains in the film. The transition between majority {110} APBs to majority {112} APBs with the increase in growth temperature indicates there is a kinet ic barrier for APB propagation along these higher index planes. 1.6 2 Annealing Effects Using the same rationale for the effect of growth temperature, post growth annealing should also drive the GaAs on Si system to its equilibrium state. Thus, it may be possible for annealing to dissolve APBs as long as the temperature is high enough to satisfy the energy required to diffuse APBs. Qualitatively, the energetic barrier for a n antiphase domain wall to move would likely be substantial, as it would require the diffusion of an entire plane of antisite defects. The effect of annealing on APBs in GaAs grown on Si(001) by MOCVD was briefly examined by Chu et al. in 1988 [115] The films were between 1.9 m and 5.5 m thick and had a high degree of antiphase disorder. The study found that APB density in the films decreased, i.e. antiphase domains grew, with 2 hr annea ls for a temperature range from 700 to 800 C. No remarks were made about the activation energy for APB diffusion. However, the results do demonstrate that the diffusion of an APB in GaAs is possible which is consistent with what has been reported for anne aling of APBs in other material systems. Further, the anneals indicate that the lowest energy state for the GaAs on Si films is in fact APB free and that annealing helps to reduce the total APB interfacial area. The evolution of APBs with annealing was re cently investigated for GaP films grown on Si by MBE [116] The study examined the ratio of antiphase crystal to main phase crystal using data obtained by X ray diff raction reciprocal space mappin g for an

PAGE 39

39 as grown sample and a sample annealed at 600 C for 10 min. It was found that the ratio of antiphase to main phase crystal increased after annealing, in addition to increased surface roughne ss as determined by atomic force microscopy (AFM). The authors theorized that this result occurred in part because the APBs were grown in along {111} type orientations. Annealing the sample caused APBs to reconfigure along {110} planes that are more energe tically stable [92] The enhanced migration of atoms in an APB near the surface of the film during this volumetric redistribution would cause pits to form and make the surface very rough. This theory contradicts the hypothesis that the equilibrium state for a film with APBs should be APB free. If APBs reconfigu red along {110} planes during annealing then they would never annihilate. More work is needed to understand the effect of annealing on APBs in GaAs and other III V compounds. 1.7 Summary and Statement of Research Goals While the epitaxy of GaAs on Si has been studied for several decades, there are still significant gaps in literature regarding the characteristics of APB defects. Many early studies took advantage of methods to suppress antiphase disorder that are not practical for industrial integration. Th e use of GaAs and other III V compounds in future electronic devices will require a more complete understanding of factors influencing the nucleation and energetic stability of APBs in GaAs on Si. This work will examine these characteristics of APBs at dif ferent stages of processing for GaAs on Si films. First, the quantitative relationship between substrate surface contamination and nucleated APB density will be developed. Second, the influence of MOCVD growth temperature on the annihilation rate of APBs a nd the associated activation energy will be investigated. Third, the effect of post growth annealing on grown in APBs and the associated

PAGE 40

40 activation energy for APB diffusion will be studied. Finally, the effect of in situ doping on grown in APB density will be examined.

PAGE 41

41 Table 1 1. GaAs APB formation energy on different habit planes as determined by DFT calculations. Values from Rubel and Baranovskii [92] GaAs APB Plane Formation Energy (meV/ 2 ) {110} 28 {111} 43 {112} 39 {113} 39 Figure 1 1 GaAs unit cell along various viewing directions. A) Offset view showing atom positions. B ) Viewed along the [110] projection showing the Ga As dumbbell structure. C) Viewed along the [ 110] projection showing the reversal of Ga As dumbbell polarity when the lattice is rotated by 90 Images made with CrystalMaker software CrystalMaker Software Ltd, Oxford, England (www.crystal maker.com).

PAGE 42

42 Figure 1 2 Plan view SEM image of GaAs layer grown on Si(001) with high APB density APBs are visible as linear features of dark contrast. Figure 1 3 Schematic drawing of dimer bonds on the Si(001) surface showing the rotation of dimers by 90 across single steps.

PAGE 43

43 Figure 1 4 <110> projection of an epitaxial GaAs layer on a Si(001) substrate with APBs A ) A n APB is nucleated at a single step and propagates along a {110} plane B) APBs may also nucleate along inclined planes, e.g. {111}, and self annihilate C) Double steps suppress the formation of APBs by resetting the atomic ordering of the epitaxial layer. Figure 1 5 <110> projection of a Si(001) surface showing the two types of terraces that will exist Type A terraces have dangling bonds perpendicular to the step edge and Type B terraces have dangling bonds parallel to the step edge.

PAGE 44

44 Fig ure 1 6 Composite XTEM image showing the meandering path of an APB (marked by arrows) through a GaAs on Si(001) layer. Figure 1 7 Arbitrary Gibbs free energy diagram for a GaAs on Si system w ith two different states: B, a GaAs layer with APBs propagating vertically through the film, and A, a GaAs layer with self annihilated APBs suppressed in the film. Arrangement A should be thermodynamically favored but the transition between the two states may be kinetically limited by an energy barrier, E a

PAGE 45

45 CHAPTER 2 EXPERIMENTAL METHODS In any study on APBs in III V films grown on Si, it is critical to know the complete background of sample processing conditions as these can have a large impact on the severity of antiphase disorder. An overview of the methods used to prepare, grow, and process all samples for the experiments in this work will be discussed in the following sections. Details that are pertinent for specific experiments will be given in the respective chapters. The latter portion of this chapter is dedicated to sample characterization methods with an emphasis on how certain techniques can be used to detect and quantify antiphase disorder. 2.1 Materials Processing 2.1.1 Substrate Preparation All substrates used in this work were commercially available double side polished 300 mm Si(001) wafers with no intentional off cut. There is typically some degree of unintentional wafer misorientation due to uncertainties in processing from the boule of u p to 0.5 [117] Prior to film growth, the native oxide and other contaminants on the Si substrates were removed in a novel process developed by Applied Materials known as a Siconi TM clean. The details of this cleaning procedure have been discussed elsewhere [118] Briefly, the substrates are placed under vacuum and transferred to a dedicated cleaning chamber where they are exposed to a dry NF 3 /NH 3 based plasma. The plasma species have high selectivity for the Si native oxide and forms reaction products that are easily sublimated at temperatures near 100 C. This Siconi TM cleaning process has several advantages over conventional substrate cleaning methods. First, it i s a low temperature

PAGE 46

46 process. Second, the substrates are transferred from the cleaning chamber to the growth chamber under vacuum. The vacuum transfer prevents any oxide reformation from exposure to the ambient. This exposure is unavoidable if dipping in HF is used to remove the native oxide. After transferring the substrates to the growth chamber, a high temperature bake is performed under hydrogen gas flow near 900 C. This pre growth bake helps to promote the formation of double steps on the Si surface. The combination of the substrate clean and pre growth bake is critical for reducing defect densities in the epitaxial GaAs films. The effect of residual surface contamination after Siconi TM cleaning on APB density will be discussed in Chapter 3. 2.1.2 GaAs Film Growth GaAs film growth in this work was performed in an Applied Materials III V MOCVD system. A simplified schematic of a typical MOCVD growth system is shown in Figure 2 1. In the MOCVD growth process, the substrate is loaded into the chamber on a rotating chuck or susceptor. The temperature of the susceptor, and thereby the temperature of the substrate, is controlled using heating elements such as lamp arrays or filaments. The metal organic precursors are contained outside of the growth chamber. Hy drogen gas is flowed through the containers to carry vaporized precursor molecules into the growth chamber. The flow rate of the individual precursors can be easily modulated. The precursor molecules contained in the hydrogen gas flow towards the heated su bstrate where they react and decompose forming solid phase material with the metallic element and gaseous organic by products. The by products are then pumped out of the chamber.

PAGE 47

47 The key parameters for the growth of GaAs on Si via MOCVD are the growth temp erature, i.e. the substrate temperature, and the ratio of the partial pressures of the group V and group III precursor gases, i.e. the V/III ratio. The growth temperature must be sufficient to decompose the precursor molecules and also can affect the growt h rate. The V/III ratio must be greater than unity in order to form purely solid phase GaAs due to the relatively high volatility of group V elements [30] The baseline recipe for GaAs film growth in this work has been developed and optimized by Applied Materials. The precursors used for Ga and As are trimethylgallium (TMGa) and tertbutylarsine (TBAs), respectively. The V/III ratio is approximately an order of magnitude greater than unit y. The chamber pressure is maintained in the typical reduced pressure MOCVD processing range of 10 100 Torr [30] A standard two step growth process is used, unless otherwise stated, w ith growth temperatures in the range of 500 to 700 C. 2.1.3 Thermal Processing The thermal processing of GaAs can be very challenging and it is one of the reasons as to why GaAs never overtook Si as the primary material of the semiconductor industry in the 1980s. Thermal treatments are essential in semiconductor processing for the di ffusion and activation of dopant impurities. The temperatures required can easily reach 850 C to over 1000 C for anneal times up to 60 min [119] These temperatures pose n o issue for Si which is only limited by its melting point of ~1400 C. However, for a binary semiconductor such as GaAs, the surfaces are prone to dissociation before the actual melting point of the solid. GaAs is known to suffer significant surface degrad ation at temperatures above 650 C [120] At this point, known as the congruent evaporation point [121] As leaves the surface preferentially over Ga leaving behind a non

PAGE 48

48 stoichiometric surface. Ga droplets can then form which etch away underlying material and r oughen the surface [120] Several strategies have been developed to protect the GaAs surface during thermal treatments and to suppress the loss of As. 2.1.3. 1 Prevention of su rface d egradation One of the methods to prevent GaAs surface degradation during annealing is to provide an overpressure of As in the annealing environment. The overpressure is typically accomplished by flowing an As based gas or by heating liquid or solid source As in a separate chamber and flowing the vapor into the compartment containing the GaAs sample [120] The higher partial pressure of As at the GaAs surface reduces the t endency of As to evaporate, thus increasing the upper temperature limit of the anneals. This technique is not popular due to the inherent safety concerns involved with handling toxic As based gases and other sources. An alternative means of providing a lo cal As overpressure at the GaAs surface is to physically place another piece of material in contact with the GaAs sample. This technique is called proximity capping. The cap material is typically another piece of GaAs, but Si may also be used. The presence of the proximity cap suppresses the outdiffusion of As to atmosphere from between the two surfaces, i.e. any evaporated As is essentially trapped near the GaAs surface. Therefore, further degradation of the surface is prevented. The proximity cap method c an be effective for temperatures up to about 900 C [120] Instead of using a proximity cap, it is also possible to protect the GaAs surface by depositing a thin dielectric f ilm encapsulant. Early encapsulating films used were Si 3 N 4 SiO 2 and AlN [120] since these materials can be easily deposited at relatively low temperatures, form good interfac es, and are stable at elevated anneal temperatures. As

PAGE 49

49 evaporation is suppressed for processing temperatures of up to 900 C for 10 s [122] More recently, Al 2 O 3 films grown by atomic layer deposition (ALD) have been used as a successful encapsulant material for InGaAs [100, 123] The interest in ALD A l 2 O 3 films was spurred by the need for an effective gate dielectric material for future MOSFET devices based on III V arsenides, e.g. GaAs and InGaAs. It has been demonstrated that the process of depositing Al 2 O 3 on GaAs via ALD has the secondary effect of cleaning the native oxide present on the GaAs surface [124, 125] For the anneals performed in this work, the maj ority of GaAs on Si samples were capped with Al 2 O 3 films deposited by a Cambridge Nano Fiji 200 ALD system in the Research Service Centers (formerly Nanoscale Research Facility) cleanroom at the University of Florida. Films thicknesses varied between 15 an d 30 nm. The precursors for Al and O were trimethylaluminum (TMAl) and water, respectively. The O source was changed to an oxygen plasma for select recipes which will be discussed in Chapter 5. The deposition temperature was between 200 C and 250 C. Addi tional annealed GaAs on Si samples were proximity capped with either a polished semi insulating GaAs wafer or another piece of GaAs on Si with the two GaAs films in contact. 2.1.3.2 Furnace anneals Long term anneals ( 4 min) in this study were performed i quartz tube furnace. Due to the long ramp rate that is characteristic of tube furnaces with heating coils, shorter anneals are not reliable. Samples were loaded surface of interest down on a Si carrier wafer in a quartz boat. The temperat ure of the furnace was controlled using an external digital controller and checked with a thermocouple prior to inserting the boat with the samples. Care was taken to ensure that the thermocouple

PAGE 50

50 measured the temperature of the furnace at the same location along the tube as where the samples would sit. During the anneals, inert Ar gas was flowed through the tube at about 3 L/min. 2.1.3.3 Rapid thermal anneals Short term anneals ( 4 min) in this study were performed in an AG Associates HeatPulse 4100 rapid thermal anneal (RTA) system. A RTA system uses high intensity lamps to heat samples inside of a quartz chamber. The quartz is transparent to the radiation emitted by the lamps which allows for much faster and direct heating of the samples [126] Samples were placed surface of interest down on a Si carrier wafer on top of a quartz tray inside the anneal chamber. The chamber was filled with inert Ar gas during the anneals. The temperature was monitored by a thermocouple placed between the Si carrier wafer w ith the samples and an additional Si wafer underneath. The temperature ramp rate used for the RTA recipes in this work was approximately 25 C/s. 2.2 Sample Characterization 2.2.1 Transmission Electron Microscopy TEM is a useful method for microstructural characterization of semiconductor materials down to the atomic scale. The ability analyze features at this scale has become increasingly more important with the ever decreasing physical length scales of transistors, now approaching the sub 10 nm regime. TE M operation is relatively complex so a brief background on the subject is necessary before beginning the discussion of APB characterization. In TEM, an electron beam is generated and accelerated towards a specimen. The operating voltage is typically between 100 and 300 kV. Along the way, the beam focused by a series of magnetic lenses and also passes through a number of different

PAGE 51

51 apertures. The specimen sits near the middle of the TEM column with lenses and apertures above and below it. The electrons in the beam are scattered by the specimen as they pass through it before they finally reach a detector. This interaction can be used to develop ima ges of the specimen since the interaction depends on the specimen structure. The exact information that can be gained from TEM imaging depends on numerous factors including the mode of TEM operation (i.e. the setup of the beam), the orientation of the spec imen, and the crystalline nature of the specimen. For the purposes of this work, only background on TEM imaging of single crystal materials is relevant. The interaction of the electron beam with a single crystal specimen generates a characteristic diffract ion pattern with spots from specific families of crystal planes. The (2 1) pattern depends on the orientation and crystal structure of the specimen. For most of the TEM in this work, specimens are imaged in cross section along <110> type crystal lographic directions. Figure 2 2 shows the diffraction pattern that is expected from a GaAs on Si sample along a [110] zone axis. Note that GaAs and Si will have similar diffraction patterns due to equivalent positioning of atoms in their crystal structure s, but the spacing of the diffraction spots, or Bragg spots, will differ due to the dissimilar values of d for their respective sets of crystal planes. The 002 type Bragg spots are forbidden reflections for Si because every atom in its crystal structure is identical which

PAGE 52

52 cancels out the intensity. These spots will appear, although weakly, for GaAs as superlattice reflections because it has two distinct Ga and As based FCC sublattices that will scatter the electron beam differently [127] The 002 Bragg spots in turn can be used to det ect antiphase disorder because these planes are sensitive to the non equivalent ordering of Ga and As atoms on the sublattices across an APB, as was discussed in Chapter 1. The primary modes of TEM operation are bright field (BF), dark field (DF), and high resolution (HR). In BF imaging, the objective aperture of the TEM is centered on the direct spot (Figure 2 2) so that it is the only contributing beam. The factors affecting contrast in BF imaging are mass thickness of the specimen and diffraction of the beam. Regions of the sample that strongly diffract the electron beam will appear dark and regions in which the beam passes through relatively unperturbed will appear bright. Thicker samples or samples with atoms of higher Z will also appear darker due to g reater scattering of the electron beam. DF is performed by using one of the Bragg spots for imaging instead of the direct spot. The sample is usually tilted such that the scattering intensity for the desired Bragg spot is enhanced, thereby leading to bette r contrast. The contrast is opposite that of BF, with strongly diffracting areas of the sample now appearing bright. This aspect of DF imaging makes it useful for the analysis of crystal defects. HRTEM imaging utilizes phase contrast, i.e. the difference i n phase of the electrons in the beam after they interact with the sample, in addition to diffraction and mass thickness contrast. The objective aperture must be placed around the direct spot and multiple Bragg spots so that they all simultaneously contribu te to imaging. Lattice fringes corresponding to atomic planes in the sample become visible in HRTEM

PAGE 53

53 images because the difference in phase between electrons depends on whether they travel directly down columns of atoms or in between them [128] The variation in intensity is periodic with a spacing that corresponds to the atomic plane spacing. High angle annular DF scanning TEM (HAADF STEM) is a special mode of TEM operation that can provide atomic resolution images with excellent Z contrast. Therefore, for III V compounds with differing atoms, it is possible to directly observe and discern the group III and group V atoms and their specific sublattice allocation. 2.2.1.1 Ident ification of antiphase boundaries DF imaging using the 002 superlattice reflection has previously been used to confirm the presence of APBs in GaAs and other III V compounds [11, 79, 82, 89, 113] Due to the reversal of the Ga As dumbbells across an APB, the g 002 reflection for the crystal on one side of the APB is equivalent to the g 00 2 reflection for the crystal on the other side [129] Therefore, when performing DF imaging using only one of these spots, the contrast from diffraction intensity should also be reversed. That is, the region inside of an antiphase domain should app ear bright relative to the main phase crystal, or vice versa. This contrast reversal can be seen in Figure 2 3 for a <110> DF XTEM image of a GaAs film grown on Si made using the g 002 reflection. The contrast can be enhanced by tilting the sample to also e xcite g 115 and g 117 type spots along with the g 002 type spot [113] For a [1 10] cross section, the g 115 and g 11 7 spots should be excited. This tilting increases the diffraction intensity of the g 002 type spots used for imaging. APBs can also be identified in high resolution TEM images (either HRTEM or HAADF STEM) by Fourier filtering the images using the 002 spots [130, 131] The filtered images show a series of lines corresponding to the (002) planes in the crystal. The (002) planes are misaligned by plane spacing across an APB. Thus, at the

PAGE 54

54 location of an APB in the filtered image, the lines are disrupted and misaligned by the same pro portion. Fourier filtering of TEM images in this work was accomplished using ImageJ software [132] Fig ure 2 4 shows the procedure for Fourier filtering a HAADF STEM image of an APB in a GaAs on Si sample (A) First, a fast Fourier transform (FFT) is taken of the original TEM image (B) Second, a mask is created encircling only th e 002 spots and the direct spot (C) Finally, the inverse FFT of this masked FFT is taken, which give s the Fourier filtered image (D) The misalignment of the ( 002 ) planes at the location of the APB is clearly visible. The procedure for processing HRTEM images and identifying APBs i s identical as long as the HRTEM image is of sufficient magnification and resolution for lattice imaging. 2.2.1.2 Preparation of samples Samples for XTEM imaging in this study were prepared using a Focused Ion Beam (FIB) system. The primary tools used were FEI Strata DB235 and FEI Helios NanoLab 600 dual beam FIB/SEMs located at the University of Florida Research Service Centers. The recipe for sample preparation is similar to that published by Schaffer et al. [133] Briefly, Ga + ions are used to mill out a thin lamella (~10 m long x 5 m tall x 1 m thick). The lamella is attached to an Omniprobe micromanipulator and separated from the rest of the sample. The lamella is then attached to a s pecialized Cu Omniprobe TEM grid where it is thinned to a final thickness of ~100 nm. It is important to protect the sample surface during milling from the ion beam to prevent amorphization. Initially, a carbon rod evaporator system was used to coat sample s with a thin C film. However, based on work by Park et al. [134] it was determined that coating the samples with Sharpie permanent marker ink was sufficient and this method was

PAGE 55

55 subsequently used. Further surface protection is afforded by depositing a 1 m thick Pt bar on top of the C layer over the area of interest in the FIB system. 2.2 .2 Chemical Etching and Scanning Electron Microscopy Chemical etching by wet or dry methods is a widely used tool in semiconductor processing. The diversity of materials present in the fabrication of any integrated circuit device require an equally diverse array of chemistries that will selectively etch away material with high spatial and temporal precision. These same etching chemistries can also be effective in the identification and quantification of defects. The defects present in a given material, such as APBs in a GaAs on Si film, are not visible to the naked eye and may not even be visible in the as grown material under a high powered microscope amongst the typical surface morphology. Selectively etching these defects would allow for a quick and accurate count of the defect densities. Wet chemical etching, in combination with a SEM for imaging, is used for the quantification of APB density in this work. Wet chemistries for etching are binary, having one component for oxidizing some material in a sample and a second component for reducing and removing the newly oxidized material. The sequence repeats itself until the etch process is stopped, typically by immersing the sample in water to neutralize and dilute the etchant. SEM is a good tool for imaging defects with the scale and distribution of APBs since it can quickly take several ima ges over a large area (> 100 m 2 ) with excellent spatial resolution (< 1 nm). Selectively etching APBs would cause etch pits to form along their entire length. Thus, the APBs will show up in secondary electron mode SEM images with dark contrast relative to the surrounding material. Fewer secondary electrons are able to escape from the material inside the etch pits and

PAGE 56

56 reach the detector, thereby giving lower intensity of signal in those areas. The SEM used for imaging in this work was a FEI Nova NanoSEM 430 at the University of Florida Research Service Centers operating at 5 kV. An etchant that is selective for APBs in GaAs has been reported on previously [11, 115, 135] The chemistry involves HNO 3 as the oxidizing agent, HF as the reducing agent, and water as the diluting agent for control of etch rate. In thi s study, a ratio of 20:1:7 of of HF(49%):HNO 3 (69%):H 2 O by volume was used. This etchant will henceforth be referred to as the HF/HNO 3 etching solution. A typical etch time of 10 s at room temperature was found to be effective in developing strong contrast of APBs in SEM (see Figure 1 3 for a representative image). The HF/HNO 3 solution is commonly used in semiconductor processing for the etching of Si. It is likely selective for APBs in GaAs due to the high associated energy of APBs. However, a more accurate description would be to say that the etchant is more selective for APBs over other defects and the GaAs film over the time scale of the etch. Some thickness of the overall GaAs film is still removed but the APBs are etched at a faster rate. Figure 2 5 sho ws BF XTEM images of a GaAs film grown on Si as grown ( A) and then after etching for 10 s in the HF/HNO 3 solution ( B) The etch pits that form at the location of APBs which give rise to the strong contrast of APBs in SEM ar e clearly visible in Figure 2 5B The as grown GaAs film thickness is about 450 nm. After etching, about 50 nm of GaAs is removed and the etch pits extend a further 100 150 nm into the film. Thus, the etch rate of the APBs (~15 20 nm/s) is three to four times that of the overall GaAs film (~5 nm/s). The selectivity of the HF/HNO 3 solution for APBs over other crystallographic defects was also confirmed with XTEM. Figure 2 6 shows a HRTEM image of a

PAGE 57

57 developed etch pit in a GaAs film on Si. At the bottom point of the etch pit, an APB can be observed extending down into the film. A stacking fault can be seen intersecting the sidewall of the etch pit. The lack of an additional smaller pit or depression at the site of the stacking fault intersection with the etch pit demonstrates that it is not attacked preferentially by the HF/HNO3 solution over the APB. Therefore, the etched features visible in SEM can unambiguously be identified as APBs. The APB density in samples etched with the HF/HNO 3 solution is quantified by tracing the line length of AP Bs in a given plan view SEM image using ImageJ software [132] Figure 2 7 shows an example of the traci ng procedure for APBs in a plan view SEM image. APB features are traced by hand using the freehand line tool in ImageJ. The total line length counted is then divided by the area of the image. Therefore the units for APB density are defined as m 1 In order to obtain a representative count of the APB density for a sample, multiple images are taken in different areas of the sample surface. A total area of at least 500 1000 m 2 is counted, depending on the relative APB density in the sample. The APB density reported for a sample is the average of the APB densities in each SEM image and the error is the standard deviation. While having an APB selective etchant is a helpful tool for identifying APBs, an isotropic non selective etch is useful if depth p rofiling of APB density is desired. There are several reported etchants for GaAs [136] A mixture of H 3 PO 4 and H 2 O 2 was previously used for the depth profiling of APB density in GaAs films [11] and was selected for the same purpose in this study. The composition used for this work was CH 3 OH:H 3 PO 4 (85%):H 2 O 2 (30%) in a ratio of 10:1:1 by volume. This etchant will henceforth be referred to as the H 3 PO 4 /H 2 O 2 solution. The etch rate of the H 3 PO 4 /H 2 O 2

PAGE 58

58 solution at room temperature for GaAs was determined to be ~120 130 nm/min. The non selectively of the etchant was also confirmed. Figure 2 8 shows a BF XTEM image of a GaAs on Si sample after the removal of ~380 nm of GaAs. The surface of the film is still relatively smooth and the defects in the film are not preferentially attacked. The HF/HNO 3 solution can then be applied to deter mine the APB density at a given depth in the film. 2.2 3 X Ray Diffraction X ray diffraction (XRD) is a method that can help to determine the crystal quality of a sample. The X rays will interact with the sample and constructively interfere when the Bragg condition is met for a set of atomic planes (Equation 2 1). This constructive interference will give a peak of intense signal for a given Bragg angle, or reflection. In theory, for an ideal crystal, the signal for these satisfied reflections should be line ar. However, scattering from perturbations in the crystal lattice, such as defects that induce local strain, lead to broadening of the peak. Specific lattice plane reflections can be analyzed that are sensitive to the presence of crystalline defects. For epitaxial GaAs grown on Si, the fundamental (004) reflection can be used to determine the threading dislocation density in the films that arises from the lattice mismatch [80, 137, 138] The threading dislocation density is related to the full width at half maximum (FWHM) of the (004) peak by [137] : (2 2) of the Burgers vector for the dislocation. For classic 60 dislocations, the value of b is

PAGE 59

59 XRD can also be used to detect the presence of antiphase disorder in a crystal [13, 116, 139] For GaAs and other III V films grown on Si, the superlattice (002) reflection is sensitive to the antiphase disorder in XRD for the same reasons as in TEM. The presence of antiphase domains in the film with inequivalent atomic ordering on (002) planes leads to broadening of the ( 002) peak. The peak broadening more specifically is related to the average separation between APBs, i.e. the average domain size. For samples with less separation between APBs, i.e. higher APB density, the peak broadening is more severe. The average domain size was determined for GaP on Si samples with additional processing of the XRD data [116, 139] The downside to XRD is that it does not provide direct imaging of APBs. It should also be noted that there is an upper limit on antiphase domain size that can be resolved by XRD which is around 300 to 500 nm [13, 116] depending on the XRD system used. Therefore, for comparing antiphase disorder between different sets of samples, selective etching and SEM is a more reliable method of quantifying APB density. 2.2 4 Ellipsometry Spectroscop ic ellipsometry can be an effective method for determining the thickness of a thin film. A multiple wavelength polarized light source is set at an angle to the sample such that the light reflecting off the film reaches a detector. The detector measures cha nges in the amplitude and phase of the light as a function of the wavelength. Light reflecting off of the sample surface and light traveling through the film can have a phase difference and interfere, thereby affecting these changes [140] Any light absorption by the layer will also affect this interference for a given wavelength of incident light. A model is fit to the data, assuming a layer composition with specified optical properties, to extract the film thickness. Ellipsometry can be used to measure

PAGE 60

60 film thicknesses from the sub nm level up to approximately a micron. XTEM can also be used to meas ure layer thicknesses, but ellipsometry is advantageous in that it is a non destructive process and measurement of several samples can be completed in a relatively short period of time. Ellipsometry is used in this study as the primary method to measure th icknesses of GaAs films after the non selective H 3 PO 4 /H 2 O 2 solution etch. The ellipsometer used for measurements was a J.A. Woollam M88 at the University of Florida Research Service Centers. The accuracy of models for fitting GaAs film thicknesses was confirmed with selected XTEM samples. 2.2 5 Secondary Ion Mass Spectrometry S econdary ion mass spectrometry (SIMS) is used to determine the chemical composition of a sample with depth sensitivity. The sample is sputtered from the surface using an ion beam, typically Cs + The ejected secondary ions are accelerated towards and analyz ed by a mass spectrometer. The spectrometer separates the ions by their mass to charge ratio to determine a number of counts for an element. Comparison to known standards allows for conversion to a more quantitative concentration value. By knowing the sput tering time and measuring the depth of the sputtered crater, the concentration of a given element can be tracked as a function of its depth in the sample. SIMS can detect atomic concentrations on the order of 10 15 cm 3 or lower which makes it ideal for ana lyzing dopant concentrations in semiconductor thin films. SIMS will be used in this study when necessary to determine the concentrations of impurity atoms, dopants or otherwise, at the GaAs/Si interface and in the GaAs film. 2.2 6 Hall Effect Hall effect m easurements are a widely use method in semiconductor research to determine the electrical properties of samples, including carrier type and concentration

PAGE 61

61 and carrier mobility Measurements are commonly made with contacts at the four corners of a square sam ple, known as the Van der Pauw configuration. A known current is applied across a pair of contacts in opposite corners in the presence of a magnetic field with a known strength The Lorentz force from the magnetic field deflects the charge carriers by some amount and induces an equivalent electric field to oppose this force. The voltage associated with this induced electric field is measured at the remaining two contacts. This voltage is called the Hall voltage V H and can be used to determine the sheet ca rrier density N s, by the relation: (2 3) where I is the applied current, B is the applied magnetic field strength, and q is the elementary electron charge. The sign of the Hall voltage will change depending on the carrier type i.e. negative for electrons and positive for holes Sheet carrier density, or sheet number, is an areal density and can be used to determine the concentration of carriers by dividing by the thickness of the film or doped layer. The carrier mobility can be calcu lated from the sheet number by: (2 4) where R s is the measured sheet resistance. The measurements are repeated for all sets of contacts, and then again with the magnetic field reversed for a more accurate determination of the Hall voltage. Hall effect measurements using the van der Pauw method are used in this study to determine carrier concentration s and mobilities in GaAs on Si films. Samples are 1 cm by 1 cm square with contacts made in the corners using pressed on In metal. For

PAGE 62

62 undoped GaAs, it can be difficult to make the ohmic contacts required for an accurate Hall effect measurement [17] Annealing is typically needed after depositing the contact material to form a more conductive alloy at the surface [141] For this work, a 2 min furnace anneal in Ar ambient at 425 C was found to be sufficient in making ohmic contacts to nominally undoped GaAs. Highly doped GaAs samples (> 10 18 cm 3 carriers) did not require any annealing step after pressing on th e In contacts.

PAGE 63

63 Figure 2 1 S implified s chematic drawing of MOCVD chamber. Precursors (TMGa, TBAs) are carried with hydrogen gas into the growth chamber through a showerhead to evenly distribute the precursor molecules acro ss the wafer The wafer sits on a heated and rotating susceptor.

PAGE 64

64 Figure 2 2 Simulated diffraction pattern for GaAs along the [110] zone axis. The g 002 and g 00 2 spots are superlattice reflections and are sensitive to antiphase disorder in the GaAs crystal. Generated with SingleCrystal software, CrystalMaker Software Ltd, Oxford, England (www.crystalmaker.com). Figure 2 3 DF XTEM image of a GaAs layer on Si(001) made using the g 002 Bragg reflection. The contrast reversal in the GaAs layer across an APB (marked by arrows) can be observed under these conditions.

PAGE 65

65 Figure 2 4 Procedure for Fourier filtering a HAADF STEM image of a GaAs on Si film. A) Original HAADF STEM image. B) FFT of the original image with 002 and direct spots marked by arrows C) A mask is created using the 002 spots and the direct spot D) Finally, the in verse FFT is generated gi ving the Fourier filtered image. The misalignment of (002) planes at the location of an APB in clearly visible.

PAGE 66

66 Figure 2 5 BF XTEM images of a GaAs on Si(001) film before and after etching with HF /HNO 3 solution. A) As grown film. B) A fter etching for 10 s in the APB selective HF/HNO 3 solution The etch rate is enhanced at the location of APBs leading to the formation of pits. About 50 nm of GaAs is removed and the etch pits extend a further 100 150 nm into the film. Figure 2 6 HRTEM image of a developed etch pit at the site of an APB in a GaAs on Si layer demonstrating the selectivity of the HF/HNO 3 etchant for APBs. A stacking fault (SF) can be seen intersecting t he sidewall of the etch pit and is not preferentially attacked by the etchant relative to the APB.

PAGE 67

67 Figure 2 7 Example of tracing procedure of APBs in plan view SEM images. A) Original SEM image of GaAs on Si film with APBs. B) After tracing APBs using the freehand line tool in ImageJ software. The APB density is defined as the total line length of traced APBs divided by the area of the image giving units of m 1 Figure 2 8 BF XTEM image of a GaAs on Si sample after the removal of ~380 nm of GaAs using the non selective H 3 PO 4 /H 2 O 2 solution The surface of the layer is still relatively smooth and defects are not preferentially attacked.

PAGE 68

68 CHAPTER 3 THE EFFECT OF INTERFACIAL CONTAMINA TION 3.1 Background The prospect of integrating III V materials such as GaAs on Si for future transistors and other electronic devices is hindered by a number of growth related defects in the epitaxy of these compounds on Si. In particular, APBs can nuclea te at the GaAs/Si interface and propagate through the film, degrading electrical and optical properties [13] Single steps on the Si surface are well established to be a source of APB nucleation. However, the role of residual contamination on the Si surface on APB formation is not as well studied Contamination present on substrate surfaces is known to contribute to higher densities of stacking faults and dislocations in epitaxial layers from disruption of the growth kinetics [13, 32, 142] APBs have been found to be associated with the presence of amorphous interfacial impu rities through XTEM analysis [33] leading to speculation that these impurities may also nucleate APBs in addition to those nucleated by single steps. The exact relationship is still unclear and the effect of surface contamination on APB density has not been quantified. The ideal Si surface for epitaxy should be pristine and free of native oxide or oth er contaminants. In practice, it is exceedingly difficult to completely remove these contaminants, most commonly oxygen and carbon, from the surface. The most widely used method is to perform a dip of the Si wafer in HF to etch away the native oxide follow ed by a high temperature anneal at 1000 C in the growth chamber to desorb any residual oxide [38, 39] This procedure has the disadvantage of needing to expose the wafer to ambient after dipping in HF and before transferring the wafer to the growth chamber. The Siconi TM etch process [118] is a low temperature method of removing

PAGE 69

69 native oxide via a fluorinated plasma and can be performed under vacuum in a chamber attached to the growth chamber. The combination of the Siconi TM clean with a pre growth anneal near 900 C leads to a dramatic reduct ion in APB density for GaAs on Si films compared to conventional preparation. Figure 3 1 shows plan view SEM images of the APB densities for GaAs films grown on nominally on axis Si(001) by MOCVD with the Siconi TM clean only ( A) and the Siconi TM clean plus a 900 C pre growth anneal (B) The GaAs film thickness for both samples is ~450 nm. The samples were etched with a HF/HNO 3 solution to selectively delineate APBs. The APB density in Figure 3 1 A is comparable to the densities expected in films with conven tional preparation of Si substrates with no deliberate off cut to suppress APB nucleation [77] Th e low APB density in Figure 3 1B is much more advantageous for achieving complete suppression of antiphase disorder via APB self annihilation. It is especially notable that the low density was achieved on a Si(001) substrate with no intentional misorientation. Despite the promising resul ts from preparing Si substrates with the Siconi TM clean and pre growth bake, there is still a need to understand the role of substrate surface contamination on APB formation. The Siconi TM process provides a useful method for controlling the relative amount of residual contamination on the Si surface. The goal of this experiment is to examine the effect of interfacial contamination on APB density in GaAs grown on Si(001) by MOCVD and to develop a quantitative relationship. The work in this chapter has been p ublished in Barrett et al. [143] 3.2 Experimental A 300 mm ( 001 ) oriented Si wafer with no intentional off cut was utilized as t he substrate in this study. The substrate was cleaned prior to growth using the Siconi TM etch process [118] The clean was varied across the wafer in order to affect the relative

PAGE 70

70 amount of residual contamination from the wafer center to the wafer edge (Figure 3 2). M odulating the residual contamination on a single wafer ensures that specimens taken from different areas of the wafer experienced similar growth conditions A pre growth bake at ~9 00 C was then performed in the growth chamber under hydrog en ambient to help promote the formation of double steps. An Applied Materials III V MOCVD system was used for the growth of the GaAs film The precursors for Ga and As were trimethylgallium (TMGa) and tertiarybutylarsine (TBAs), respectively. GaAs growth was carried out using the conventional two step growth method for III V epitaxy to improve crystal quality [26, 75] The GaAs film was first grown at a low temperature of 530 C for ~200 nm followed by a high temperature layer grown at ~610 C for an additional ~200 nm The total film thickness was ~450 nm. Figure 3 3 shows a schematic diagram of the GaAs film structure. The cha mber was maintained at reduced pressure and the V/III ratio was in the typical process range for MOCVD. O xygen and carbon concentration depth profiles in the GaAs film were obtained using SIMS to confirm a variation in interfacial contamination across the wafer. Samples were selected from the wafer center to the wafer edge. The samples for SIMS were sputtered using Cs + ions. The depth resolution was ~1 nm. The amount of interfacial oxygen and carbon contamination for the samples was calculated by integrati ng the corresponding peaks at the GaAs/Si interface to obtain a dose. The APB density in the GaAs film was also measured across the wafer. Samples were taken in sequential order from the wafer center to wafer edge. The selective HF(49%):HNO 3 (69%):H 2 O (10:1 :7 by volume) etchant was used to delineate APBs prior

PAGE 71

71 to imaging in SEM The samples were dipped in the HF/HNO 3 solution at room temperature for 10 s Plan view SEM images were collected in various areas of the samples to obtain accurate counts of APB den sity. The APBs in SEM images were traced using ImageJ software [132] The APB density is defined as the line length of 1 Analysis of the GaAs/Si interface and associated defects was accomplished using XTEM. Samples for XTEM were prepared using a FEI Strata DB235 dual beam FIB/SEM HRTEM imaging of the GaAs layer and GaAs/Si interface was performed using a JEOL 2010F operating at 200 kV Additional microscopy with a nnular bright field scanning transmission electron microscopy (ABF STEM) and high angle annular dark field scanning transmission electron microscopy (H AADF STEM) imaging w as carried out using a probe aberration c orrected JEM ARM200cF operating at 200 kV. The resolution of STEM imaging on this instrument is ~ 0.78 . HAADF STEM images were acquired with a probe convergence semi angle of 11 mrad and collect ion angles of 76 174.6 mrad. 3.3 Results and Discussion 3.3.1 SIMS Measurements SIMS measurements of oxygen and carbon concentrations in the GaAs film were made for samples at the center, +6 cm from the center, +8 cm from the center, +10 cm from the center and edge of the wafer. The depth profiles were first normalized to the actual GaAs layer thickness as determined by XTEM. Figure 3 4 shows BF TEM images of the GaAs layer at the center of the wafer ( A) and the edge of the wafer (B ). The film thickness va ries from ~455 nm at the center to ~445 nm at the edge. Film thicknesses for intermediately located samples on the wafer were then linearly

PAGE 72

72 interpolated from this difference. There is also a notable difference in defect densities for the GaAs layer at the center versus the edge. Figure 3 5A and 3 5B show the interfacial peaks for the selected SIMS samples of oxygen and carbon, respectively The re is some widening of the peaks, i.e. distortion of the true spatial distribution of oxygen and carbon near the i nterface, due to limitations of the SIMS depth resolution particularly from ion beam mixing The peak oxygen concentration in Figure 3 5A at the center of the wafer is notably higher than the concentration near the middle of the wafer at +6 and +8 cm from the center. However, there is an overall trend of increasing oxygen concentration from center to edge and the peak concentration is greatest at the edge of the wafer. The interfacial carbon profiles determined by SIMS are shown in Figure 3 5B Carbon also has a trend of increasing peak concentration from center to edge. There is a noticeable shoulder to the carbon depth profile which likely stems from incorporation of the MOCVD precursor reaction products in the GaAs layer [144] Ho wever, there is still a distinguishable interfacial peak as highlighted by the inset in Figure 3 5 B that likely arises from residual carbon contaminants on the Si surface prior to growth This area will be used to determine the interfacial carbon dose. The SIMS measurements ind icate that the variation in the pre clean procedure was effective in changing the amount of residual oxygen and carbon contamination on the Si wafer surface from center to edge. Figure 3 6 shows the interfacial carbon and oxygen doses as a function of posi tion on the wafer. The interfacial doses are calculated by integrating the area under the respective interfacial peaks from Figure 3 5. This process helps to correct for distortion in oxygen and carbon distribution near the interface from the SIMS resoluti on.

PAGE 73

73 Oxygen and carbon contaminants are expected to be concentrated at the GaAs/Si interface with negligible dispersion into the GaAs layer. Error bars for the interfacial doses were obtained by shifting the position of the interface 15 nm from the depth d etermined by XTEM. The results of the calculated integrated doses corroborate the trend observed in Figure 3 5 of increasing amounts of oxygen and carbon interfacial contamination from center to edge of the wafer. The interfacial carbon dose is over two or ders of magnitude greater than that of the interfacial oxygen dose across the wafer. The oxygen dose increases from 7.3x10 11 atoms/cm 2 at the center to 4.0x10 12 atoms/cm 2 at the edge, while the carbon dose increases from 1.9x10 14 atoms/cm 2 at the center to 1.7x10 15 atoms/cm 2 at the edge. The carbon dose at the edge corresponds to ~3 monolayers (6x10 14 atoms/cm 2 ). 3.3.2 APB Density Measurements After determining that there was a variation in residual oxygen and carbon contamination across the wafer, the APB density was also measured at corresponding locations using the selective HF/HNO 3 etch. Figure 3 7 shows representative plan view SEM images of the APB density in the GaAs film at the center of the wafer ( A) +8 cm from the center ( B) and at the edge of th e wafer ( C) The APB density increases from 0. 1 to 3.2 1 over this distance and there is a marked increase of the average antiphase domain size from center to edge. The domains at the center of the wafer are on the order of a few hundred nanometers in diameter and the APBs demonstrate faceting along crystallogr aphic planes. The APBs at the edge of the wafer tend to curve and extend for relatively long distances such that it becomes difficult to distinguish individual domains.

PAGE 74

74 Figure 3 8 shows all of the APB density counts for samples measured in 2 cm increments from the center of the wafer to the edge. The increase in APB density corresponds well to the increase in interfacial contamination across the wafer. There is a large jump in the APB density from +6 cm from the center to +10 cm from the center of the wafer The APB density begins to plateau at the edge, suggesting that there is a maximum value that can be reached. This maximum value likely stems from the self annihilation of APBs once they are close enough to interact. 3.3.3 XTEM Characterization of Interface The GaAs/Si interface and associated structural quality of the GaAs layer was examined with XTEM. Samples were prepared from the center of the wafer, +8 cm from the center, and the edge of the wafer. Distinct amorphous particles were observed at the GaAs/Si interface only for the edge sample which had the highest oxygen and carbon interfacial doses. Figure 3 9 is a HRTEM image showing the spatial distribution of these particles along the interface in an edge sample. The particles were associated with the formation of several defects including stacking faults. Examining the particles at higher magnifications demonstrated their physical relationship with defects. Figure 3 10 shows a series of HRTEM images of amorphous interfacial particles. The part icles are a source of nucleation for stacking faults and twins. It is notable that in Figure 3 10D the stacking fault nucleated by the particle is terminated ~15 nm away from the interface. This image is strikingly similar to the observation made by Lilie ntal Weber et al. [33] in which a stacking fault was terminated by an APB above an amorphous layer The authors speculated that both the stacking fault and APB were nucleated by the amorphous layer. The termination of stacking faults and other defects at the location of APBs has been observed in additional studies [11, 130] This termination likely due to

PAGE 75

75 the high energy and degree of lattice disorde r associated with an APB. Therefore, it is plausible to conclude that th e stacking fault in Figure 3 10D is terminated by an APB. The close proximity of the APB with the interfacial particle, in turn, gives evidence for the nucleation of APBs by contaminan t particles. Further analysis of the GaAs/Si interface for a sample from the edge of the wafer was performed with STEM imaging. Figure 3 11 shows a region of the interface in ABF STEM ( A) and HAADF STEM ( B) An amorphous particle is clearly observed at t he G aAs/Si interface in Figure 3 11A (indicated by the arrow). It is difficult to tell whether the Ga As dumbbells are reversed in any region of Figure 3 11B Even though HAADF STEM has excellent Z contrast, Ga and As only have an atomic number difference of 3. After processing the HAADF STEM image in Figure 3 11 B via Fourier filtering with the 002 spots, the presence of an APB becomes evident (Fi gure 3 11C ). The (002) atomic planes are disrupted by ~ plane spacing at the location of the APB. The APB exten ds back towards the interface to the exact location of the amorphous particle. This observation corroborates the findings from HRTEM imaging (Figure 3 10 D ) indicating the direct nucleation of APBs by interfacial contaminant particles. These APBs would form in addition to those nucleated by single steps on the Si surface and contribute to an overall greater APB density in the GaAs layer compared to samples with lower amounts of interfacial contamination, e.g. the center of the wafer in this study. 3.3.4 Disc ussion of Source of Particles The chemical identity of the particles observed in XTEM, i.e. whether they are oxygen or carbon based, is still unclear. Energy dispersive spectroscopy (EDS) and electron energy loss spectroscopy (EELS) scans were inconclusive It is possible to gain further insight by performing a basic calculation of the expected spatial distribution

PAGE 76

76 of particles along the GaAs/Si interface if they were oxygen or carbon based particles (see Appendix A.1 for full calculation). First the typica l particle size is measured from the XTEM observations in Figure 3 10 and found to be ~20 nm long with a height of ~2 nm. The particles are assumed to be circular and so their total volume can be calculated. A further assumption is made that if the particl es are oxygen based they are SiO 2 and if they are carbon based then they are SiC. Using the known values of atomic density for SiO 2 and SiC, the number of oxygen and carbon atoms per particle can be determined. The oxygen and carbon interfacial doses for t he edge sample were 4.0x10 12 atoms/cm 2 and 1.7x10 15 atoms/cm 2 respectively, as determined by the SIMS analysis. Thus, the areal density of particles along the interface can be found by dividing the interfacial doses of oxygen and carbon by the corresponding number of atoms per particle. Finally, the expected spacing of particles along the interface can be obtained by multiplying the particle areal density by the thickness of the lamella produced by FIB for XTEM imaging of ~50 nm. If the particle s are carbon based SiC, the expected distribution is ~40 nm which is consistent with the observations from Figure 3 9. If the particles are oxygen based SiO 2 the expected spacing is ~17 m meaning that it would be fortuitous to observe even a single inter facial particle in a given XTEM sample (observable area is typically ~10 m or less). It appears that carbon is the only source of contamination that is present in significant enough quantity to account for the presence of the amorphous particles at the Ga As/Si interface in the edge sample. 3.3.5 Correlation or Causation? It is now concluded that the interfacial particles associated with the formation of APBs and other defects such as stacking faults in this study are likely composed of

PAGE 77

77 carbon. Figure 3 12 shows the APB density in samples from the center to edge of the wafer versus the corresponding interfacial carbon doses. There is a logarithmic correlation between APB density and interfacial carbon dose. The plateau in APB density after ~4x10 14 atoms/cm 2 of carbon may stem from an upper limit on the total APB density in the layer from the inherent self annihilation of neighboring APBs. Otherwise, it is plausible to assume that APB density would increase on an approximate linear trend with increasing inter facial carbon dose, i.e. the trend of APB density prior to ~4x10 14 atoms/cm 2 of carbon. This quantitative correlation between APB density in GaAs on Si films and interfacial carbon underscore the importance of reducing the amounts of residual contamination on substrates prior to epitaxy. Extrapolating back to zero APB density in Figure 3 12, interfacial carbon should ideally kept below a density of ~2x10 14 atoms/cm 2 Some amount of carbon contamination from incorporation of precursor reaction products or de sorption from the chamber walls during the pre growth bake, in addition to any residual carbon contamination after the Siconi TM clean, is unavoidable. Some APBs will undoubtedly form because of single steps on the Si surface, but lowering the amount of res idual contamination allows for suppression of antiphase disorder via APB self annihilation with minimal GaAs layer thickness. While there is a clear quantitative correlation between APB density and interfacial contamination, in addition to qualitative ass ociation in XTEM, it cannot be definitively stated that the contaminant particles are in fact nucleating APBs. It is possible that the amorphous particles disrupt the atomic ordering of the epitaxial GaAs film in a manner similar to single steps. However, an alternative explanation for the correlation is that the interfacial contaminant particles are associated with steps on the Si surface. Carbon is

PAGE 78

78 known to preferentially agglomerate at the site of steps on a Si(001) surface [145] It follows that if these steps wit h agglomerated carbon particles are single steps, they are essentially passivated, i.e. the dangling Si bonds are satisfied. Therefore these steps would be less likely to reconstruct into double steps during the pre growth bake. If GaAs growth then occurs in islands on either side of this passivated single step, an APB would form when the islands coalesce due to the rotation of the GaAs lattice. This APB would then be approximately located directly above the carbon particle leading to a structural associati on a s observed in XTEM (Figure 3 10D and Figure 3 11). In other words, carbon contamination on a substrate may not directly cause the nucleation of APBs, but regions with higher amounts of residual contamination will have a greater density of single steps and thus a greater APB density. Further study is needed to investigate the Si(001) surface at different stages of the pre growth processing, i.e. after the Siconi TM clean and after the pre growth bake. 3.4 Summary The effect of interfacial contamination on APB density was examined for GaAs grown on Si(001) by MOCVD. The pre clean was varied across the substrate to affect the amount of residual contamination from the wafer center to the wafer edge. SIMS analysis confirmed an increase in both interfacial oxyg en and carbon from center to edge. It was shown that there is a quantitative correlation between the APB density and the amount of interfacial contamination. Over a range of interfacial carbon dose determined by SIMS of 1.9x10 14 atoms/cm 2 to 1.7x10 15 atoms/cm 2 from center to edge, the APB density increased 1 Examination of the GaAs/Si interface with XTEM showed the physical association of APBs with amorphous interfacial particles. The particles are likely carbon based, such as SiC It is unclear if the higher

PAGE 79

79 APB densities are an effect of contaminant particles directly nucleating APBs or an indirect effect from agglomerated carbon passivating single steps on the Si surface. The results of this study demonstrate the importance of r educing residual contamination on Si substrates prior to GaAs epitaxy. If the APB density is minimized, then the complete suppression of antiphase disorder via APB self annihilation can be achieved with minimal GaAs layer thickness. In addition, this study shows that if the pre clean is not effective, the antiphase disorder will still be significant regardless of a pre growth bake of the substrate. When the same pre clean procedure is used, but no pre growth bake, the APB density is also relatively high (Fi gure 3 1). Therefore, the combination of an effective substrate clean and the pre growth bake is critical for reducing the density of nucleated APBs in GaAs on Si films.

PAGE 80

80 Figure 3 1 Representative plan view SEM images of G aAs film grown on Si(001) with varying pre growth conditions. A) Without a pre growth bake and B) w ith a pre growth bake at 900 C. APBs are visible as linear features with dark contrast after selective etching. Figure 3 2 Illustration showing the variation in pre cleaning procedure to affect the amount of residual contamination across the Si(001) wafer from center to edge.

PAGE 81

81 Figure 3 3 Diagram of the GaAs layer structure. Initial layer of 2 00 nm is first grown at a low temperature of 530 C followed by 200 nm at 610 C. The two step growth process is commonly used for III V epitaxial films to improve crystal quality. Figure 3 4 BF XTEM images of GaAs layer at different wafer positions. A) At the center of the wafer and B) at the edge of the wafer. There is a ~10 nm decrease in film thickness from center to edge and defect density is qualitatively higher at the edge.

PAGE 82

82 Figure 3 5 SIMS profiles of contamination sources at the GaAs/Si interface with varying wafer position. A) Oxygen interfacial contamination. B) Carbon interfacial contamination. There is an overall trend of increasing interfacial oxyg en and carbon from center to edge. The carbon profile has a large shoulder, likely from incorporation of MOCVD precursor reaction products; however, there is still a distinguishable interfacial peak shown in the inset. Figure 3 6 Interfacial oxygen and carbon doses determined by integrating the respective interfacial peaks in SIMS as a function of wafer position. Dose tends to increase from center to edge for both oxygen and carbon. Carbon dose is over two orders of magnitu de greater than oxygen dose.

PAGE 83

83 Figure 3 7 Representative plan view SEM images of the APB density in the GaAs film at different wafer positions. A) Center of the wafer, B) +8 cm from the center, and C) at the edge of the wafer The APB density increases from 0. 1 to 3.2 1 over this distance and there is a marked increase in antiphase domain size. Figure 3 8 Plot of all APB density counts as a function of wafer position. There is a consistent trend of increasing APB density from center to edge that correlates with the increase in contamination There is a large jump in APB density from +6 to +10 cm from the wafer center. The APB density appears to reach a maximum value towards the ed ge of the wafer.

PAGE 84

84 Figure 3 9 HRTEM image of GaAs/Si interface of a sample from the wafer edge. Amorphous particles are observed at the interface (marked by arrows). The particles are associated with the formation of defect s including stacking faults. Figure 3 10 HRTEM images at higher magnification of a series of amorphous particles observe d at the GaAs/Si interface A stacking fault in D) is terminated ~15 nm away from the interface (marked by arrow) which may correspond with the presence of an APB.

PAGE 85

85 Figure 3 11 STEM images of the GaAs/Si interface in an edge sample showing an APB associated with an amorp hous particle. A) ABF STEM image with amorphous particle marked by an arrow. B) HAADF STEM image of same area. C) The Fourier filtered HAADF image using the 002 spots reveals an APB that is directly emanating from the location of the particle, evidenced by the misordering of the (002) planes. Figure 3 12 Plot of APB density as a function of interfacial carbon dose for measured samples. APB density increases linearly with carbon dose and then plateaus after ~4x10 14 atoms/cm 2 The plateau may stem from an upper limit on the total APB density in the layer from self annihilation.

PAGE 86

86 CHAPTER 4 THE EFFECT OF GROWTH TEMPERATURE 4.1 Background The recent drive for alternative materials to replace Si in future transistor devices has renewed interest in the III V compound semiconductors. III V materials such as InGaAs and InAs have much better electron transport properties than Si [6] The long standing barrier to the implementation of III V materials in commercial appli cations has been cost. The renewed interest has prompted a need for economical integration of III V materials on Si. Perhaps the simplest method is to epitaxially grow a layer stack on top of a Si substrate. GaAs is a good candidate as an initial buffer la yer because it has an intermediate lattice constant between that of Si and InAs. Unfortunately, the epitaxy of GaAs and other III Vs on Si is typically plagued by a number of defects such as APBs. APBs in epitaxial GaAs films are nucleated at the site of single steps on the Si(001) surface. These APBs tend to nucleate along {110} planes that are the lowest energy of formation [92] and thus can propagate entirely through a film. The most common strategy for reducing the density of APBs is to use an intentionally misoriented substrate along with a high temperature pre growth anneal at 1000 C [13, 67] The anneal promotes the transition of single steps into double steps that reset the ordering of atoms in the epitaxial layer and prevent APB formation. However, the use of misoriented substrates is not practical in conventional CMOS processing because it would require a complete respecification of existing tools and procedures. High temperature anneals should also be avoided if possible to prevent unintentional effects to a ny other devices on the chip. Therefore, other methods of suppressing antiphase disorder in GaAs on Si films must be explored and understood.

PAGE 87

87 It has been observed that while APBs tend to nucleate along {110} orientations, they readily change their plane o f propagation in the layer during growth [82, 146] Thus, it is possible for the antiphase disorder in a film to be completely suppressed if all nucleated APBs kink over and self annihilate. Georgakilas et al. [11] analyzed the through thickness APB density for a GaAs on Si film and found that the APB density decreased further away from the GaAs/Si interface. It can then be inferred that APBs must be self annihilating via propagation along inclined planes, even if they are nominally higher formation energy. Possible APB habit planes that would be conducive for annihilation are {111}, {112}, and {113} planes. Figure 4 1 shows an illustration of these APBs in a GaAs on Si film. APBs along {111} planes are non stoichiometric, i.e. composed of all the same type wrong bond, and have the highest relative formation energy [92] On the other hand, APBs along {112} planes maintain stoichiometry, similar to {110} APBs, and so may be favorable for APB annihilat ion. One factor that may influence the tendency of APBs to kink over from {110} type planes to higher index planes is the growth temperature of the film. Increasing the thermal energy available should help drive the system to an equilibrium configuration. APBs are high energy defects and thus their existence in a film should be energetically unfavorable. There is evidence that higher growth temperatures lead to a greater proportion of APBs propagating along higher index planes and self annihilating. It was found for GaP films grown on Si via MOCVD that low bulk growth temperatures favored {110} APBs, but at higher bulk growth temperatures the APBs became majority {111} and {112} oriented and completely self annihilated [113, 114] Unfortunately, these studies did not quantify the effect of growth temperature on APB annihilation. It is also

PAGE 88

88 unclear why APB propagation along orientations that are nominally higher in formation energy becomes favorable at these higher growth temperatures. It is plausible that there is a kinetic barrier to APB propagation along higher index planes, but that the process is thermodynamically favored overall due to the consequential reduction in total APB interfacial area. The goal of this work is to conclusively analyze the effect of growth temperature on APB annihilation in GaAs grown on Si(001) by MOCVD and t o gain insight on the associated APB energetics. The work in this chapter has been published in Barrett et al. [147] 4.2 Experimental The substrates used in this work were 300 mm Si(001) wafers with no intentional off cut. The wafers were cleaned of native oxide and other contaminants using the Siconi TM process [118] and then transferred under vacuum to the growth chamber. Each wafer was then subjected to a ~900 C bak e to help promote the formation of double steps. This preparation procedure reduces the APB density in GaAs films by over an order of magnitude (see Chapter 3). Starting with a low APB density is advantageous both from an engineering standpoint and for thi s study since it becomes easier to track changes in APB density over the thickness of the film. GaAs film growth was performed in an Applied Materials III V MOCVD system. Layers were grown in the conventional two step manner to improve crystal quality [26, 75] The precursors for Ga and As were trimethylgallium (TMGa) and tertiarybutylarsine (TBAs), respectively. The V/III ratio was in the typical process range for MOCVD and the chamber was maintained at standard reduced pressure. Figure 4 2 shows the layer structure used for the GaAs films in this study. First, an identical nucleation layer of 200 nm of GaAs was grown at 530 C. This initial layer was important to set the same

PAGE 89

89 approximate starting APB density across all of the samples. N ext, a bulk layer of 600 nm of GaAs was grown for temperatures ranging from 530 C to 650 C to analyze the effect of increasing the growth temperature on APB propagation and annihilation. The sample grown completely at 530 C is used as a control. Growth at higher temperatures was not feasible due to cracking of the GaAs film upon cooling. No other growth conditions besides growth temperature were changed, i.e. this study only investigates cknesses for the samples varied over a small range from the nominal 800 nm thickness due to differences in the growth rate at different bulk growth temperatures. In order to analyze the through thickness APB density for the samples, a two part etching process was used similar to that discussed by Georgakilas et al. [11] A non se lective CH 3 OH:H 3 PO 4 (85%):H 2 O 2 (30%) (10:1:1) solution was first used to etch the GaAs films down to different depth points. Th e remaining film thickness at each point was measured with a JA Woollam M88 ellipsometer The ellipsometry measurements were compared with selected XTEM samples and found to be in good agreement with less than 10 nm of variation. Samples were then dipped i n a HF(49%):HNO 3 (69%):H 2 O (10:1:7) solution for 10 s to selectively etch down APBs and improve contrast in SEM. Thus, the APB density as a function of depth in the GaAs films could be determined. Plan view SEM images were collected in various areas of the samples to obtain accurate counts of APB density. APB features in SEM were traced using ImageJ software [132 ] The APB density is defined as the line length of APBs per unit area, 1 The detection limit of antiphase domains is on the order of 0.1 to

PAGE 90

90 APB lengths that were collected. The evolution of APB propagation in the samples was analyzed with DF XTEM imaging using a JEOL 200CX TEM operating at 200 kV XTEM specimens were prepared in <110> orientations using a FEI Strata DB325 dualbeam FIB/SEM instrument. DF imaging was performed wit h the 002 and 00 2 superlattice reflections in order to have sensitivity for crystal orientation across antiphase domains [129, 148] The contrast in DF imaging was improved by further tilting of the samples off of the <110> zone axis to excite 115 and 11 7 type spots along with the 002 type spot. These tilting conditions increase the intensity of the 002 reflection [113] Using these DF imaging conditions, antiphase domains will have opposing contrast relative to the main phase crystal, i.e. one will be bright and the other will be da rk. This contrast allows for the direct identification of APBs and their habit planes. 4.3 Results 4.3.1 APB Density Depth Profiles After etching the samples down to different film thickness and gathering a series of images of the APB density in SEM, it was readily apparent that number of antiphase domains visible, along with the overall APB density, began to increase with decreasing GaAs film thickness. An example of this trend can be seen in the sample with the bulk layer grown at 570 C. Figure 4 3 sho ws representative plan view SEM images of the surface of this sample after selectively etching APBs at different film thicknesses. The type of discrete domains was typical for all samples. The decrease in the number of antiphase domains visible with increa sing distance from the GaAs/Si interface indicates that the bounding APBs must be kinking over and self annihilating, thereby suppressing

PAGE 91

91 the domains within the film. The APB density from a film thickness of 240 nm ( D) to a film thickness of 745 nm ( A) decreases from 0.33 1 to 0.047 1 Similar trends were observed in the other samples with different bulk growth temperatures. The through thickness APB density depth profiles for all samples with varied bulk growth temperature are shown in Figure 4 4. Note the logarithmic scale used for APB density. There is a marked decrease in APB density for all samples with increasing distance from the GaAs/Si interface. This decrease in APB density is facilitated by a reduction in the total line length of APBs i n the (001) growth plane due to APBs kinking to inclined planes and self annihilating. The control sample grown completely at 530 C also experienced a moderate decrease in APB density, indicating that there is an inherent tendency for APBs to self annihil ate during growth. However, there is a variation in the magnitude of APB density reduction with GaAs film thickness that appears to be temperature dependent. Increasing the bulk growth temperature to 570, 590, and 610 C causes an increase in the rate of A PB annihilation with respect to film thickness. After this point, i.e. for the samples grown at 630 and 650 C, the APB density depth profiles follow the same approximate curve. Thus, there is no further enhancement of APB annihilation for bulk growth temp eratures greater than 610 C. 4.3.2 APB Annihilation Rates A quantitative measure of APB annihilation rate for the samples grown at different bulk growth temperatures can be obtained by fitting the respective APB density depth profiles in Figure 4 4. The se profiles follow an approximate exponential decay, as evidenced by the linear relationships on a semi log plot. This character allows for a trivial determination of the APB annihilation rates. Figure 4 5 shows an example of the fitting procedure for the sample with bulk growth at 610 C. The profile is fit to a

PAGE 92

92 standard exponential decay function and the APB annihilation rate is taken as the exponential decay rate. The units of APB annihilation rate are then 1 This procedure was repeated for the depth profiles of all samples to extract the respective rates. Figure 4 6 shows the extracted APB annihilation rates with respect to film thickness for all samples versus the bulk growth temperature. The APB annihilation rate increases by approximately a facto r of four from 2.6 1 to 10.7 1 over a bulk growth temperature range of 530 C to 610 C. After 610 C, the annihilation rate saturates at ~10 11 1 corresponding to the overlap of APB density depth profiles in Figure 4 4. The strong correlation be tween APB annihilation rate and the bulk growth temperature, combined with the knowledge that APB annihilation occurs via APBs kinking to higher index planes, indicates that the kinking of APBs is a thermally activated process. At higher growth temperature s, a greater portion of APBs are able to change their plane of propagation and self annihilate. The saturation of APB annihilation rate at growth temperatures above 610 C may stem from geometric factors. The film thickness required for an APB to self anni hilate is dependent on its plane of propagation. Therefore, if a given set of planes is preferred energetically to achieve annihilation, e.g. {112} over {111}, and a majority of APBs are already kinked over to these planes, then there will be no significan t change in the APB annihilation rate with respect to film thickness at higher temperatures. An Arrhenius relationship can be used to determine the associated activation energy for APBs kinking to higher index planes during growth. Figure 4 7 is an Arrhen ius plot of the APB annihilation rates for the samples with bulk growth from 530 C to 610 C, i.e. before the saturation of APB annihilation rate. The activation energy

PAGE 93

93 found from fitting this plot is 1.1 eV. This activation energy likely corresponds to t he energy barrier for APBs to form along a higher index plane configuration, e.g. {112}, versus a {110} configuration. APBs along {112} planes have a higher wrong bond density compared to {110} APBs [92] It is unknown how this activation energy compares to the activation energy for the GaAs growth process, i.e. growth rate. However, based on previous literature for the MOCVD growth of GaAs using TMGa and TBAs under similar conditions, the growth rate acti vation energy is likely between 0.5 to 1.5 eV [29, 30, 149] In addition, the consistency of the film thicknesses at different bulk growth temperatures indicates that the growth process is not kinetically limited Th us the process of APB kinking is unlikely to be affected by GaAs fi lm growth reactions. 4.3.3 XTEM Analysis The evolution of APBs in the GaAs layer was investigated using DF XTEM imaging. Due to the overall low density of APBs in the samples, it is difficult to capture entire domains in a single XTEM sample. However, a f ew examples of antiphase domains were found in the samples that were imaged. Figure 4 8 shows an antiphase domain in the sample with bulk growth at 630 C imaged using the g 002 reflection ( A) and the g 00 2 reflection ( B) The reversal of contrast in the im age from Figure 4 8A to Figure 4 8B using the opposing reflections confirms that the feature is an antiphase domain. The size of the domain is on the order of those observed in plan view SEM (Figure 4 3). It can be observed that, while the domain is not co mpletely suppressed in the film, the diameter decreases with increasing distance from the GaAs/Si interface. This observation is consistent with the quantitative trend of the APB density depth profiles (Figure 4 4). The APBs on either side of the domain ap pear to be faceted both along {110} planes and higher index planes leading to a net decrease in width.

PAGE 94

94 Another example of an antiphase domain in DF XTEM is shown in Figure 4 9 for the sample with bulk growth at 610 C. The image contrast in the GaAs laye r is again reversed when using the g 002 reflection ( A) and the g 00 2 reflection ( B) confirming that this is an antiphase domain. This domain is suppressed within the GaAs layer. The APBs appear to nucleate along {110} planes before eventually kinking over to what are macroscopically {112} planes and self annihilating. This observation i s consistent with the work by Volz et al. [114] and Beyer et al. [95] that showed preferential propagation along {112} planes a t higher growth temperatures for APBs in GaP on Si grown by MOCVD. It is important to note that Figure 4 9 is not necessarily representative of typical APB annihilation for GaAs on Si, but it is an example of an energetically allowable pathway. The kinking of APBs from {110} planes to {112} planes as seen in Figure 4 9 is likely limited by a kinetic barrier, i.e. the 1.1 eV energy barrier found for the relationship between APB annihilation rate and growth temperature (Figure 4 7). 4.4 Discussion 4.4.1 Driv ing Force for Annihilation The results of the through thickness APB density analysis for the GaAs on Si films in this study demonstrate that APB annihilation during growth is dependent on the growth temperature. This relationship is not surprising given t hat APBs are defects with large associated energies and therefore should be inherently unstable in a crystal. The equilibrium configuration for a GaAs crystal should be free of APBs. However, APBs are forced to form because of single steps on the Si surfac e. At lower growth temperatures, APBs nucleate along {110} planes that are lower formation energy. These APBs would otherwise propagate entirely through a film. Increasing the growth temperature helps to overcome a kinetic energy barrier for APBs to kink o ver to higher index planes, such as

PAGE 95

95 {112}. In turn, this configuration leads to the self annihilation of APBs and a total reduction in APB density. The energetic driving force for APBs kinking during growth can be understood using a simple schematic diagr am of atoms in the (001) growth plane, as seen in Figure 4 10. A typical annihilating antiphase domain is shown in cross section in Figure 4 10A similar to the domains seen in DF XTEM (Figure 4 8 and Figure 4 9). Two (001) slices are taken at a point wher e the APBs lie along {110} planes ( B) and where they lie along inclined planes ( C) Each atom depicted in Figure 4 10B and Figure 4 10C is involved in an APB. The kinking of APBs to inclined planes facilitates a reduction in the number of atoms involved in an APB with each successive monolayer of growth. A significant reduction in the contribution of APBs to the overall free energy of the crystal can then be achieved, as seen by comparing the total APB line length in Figure 4 10B versus Figure 4 10C The do main will eventually be completely suppressed within the film. 4.4.2 Discussion of Mechanism Another explanation for the dependence of APB annihilation rate on growth temperature is not a pure kinking of APBs to higher index planes such as {112}, but a ste pwise kinking of APBs mediated by jumps to {110} planes. In addition to showing the preferential annihilation of APBs along {112} planes [95] Beyer et al. [131] observed that in similarly grown GaP on Si samples APBs oriented along {110} planes displayed a tendency to jump one or more atomic planes, ther eby creating a finite thickness to the APB. It follows that increasing growth temperature can increase the jump probability for APB planes and enhance the annihilation rate. However, the jumping mechanism inherently forms segments of APB facets that should be energetically unfavorable, e.g non stoichiometric {111} or {001}. APBs along {001} also have greater wrong bond

PAGE 96

96 density than {110}, {111}, or {112} [91] It is possible that these segments are relatively smal l enough in length that the APB jump is still rendered favorable. Otherwise, it is unclear how a macroscopic {112} APB as observed in this study ( Figure 4 9 ) would be manifested via the stepwise kinking mechanism. Considering the geometry of an APB in a {1 12} orientation, construction with {111} and {110} facets would never meet the shallow angle of a {112} plane on average. Construction with {110} and {001} facets requires that the unfavorable {001} segments must be longer than the {110} segments. APB annihilation may also be influenced by rotation of the lattice due to the differing lengths of Ga Ga and As As wrong bon ds versus Ga As bonds. Vajargah et al. [130] measured phase changes HAADF STEM images of a GaSb on Si layer and observed rotation of the lattice at an APB. The rotation alterna ted in sequential segments which allowed the APB to initially preserve a normal direction relative to the substrate before the APB eventually kinked over to inclined planes. The authors posit that faceting along higher index planes is influenced by this la ttice rotation and local relaxation at the APB and thus it will promote annihilation. However, since the alternating rotation was seen in both the curved and normal sections of the APB, it is unclear how much effect the rotation has on the actual APB kinki ng process. As well, the change in APB formation energies along different planes due to the lattice rotation and relaxation is unknown. There may be a relationship between the local relaxation and lattice rotation at an APB and the tendency to kink to high er index planes. 4.4.3 Comparison with Other Samples The results of the APB density depth profiles for the samples in this study can be compared to those of samples grown at different times. Figure 4 11 shows a series of APB density depth profiles for sam ples with different bulk growth temperatures, along

PAGE 97

97 with the profile of the sample with bulk growth at 610 C from this work. The other samples include two proof of concept (POC) samples with bulk growth at 590 C and 630 C; the growth conditions are othe rwise identical to the samples in this study. Another sample is shown that was grown with a 400 nm nucleation layer and 600 nm bulk layer at 610 C for a total GaAs film thickness of 1 m; the growth conditions are otherwise identical. Note the shifted pos itioning of the bulk layer demarcation for this 1 m sample as well as the lower initial APB density prior to the onset of the bulk layer in Figure 4 11. The POC sample with bulk growth at 590 C experiences much slower APB annihilation with respect to GaA s film thickness compared to the other samples which is consistent with the findings of this work. What is also interesting to point out is that the APB annihilation rates for the POC sample at 630 C and the 1 m sample at 610 C are similar to that of th e sample with 610 C bulk growth from this study, i.e. they follow the same approximate slope of decay. The similarity of the APB annihilation rates for the samples with bulk growth at 610 C and above demonstrate that there is a consistent saturation poin t for the APB annihilation rate. Further, the similarity of the rate of decay for the 1 m sample shows that the starting APB density prior to the growth of the bulk layer does not affect the APB annihilation rate. 4.5 Summary In conclusion, the annihilation rate of APBs with respect to film thickness for GaAs films grown on Si(001) by MOCVD has been shown to have a dependence on the bulk growth temperature. Increasing the bulk growth temperature from 530 C to 610 C increased the exponential decay rate of APB density with respect to GaAs film thickness 1 1 The annihilat ion rate saturated after 610 C. Higher growth

PAGE 98

98 temperatures likely remove kinetic barriers for the kinking of APBs from {110} type planes to higher index planes, such as {112}, that are nominally higher in formation energy but facilitate self annihilation. An activation energy for the APB annihilation process during growth of 1.1 eV was found using an Arrhenius relationship. The driving for ce for APBs kinking to higher index planes stems from a reduction in the total line length of APBs in the (001) growth plane, i.e. a reduction in the number of adatoms involved in an APB for each subsequent monolayer of growth. DF XTEM imaging showed a ten dency for antiphase domains to shrink in width during growth and that {112} is a possible habit plane for APB propagation to facilitate annihilation at higher growth temperature. These findings are consistent with prior studies showing a preference for APB propagation along {112} planes over {110} at higher growth temperatures. This work demonstrates a quantitative relationship between APB annihilation rate and growth temperature.

PAGE 99

99 Figure 4 1 Illustration of APBs oriented a long various habit planes in a GaAs on Si film. Inclined habit planes such as {111}, {112}, and {113} favor APB self annihilation. Note the stoichiometric nature of {110} and {112} type APBs. Figure 4 2 Illustration of GaA s layer structure for the growth temperature experiment. A 200 nm layer was first grown at 530 C for all samples to set the APB density. Next a 600 nm bulk layer was grown with varying temperature from 530 to 650 C

PAGE 100

100 Figure 4 3 Representative plan view SEM images of the APB density in the sample with bulk growth at 570 C at varying film thickness points: A) 745 nm, B) 550 nm C) 400 nm and D) 240 nm. There is a marked increase in APB density and th e number of domains visible closer to the GaAs/Si interface, indicating that APBs are annihilating during growth and suppressing antiphase domains

PAGE 101

101 Figure 4 4 Depth profiles of APB density for all samples with varying bul k growth temperature. APB density for all samples decreases with increasing distance from the GaAs/Si interface. The rate of decrease with respect to film thickness increases with increasing bulk growth temperature, but plateaus after 610 C. Figure 4 5 Example of fitting procedure for APB density depth profile curves for the sample with bulk growth at 610 C. The curves follow an approximate exponential decay. The APB annihilation rate is taken as the exponential decay rat e.

PAGE 102

102 Figure 4 6 Plot of the extracted APB annihilation rates, i.e. exponential decay rate, for all samples versus the bulk growth temperature. The APB annihilation rate increases by ~4x from 530 to 610 C before it plateaus The annihilation rate is likely limited by the APB habit planes at higher growth temperatures. Figure 4 7 Arrhenius plot of APB annihilation rates for samples with bulk growth from 530 to 610 C. The extracted activation energy is 1.1 eV which likely corresponds to the energy barrier for APBs to kink to inclined habit planes during growth and self annihilate.

PAGE 103

103 Figure 4 8 Pair of DF XTEM images showing an antiphase domain in a sample with bulk growth at 630 C A) DF image using the g 002 reflection and B) DF image using the g 00 2 reflection The reversal of contrast from A to B demonstrates that this feature is an antiphase domain. There is a net decrease in domain width with increasing GaA s film thickness. Figure 4 9 Pair of DF XTEM images showing an antiphase domain in the sample with bulk growth at 610 C A) DF image using the g 002 reflection and B) DF image using the g 00 2 reflection. The reversal of co ntrast from A to B demonstrates that this feature is an antiphase domain. The domain is suppressed within the film due to the self annihilation of APBs along {112} planes.

PAGE 104

104 Figure 4 10 Illustration of an annihilating antiphase domain A) Cross section of the domain, B) arbitrary (001) section of APBs in the bulk of the domain, and C) arbitrary (001) section of APBs in the annihilating portion of the domain. The kinking of APBs leads to a net reduction in the number of atoms involved in an APB for (001) sections from B to C Figure 4 11 Comparison of APB density depth profiles for samples grown at different times with similar conditions as a proof of concept (POC) and with different lay er thicknesses (1 m sample) compared to data from this study The APB annihilation rates for the high temperature bulk growth samples ( 610 C) are qualitatively similar which demonstrates that there is a consistent saturation point for the annihilation rate, likely stemming from geometric factors.

PAGE 105

105 CHAPTER 5 THE EFFECT OF POST GROWTH ANNEALING The following chapter discusses the effect of post growth annealing on APBs in GaAs films grown on Si. The chapter is broken down into four major sections. The first section discusses the main experiment regarding the annealing effects on APBs in GaAs on Si samples with standard preparation and growth conditions, i.e. Siconi TM clean, pre growth bake, and two step growth process. The second section discusses the r esults of annealing samples with varying substrate preparation and growth procedures. The third section discusses the results of varying the annealing conditions on APBs, specifically the type of surface protection used. The fourth and final section will d iscuss the effect of APB density on Si diffusion from the substrate during annealing. Samples for annealing were first encapsulated with Al 2 O 3 deposited by ALD. Unless otherwise stated, the recipe for Al 2 O 3 deposition was a hybrid exposure/plasma mode using water as the precursor for oxygen for the first 0.5 nm (5 cycles) and then oxygen plasma as the oxygen source for the remainder of film growth. The full recipe is given in Appendix A.2. Films thicker than 10 nm made using water as the precursor for the entirety of film growth are prone to having trapped hydrogen [150 152] The trapped hydrogen can bubble and cause blistering and delamination of the film at elevated temperatures, i.e. during annealing. 5. 1 Dissol ution of Antiphase Domains by Annealing 5.1.1 Background GaAs on Si will play a vital role in the implementation of III V materials in future electronic devices. Unfortunately, the crystal quality of GaAs layers grown epitaxially on

PAGE 106

106 Si is typically much worse than bulk GaAs. Threading dislocations from the lattice mismatch and APBs from single steps on the Si surface degrade the electronic and optical properties of the GaAs layer. Bulk GaAs wafers are too expensive and too small relative to Si wafers for use in large scale commercial applications. There are avenues for reducing the density of these defects during the growth of the GaAs layers. For threading dislocations, strained layer superlattices [46 51] and growth on patterned substrates [52 59] known as aspect ratio trapping, have been shown to be effective. For APBs, careful preparation of the Si substrate can reduce the nucleated density [143] and further tuning of the growth temperature can encourage self annihilation [113, 147] In addition to changing the growth conditions, defect densities can be reduced via post growth processing. Annealing is a common method to reduce defect densities in epitaxial layers, including GaAs on Si, and to improve the overall crystal quality. Post growth annealing by furnace or RTA has seen extensive use in the semiconductor research field and industry. RTA is used most often over conventional furnace annealing when there are thermal budget restrictions for processing, such as for the activatio n of dopants without excessive diffusion of the dopant species. However, both techniques can be equally effective in the reduction of grown in defects for GaAs grown on Si by MBE or MOCVD. The elevated temperatures allow for threading dislocations and atom s in stacking faults and twins to become mobile. These defects add to the free energy of the crystal, so their elimination should be thermodynamically favorable. Furnace annealing in the range of 800 to 850 C for 10 to 30 min leads to lower densities of d islocations, stacking

PAGE 107

107 faults, and twins in GaAs on Si layers [153 156] RTA is able to accomplish similar decreases in defect densities for treatments at 800 to 950 C for 10 s [83, 157, 158] Further studies on the activation energy for threading dislocation reduction has also been performed for GaAs on Si. Yamaguchi et al. [159] interrupted the GaAs growth and performed in situ anneals, i.e. in the growth chamber, at 700 to 900 C under As overpressure and found an activa tion energy for dislocation density reduction of 1.35 eV. Ayers et al. [138] performed post growth in situ anneals from 675 to 900 C under As overpressure and found an associated activation energy for threading dislocation density reduction of 1.91 eV. The effect of annealing on APBs in III V films, including GaAs, on Si is not as well studied. Chu et al. [115] found that antiphase domains grew, i.e. APB density was reduced, for highly defective GaAs on Si films for furnace anneals from 70 0 to 800 C for 2 hrs. No attempt at calculating an activation energy for APB migration was made. Guo et al. [116] investigated the evolution of APBs in GaP films grown on Si using X ray methods for a single anneal at 600 C for 10 min. The effect of anneal temperature on APB density was not quantified. The authors concluded that APBs reconst ructed from grown in {111} type orientations to {110} planes. However, this result would indicate that APBs can never be eliminated in III V films. The goal of this study is to systematically examine the effect of post growth annealing on APBs in GaAs grow n on Si and determine the associated energetics of APB mobility. 5.1.2 Experimental Samples for this study were taken from a single GaAs on Si wafer to ensure that all had experienced similar growth conditions prior to annealing. A 300 mm Si(001) wafer wi th no intentional misorientation was used as the substrate for GaAs growth. The

PAGE 108

108 Si wafer was cleaned using the Siconi TM process [118] prior to transfer to the growth chamber. A pre growth bake at ~900 C was then performed in the growth chamber to he lp promote the formation of double steps. This substrate preparation procedure is effective in reducing the density of nucleated APBs [143] GaAs growth was performed in an Applied Materials III V MOCVD reactor. The precursors for Ga and As were trimethylgallium (TMGa) and tertiarybutylarsine (TBAs), respectively. The V/III ratio was within the standard MOCVD process range and the chamber was maintained at typical reduced pressure ambient during film growth. GaAs was grown in a standard two step process. The layer structure is similar to that shown in Figure 3 3. An initial layer of 200 nm of GaAs was grown at 530 C followed by an additional 200 nm grown at 610 C. The higher growth temperature layer has the added benefit of accelerating APB annihilation via kinking of APBs to inclined planes [147] For the annealing treatments, samples were encapsulated with 20 nm of Al 2 O 3 deposited at 200 C by ALD to protect against surface degradation. Figure 5 1 shows a BF XTEM image of a GaAs on Si sample annealed at 600 C for 30 min using this encapsulant. The surface is still very smooth and free of any visible pits after annealing and removing the cap. Anneals for this study were performed at temperatures from 550 to 700 C for a range of times. Longer anneals (> 5 min) were performed in a tube furnace with Ar ambient and shorter anneals (< 5 min) were performed in RTA with Ar ambie nt Test anneals with overlapping times were done to ensure consistency between the furnace and RTA. After annealing, the Al 2 O 3 caps were removed by dipping the samples in concentrated pure HF(49%) for 30 s.

PAGE 109

109 The APB densities in as grown and annealed samp les were determined by selective etching of APBs using HF(49%):HNO 3 (69%):H 2 O (20:1:7) for 10 s and then imaging in SEM. APB features in SEM images were traced using ImageJ software [132] The APB density is defined as the total APB line length measured of the total observed surface area, giving units of m 1 Depth profiling of APB density was also conducte d. Samples for depth profiling were simultaneously annealed side by side for each anneal condition. After removing the Al 2 O 3 caps, a non selective etchant of CH 3 OH:H 3 PO 4 (85%):H 2 O 2 (30%) (10:1:1) was used to etch the GaAs layer in the samples down to various thicknesses. The remaining GaAs film thickness was measured using a JA Woollam M88 ellipsometer. Ellipsometry measurements were in good agreement with select film thickness measurements by XTEM with less than 10 nm of variation. The selective HF/HNO 3 was then used to delineate the APBs for imaging in SEM. 5.1.3 Results and Discussion A nnealing treatments had a noticeable effect on the density of APBs at the surface of the GaAs layer. Figure 5 2 shows representative plan view SEM images of the GaAs surface after etching with the APB selective HF/HNO 3 solution for an as grown sample ( A) and a 650 C, 7.5 min furnace annealed sample ( B) The APB density 1 1 after annealing and there is also a clear decrease in the number of antiphase domains present in a given area. The corresponding histograms of domain size, i.e. APB loop length for a domain, for all domains counted in the as grown sample and the sample an nealed at 650 C for 7.5 min are shown in Figure 5 2C and Figure 5 2D respectively. There is a significant

PAGE 110

110 decrease in the average domain size after annealing. It can then be inferred that the APBs must be mobile during the anneal and that the net motion is towards their centers of curvature. This motion eventually leads to self annihilation. Figure 5 3 shows the surface APB densities of the GaAs on Si samples for all annealing conditions. There is a clear and consistent trend of decreasing APB density wi th increasing anneal time and temperature. For the higher temperature anneals at 650 C and 700 C the APB density decreases very rapidly but plateaus at a nonzero value. These low densities represent an average of one to two antiphase domains present in a 2 It is unclear why some domains are still present for relatively long and high temperature anneals. Regardless, over 90% of domains at the surface are eliminated and the APB density is reduced by an order of magnit ude. In order to gain further insight on the evolution of APBs during post growth anneals, depth profiles of the APB density in GaAs on Si samples were obtained for selected anneal conditions of varying time and temperature. Figure 5 4 shows the APB densi ty depth profiles for these annealed samples along with an as grown sample. In the as grown sample, the APB density starts very high near the GaAs/Si interface but then drops significantly with the onset of the higher temperature bulk growth layer. This de crease is consistent with the tendency of APBs to kink over from vertical {110} type planes to inclined higher index planes, e.g. {111} or {112}, and self annihilate at higher growth temperatures [114, 147] After annealing at 600 C for 10 min, it is observed that the APB density closer to the surface of the GaAs layer begins to decrease relative to the as grown sample. In the middle of the GaAs layer, the APB density remains

PAGE 111

111 appro ximately the same. Thus, it is hypothesized that APBs begin self annihilating from the surface down during annealing. Further annealing for longer time (30 min) or at a higher temperature (650 C) reduces the APB density throughout the GaAs layer. The the rmally activated motion and subsequent annihilation of APBs observed for GaAs on Si in this study is similar to the motion of APBs that has been investigated in metals and metal oxides. According to the theory established by Allen and Cahn [110] APBs in binary alloys evolve to reduce their curvature and the density of APBs changes with annealing time as t 1/2 This time dependence of APB densit y with annealing has been observed experimentally for epitaxial Fe 3 O 4 films grown on MgO [112] and for bulk Cu Pd alloys [160] A similar treatment of the APB density evolution data in this study is applied to determine the activation energy of APB motion in GaAs on Si. For simplicity, the inverse square of the surface APB density (see Figure 5 3 ) is plotted against the annealing time. The rate constant for APB motion, k, for the Arrhenius equation k=k 0 *exp( E a /k B T) is extracted from the slope a 2 /s. Figure 5 5A shows this procedure for the APB density versus annealing time data at 700 C. This procedure was repeated for the surface APB density data for all other annealing temperatures. Figure 5 5 B shows the Arrhenius plot of the natural log of the extracted rate constants versus inverse annealing temperature. From the slope of this curve, the activation energy is determined to be 3.8 eV. While it is not immediately clear what atomic exchange proces s facilitates the movement of an APB during annealing, a plausible mechanism can be proposed. The requirements for APB motion involve the breaking of the original antisite bond, either Ga Ga or As As, and then the diffusion of the Ga or As atom to an adjac ent antisite.

PAGE 112

112 This process must also occur nearly simultaneously for numerous antisite bonds for the entire APB plane to move. The activation energy of 3.8 eV determined in this study is very close to the activation energy that has been reported for Ga sel f diffusion via Ga vacancies in GaAs of ~4 eV [161 163] The self diffusion of As in GaAs is not as well studied, but is also believed to occur via vacancies [164] Correspondingly, there is a greater thermal equilibrium concentration of vacancies at elevated temperatures. It follows that the motion of an APB during annealing is likely enabled by the presence of nearby vacancies and that the rate limiting step is the diffusion of Ga atoms. Th e APBs will have a tendency to migrate towards their centers of curvature in order to reduce the total APB interfacial area in the GaAs layer and consequently lower the overall free energy of the system. Figure 5 6 illustrates the possible mechanism for th e movement of an APB segment from a {110} configuration to {111} from the diffusion of a Ga vacancy. It should also be noted that this activation energy for APB diffusion in GaAs on Si is much higher than the activation energy previously determined for APB kinking during growth of 1.1 eV [147] Thus, the reduction of APB density during film growth is expected to be dominated by APB kinking from adatoms aligning along higher ind ex planes, rather than the diffusive motion of existing APBs. Both processes have a large energetic driving force to reduce the APB interfacial area and decrease the proportion of antiphase to main phase in the GaAs layer. These findings illustrate the nee d for additional computational analysis of APB systems in GaAs and other III V materials. Future work should also investigate the effect of in situ annealing with an As overpressure and whether a similar reduction in APB density would occur.

PAGE 113

113 5.1.3 Summary In summary, it was shown that APBs in GaAs grown on planar Si(001) can be removed by ex situ post growth annealing. APB evolution at the surface of the GaAs layer followed the t 1/2 time dependence established by Allen and Cahn [110] At annealing temperatures of 650 C and above, the surface APB density was decreased by an order of magnitude in less than 10 min. The activation energy for APB mi gration at the surface during annealing was determined to be 3.8 eV. The motion of APBs is believed to be limited by Ga diffusion via vacancies. Depth profiles of APB density for annealed samples confirmed that the dissolution process starts at the surface and that the APBs were being removed throughout the GaAs layer. The energetic driving force for the migration of APBs can be understood by a reduction in the total APB interfacial area. Since it is difficult to prevent the nucleation of APBs on Si(001) wa fers with no intentional off cut or high temperature pre growth treatments, the ability to greatly reduce the grown in APB density with relatively little additional processing is advantageous. 5.2 Effect of Sample Conditions 5.2.1 Background The results d iscussed in section 5.1 are considered to be for a low starting APB density sample with standard growth conditions, i.e. high temperature bulk growth. This section will discuss the results of annealing samples with varying starting APB densities and growth conditions. The samples include two samples with high starting APB densities, one with large domains and extended APBs and one with mostly discrete domains, and a low density sample with low temperature bulk growth.

PAGE 114

114 5.2.2 Experimental All additional samp les were grown on 300 mm Si(001) wafers with no intentional off cut in an Applied Materials III V MOCVD system. The growth conditions, unless otherwise stated, are identical to those stated in section 5.1.2. All samples were cleaned using the Siconi TM proc ess prior to growth. The layer structure for the two high density samples is the same as that shown in Figure 3 3, i.e. 200 nm grown at 530 C followed by 200 nm grown at 610 C. The high APB density in these samples was caused by changing the pre growth b ake conditions. The layer structure for the low density, low temperature bulk growth sample is similar to that shown in Figure 4 2 except that the total ~800 nm of growth was completed at 530 C. A standard pre growth bake at ~900 C was used for this samp le. The processing conditions for annealing and characterization are also identical to those stated in section 5.1.2. Briefly, samples were encapsulated with 20 nm of Al 2 O 3 by ALD and annealed in a tube furnace with Ar ambient. APBs were selectively etche d for imaging in SEM using the HF/HNO 3 solution. Non selective etching with H 3 PO 4 /H 2 O 2 was used for depth profiling of APB density. 5.2.3 Results 5.2.3.1 High density samples Figure 5 7 shows representative plan view SEM images of the high starting APB d ensity sample with the largest domains as grown ( A) and then after annealing at 700 C for 30 min ( B) and 4 hrs ( C) The APB density decreases in these images from the as grown value of 3.3 m 1 to 2.6 m 1 to 2.4 m 1 respectively. Along with the overall decrease in APB density, the APBs appear to become more rounded with annealing compared to as grown. This observation is not surprising if the evolution of APBs is

PAGE 115

115 driven by a reduction in curvature, following the All en Cahn theory [110] The APBs that loop around an antiphase domain should follow a much less tortuous path and smooth out over time. Figure 5 8 shows the surface APB density results of all anneals for this sample at 650 and 700 C for various times. The largest decrease occurs within the first 30 min of annealing which agrees with the qualitative observations in the SEM images. The APB density appears to plateau after this point. This appearance of a plateau is likely due to the annihilation of smaller domains within the first 30 min of annealing and the comparatively much slower annihilation of larger domains since the APBs must diffuse over a longer distance to annihilate the domains. Anneals performed on separate occasions for this high density sample led to much more noticeable rounding of APBs. Figure 5 9 shows representative plan view SEM images of the APB density after annealing at 700C for 2 hrs ( A) and 750 C for 1 hr ( B) Some domains appear to be nearly circular in shape. It should be noted that these specific specimens were encapsulated with Al 2 O 3 entirely with water as the oxygen precursor. Therefore, it is possible that there was a cap failure that led to disparate annealing conditions. However, it is interesting that this degree of rounding is possible. Also, this rounding again would be consistent with Allen Cahn theory for the minimization of free energy. If annealing was extende d for infinite time and temperature, it is expected that the antiphase domains would all decrease in diameter and eventually become circular before completely annihilating. Representative plan view SEM images of the second high density sample with more dis crete domains are shown in Figure 5 10 for the sample as grown ( A) and then after annealing at 700 C for 1 hr ( B) and 4 hrs ( C) The decrease in APB density with

PAGE 116

116 annealing is noticeable along with the decrease in number of antiphase domains visible. The A PBs also become more rounded with annealing, similar to the previous high density sample. Figure 5 11 shows a plot of the surface APB densities for all of the annealing conditions for this sample. There is a consistent trend of decreasing APB density with increasing anneal time and temperature. However, the APB density decrease appears to slow after ~1 hr of annealing time. To test whether this was an effect of smaller domains being eliminated initially and slower annihilation of larger domains, counts of d omain size versus annealing were made for select conditions. Figure 5 12A shows a graph of domain counts for as grown and after annealing at 700 C for 30 min and 4 hrs. The domain counts are separated into categories of small domains for those less than 2 m in APB loop length and large domains for those greater than 2 m in APB loop length. Figure 5 12 B shows how the ratio of small to large domains changes with annealing conditions. From these two plots, it can be concluded that small domains are in fact annihilated much more rapidly than large domains. This result is intuitive based on the longer distance APBs looping around a larger domain would have to diffuse in order to annihilate the domain. The much quicker annihilation of smaller domains leads to p lateaus in APB density decreases for extended annealing times. 5.2.3.2 Low temperature bulk growth sample The low starting APB density sample with low temperature bulk growth is the same sample used in the growth temperature study in Chapter 4 with bulk g rowth at 530 C. Figure 5 13 shows a representative plan view SEM image of the as grown surface APB density for this sample. Anneals were performed on this sample in order to

PAGE 117

117 determine whether a similar decrease in APB density as seen by raising the bulk g rowth temperature could be replicated through post growth annealing. To this end, analysis of this sample focused on through thickness APB density. Figure 5 14 shows the through thickness APB density depth profiles for the low temperature bulk growth samp le as grown and after annealing up to 30 min at 600 to 650 C. At first glance, it is apparent that the APB density for the annealed samples is always lower compared to the as grown sample and the degree of decrease correlates with increasing anneal time a nd temperature. However, the profiles do not indicate a steady decrease in APB density, i.e. APB annihilation from the surface down, as was seen for the annealed sample in section 5.1.3 (Figure 5 4), nor do they match the profiles for samples grown at high er bulk growth temperatures in section 4.3.1 (Figure 4 4). In some cases for the annealed profiles, e.g. 650 C for 30 min, the APB density is markedly lower towards the GaAs/Si interface than at the surface of the film. This unusual trend indicates that A PBs may be being pinned at the GaAs/Al 2 O 3 interface for this sample and antiphase domains are being pinched off in the bulk of the GaAs film. Further investigation of APB evolution in this sample after annealing was performed with XTEM. Figure 5 15 shows a pair of DF XTEM images of the low temperature growth sample after annealing at 600 C for 30 min under a g 002 condition ( A) and a g 00 2 condition ( B) A domain can be observed near the surface of the GaAs film. The reversal of contrast from the g 002 to g 00 2 condition indicates this feature is an antiphase domain. The antiphase domain is terminated towards the middle of the GaAs film. This occurrence can only be an effect of the anneal, as the spontaneous nucleation of an APB during growth would be highly energetically unfavorable. The shape of this

PAGE 118

118 domain, particularly the wider diameter at the GaAs surface, indicates that the APBs may have been pinned at the GaAs/Al 2 O 3 interface during the anneal. At the GaAs/Si interface, another domain like feature is observed directly under the pinched off portion of the previously mentioned antiphase domain. This domain is likely indicative of where the first antiphase domain nucleated. The antiphase domain was then pinched off during the anneal, leaving two separate portions of antiphase domains in the GaAs layer. These observations support the results of the APB density depth profiles. 5.2.4 Discussion The evolution of the high APB density samples during annealing was consistent with the results from section 5.1 for the low density sample with standard growth conditions. The surface APB density decreased with increasing annealing time and temperature. APBs were observed to smooth out with annealing, i.e. forming less tortuous loops around antiphase domains, indicatin g that they were evolving to reduce their curvature. This reduction in curvature and in turn the reduction in APB density lowers the free energy of the system. Eventually, antiphase domains are eliminated due to APB self annihilation. The evolution of APB density in the low density sample with low temperature bulk growth was more curious. Through thickness APB density depth profiles showed that APB annihilation occurred more rapidly and efficiently in the bulk of the GaAs film, rather than from the surface down. The XTEM analysis supports a conclusion that APBs are being pinned at the GaAs/Al 2 O 3 during annealing leading to pinched off domains. The results indicate that the grown in character of APBs and the surface conditions during annealing may play a rol e in APB mobility and annihilation. APBs for this sample with growth completely at a low temperature should tend to {110} type configurations,

PAGE 119

119 rather than the inclined habit planes that are preferred at higher growth temperatures [114, 147] (see Chapter 4). It is possible that {110} APBs are more difficult to diffuse, i.e. they are more energetically stable, with high temperature post growth treatments compared to higher index plane APBs, e.g. {112}. This hypothesis is supported by the lower formation energy of {110} APBs over higher index planes [92] Thus, it may be required for more efficient annihilation of APBs with annealing for the APBs to already be configured along relatively energetically unstable planes, i.e. higher index plane s, due to high bulk growth temperatures. The effect of the surface conditions during annealing will be expanded upon in section 5.3; briefly, {110} APBs may be even more stable at the surface of the GaAs film during annealing compared to the bulk, leading to the appearance of APB pinning at the surface. 5.2.5 Summary APBs are driven to annihilation with increasing annealing time and temperature regardless of the starting APB density for GaAs films grown on Si. The APB density in high starting density sample s decrease with increasing annealing time and temperature. The results support the notion that the equilibrium state for the film should be APB free. The driving force for APB evolution with annealing is a reduction in curvature and a corresponding decreas e in the total APB interfacial area. It was also observed that the sample growth conditions, particularly the growth temperature, appear to affect the nature of APB annihilation with post growth annealing. That is, APB annihilation occurs with annealing mo re readily if the APBs are already configured along higher index planes, e.g. {112}, rather than {110} planes. Through thickness APB density depth profiles of a sample with low temperature bulk growth indicated that APBs were being pinned at the surface of the GaAs film. XTEM analysis supported this theory and

PAGE 120

120 showed an antiphase domain that had the appearance of propagating from the surface down and was terminated in the middle of the film. The surface conditions during annealing may also play a role in th is pinning effect. 5.3 Effect of Annealing Conditions 5.3.1 Background Protecting the surface against degradation is an important aspect of annealing GaAs films or substrates. Arsenic has a relatively high vapor pressure and is prone to leave the surface, leading to the formation of pits [120] The Al 2 O 3 films used in this study provide a diffusion barrier against As outgassing. An As overpressure can also be provided to prevent the loss of As during annealing by flowing As vapor or gas. A more practical means of providing an As overpressure is to use a p roximity cap, i.e. place another piece of GaAs on top of the sample. The loss of As is limited by outdiffusion from between the two contacted surfaces which leads to a local As overpressure [165] Proximity capping has the additional advant age over depositing encapsulant materials of reducing the processing time needed for preparing samples for annealing. The effect of the surface conditions during annealing, i.e. the type of cap used, on the stability of APBs in GaAs films grown on Si was i nvestigated in this experiment. 5.3.2 Experimental Two separate samples were used for this experiment. The first sample was the low temperature bulk growth sample from Chapter 4, the growth temperature study. This low temperature bulk growth sample has the layer structure shown in Figure 4 2. Growth was co nducted completely at 530 C for a nominal film thickness of ~800 nm. The second sample for this experiment was the same used in section 5.1 for the annealing study. This high temperature bulk growth sample has the layer structure

PAGE 121

121 shown in Figure 3 3. Bulk growth for 200 nm was conducted at 610 C and the total film thickness was ~400 nm. These samples will henceforth be referred to as the low temperature sample and high temperature sample, respectively. The samples were chosen because they have comparable surface APB densities after growth. Surface protection during annealing was provided either by a 20 nm Al 2 O 3 cap deposited by ALD or a proximity cap. The proximity cap was a sacrificial GaAs coupon from a semi insulating GaAs wafer. The presence of native oxide on the GaAs proximity cap and the sample was found to not have a significant effect on APB density after annealing. Therefore, the native oxide was not stripped from the samples or the cap prior to the anneals discussed here. APB density after anne aling was analyzed using the selective HF/HNO 3 etchant and SEM. Anneals were performed at 600 for times up to 30 min in a tube furnace with flowing Ar ambient. 5.3.3 Results Figure 5 16 shows representative plan view SEM images of the APB density in the lo w temperature sample as grown ( A) and then after annealing at 600 C for 30 min with a proximity cap ( B) and an Al 2 O 3 cap ( C) There is a large disparity in the APB density and the number of domains visible at the surface after annealing this sample with t he different caps. The APB density in the proximity capped sample is nearly the same as the as grown sample. On the other hand, significant reduction in APB density is observed for the sample that was capped with Al 2 O 3 during the anneal. Therefore, the pro ximity cap conditions appear to worsen the pinning of APBs at the GaAs surface during annealing. This APB pinning was already suspected for this sample when an Al 2 O 3 cap was used (see section 5.2). The surface APB densities after the full series of

PAGE 122

122 anneals at 600 C for the low temperature sample with different caps is shown in Figure 5 17. The APB density for an Al 2 O 3 cap experiences a steady decay with increasing annealing time, while the APB density for the proximity cap does not change. In other words, there is negligible mobility of APBs at the surface of the GaAs film during annealing for the low temperature sample when a proximity cap is used. The effect of the proximity cap on APB dissolution during annealing was also examined for the high temperatur e sample. Figure 5 18 shows the results of the surface APB densities for this sample after annealing compared to the low temperature sample under the same conditions. The APBs in the high temperature sample are clearly still mobile at the surface even with a proximity cap. This result provides further evidence that the grown in nature of the APBs, i.e. the APB habit planes, affects their stability during annealing. In fact, the difference in grown in APB habit planes can be seen by directly comparing the as grown samples. Figure 5 19 shows representative plan view SEM images of antiphase domains in the low temperature sample ( A) and the high temperature sample ( B) The APBs in the low temperature sample have a tendency to facet along {110} orientations which is expected at low growth temperatures due to the low formation energy of these APB planes. On the other hand, the APBs in the high temperature sample are not as obviously faceted, giving the appearance of antiphase domains with rounder shapes. These APBs in the high temperature sample are likely already propagating on higher index planes when they reach the GaAs surface. 5.3.4 Discussion The surface conditions during the anneal appear to have an effect on the mobility of APBs. The surface conditions of a GaAs film annealed with a proximity cap can be considered to be As rich due to the local As overpressure. Self diffusion in GaAs is

PAGE 123

123 known to be affected by whether the local conditions are intrinsic, Ga rich, or As rich [163, 166] This effect arises from the relative stability of Ga vacancies that are the primary mediators of diffusion. As rich conditions in particular support the formation of Ga vacancies [161] The stability of antisite defects, which are the constituents of APBs, is also affected by the local conditions. Recent DFT calculations by Chroneos et al. [99] showed that the formation energy of As Ga and Ga As antisite defects change whether the conditions are Ga rich, As ric h, or stoichiometric. As conditions become more As rich the formation energy of As Ga is lowered and for more Ga rich conditions the formation energy of Ga As is lowered. Thus, it is logical to conclude that the relative stability of point defects, i.e. vaca ncies and antisite defects, under given annealing conditions will affect the mobility and subsequent annihilation of APBs. The magnitude of this effect is also apparently influenced by the grown in nature of APBs, i.e. the APB habit planes. 5.3.5 Summary T he mobility and subsequent dissolution of APBs in GaAs on Si films during annealing is affected by the surface conditions. APB pinning at the GaAs surface was exacerbated when a proximity cap was used compared to an Al 2 O 3 cap. This effect likely stems from the influence of the local As overpressure with the proximity cap conditions on the relative formation energy and stability of point defects, such as the antisite defects that compose APBs. A secondary effect on APB mobility is the grown in APB habit plan e. These results demonstrate that the most efficient APB dissolution for post growth annealing will occur for samples grown at a high bulk growth temperature and that are processed with Al 2 O 3 encapsulating films instead of proximity caps.

PAGE 124

124 5. 4 Effect of APB Density on Si Diffusion 5. 4.1 Background The growth of GaAs on Si by definition creates a diffusion couple. There is a nonzero diffusivity of Ga and As atoms into Si and for Si atoms into GaAs at elevated temperatures [167] In particular, the diffusion of Si atoms into the GaAs layer from the substrate can be enhanced by the presence of e xtended defects stemming from the GaAs/Si interface, e.g. grain boundaries and dislocations [13, 44] Si concentration in GaAs epitaxial layers grown on Si has been correlated to the dislocation density in the layers [168] Further, the diffusion coefficient of Si in GaAs for annealed GaAs on Si layers is enhanced for thinner films, i.e. those with higher defect densities [169] Impurity atoms are known to be preferentially segregated to APBs in order to decrease the degree of crystal distorti on and total free energy [11] It follow s that the APB density in epitaxial GaAs layers may have an effect on the diffusion of Si from the substrate during annealing treatments. 5.4.2 Experimental Two GaAs on Si samples with varying APB density were chosen to compare the effect of APBs on Si di ffusion from the substrate. Both samples were grown on 300 mm (001) oriented Si wafers with no intentional off cut. The wafers were cleaned using the Siconi TM process [118] prior to GaAs growth. After transferring the wafers to the growth chamber, t he pre growth bake at 900 C was varied to affect the step doubling, and thus the nucleated APB density. GaAs films with a nominal thickness of 400 nm were grown in an Applied Materials III V MOCVD system. The growth conditions for both samples were identi cal to those stated in section 5.1.2. The layer structure is the same as that

PAGE 125

125 shown in Figure 3 3. First, 200 nm of GaAs is grown at 530 C followed by 200 nm of GaAs grown at 610 C. Annealing was performed in a tube furnace at 650 C for 15 min and 2 hr s with flowing Ar ambient. Annealed samples were capped with 20 nm of Al 2 O 3 deposited by ALD. The APB density in the samples was analyzed using the selective HF/HNO 3 etchant and SEM. The GaAs layers were compared in XTEM after cutting out samples in a FIB and imaging with a JEOL 2010F TEM operating at 200 kV. SIMS was performed to determine the amount of Si diffusion in the layers as a function of annealing time. Samples for SIMS were sputtered with Cs + ions. XRD was performed with a Panalytical Xpert MRD s ystem to compare the structural quality of the layers. 5.4.3 Results and Discussion Figure 5 20 shows the as grown surface APB densities in the low density ( A) and high density ( B) samples. The low density sample has a starting APB density of 0.10 m 1 and the high density sample has a starting APB density of 2.8 m 1 Figure 5 21 shows BF XTEM images of the low density sample ( A) and the high density sample ( B) There is no apparent difference in the density of defects besides for APBs. The GaAs layer thicknesses are also approximately the same. The difference in APB density is large enough to analyze the effect of APBs on Si diffusion from the substrate. The SIMS profiles of Si concentration in the GaAs layers is shown in Figure 5 22 for the low densi ty sample ( A) and the high density sample ( B) for as grown conditions and after annealing at 650 C for 15 min and 2 hrs. For the low density sample, it is difficult to differentiate the Si concentration profiles for each condition. The effect of annealing is most obvious for the sample annealed for 2 hrs. At the location of the

PAGE 126

126 GaAs/Si interface (depth of ~400 nm) for the low density sample, there is an apparent shoulder in the Si concentration profile after annealing at 650 C for 2 hrs. The total Si conc entration in the GaAs layer near the interface tends to increase with increasing annealing time. The effect of annealing is more pronounced for the high density sample. The total Si concentration in the GaAs layer increases with increasing annealing time. Figure 5 23 compares the Si concentration profiles after annealing at 650 C for 2 hrs for the low density and high density samples. It is apparent that Si diffusion from the substrate is enhanced for the sample with higher APB density compared to the samp le with low APB density. To further rule out the effect of other defects on the enhanced Si diffusion from the substrate, XRD analysis was performed. Figure 5 24 shows the (002) ( A) and (004) (B) rocking curves for the low and high density samples. The (0 02) rocking curve should be sensitive to the degree of antiphase disorder in the samples. However, the profiles are very similar for both samples. It is likely that most of the antiphase domains in the high density sample are too large to be detected by XR D (>300 nm) [13, 116] The (004) peak is sensitive to defect density and the overall crystalline order in the GaAs layer without contribution from antiphase domains. The (004) profiles for both samples are also very similar in peak intensity and shape. The similarity between these two profiles indicates that the defect density (other than antiphase domains) and crystal quality is also very similar. Thus, the difference in Si diffusion from the substrate (Figure 5 23) is less likely to be from factors other than the difference in APB density between the samples.

PAGE 127

127 The caveat to th ese results is that APBs are simultaneously being dissolved during annealing while Si is diffusing into the GaAs layer from the substrate along APBs. Based on the results of the annealing experiments previously discussed in sections 5.1 and 5.2, it can be assumed that APBs are still prevalent in the high density sample, while they are likely most dissolved in the low density sample after 2 hrs of annealing at 650 C. It is unclear what role this effect plays on Si diffusion, although it would likely lead to an even greater difference in Si diffusion for the high versus low APB density samples. 5.4.4 Summary The diffusion of Si from the substrate for GaAs on Si films is enhanced during annealing treatments by the presence of APBs. SIMS analysis showed that after annealing at 650 C for 2 hrs, there was markedly higher Si concentration in the GaAs layer for the sample with a starting APB density of 2.8 m 1 compared to the sample with a starting APB density of 0.10 m 1 XRD analysis indicated that the difference in Si diffusion from the substrate between the two samples is more likely to be from the difference in APB density than from any other defects, e.g. threading dislocations. 5.5 Summary of Findings for the Chapter It is possible to annihilate APBs in GaAs layers grown on Si through post growth annealing treatments. APBs diffuse in order to reduce their curvature and minimize the total APB interfacial area. It is unclear what the exact atomic mechanism is that facili tates the motion of an APB. However, APBs are clearly energetically unstable in GaAs and there is an energetic driving force for their dissolution. The stability of APBs in GaAs is affected by the grown in nature of the APBs, i.e. the APB habit planes, and

PAGE 128

128 the surface conditions during annealing, i.e. when an Al 2 O 3 cap is used versus a proximity cap. Si diffusion from the substrate is enhanced during annealing by the presence of APBs. The Si Ga and Si As bonds are energetically preferred over the wrong Ga G a and As As bonds at an APB.

PAGE 129

129 Figure 5 1 BF XTEM image of of a GaAs on Si sample annealed at 600 C for 30 min with 20 nm of Al 2 O 3 as a capping layer There are no visible signs of surface degradation. Figure 5 2 Evolution of APB density and domain size with annealing. Representative plan view SEM images of the surface APB density in the GaAs layer A) as grown and B) after annealing at 650 C for 7.5 min. There is a clear reduction in the n umber of antiphase domains visible after annealing. The corresponding histograms of domain size C ) as grown and D) after annealing show that the average domain size is also decreasing.

PAGE 130

130 Figure 5 3 Plot of surface APB densities for samples with varying annealing times and temperatures. There is a consistent trend of decreasing APB density and increasing annealing time and temperature. Figure 5 4 Depth profiles of APB density for samples as grown and with varying annealing conditions. After annealing at 600 C for 10 min, there is a noticeable decrease in APB density towards the GaAs surface. After annealing for longer time or at higher temperature, the APB density decreases throughout th e GaAs layer. The annihilation of APBs during annealing may propagate from the surface down.

PAGE 131

131 Figure 5 5 Example of the fitting procedure for APB density versus annealing time A) D ata for sample annealed at 700 C. The r ate, k, is extracted from the slope. B) The Arrhenius plot of all extracted rates for varying anneal temperature is shown. The corresponding activation energy is 3.8 eV.

PAGE 132

132 Figure 5 6 Illustration of a possible mechanism for the movement of an APB segment from a {110} configuration to a {111} configuration via the diffusion of a Ga vacancy (V Ga ) A) Original structure with {110} APB, B) generation of Ga vacancy, C) diffusion of Ga vacan cy to neighboring site, and D) further diffusion of Ga vacancy and consequential development of a {111} kink

PAGE 133

133 Figure 5 7 Representative plan view SEM images of APB density in highly defective sample as grown and after annealing A) Sample as grown, B) after annealing at 700 C for 30 min and C) after annealing at 700 C for 4 hrs. The APB density decreases from 3.3 m 1 to 2.6 m 1 to 2.4 m 1 respectively and APBs become noticeably rounder with annealing. Figure 5 8 Plot of surface APB densities for high starting APB density sample for various annealing conditions. There is a large decrease in APB density after annealing for 30 min, but then the density plateaus. The plateau is likel y an artifact of smaller domains being annihilated from 0 30 min of annealing and the slower motion and annihilation of large APBs and domains.

PAGE 134

134 Figure 5 9 Representative plan view SEM images of high starting APB density sample after annealing at 700 C for 2 hrs A) and 750 C for 1 hr B) There is much more noticeable rounding of APBs af ter annealing for these samples and some domains even appear circular. T he rounding demonstrates the evolution of APBs to minimize free energy in accordance with Allen Cahn theory. Figure 5 10 Representative plan view SEM images of the APB density in a highly defective sample with more discret e domains as grown A) after annealing at 700 C for 1 hr B) and then 4 hrs C) There is a noticeable rounding of APBs and an overall decrease in APB density.

PAGE 135

135 Figure 5 11 Plot of surface APB densities for sample with high starting density and discrete domains after various annealing treatments. There is a consistent trend of decreasing APB density with increasing annealing time and temperature. The rate of APB density decrease appears to slow after ~1 hr of annealing time likely from the faster elimination of smaller domains. Figure 5 12 Evolution of large and small antiphase domains in high density sample with annealing. A) Histogram of domain size as grown and after annealing separated into catego ries of small and large domains B) The ratio of small to large domains decreases with annealing indicating that small domains are in fact annihilated much more rapidly than large domains. This difference leads to apparent plateaus in APB densit y for extended annealing times.

PAGE 136

136 Figure 5 13 Representative plan view SEM image of APB density in a sample grown entirely at a low temperature of 530 C Figure 5 14 APB density depth profiles for the low temperature growth sample as grown and after various annealing treatments. The APB density in annealed samples is consistently lower compared to the as grown profile, but there is not a steady decrease in density from the surface down. The APB density is sometimes higher towards the surface indicating that APBs may be being pinned at the GaAs/Al 2 O 3 interface during annealing.

PAGE 137

137 Figure 5 15 Pinched off antiphase domain in t he low temperature growth sample after annealing at 600 C for 30 min A) DF image using the g 002 reflection and B) DF image using the g 00 2 reflection The antiphase domain is visible near the surface of the GaAs film and is terminated towards the middle of the film. Another possible domain is observed near the GaAs/Si interface (marked by arrow) that may have been connected to the domain near the surface prior to annealing. Figure 5 16 Representative p lan view SEM images of the APB density in the low temperature growth sample after annealing with different capping conditions. A) Sample as grown, B) after annealing at 600 C for 30 min with a proximity cap, and C) after annealing at 600 C for 30 min wit h an Al 2 O 3 cap. There is no apparent difference in APB density after annealing with the proximity cap which indicates that the pinning of APBs at the GaAs surface is greater under these conditions. There is a noticeable decrease in APB density when using t he Al 2 O 3 cap.

PAGE 138

138 Figure 5 17 Plot of surface APB densities for the low temperature sample after annealing with a proximity cap and Al 2 O 3 cap at 600 C for various times. There is a steady decrease in APB density when using the Al 2 O 3 cap and no decrease when using the proximity cap supporting the notion that APB pinning at the GaAs surface is affected by the surface conditions. Figure 5 18 Plot of surface APB densities for the low temperature growth sample compared with a high temperature growth sample (610 C ) for similar annealing conditions with a proximity cap. APB density decreases for the high temperat ure growth sample even with a proximity cap. Therefore, the stability of APBs during annealing may also be affected by the grown in nature of APBs, i.e. {110} versus inclined habit planes.

PAGE 139

139 Figure 5 19 Representative plan view SEM images showing the grown in nature of APBs as a function of growth temperature. A) APBs in the low temperature growth sample and B) APBs in the high temperature growth sample. The APBs in the low temperature growth sample tend to be faceted along {110} orientations while the APBs in the high temperature growth sample tend to be more rounded. Figure 5 20 Representative plan view SEM images of the as grown surface APB densities in the samples for the Si diffusion study. A) L ow density sample and B) high density sample The APB densities are 0.10 m 1 and 2.8 m 1 respectively.

PAGE 140

140 Figure 5 21 BF XTEM images of samples for the Si diffusion study. A) Low A PB density s ample and B) high APB density sample There is no apparent difference in defect densities besides that of APBs and the GaAs layer thicknesses are approximately the same. Figure 5 22 SIMS profiles of Si concentr ation in the GaAs layers for samples with different starting APB densities as grown and after annealing A) L ow APB density sample and B) high APB density sample. Si profiles are shown for as grown conditions and after annealing at 650 C for 15 min and 2 hrs.

PAGE 141

141 Figure 5 23 Comparison of Si concentration profiles for the low and high APB density samples after annealing at 650 C for 2 hrs. There is noticeably more diffusion of Si from the substrate for the high APB density sample compared to the low APB density sample. Figure 5 24 XRD rocking curves of the low and high APB density samples as grown. A) Profiles for the (002) peak and B) profiles for the (004) peak There is no difference in t he (002) peaks that should be sensitive to antiphase disorder likely because the domains in the high density sample are larger than the XRD detection limit (~300 nm). There is also no apparent difference in the (004) peaks which indicates that the density of defects other than APBs, e.g. dislocations, and the overall crystalline quality are similar between the two samples.

PAGE 142

142 CHAPTER 6 THE EFFECT OF DOPING 6.1 Background The electrical and optical properties of GaAs layers grown epitaxially on Si(001) are degraded by the formation of APBs. These APBs nucleate at single steps on the Si surface and form planar defects composed entirely of antisite bonds, i.e. Ga Ga and As As The inherent formation of antisite bonds contributes to the large formation energy of APBs [92] The presence of APBs in a GaAs layer should be energetically unfavorable. It has been demonstrated that higher growth temperatures lead enhanced annihilation of APBs [114, 147] and that post growth annealing is capable of dissolving antiphase domains (see Chapter 5). These results indicate that the kinetic stability of APBs can be affected and overcome by increasing the thermal energy availa ble. The lowest energy configuration for a GaAs layer should be free of APBs. It may be possible to affect the energetic stability of APBs either during growth or during post growth thermal treatments through other avenues. One option is to alter the rel ative stability of the antisite defects that constitute APBs. Walukiewicz [97] hypothesized that p type or n type shifts in the Fermi level for GaAs, i.e. doping, will favor the formation of antisite defects and vacancies to compensate the charge from the dopant species. The formation energy of Ga As antisite defects (and Ga vacancies) is lowered as the Fermi level approaches the conduction band edge and the formation energy of As Ga antisite defects (and As vacancies) is lowered towards the valence band edge. The work by Walukiewicz was corroborated later on by DFT studies by Chroneos et al. [99] The authors focused on the formation energy of Ga As and As Ga antis ite defects as a function of the Fermi level. The formation energy of As Ga defects is

PAGE 143

143 decreased near either the valence or conduction band edge. The formation energy of Ga As decreases moving from the valence band edge to the conduction band edge, at which point it is lower overall in energy compared to As Ga It follows from the effect of Fermi level shifts on antisite formation energies that doping may also change the energetic stability of APBs in GaAs. This section investigates the effect of doping on the grown in APB density for GaAs films grown on Si. In situ doping was chosen to avoid any possible effects of damage and regrowth from ex situ doping, i.e. ion implantation, on APBs. 6.2 Experimental All samples in this experiment were grown on 300 mm lig htly p doped Si(001) wafers with no intentional off cut. The wafers were cleaned prior to GaAs growth using the Siconi TM process [118] The wafers were then transferred under vacuum to an Applied Materials III V MOCVD system. A ~900 C bake was perf ormed to promote the formation of double steps on the Si surface. GaAs film growth proceeded in the standard two step process with a low temperature nucleation layer and high temperature bulk layer. The layer structure is similar to that depicted in Figure 3 3 with 200 nm of GaAs grown at 530 C followed by 200 nm of GaAs grown at 610 C. The precursors for Ga and As were trimethylgallium (TMGa) and tertiarybutylarsine (TBAs), respectively. The V/III ratio was within the standard MOCVD process range and the chamber was maintained at typical reduced pressure ambient during film growth P doping was accomplished by the addition of diethylzinc (DEZn) flow during growth and n doping was accomplished by the addition of disilane flow. One sample was p doped, one w as n doped, and one was nominally undoped for the control.

PAGE 144

144 As grown layer thickness measurements for the samples were obtained using ellipsometry. Electrical characterization of the samples was performed using Hall effect measurements. Contacts were made using pressed on In wire. The nominally undoped sample was further annealed at 425 C for 2 min in a tube furnace with Ar ambient to make good electrical contact. The APB density in the samples was measured using the selective HF(49%):HNO 3 (69%):H 2 O (20:1:7 ) etchant and SEM. APB features in SEM images were traced using ImageJ software [132] 6.3 Results The sheet number, i.e. areal density, of active charge carriers in the GaAs films was first measured by Hall effect. The sheet number was then converted to a volumetric density by dividing the sheet number by the GaAs layer thickness. The n type sample was de termined to possess 2.7x10 18 electrons/cm 3 the p type sample had 2.5x10 18 holes/cm 3 and the undoped sample had 4.3x10 16 electrons/cm 3 Unintentional doping from background contamination in MOCVD, e.g. carbon or silicon, is unavoidable. Therefore, even the nominally undoped sample has some amount of extrinsic doping. It is important to reiterate that the undoped sample was measured after annealing the In contacts and so the sheet number measurement may be unreliable. The charge carrier concentration can be used to determine the Fermi level of each sample through the relation [170] : (6 1) where E F is the extrinsic Fermi level, E Fi is the intrinsic Fermi level, N d is the concentration of donors (or acceptors), n i is the intrinsic carrier concentration, k is type and p type samples are both

PAGE 145

145 ~0.7 eV deviated from the intrinsic Fermi level, placing the extrinsic Fermi level for these samples near the conduction band edge and valence band edge, respectively, for GaAs. The extrinsic Fermi level of the nominally undoped sample is calculated to be ~0.6 eV above the intrinsi c Fermi level; however, this calculation has a large dependence on the accuracy of the sheet number measurement from Hall effect. Figure 6 1 shows representative plan view SEM images of the surface APB densities for the p type sample ( A) undoped sample ( B) and n type sample ( C) The APB densities are 0.33 m 1 0.50 m 1 and 1.2 m 1 respectively. The APB density of the n type sample is noticeably higher than that of the p type or undoped samples simply from comparing the SEM images. Figure 6 2 plots t he APB density of the samples versus the dopant type ( A) and the calculated Fermi level ( B) There is a clear trend of increasing APB density with increasing n type character. The jump in APB density is much more sudden on Figure 6 2 B when representing the data against the calculated Fermi level. This difference may arise from inaccurate determination of the undoped sample Fermi level. Regardless, the n type sample APB density is much higher (~4x) that of the p type sample, indicating that the Fermi level does have an effect on the stability of APBs during growth. 6.4 Discussion The substrate surface preparation procedure was identical for all samples, so they can be presumed to have the same nucleated APB density. The growth condition s for the samples in this work were also identical except for the addition of the dopant precursors. Further, Zn and Si are both preferentially dissolved on the Ga sublattice [163] so there is no disparity in the physical incorporation of the dopant atoms on lattice

PAGE 146

146 sites. Therefore, the difference in surface APB densities most likely arises from an effect of the dopant precursor molecule reaction on the APB kinetic stability or from the Fermi level shift due to the incorporation of dopant atoms on lattice sites. It would difficult to decouple these two possible effects experimentally, unless a precursor for Si with similar number and structure of alkyl groups to DEZn can be used. The trend of increasing APB density with increasing n type character in the samples correlates with the decrease in antisite formation energies from the work of Walukiewicz [97] and Chroneos et al. [99] Therefore, it is plausible that the decrease in antisite formation energies stabilizes APBs during growth instead of facilitating APBs kinking to inclined plane s and self annihilating. Previous work showed that at the growth temperature for the GaAs films in this study that APB propagation along inclined planes should be preferred for nominally undoped conditions [147] P type doping may have a small effect on enhancing APB annihilation, i.e. making APBs less energetically stable. 6.5 Summary In situ doping in MOCVD affects the grown in APB density for GaAs films grown on Si. N type d oping with Si caused a >2x increase in surface APB density compared to a nominally undoped sample. P type doping with Zn led to a slight decrease in APB density compared to the undoped sample. The results indicate that the lowering of antisite defect forma tion energies with more n type Fermi level shifts works to stabilize APBs during growth. Therefore, fewer APBs kink to inclined planes and self annihilate. The results of this work also show that intentional n type doping should be avoided in GaAs layers g rown on Si for the purposes of reducing grown in APB density. Rather, intentional p type doping may be optimal for the lowest APB density.

PAGE 147

147 Figure 6 1 Representative plan view SEM samples of the APB density in doped and und oped samples as grown. A) p type sample B) nominally undoped sample and C) n type sample The measured carrier concentrations are 2.5x10 18 holes/cm 3 4.3x10 16 electrons/cms 3 and 2.7x10 18 electrons/cm 3 respectively. APB density is significantly higher for the n type sample compared to either the p type or undoped samples. Figure 6 2 Plots of APB density as a function of doping conditions and Fermi level for the p type, undoped, and n type samples. A) APB density versus dopant type and B) APB density versus calculated Fermi level. APB density increases with increasing n type character.

PAGE 148

148 CHAPTER 7 CONCLUSIONS AND FUTURE WORK 7.1 Major Findings The results of the experiments in this work support the conclusion that APBs i n GaAs on Si are inherently unstable defects. APBs are forced to nucleate in epitaxial GaAs films because of single steps on the Si surface. The elimination of APBs is only feasible through processes that are kinetically unfavorable, e.g. kinking of APBs t o inclined planes during growth and the diffusive motion of APBs post growth. However, there is a significant thermodynamic driving force for the elimination of APBs due to their large associated energies. Therefore, the kinetic barriers to APB annihilatio n can be overcome by increasing the thermal energy available and affecting the APB energetics. The GaAs on Si system approaches its lowest energy configuration which is free of APBs. Growth temperature in MOCVD was found to play a large role in the tenden cy of APBs in GaAs films to kink to inclined planes and self annihilate during growth. APB nucleation along {110} planes is favored energetically. Self annihilation can only occur if APBs propagate along higher index planes such as {111} or {112} that are nominally higher in formation energy. Increasing the growth temperature from 530 to 610 C increased the APB annihilation rate with respect to film thickness by a factor of four. An associated activation energy of 1.1 eV was found. APBs along {112} planes and other higher index plane configurations were observed in DF XTEM. The higher growth temperatures likely remove kinetic barriers for the propagation of APBs along these planes and facilitate annihilation. The kinking of APBs to inclined planes allows fo r a reduction in the total line length of APBs in the (001) growth plane with each

PAGE 149

1 49 consecutive monolayer of growth. The rate of APB annihilation with respect to film thickness is ultimately limited by the preferred habit plane for APB kinking. It was foun d that APB annihilation and suppression of antiphase domains can be achieved with post growth annealing. Diffusive motion of APBs occurred at annealing temperatures as low as 550 C and had an Arrhenius dependence up to 700 C. The activation energy for AP B diffusion was determined to be 3.8 eV. The motion of an APB is likely mediated by the presence of vacancies and the rate limiting step may be Ga self diffusion. Decreases in APB density with annealing were observed regardless of the initial starting dens ity. The evolution of APBs during annealing is qualitatively similar to the well known phenomenon of grain growth and appears to be governed by similar kinetics. The energetic driving force for APB diffusion is a reduction in curvature and total interfacia domains. Annealing brings the GaAs layer closer to an equilibrium structure that is APB free and not with {110} oriented APBs. The grown in structure of APBs, i.e. their grown in habit planes, as well as the surface conditions during annealing may affect the energetics of APB diffusion. The energetics of APB propagation during growth can also be affected by in situ doping. The APB density in GaAs on Si samples grown with identi cal conditions, besides the addition of dopant precursors, varied based on the calculated Fermi level. Heavy n type doping (>10 18 electrons/cm 3 ) led to a more than 2x increase in APB density compared to a nominally undoped reference sample. Conversely, heavy p type doping (>10 18 holes/cm 3 ) led to a slight decrease in APB density compared to the undoped sample. The increase in APB density with more n type Fermi level shifts

PAGE 150

150 correlates to decreases in antisite formation energies in GaAs. The decrease in an tisite formation energy has a broader effect of stabilizing APBs during growth and hindering self annihilation. The effect of interfacial contamination on nucleated APB density in GaAs films grown on Si was also studied in depth. Carbon was determined to be the primary source of contamination. Oxygen was also a source of contamination but was not present in as much quantity. It was found that the APB density increased with increasing interfacial carbon dose. Amorphous contaminant particles were observed at the GaAs/Si interface in XTEM and were directly associated with the formation of APBs as well as other crystalline defects. It is possible that there is a direct nucleation effect of amorphous particles for APBs. However, it is more likely that residual c ontamination essentially passivates single steps on the Si surface and prevents their reconstruction to double steps during high temperature pre growth anneals. Therefore, areas of the substrate surface with more contamination are associated with greater n umber of single steps which in turn lead to higher nucleated APB densities. 7.2 Future Work The future experimental work for this investigation of APB energetics in GaAs on Si centers on the effect of doping. Ideally, it would be better to grow a full se ries of doped samples with varying p type and n type doping levels to develop a more quantitative relationship of the effect on grown in APB density. Further, annealing experiments of the doped samples can be performed to study the effect of Fermi level sh ifts on the energetics of APB diffusion. It would be interesting to see if there is a change in the activation energy for APB diffusion during annealing. These experiments

PAGE 151

151 would provide greater insight on how antisite defects and their respective formation energies play a role in APB motion. There is an additional need for computational work to support the experimental findings presented here and gain insight on the atomic mechanisms that facilitate APB annihilation. The most recent DFT studies on APB ener getics in GaAs only investigated simple systems of parallel APB planes [92] More complex APB systems should be modeled, e.g a {110} APB with a {112} kink, to determine which configurations are energetically feasible for APB propagation and annihilation during growth. The diffusion process of APBs during annea ling should also be modeled. It is not clear what atomic exchange process allows for the net diffusive motion of an APB and would also be energetically favorable. An additional nuance to these computational studies would be to add the effect of Fermi level shifts on the APB energetics.

PAGE 152

152 APPEN DIX SUPPLEMENTARY INFORMATION A.1 Particle Distribution Calculation Assume cylindrical particles Radius = 10 nm Height = 2 nm Volume = r 2 h = 6.3x10 19 cm 3 If particles are SiC Want atoms/cm 3 Density = 3.21 g/cm 3 Molar mass = 40.1 g/mol 23 atoms/mol => 9.6x10 22 atoms/cm 3 One half are C atoms in SiC => 4.8x10 22 C atoms/cm 3 Want atoms/particle (4.8x10 22 C atoms/cm 3 )*(6.3x10 19 cm 3 [volume per particle]) => 3.0x10 4 C atoms/particle Want areal density of interfacial particles (1.7x10 15 C atoms/cm 2 [SIMS dose])/(3.0x10 4 atoms/particle) => 5.7x10 10 particles/cm 2 = 5.7x10 4 particles/nm 2 Assume lamella thickness of 50 nm (50 nm)*(5.7x10 4 particles/nm 2 ) => 0.028 particles/nm => 1 particle every ~36 nm If particles are SiO 2 Want atoms/cm 3 Density = 2.65 g/cm 3 Molar mass = 60.1 g/mol 23 atoms/mol => 8.0x10 22 atoms/cm 3 Two thirds are O atoms in SiO 2 => 5.3x10 22 C atoms/cm 3 Want atoms/particle (5.3x10 22 O atoms/cm 3 )*(6.3x10 19 cm 3 [volume per particle]) => 3.3x10 4 O atoms/particle

PAGE 153

153 Want areal density of interfacial particles (4.0x10 12 O atoms/cm 2 [SIMS dose])/(3.3x10 4 atoms/particle) => 1.2x10 8 particles/cm 2 = 1.2x10 6 particles/nm 2 Assume l amella thickness of 50 nm (50 nm)*(1.2x10 6 particles/nm 2 ) => 6.0x10 5 particles/nm => 1 particle every ~17000 nm or 17 m A.2 Hybrid Al 2 O 3 ALD Recipe Step Intruction # Value Units/Comment 1 flow 0 20 sccm 2 flow 1 40 sccm 3 heater 16 150 C 4 heater 17 150 C 5 stabilize 16 6 stabilize 17 7 heater 12 195 C 8 heater 13 195 C 9 heater 14 195 C 10 heater 15 200 C 11 stabilize 12 12 stabilize 13 13 stabilize 14 14 stabilize 15 15 wait 600 sec 16 flow 0 20 sccm 17 flow 1 60 sccm 18 wait 5 sec 19 stopvalve 0 closed 20 wait 1 sec 21 pulse Al 0.3 sec 22 wait 60 sec 23 stopvalve 1 open 24 flow 0 40 sccm 25 flow 1 140 sccm 26 wait 90 sec 27 flow 0 20 sccm 28 flow 1 60 sccm 29 wait 5 sec 30 stopvalve 0 closed 31 wait 1 sec 32 pulse 0 0.3 sec

PAGE 154

154 33 wait 60 sec 34 stopvalve 1 open 35 flow 0 40 sccm 36 flow 1 140 sccm 37 wait 90 sec 38 flow 0 20 sccm 39 flow 1 60 sccm 40 wait 5 sec 41 goto 18 5 cycles//end of exposure mode 0.5nm 42 wait 30 sec 43 flow 0 20 sccm//go to has to come back here42 44 flow 1 60 sccm 45 wait 30 sec 46 stopvalve 0 closed//start hybrid exposure 47 wait 1 sec 48 pulse Al 0.3 sec 49 wait 60 sec 50 stopvalve 1 open 51 flow 0 40 sccm//purge cycle after Al pulse 52 flow 1 140 sccm 53 wait 90 sec//end purge cycle 54 flow 0 60 sccm 55 flow 1 200 sccm 56 MFCvalve 3 1 57 wait 5 sec 58 flow 3 20 sccm 59 wait 5 sec 60 plasma 300 Watts 61 wait 60 sec 62 plasma 0 Watts 63 flow 3 0 sccm 64 wait 2 sec 65 MFCvalve 3 0 66 wait 60 sec 67 goto 42 200 cycles//end of hybrid exposure 68 flow 0 20 sccm 69 flow 1 40 sccm 70 wait 30 sec 71 heater 12 195 C

PAGE 155

155 72 heater 13 195 C 73 heater 14 195 C 74 heater 15 200 C 75 wait 5 sec 76 flow 0 0 sccm 77 flow 1 0 sccm 78 wait 10 sec 79 doorpurge 0 80 wait 5 sec

PAGE 156

156 LIST OF REFERENCES [1] Electronics vol. 38, no. 8, p. 56, 1965. [2] IEEE J. Solid State Circuits vol. 9, no. 5, pp. 256 268, Oct. 1974. [3] ice scaling limits of Si Proc. IEEE vol. 89, no. 3, pp. 259 288, Mar. 2001. [4] Computer vol. 36, no. 12, pp. 68 75, 2 003. [5] Proceedings of the 2013 ACM International Symposium on International Symposium on Physical Design New York, NY, USA, 2013, pp. 66 67. [6] scale el ectronics with III V compound Nature vol. 479, no. 7373, pp. 317 323, Nov. 2011. [7] Optical Fiber Communication Conferen ce. Optical Society of America 2014. [8] MRS Proc. vol. 91, Jan. 1987. [9] S. E. Grillo, M. Ducarroir, M. Nadal, E. Tourni, and J. P. Faurie, J. Phys. Appl. Phys. vol. 36, no. 1, p. L5, 2003. [10] MRS Proc. vol. 91, Jan. 1987. [11] of antiphase domain boundaries in GaAs on Si grown by molecular beam J. Mater. Res. vol. 8, no. 8, pp. 1908 1921, 1993. [12] of CMOS Logic Devices with Ge and III MRS Bull. vol. 34, no. 7, pp. 485 492, Jul. 2009. [13] J. Appl. Phys. vol. 68, no. 7, pp. R31 R58 Oct. 1990.

PAGE 157

157 [14] V/Ge channel MOS device technologies Jpn. J. Appl. Phys. vol. 54, no. 6S1, p. 06FA01, Jun. 2015. [15] insulating GaAs layers grown by molecular beam epitaxy, Appl. Phys. Lett. vol. 42, no. 2, pp. 178 180, Jan. 1983. [16] Physica vol. 20, no. 7, pp. 893 909, Jan. 1954. [17] VLSI Electron. Microstruct. Sci. vol. 11, pp. 1 39, Jan. 1985. [18] P. H. Holloway and G. E. McGuire, Handbook of Compound Semiconductors: Growth, Processing, Characterization, and Devices Elsevier, 1996. [19] nd MBE growth technologies for III V CS MANTECH Conference, USA 2013, pp. 105 108. [20] Appl. Phys. Lett. vol. 44, no. 12, pp. 1149 1151, Jun. 1984. [21] J. Cryst. Growth vol. 68, no. 1, pp. 21 26, Sep. 1984. [22] T. Li, M. Mastro, and A. Dadgar, III V Compound Semiconductors: Integration with Silicon Ba sed Microelectronics CRC Press, 2011. [23] strain relaxation during molecular beam epitaxy of highly strained J. Cryst. Growth vol. 135, n o. 1 2, pp. 97 112, Jan. 1994. [24] J. E. Ayers, Heteroepitaxy of Semiconductors: Theory, Growth, and Characterization CRC Press, 2007. [25] W. D. Callister and D. G. Rethwisch, Fundamentals of materials science and engineering: an integrated approach Jo hn Wiley & Sons, 2008. [26] J. Appl. Phys. vol. 61, no. 5, pp. 1856 1859, Mar. 1987. [27] D. K. Biegelsen, F. A. Ponce, B. S. K rusor, J. C. Tramontana, and R. D. Yingling, thickness samples for molecular beam epitaxial growth studies of GaAs/Si heteroepitaxy, Appl. Phys. Lett. vol. 52, no. 21, pp. 1779 1781, May 1988.

PAGE 158

158 [28] Phys. Uspekhi vol. 51, no. 5, pp. 437 456, May 2008. [29] ng J. Cryst. Growth vol. 93, no. 1 4, pp. 15 19, 1988. [30] G. B. Stringfellow, Organometallic Vapor Phase Epitaxy: Theory and Practice San Diego, CA: Academic Press, 1989. [31] S. J. Rosner, S. M. Koch, J. S. H layers of MBE MRS Proc. vol. 67, Jan. 1986. [32] A. E. Blakeslee, M. M. Al MRS Proc. vol. 91, Ja n. 1987. [33] Z. Liliental Weber, E. R. Weber, L. Parechanian use of convergent beam electron diffraction for identification of antiphase Ultramicroscopy vol. 26, no. 1 2, pp. 59 63, 1988. [ 34] Sol. Energy Mater. Sol. Cells vol. 66, no. 1 4, pp. 479 486, Feb. 2001. [35] Energy Electron Diffraction and Auger J. Electrochem. Soc. vol. 119, no. 6, pp. 772 775, Jun. 1972. [36] J. Electrochem. Soc. vol. 133, no. 4, pp. 666 6 71, Apr. 1986. [37] MRS Proc. vol. 67, Jan. 1986. [38] mismatched growth of III V Prog. Cryst. Growth Charact. Mater. vol. 22, no. 1 2, pp. 53 141, 1991. [39] J. Cryst. Growth vol. 315, no. 1, pp. 10 15, Jan. 2011. [40] on nonpolar epitaxy: Sublattice ordering in the nucleation and growth of GaP on Si(211) surfaces, J. Vac. Sci. Technol. vol. 21, no. 2, pp. 534 539, Jul. 1982.

PAGE 159

159 [41] S. J. Appl. Phys. vol. 55, no. 8, pp. 2916 2927, Apr. 1984. [42] J. Appl. Phys. vol. 58, n o. 6, pp. 2195 2203, Sep. 1985. [43] MRS Proc. vol. 67, Jan. 1986. [44] D. B. Holt and B. G. Yacobi, Extended Defects in Semiconductors: Electronic Properties, Device Effects and Structures C ambridge University Press, 2007. [45] J. Appl. Phys. vol. 104, no. 9, p. 93520, Nov. 2008. [46] L. Goldstein, F. Glas, molecular beam epitaxy and characterization of InAs/GaAs strained layer superlattices, Appl. Phys. Lett. vol. 47, no. 10, pp. 1099 1101, Nov. 1985. [47] T. Soga, S. Hattori, S. Sakai, M. Takeyasu, epitaxially grown GaAs on Si substrates with III V compounds intermediate layers by metalorganic chemical vapor deposition, J. Appl. Phys. vol. 57, no. 10, pp. 4578 4582, May 1985. [48] H. Okamoto, Y. Watanabe, Y. Kado GaAs on Si by Thermal Cycles and InGaAs/GaAs Strained Jpn. J. Appl. Phys. vol. 26, no. 12A, p. L1950, Dec. 1987. [49] N. El f strained layer superlattices in reducing defects in GaAs epilayers grown on silicon substrates, Appl. Phys. Lett. vol. 51, no. 20, pp. 1608 1610, Nov. 1987. [50] N. A. El Masry, J. C. Tarn, and N. H. Karam, Interactions of dislocations in GaAs grown o n Si substrates with InGaAs GaAsP strained layered superlattices, J. Appl. Phys. vol. 64, no. 7, pp. 3672 3677, Oct. 1988. [51] layer superlattice effects on dislocation density reduction in G Appl. Phys. Lett. vol. 54, no. 1, pp. 24 26, Jan. 1989. [52] J. Z. Li, J. Bai, J. Appl. Phys. Lett. vol. 91, no. 2, p. 21114, 20 07. [53] J. Appl. Phys. vol. 103, no. 10, p. 106102, May 2008.

PAGE 160

160 [54] V Materia Cryst. Growth Des. vol. 12, no. 10, pp. 4696 4702, Oct. 2012. [55] Axis Si J. Electrochem. Soc. vol. 159, no. 3, pp. H260 H265, 2012. [56] selective area metal organic vapor phase epitaxy in sub 50 nm width trenches: J. Appl. Phys. vol. 115, no. 2, p. 23710, Jan. 2014. [57] organic chemical vapor Appl. Phys. Lett. vol. 105, no. 6, p. 62101, Aug. 2014. [58] on Si epitaxy by aspect ratio trapping: J. Appl. Phys. vol. 118, no. 10, p. 105307, Sep. 2015. [59] on J. Appl. Phys. vol. 120, no. 8, p. 85308, Aug. 2016. [60] J. Phys. Chem. Solids vol. 30, no. 6, pp. 1297 1308, Jun. 1969. [61] tiphase domain structures in GaP and GaAs epitaxial layers J. Cryst. Growth vol. 38, no. 2, pp. 249 254, May 1977. [62] subsurface strain S urf. Sci. vol. 74, no. 1, pp. 21 33, May 1978. [63] Phys. Rev. Lett. vol. 43, no. 1, pp. 43 47, Jul. 1979. [64] on J. Cryst. Growth vol. 81, no. 1 4, pp. 193 204, Feb. 1987. [65] Phys. Rev. Lett. vol. 59, no. 15, pp. 1691 1694, Oct. 1987. [66] es of Stepped Jpn. J. Appl. Phys. vol. 13, no. S2, p. 389, Jan. 1974.

PAGE 161

161 [67] Surf. Sci. vol. 93, no. 1, pp. 145 158, Mar. 1980. [68] Phys. Rev. Lett. vol. 59, no. 19, pp. 2169 2172, Nov. 1987. [69] temperature phase diagram Phys. Rev. Lett. vol. 64, no. 20, pp. 2406 2409, May 1990. [70] Phys. Rev. Lett. vol. 67, no. 10, pp. 1290 1293, Sep. 1991. [71] T. Sakamoto and G. Hashi 21 Single Domain Structure Obtained Jpn. J. Appl. Phys. vol. 25, no. 1A, p. L78, Jan. 1986. [72] Phys. Rev. Lett. vol. 57, no. 24, pp. 3 054 3057, Dec. 1986. [73] hydrogen Surf. Sci. vol. 545, no. 1 2, pp. 70 84, Nov. 2003. [74] K. Volz, W. Stolz, A. Dadgar, and A. Krost, Growth of III/Vs on Silicon: Nitrides, Phosphides, Arsenides and Handbook of Crystal Growth (Second Edition) T. F. Kuech, Ed. Boston: North Holland, 2015, pp. 1249 1300. [75] M. Akiyama, Y. Kawarada, T. Ueda, S. Nishi, and K. Kamini J. Cryst. Growth vol. 77, no. 1 3, pp. 490 497, Sep. 1986. [76] J. C ryst. Growth vol. 77, no. 1 3, pp. 498 502, Sep. 1986. [77] epitaxial growth of a GaAs layer free from antiphase domains on an exactly (100) oriented Si substrate preheated at 1000 J. Cryst. Growth vol. 83, no. 3, pp. 431 436, Jun. 1987. [78] and crystallographic tilt in GaP layers grown on misoriented Si(001) substrates by metalorganic vapor phase epit J. Appl. Phys. vol. 107, no. 6, p. 63506, Mar. 2010.

PAGE 162

162 [79] J. Cryst. Growth vol. 363, pp. 258 263, Jan. 2013. [80] Y. Wan step growth of metamorphic GaAs on Si(001) by low J. Vac. Sci. Technol. B vol. 31, no. 5, p. 51211, Sep. 2013. [81] C. B. Carter, N. H. Cho, S. Mckernan, and D. K. Wagne MRS Proc. vol. 91, Jan. 1987. [82] annihilation of antiphase domains in GaAs/Si heterostructures, Appl. Phys. Lett. vol. 55, no 5, pp. 445 447, Jul. 1989. [83] Z. Liliental MRS Proc. vol. 148, Jan. 1989. [84] A. Georgakilas, P. Panayotatos, J. Stoemenos, J. L. Mourrain, and A. Christou, ations in optimized GaAs films grown on Si by molecular J. Appl. Phys. vol. 71, no. 6, p. 2679, 1992. [85] N. J. Mater. Sc i. vol. 36, no. 17, pp. 4209 4222, Sep. 2001. [86] J. Appl. Phys. vol. 75, no. 11, pp. 7258 7265, Jun. 1994. [8 7] J. Appl. Phys. vol. 93, no. 7, p. 3859, 2003. [88] free GaAs epitaxy on Si by using midpatterned growth: Application to Si based wavelength selective J. Appl. Phys. vol. 104, no. 11, p. 113114, Dec. 2008. [89] GaAs/Ge superlattices grown by me J. Cryst. Growth vol. 435, pp. 50 55, Feb. 2016. [90] J. Vac. Sci. Technol. B vol. 4, no. 4, pp. 874 877, Jul. 1986. [ 91] Phys. Rev. B vol. 45, no. 19, pp. 11192 11201, May 1992.

PAGE 163

163 [92] Int J. Mol. Sci. vol. 10, no. 12, pp. 5104 5114, Nov. 2009. [93] N. H. Cho, B. C. D. Cooman, C. B. Carter, R. Fletcher, and D. K. Wagner, Appl. Phys. Lett. vol. 47, no. 8, pp. 879 881, Oct. 1985. [94] D. Cohen and C. B. Cart J. Microsc. vol. 208, no. 2, pp. 84 99, 2002. [95] J. Appl. Phys. vol. 109, no. 8, p. 83529, Apr. 2011. [96] M. Grundmann, The physics of semiconductors: an introduction including devices and nanophysics [97] Appl. P hys. Lett. vol. 54, no. 21, p. 2094, 1989. [98] Phys. Rev. B vol. 37, no. 9, pp. 4760 4763, Mar. 1988. [99] III J. Appl. Phys. vol. 116, no. 2, p. 23505, Jul. 2014. [100] Defect Interactions in Si Doped I Florida, Gainesville, Fla., 2015. [101] Proc. R. Soc. Lond. Ser. Contain. Pap. Math. Phys. Character vol. 145, no. 855, pp. 699 730, 1 934. [102] J. Chem. Phys. vol. 6, no. 3, pp. 130 144, Mar. 1938. [103] Proc. Phys. Soc. vol. 52, no. 1, p. 105, 1940. [104] A. B. Glosso Phase Domain Boundaries in Ordered Copper Proc. R. Soc. Lond. Math. Phys. Eng. Sci. vol. 250, no. 1260, pp. 132 146, Feb. 1959. [105] ervation of Antiphase Domain J. Appl. Phys. vol. 31, no. 9, pp. 1687 1687, Sep. 1960.

PAGE 164

164 [106] Phase Domains in Cu3Au by Means of Electron Diffraction and Electron J. Phys. Soc. Jpn. vol. 17, no. 6, pp. 1030 1041, Jun. 1962. [107] Acta Metall. vol. 27, no. 7, pp. 1261 1269, Jul. 1979. [108] A. Putnis, An Introduction to Mineral Sciences Cambridge University Press, 1992. [109] D. A. Porter, K. E. Easterling, and M. Sherif, Phase Transformations in Metals and Alloys, Third Edition (Revised Reprint) CRC Press, 2009. [110] Acta Metall. vol. 27, no. 6, pp. 1085 1095, Jun. 1979. [111] A. V. Ramos, J. B. Moussy, M. netotransport properties of J. Appl. Phys. vol. 100, no. 10, p. 103902, Nov. 2006. [112] antipha se domain boundaries in ${ \ mathrm{Fe}}_{3}{ \ Phys. Rev. B vol. 68, no. 1, p. 14428, Jul. 2003. [113] Correlation of morphology, anti phase domain structure an d MOVPE growth J. Cryst. Growth vol. 310, no. 7 9, pp. 1595 1601, Apr. 2008. [114] nucleation on exact Si (0 0 1) substrates J. Cryst. Growth vol. 315, no. 1, pp. 37 47 Jan. 2011. [115] S. N. G. Chu, S. Nakahara, S. J. Pearton, T. Boone, and S. M. Vernon, on silicon on J. Appl. Phys. vol. 64, no. 6, p. 2981, 1988. [116] W. Guo, A. Bo boundaries in GaP/Si epilayers evidenced by advanced X Appl. Surf. Sci. vol. 258, no. 7, pp. 2808 2815, Jan. 2012. [117] i phase boundaries Free Appl. Phys. Lett. vol. 107, no. 21, p. 212105, Nov. 2015.

PAGE 165

165 [118] Ni silicide (Siconi) cleaning at 65nm to J. Vac. Sci. Technol. B vol. 28, no. 1, pp. 56 61, Jan. 2010. [119] S. A. Campbell, Fabrication engineering at the micro and nanoscale 3rd ed. New York: Oxford University Press, 2008. [120] W. M. Duncan and G. H. Westph VLSI Electron. Microstruct. Sci. vol. 11, pp. 41 86, Jan. 1985. [121] evaporation temperature of GaAs(001) controlled by As Appl. Phys. Lett. vol. 97, no. 12, p. 121912, Sep. 2010. [122] V. Borisenko and P. J. Hesketh, Rapid Thermal Processing of Semiconductors Springer Science & Business Media, 2013. [123] Review of Doping Issues in III ECS Trans. vol. 53, no. 3, pp. 97 105, May 2013. [124] cl eaning by Appl. Phys. Lett. vol. 92, no. 7, p. 71901, Feb. 2008. [125] k dielectrics on III V Prog. Cryst. Growth Charact. Mater. vol. 62, no. 4, pp. 1 21, D ec. 2016. [126] University of Florida, Gainesville, Fla., 2012. [127] D. B. Williams and C. B. Carter, Transmission Electron Microscopy: A Textbook for Materials Science Sprin ger Science & Business Media, 2009. [128] R. F. Egerton, Physical Principles of Electron Microscopy: An Introduction to TEM, SEM, and AEM Springer Science & Business Media, 2006. [129] T. S. Kuan and C. GaAs J. Appl. Phys. vol. 54, no. 8, pp. 4408 4413, Aug. 1983. [130] S. H. Vajargah, S. Y. Woo, S. Ghanad resolution study of polarity reversal in GaSb grown on Si by scanning transmission electron J. Appl. Phys. vol. 112, no. 9, p. 93101, Nov. 2012. [131] phase Appl. Phys. Lett. vol. 103, no. 3, p. 32107, Jul. 2013.

PAGE 166

166 [132] C Nat. Methods vol. 9, no. 7, pp. 671 675, Jul. 2012. [133] resolution STEM at low Ultramicroscopy vol. 114, pp. 62 71, Mar. 2012. [134] J. Microsc. vol. 255, no. 3, pp. 180 187, Sep. 2014. [135] based structures on exactly (001) oriented Ge/Si virtual substrates: reduction of the structural defect density and observation of electroluminescence at room J. Cryst. Growth vol. 265, no. 1 2, pp. 53 59, Apr. 2004. [136] [137] he effect of rapid thermal annealing on crystalline quality of GaAs grown on Si by molecular beam epitaxy, J. Vac. Sci. Technol. B vol. 5, no. 3, pp. 822 826, May 1987. [138] growth thermal annealin g J. Cryst. Growth vol. 125, no. 1 2, pp. 329 335, Nov. 1992. [139] ray study of antiphase domains and J. Cryst. Growth vol. 323, no. 1, pp. 409 412, May 2011. [140] angle spectroscopic ellipsometry (VASE): I. Basic theory and typical Opt. Metrol. vol. CR72, pp 3 28, 1999. [141] D. C. Look, Electrical Characterization of GaAs Materials and Devices Wiley, 1992. [142] J. Mater. Sci. vol. 39, no. 1, pp. 361 363, 2004. [143] J. Mater. Sci. vol. 51, no. 1, pp. 449 456, Aug. 2015. [144] T. F. Kuech and J. Cryst. Growth vol. 68, no. 1, pp. 148 156, Sep. 1984.

PAGE 167

167 [145] Surf. Sci. vol. 411, no. 1 2, pp. 61 69, Aug. 1998. [146] J. Cryst. Growth vol. 132, no. 3 4, pp. 477 482, Sep. 1993. [147] C. S. C. Barrett, T. P. Martin, X. temperature on antiphase domain boundary annihilation rate in MOCVD grown GaAs on J. Cryst. Growth vol. 450, pp. 39 44, Sep. 2016. [148] J. B. Posthill, J. C. L. Tarn, K. Das, T. P. Humphreys, and N. R. Parikh, boundaries in GaAs on silicon by transmission Appl. Phys. Lett. vol. 53, no. 13, p. 1207, 1988. [149] selectively doped AlGaAs/GaAs heterostructures by metalorgan ic vapor phase J. Appl. Phys. vol. 67, no. 12, pp. 7576 7582, Jun. 1990. [150] 2011 37th IEEE Photovoltaic Specialists Conference 2011, pp. 003562 003567. [151] 2012 38th IEEE Photovoltaic Specialists Conference 2012, pp. 001135 001138. [152] Al2O3 layers grown on silicon and its effect on metal insulator semiconductor J. Vac. Sci. Technol. Vac. Surf. Films vol. 31, no. 1, p. 01A128, Nov. 2012. [153] Appl. Phys. Lett. vol. 50, no. 15, pp. 992 994, Apr. 1987. [154] erface after Appl. Phys. Lett. vol. 51, no. 2, pp. 130 132, Jul. 1987. [155] M. M. Al MRS Proc. vol. 116, Jan. 1988. [156] R. W. Kaliski, substrate orientation on metalorganic chemical vapor deposition GaAs on silicon J. Appl. Phys. vol. 64, no. 3, pp. 1196 1200, Aug. 1988.

PAGE 168

168 [157] N. Chand, R. People, F. A. Baiocc improvement in crystalline quality of molecular beam epitaxially grown GaAs on Appl. Phys. Lett. vol. 49, no. 13, pp. 815 817, Sep. 1986. [158] J. M. Baribeau, D. C. Hought MRS Proc. vol. 91, Jan. 1987. [159] J. A ppl. Phys. vol. 68, no. 9, pp. 4518 4522, Nov. 1990. [160] domain structure and coarsening at Cu Phys. Rev. B vol. 57, no. 19, p. 12497, 1998. [ 161] diffusion in GaAs, J. Appl. Phys. vol. 69, no. 5, pp. 3022 3031, Mar. 1991. [162] diffusion in GaAs isotope Phys. Rev. Lett. vol. 76, no. 13, p. 234 2, 1996. [163] Defect Diffus. Forum vol. 143 147, pp. 1079 1094, 1997. [164] Diffusion in gallium J. Electron. Mater. vol. 12, no. 5, pp. 863 877, Sep. 1983. [165] proximity caps during the annealing of UV J. Appl. Phys. vol. 101, no. 11, p. 114321, Jun. 2007. [166] formation energies in GaAs: Application to Ga self Phys. Rev. Lett. vol. 67, no. 17, pp. 2339 2342, Oct. 1991. [167] G. Radhakrishnan, O. McC spectrometry study of ex situ annealing of epitaxial GaAs grown on Si Appl. Phys. Lett. vol. 52, no. 9, pp. 731 732, Feb. 1988. [168] A. Freundlich, A. Leycuras, J. C. Grenet, and C. Grattepa related Si diffusion in GaAs on Si, Appl. Phys. Lett. vol. 53, no. 26, pp. 2635 2637, Dec. 1988. [169] GaAs on Si grown by metalorganic chemical vapor depositio n, Appl. Phys. Lett. vol. 51, no. 9, pp. 682 684, Aug. 1987.

PAGE 169

169 [170] S. O. Kasap, Principles of electronic materials and devices 3rd ed. Boston: McGraw Hill, 2006.

PAGE 170

170 BIOGRAPHICAL SKETCH Caleb Shuan Chia Barrett is the son of Michael and Shau Ling Barre tt. He grew up in Greenbelt, MD and graduated from the Science and Technology Program at Eleanor Roosevelt High School. He then attended the University o f Maryland where he majored in m aterials science and e ngineering. After graduating in 2013, he moved on to the University of Florida to pursue his Ph.D. in materials science and e ngineering under Dr. Kevin Jones. Along the way, he aterials s cience and e ngineering in 2015. He has accepted an offer to work at Intel in Hillsboro, OR, after the completion of his Ph.D. to fund his future car collection and world explorations